^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) // SPDX-License-Identifier: GPL-2.0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * Copyright (C) 2007 Oracle. All rights reserved.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) #include <crypto/hash.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) #include <linux/kernel.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) #include <linux/bio.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) #include <linux/file.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) #include <linux/fs.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) #include <linux/pagemap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) #include <linux/highmem.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) #include <linux/time.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) #include <linux/init.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) #include <linux/string.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) #include <linux/backing-dev.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) #include <linux/writeback.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) #include <linux/compat.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) #include <linux/xattr.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) #include <linux/posix_acl.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) #include <linux/falloc.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) #include <linux/slab.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) #include <linux/ratelimit.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) #include <linux/btrfs.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) #include <linux/blkdev.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) #include <linux/posix_acl_xattr.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) #include <linux/uio.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) #include <linux/magic.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) #include <linux/iversion.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) #include <linux/swap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) #include <linux/migrate.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) #include <linux/sched/mm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) #include <linux/iomap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) #include <asm/unaligned.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) #include "misc.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) #include "ctree.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) #include "disk-io.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) #include "transaction.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) #include "btrfs_inode.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) #include "print-tree.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) #include "ordered-data.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) #include "xattr.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) #include "tree-log.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) #include "volumes.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) #include "compression.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) #include "locking.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) #include "free-space-cache.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) #include "inode-map.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) #include "props.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) #include "qgroup.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) #include "delalloc-space.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) #include "block-group.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) #include "space-info.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) struct btrfs_iget_args {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) u64 ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) struct btrfs_root *root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) struct btrfs_dio_data {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) u64 reserve;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) loff_t length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) ssize_t submitted;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) struct extent_changeset *data_reserved;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) bool sync;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) static const struct inode_operations btrfs_dir_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) static const struct inode_operations btrfs_symlink_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) static const struct inode_operations btrfs_special_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) static const struct inode_operations btrfs_file_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) static const struct address_space_operations btrfs_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) static const struct file_operations btrfs_dir_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) static struct kmem_cache *btrfs_inode_cachep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) struct kmem_cache *btrfs_trans_handle_cachep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) struct kmem_cache *btrfs_path_cachep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) struct kmem_cache *btrfs_free_space_cachep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) struct kmem_cache *btrfs_free_space_bitmap_cachep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) static int btrfs_setsize(struct inode *inode, struct iattr *attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) static int btrfs_truncate(struct inode *inode, bool skip_writeback);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) static noinline int cow_file_range(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) u64 start, u64 end, int *page_started,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) unsigned long *nr_written, int unlock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) static struct extent_map *create_io_em(struct btrfs_inode *inode, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) u64 len, u64 orig_start, u64 block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) u64 block_len, u64 orig_block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) u64 ram_bytes, int compress_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) int type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) static void __endio_write_update_ordered(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) const u64 offset, const u64 bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) const bool uptodate);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) * Cleanup all submitted ordered extents in specified range to handle errors
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) * from the btrfs_run_delalloc_range() callback.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) * NOTE: caller must ensure that when an error happens, it can not call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) * extent_clear_unlock_delalloc() to clear both the bits EXTENT_DO_ACCOUNTING
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) * and EXTENT_DELALLOC simultaneously, because that causes the reserved metadata
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) * to be released, which we want to happen only when finishing the ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) * extent (btrfs_finish_ordered_io()).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) static inline void btrfs_cleanup_ordered_extents(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) u64 offset, u64 bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) unsigned long index = offset >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) unsigned long end_index = (offset + bytes - 1) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) u64 page_start = page_offset(locked_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) u64 page_end = page_start + PAGE_SIZE - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) while (index <= end_index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) page = find_get_page(inode->vfs_inode.i_mapping, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) index++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) ClearPagePrivate2(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) * In case this page belongs to the delalloc range being instantiated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) * then skip it, since the first page of a range is going to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) * properly cleaned up by the caller of run_delalloc_range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) if (page_start >= offset && page_end <= (offset + bytes - 1)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) offset += PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) bytes -= PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) return __endio_write_update_ordered(inode, offset, bytes, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) static int btrfs_dirty_inode(struct inode *inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) struct inode *inode, struct inode *dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) const struct qstr *qstr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) err = btrfs_init_acl(trans, inode, dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) if (!err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) err = btrfs_xattr_security_init(trans, inode, dir, qstr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) * this does all the hard work for inserting an inline extent into
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) * the btree. The caller should have done a btrfs_drop_extents so that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) * no overlapping inline items exist in the btree
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) static int insert_inline_extent(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) struct btrfs_path *path, int extent_inserted,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) struct btrfs_root *root, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) u64 start, size_t size, size_t compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) int compress_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165) struct page **compressed_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) struct page *page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) char *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) unsigned long ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) struct btrfs_file_extent_item *ei;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) size_t cur_size = size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) unsigned long offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) ASSERT((compressed_size > 0 && compressed_pages) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) (compressed_size == 0 && !compressed_pages));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) if (compressed_size && compressed_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) cur_size = compressed_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182) inode_add_bytes(inode, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) if (!extent_inserted) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) size_t datasize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) key.objectid = btrfs_ino(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) key.offset = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) key.type = BTRFS_EXTENT_DATA_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) datasize = btrfs_file_extent_calc_inline_size(cur_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) path->leave_spinning = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194) ret = btrfs_insert_empty_item(trans, root, path, &key,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) datasize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) ei = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) struct btrfs_file_extent_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) btrfs_set_file_extent_generation(leaf, ei, trans->transid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) btrfs_set_file_extent_encryption(leaf, ei, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) btrfs_set_file_extent_other_encoding(leaf, ei, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) btrfs_set_file_extent_ram_bytes(leaf, ei, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) ptr = btrfs_file_extent_inline_start(ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) if (compress_type != BTRFS_COMPRESS_NONE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) struct page *cpage;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) int i = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) while (compressed_size > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213) cpage = compressed_pages[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214) cur_size = min_t(unsigned long, compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217) kaddr = kmap_atomic(cpage);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218) write_extent_buffer(leaf, kaddr, ptr, cur_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219) kunmap_atomic(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) i++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) ptr += cur_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223) compressed_size -= cur_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) btrfs_set_file_extent_compression(leaf, ei,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) compress_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) page = find_get_page(inode->i_mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229) start >> PAGE_SHIFT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) btrfs_set_file_extent_compression(leaf, ei, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231) kaddr = kmap_atomic(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232) offset = offset_in_page(start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) write_extent_buffer(leaf, kaddr + offset, ptr, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234) kunmap_atomic(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) btrfs_mark_buffer_dirty(leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) * We align size to sectorsize for inline extents just for simplicity
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) * sake.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244) size = ALIGN(size, root->fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245) ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode), start, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 248)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 249) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 250) * we're an inline extent, so nobody can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251) * extend the file past i_size without locking
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252) * a page we already have locked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) * We must do any isize and inode updates
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255) * before we unlock the pages. Otherwise we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) * could end up racing with unlink.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258) BTRFS_I(inode)->disk_i_size = inode->i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261) fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267) * conditionally insert an inline extent into the file. This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) * does the checks required to make sure the data is small enough
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) * to fit as an inline extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271) static noinline int cow_file_range_inline(struct btrfs_inode *inode, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) u64 end, size_t compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 273) int compress_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 274) struct page **compressed_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 275) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 276) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 277) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 278) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 279) u64 isize = i_size_read(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 280) u64 actual_end = min(end + 1, isize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 281) u64 inline_len = actual_end - start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 282) u64 aligned_end = ALIGN(end, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 283) u64 data_len = inline_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 284) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 285) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 286) int extent_inserted = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 287) u32 extent_item_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 288)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 289) if (compressed_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 290) data_len = compressed_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 292) if (start > 0 ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 293) actual_end > fs_info->sectorsize ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 294) data_len > BTRFS_MAX_INLINE_DATA_SIZE(fs_info) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 295) (!compressed_size &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 296) (actual_end & (fs_info->sectorsize - 1)) == 0) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 297) end + 1 < isize ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 298) data_len > fs_info->max_inline) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 299) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 300) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 302) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 303) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 304) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 305)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 306) trans = btrfs_join_transaction(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 307) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 308) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 309) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 311) trans->block_rsv = &inode->block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 312)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 313) if (compressed_size && compressed_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 314) extent_item_size = btrfs_file_extent_calc_inline_size(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 315) compressed_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 316) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 317) extent_item_size = btrfs_file_extent_calc_inline_size(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 318) inline_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 320) ret = __btrfs_drop_extents(trans, root, inode, path, start, aligned_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 321) NULL, 1, 1, extent_item_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 322) &extent_inserted);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 323) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 324) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 325) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 326) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 327)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 328) if (isize > actual_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 329) inline_len = min_t(u64, isize, actual_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 330) ret = insert_inline_extent(trans, path, extent_inserted,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 331) root, &inode->vfs_inode, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 332) inline_len, compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 333) compress_type, compressed_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 334) if (ret && ret != -ENOSPC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 335) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 336) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 337) } else if (ret == -ENOSPC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 338) ret = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 339) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 340) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 341)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 342) set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 343) btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 344) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 345) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 346) * Don't forget to free the reserved space, as for inlined extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 347) * it won't count as data extent, free them directly here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 348) * And at reserve time, it's always aligned to page size, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 349) * just free one page here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 350) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 351) btrfs_qgroup_free_data(inode, NULL, 0, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 352) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 353) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 354) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 355) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 357) struct async_extent {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 358) u64 start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 359) u64 ram_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 360) u64 compressed_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 361) struct page **pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 362) unsigned long nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 363) int compress_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 364) struct list_head list;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 365) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 366)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 367) struct async_chunk {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 368) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 369) struct page *locked_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 370) u64 start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 371) u64 end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 372) unsigned int write_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 373) struct list_head extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 374) struct cgroup_subsys_state *blkcg_css;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 375) struct btrfs_work work;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 376) atomic_t *pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 377) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 379) struct async_cow {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 380) /* Number of chunks in flight; must be first in the structure */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 381) atomic_t num_chunks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 382) struct async_chunk chunks[];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 383) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 384)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 385) static noinline int add_async_extent(struct async_chunk *cow,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 386) u64 start, u64 ram_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 387) u64 compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 388) struct page **pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 389) unsigned long nr_pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 390) int compress_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 391) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 392) struct async_extent *async_extent;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 393)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 394) async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 395) BUG_ON(!async_extent); /* -ENOMEM */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 396) async_extent->start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 397) async_extent->ram_size = ram_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 398) async_extent->compressed_size = compressed_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 399) async_extent->pages = pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 400) async_extent->nr_pages = nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 401) async_extent->compress_type = compress_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 402) list_add_tail(&async_extent->list, &cow->extents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 403) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 404) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 405)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 406) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 407) * Check if the inode has flags compatible with compression
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 408) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 409) static inline bool inode_can_compress(struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 410) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 411) if (inode->flags & BTRFS_INODE_NODATACOW ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 412) inode->flags & BTRFS_INODE_NODATASUM)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 413) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 414) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 415) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 416)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 417) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 418) * Check if the inode needs to be submitted to compression, based on mount
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 419) * options, defragmentation, properties or heuristics.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 420) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 421) static inline int inode_need_compress(struct btrfs_inode *inode, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 422) u64 end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 423) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 424) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 425)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 426) if (!inode_can_compress(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 427) WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 428) KERN_ERR "BTRFS: unexpected compression for ino %llu\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 429) btrfs_ino(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 430) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 431) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 432) /* force compress */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 433) if (btrfs_test_opt(fs_info, FORCE_COMPRESS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 434) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 435) /* defrag ioctl */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 436) if (inode->defrag_compress)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 437) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 438) /* bad compression ratios */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 439) if (inode->flags & BTRFS_INODE_NOCOMPRESS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 440) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 441) if (btrfs_test_opt(fs_info, COMPRESS) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 442) inode->flags & BTRFS_INODE_COMPRESS ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 443) inode->prop_compress)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 444) return btrfs_compress_heuristic(&inode->vfs_inode, start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 445) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 446) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 447)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 448) static inline void inode_should_defrag(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 449) u64 start, u64 end, u64 num_bytes, u64 small_write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 450) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 451) /* If this is a small write inside eof, kick off a defrag */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 452) if (num_bytes < small_write &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 453) (start > 0 || end + 1 < inode->disk_i_size))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 454) btrfs_add_inode_defrag(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 455) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 456)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 457) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 458) * we create compressed extents in two phases. The first
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 459) * phase compresses a range of pages that have already been
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 460) * locked (both pages and state bits are locked).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 461) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 462) * This is done inside an ordered work queue, and the compression
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 463) * is spread across many cpus. The actual IO submission is step
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 464) * two, and the ordered work queue takes care of making sure that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 465) * happens in the same order things were put onto the queue by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 466) * writepages and friends.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 467) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 468) * If this code finds it can't get good compression, it puts an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 469) * entry onto the work queue to write the uncompressed bytes. This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 470) * makes sure that both compressed inodes and uncompressed inodes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 471) * are written in the same order that the flusher thread sent them
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 472) * down.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 473) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 474) static noinline int compress_file_range(struct async_chunk *async_chunk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 475) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 476) struct inode *inode = async_chunk->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 477) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 478) u64 blocksize = fs_info->sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 479) u64 start = async_chunk->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 480) u64 end = async_chunk->end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 481) u64 actual_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 482) u64 i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 483) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 484) struct page **pages = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 485) unsigned long nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 486) unsigned long total_compressed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 487) unsigned long total_in = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 488) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 489) int will_compress;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 490) int compress_type = fs_info->compress_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 491) int compressed_extents = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 492) int redirty = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 493)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 494) inode_should_defrag(BTRFS_I(inode), start, end, end - start + 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 495) SZ_16K);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 496)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 497) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 498) * We need to save i_size before now because it could change in between
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 499) * us evaluating the size and assigning it. This is because we lock and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 500) * unlock the page in truncate and fallocate, and then modify the i_size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 501) * later on.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 502) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 503) * The barriers are to emulate READ_ONCE, remove that once i_size_read
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 504) * does that for us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 505) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 506) barrier();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 507) i_size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 508) barrier();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 509) actual_end = min_t(u64, i_size, end + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 510) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 511) will_compress = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 512) nr_pages = (end >> PAGE_SHIFT) - (start >> PAGE_SHIFT) + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 513) BUILD_BUG_ON((BTRFS_MAX_COMPRESSED % PAGE_SIZE) != 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 514) nr_pages = min_t(unsigned long, nr_pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 515) BTRFS_MAX_COMPRESSED / PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 517) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 518) * we don't want to send crud past the end of i_size through
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 519) * compression, that's just a waste of CPU time. So, if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 520) * end of the file is before the start of our current
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 521) * requested range of bytes, we bail out to the uncompressed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 522) * cleanup code that can deal with all of this.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 523) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 524) * It isn't really the fastest way to fix things, but this is a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 525) * very uncommon corner.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 526) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 527) if (actual_end <= start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 528) goto cleanup_and_bail_uncompressed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 529)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 530) total_compressed = actual_end - start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 531)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 532) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 533) * skip compression for a small file range(<=blocksize) that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 534) * isn't an inline extent, since it doesn't save disk space at all.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 535) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 536) if (total_compressed <= blocksize &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 537) (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 538) goto cleanup_and_bail_uncompressed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 540) total_compressed = min_t(unsigned long, total_compressed,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 541) BTRFS_MAX_UNCOMPRESSED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 542) total_in = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 543) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 544)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 545) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 546) * we do compression for mount -o compress and when the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 547) * inode has not been flagged as nocompress. This flag can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 548) * change at any time if we discover bad compression ratios.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 549) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 550) if (inode_need_compress(BTRFS_I(inode), start, end)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 551) WARN_ON(pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 552) pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 553) if (!pages) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 554) /* just bail out to the uncompressed code */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 555) nr_pages = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 556) goto cont;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 557) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 558)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 559) if (BTRFS_I(inode)->defrag_compress)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 560) compress_type = BTRFS_I(inode)->defrag_compress;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 561) else if (BTRFS_I(inode)->prop_compress)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 562) compress_type = BTRFS_I(inode)->prop_compress;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 564) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 565) * we need to call clear_page_dirty_for_io on each
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 566) * page in the range. Otherwise applications with the file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 567) * mmap'd can wander in and change the page contents while
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 568) * we are compressing them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 569) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 570) * If the compression fails for any reason, we set the pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 571) * dirty again later on.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 572) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 573) * Note that the remaining part is redirtied, the start pointer
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 574) * has moved, the end is the original one.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 575) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 576) if (!redirty) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 577) extent_range_clear_dirty_for_io(inode, start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 578) redirty = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 579) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 580)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 581) /* Compression level is applied here and only here */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 582) ret = btrfs_compress_pages(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 583) compress_type | (fs_info->compress_level << 4),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 584) inode->i_mapping, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 585) pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 586) &nr_pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 587) &total_in,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 588) &total_compressed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 589)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 590) if (!ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 591) unsigned long offset = offset_in_page(total_compressed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 592) struct page *page = pages[nr_pages - 1];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 593) char *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 594)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 595) /* zero the tail end of the last page, we might be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 596) * sending it down to disk
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 597) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 598) if (offset) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 599) kaddr = kmap_atomic(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 600) memset(kaddr + offset, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 601) PAGE_SIZE - offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 602) kunmap_atomic(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 603) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 604) will_compress = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 605) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 606) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 607) cont:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 608) if (start == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 609) /* lets try to make an inline extent */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 610) if (ret || total_in < actual_end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 611) /* we didn't compress the entire range, try
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 612) * to make an uncompressed inline extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 613) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 614) ret = cow_file_range_inline(BTRFS_I(inode), start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 615) 0, BTRFS_COMPRESS_NONE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 616) NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 617) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 618) /* try making a compressed inline extent */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 619) ret = cow_file_range_inline(BTRFS_I(inode), start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 620) total_compressed,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 621) compress_type, pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 622) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 623) if (ret <= 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 624) unsigned long clear_flags = EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 625) EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 626) EXTENT_DO_ACCOUNTING;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 627) unsigned long page_error_op;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 629) page_error_op = ret < 0 ? PAGE_SET_ERROR : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 630)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 631) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 632) * inline extent creation worked or returned error,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 633) * we don't need to create any more async work items.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 634) * Unlock and free up our temp pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 635) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 636) * We use DO_ACCOUNTING here because we need the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 637) * delalloc_release_metadata to be done _after_ we drop
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 638) * our outstanding extent for clearing delalloc for this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 639) * range.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 640) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 641) extent_clear_unlock_delalloc(BTRFS_I(inode), start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 642) NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 643) clear_flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 644) PAGE_UNLOCK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 645) PAGE_CLEAR_DIRTY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 646) PAGE_SET_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 647) page_error_op |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 648) PAGE_END_WRITEBACK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 649)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 650) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 651) * Ensure we only free the compressed pages if we have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 652) * them allocated, as we can still reach here with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 653) * inode_need_compress() == false.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 654) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 655) if (pages) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 656) for (i = 0; i < nr_pages; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 657) WARN_ON(pages[i]->mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 658) put_page(pages[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 659) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 660) kfree(pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 661) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 662) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 663) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 664) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 665)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 666) if (will_compress) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 667) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 668) * we aren't doing an inline extent round the compressed size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 669) * up to a block size boundary so the allocator does sane
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 670) * things
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 671) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 672) total_compressed = ALIGN(total_compressed, blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 673)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 674) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 675) * one last check to make sure the compression is really a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 676) * win, compare the page count read with the blocks on disk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 677) * compression must free at least one sector size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 678) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 679) total_in = ALIGN(total_in, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 680) if (total_compressed + blocksize <= total_in) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 681) compressed_extents++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 682)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 683) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 684) * The async work queues will take care of doing actual
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 685) * allocation on disk for these compressed pages, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 686) * will submit them to the elevator.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 687) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 688) add_async_extent(async_chunk, start, total_in,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 689) total_compressed, pages, nr_pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 690) compress_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 691)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 692) if (start + total_in < end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 693) start += total_in;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 694) pages = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 695) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 696) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 697) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 698) return compressed_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 699) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 700) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 701) if (pages) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 702) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 703) * the compression code ran but failed to make things smaller,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 704) * free any pages it allocated and our page pointer array
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 705) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 706) for (i = 0; i < nr_pages; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 707) WARN_ON(pages[i]->mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 708) put_page(pages[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 709) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 710) kfree(pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 711) pages = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 712) total_compressed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 713) nr_pages = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 714)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 715) /* flag the file so we don't compress in the future */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 716) if (!btrfs_test_opt(fs_info, FORCE_COMPRESS) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 717) !(BTRFS_I(inode)->prop_compress)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 718) BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 719) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 720) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 721) cleanup_and_bail_uncompressed:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 722) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 723) * No compression, but we still need to write the pages in the file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 724) * we've been given so far. redirty the locked page if it corresponds
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 725) * to our extent and set things up for the async work queue to run
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 726) * cow_file_range to do the normal delalloc dance.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 727) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 728) if (async_chunk->locked_page &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 729) (page_offset(async_chunk->locked_page) >= start &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 730) page_offset(async_chunk->locked_page)) <= end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 731) __set_page_dirty_nobuffers(async_chunk->locked_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 732) /* unlocked later on in the async handlers */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 733) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 735) if (redirty)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 736) extent_range_redirty_for_io(inode, start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 737) add_async_extent(async_chunk, start, end - start + 1, 0, NULL, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 738) BTRFS_COMPRESS_NONE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 739) compressed_extents++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 740)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 741) return compressed_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 742) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 743)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 744) static void free_async_extent_pages(struct async_extent *async_extent)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 745) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 746) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 747)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 748) if (!async_extent->pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 749) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 750)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 751) for (i = 0; i < async_extent->nr_pages; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 752) WARN_ON(async_extent->pages[i]->mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 753) put_page(async_extent->pages[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 754) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 755) kfree(async_extent->pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 756) async_extent->nr_pages = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 757) async_extent->pages = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 758) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 759)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 760) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 761) * phase two of compressed writeback. This is the ordered portion
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 762) * of the code, which only gets called in the order the work was
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 763) * queued. We walk all the async extents created by compress_file_range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 764) * and send them down to the disk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 765) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 766) static noinline void submit_compressed_extents(struct async_chunk *async_chunk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 767) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 768) struct btrfs_inode *inode = BTRFS_I(async_chunk->inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 769) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 770) struct async_extent *async_extent;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 771) u64 alloc_hint = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 772) struct btrfs_key ins;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 773) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 774) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 775) struct extent_io_tree *io_tree = &inode->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 776) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 777)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 778) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 779) while (!list_empty(&async_chunk->extents)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 780) async_extent = list_entry(async_chunk->extents.next,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 781) struct async_extent, list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 782) list_del(&async_extent->list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 783)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 784) retry:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 785) lock_extent(io_tree, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 786) async_extent->start + async_extent->ram_size - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 787) /* did the compression code fall back to uncompressed IO? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 788) if (!async_extent->pages) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 789) int page_started = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 790) unsigned long nr_written = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 791)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 792) /* allocate blocks */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 793) ret = cow_file_range(inode, async_chunk->locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 794) async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 795) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 796) async_extent->ram_size - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 797) &page_started, &nr_written, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 798)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 799) /* JDM XXX */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 801) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 802) * if page_started, cow_file_range inserted an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 803) * inline extent and took care of all the unlocking
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 804) * and IO for us. Otherwise, we need to submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 805) * all those pages down to the drive.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 806) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 807) if (!page_started && !ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 808) extent_write_locked_range(&inode->vfs_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 809) async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 810) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 811) async_extent->ram_size - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 812) WB_SYNC_ALL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 813) else if (ret && async_chunk->locked_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 814) unlock_page(async_chunk->locked_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 815) kfree(async_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 816) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 817) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 818) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 819)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 820) ret = btrfs_reserve_extent(root, async_extent->ram_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 821) async_extent->compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 822) async_extent->compressed_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 823) 0, alloc_hint, &ins, 1, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 824) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 825) free_async_extent_pages(async_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 826)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 827) if (ret == -ENOSPC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 828) unlock_extent(io_tree, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 829) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 830) async_extent->ram_size - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 831)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 832) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 833) * we need to redirty the pages if we decide to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 834) * fallback to uncompressed IO, otherwise we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 835) * will not submit these pages down to lower
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 836) * layers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 837) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 838) extent_range_redirty_for_io(&inode->vfs_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 839) async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 840) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 841) async_extent->ram_size - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 843) goto retry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 844) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 845) goto out_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 846) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 847) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 848) * here we're doing allocation and writeback of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 849) * compressed pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 850) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 851) em = create_io_em(inode, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 852) async_extent->ram_size, /* len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 853) async_extent->start, /* orig_start */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 854) ins.objectid, /* block_start */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 855) ins.offset, /* block_len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 856) ins.offset, /* orig_block_len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 857) async_extent->ram_size, /* ram_bytes */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 858) async_extent->compress_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 859) BTRFS_ORDERED_COMPRESSED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 860) if (IS_ERR(em))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 861) /* ret value is not necessary due to void function */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 862) goto out_free_reserve;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 863) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 864)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 865) ret = btrfs_add_ordered_extent_compress(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 866) async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 867) ins.objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 868) async_extent->ram_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 869) ins.offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 870) BTRFS_ORDERED_COMPRESSED,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 871) async_extent->compress_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 872) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 873) btrfs_drop_extent_cache(inode, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 874) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 875) async_extent->ram_size - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 876) goto out_free_reserve;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 877) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 878) btrfs_dec_block_group_reservations(fs_info, ins.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 879)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 880) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 881) * clear dirty, set writeback and unlock the pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 882) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 883) extent_clear_unlock_delalloc(inode, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 884) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 885) async_extent->ram_size - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 886) NULL, EXTENT_LOCKED | EXTENT_DELALLOC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 887) PAGE_UNLOCK | PAGE_CLEAR_DIRTY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 888) PAGE_SET_WRITEBACK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 889) if (btrfs_submit_compressed_write(inode, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 890) async_extent->ram_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 891) ins.objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 892) ins.offset, async_extent->pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 893) async_extent->nr_pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 894) async_chunk->write_flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 895) async_chunk->blkcg_css)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 896) struct page *p = async_extent->pages[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 897) const u64 start = async_extent->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 898) const u64 end = start + async_extent->ram_size - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 899)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 900) p->mapping = inode->vfs_inode.i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 901) btrfs_writepage_endio_finish_ordered(p, start, end, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 902)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 903) p->mapping = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 904) extent_clear_unlock_delalloc(inode, start, end, NULL, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 905) PAGE_END_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 906) PAGE_SET_ERROR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 907) free_async_extent_pages(async_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 908) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 909) alloc_hint = ins.objectid + ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 910) kfree(async_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 911) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 912) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 913) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 914) out_free_reserve:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 915) btrfs_dec_block_group_reservations(fs_info, ins.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 916) btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 917) out_free:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 918) extent_clear_unlock_delalloc(inode, async_extent->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 919) async_extent->start +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 920) async_extent->ram_size - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 921) NULL, EXTENT_LOCKED | EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 922) EXTENT_DELALLOC_NEW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 923) EXTENT_DEFRAG | EXTENT_DO_ACCOUNTING,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 924) PAGE_UNLOCK | PAGE_CLEAR_DIRTY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 925) PAGE_SET_WRITEBACK | PAGE_END_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 926) PAGE_SET_ERROR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 927) free_async_extent_pages(async_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 928) kfree(async_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 929) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 930) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 932) static u64 get_extent_allocation_hint(struct btrfs_inode *inode, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 933) u64 num_bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 934) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 935) struct extent_map_tree *em_tree = &inode->extent_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 936) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 937) u64 alloc_hint = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 938)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 939) read_lock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 940) em = search_extent_mapping(em_tree, start, num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 941) if (em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 942) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 943) * if block start isn't an actual block number then find the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 944) * first block in this inode and use that as a hint. If that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 945) * block is also bogus then just don't worry about it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 946) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 947) if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 948) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 949) em = search_extent_mapping(em_tree, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 950) if (em && em->block_start < EXTENT_MAP_LAST_BYTE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 951) alloc_hint = em->block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 952) if (em)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 953) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 954) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 955) alloc_hint = em->block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 956) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 957) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 958) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 959) read_unlock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 960)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 961) return alloc_hint;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 962) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 963)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 964) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 965) * when extent_io.c finds a delayed allocation range in the file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 966) * the call backs end up in this code. The basic idea is to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 967) * allocate extents on disk for the range, and create ordered data structs
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 968) * in ram to track those extents.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 969) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 970) * locked_page is the page that writepage had locked already. We use
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 971) * it to make sure we don't do extra locks or unlocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 972) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 973) * *page_started is set to one if we unlock locked_page and do everything
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 974) * required to start IO on it. It may be clean and already done with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 975) * IO when we return.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 976) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 977) static noinline int cow_file_range(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 978) struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 979) u64 start, u64 end, int *page_started,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 980) unsigned long *nr_written, int unlock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 981) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 982) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 983) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 984) u64 alloc_hint = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 985) u64 num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 986) unsigned long ram_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 987) u64 cur_alloc_size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 988) u64 min_alloc_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 989) u64 blocksize = fs_info->sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 990) struct btrfs_key ins;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 991) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 992) unsigned clear_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 993) unsigned long page_ops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 994) bool extent_reserved = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 995) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 996)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 997) if (btrfs_is_free_space_inode(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 998) WARN_ON_ONCE(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 999) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1000) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1001) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1002)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1003) num_bytes = ALIGN(end - start + 1, blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1004) num_bytes = max(blocksize, num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1005) ASSERT(num_bytes <= btrfs_super_total_bytes(fs_info->super_copy));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1006)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1007) inode_should_defrag(inode, start, end, num_bytes, SZ_64K);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1008)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1009) if (start == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1010) /* lets try to make an inline extent */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1011) ret = cow_file_range_inline(inode, start, end, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1012) BTRFS_COMPRESS_NONE, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1013) if (ret == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1014) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1015) * We use DO_ACCOUNTING here because we need the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1016) * delalloc_release_metadata to be run _after_ we drop
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1017) * our outstanding extent for clearing delalloc for this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1018) * range.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1019) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1020) extent_clear_unlock_delalloc(inode, start, end, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1021) EXTENT_LOCKED | EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1022) EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1023) EXTENT_DO_ACCOUNTING, PAGE_UNLOCK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1024) PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1025) PAGE_END_WRITEBACK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1026) *nr_written = *nr_written +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1027) (end - start + PAGE_SIZE) / PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1028) *page_started = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1029) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1030) } else if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1031) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1032) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1033) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1034)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1035) alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1036) btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1038) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1039) * Relocation relies on the relocated extents to have exactly the same
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1040) * size as the original extents. Normally writeback for relocation data
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1041) * extents follows a NOCOW path because relocation preallocates the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1042) * extents. However, due to an operation such as scrub turning a block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1043) * group to RO mode, it may fallback to COW mode, so we must make sure
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1044) * an extent allocated during COW has exactly the requested size and can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1045) * not be split into smaller extents, otherwise relocation breaks and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1046) * fails during the stage where it updates the bytenr of file extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1047) * items.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1048) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1049) if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1050) min_alloc_size = num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1051) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1052) min_alloc_size = fs_info->sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1054) while (num_bytes > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1055) cur_alloc_size = num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1056) ret = btrfs_reserve_extent(root, cur_alloc_size, cur_alloc_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1057) min_alloc_size, 0, alloc_hint,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1058) &ins, 1, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1059) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1060) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1061) cur_alloc_size = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1062) extent_reserved = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1064) ram_size = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1065) em = create_io_em(inode, start, ins.offset, /* len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1066) start, /* orig_start */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1067) ins.objectid, /* block_start */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1068) ins.offset, /* block_len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1069) ins.offset, /* orig_block_len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1070) ram_size, /* ram_bytes */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1071) BTRFS_COMPRESS_NONE, /* compress_type */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1072) BTRFS_ORDERED_REGULAR /* type */);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1073) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1074) ret = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1075) goto out_reserve;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1076) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1077) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1078)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1079) ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1080) ram_size, cur_alloc_size, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1081) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1082) goto out_drop_extent_cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1084) if (root->root_key.objectid ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1085) BTRFS_DATA_RELOC_TREE_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1086) ret = btrfs_reloc_clone_csums(inode, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1087) cur_alloc_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1088) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1089) * Only drop cache here, and process as normal.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1090) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1091) * We must not allow extent_clear_unlock_delalloc()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1092) * at out_unlock label to free meta of this ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1093) * extent, as its meta should be freed by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1094) * btrfs_finish_ordered_io().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1095) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1096) * So we must continue until @start is increased to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1097) * skip current ordered extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1098) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1099) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1100) btrfs_drop_extent_cache(inode, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1101) start + ram_size - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1104) btrfs_dec_block_group_reservations(fs_info, ins.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1106) /* we're not doing compressed IO, don't unlock the first
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1107) * page (which the caller expects to stay locked), don't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1108) * clear any dirty bits and don't set any writeback bits
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1109) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1110) * Do set the Private2 bit so we know this page was properly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1111) * setup for writepage
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1112) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1113) page_ops = unlock ? PAGE_UNLOCK : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1114) page_ops |= PAGE_SET_PRIVATE2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1116) extent_clear_unlock_delalloc(inode, start, start + ram_size - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1117) locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1118) EXTENT_LOCKED | EXTENT_DELALLOC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1119) page_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1120) if (num_bytes < cur_alloc_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1121) num_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1122) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1123) num_bytes -= cur_alloc_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1124) alloc_hint = ins.objectid + ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1125) start += cur_alloc_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1126) extent_reserved = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1127)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1128) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1129) * btrfs_reloc_clone_csums() error, since start is increased
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1130) * extent_clear_unlock_delalloc() at out_unlock label won't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1131) * free metadata of current ordered extent, we're OK to exit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1132) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1133) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1134) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1135) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1136) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1137) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1139) out_drop_extent_cache:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1140) btrfs_drop_extent_cache(inode, start, start + ram_size - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1141) out_reserve:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1142) btrfs_dec_block_group_reservations(fs_info, ins.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1143) btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1144) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1145) clear_bits = EXTENT_LOCKED | EXTENT_DELALLOC | EXTENT_DELALLOC_NEW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1146) EXTENT_DEFRAG | EXTENT_CLEAR_META_RESV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1147) page_ops = PAGE_UNLOCK | PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1148) PAGE_END_WRITEBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1149) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1150) * If we reserved an extent for our delalloc range (or a subrange) and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1151) * failed to create the respective ordered extent, then it means that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1152) * when we reserved the extent we decremented the extent's size from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1153) * the data space_info's bytes_may_use counter and incremented the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1154) * space_info's bytes_reserved counter by the same amount. We must make
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1155) * sure extent_clear_unlock_delalloc() does not try to decrement again
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1156) * the data space_info's bytes_may_use counter, therefore we do not pass
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1157) * it the flag EXTENT_CLEAR_DATA_RESV.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1158) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1159) if (extent_reserved) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1160) extent_clear_unlock_delalloc(inode, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1161) start + cur_alloc_size - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1162) locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1163) clear_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1164) page_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1165) start += cur_alloc_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1166) if (start >= end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1167) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1168) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1169) extent_clear_unlock_delalloc(inode, start, end, locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1170) clear_bits | EXTENT_CLEAR_DATA_RESV,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1171) page_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1172) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1173) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1174)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1175) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1176) * work queue call back to started compression on a file and pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1177) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1178) static noinline void async_cow_start(struct btrfs_work *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1179) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1180) struct async_chunk *async_chunk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1181) int compressed_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1183) async_chunk = container_of(work, struct async_chunk, work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1184)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1185) compressed_extents = compress_file_range(async_chunk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1186) if (compressed_extents == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1187) btrfs_add_delayed_iput(async_chunk->inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1188) async_chunk->inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1189) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1190) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1192) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1193) * work queue call back to submit previously compressed pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1194) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1195) static noinline void async_cow_submit(struct btrfs_work *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1196) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1197) struct async_chunk *async_chunk = container_of(work, struct async_chunk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1198) work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1199) struct btrfs_fs_info *fs_info = btrfs_work_owner(work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1200) unsigned long nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1201)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1202) nr_pages = (async_chunk->end - async_chunk->start + PAGE_SIZE) >>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1203) PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1204)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1205) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1206) * ->inode could be NULL if async_chunk_start has failed to compress,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1207) * in which case we don't have anything to submit, yet we need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1208) * always adjust ->async_delalloc_pages as its paired with the init
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1209) * happening in cow_file_range_async
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1210) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1211) if (async_chunk->inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1212) submit_compressed_extents(async_chunk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1213)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1214) /* atomic_sub_return implies a barrier */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1215) if (atomic_sub_return(nr_pages, &fs_info->async_delalloc_pages) <
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1216) 5 * SZ_1M)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1217) cond_wake_up_nomb(&fs_info->async_submit_wait);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1218) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1219)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1220) static noinline void async_cow_free(struct btrfs_work *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1221) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1222) struct async_chunk *async_chunk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1223)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1224) async_chunk = container_of(work, struct async_chunk, work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1225) if (async_chunk->inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1226) btrfs_add_delayed_iput(async_chunk->inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1227) if (async_chunk->blkcg_css)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1228) css_put(async_chunk->blkcg_css);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1229) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1230) * Since the pointer to 'pending' is at the beginning of the array of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1231) * async_chunk's, freeing it ensures the whole array has been freed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1232) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1233) if (atomic_dec_and_test(async_chunk->pending))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1234) kvfree(async_chunk->pending);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1235) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1236)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1237) static int cow_file_range_async(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1238) struct writeback_control *wbc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1239) struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1240) u64 start, u64 end, int *page_started,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1241) unsigned long *nr_written)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1242) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1243) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1244) struct cgroup_subsys_state *blkcg_css = wbc_blkcg_css(wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1245) struct async_cow *ctx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1246) struct async_chunk *async_chunk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1247) unsigned long nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1248) u64 cur_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1249) u64 num_chunks = DIV_ROUND_UP(end - start, SZ_512K);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1250) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1251) bool should_compress;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1252) unsigned nofs_flag;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1253) const unsigned int write_flags = wbc_to_write_flags(wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1254)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1255) unlock_extent(&inode->io_tree, start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1256)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1257) if (inode->flags & BTRFS_INODE_NOCOMPRESS &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1258) !btrfs_test_opt(fs_info, FORCE_COMPRESS)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1259) num_chunks = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1260) should_compress = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1261) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1262) should_compress = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1263) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1264)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1265) nofs_flag = memalloc_nofs_save();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1266) ctx = kvmalloc(struct_size(ctx, chunks, num_chunks), GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1267) memalloc_nofs_restore(nofs_flag);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1268)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1269) if (!ctx) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1270) unsigned clear_bits = EXTENT_LOCKED | EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1271) EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1272) EXTENT_DO_ACCOUNTING;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1273) unsigned long page_ops = PAGE_UNLOCK | PAGE_CLEAR_DIRTY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1274) PAGE_SET_WRITEBACK | PAGE_END_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1275) PAGE_SET_ERROR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1276)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1277) extent_clear_unlock_delalloc(inode, start, end, locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1278) clear_bits, page_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1279) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1280) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1282) async_chunk = ctx->chunks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1283) atomic_set(&ctx->num_chunks, num_chunks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1284)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1285) for (i = 0; i < num_chunks; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1286) if (should_compress)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1287) cur_end = min(end, start + SZ_512K - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1288) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1289) cur_end = end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1290)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1291) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1292) * igrab is called higher up in the call chain, take only the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1293) * lightweight reference for the callback lifetime
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1294) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1295) ihold(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1296) async_chunk[i].pending = &ctx->num_chunks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1297) async_chunk[i].inode = &inode->vfs_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1298) async_chunk[i].start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1299) async_chunk[i].end = cur_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1300) async_chunk[i].write_flags = write_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1301) INIT_LIST_HEAD(&async_chunk[i].extents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1302)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1303) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1304) * The locked_page comes all the way from writepage and its
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1305) * the original page we were actually given. As we spread
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1306) * this large delalloc region across multiple async_chunk
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1307) * structs, only the first struct needs a pointer to locked_page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1308) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1309) * This way we don't need racey decisions about who is supposed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1310) * to unlock it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1311) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1312) if (locked_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1313) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1314) * Depending on the compressibility, the pages might or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1315) * might not go through async. We want all of them to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1316) * be accounted against wbc once. Let's do it here
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1317) * before the paths diverge. wbc accounting is used
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1318) * only for foreign writeback detection and doesn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1319) * need full accuracy. Just account the whole thing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1320) * against the first page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1321) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1322) wbc_account_cgroup_owner(wbc, locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1323) cur_end - start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1324) async_chunk[i].locked_page = locked_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1325) locked_page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1326) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1327) async_chunk[i].locked_page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1328) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1330) if (blkcg_css != blkcg_root_css) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1331) css_get(blkcg_css);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1332) async_chunk[i].blkcg_css = blkcg_css;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1333) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1334) async_chunk[i].blkcg_css = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1335) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1336)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1337) btrfs_init_work(&async_chunk[i].work, async_cow_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1338) async_cow_submit, async_cow_free);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1339)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1340) nr_pages = DIV_ROUND_UP(cur_end - start, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1341) atomic_add(nr_pages, &fs_info->async_delalloc_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1342)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1343) btrfs_queue_work(fs_info->delalloc_workers, &async_chunk[i].work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1344)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1345) *nr_written += nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1346) start = cur_end + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1347) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1348) *page_started = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1349) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1350) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1351)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1352) static noinline int csum_exist_in_range(struct btrfs_fs_info *fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1353) u64 bytenr, u64 num_bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1354) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1355) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1356) struct btrfs_ordered_sum *sums;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1357) LIST_HEAD(list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1358)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1359) ret = btrfs_lookup_csums_range(fs_info->csum_root, bytenr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1360) bytenr + num_bytes - 1, &list, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1361) if (ret == 0 && list_empty(&list))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1362) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1364) while (!list_empty(&list)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1365) sums = list_entry(list.next, struct btrfs_ordered_sum, list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1366) list_del(&sums->list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1367) kfree(sums);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1368) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1369) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1370) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1371) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1372) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1373)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1374) static int fallback_to_cow(struct btrfs_inode *inode, struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1375) const u64 start, const u64 end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1376) int *page_started, unsigned long *nr_written)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1377) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1378) const bool is_space_ino = btrfs_is_free_space_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1379) const bool is_reloc_ino = (inode->root->root_key.objectid ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1380) BTRFS_DATA_RELOC_TREE_OBJECTID);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1381) const u64 range_bytes = end + 1 - start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1382) struct extent_io_tree *io_tree = &inode->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1383) u64 range_start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1384) u64 count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1385)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1386) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1387) * If EXTENT_NORESERVE is set it means that when the buffered write was
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1388) * made we had not enough available data space and therefore we did not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1389) * reserve data space for it, since we though we could do NOCOW for the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1390) * respective file range (either there is prealloc extent or the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1391) * has the NOCOW bit set).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1392) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1393) * However when we need to fallback to COW mode (because for example the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1394) * block group for the corresponding extent was turned to RO mode by a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1395) * scrub or relocation) we need to do the following:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1396) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1397) * 1) We increment the bytes_may_use counter of the data space info.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1398) * If COW succeeds, it allocates a new data extent and after doing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1399) * that it decrements the space info's bytes_may_use counter and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1400) * increments its bytes_reserved counter by the same amount (we do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1401) * this at btrfs_add_reserved_bytes()). So we need to increment the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1402) * bytes_may_use counter to compensate (when space is reserved at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1403) * buffered write time, the bytes_may_use counter is incremented);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1404) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1405) * 2) We clear the EXTENT_NORESERVE bit from the range. We do this so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1406) * that if the COW path fails for any reason, it decrements (through
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1407) * extent_clear_unlock_delalloc()) the bytes_may_use counter of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1408) * data space info, which we incremented in the step above.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1409) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1410) * If we need to fallback to cow and the inode corresponds to a free
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1411) * space cache inode or an inode of the data relocation tree, we must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1412) * also increment bytes_may_use of the data space_info for the same
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1413) * reason. Space caches and relocated data extents always get a prealloc
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1414) * extent for them, however scrub or balance may have set the block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1415) * group that contains that extent to RO mode and therefore force COW
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1416) * when starting writeback.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1417) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1418) count = count_range_bits(io_tree, &range_start, end, range_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1419) EXTENT_NORESERVE, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1420) if (count > 0 || is_space_ino || is_reloc_ino) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1421) u64 bytes = count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1422) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1423) struct btrfs_space_info *sinfo = fs_info->data_sinfo;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1424)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1425) if (is_space_ino || is_reloc_ino)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1426) bytes = range_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1427)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1428) spin_lock(&sinfo->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1429) btrfs_space_info_update_bytes_may_use(fs_info, sinfo, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1430) spin_unlock(&sinfo->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1431)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1432) if (count > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1433) clear_extent_bit(io_tree, start, end, EXTENT_NORESERVE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1434) 0, 0, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1435) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1436)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1437) return cow_file_range(inode, locked_page, start, end, page_started,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1438) nr_written, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1439) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1441) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1442) * when nowcow writeback call back. This checks for snapshots or COW copies
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1443) * of the extents that exist in the file, and COWs the file as required.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1444) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1445) * If no cow copies or snapshots exist, we write directly to the existing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1446) * blocks on disk
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1447) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1448) static noinline int run_delalloc_nocow(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1449) struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1450) const u64 start, const u64 end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1451) int *page_started, int force,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1452) unsigned long *nr_written)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1453) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1454) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1455) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1456) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1457) u64 cow_start = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1458) u64 cur_offset = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1459) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1460) bool check_prev = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1461) const bool freespace_inode = btrfs_is_free_space_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1462) u64 ino = btrfs_ino(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1463) bool nocow = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1464) u64 disk_bytenr = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1465)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1466) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1467) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1468) extent_clear_unlock_delalloc(inode, start, end, locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1469) EXTENT_LOCKED | EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1470) EXTENT_DO_ACCOUNTING |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1471) EXTENT_DEFRAG, PAGE_UNLOCK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1472) PAGE_CLEAR_DIRTY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1473) PAGE_SET_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1474) PAGE_END_WRITEBACK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1475) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1476) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1477)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1478) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1479) struct btrfs_key found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1480) struct btrfs_file_extent_item *fi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1481) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1482) u64 extent_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1483) u64 extent_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1484) u64 num_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1485) u64 disk_num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1486) u64 ram_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1487) int extent_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1489) nocow = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1490)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1491) ret = btrfs_lookup_file_extent(NULL, root, path, ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1492) cur_offset, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1493) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1494) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1495)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1496) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1497) * If there is no extent for our range when doing the initial
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1498) * search, then go back to the previous slot as it will be the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1499) * one containing the search offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1500) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1501) if (ret > 0 && path->slots[0] > 0 && check_prev) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1502) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1503) btrfs_item_key_to_cpu(leaf, &found_key,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1504) path->slots[0] - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1505) if (found_key.objectid == ino &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1506) found_key.type == BTRFS_EXTENT_DATA_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1507) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1508) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1509) check_prev = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1510) next_slot:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1511) /* Go to next leaf if we have exhausted the current one */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1512) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1513) if (path->slots[0] >= btrfs_header_nritems(leaf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1514) ret = btrfs_next_leaf(root, path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1515) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1516) if (cow_start != (u64)-1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1517) cur_offset = cow_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1518) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1519) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1520) if (ret > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1521) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1522) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1523) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1525) btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1527) /* Didn't find anything for our INO */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1528) if (found_key.objectid > ino)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1529) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1530) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1531) * Keep searching until we find an EXTENT_ITEM or there are no
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1532) * more extents for this inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1533) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1534) if (WARN_ON_ONCE(found_key.objectid < ino) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1535) found_key.type < BTRFS_EXTENT_DATA_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1536) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1537) goto next_slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1538) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1540) /* Found key is not EXTENT_DATA_KEY or starts after req range */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1541) if (found_key.type > BTRFS_EXTENT_DATA_KEY ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1542) found_key.offset > end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1543) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1544)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1545) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1546) * If the found extent starts after requested offset, then
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1547) * adjust extent_end to be right before this extent begins
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1548) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1549) if (found_key.offset > cur_offset) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1550) extent_end = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1551) extent_type = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1552) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1553) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1554)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1555) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1556) * Found extent which begins before our range and potentially
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1557) * intersect it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1558) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1559) fi = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1560) struct btrfs_file_extent_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1561) extent_type = btrfs_file_extent_type(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1562)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1563) ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1564) if (extent_type == BTRFS_FILE_EXTENT_REG ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1565) extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1566) disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1567) extent_offset = btrfs_file_extent_offset(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1568) extent_end = found_key.offset +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1569) btrfs_file_extent_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1570) disk_num_bytes =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1571) btrfs_file_extent_disk_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1572) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1573) * If the extent we got ends before our current offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1574) * skip to the next extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1575) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1576) if (extent_end <= cur_offset) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1577) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1578) goto next_slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1579) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1580) /* Skip holes */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1581) if (disk_bytenr == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1582) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1583) /* Skip compressed/encrypted/encoded extents */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1584) if (btrfs_file_extent_compression(leaf, fi) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1585) btrfs_file_extent_encryption(leaf, fi) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1586) btrfs_file_extent_other_encoding(leaf, fi))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1587) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1588) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1589) * If extent is created before the last volume's snapshot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1590) * this implies the extent is shared, hence we can't do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1591) * nocow. This is the same check as in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1592) * btrfs_cross_ref_exist but without calling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1593) * btrfs_search_slot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1594) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1595) if (!freespace_inode &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1596) btrfs_file_extent_generation(leaf, fi) <=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1597) btrfs_root_last_snapshot(&root->root_item))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1598) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1599) if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1600) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1601) /* If extent is RO, we must COW it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1602) if (btrfs_extent_readonly(fs_info, disk_bytenr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1603) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1604) ret = btrfs_cross_ref_exist(root, ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1605) found_key.offset -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1606) extent_offset, disk_bytenr, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1607) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1608) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1609) * ret could be -EIO if the above fails to read
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1610) * metadata.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1611) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1612) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1613) if (cow_start != (u64)-1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1614) cur_offset = cow_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1615) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1618) WARN_ON_ONCE(freespace_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1619) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1620) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1621) disk_bytenr += extent_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1622) disk_bytenr += cur_offset - found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1623) num_bytes = min(end + 1, extent_end) - cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1624) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1625) * If there are pending snapshots for this root, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1626) * fall into common COW way
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1627) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1628) if (!freespace_inode && atomic_read(&root->snapshot_force_cow))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1629) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1630) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1631) * force cow if csum exists in the range.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1632) * this ensure that csum for a given extent are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1633) * either valid or do not exist.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1634) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1635) ret = csum_exist_in_range(fs_info, disk_bytenr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1636) num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1637) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1638) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1639) * ret could be -EIO if the above fails to read
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1640) * metadata.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1641) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1642) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1643) if (cow_start != (u64)-1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1644) cur_offset = cow_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1645) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1646) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1647) WARN_ON_ONCE(freespace_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1648) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1649) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1650) if (!btrfs_inc_nocow_writers(fs_info, disk_bytenr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1651) goto out_check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1652) nocow = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1653) } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1654) extent_end = found_key.offset + ram_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1655) extent_end = ALIGN(extent_end, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1656) /* Skip extents outside of our requested range */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1657) if (extent_end <= start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1658) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1659) goto next_slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1660) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1661) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1662) /* If this triggers then we have a memory corruption */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1663) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1664) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1665) out_check:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1666) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1667) * If nocow is false then record the beginning of the range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1668) * that needs to be COWed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1669) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1670) if (!nocow) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1671) if (cow_start == (u64)-1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1672) cow_start = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1673) cur_offset = extent_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1674) if (cur_offset > end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1675) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1676) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1677) goto next_slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1678) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1679)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1680) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1681)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1682) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1683) * COW range from cow_start to found_key.offset - 1. As the key
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1684) * will contain the beginning of the first extent that can be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1685) * NOCOW, following one which needs to be COW'ed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1686) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1687) if (cow_start != (u64)-1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1688) ret = fallback_to_cow(inode, locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1689) cow_start, found_key.offset - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1690) page_started, nr_written);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1691) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1692) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1693) cow_start = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1694) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1695)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1696) if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1697) u64 orig_start = found_key.offset - extent_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1698) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1699)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1700) em = create_io_em(inode, cur_offset, num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1701) orig_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1702) disk_bytenr, /* block_start */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1703) num_bytes, /* block_len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1704) disk_num_bytes, /* orig_block_len */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1705) ram_bytes, BTRFS_COMPRESS_NONE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1706) BTRFS_ORDERED_PREALLOC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1707) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1708) ret = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1709) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1710) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1711) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1712) ret = btrfs_add_ordered_extent(inode, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1713) disk_bytenr, num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1714) num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1715) BTRFS_ORDERED_PREALLOC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1716) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1717) btrfs_drop_extent_cache(inode, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1718) cur_offset + num_bytes - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1719) 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1720) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1721) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1722) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1723) ret = btrfs_add_ordered_extent(inode, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1724) disk_bytenr, num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1725) num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1726) BTRFS_ORDERED_NOCOW);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1727) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1728) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1731) if (nocow)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1732) btrfs_dec_nocow_writers(fs_info, disk_bytenr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1733) nocow = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1735) if (root->root_key.objectid ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1736) BTRFS_DATA_RELOC_TREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1737) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1738) * Error handled later, as we must prevent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1739) * extent_clear_unlock_delalloc() in error handler
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1740) * from freeing metadata of created ordered extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1741) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1742) ret = btrfs_reloc_clone_csums(inode, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1743) num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1744)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1745) extent_clear_unlock_delalloc(inode, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1746) cur_offset + num_bytes - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1747) locked_page, EXTENT_LOCKED |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1748) EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1749) EXTENT_CLEAR_DATA_RESV,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1750) PAGE_UNLOCK | PAGE_SET_PRIVATE2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1751)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1752) cur_offset = extent_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1753)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1754) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1755) * btrfs_reloc_clone_csums() error, now we're OK to call error
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1756) * handler, as metadata for created ordered extent will only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1757) * be freed by btrfs_finish_ordered_io().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1758) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1759) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1760) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1761) if (cur_offset > end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1762) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1763) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1764) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1765)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1766) if (cur_offset <= end && cow_start == (u64)-1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1767) cow_start = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1768)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1769) if (cow_start != (u64)-1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1770) cur_offset = end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1771) ret = fallback_to_cow(inode, locked_page, cow_start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1772) page_started, nr_written);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1773) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1774) goto error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1775) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1776)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1777) error:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1778) if (nocow)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1779) btrfs_dec_nocow_writers(fs_info, disk_bytenr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1780)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1781) if (ret && cur_offset < end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1782) extent_clear_unlock_delalloc(inode, cur_offset, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1783) locked_page, EXTENT_LOCKED |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1784) EXTENT_DELALLOC | EXTENT_DEFRAG |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1785) EXTENT_DO_ACCOUNTING, PAGE_UNLOCK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1786) PAGE_CLEAR_DIRTY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1787) PAGE_SET_WRITEBACK |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1788) PAGE_END_WRITEBACK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1789) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1790) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1791) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1792)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1793) static inline int need_force_cow(struct btrfs_inode *inode, u64 start, u64 end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1794) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1795)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1796) if (!(inode->flags & BTRFS_INODE_NODATACOW) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1797) !(inode->flags & BTRFS_INODE_PREALLOC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1798) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1799)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1800) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1801) * @defrag_bytes is a hint value, no spinlock held here,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1802) * if is not zero, it means the file is defragging.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1803) * Force cow if given extent needs to be defragged.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1804) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1805) if (inode->defrag_bytes &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1806) test_range_bit(&inode->io_tree, start, end, EXTENT_DEFRAG, 0, NULL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1807) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1809) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1810) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1811)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1812) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1813) * Function to process delayed allocation (create CoW) for ranges which are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1814) * being touched for the first time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1815) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1816) int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1817) u64 start, u64 end, int *page_started, unsigned long *nr_written,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1818) struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1819) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1820) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1821) int force_cow = need_force_cow(inode, start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1822)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1823) if (inode->flags & BTRFS_INODE_NODATACOW && !force_cow) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1824) ret = run_delalloc_nocow(inode, locked_page, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1825) page_started, 1, nr_written);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1826) } else if (inode->flags & BTRFS_INODE_PREALLOC && !force_cow) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1827) ret = run_delalloc_nocow(inode, locked_page, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1828) page_started, 0, nr_written);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1829) } else if (!inode_can_compress(inode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1830) !inode_need_compress(inode, start, end)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1831) ret = cow_file_range(inode, locked_page, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1832) page_started, nr_written, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1833) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1834) set_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, &inode->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1835) ret = cow_file_range_async(inode, wbc, locked_page, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1836) page_started, nr_written);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1837) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1838) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1839) btrfs_cleanup_ordered_extents(inode, locked_page, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1840) end - start + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1841) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1842) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1843)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1844) void btrfs_split_delalloc_extent(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1845) struct extent_state *orig, u64 split)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1846) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1847) u64 size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1849) /* not delalloc, ignore it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1850) if (!(orig->state & EXTENT_DELALLOC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1851) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1852)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1853) size = orig->end - orig->start + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1854) if (size > BTRFS_MAX_EXTENT_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1855) u32 num_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1856) u64 new_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1857)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1858) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1859) * See the explanation in btrfs_merge_delalloc_extent, the same
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1860) * applies here, just in reverse.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1861) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1862) new_size = orig->end - split + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1863) num_extents = count_max_extents(new_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1864) new_size = split - orig->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1865) num_extents += count_max_extents(new_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1866) if (count_max_extents(size) >= num_extents)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1867) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1868) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1870) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1871) btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1872) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1875) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1876) * Handle merged delayed allocation extents so we can keep track of new extents
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1877) * that are just merged onto old extents, such as when we are doing sequential
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1878) * writes, so we can properly account for the metadata space we'll need.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1879) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1880) void btrfs_merge_delalloc_extent(struct inode *inode, struct extent_state *new,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1881) struct extent_state *other)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1882) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1883) u64 new_size, old_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1884) u32 num_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1885)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1886) /* not delalloc, ignore it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1887) if (!(other->state & EXTENT_DELALLOC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1888) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1889)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1890) if (new->start > other->start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1891) new_size = new->end - other->start + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1892) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1893) new_size = other->end - new->start + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1894)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1895) /* we're not bigger than the max, unreserve the space and go */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1896) if (new_size <= BTRFS_MAX_EXTENT_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1897) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1898) btrfs_mod_outstanding_extents(BTRFS_I(inode), -1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1899) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1900) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1901) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1902)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1903) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1904) * We have to add up either side to figure out how many extents were
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1905) * accounted for before we merged into one big extent. If the number of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1906) * extents we accounted for is <= the amount we need for the new range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1907) * then we can return, otherwise drop. Think of it like this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1908) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1909) * [ 4k][MAX_SIZE]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1910) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1911) * So we've grown the extent by a MAX_SIZE extent, this would mean we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1912) * need 2 outstanding extents, on one side we have 1 and the other side
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1913) * we have 1 so they are == and we can return. But in this case
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1914) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1915) * [MAX_SIZE+4k][MAX_SIZE+4k]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1916) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1917) * Each range on their own accounts for 2 extents, but merged together
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1918) * they are only 3 extents worth of accounting, so we need to drop in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1919) * this case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1920) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1921) old_size = other->end - other->start + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1922) num_extents = count_max_extents(old_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1923) old_size = new->end - new->start + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1924) num_extents += count_max_extents(old_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1925) if (count_max_extents(new_size) >= num_extents)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1926) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1927)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1928) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1929) btrfs_mod_outstanding_extents(BTRFS_I(inode), -1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1930) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1931) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1932)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1933) static void btrfs_add_delalloc_inodes(struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1934) struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1935) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1936) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1937)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1938) spin_lock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1939) if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1940) list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1941) &root->delalloc_inodes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1942) set_bit(BTRFS_INODE_IN_DELALLOC_LIST,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1943) &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1944) root->nr_delalloc_inodes++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1945) if (root->nr_delalloc_inodes == 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1946) spin_lock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1947) BUG_ON(!list_empty(&root->delalloc_root));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1948) list_add_tail(&root->delalloc_root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1949) &fs_info->delalloc_roots);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1950) spin_unlock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1951) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1952) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1953) spin_unlock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1954) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1955)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1956)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1957) void __btrfs_del_delalloc_inode(struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1958) struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1959) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1960) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1962) if (!list_empty(&inode->delalloc_inodes)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1963) list_del_init(&inode->delalloc_inodes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1964) clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1965) &inode->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1966) root->nr_delalloc_inodes--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1967) if (!root->nr_delalloc_inodes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1968) ASSERT(list_empty(&root->delalloc_inodes));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1969) spin_lock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1970) BUG_ON(list_empty(&root->delalloc_root));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1971) list_del_init(&root->delalloc_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1972) spin_unlock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1973) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1974) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1975) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1977) static void btrfs_del_delalloc_inode(struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1978) struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1979) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1980) spin_lock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1981) __btrfs_del_delalloc_inode(root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1982) spin_unlock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1983) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1984)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1985) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1986) * Properly track delayed allocation bytes in the inode and to maintain the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1987) * list of inodes that have pending delalloc work to be done.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1988) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1989) void btrfs_set_delalloc_extent(struct inode *inode, struct extent_state *state,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1990) unsigned *bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1991) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1992) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1994) if ((*bits & EXTENT_DEFRAG) && !(*bits & EXTENT_DELALLOC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1995) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1996) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1997) * set_bit and clear bit hooks normally require _irqsave/restore
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1998) * but in this case, we are only testing for the DELALLOC
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1999) * bit, which is only set or cleared with irqs on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2000) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2001) if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2002) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2003) u64 len = state->end + 1 - state->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2004) u32 num_extents = count_max_extents(len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2005) bool do_list = !btrfs_is_free_space_inode(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2006)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2007) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2008) btrfs_mod_outstanding_extents(BTRFS_I(inode), num_extents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2009) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2010)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2011) /* For sanity tests */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2012) if (btrfs_is_testing(fs_info))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2013) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2015) percpu_counter_add_batch(&fs_info->delalloc_bytes, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2016) fs_info->delalloc_batch);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2017) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2018) BTRFS_I(inode)->delalloc_bytes += len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2019) if (*bits & EXTENT_DEFRAG)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2020) BTRFS_I(inode)->defrag_bytes += len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2021) if (do_list && !test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2022) &BTRFS_I(inode)->runtime_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2023) btrfs_add_delalloc_inodes(root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2024) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2025) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2026)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2027) if (!(state->state & EXTENT_DELALLOC_NEW) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2028) (*bits & EXTENT_DELALLOC_NEW)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2029) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2030) BTRFS_I(inode)->new_delalloc_bytes += state->end + 1 -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2031) state->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2032) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2033) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2034) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2035)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2036) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2037) * Once a range is no longer delalloc this function ensures that proper
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2038) * accounting happens.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2039) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2040) void btrfs_clear_delalloc_extent(struct inode *vfs_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2041) struct extent_state *state, unsigned *bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2042) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2043) struct btrfs_inode *inode = BTRFS_I(vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2044) struct btrfs_fs_info *fs_info = btrfs_sb(vfs_inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2045) u64 len = state->end + 1 - state->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2046) u32 num_extents = count_max_extents(len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2047)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2048) if ((state->state & EXTENT_DEFRAG) && (*bits & EXTENT_DEFRAG)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2049) spin_lock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2050) inode->defrag_bytes -= len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2051) spin_unlock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2052) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2054) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2055) * set_bit and clear bit hooks normally require _irqsave/restore
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2056) * but in this case, we are only testing for the DELALLOC
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2057) * bit, which is only set or cleared with irqs on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2058) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2059) if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2060) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2061) bool do_list = !btrfs_is_free_space_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2062)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2063) spin_lock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2064) btrfs_mod_outstanding_extents(inode, -num_extents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2065) spin_unlock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2066)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2067) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2068) * We don't reserve metadata space for space cache inodes so we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2069) * don't need to call delalloc_release_metadata if there is an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2070) * error.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2071) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2072) if (*bits & EXTENT_CLEAR_META_RESV &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2073) root != fs_info->tree_root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2074) btrfs_delalloc_release_metadata(inode, len, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2075)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2076) /* For sanity tests. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2077) if (btrfs_is_testing(fs_info))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2078) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2079)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2080) if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2081) do_list && !(state->state & EXTENT_NORESERVE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2082) (*bits & EXTENT_CLEAR_DATA_RESV))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2083) btrfs_free_reserved_data_space_noquota(fs_info, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2084)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2085) percpu_counter_add_batch(&fs_info->delalloc_bytes, -len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2086) fs_info->delalloc_batch);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2087) spin_lock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2088) inode->delalloc_bytes -= len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2089) if (do_list && inode->delalloc_bytes == 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2090) test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2091) &inode->runtime_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2092) btrfs_del_delalloc_inode(root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2093) spin_unlock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2094) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2095)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2096) if ((state->state & EXTENT_DELALLOC_NEW) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2097) (*bits & EXTENT_DELALLOC_NEW)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2098) spin_lock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2099) ASSERT(inode->new_delalloc_bytes >= len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2100) inode->new_delalloc_bytes -= len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2101) spin_unlock(&inode->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2103) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2104)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2105) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2106) * btrfs_bio_fits_in_stripe - Checks whether the size of the given bio will fit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2107) * in a chunk's stripe. This function ensures that bios do not span a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2108) * stripe/chunk
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2109) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2110) * @page - The page we are about to add to the bio
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2111) * @size - size we want to add to the bio
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2112) * @bio - bio we want to ensure is smaller than a stripe
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2113) * @bio_flags - flags of the bio
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2114) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2115) * return 1 if page cannot be added to the bio
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2116) * return 0 if page can be added to the bio
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2117) * return error otherwise
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2118) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2119) int btrfs_bio_fits_in_stripe(struct page *page, size_t size, struct bio *bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2120) unsigned long bio_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2121) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2122) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2123) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2124) u64 logical = (u64)bio->bi_iter.bi_sector << 9;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2125) u64 length = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2126) u64 map_length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2127) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2128) struct btrfs_io_geometry geom;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2129)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2130) if (bio_flags & EXTENT_BIO_COMPRESSED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2131) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2132)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2133) length = bio->bi_iter.bi_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2134) map_length = length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2135) ret = btrfs_get_io_geometry(fs_info, btrfs_op(bio), logical, map_length,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2136) &geom);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2137) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2138) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2140) if (geom.len < length + size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2141) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2142) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2143) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2145) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2146) * in order to insert checksums into the metadata in large chunks,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2147) * we wait until bio submission time. All the pages in the bio are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2148) * checksummed and sums are attached onto the ordered extent record.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2149) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2150) * At IO completion time the cums attached on the ordered extent record
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2151) * are inserted into the btree
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2152) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2153) static blk_status_t btrfs_submit_bio_start(void *private_data, struct bio *bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2154) u64 bio_offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2155) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2156) struct inode *inode = private_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2158) return btrfs_csum_one_bio(BTRFS_I(inode), bio, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2159) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2160)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2161) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2162) * extent_io.c submission hook. This does the right thing for csum calculation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2163) * on write, or reading the csums from the tree before a read.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2164) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2165) * Rules about async/sync submit,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2166) * a) read: sync submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2167) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2168) * b) write without checksum: sync submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2169) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2170) * c) write with checksum:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2171) * c-1) if bio is issued by fsync: sync submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2172) * (sync_writers != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2173) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2174) * c-2) if root is reloc root: sync submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2175) * (only in case of buffered IO)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2176) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2177) * c-3) otherwise: async submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2178) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2179) blk_status_t btrfs_submit_data_bio(struct inode *inode, struct bio *bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2180) int mirror_num, unsigned long bio_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2181)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2182) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2183) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2184) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2185) enum btrfs_wq_endio_type metadata = BTRFS_WQ_ENDIO_DATA;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2186) blk_status_t ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2187) int skip_sum;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2188) int async = !atomic_read(&BTRFS_I(inode)->sync_writers);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2189)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2190) skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2192) if (btrfs_is_free_space_inode(BTRFS_I(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2193) metadata = BTRFS_WQ_ENDIO_FREE_SPACE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2194)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2195) if (bio_op(bio) != REQ_OP_WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2196) ret = btrfs_bio_wq_end_io(fs_info, bio, metadata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2197) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2198) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2200) if (bio_flags & EXTENT_BIO_COMPRESSED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2201) ret = btrfs_submit_compressed_read(inode, bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2202) mirror_num,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2203) bio_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2204) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2205) } else if (!skip_sum) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2206) ret = btrfs_lookup_bio_sums(inode, bio, (u64)-1, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2207) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2208) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2209) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2210) goto mapit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2211) } else if (async && !skip_sum) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2212) /* csum items have already been cloned */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2213) if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2214) goto mapit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2215) /* we're doing a write, do the async checksumming */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2216) ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, bio_flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2217) 0, inode, btrfs_submit_bio_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2218) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2219) } else if (!skip_sum) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2220) ret = btrfs_csum_one_bio(BTRFS_I(inode), bio, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2221) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2222) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2223) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2224)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2225) mapit:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2226) ret = btrfs_map_bio(fs_info, bio, mirror_num);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2227)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2228) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2229) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2230) bio->bi_status = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2231) bio_endio(bio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2232) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2233) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2234) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2235)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2236) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2237) * given a list of ordered sums record them in the inode. This happens
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2238) * at IO completion time based on sums calculated at bio submission time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2239) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2240) static int add_pending_csums(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2241) struct list_head *list)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2242) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2243) struct btrfs_ordered_sum *sum;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2244) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2246) list_for_each_entry(sum, list, list) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2247) trans->adding_csums = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2248) ret = btrfs_csum_file_blocks(trans, trans->fs_info->csum_root, sum);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2249) trans->adding_csums = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2250) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2251) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2252) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2253) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2254) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2255)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2256) static int btrfs_find_new_delalloc_bytes(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2257) const u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2258) const u64 len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2259) struct extent_state **cached_state)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2260) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2261) u64 search_start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2262) const u64 end = start + len - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2264) while (search_start < end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2265) const u64 search_len = end - search_start + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2266) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2267) u64 em_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2268) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2269)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2270) em = btrfs_get_extent(inode, NULL, 0, search_start, search_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2271) if (IS_ERR(em))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2272) return PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2273)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2274) if (em->block_start != EXTENT_MAP_HOLE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2275) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2276)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2277) em_len = em->len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2278) if (em->start < search_start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2279) em_len -= search_start - em->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2280) if (em_len > search_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2281) em_len = search_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2282)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2283) ret = set_extent_bit(&inode->io_tree, search_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2284) search_start + em_len - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2285) EXTENT_DELALLOC_NEW,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2286) NULL, cached_state, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2287) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2288) search_start = extent_map_end(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2289) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2290) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2291) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2292) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2293) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2294) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2295)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2296) int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2297) unsigned int extra_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2298) struct extent_state **cached_state)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2299) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2300) WARN_ON(PAGE_ALIGNED(end));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2302) if (start >= i_size_read(&inode->vfs_inode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2303) !(inode->flags & BTRFS_INODE_PREALLOC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2304) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2305) * There can't be any extents following eof in this case so just
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2306) * set the delalloc new bit for the range directly.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2307) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2308) extra_bits |= EXTENT_DELALLOC_NEW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2309) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2310) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2312) ret = btrfs_find_new_delalloc_bytes(inode, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2313) end + 1 - start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2314) cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2315) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2316) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2317) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2318)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2319) return set_extent_delalloc(&inode->io_tree, start, end, extra_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2320) cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2321) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2322)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2323) /* see btrfs_writepage_start_hook for details on why this is required */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2324) struct btrfs_writepage_fixup {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2325) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2326) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2327) struct btrfs_work work;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2328) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2330) static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2331) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2332) struct btrfs_writepage_fixup *fixup;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2333) struct btrfs_ordered_extent *ordered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2334) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2335) struct extent_changeset *data_reserved = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2336) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2337) struct btrfs_inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2338) u64 page_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2339) u64 page_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2340) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2341) bool free_delalloc_space = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2342)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2343) fixup = container_of(work, struct btrfs_writepage_fixup, work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2344) page = fixup->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2345) inode = BTRFS_I(fixup->inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2346) page_start = page_offset(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2347) page_end = page_offset(page) + PAGE_SIZE - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2349) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2350) * This is similar to page_mkwrite, we need to reserve the space before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2351) * we take the page lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2352) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2353) ret = btrfs_delalloc_reserve_space(inode, &data_reserved, page_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2354) PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2355) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2356) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2357)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2358) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2359) * Before we queued this fixup, we took a reference on the page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2360) * page->mapping may go NULL, but it shouldn't be moved to a different
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2361) * address space.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2362) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2363) if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2364) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2365) * Unfortunately this is a little tricky, either
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2366) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2367) * 1) We got here and our page had already been dealt with and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2368) * we reserved our space, thus ret == 0, so we need to just
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2369) * drop our space reservation and bail. This can happen the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2370) * first time we come into the fixup worker, or could happen
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2371) * while waiting for the ordered extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2372) * 2) Our page was already dealt with, but we happened to get an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2373) * ENOSPC above from the btrfs_delalloc_reserve_space. In
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2374) * this case we obviously don't have anything to release, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2375) * because the page was already dealt with we don't want to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2376) * mark the page with an error, so make sure we're resetting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2377) * ret to 0. This is why we have this check _before_ the ret
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2378) * check, because we do not want to have a surprise ENOSPC
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2379) * when the page was already properly dealt with.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2380) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2381) if (!ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2382) btrfs_delalloc_release_extents(inode, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2383) btrfs_delalloc_release_space(inode, data_reserved,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2384) page_start, PAGE_SIZE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2385) true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2386) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2387) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2388) goto out_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2389) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2391) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2392) * We can't mess with the page state unless it is locked, so now that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2393) * it is locked bail if we failed to make our space reservation.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2394) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2395) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2396) goto out_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2397)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2398) lock_extent_bits(&inode->io_tree, page_start, page_end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2399)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2400) /* already ordered? We're done */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2401) if (PagePrivate2(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2402) goto out_reserved;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2403)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2404) ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2405) if (ordered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2406) unlock_extent_cached(&inode->io_tree, page_start, page_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2407) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2408) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2409) btrfs_start_ordered_extent(ordered, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2410) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2411) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2412) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2413)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2414) ret = btrfs_set_extent_delalloc(inode, page_start, page_end, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2415) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2416) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2417) goto out_reserved;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2419) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2420) * Everything went as planned, we're now the owner of a dirty page with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2421) * delayed allocation bits set and space reserved for our COW
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2422) * destination.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2423) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2424) * The page was dirty when we started, nothing should have cleaned it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2425) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2426) BUG_ON(!PageDirty(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2427) free_delalloc_space = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2428) out_reserved:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2429) btrfs_delalloc_release_extents(inode, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2430) if (free_delalloc_space)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2431) btrfs_delalloc_release_space(inode, data_reserved, page_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2432) PAGE_SIZE, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2433) unlock_extent_cached(&inode->io_tree, page_start, page_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2434) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2435) out_page:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2436) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2437) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2438) * We hit ENOSPC or other errors. Update the mapping and page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2439) * to reflect the errors and clean the page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2440) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2441) mapping_set_error(page->mapping, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2442) end_extent_writepage(page, ret, page_start, page_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2443) clear_page_dirty_for_io(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2444) SetPageError(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2445) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2446) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2447) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2448) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2449) kfree(fixup);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2450) extent_changeset_free(data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2451) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2452) * As a precaution, do a delayed iput in case it would be the last iput
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2453) * that could need flushing space. Recursing back to fixup worker would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2454) * deadlock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2455) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2456) btrfs_add_delayed_iput(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2457) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2458)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2459) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2460) * There are a few paths in the higher layers of the kernel that directly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2461) * set the page dirty bit without asking the filesystem if it is a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2462) * good idea. This causes problems because we want to make sure COW
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2463) * properly happens and the data=ordered rules are followed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2464) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2465) * In our case any range that doesn't have the ORDERED bit set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2466) * hasn't been properly setup for IO. We kick off an async process
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2467) * to fix it up. The async helper will wait for ordered extents, set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2468) * the delalloc bit and make it safe to write the page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2469) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2470) int btrfs_writepage_cow_fixup(struct page *page, u64 start, u64 end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2471) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2472) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2473) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2474) struct btrfs_writepage_fixup *fixup;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2475)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2476) /* this page is properly in the ordered list */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2477) if (TestClearPagePrivate2(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2478) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2479)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2480) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2481) * PageChecked is set below when we create a fixup worker for this page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2482) * don't try to create another one if we're already PageChecked()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2483) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2484) * The extent_io writepage code will redirty the page if we send back
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2485) * EAGAIN.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2486) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2487) if (PageChecked(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2488) return -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2489)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2490) fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2491) if (!fixup)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2492) return -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2493)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2494) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2495) * We are already holding a reference to this inode from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2496) * write_cache_pages. We need to hold it because the space reservation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2497) * takes place outside of the page lock, and we can't trust
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2498) * page->mapping outside of the page lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2499) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2500) ihold(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2501) SetPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2502) get_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2503) btrfs_init_work(&fixup->work, btrfs_writepage_fixup_worker, NULL, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2504) fixup->page = page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2505) fixup->inode = inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2506) btrfs_queue_work(fs_info->fixup_workers, &fixup->work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2507)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2508) return -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2509) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2510)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2511) static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2512) struct btrfs_inode *inode, u64 file_pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2513) struct btrfs_file_extent_item *stack_fi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2514) u64 qgroup_reserved)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2515) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2516) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2517) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2518) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2519) struct btrfs_key ins;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2520) u64 disk_num_bytes = btrfs_stack_file_extent_disk_num_bytes(stack_fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2521) u64 disk_bytenr = btrfs_stack_file_extent_disk_bytenr(stack_fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2522) u64 num_bytes = btrfs_stack_file_extent_num_bytes(stack_fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2523) u64 ram_bytes = btrfs_stack_file_extent_ram_bytes(stack_fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2524) int extent_inserted = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2525) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2527) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2528) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2529) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2530)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2531) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2532) * we may be replacing one extent in the tree with another.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2533) * The new extent is pinned in the extent map, and we don't want
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2534) * to drop it from the cache until it is completely in the btree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2535) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2536) * So, tell btrfs_drop_extents to leave this extent in the cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2537) * the caller is expected to unpin it and allow it to be merged
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2538) * with the others.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2539) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2540) ret = __btrfs_drop_extents(trans, root, inode, path, file_pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2541) file_pos + num_bytes, NULL, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2542) 1, sizeof(*stack_fi), &extent_inserted);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2543) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2544) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2546) if (!extent_inserted) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2547) ins.objectid = btrfs_ino(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2548) ins.offset = file_pos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2549) ins.type = BTRFS_EXTENT_DATA_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2550)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2551) path->leave_spinning = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2552) ret = btrfs_insert_empty_item(trans, root, path, &ins,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2553) sizeof(*stack_fi));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2554) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2555) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2556) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2557) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2558) btrfs_set_stack_file_extent_generation(stack_fi, trans->transid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2559) write_extent_buffer(leaf, stack_fi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2560) btrfs_item_ptr_offset(leaf, path->slots[0]),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2561) sizeof(struct btrfs_file_extent_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2562)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2563) btrfs_mark_buffer_dirty(leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2564) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2565)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2566) inode_add_bytes(&inode->vfs_inode, num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2567)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2568) ins.objectid = disk_bytenr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2569) ins.offset = disk_num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2570) ins.type = BTRFS_EXTENT_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2571)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2572) ret = btrfs_inode_set_file_extent_range(inode, file_pos, ram_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2573) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2574) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2575)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2576) ret = btrfs_alloc_reserved_file_extent(trans, root, btrfs_ino(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2577) file_pos, qgroup_reserved, &ins);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2578) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2579) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2580)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2581) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2582) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2583)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2584) static void btrfs_release_delalloc_bytes(struct btrfs_fs_info *fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2585) u64 start, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2586) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2587) struct btrfs_block_group *cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2588)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2589) cache = btrfs_lookup_block_group(fs_info, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2590) ASSERT(cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2591)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2592) spin_lock(&cache->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2593) cache->delalloc_bytes -= len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2594) spin_unlock(&cache->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2595)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2596) btrfs_put_block_group(cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2597) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2598)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2599) static int insert_ordered_extent_file_extent(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2600) struct btrfs_ordered_extent *oe)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2601) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2602) struct btrfs_file_extent_item stack_fi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2603) u64 logical_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2605) memset(&stack_fi, 0, sizeof(stack_fi));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2606) btrfs_set_stack_file_extent_type(&stack_fi, BTRFS_FILE_EXTENT_REG);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2607) btrfs_set_stack_file_extent_disk_bytenr(&stack_fi, oe->disk_bytenr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2608) btrfs_set_stack_file_extent_disk_num_bytes(&stack_fi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2609) oe->disk_num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2610) if (test_bit(BTRFS_ORDERED_TRUNCATED, &oe->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2611) logical_len = oe->truncated_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2612) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2613) logical_len = oe->num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2614) btrfs_set_stack_file_extent_num_bytes(&stack_fi, logical_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2615) btrfs_set_stack_file_extent_ram_bytes(&stack_fi, logical_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2616) btrfs_set_stack_file_extent_compression(&stack_fi, oe->compress_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2617) /* Encryption and other encoding is reserved and all 0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2618)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2619) return insert_reserved_file_extent(trans, BTRFS_I(oe->inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2620) oe->file_offset, &stack_fi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2621) oe->qgroup_rsv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2622) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2623)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2624) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2625) * As ordered data IO finishes, this gets called so we can finish
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2626) * an ordered extent if the range of bytes in the file it covers are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2627) * fully written.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2628) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2629) static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2630) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2631) struct inode *inode = ordered_extent->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2632) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2633) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2634) struct btrfs_trans_handle *trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2635) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2636) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2637) u64 start, end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2638) int compress_type = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2639) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2640) u64 logical_len = ordered_extent->num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2641) bool freespace_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2642) bool truncated = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2643) bool range_locked = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2644) bool clear_new_delalloc_bytes = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2645) bool clear_reserved_extent = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2646) unsigned int clear_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2647)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2648) start = ordered_extent->file_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2649) end = start + ordered_extent->num_bytes - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2651) if (!test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2652) !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2653) !test_bit(BTRFS_ORDERED_DIRECT, &ordered_extent->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2654) clear_new_delalloc_bytes = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2655)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2656) freespace_inode = btrfs_is_free_space_inode(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2657)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2658) if (test_bit(BTRFS_ORDERED_IOERR, &ordered_extent->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2659) ret = -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2660) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2661) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2662)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2663) btrfs_free_io_failure_record(BTRFS_I(inode), start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2664)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2665) if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered_extent->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2666) truncated = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2667) logical_len = ordered_extent->truncated_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2668) /* Truncated the entire extent, don't bother adding */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2669) if (!logical_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2670) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2671) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2672)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2673) if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2674) BUG_ON(!list_empty(&ordered_extent->list)); /* Logic error */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2675)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2676) btrfs_inode_safe_disk_i_size_write(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2677) if (freespace_inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2678) trans = btrfs_join_transaction_spacecache(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2679) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2680) trans = btrfs_join_transaction(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2681) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2682) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2683) trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2684) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2685) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2686) trans->block_rsv = &BTRFS_I(inode)->block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2687) ret = btrfs_update_inode_fallback(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2688) if (ret) /* -ENOMEM or corruption */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2689) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2690) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2691) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2692)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2693) range_locked = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2694) lock_extent_bits(io_tree, start, end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2695)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2696) if (freespace_inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2697) trans = btrfs_join_transaction_spacecache(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2698) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2699) trans = btrfs_join_transaction(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2700) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2701) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2702) trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2703) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2704) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2705)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2706) trans->block_rsv = &BTRFS_I(inode)->block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2707)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2708) if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2709) compress_type = ordered_extent->compress_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2710) if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2711) BUG_ON(compress_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2712) ret = btrfs_mark_extent_written(trans, BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2713) ordered_extent->file_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2714) ordered_extent->file_offset +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2715) logical_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2716) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2717) BUG_ON(root == fs_info->tree_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2718) ret = insert_ordered_extent_file_extent(trans, ordered_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2719) if (!ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2720) clear_reserved_extent = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2721) btrfs_release_delalloc_bytes(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2722) ordered_extent->disk_bytenr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2723) ordered_extent->disk_num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2724) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2725) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2726) unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2727) ordered_extent->file_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2728) ordered_extent->num_bytes, trans->transid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2729) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2730) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2731) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2732) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2733)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2734) ret = add_pending_csums(trans, &ordered_extent->list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2735) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2736) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2737) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2738) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2739)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2740) btrfs_inode_safe_disk_i_size_write(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2741) ret = btrfs_update_inode_fallback(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2742) if (ret) { /* -ENOMEM or corruption */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2743) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2744) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2745) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2746) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2747) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2748) clear_bits = EXTENT_DEFRAG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2749) if (range_locked)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2750) clear_bits |= EXTENT_LOCKED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2751) if (clear_new_delalloc_bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2752) clear_bits |= EXTENT_DELALLOC_NEW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2753) clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, clear_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2754) (clear_bits & EXTENT_LOCKED) ? 1 : 0, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2755) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2756)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2757) if (trans)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2758) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2759)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2760) if (ret || truncated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2761) u64 unwritten_start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2763) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2764) * If we failed to finish this ordered extent for any reason we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2765) * need to make sure BTRFS_ORDERED_IOERR is set on the ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2766) * extent, and mark the inode with the error if it wasn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2767) * already set. Any error during writeback would have already
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2768) * set the mapping error, so we need to set it if we're the ones
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2769) * marking this ordered extent as failed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2770) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2771) if (ret && !test_and_set_bit(BTRFS_ORDERED_IOERR,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2772) &ordered_extent->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2773) mapping_set_error(ordered_extent->inode->i_mapping, -EIO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2774)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2775) if (truncated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2776) unwritten_start += logical_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2777) clear_extent_uptodate(io_tree, unwritten_start, end, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2778)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2779) /* Drop the cache for the part of the extent we didn't write. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2780) btrfs_drop_extent_cache(BTRFS_I(inode), unwritten_start, end, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2781)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2782) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2783) * If the ordered extent had an IOERR or something else went
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2784) * wrong we need to return the space for this ordered extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2785) * back to the allocator. We only free the extent in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2786) * truncated case if we didn't write out the extent at all.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2787) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2788) * If we made it past insert_reserved_file_extent before we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2789) * errored out then we don't need to do this as the accounting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2790) * has already been done.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2791) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2792) if ((ret || !logical_len) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2793) clear_reserved_extent &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2794) !test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2795) !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2796) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2797) * Discard the range before returning it back to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2798) * free space pool
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2799) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2800) if (ret && btrfs_test_opt(fs_info, DISCARD_SYNC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2801) btrfs_discard_extent(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2802) ordered_extent->disk_bytenr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2803) ordered_extent->disk_num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2804) NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2805) btrfs_free_reserved_extent(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2806) ordered_extent->disk_bytenr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2807) ordered_extent->disk_num_bytes, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2808) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2809) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2810)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2811) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2812) * This needs to be done to make sure anybody waiting knows we are done
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2813) * updating everything for this ordered extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2814) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2815) btrfs_remove_ordered_extent(BTRFS_I(inode), ordered_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2816)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2817) /* once for us */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2818) btrfs_put_ordered_extent(ordered_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2819) /* once for the tree */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2820) btrfs_put_ordered_extent(ordered_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2821)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2822) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2823) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2825) static void finish_ordered_fn(struct btrfs_work *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2826) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2827) struct btrfs_ordered_extent *ordered_extent;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2828) ordered_extent = container_of(work, struct btrfs_ordered_extent, work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2829) btrfs_finish_ordered_io(ordered_extent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2830) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2831)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2832) void btrfs_writepage_endio_finish_ordered(struct page *page, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2833) u64 end, int uptodate)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2834) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2835) struct btrfs_inode *inode = BTRFS_I(page->mapping->host);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2836) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2837) struct btrfs_ordered_extent *ordered_extent = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2838) struct btrfs_workqueue *wq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2839)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2840) trace_btrfs_writepage_end_io_hook(page, start, end, uptodate);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2841)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2842) ClearPagePrivate2(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2843) if (!btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2844) end - start + 1, uptodate))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2845) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2847) if (btrfs_is_free_space_inode(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2848) wq = fs_info->endio_freespace_worker;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2849) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2850) wq = fs_info->endio_write_workers;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2851)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2852) btrfs_init_work(&ordered_extent->work, finish_ordered_fn, NULL, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2853) btrfs_queue_work(wq, &ordered_extent->work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2854) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2856) static int check_data_csum(struct inode *inode, struct btrfs_io_bio *io_bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2857) int icsum, struct page *page, int pgoff, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2858) size_t len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2859) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2860) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2861) SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2862) char *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2863) u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2864) u8 *csum_expected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2865) u8 csum[BTRFS_CSUM_SIZE];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2866)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2867) csum_expected = ((u8 *)io_bio->csum) + icsum * csum_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2869) kaddr = kmap_atomic(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2870) shash->tfm = fs_info->csum_shash;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2872) crypto_shash_digest(shash, kaddr + pgoff, len, csum);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2873)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2874) if (memcmp(csum, csum_expected, csum_size))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2875) goto zeroit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2876)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2877) kunmap_atomic(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2878) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2879) zeroit:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2880) btrfs_print_data_csum_error(BTRFS_I(inode), start, csum, csum_expected,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2881) io_bio->mirror_num);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2882) if (io_bio->device)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2883) btrfs_dev_stat_inc_and_print(io_bio->device,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2884) BTRFS_DEV_STAT_CORRUPTION_ERRS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2885) memset(kaddr + pgoff, 1, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2886) flush_dcache_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2887) kunmap_atomic(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2888) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2889) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2890)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2891) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2892) * when reads are done, we need to check csums to verify the data is correct
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2893) * if there's a match, we allow the bio to finish. If not, the code in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2894) * extent_io.c will try to find good copies for us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2895) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2896) int btrfs_verify_data_csum(struct btrfs_io_bio *io_bio, u64 phy_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2897) struct page *page, u64 start, u64 end, int mirror)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2898) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2899) size_t offset = start - page_offset(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2900) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2901) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2902) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2903)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2904) if (PageChecked(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2905) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2906) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2907) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2909) if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2910) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2911)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2912) if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2913) test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2914) clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2915) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2916) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2918) phy_offset >>= inode->i_sb->s_blocksize_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2919) return check_data_csum(inode, io_bio, phy_offset, page, offset, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2920) (size_t)(end - start + 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2921) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2922)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2923) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2924) * btrfs_add_delayed_iput - perform a delayed iput on @inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2925) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2926) * @inode: The inode we want to perform iput on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2927) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2928) * This function uses the generic vfs_inode::i_count to track whether we should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2929) * just decrement it (in case it's > 1) or if this is the last iput then link
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2930) * the inode to the delayed iput machinery. Delayed iputs are processed at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2931) * transaction commit time/superblock commit/cleaner kthread.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2932) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2933) void btrfs_add_delayed_iput(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2934) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2935) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2936) struct btrfs_inode *binode = BTRFS_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2937)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2938) if (atomic_add_unless(&inode->i_count, -1, 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2939) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2940)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2941) atomic_inc(&fs_info->nr_delayed_iputs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2942) spin_lock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2943) ASSERT(list_empty(&binode->delayed_iput));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2944) list_add_tail(&binode->delayed_iput, &fs_info->delayed_iputs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2945) spin_unlock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2946) if (!test_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2947) wake_up_process(fs_info->cleaner_kthread);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2948) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2949)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2950) static void run_delayed_iput_locked(struct btrfs_fs_info *fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2951) struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2952) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2953) list_del_init(&inode->delayed_iput);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2954) spin_unlock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2955) iput(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2956) if (atomic_dec_and_test(&fs_info->nr_delayed_iputs))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2957) wake_up(&fs_info->delayed_iputs_wait);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2958) spin_lock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2959) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2960)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2961) static void btrfs_run_delayed_iput(struct btrfs_fs_info *fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2962) struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2963) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2964) if (!list_empty(&inode->delayed_iput)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2965) spin_lock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2966) if (!list_empty(&inode->delayed_iput))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2967) run_delayed_iput_locked(fs_info, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2968) spin_unlock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2969) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2970) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2972) void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2973) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2974)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2975) spin_lock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2976) while (!list_empty(&fs_info->delayed_iputs)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2977) struct btrfs_inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2978)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2979) inode = list_first_entry(&fs_info->delayed_iputs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2980) struct btrfs_inode, delayed_iput);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2981) run_delayed_iput_locked(fs_info, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2982) cond_resched_lock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2983) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2984) spin_unlock(&fs_info->delayed_iput_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2985) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2986)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2987) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2988) * btrfs_wait_on_delayed_iputs - wait on the delayed iputs to be done running
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2989) * @fs_info - the fs_info for this fs
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2990) * @return - EINTR if we were killed, 0 if nothing's pending
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2991) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2992) * This will wait on any delayed iputs that are currently running with KILLABLE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2993) * set. Once they are all done running we will return, unless we are killed in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2994) * which case we return EINTR. This helps in user operations like fallocate etc
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2995) * that might get blocked on the iputs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2996) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2997) int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2998) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2999) int ret = wait_event_killable(fs_info->delayed_iputs_wait,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3000) atomic_read(&fs_info->nr_delayed_iputs) == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3001) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3002) return -EINTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3003) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3004) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3005)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3006) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3007) * This creates an orphan entry for the given inode in case something goes wrong
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3008) * in the middle of an unlink.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3009) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3010) int btrfs_orphan_add(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3011) struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3012) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3013) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3015) ret = btrfs_insert_orphan_item(trans, inode->root, btrfs_ino(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3016) if (ret && ret != -EEXIST) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3017) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3018) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3019) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3020)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3021) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3022) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3024) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3025) * We have done the delete so we can go ahead and remove the orphan item for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3026) * this particular inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3027) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3028) static int btrfs_orphan_del(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3029) struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3030) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3031) return btrfs_del_orphan_item(trans, inode->root, btrfs_ino(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3032) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3033)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3034) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3035) * this cleans up any orphans that may be left on the list from the last use
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3036) * of this root.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3037) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3038) int btrfs_orphan_cleanup(struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3039) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3040) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3041) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3042) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3043) struct btrfs_key key, found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3044) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3045) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3046) u64 last_objectid = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3047) int ret = 0, nr_unlink = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3048)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3049) if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3050) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3051)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3052) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3053) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3054) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3055) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3056) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3057) path->reada = READA_BACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3058)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3059) key.objectid = BTRFS_ORPHAN_OBJECTID;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3060) key.type = BTRFS_ORPHAN_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3061) key.offset = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3062)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3063) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3064) ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3065) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3066) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3068) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3069) * if ret == 0 means we found what we were searching for, which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3070) * is weird, but possible, so only screw with path if we didn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3071) * find the key and see if we have stuff that matches
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3072) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3073) if (ret > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3074) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3075) if (path->slots[0] == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3076) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3077) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3078) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3079)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3080) /* pull out the item */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3081) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3082) btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3084) /* make sure the item matches what we want */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3085) if (found_key.objectid != BTRFS_ORPHAN_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3086) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3087) if (found_key.type != BTRFS_ORPHAN_ITEM_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3088) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3089)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3090) /* release the path since we're done with it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3091) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3092)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3093) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3094) * this is where we are basically btrfs_lookup, without the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3095) * crossing root thing. we store the inode number in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3096) * offset of the orphan item.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3097) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3098)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3099) if (found_key.offset == last_objectid) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3100) btrfs_err(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3101) "Error removing orphan entry, stopping orphan cleanup");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3102) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3103) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3104) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3106) last_objectid = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3108) found_key.objectid = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3109) found_key.type = BTRFS_INODE_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3110) found_key.offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3111) inode = btrfs_iget(fs_info->sb, last_objectid, root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3112) ret = PTR_ERR_OR_ZERO(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3113) if (ret && ret != -ENOENT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3114) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3116) if (ret == -ENOENT && root == fs_info->tree_root) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3117) struct btrfs_root *dead_root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3118) int is_dead_root = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3119)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3120) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3121) * this is an orphan in the tree root. Currently these
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3122) * could come from 2 sources:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3123) * a) a snapshot deletion in progress
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3124) * b) a free space cache inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3125) * We need to distinguish those two, as the snapshot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3126) * orphan must not get deleted.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3127) * find_dead_roots already ran before us, so if this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3128) * is a snapshot deletion, we should find the root
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3129) * in the fs_roots radix tree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3130) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3132) spin_lock(&fs_info->fs_roots_radix_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3133) dead_root = radix_tree_lookup(&fs_info->fs_roots_radix,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3134) (unsigned long)found_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3135) if (dead_root && btrfs_root_refs(&dead_root->root_item) == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3136) is_dead_root = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3137) spin_unlock(&fs_info->fs_roots_radix_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3139) if (is_dead_root) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3140) /* prevent this orphan from being found again */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3141) key.offset = found_key.objectid - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3142) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3143) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3145) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3147) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3148) * If we have an inode with links, there are a couple of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3149) * possibilities. Old kernels (before v3.12) used to create an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3150) * orphan item for truncate indicating that there were possibly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3151) * extent items past i_size that needed to be deleted. In v3.12,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3152) * truncate was changed to update i_size in sync with the extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3153) * items, but the (useless) orphan item was still created. Since
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3154) * v4.18, we don't create the orphan item for truncate at all.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3155) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3156) * So, this item could mean that we need to do a truncate, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3157) * only if this filesystem was last used on a pre-v3.12 kernel
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3158) * and was not cleanly unmounted. The odds of that are quite
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3159) * slim, and it's a pain to do the truncate now, so just delete
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3160) * the orphan item.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3161) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3162) * It's also possible that this orphan item was supposed to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3163) * deleted but wasn't. The inode number may have been reused,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3164) * but either way, we can delete the orphan item.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3165) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3166) if (ret == -ENOENT || inode->i_nlink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3167) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3168) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3169) trans = btrfs_start_transaction(root, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3170) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3171) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3172) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3173) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3174) btrfs_debug(fs_info, "auto deleting %Lu",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3175) found_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3176) ret = btrfs_del_orphan_item(trans, root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3177) found_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3178) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3179) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3180) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3181) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3182) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3183)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3184) nr_unlink++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3185)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3186) /* this will do delete_inode and everything for us */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3187) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3188) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3189) /* release the path since we're done with it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3190) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3192) root->orphan_cleanup_state = ORPHAN_CLEANUP_DONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3194) if (test_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3195) trans = btrfs_join_transaction(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3196) if (!IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3197) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3198) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3200) if (nr_unlink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3201) btrfs_debug(fs_info, "unlinked %d orphans", nr_unlink);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3202)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3203) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3204) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3205) btrfs_err(fs_info, "could not do orphan cleanup %d", ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3206) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3207) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3208) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3209)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3210) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3211) * very simple check to peek ahead in the leaf looking for xattrs. If we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3212) * don't find any xattrs, we know there can't be any acls.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3213) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3214) * slot is the slot the inode is in, objectid is the objectid of the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3215) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3216) static noinline int acls_after_inode_item(struct extent_buffer *leaf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3217) int slot, u64 objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3218) int *first_xattr_slot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3219) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3220) u32 nritems = btrfs_header_nritems(leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3221) struct btrfs_key found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3222) static u64 xattr_access = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3223) static u64 xattr_default = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3224) int scanned = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3225)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3226) if (!xattr_access) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3227) xattr_access = btrfs_name_hash(XATTR_NAME_POSIX_ACL_ACCESS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3228) strlen(XATTR_NAME_POSIX_ACL_ACCESS));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3229) xattr_default = btrfs_name_hash(XATTR_NAME_POSIX_ACL_DEFAULT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3230) strlen(XATTR_NAME_POSIX_ACL_DEFAULT));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3231) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3232)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3233) slot++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3234) *first_xattr_slot = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3235) while (slot < nritems) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3236) btrfs_item_key_to_cpu(leaf, &found_key, slot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3237)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3238) /* we found a different objectid, there must not be acls */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3239) if (found_key.objectid != objectid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3240) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3241)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3242) /* we found an xattr, assume we've got an acl */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3243) if (found_key.type == BTRFS_XATTR_ITEM_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3244) if (*first_xattr_slot == -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3245) *first_xattr_slot = slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3246) if (found_key.offset == xattr_access ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3247) found_key.offset == xattr_default)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3248) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3249) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3250)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3251) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3252) * we found a key greater than an xattr key, there can't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3253) * be any acls later on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3254) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3255) if (found_key.type > BTRFS_XATTR_ITEM_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3256) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3257)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3258) slot++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3259) scanned++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3260)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3261) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3262) * it goes inode, inode backrefs, xattrs, extents,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3263) * so if there are a ton of hard links to an inode there can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3264) * be a lot of backrefs. Don't waste time searching too hard,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3265) * this is just an optimization
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3266) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3267) if (scanned >= 8)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3268) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3269) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3270) /* we hit the end of the leaf before we found an xattr or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3271) * something larger than an xattr. We have to assume the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3272) * has acls
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3273) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3274) if (*first_xattr_slot == -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3275) *first_xattr_slot = slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3276) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3277) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3278)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3279) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3280) * read an inode from the btree into the in-memory inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3281) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3282) static int btrfs_read_locked_inode(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3283) struct btrfs_path *in_path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3284) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3285) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3286) struct btrfs_path *path = in_path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3287) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3288) struct btrfs_inode_item *inode_item;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3289) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3290) struct btrfs_key location;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3291) unsigned long ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3292) int maybe_acls;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3293) u32 rdev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3294) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3295) bool filled = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3296) int first_xattr_slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3297)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3298) ret = btrfs_fill_inode(inode, &rdev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3299) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3300) filled = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3302) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3303) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3304) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3305) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3306) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3307)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3308) memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3309)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3310) ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3311) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3312) if (path != in_path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3313) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3314) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3315) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3316)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3317) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3318)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3319) if (filled)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3320) goto cache_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3322) inode_item = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3323) struct btrfs_inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3324) inode->i_mode = btrfs_inode_mode(leaf, inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3325) set_nlink(inode, btrfs_inode_nlink(leaf, inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3326) i_uid_write(inode, btrfs_inode_uid(leaf, inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3327) i_gid_write(inode, btrfs_inode_gid(leaf, inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3328) btrfs_i_size_write(BTRFS_I(inode), btrfs_inode_size(leaf, inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3329) btrfs_inode_set_file_extent_range(BTRFS_I(inode), 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3330) round_up(i_size_read(inode), fs_info->sectorsize));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3331)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3332) inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, &inode_item->atime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3333) inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, &inode_item->atime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3335) inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, &inode_item->mtime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3336) inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, &inode_item->mtime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3337)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3338) inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, &inode_item->ctime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3339) inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, &inode_item->ctime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3340)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3341) BTRFS_I(inode)->i_otime.tv_sec =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3342) btrfs_timespec_sec(leaf, &inode_item->otime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3343) BTRFS_I(inode)->i_otime.tv_nsec =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3344) btrfs_timespec_nsec(leaf, &inode_item->otime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3345)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3346) inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3347) BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3348) BTRFS_I(inode)->last_trans = btrfs_inode_transid(leaf, inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3350) inode_set_iversion_queried(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3351) btrfs_inode_sequence(leaf, inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3352) inode->i_generation = BTRFS_I(inode)->generation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3353) inode->i_rdev = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3354) rdev = btrfs_inode_rdev(leaf, inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3356) BTRFS_I(inode)->index_cnt = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3357) BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3358)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3359) cache_index:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3360) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3361) * If we were modified in the current generation and evicted from memory
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3362) * and then re-read we need to do a full sync since we don't have any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3363) * idea about which extents were modified before we were evicted from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3364) * cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3365) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3366) * This is required for both inode re-read from disk and delayed inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3367) * in delayed_nodes_tree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3368) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3369) if (BTRFS_I(inode)->last_trans == fs_info->generation)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3370) set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3371) &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3373) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3374) * We don't persist the id of the transaction where an unlink operation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3375) * against the inode was last made. So here we assume the inode might
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3376) * have been evicted, and therefore the exact value of last_unlink_trans
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3377) * lost, and set it to last_trans to avoid metadata inconsistencies
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3378) * between the inode and its parent if the inode is fsync'ed and the log
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3379) * replayed. For example, in the scenario:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3380) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3381) * touch mydir/foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3382) * ln mydir/foo mydir/bar
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3383) * sync
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3384) * unlink mydir/bar
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3385) * echo 2 > /proc/sys/vm/drop_caches # evicts inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3386) * xfs_io -c fsync mydir/foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3387) * <power failure>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3388) * mount fs, triggers fsync log replay
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3389) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3390) * We must make sure that when we fsync our inode foo we also log its
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3391) * parent inode, otherwise after log replay the parent still has the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3392) * dentry with the "bar" name but our inode foo has a link count of 1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3393) * and doesn't have an inode ref with the name "bar" anymore.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3394) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3395) * Setting last_unlink_trans to last_trans is a pessimistic approach,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3396) * but it guarantees correctness at the expense of occasional full
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3397) * transaction commits on fsync if our inode is a directory, or if our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3398) * inode is not a directory, logging its parent unnecessarily.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3399) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3400) BTRFS_I(inode)->last_unlink_trans = BTRFS_I(inode)->last_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3401)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3402) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3403) * Same logic as for last_unlink_trans. We don't persist the generation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3404) * of the last transaction where this inode was used for a reflink
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3405) * operation, so after eviction and reloading the inode we must be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3406) * pessimistic and assume the last transaction that modified the inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3407) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3408) BTRFS_I(inode)->last_reflink_trans = BTRFS_I(inode)->last_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3409)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3410) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3411) if (inode->i_nlink != 1 ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3412) path->slots[0] >= btrfs_header_nritems(leaf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3413) goto cache_acl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3414)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3415) btrfs_item_key_to_cpu(leaf, &location, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3416) if (location.objectid != btrfs_ino(BTRFS_I(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3417) goto cache_acl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3419) ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3420) if (location.type == BTRFS_INODE_REF_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3421) struct btrfs_inode_ref *ref;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3422)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3423) ref = (struct btrfs_inode_ref *)ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3424) BTRFS_I(inode)->dir_index = btrfs_inode_ref_index(leaf, ref);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3425) } else if (location.type == BTRFS_INODE_EXTREF_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3426) struct btrfs_inode_extref *extref;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3427)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3428) extref = (struct btrfs_inode_extref *)ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3429) BTRFS_I(inode)->dir_index = btrfs_inode_extref_index(leaf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3430) extref);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3431) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3432) cache_acl:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3433) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3434) * try to precache a NULL acl entry for files that don't have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3435) * any xattrs or acls
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3436) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3437) maybe_acls = acls_after_inode_item(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3438) btrfs_ino(BTRFS_I(inode)), &first_xattr_slot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3439) if (first_xattr_slot != -1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3440) path->slots[0] = first_xattr_slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3441) ret = btrfs_load_inode_props(inode, path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3442) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3443) btrfs_err(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3444) "error loading props for ino %llu (root %llu): %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3445) btrfs_ino(BTRFS_I(inode)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3446) root->root_key.objectid, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3447) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3448) if (path != in_path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3449) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3450)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3451) if (!maybe_acls)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3452) cache_no_acl(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3454) switch (inode->i_mode & S_IFMT) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3455) case S_IFREG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3456) inode->i_mapping->a_ops = &btrfs_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3457) inode->i_fop = &btrfs_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3458) inode->i_op = &btrfs_file_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3459) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3460) case S_IFDIR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3461) inode->i_fop = &btrfs_dir_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3462) inode->i_op = &btrfs_dir_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3463) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3464) case S_IFLNK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3465) inode->i_op = &btrfs_symlink_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3466) inode_nohighmem(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3467) inode->i_mapping->a_ops = &btrfs_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3468) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3469) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3470) inode->i_op = &btrfs_special_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3471) init_special_inode(inode, inode->i_mode, rdev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3472) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3473) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3474)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3475) btrfs_sync_inode_flags_to_i_flags(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3476) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3477) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3478)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3479) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3480) * given a leaf and an inode, copy the inode fields into the leaf
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3481) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3482) static void fill_inode_item(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3483) struct extent_buffer *leaf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3484) struct btrfs_inode_item *item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3485) struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3486) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3487) struct btrfs_map_token token;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3489) btrfs_init_map_token(&token, leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3490)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3491) btrfs_set_token_inode_uid(&token, item, i_uid_read(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3492) btrfs_set_token_inode_gid(&token, item, i_gid_read(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3493) btrfs_set_token_inode_size(&token, item, BTRFS_I(inode)->disk_i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3494) btrfs_set_token_inode_mode(&token, item, inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3495) btrfs_set_token_inode_nlink(&token, item, inode->i_nlink);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3496)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3497) btrfs_set_token_timespec_sec(&token, &item->atime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3498) inode->i_atime.tv_sec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3499) btrfs_set_token_timespec_nsec(&token, &item->atime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3500) inode->i_atime.tv_nsec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3501)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3502) btrfs_set_token_timespec_sec(&token, &item->mtime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3503) inode->i_mtime.tv_sec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3504) btrfs_set_token_timespec_nsec(&token, &item->mtime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3505) inode->i_mtime.tv_nsec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3506)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3507) btrfs_set_token_timespec_sec(&token, &item->ctime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3508) inode->i_ctime.tv_sec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3509) btrfs_set_token_timespec_nsec(&token, &item->ctime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3510) inode->i_ctime.tv_nsec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3512) btrfs_set_token_timespec_sec(&token, &item->otime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3513) BTRFS_I(inode)->i_otime.tv_sec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3514) btrfs_set_token_timespec_nsec(&token, &item->otime,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3515) BTRFS_I(inode)->i_otime.tv_nsec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3517) btrfs_set_token_inode_nbytes(&token, item, inode_get_bytes(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3518) btrfs_set_token_inode_generation(&token, item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3519) BTRFS_I(inode)->generation);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3520) btrfs_set_token_inode_sequence(&token, item, inode_peek_iversion(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3521) btrfs_set_token_inode_transid(&token, item, trans->transid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3522) btrfs_set_token_inode_rdev(&token, item, inode->i_rdev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3523) btrfs_set_token_inode_flags(&token, item, BTRFS_I(inode)->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3524) btrfs_set_token_inode_block_group(&token, item, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3525) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3527) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3528) * copy everything in the in-memory inode into the btree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3529) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3530) static noinline int btrfs_update_inode_item(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3531) struct btrfs_root *root, struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3532) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3533) struct btrfs_inode_item *inode_item;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3534) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3535) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3536) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3537)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3538) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3539) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3540) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3542) path->leave_spinning = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3543) ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3544) 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3545) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3546) if (ret > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3547) ret = -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3548) goto failed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3549) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3550)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3551) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3552) inode_item = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3553) struct btrfs_inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3554)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3555) fill_inode_item(trans, leaf, inode_item, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3556) btrfs_mark_buffer_dirty(leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3557) btrfs_set_inode_last_trans(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3558) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3559) failed:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3560) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3561) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3562) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3564) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3565) * copy everything in the in-memory inode into the btree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3566) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3567) noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3568) struct btrfs_root *root, struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3569) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3570) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3571) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3572)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3573) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3574) * If the inode is a free space inode, we can deadlock during commit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3575) * if we put it into the delayed code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3576) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3577) * The data relocation inode should also be directly updated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3578) * without delay
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3579) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3580) if (!btrfs_is_free_space_inode(BTRFS_I(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3581) && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3582) && !test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3583) btrfs_update_root_times(trans, root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3584)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3585) ret = btrfs_delayed_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3586) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3587) btrfs_set_inode_last_trans(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3588) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3589) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3591) return btrfs_update_inode_item(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3592) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3593)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3594) noinline int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3595) struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3596) struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3597) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3598) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3599)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3600) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3601) if (ret == -ENOSPC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3602) return btrfs_update_inode_item(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3603) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3604) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3605)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3606) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3607) * unlink helper that gets used here in inode.c and in the tree logging
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3608) * recovery code. It remove a link in a directory with a given name, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3609) * also drops the back refs in the inode to the directory
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3610) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3611) static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3612) struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3613) struct btrfs_inode *dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3614) struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3615) const char *name, int name_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3616) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3617) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3618) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3619) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3620) struct btrfs_dir_item *di;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3621) u64 index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3622) u64 ino = btrfs_ino(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3623) u64 dir_ino = btrfs_ino(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3625) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3626) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3627) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3628) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3629) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3630)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3631) path->leave_spinning = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3632) di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3633) name, name_len, -1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3634) if (IS_ERR_OR_NULL(di)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3635) ret = di ? PTR_ERR(di) : -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3636) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3637) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3638) ret = btrfs_delete_one_dir_name(trans, root, path, di);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3639) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3640) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3641) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3643) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3644) * If we don't have dir index, we have to get it by looking up
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3645) * the inode ref, since we get the inode ref, remove it directly,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3646) * it is unnecessary to do delayed deletion.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3647) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3648) * But if we have dir index, needn't search inode ref to get it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3649) * Since the inode ref is close to the inode item, it is better
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3650) * that we delay to delete it, and just do this deletion when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3651) * we update the inode item.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3652) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3653) if (inode->dir_index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3654) ret = btrfs_delayed_delete_inode_ref(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3655) if (!ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3656) index = inode->dir_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3657) goto skip_backref;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3658) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3659) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3660)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3661) ret = btrfs_del_inode_ref(trans, root, name, name_len, ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3662) dir_ino, &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3663) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3664) btrfs_info(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3665) "failed to delete reference to %.*s, inode %llu parent %llu",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3666) name_len, name, ino, dir_ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3667) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3668) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3669) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3670) skip_backref:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3671) ret = btrfs_delete_delayed_dir_index(trans, dir, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3672) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3673) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3674) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3675) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3676)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3677) ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3678) dir_ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3679) if (ret != 0 && ret != -ENOENT) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3680) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3681) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3682) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3683)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3684) ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len, dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3685) index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3686) if (ret == -ENOENT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3687) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3688) else if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3689) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3690)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3691) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3692) * If we have a pending delayed iput we could end up with the final iput
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3693) * being run in btrfs-cleaner context. If we have enough of these built
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3694) * up we can end up burning a lot of time in btrfs-cleaner without any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3695) * way to throttle the unlinks. Since we're currently holding a ref on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3696) * the inode we can run the delayed iput here without any issues as the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3697) * final iput won't be done until after we drop the ref we're currently
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3698) * holding.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3699) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3700) btrfs_run_delayed_iput(fs_info, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3701) err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3702) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3703) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3704) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3705)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3706) btrfs_i_size_write(dir, dir->vfs_inode.i_size - name_len * 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3707) inode_inc_iversion(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3708) inode_inc_iversion(&dir->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3709) inode->vfs_inode.i_ctime = dir->vfs_inode.i_mtime =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3710) dir->vfs_inode.i_ctime = current_time(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3711) ret = btrfs_update_inode(trans, root, &dir->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3712) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3713) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3714) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3715)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3716) int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3717) struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3718) struct btrfs_inode *dir, struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3719) const char *name, int name_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3720) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3721) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3722) ret = __btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3723) if (!ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3724) drop_nlink(&inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3725) ret = btrfs_update_inode(trans, root, &inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3726) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3727) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3728) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3729)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3730) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3731) * helper to start transaction for unlink and rmdir.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3732) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3733) * unlink and rmdir are special in btrfs, they do not always free space, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3734) * if we cannot make our reservations the normal way try and see if there is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3735) * plenty of slack room in the global reserve to migrate, otherwise we cannot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3736) * allow the unlink to occur.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3737) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3738) static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3739) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3740) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3741)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3742) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3743) * 1 for the possible orphan item
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3744) * 1 for the dir item
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3745) * 1 for the dir index
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3746) * 1 for the inode ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3747) * 1 for the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3748) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3749) return btrfs_start_transaction_fallback_global_rsv(root, 5);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3750) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3751)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3752) static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3753) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3754) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3755) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3756) struct inode *inode = d_inode(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3757) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3758)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3759) trans = __unlink_start_trans(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3760) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3761) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3763) btrfs_record_unlink_dir(trans, BTRFS_I(dir), BTRFS_I(d_inode(dentry)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3764) 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3765)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3766) ret = btrfs_unlink_inode(trans, root, BTRFS_I(dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3767) BTRFS_I(d_inode(dentry)), dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3768) dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3769) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3770) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3771)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3772) if (inode->i_nlink == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3773) ret = btrfs_orphan_add(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3774) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3775) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3776) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3777)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3778) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3779) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3780) btrfs_btree_balance_dirty(root->fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3781) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3782) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3783)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3784) static int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3785) struct inode *dir, struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3786) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3787) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3788) struct btrfs_inode *inode = BTRFS_I(d_inode(dentry));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3789) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3790) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3791) struct btrfs_dir_item *di;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3792) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3793) const char *name = dentry->d_name.name;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3794) int name_len = dentry->d_name.len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3795) u64 index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3796) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3797) u64 objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3798) u64 dir_ino = btrfs_ino(BTRFS_I(dir));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3799)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3800) if (btrfs_ino(inode) == BTRFS_FIRST_FREE_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3801) objectid = inode->root->root_key.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3802) } else if (btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3803) objectid = inode->location.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3804) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3805) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3806) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3807) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3809) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3810) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3811) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3812)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3813) di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3814) name, name_len, -1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3815) if (IS_ERR_OR_NULL(di)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3816) ret = di ? PTR_ERR(di) : -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3817) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3818) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3819)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3820) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3821) btrfs_dir_item_key_to_cpu(leaf, di, &key);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3822) WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3823) ret = btrfs_delete_one_dir_name(trans, root, path, di);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3824) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3825) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3826) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3827) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3828) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3830) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3831) * This is a placeholder inode for a subvolume we didn't have a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3832) * reference to at the time of the snapshot creation. In the meantime
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3833) * we could have renamed the real subvol link into our snapshot, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3834) * depending on btrfs_del_root_ref to return -ENOENT here is incorret.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3835) * Instead simply lookup the dir_index_item for this entry so we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3836) * remove it. Otherwise we know we have a ref to the root and we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3837) * call btrfs_del_root_ref, and it _shouldn't_ fail.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3838) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3839) if (btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3840) di = btrfs_search_dir_index_item(root, path, dir_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3841) name, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3842) if (IS_ERR_OR_NULL(di)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3843) if (!di)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3844) ret = -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3845) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3846) ret = PTR_ERR(di);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3847) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3848) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3849) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3850)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3851) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3852) btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3853) index = key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3854) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3855) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3856) ret = btrfs_del_root_ref(trans, objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3857) root->root_key.objectid, dir_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3858) &index, name, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3859) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3860) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3861) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3862) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3863) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3864)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3865) ret = btrfs_delete_delayed_dir_index(trans, BTRFS_I(dir), index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3866) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3867) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3868) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3869) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3870)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3871) btrfs_i_size_write(BTRFS_I(dir), dir->i_size - name_len * 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3872) inode_inc_iversion(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3873) dir->i_mtime = dir->i_ctime = current_time(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3874) ret = btrfs_update_inode_fallback(trans, root, dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3875) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3876) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3877) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3878) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3879) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3880) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3881)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3882) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3883) * Helper to check if the subvolume references other subvolumes or if it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3884) * default.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3885) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3886) static noinline int may_destroy_subvol(struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3887) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3888) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3889) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3890) struct btrfs_dir_item *di;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3891) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3892) u64 dir_id;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3893) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3894)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3895) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3896) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3897) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3899) /* Make sure this root isn't set as the default subvol */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3900) dir_id = btrfs_super_root_dir(fs_info->super_copy);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3901) di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3902) dir_id, "default", 7, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3903) if (di && !IS_ERR(di)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3904) btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3905) if (key.objectid == root->root_key.objectid) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3906) ret = -EPERM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3907) btrfs_err(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3908) "deleting default subvolume %llu is not allowed",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3909) key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3910) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3911) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3912) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3913) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3914)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3915) key.objectid = root->root_key.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3916) key.type = BTRFS_ROOT_REF_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3917) key.offset = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3919) ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3920) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3921) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3922) BUG_ON(ret == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3924) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3925) if (path->slots[0] > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3926) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3927) btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3928) if (key.objectid == root->root_key.objectid &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3929) key.type == BTRFS_ROOT_REF_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3930) ret = -ENOTEMPTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3931) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3932) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3933) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3934) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3935) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3936)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3937) /* Delete all dentries for inodes belonging to the root */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3938) static void btrfs_prune_dentries(struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3939) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3940) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3941) struct rb_node *node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3942) struct rb_node *prev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3943) struct btrfs_inode *entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3944) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3945) u64 objectid = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3946)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3947) if (!test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3948) WARN_ON(btrfs_root_refs(&root->root_item) != 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3949)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3950) spin_lock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3951) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3952) node = root->inode_tree.rb_node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3953) prev = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3954) while (node) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3955) prev = node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3956) entry = rb_entry(node, struct btrfs_inode, rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3957)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3958) if (objectid < btrfs_ino(entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3959) node = node->rb_left;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3960) else if (objectid > btrfs_ino(entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3961) node = node->rb_right;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3962) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3963) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3964) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3965) if (!node) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3966) while (prev) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3967) entry = rb_entry(prev, struct btrfs_inode, rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3968) if (objectid <= btrfs_ino(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3969) node = prev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3970) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3971) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3972) prev = rb_next(prev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3973) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3974) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3975) while (node) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3976) entry = rb_entry(node, struct btrfs_inode, rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3977) objectid = btrfs_ino(entry) + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3978) inode = igrab(&entry->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3979) if (inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3980) spin_unlock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3981) if (atomic_read(&inode->i_count) > 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3982) d_prune_aliases(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3983) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3984) * btrfs_drop_inode will have it removed from the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3985) * cache when its usage count hits zero.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3986) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3987) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3988) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3989) spin_lock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3990) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3991) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3992)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3993) if (cond_resched_lock(&root->inode_lock))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3994) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3995)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3996) node = rb_next(node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3997) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3998) spin_unlock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3999) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4000)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4001) int btrfs_delete_subvolume(struct inode *dir, struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4002) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4003) struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4004) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4005) struct inode *inode = d_inode(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4006) struct btrfs_root *dest = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4007) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4008) struct btrfs_block_rsv block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4009) u64 root_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4010) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4011) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4012)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4013) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4014) * Don't allow to delete a subvolume with send in progress. This is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4015) * inside the inode lock so the error handling that has to drop the bit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4016) * again is not run concurrently.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4017) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4018) spin_lock(&dest->root_item_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4019) if (dest->send_in_progress) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4020) spin_unlock(&dest->root_item_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4021) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4022) "attempt to delete subvolume %llu during send",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4023) dest->root_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4024) return -EPERM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4025) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4026) root_flags = btrfs_root_flags(&dest->root_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4027) btrfs_set_root_flags(&dest->root_item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4028) root_flags | BTRFS_ROOT_SUBVOL_DEAD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4029) spin_unlock(&dest->root_item_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4030)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4031) down_write(&fs_info->subvol_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4032)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4033) err = may_destroy_subvol(dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4034) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4035) goto out_up_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4036)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4037) btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4038) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4039) * One for dir inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4040) * two for dir entries,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4041) * two for root ref/backref.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4042) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4043) err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 5, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4044) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4045) goto out_up_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4046)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4047) trans = btrfs_start_transaction(root, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4048) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4049) err = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4050) goto out_release;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4051) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4052) trans->block_rsv = &block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4053) trans->bytes_reserved = block_rsv.size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4054)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4055) btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4056)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4057) ret = btrfs_unlink_subvol(trans, dir, dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4058) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4059) err = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4060) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4061) goto out_end_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4062) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4064) btrfs_record_root_in_trans(trans, dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4065)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4066) memset(&dest->root_item.drop_progress, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4067) sizeof(dest->root_item.drop_progress));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4068) dest->root_item.drop_level = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4069) btrfs_set_root_refs(&dest->root_item, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4070)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4071) if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4072) ret = btrfs_insert_orphan_item(trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4073) fs_info->tree_root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4074) dest->root_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4075) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4076) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4077) err = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4078) goto out_end_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4079) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4080) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4081)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4082) ret = btrfs_uuid_tree_remove(trans, dest->root_item.uuid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4083) BTRFS_UUID_KEY_SUBVOL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4084) dest->root_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4085) if (ret && ret != -ENOENT) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4086) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4087) err = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4088) goto out_end_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4089) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4090) if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4091) ret = btrfs_uuid_tree_remove(trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4092) dest->root_item.received_uuid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4093) BTRFS_UUID_KEY_RECEIVED_SUBVOL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4094) dest->root_key.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4095) if (ret && ret != -ENOENT) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4096) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4097) err = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4098) goto out_end_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4099) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4100) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4102) free_anon_bdev(dest->anon_dev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4103) dest->anon_dev = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4104) out_end_trans:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4105) trans->block_rsv = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4106) trans->bytes_reserved = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4107) ret = btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4108) if (ret && !err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4109) err = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4110) inode->i_flags |= S_DEAD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4111) out_release:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4112) btrfs_subvolume_release_metadata(root, &block_rsv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4113) out_up_write:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4114) up_write(&fs_info->subvol_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4115) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4116) spin_lock(&dest->root_item_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4117) root_flags = btrfs_root_flags(&dest->root_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4118) btrfs_set_root_flags(&dest->root_item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4119) root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4120) spin_unlock(&dest->root_item_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4121) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4122) d_invalidate(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4123) btrfs_prune_dentries(dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4124) ASSERT(dest->send_in_progress == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4125)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4126) /* the last ref */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4127) if (dest->ino_cache_inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4128) iput(dest->ino_cache_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4129) dest->ino_cache_inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4130) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4131) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4132)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4133) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4134) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4135)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4136) static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4137) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4138) struct inode *inode = d_inode(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4139) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4140) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4141) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4142) u64 last_unlink_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4143)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4144) if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4145) return -ENOTEMPTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4146) if (btrfs_ino(BTRFS_I(inode)) == BTRFS_FIRST_FREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4147) return btrfs_delete_subvolume(dir, dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4148)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4149) trans = __unlink_start_trans(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4150) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4151) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4152)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4153) if (unlikely(btrfs_ino(BTRFS_I(inode)) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4154) err = btrfs_unlink_subvol(trans, dir, dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4155) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4156) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4158) err = btrfs_orphan_add(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4159) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4160) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4161)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4162) last_unlink_trans = BTRFS_I(inode)->last_unlink_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4164) /* now the directory is empty */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4165) err = btrfs_unlink_inode(trans, root, BTRFS_I(dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4166) BTRFS_I(d_inode(dentry)), dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4167) dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4168) if (!err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4169) btrfs_i_size_write(BTRFS_I(inode), 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4170) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4171) * Propagate the last_unlink_trans value of the deleted dir to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4172) * its parent directory. This is to prevent an unrecoverable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4173) * log tree in the case we do something like this:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4174) * 1) create dir foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4175) * 2) create snapshot under dir foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4176) * 3) delete the snapshot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4177) * 4) rmdir foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4178) * 5) mkdir foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4179) * 6) fsync foo or some file inside foo
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4180) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4181) if (last_unlink_trans >= trans->transid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4182) BTRFS_I(dir)->last_unlink_trans = last_unlink_trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4183) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4184) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4185) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4186) btrfs_btree_balance_dirty(root->fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4187)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4188) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4189) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4190)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4191) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4192) * Return this if we need to call truncate_block for the last bit of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4193) * truncate.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4194) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4195) #define NEED_TRUNCATE_BLOCK 1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4196)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4197) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4198) * this can truncate away extent items, csum items and directory items.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4199) * It starts at a high offset and removes keys until it can't find
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4200) * any higher than new_size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4201) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4202) * csum items that cross the new i_size are truncated to the new size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4203) * as well.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4204) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4205) * min_type is the minimum key type to truncate down to. If set to 0, this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4206) * will kill all the items on this inode, including the INODE_ITEM_KEY.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4207) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4208) int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4209) struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4210) struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4211) u64 new_size, u32 min_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4212) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4213) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4214) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4215) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4216) struct btrfs_file_extent_item *fi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4217) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4218) struct btrfs_key found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4219) u64 extent_start = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4220) u64 extent_num_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4221) u64 extent_offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4222) u64 item_end = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4223) u64 last_size = new_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4224) u32 found_type = (u8)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4225) int found_extent;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4226) int del_item;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4227) int pending_del_nr = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4228) int pending_del_slot = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4229) int extent_type = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4230) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4231) u64 ino = btrfs_ino(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4232) u64 bytes_deleted = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4233) bool be_nice = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4234) bool should_throttle = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4235) const u64 lock_start = ALIGN_DOWN(new_size, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4236) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4237)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4238) BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4239)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4240) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4241) * For non-free space inodes and non-shareable roots, we want to back
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4242) * off from time to time. This means all inodes in subvolume roots,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4243) * reloc roots, and data reloc roots.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4244) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4245) if (!btrfs_is_free_space_inode(BTRFS_I(inode)) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4246) test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4247) be_nice = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4248)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4249) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4250) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4251) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4252) path->reada = READA_BACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4254) if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4255) lock_extent_bits(&BTRFS_I(inode)->io_tree, lock_start, (u64)-1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4256) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4257)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4258) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4259) * We want to drop from the next block forward in case this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4260) * new size is not block aligned since we will be keeping the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4261) * last block of the extent just the way it is.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4262) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4263) btrfs_drop_extent_cache(BTRFS_I(inode), ALIGN(new_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4264) fs_info->sectorsize),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4265) (u64)-1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4266) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4267)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4268) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4269) * This function is also used to drop the items in the log tree before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4270) * we relog the inode, so if root != BTRFS_I(inode)->root, it means
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4271) * it is used to drop the logged items. So we shouldn't kill the delayed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4272) * items.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4273) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4274) if (min_type == 0 && root == BTRFS_I(inode)->root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4275) btrfs_kill_delayed_inode_items(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4276)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4277) key.objectid = ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4278) key.offset = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4279) key.type = (u8)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4280)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4281) search_again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4282) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4283) * with a 16K leaf size and 128MB extents, you can actually queue
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4284) * up a huge file in a single leaf. Most of the time that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4285) * bytes_deleted is > 0, it will be huge by the time we get here
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4286) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4287) if (be_nice && bytes_deleted > SZ_32M &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4288) btrfs_should_end_transaction(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4289) ret = -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4290) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4291) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4292)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4293) ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4294) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4295) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4297) if (ret > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4298) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4299) /* there are no items in the tree for us to truncate, we're
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4300) * done
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4301) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4302) if (path->slots[0] == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4303) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4304) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4305) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4306)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4307) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4308) u64 clear_start = 0, clear_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4309)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4310) fi = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4311) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4312) btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4313) found_type = found_key.type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4314)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4315) if (found_key.objectid != ino)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4316) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4317)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4318) if (found_type < min_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4319) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4320)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4321) item_end = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4322) if (found_type == BTRFS_EXTENT_DATA_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4323) fi = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4324) struct btrfs_file_extent_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4325) extent_type = btrfs_file_extent_type(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4326) if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4327) item_end +=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4328) btrfs_file_extent_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4330) trace_btrfs_truncate_show_fi_regular(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4331) BTRFS_I(inode), leaf, fi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4332) found_key.offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4333) } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4334) item_end += btrfs_file_extent_ram_bytes(leaf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4335) fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4336)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4337) trace_btrfs_truncate_show_fi_inline(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4338) BTRFS_I(inode), leaf, fi, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4339) found_key.offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4340) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4341) item_end--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4342) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4343) if (found_type > min_type) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4344) del_item = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4345) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4346) if (item_end < new_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4347) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4348) if (found_key.offset >= new_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4349) del_item = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4350) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4351) del_item = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4352) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4353) found_extent = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4354) /* FIXME, shrink the extent if the ref count is only 1 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4355) if (found_type != BTRFS_EXTENT_DATA_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4356) goto delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4357)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4358) if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4359) u64 num_dec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4360)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4361) clear_start = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4362) extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4363) if (!del_item) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4364) u64 orig_num_bytes =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4365) btrfs_file_extent_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4366) extent_num_bytes = ALIGN(new_size -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4367) found_key.offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4368) fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4369) clear_start = ALIGN(new_size, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4370) btrfs_set_file_extent_num_bytes(leaf, fi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4371) extent_num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4372) num_dec = (orig_num_bytes -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4373) extent_num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4374) if (test_bit(BTRFS_ROOT_SHAREABLE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4375) &root->state) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4376) extent_start != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4377) inode_sub_bytes(inode, num_dec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4378) btrfs_mark_buffer_dirty(leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4379) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4380) extent_num_bytes =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4381) btrfs_file_extent_disk_num_bytes(leaf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4382) fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4383) extent_offset = found_key.offset -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4384) btrfs_file_extent_offset(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4385)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4386) /* FIXME blocksize != 4096 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4387) num_dec = btrfs_file_extent_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4388) if (extent_start != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4389) found_extent = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4390) if (test_bit(BTRFS_ROOT_SHAREABLE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4391) &root->state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4392) inode_sub_bytes(inode, num_dec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4393) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4394) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4395) clear_len = num_dec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4396) } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4397) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4398) * we can't truncate inline items that have had
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4399) * special encodings
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4400) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4401) if (!del_item &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4402) btrfs_file_extent_encryption(leaf, fi) == 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4403) btrfs_file_extent_other_encoding(leaf, fi) == 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4404) btrfs_file_extent_compression(leaf, fi) == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4405) u32 size = (u32)(new_size - found_key.offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4406)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4407) btrfs_set_file_extent_ram_bytes(leaf, fi, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4408) size = btrfs_file_extent_calc_inline_size(size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4409) btrfs_truncate_item(path, size, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4410) } else if (!del_item) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4411) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4412) * We have to bail so the last_size is set to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4413) * just before this extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4414) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4415) ret = NEED_TRUNCATE_BLOCK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4416) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4417) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4418) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4419) * Inline extents are special, we just treat
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4420) * them as a full sector worth in the file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4421) * extent tree just for simplicity sake.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4422) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4423) clear_len = fs_info->sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4424) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4425)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4426) if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4427) inode_sub_bytes(inode, item_end + 1 - new_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4428) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4429) delete:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4430) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4431) * We use btrfs_truncate_inode_items() to clean up log trees for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4432) * multiple fsyncs, and in this case we don't want to clear the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4433) * file extent range because it's just the log.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4434) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4435) if (root == BTRFS_I(inode)->root) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4436) ret = btrfs_inode_clear_file_extent_range(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4437) clear_start, clear_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4438) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4439) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4440) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4441) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4442) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4444) if (del_item)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4445) last_size = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4446) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4447) last_size = new_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4448) if (del_item) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4449) if (!pending_del_nr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4450) /* no pending yet, add ourselves */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4451) pending_del_slot = path->slots[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4452) pending_del_nr = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4453) } else if (pending_del_nr &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4454) path->slots[0] + 1 == pending_del_slot) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4455) /* hop on the pending chunk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4456) pending_del_nr++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4457) pending_del_slot = path->slots[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4458) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4459) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4460) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4461) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4462) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4463) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4464) should_throttle = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4465)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4466) if (found_extent &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4467) root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4468) struct btrfs_ref ref = { 0 };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4469)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4470) bytes_deleted += extent_num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4471)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4472) btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4473) extent_start, extent_num_bytes, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4474) ref.real_root = root->root_key.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4475) btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4476) ino, extent_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4477) ret = btrfs_free_extent(trans, &ref);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4478) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4479) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4480) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4481) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4482) if (be_nice) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4483) if (btrfs_should_throttle_delayed_refs(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4484) should_throttle = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4485) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4486) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4487)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4488) if (found_type == BTRFS_INODE_ITEM_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4489) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4490)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4491) if (path->slots[0] == 0 ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4492) path->slots[0] != pending_del_slot ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4493) should_throttle) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4494) if (pending_del_nr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4495) ret = btrfs_del_items(trans, root, path,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4496) pending_del_slot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4497) pending_del_nr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4498) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4499) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4500) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4501) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4502) pending_del_nr = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4503) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4504) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4505)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4506) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4507) * We can generate a lot of delayed refs, so we need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4508) * throttle every once and a while and make sure we're
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4509) * adding enough space to keep up with the work we are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4510) * generating. Since we hold a transaction here we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4511) * can't flush, and we don't want to FLUSH_LIMIT because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4512) * we could have generated too many delayed refs to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4513) * actually allocate, so just bail if we're short and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4514) * let the normal reservation dance happen higher up.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4515) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4516) if (should_throttle) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4517) ret = btrfs_delayed_refs_rsv_refill(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4518) BTRFS_RESERVE_NO_FLUSH);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4519) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4520) ret = -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4521) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4522) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4523) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4524) goto search_again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4525) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4526) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4527) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4528) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4529) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4530) if (ret >= 0 && pending_del_nr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4531) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4532)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4533) err = btrfs_del_items(trans, root, path, pending_del_slot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4534) pending_del_nr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4535) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4536) btrfs_abort_transaction(trans, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4537) ret = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4538) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4539) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4540) if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4541) ASSERT(last_size >= new_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4542) if (!ret && last_size > new_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4543) last_size = new_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4544) btrfs_inode_safe_disk_i_size_write(inode, last_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4545) unlock_extent_cached(&BTRFS_I(inode)->io_tree, lock_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4546) (u64)-1, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4547) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4548)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4549) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4550) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4551) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4552)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4553) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4554) * btrfs_truncate_block - read, zero a chunk and write a block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4555) * @inode - inode that we're zeroing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4556) * @from - the offset to start zeroing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4557) * @len - the length to zero, 0 to zero the entire range respective to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4558) * offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4559) * @front - zero up to the offset instead of from the offset on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4560) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4561) * This will find the block for the "from" offset and cow the block and zero the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4562) * part we want to zero. This is used with truncate and hole punching.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4563) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4564) int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4565) int front)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4566) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4567) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4568) struct address_space *mapping = inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4569) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4570) struct btrfs_ordered_extent *ordered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4571) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4572) struct extent_changeset *data_reserved = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4573) char *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4574) bool only_release_metadata = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4575) u32 blocksize = fs_info->sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4576) pgoff_t index = from >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4577) unsigned offset = from & (blocksize - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4578) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4579) gfp_t mask = btrfs_alloc_write_mask(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4580) size_t write_bytes = blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4581) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4582) u64 block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4583) u64 block_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4584)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4585) if (IS_ALIGNED(offset, blocksize) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4586) (!len || IS_ALIGNED(len, blocksize)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4587) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4588)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4589) block_start = round_down(from, blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4590) block_end = block_start + blocksize - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4591)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4592) ret = btrfs_check_data_free_space(BTRFS_I(inode), &data_reserved,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4593) block_start, blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4594) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4595) if (btrfs_check_nocow_lock(BTRFS_I(inode), block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4596) &write_bytes) > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4597) /* For nocow case, no need to reserve data space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4598) only_release_metadata = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4599) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4600) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4601) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4602) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4603) ret = btrfs_delalloc_reserve_metadata(BTRFS_I(inode), blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4604) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4605) if (!only_release_metadata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4606) btrfs_free_reserved_data_space(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4607) data_reserved, block_start, blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4608) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4609) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4610) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4611) page = find_or_create_page(mapping, index, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4612) if (!page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4613) btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4614) block_start, blocksize, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4615) btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4616) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4617) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4618) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4619)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4620) if (!PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4621) ret = btrfs_readpage(NULL, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4622) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4623) if (page->mapping != mapping) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4624) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4625) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4626) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4627) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4628) if (!PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4629) ret = -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4630) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4631) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4632) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4633) wait_on_page_writeback(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4634)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4635) lock_extent_bits(io_tree, block_start, block_end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4636) set_page_extent_mapped(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4637)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4638) ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode), block_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4639) if (ordered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4640) unlock_extent_cached(io_tree, block_start, block_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4641) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4642) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4643) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4644) btrfs_start_ordered_extent(ordered, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4645) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4646) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4647) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4649) clear_extent_bit(&BTRFS_I(inode)->io_tree, block_start, block_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4650) EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4651) 0, 0, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4652)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4653) ret = btrfs_set_extent_delalloc(BTRFS_I(inode), block_start, block_end, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4654) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4655) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4656) unlock_extent_cached(io_tree, block_start, block_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4657) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4658) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4659) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4660)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4661) if (offset != blocksize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4662) if (!len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4663) len = blocksize - offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4664) kaddr = kmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4665) if (front)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4666) memset(kaddr + (block_start - page_offset(page)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4667) 0, offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4668) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4669) memset(kaddr + (block_start - page_offset(page)) + offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4670) 0, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4671) flush_dcache_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4672) kunmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4673) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4674) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4675) set_page_dirty(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4676) unlock_extent_cached(io_tree, block_start, block_end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4677)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4678) if (only_release_metadata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4679) set_extent_bit(&BTRFS_I(inode)->io_tree, block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4680) block_end, EXTENT_NORESERVE, NULL, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4681) GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4682)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4683) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4684) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4685) if (only_release_metadata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4686) btrfs_delalloc_release_metadata(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4687) blocksize, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4688) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4689) btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4690) block_start, blocksize, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4691) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4692) btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4693) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4694) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4695) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4696) if (only_release_metadata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4697) btrfs_check_nocow_unlock(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4698) extent_changeset_free(data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4699) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4700) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4702) static int maybe_insert_hole(struct btrfs_root *root, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4703) u64 offset, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4704) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4705) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4706) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4707) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4708)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4709) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4710) * Still need to make sure the inode looks like it's been updated so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4711) * that any holes get logged if we fsync.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4712) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4713) if (btrfs_fs_incompat(fs_info, NO_HOLES)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4714) BTRFS_I(inode)->last_trans = fs_info->generation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4715) BTRFS_I(inode)->last_sub_trans = root->log_transid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4716) BTRFS_I(inode)->last_log_commit = root->last_log_commit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4717) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4718) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4719)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4720) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4721) * 1 - for the one we're dropping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4722) * 1 - for the one we're adding
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4723) * 1 - for updating the inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4724) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4725) trans = btrfs_start_transaction(root, 3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4726) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4727) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4728)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4729) ret = btrfs_drop_extents(trans, root, inode, offset, offset + len, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4730) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4731) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4732) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4733) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4734) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4736) ret = btrfs_insert_file_extent(trans, root, btrfs_ino(BTRFS_I(inode)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4737) offset, 0, 0, len, 0, len, 0, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4738) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4739) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4740) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4741) btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4742) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4743) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4744) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4745)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4746) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4747) * This function puts in dummy file extents for the area we're creating a hole
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4748) * for. So if we are truncating this file to a larger size we need to insert
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4749) * these file extents so that btrfs_get_extent will return a EXTENT_MAP_HOLE for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4750) * the range between oldsize and size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4751) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4752) int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4753) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4754) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4755) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4756) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4757) struct extent_map *em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4758) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4759) struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4760) u64 hole_start = ALIGN(oldsize, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4761) u64 block_end = ALIGN(size, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4762) u64 last_byte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4763) u64 cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4764) u64 hole_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4765) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4766)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4767) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4768) * If our size started in the middle of a block we need to zero out the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4769) * rest of the block before we expand the i_size, otherwise we could
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4770) * expose stale data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4771) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4772) err = btrfs_truncate_block(inode, oldsize, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4773) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4774) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4775)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4776) if (size <= hole_start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4777) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4778)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4779) btrfs_lock_and_flush_ordered_range(BTRFS_I(inode), hole_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4780) block_end - 1, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4781) cur_offset = hole_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4782) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4783) em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4784) block_end - cur_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4785) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4786) err = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4787) em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4788) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4789) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4790) last_byte = min(extent_map_end(em), block_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4791) last_byte = ALIGN(last_byte, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4792) hole_size = last_byte - cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4793)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4794) if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4795) struct extent_map *hole_em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4796)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4797) err = maybe_insert_hole(root, inode, cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4798) hole_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4799) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4800) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4801)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4802) err = btrfs_inode_set_file_extent_range(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4803) cur_offset, hole_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4804) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4805) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4806)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4807) btrfs_drop_extent_cache(BTRFS_I(inode), cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4808) cur_offset + hole_size - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4809) hole_em = alloc_extent_map();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4810) if (!hole_em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4811) set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4812) &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4813) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4814) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4815) hole_em->start = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4816) hole_em->len = hole_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4817) hole_em->orig_start = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4818)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4819) hole_em->block_start = EXTENT_MAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4820) hole_em->block_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4821) hole_em->orig_block_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4822) hole_em->ram_bytes = hole_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4823) hole_em->compress_type = BTRFS_COMPRESS_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4824) hole_em->generation = fs_info->generation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4825)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4826) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4827) write_lock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4828) err = add_extent_mapping(em_tree, hole_em, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4829) write_unlock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4830) if (err != -EEXIST)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4831) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4832) btrfs_drop_extent_cache(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4833) cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4834) cur_offset +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4835) hole_size - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4836) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4837) free_extent_map(hole_em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4838) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4839) err = btrfs_inode_set_file_extent_range(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4840) cur_offset, hole_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4841) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4842) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4843) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4844) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4845) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4846) em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4847) cur_offset = last_byte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4848) if (cur_offset >= block_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4849) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4850) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4851) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4852) unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4853) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4854) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4856) static int btrfs_setsize(struct inode *inode, struct iattr *attr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4857) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4858) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4859) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4860) loff_t oldsize = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4861) loff_t newsize = attr->ia_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4862) int mask = attr->ia_valid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4863) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4864)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4865) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4866) * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4867) * special case where we need to update the times despite not having
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4868) * these flags set. For all other operations the VFS set these flags
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4869) * explicitly if it wants a timestamp update.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4870) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4871) if (newsize != oldsize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4872) inode_inc_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4873) if (!(mask & (ATTR_CTIME | ATTR_MTIME)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4874) inode->i_ctime = inode->i_mtime =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4875) current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4876) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4877)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4878) if (newsize > oldsize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4879) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4880) * Don't do an expanding truncate while snapshotting is ongoing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4881) * This is to ensure the snapshot captures a fully consistent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4882) * state of this file - if the snapshot captures this expanding
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4883) * truncation, it must capture all writes that happened before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4884) * this truncation.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4885) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4886) btrfs_drew_write_lock(&root->snapshot_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4887) ret = btrfs_cont_expand(inode, oldsize, newsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4888) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4889) btrfs_drew_write_unlock(&root->snapshot_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4890) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4891) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4892)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4893) trans = btrfs_start_transaction(root, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4894) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4895) btrfs_drew_write_unlock(&root->snapshot_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4896) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4897) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4899) i_size_write(inode, newsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4900) btrfs_inode_safe_disk_i_size_write(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4901) pagecache_isize_extended(inode, oldsize, newsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4902) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4903) btrfs_drew_write_unlock(&root->snapshot_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4904) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4905) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4906)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4907) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4908) * We're truncating a file that used to have good data down to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4909) * zero. Make sure any new writes to the file get on disk
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4910) * on close.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4911) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4912) if (newsize == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4913) set_bit(BTRFS_INODE_FLUSH_ON_CLOSE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4914) &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4915)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4916) truncate_setsize(inode, newsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4918) inode_dio_wait(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4919)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4920) ret = btrfs_truncate(inode, newsize == oldsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4921) if (ret && inode->i_nlink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4922) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4924) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4925) * Truncate failed, so fix up the in-memory size. We
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4926) * adjusted disk_i_size down as we removed extents, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4927) * wait for disk_i_size to be stable and then update the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4928) * in-memory size to match.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4929) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4930) err = btrfs_wait_ordered_range(inode, 0, (u64)-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4931) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4932) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4933) i_size_write(inode, BTRFS_I(inode)->disk_i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4934) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4935) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4936)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4937) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4938) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4939)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4940) static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4941) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4942) struct inode *inode = d_inode(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4943) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4944) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4945)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4946) if (btrfs_root_readonly(root))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4947) return -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4948)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4949) err = setattr_prepare(dentry, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4950) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4951) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4952)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4953) if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4954) err = btrfs_setsize(inode, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4955) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4956) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4957) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4958)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4959) if (attr->ia_valid) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4960) setattr_copy(inode, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4961) inode_inc_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4962) err = btrfs_dirty_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4963)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4964) if (!err && attr->ia_valid & ATTR_MODE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4965) err = posix_acl_chmod(inode, inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4966) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4968) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4969) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4970)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4971) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4972) * While truncating the inode pages during eviction, we get the VFS calling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4973) * btrfs_invalidatepage() against each page of the inode. This is slow because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4974) * the calls to btrfs_invalidatepage() result in a huge amount of calls to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4975) * lock_extent_bits() and clear_extent_bit(), which keep merging and splitting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4976) * extent_state structures over and over, wasting lots of time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4977) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4978) * Therefore if the inode is being evicted, let btrfs_invalidatepage() skip all
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4979) * those expensive operations on a per page basis and do only the ordered io
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4980) * finishing, while we release here the extent_map and extent_state structures,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4981) * without the excessive merging and splitting.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4982) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4983) static void evict_inode_truncate_pages(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4984) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4985) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4986) struct extent_map_tree *map_tree = &BTRFS_I(inode)->extent_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4987) struct rb_node *node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4989) ASSERT(inode->i_state & I_FREEING);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4990) truncate_inode_pages_final(&inode->i_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4991)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4992) write_lock(&map_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4993) while (!RB_EMPTY_ROOT(&map_tree->map.rb_root)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4994) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4995)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4996) node = rb_first_cached(&map_tree->map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4997) em = rb_entry(node, struct extent_map, rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4998) clear_bit(EXTENT_FLAG_PINNED, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4999) clear_bit(EXTENT_FLAG_LOGGING, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5000) remove_extent_mapping(map_tree, em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5001) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5002) if (need_resched()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5003) write_unlock(&map_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5004) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5005) write_lock(&map_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5006) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5007) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5008) write_unlock(&map_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5009)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5010) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5011) * Keep looping until we have no more ranges in the io tree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5012) * We can have ongoing bios started by readahead that have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5013) * their endio callback (extent_io.c:end_bio_extent_readpage)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5014) * still in progress (unlocked the pages in the bio but did not yet
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5015) * unlocked the ranges in the io tree). Therefore this means some
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5016) * ranges can still be locked and eviction started because before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5017) * submitting those bios, which are executed by a separate task (work
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5018) * queue kthread), inode references (inode->i_count) were not taken
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5019) * (which would be dropped in the end io callback of each bio).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5020) * Therefore here we effectively end up waiting for those bios and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5021) * anyone else holding locked ranges without having bumped the inode's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5022) * reference count - if we don't do it, when they access the inode's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5023) * io_tree to unlock a range it may be too late, leading to an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5024) * use-after-free issue.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5025) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5026) spin_lock(&io_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5027) while (!RB_EMPTY_ROOT(&io_tree->state)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5028) struct extent_state *state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5029) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5030) u64 start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5031) u64 end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5032) unsigned state_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5033)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5034) node = rb_first(&io_tree->state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5035) state = rb_entry(node, struct extent_state, rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5036) start = state->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5037) end = state->end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5038) state_flags = state->state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5039) spin_unlock(&io_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5040)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5041) lock_extent_bits(io_tree, start, end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5042)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5043) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5044) * If still has DELALLOC flag, the extent didn't reach disk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5045) * and its reserved space won't be freed by delayed_ref.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5046) * So we need to free its reserved space here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5047) * (Refer to comment in btrfs_invalidatepage, case 2)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5048) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5049) * Note, end is the bytenr of last byte, so we need + 1 here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5050) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5051) if (state_flags & EXTENT_DELALLOC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5052) btrfs_qgroup_free_data(BTRFS_I(inode), NULL, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5053) end - start + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5054)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5055) clear_extent_bit(io_tree, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5056) EXTENT_LOCKED | EXTENT_DELALLOC |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5057) EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 1, 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5058) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5059)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5060) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5061) spin_lock(&io_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5062) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5063) spin_unlock(&io_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5064) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5065)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5066) static struct btrfs_trans_handle *evict_refill_and_join(struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5067) struct btrfs_block_rsv *rsv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5068) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5069) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5070) struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5071) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5072) u64 delayed_refs_extra = btrfs_calc_insert_metadata_size(fs_info, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5073) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5074)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5075) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5076) * Eviction should be taking place at some place safe because of our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5077) * delayed iputs. However the normal flushing code will run delayed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5078) * iputs, so we cannot use FLUSH_ALL otherwise we'll deadlock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5079) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5080) * We reserve the delayed_refs_extra here again because we can't use
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5081) * btrfs_start_transaction(root, 0) for the same deadlocky reason as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5082) * above. We reserve our extra bit here because we generate a ton of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5083) * delayed refs activity by truncating.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5084) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5085) * If we cannot make our reservation we'll attempt to steal from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5086) * global reserve, because we really want to be able to free up space.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5087) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5088) ret = btrfs_block_rsv_refill(root, rsv, rsv->size + delayed_refs_extra,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5089) BTRFS_RESERVE_FLUSH_EVICT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5090) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5091) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5092) * Try to steal from the global reserve if there is space for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5093) * it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5094) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5095) if (btrfs_check_space_for_delayed_refs(fs_info) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5096) btrfs_block_rsv_migrate(global_rsv, rsv, rsv->size, 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5097) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5098) "could not allocate space for delete; will truncate on mount");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5099) return ERR_PTR(-ENOSPC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5100) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5101) delayed_refs_extra = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5104) trans = btrfs_join_transaction(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5105) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5106) return trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5108) if (delayed_refs_extra) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5109) trans->block_rsv = &fs_info->trans_block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5110) trans->bytes_reserved = delayed_refs_extra;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5111) btrfs_block_rsv_migrate(rsv, trans->block_rsv,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5112) delayed_refs_extra, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5113) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5114) return trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5115) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5116)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5117) void btrfs_evict_inode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5118) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5119) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5120) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5121) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5122) struct btrfs_block_rsv *rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5123) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5124)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5125) trace_btrfs_inode_evict(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5126)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5127) if (!root) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5128) clear_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5129) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5130) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5132) evict_inode_truncate_pages(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5133)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5134) if (inode->i_nlink &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5135) ((btrfs_root_refs(&root->root_item) != 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5136) root->root_key.objectid != BTRFS_ROOT_TREE_OBJECTID) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5137) btrfs_is_free_space_inode(BTRFS_I(inode))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5138) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5140) if (is_bad_inode(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5141) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5142)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5143) btrfs_free_io_failure_record(BTRFS_I(inode), 0, (u64)-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5145) if (test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5146) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5147)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5148) if (inode->i_nlink > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5149) BUG_ON(btrfs_root_refs(&root->root_item) != 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5150) root->root_key.objectid != BTRFS_ROOT_TREE_OBJECTID);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5151) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5152) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5154) ret = btrfs_commit_inode_delayed_inode(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5155) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5156) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5158) rsv = btrfs_alloc_block_rsv(fs_info, BTRFS_BLOCK_RSV_TEMP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5159) if (!rsv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5160) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5161) rsv->size = btrfs_calc_metadata_size(fs_info, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5162) rsv->failfast = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5164) btrfs_i_size_write(BTRFS_I(inode), 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5166) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5167) trans = evict_refill_and_join(root, rsv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5168) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5169) goto free_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5170)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5171) trans->block_rsv = rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5172)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5173) ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5174) trans->block_rsv = &fs_info->trans_block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5175) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5176) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5177) if (ret && ret != -ENOSPC && ret != -EAGAIN)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5178) goto free_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5179) else if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5180) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5181) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5183) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5184) * Errors here aren't a big deal, it just means we leave orphan items in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5185) * the tree. They will be cleaned up on the next mount. If the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5186) * number gets reused, cleanup deletes the orphan item without doing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5187) * anything, and unlink reuses the existing orphan item.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5188) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5189) * If it turns out that we are dropping too many of these, we might want
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5190) * to add a mechanism for retrying these after a commit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5191) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5192) trans = evict_refill_and_join(root, rsv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5193) if (!IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5194) trans->block_rsv = rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5195) btrfs_orphan_del(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5196) trans->block_rsv = &fs_info->trans_block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5197) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5198) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5200) if (!(root == fs_info->tree_root ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5201) root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5202) btrfs_return_ino(root, btrfs_ino(BTRFS_I(inode)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5203)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5204) free_rsv:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5205) btrfs_free_block_rsv(fs_info, rsv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5206) no_delete:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5207) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5208) * If we didn't successfully delete, the orphan item will still be in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5209) * the tree and we'll retry on the next mount. Again, we might also want
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5210) * to retry these periodically in the future.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5211) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5212) btrfs_remove_delayed_node(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5213) clear_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5214) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5215)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5216) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5217) * Return the key found in the dir entry in the location pointer, fill @type
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5218) * with BTRFS_FT_*, and return 0.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5219) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5220) * If no dir entries were found, returns -ENOENT.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5221) * If found a corrupted location in dir entry, returns -EUCLEAN.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5222) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5223) static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5224) struct btrfs_key *location, u8 *type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5225) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5226) const char *name = dentry->d_name.name;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5227) int namelen = dentry->d_name.len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5228) struct btrfs_dir_item *di;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5229) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5230) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5231) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5232)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5233) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5234) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5235) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5236)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5237) di = btrfs_lookup_dir_item(NULL, root, path, btrfs_ino(BTRFS_I(dir)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5238) name, namelen, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5239) if (IS_ERR_OR_NULL(di)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5240) ret = di ? PTR_ERR(di) : -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5241) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5242) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5243)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5244) btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5245) if (location->type != BTRFS_INODE_ITEM_KEY &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5246) location->type != BTRFS_ROOT_ITEM_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5247) ret = -EUCLEAN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5248) btrfs_warn(root->fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5249) "%s gets something invalid in DIR_ITEM (name %s, directory ino %llu, location(%llu %u %llu))",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5250) __func__, name, btrfs_ino(BTRFS_I(dir)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5251) location->objectid, location->type, location->offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5252) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5253) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5254) *type = btrfs_dir_type(path->nodes[0], di);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5255) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5256) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5257) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5258) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5259)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5260) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5261) * when we hit a tree root in a directory, the btrfs part of the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5262) * needs to be changed to reflect the root directory of the tree root. This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5263) * is kind of like crossing a mount point.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5264) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5265) static int fixup_tree_root_location(struct btrfs_fs_info *fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5266) struct inode *dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5267) struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5268) struct btrfs_key *location,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5269) struct btrfs_root **sub_root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5270) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5271) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5272) struct btrfs_root *new_root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5273) struct btrfs_root_ref *ref;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5274) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5275) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5276) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5277) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5278)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5279) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5280) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5281) err = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5282) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5283) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5284)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5285) err = -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5286) key.objectid = BTRFS_I(dir)->root->root_key.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5287) key.type = BTRFS_ROOT_REF_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5288) key.offset = location->objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5289)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5290) ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5291) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5292) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5293) err = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5294) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5295) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5297) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5298) ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5299) if (btrfs_root_ref_dirid(leaf, ref) != btrfs_ino(BTRFS_I(dir)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5300) btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5301) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5302)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5303) ret = memcmp_extent_buffer(leaf, dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5304) (unsigned long)(ref + 1),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5305) dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5306) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5307) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5308)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5309) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5310)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5311) new_root = btrfs_get_fs_root(fs_info, location->objectid, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5312) if (IS_ERR(new_root)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5313) err = PTR_ERR(new_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5314) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5315) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5316)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5317) *sub_root = new_root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5318) location->objectid = btrfs_root_dirid(&new_root->root_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5319) location->type = BTRFS_INODE_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5320) location->offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5321) err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5322) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5323) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5324) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5325) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5326)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5327) static void inode_tree_add(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5328) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5329) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5330) struct btrfs_inode *entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5331) struct rb_node **p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5332) struct rb_node *parent;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5333) struct rb_node *new = &BTRFS_I(inode)->rb_node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5334) u64 ino = btrfs_ino(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5336) if (inode_unhashed(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5337) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5338) parent = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5339) spin_lock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5340) p = &root->inode_tree.rb_node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5341) while (*p) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5342) parent = *p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5343) entry = rb_entry(parent, struct btrfs_inode, rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5344)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5345) if (ino < btrfs_ino(entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5346) p = &parent->rb_left;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5347) else if (ino > btrfs_ino(entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5348) p = &parent->rb_right;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5349) else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5350) WARN_ON(!(entry->vfs_inode.i_state &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5351) (I_WILL_FREE | I_FREEING)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5352) rb_replace_node(parent, new, &root->inode_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5353) RB_CLEAR_NODE(parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5354) spin_unlock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5355) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5356) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5357) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5358) rb_link_node(new, parent, p);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5359) rb_insert_color(new, &root->inode_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5360) spin_unlock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5361) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5362)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5363) static void inode_tree_del(struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5364) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5365) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5366) int empty = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5367)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5368) spin_lock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5369) if (!RB_EMPTY_NODE(&inode->rb_node)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5370) rb_erase(&inode->rb_node, &root->inode_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5371) RB_CLEAR_NODE(&inode->rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5372) empty = RB_EMPTY_ROOT(&root->inode_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5373) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5374) spin_unlock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5375)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5376) if (empty && btrfs_root_refs(&root->root_item) == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5377) spin_lock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5378) empty = RB_EMPTY_ROOT(&root->inode_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5379) spin_unlock(&root->inode_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5380) if (empty)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5381) btrfs_add_dead_root(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5382) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5383) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5384)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5385)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5386) static int btrfs_init_locked_inode(struct inode *inode, void *p)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5387) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5388) struct btrfs_iget_args *args = p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5389)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5390) inode->i_ino = args->ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5391) BTRFS_I(inode)->location.objectid = args->ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5392) BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5393) BTRFS_I(inode)->location.offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5394) BTRFS_I(inode)->root = btrfs_grab_root(args->root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5395) BUG_ON(args->root && !BTRFS_I(inode)->root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5396) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5397) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5399) static int btrfs_find_actor(struct inode *inode, void *opaque)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5400) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5401) struct btrfs_iget_args *args = opaque;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5402)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5403) return args->ino == BTRFS_I(inode)->location.objectid &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5404) args->root == BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5405) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5406)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5407) static struct inode *btrfs_iget_locked(struct super_block *s, u64 ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5408) struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5409) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5410) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5411) struct btrfs_iget_args args;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5412) unsigned long hashval = btrfs_inode_hash(ino, root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5413)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5414) args.ino = ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5415) args.root = root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5416)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5417) inode = iget5_locked(s, hashval, btrfs_find_actor,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5418) btrfs_init_locked_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5419) (void *)&args);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5420) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5421) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5422)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5423) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5424) * Get an inode object given its inode number and corresponding root.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5425) * Path can be preallocated to prevent recursing back to iget through
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5426) * allocator. NULL is also valid but may require an additional allocation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5427) * later.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5428) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5429) struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5430) struct btrfs_root *root, struct btrfs_path *path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5431) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5432) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5433)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5434) inode = btrfs_iget_locked(s, ino, root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5435) if (!inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5436) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5437)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5438) if (inode->i_state & I_NEW) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5439) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5441) ret = btrfs_read_locked_inode(inode, path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5442) if (!ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5443) inode_tree_add(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5444) unlock_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5445) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5446) iget_failed(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5447) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5448) * ret > 0 can come from btrfs_search_slot called by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5449) * btrfs_read_locked_inode, this means the inode item
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5450) * was not found.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5451) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5452) if (ret > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5453) ret = -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5454) inode = ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5455) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5456) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5458) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5459) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5460)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5461) struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5462) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5463) return btrfs_iget_path(s, ino, root, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5464) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5465)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5466) static struct inode *new_simple_dir(struct super_block *s,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5467) struct btrfs_key *key,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5468) struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5469) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5470) struct inode *inode = new_inode(s);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5471)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5472) if (!inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5473) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5474)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5475) BTRFS_I(inode)->root = btrfs_grab_root(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5476) memcpy(&BTRFS_I(inode)->location, key, sizeof(*key));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5477) set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5478)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5479) inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5480) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5481) * We only need lookup, the rest is read-only and there's no inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5482) * associated with the dentry
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5483) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5484) inode->i_op = &simple_dir_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5485) inode->i_opflags &= ~IOP_XATTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5486) inode->i_fop = &simple_dir_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5487) inode->i_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5488) inode->i_mtime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5489) inode->i_atime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5490) inode->i_ctime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5491) BTRFS_I(inode)->i_otime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5492)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5493) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5494) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5495)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5496) static inline u8 btrfs_inode_type(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5497) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5498) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5499) * Compile-time asserts that generic FT_* types still match
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5500) * BTRFS_FT_* types
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5501) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5502) BUILD_BUG_ON(BTRFS_FT_UNKNOWN != FT_UNKNOWN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5503) BUILD_BUG_ON(BTRFS_FT_REG_FILE != FT_REG_FILE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5504) BUILD_BUG_ON(BTRFS_FT_DIR != FT_DIR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5505) BUILD_BUG_ON(BTRFS_FT_CHRDEV != FT_CHRDEV);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5506) BUILD_BUG_ON(BTRFS_FT_BLKDEV != FT_BLKDEV);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5507) BUILD_BUG_ON(BTRFS_FT_FIFO != FT_FIFO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5508) BUILD_BUG_ON(BTRFS_FT_SOCK != FT_SOCK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5509) BUILD_BUG_ON(BTRFS_FT_SYMLINK != FT_SYMLINK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5510)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5511) return fs_umode_to_ftype(inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5512) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5513)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5514) struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5515) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5516) struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5517) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5518) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5519) struct btrfs_root *sub_root = root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5520) struct btrfs_key location;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5521) u8 di_type = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5522) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5524) if (dentry->d_name.len > BTRFS_NAME_LEN)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5525) return ERR_PTR(-ENAMETOOLONG);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5527) ret = btrfs_inode_by_name(dir, dentry, &location, &di_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5528) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5529) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5530)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5531) if (location.type == BTRFS_INODE_ITEM_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5532) inode = btrfs_iget(dir->i_sb, location.objectid, root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5533) if (IS_ERR(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5534) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5536) /* Do extra check against inode mode with di_type */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5537) if (btrfs_inode_type(inode) != di_type) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5538) btrfs_crit(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5539) "inode mode mismatch with dir: inode mode=0%o btrfs type=%u dir type=%u",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5540) inode->i_mode, btrfs_inode_type(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5541) di_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5542) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5543) return ERR_PTR(-EUCLEAN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5544) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5545) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5546) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5547)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5548) ret = fixup_tree_root_location(fs_info, dir, dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5549) &location, &sub_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5550) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5551) if (ret != -ENOENT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5552) inode = ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5553) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5554) inode = new_simple_dir(dir->i_sb, &location, sub_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5555) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5556) inode = btrfs_iget(dir->i_sb, location.objectid, sub_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5557) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5558) if (root != sub_root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5559) btrfs_put_root(sub_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5560)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5561) if (!IS_ERR(inode) && root != sub_root) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5562) down_read(&fs_info->cleanup_work_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5563) if (!sb_rdonly(inode->i_sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5564) ret = btrfs_orphan_cleanup(sub_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5565) up_read(&fs_info->cleanup_work_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5566) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5567) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5568) inode = ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5569) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5570) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5571)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5572) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5573) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5574)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5575) static int btrfs_dentry_delete(const struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5576) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5577) struct btrfs_root *root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5578) struct inode *inode = d_inode(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5579)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5580) if (!inode && !IS_ROOT(dentry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5581) inode = d_inode(dentry->d_parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5582)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5583) if (inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5584) root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5585) if (btrfs_root_refs(&root->root_item) == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5586) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5587)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5588) if (btrfs_ino(BTRFS_I(inode)) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5589) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5590) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5591) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5592) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5593)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5594) static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5595) unsigned int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5596) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5597) struct inode *inode = btrfs_lookup_dentry(dir, dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5598)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5599) if (inode == ERR_PTR(-ENOENT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5600) inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5601) return d_splice_alias(inode, dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5602) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5603)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5604) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5605) * All this infrastructure exists because dir_emit can fault, and we are holding
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5606) * the tree lock when doing readdir. For now just allocate a buffer and copy
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5607) * our information into that, and then dir_emit from the buffer. This is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5608) * similar to what NFS does, only we don't keep the buffer around in pagecache
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5609) * because I'm afraid I'll mess that up. Long term we need to make filldir do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5610) * copy_to_user_inatomic so we don't have to worry about page faulting under the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5611) * tree lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5612) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5613) static int btrfs_opendir(struct inode *inode, struct file *file)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5614) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5615) struct btrfs_file_private *private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5616)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5617) private = kzalloc(sizeof(struct btrfs_file_private), GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5618) if (!private)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5619) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5620) private->filldir_buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5621) if (!private->filldir_buf) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5622) kfree(private);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5623) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5624) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5625) file->private_data = private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5626) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5627) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5629) struct dir_entry {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5630) u64 ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5631) u64 offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5632) unsigned type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5633) int name_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5634) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5635)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5636) static int btrfs_filldir(void *addr, int entries, struct dir_context *ctx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5637) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5638) while (entries--) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5639) struct dir_entry *entry = addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5640) char *name = (char *)(entry + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5641)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5642) ctx->pos = get_unaligned(&entry->offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5643) if (!dir_emit(ctx, name, get_unaligned(&entry->name_len),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5644) get_unaligned(&entry->ino),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5645) get_unaligned(&entry->type)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5646) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5647) addr += sizeof(struct dir_entry) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5648) get_unaligned(&entry->name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5649) ctx->pos++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5650) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5651) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5652) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5653)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5654) static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5655) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5656) struct inode *inode = file_inode(file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5657) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5658) struct btrfs_file_private *private = file->private_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5659) struct btrfs_dir_item *di;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5660) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5661) struct btrfs_key found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5662) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5663) void *addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5664) struct list_head ins_list;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5665) struct list_head del_list;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5666) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5667) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5668) int slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5669) char *name_ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5670) int name_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5671) int entries = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5672) int total_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5673) bool put = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5674) struct btrfs_key location;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5675)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5676) if (!dir_emit_dots(file, ctx))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5677) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5678)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5679) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5680) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5681) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5682)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5683) addr = private->filldir_buf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5684) path->reada = READA_FORWARD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5685)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5686) INIT_LIST_HEAD(&ins_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5687) INIT_LIST_HEAD(&del_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5688) put = btrfs_readdir_get_delayed_items(inode, &ins_list, &del_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5689)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5690) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5691) key.type = BTRFS_DIR_INDEX_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5692) key.offset = ctx->pos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5693) key.objectid = btrfs_ino(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5694)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5695) ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5696) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5697) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5698)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5699) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5700) struct dir_entry *entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5702) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5703) slot = path->slots[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5704) if (slot >= btrfs_header_nritems(leaf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5705) ret = btrfs_next_leaf(root, path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5706) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5707) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5708) else if (ret > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5709) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5710) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5711) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5712)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5713) btrfs_item_key_to_cpu(leaf, &found_key, slot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5714)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5715) if (found_key.objectid != key.objectid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5716) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5717) if (found_key.type != BTRFS_DIR_INDEX_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5718) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5719) if (found_key.offset < ctx->pos)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5720) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5721) if (btrfs_should_delete_dir_index(&del_list, found_key.offset))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5722) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5723) di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5724) name_len = btrfs_dir_name_len(leaf, di);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5725) if ((total_len + sizeof(struct dir_entry) + name_len) >=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5726) PAGE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5727) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5728) ret = btrfs_filldir(private->filldir_buf, entries, ctx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5729) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5730) goto nopos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5731) addr = private->filldir_buf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5732) entries = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5733) total_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5734) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5735) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5737) entry = addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5738) put_unaligned(name_len, &entry->name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5739) name_ptr = (char *)(entry + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5740) read_extent_buffer(leaf, name_ptr, (unsigned long)(di + 1),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5741) name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5742) put_unaligned(fs_ftype_to_dtype(btrfs_dir_type(leaf, di)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5743) &entry->type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5744) btrfs_dir_item_key_to_cpu(leaf, di, &location);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5745) put_unaligned(location.objectid, &entry->ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5746) put_unaligned(found_key.offset, &entry->offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5747) entries++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5748) addr += sizeof(struct dir_entry) + name_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5749) total_len += sizeof(struct dir_entry) + name_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5750) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5751) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5752) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5753) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5755) ret = btrfs_filldir(private->filldir_buf, entries, ctx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5756) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5757) goto nopos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5758)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5759) ret = btrfs_readdir_delayed_dir_index(ctx, &ins_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5760) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5761) goto nopos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5763) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5764) * Stop new entries from being returned after we return the last
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5765) * entry.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5766) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5767) * New directory entries are assigned a strictly increasing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5768) * offset. This means that new entries created during readdir
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5769) * are *guaranteed* to be seen in the future by that readdir.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5770) * This has broken buggy programs which operate on names as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5771) * they're returned by readdir. Until we re-use freed offsets
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5772) * we have this hack to stop new entries from being returned
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5773) * under the assumption that they'll never reach this huge
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5774) * offset.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5775) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5776) * This is being careful not to overflow 32bit loff_t unless the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5777) * last entry requires it because doing so has broken 32bit apps
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5778) * in the past.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5779) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5780) if (ctx->pos >= INT_MAX)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5781) ctx->pos = LLONG_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5782) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5783) ctx->pos = INT_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5784) nopos:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5785) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5786) err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5787) if (put)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5788) btrfs_readdir_put_delayed_items(inode, &ins_list, &del_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5789) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5790) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5791) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5792)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5793) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5794) * This is somewhat expensive, updating the tree every time the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5795) * inode changes. But, it is most likely to find the inode in cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5796) * FIXME, needs more benchmarking...there are no reasons other than performance
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5797) * to keep or drop this code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5798) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5799) static int btrfs_dirty_inode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5800) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5801) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5802) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5803) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5804) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5806) if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5807) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5809) trans = btrfs_join_transaction(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5810) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5811) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5812)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5813) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5814) if (ret && (ret == -ENOSPC || ret == -EDQUOT)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5815) /* whoops, lets try again with the full transaction */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5816) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5817) trans = btrfs_start_transaction(root, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5818) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5819) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5820)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5821) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5822) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5823) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5824) if (BTRFS_I(inode)->delayed_node)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5825) btrfs_balance_delayed_items(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5826)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5827) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5828) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5830) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5831) * This is a copy of file_update_time. We need this so we can return error on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5832) * ENOSPC for updating the inode in the case of file write and mmap writes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5833) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5834) static int btrfs_update_time(struct inode *inode, struct timespec64 *now,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5835) int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5836) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5837) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5838) bool dirty = flags & ~S_VERSION;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5839)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5840) if (btrfs_root_readonly(root))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5841) return -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5843) if (flags & S_VERSION)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5844) dirty |= inode_maybe_inc_iversion(inode, dirty);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5845) if (flags & S_CTIME)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5846) inode->i_ctime = *now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5847) if (flags & S_MTIME)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5848) inode->i_mtime = *now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5849) if (flags & S_ATIME)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5850) inode->i_atime = *now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5851) return dirty ? btrfs_dirty_inode(inode) : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5852) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5853)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5854) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5855) * find the highest existing sequence number in a directory
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5856) * and then set the in-memory index_cnt variable to reflect
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5857) * free sequence numbers
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5858) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5859) static int btrfs_set_inode_index_count(struct btrfs_inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5860) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5861) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5862) struct btrfs_key key, found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5863) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5864) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5865) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5866)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5867) key.objectid = btrfs_ino(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5868) key.type = BTRFS_DIR_INDEX_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5869) key.offset = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5870)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5871) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5872) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5873) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5875) ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5876) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5877) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5878) /* FIXME: we should be able to handle this */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5879) if (ret == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5880) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5881) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5882)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5883) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5884) * MAGIC NUMBER EXPLANATION:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5885) * since we search a directory based on f_pos we have to start at 2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5886) * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5887) * else has to start at 2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5888) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5889) if (path->slots[0] == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5890) inode->index_cnt = 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5891) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5892) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5893)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5894) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5895)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5896) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5897) btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5899) if (found_key.objectid != btrfs_ino(inode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5900) found_key.type != BTRFS_DIR_INDEX_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5901) inode->index_cnt = 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5902) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5903) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5904)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5905) inode->index_cnt = found_key.offset + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5906) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5907) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5908) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5909) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5910)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5911) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5912) * helper to find a free sequence number in a given directory. This current
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5913) * code is very simple, later versions will do smarter things in the btree
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5914) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5915) int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5916) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5917) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5919) if (dir->index_cnt == (u64)-1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5920) ret = btrfs_inode_delayed_dir_index_count(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5921) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5922) ret = btrfs_set_inode_index_count(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5923) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5924) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5925) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5926) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5927)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5928) *index = dir->index_cnt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5929) dir->index_cnt++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5930)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5931) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5932) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5933)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5934) static int btrfs_insert_inode_locked(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5935) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5936) struct btrfs_iget_args args;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5937)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5938) args.ino = BTRFS_I(inode)->location.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5939) args.root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5940)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5941) return insert_inode_locked4(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5942) btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5943) btrfs_find_actor, &args);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5944) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5945)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5946) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5947) * Inherit flags from the parent inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5948) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5949) * Currently only the compression flags and the cow flags are inherited.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5950) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5951) static void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5952) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5953) unsigned int flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5954)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5955) if (!dir)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5956) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5957)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5958) flags = BTRFS_I(dir)->flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5959)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5960) if (flags & BTRFS_INODE_NOCOMPRESS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5961) BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5962) BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5963) } else if (flags & BTRFS_INODE_COMPRESS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5964) BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5965) BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5966) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5968) if (flags & BTRFS_INODE_NODATACOW) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5969) BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5970) if (S_ISREG(inode->i_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5971) BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5972) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5973)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5974) btrfs_sync_inode_flags_to_i_flags(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5975) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5977) static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5978) struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5979) struct inode *dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5980) const char *name, int name_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5981) u64 ref_objectid, u64 objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5982) umode_t mode, u64 *index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5983) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5984) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5985) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5986) struct btrfs_inode_item *inode_item;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5987) struct btrfs_key *location;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5988) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5989) struct btrfs_inode_ref *ref;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5990) struct btrfs_key key[2];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5991) u32 sizes[2];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5992) int nitems = name ? 2 : 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5993) unsigned long ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5994) unsigned int nofs_flag;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5995) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5996)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5997) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5998) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5999) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6000)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6001) nofs_flag = memalloc_nofs_save();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6002) inode = new_inode(fs_info->sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6003) memalloc_nofs_restore(nofs_flag);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6004) if (!inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6005) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6006) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6007) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6008)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6009) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6010) * O_TMPFILE, set link count to 0, so that after this point,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6011) * we fill in an inode item with the correct link count.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6012) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6013) if (!name)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6014) set_nlink(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6015)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6016) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6017) * we have to initialize this early, so we can reclaim the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6018) * number if we fail afterwards in this function.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6019) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6020) inode->i_ino = objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6022) if (dir && name) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6023) trace_btrfs_inode_request(dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6024)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6025) ret = btrfs_set_inode_index(BTRFS_I(dir), index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6026) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6027) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6028) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6029) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6030) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6031) } else if (dir) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6032) *index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6033) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6034) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6035) * index_cnt is ignored for everything but a dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6036) * btrfs_set_inode_index_count has an explanation for the magic
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6037) * number
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6038) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6039) BTRFS_I(inode)->index_cnt = 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6040) BTRFS_I(inode)->dir_index = *index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6041) BTRFS_I(inode)->root = btrfs_grab_root(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6042) BTRFS_I(inode)->generation = trans->transid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6043) inode->i_generation = BTRFS_I(inode)->generation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6044)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6045) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6046) * We could have gotten an inode number from somebody who was fsynced
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6047) * and then removed in this same transaction, so let's just set full
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6048) * sync since it will be a full sync anyway and this will blow away the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6049) * old info in the log.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6050) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6051) set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6052)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6053) key[0].objectid = objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6054) key[0].type = BTRFS_INODE_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6055) key[0].offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6056)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6057) sizes[0] = sizeof(struct btrfs_inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6058)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6059) if (name) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6060) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6061) * Start new inodes with an inode_ref. This is slightly more
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6062) * efficient for small numbers of hard links since they will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6063) * be packed into one item. Extended refs will kick in if we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6064) * add more hard links than can fit in the ref item.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6065) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6066) key[1].objectid = objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6067) key[1].type = BTRFS_INODE_REF_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6068) key[1].offset = ref_objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6069)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6070) sizes[1] = name_len + sizeof(*ref);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6071) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6072)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6073) location = &BTRFS_I(inode)->location;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6074) location->objectid = objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6075) location->offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6076) location->type = BTRFS_INODE_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6077)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6078) ret = btrfs_insert_inode_locked(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6079) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6080) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6081) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6082) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6084) path->leave_spinning = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6085) ret = btrfs_insert_empty_items(trans, root, path, key, sizes, nitems);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6086) if (ret != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6087) goto fail_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6088)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6089) inode_init_owner(inode, dir, mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6090) inode_set_bytes(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6091)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6092) inode->i_mtime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6093) inode->i_atime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6094) inode->i_ctime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6095) BTRFS_I(inode)->i_otime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6097) inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6098) struct btrfs_inode_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6099) memzero_extent_buffer(path->nodes[0], (unsigned long)inode_item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6100) sizeof(*inode_item));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6101) fill_inode_item(trans, path->nodes[0], inode_item, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6102)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6103) if (name) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6104) ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6105) struct btrfs_inode_ref);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6106) btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6107) btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6108) ptr = (unsigned long)(ref + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6109) write_extent_buffer(path->nodes[0], name, ptr, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6110) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6111)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6112) btrfs_mark_buffer_dirty(path->nodes[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6113) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6114)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6115) btrfs_inherit_iflags(inode, dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6116)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6117) if (S_ISREG(mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6118) if (btrfs_test_opt(fs_info, NODATASUM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6119) BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6120) if (btrfs_test_opt(fs_info, NODATACOW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6121) BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6122) BTRFS_INODE_NODATASUM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6123) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6124)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6125) inode_tree_add(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6126)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6127) trace_btrfs_inode_new(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6128) btrfs_set_inode_last_trans(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6129)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6130) btrfs_update_root_times(trans, root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6132) ret = btrfs_inode_inherit_props(trans, inode, dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6133) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6134) btrfs_err(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6135) "error inheriting props for ino %llu (root %llu): %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6136) btrfs_ino(BTRFS_I(inode)), root->root_key.objectid, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6138) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6140) fail_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6141) discard_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6142) fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6143) if (dir && name)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6144) BTRFS_I(dir)->index_cnt--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6145) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6146) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6147) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6148)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6149) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6150) * utility function to add 'inode' into 'parent_inode' with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6151) * a give name and a given sequence number.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6152) * if 'add_backref' is true, also insert a backref from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6153) * inode to the parent directory.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6154) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6155) int btrfs_add_link(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6156) struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6157) const char *name, int name_len, int add_backref, u64 index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6158) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6159) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6160) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6161) struct btrfs_root *root = parent_inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6162) u64 ino = btrfs_ino(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6163) u64 parent_ino = btrfs_ino(parent_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6164)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6165) if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6166) memcpy(&key, &inode->root->root_key, sizeof(key));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6167) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6168) key.objectid = ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6169) key.type = BTRFS_INODE_ITEM_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6170) key.offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6171) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6172)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6173) if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6174) ret = btrfs_add_root_ref(trans, key.objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6175) root->root_key.objectid, parent_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6176) index, name, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6177) } else if (add_backref) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6178) ret = btrfs_insert_inode_ref(trans, root, name, name_len, ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6179) parent_ino, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6180) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6181)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6182) /* Nothing to clean up yet */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6183) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6184) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6185)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6186) ret = btrfs_insert_dir_item(trans, name, name_len, parent_inode, &key,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6187) btrfs_inode_type(&inode->vfs_inode), index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6188) if (ret == -EEXIST || ret == -EOVERFLOW)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6189) goto fail_dir_item;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6190) else if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6191) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6192) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6193) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6194)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6195) btrfs_i_size_write(parent_inode, parent_inode->vfs_inode.i_size +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6196) name_len * 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6197) inode_inc_iversion(&parent_inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6198) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6199) * If we are replaying a log tree, we do not want to update the mtime
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6200) * and ctime of the parent directory with the current time, since the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6201) * log replay procedure is responsible for setting them to their correct
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6202) * values (the ones it had when the fsync was done).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6203) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6204) if (!test_bit(BTRFS_FS_LOG_RECOVERING, &root->fs_info->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6205) struct timespec64 now = current_time(&parent_inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6206)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6207) parent_inode->vfs_inode.i_mtime = now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6208) parent_inode->vfs_inode.i_ctime = now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6209) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6210) ret = btrfs_update_inode(trans, root, &parent_inode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6211) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6212) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6213) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6214)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6215) fail_dir_item:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6216) if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6217) u64 local_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6218) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6219) err = btrfs_del_root_ref(trans, key.objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6220) root->root_key.objectid, parent_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6221) &local_index, name, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6222) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6223) btrfs_abort_transaction(trans, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6224) } else if (add_backref) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6225) u64 local_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6226) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6227)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6228) err = btrfs_del_inode_ref(trans, root, name, name_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6229) ino, parent_ino, &local_index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6230) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6231) btrfs_abort_transaction(trans, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6232) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6233)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6234) /* Return the original error code */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6235) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6236) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6237)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6238) static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6239) struct btrfs_inode *dir, struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6240) struct btrfs_inode *inode, int backref, u64 index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6241) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6242) int err = btrfs_add_link(trans, dir, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6243) dentry->d_name.name, dentry->d_name.len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6244) backref, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6245) if (err > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6246) err = -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6247) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6249)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6250) static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6251) umode_t mode, dev_t rdev)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6252) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6253) struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6254) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6255) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6256) struct inode *inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6257) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6258) u64 objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6259) u64 index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6260)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6261) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6262) * 2 for inode item and ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6263) * 2 for dir items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6264) * 1 for xattr if selinux is on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6265) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6266) trans = btrfs_start_transaction(root, 5);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6267) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6268) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6269)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6270) err = btrfs_find_free_ino(root, &objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6271) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6272) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6273)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6274) inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6275) dentry->d_name.len, btrfs_ino(BTRFS_I(dir)), objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6276) mode, &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6277) if (IS_ERR(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6278) err = PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6279) inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6280) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6281) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6282)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6283) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6284) * If the active LSM wants to access the inode during
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6285) * d_instantiate it needs these. Smack checks to see
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6286) * if the filesystem supports xattrs by looking at the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6287) * ops vector.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6288) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6289) inode->i_op = &btrfs_special_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6290) init_special_inode(inode, inode->i_mode, rdev);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6292) err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6293) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6294) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6295)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6296) err = btrfs_add_nondir(trans, BTRFS_I(dir), dentry, BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6297) 0, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6298) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6299) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6300)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6301) btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6302) d_instantiate_new(dentry, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6303)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6304) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6305) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6306) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6307) if (err && inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6308) inode_dec_link_count(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6309) discard_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6311) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6312) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6313)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6314) static int btrfs_create(struct inode *dir, struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6315) umode_t mode, bool excl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6316) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6317) struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6318) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6319) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6320) struct inode *inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6321) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6322) u64 objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6323) u64 index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6324)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6325) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6326) * 2 for inode item and ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6327) * 2 for dir items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6328) * 1 for xattr if selinux is on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6329) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6330) trans = btrfs_start_transaction(root, 5);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6331) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6332) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6333)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6334) err = btrfs_find_free_ino(root, &objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6335) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6336) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6337)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6338) inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6339) dentry->d_name.len, btrfs_ino(BTRFS_I(dir)), objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6340) mode, &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6341) if (IS_ERR(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6342) err = PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6343) inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6344) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6345) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6346) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6347) * If the active LSM wants to access the inode during
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6348) * d_instantiate it needs these. Smack checks to see
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6349) * if the filesystem supports xattrs by looking at the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6350) * ops vector.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6351) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6352) inode->i_fop = &btrfs_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6353) inode->i_op = &btrfs_file_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6354) inode->i_mapping->a_ops = &btrfs_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6356) err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6357) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6358) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6359)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6360) err = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6361) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6362) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6364) err = btrfs_add_nondir(trans, BTRFS_I(dir), dentry, BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6365) 0, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6366) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6367) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6368)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6369) d_instantiate_new(dentry, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6370)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6371) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6372) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6373) if (err && inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6374) inode_dec_link_count(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6375) discard_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6376) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6377) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6378) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6379) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6380)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6381) static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6382) struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6383) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6384) struct btrfs_trans_handle *trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6385) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6386) struct inode *inode = d_inode(old_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6387) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6388) u64 index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6389) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6390) int drop_inode = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6392) /* do not allow sys_link's with other subvols of the same device */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6393) if (root->root_key.objectid != BTRFS_I(inode)->root->root_key.objectid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6394) return -EXDEV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6395)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6396) if (inode->i_nlink >= BTRFS_LINK_MAX)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6397) return -EMLINK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6399) err = btrfs_set_inode_index(BTRFS_I(dir), &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6400) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6401) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6402)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6403) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6404) * 2 items for inode and inode ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6405) * 2 items for dir items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6406) * 1 item for parent inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6407) * 1 item for orphan item deletion if O_TMPFILE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6408) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6409) trans = btrfs_start_transaction(root, inode->i_nlink ? 5 : 6);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6410) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6411) err = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6412) trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6413) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6414) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6415)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6416) /* There are several dir indexes for this inode, clear the cache. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6417) BTRFS_I(inode)->dir_index = 0ULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6418) inc_nlink(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6419) inode_inc_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6420) inode->i_ctime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6421) ihold(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6422) set_bit(BTRFS_INODE_COPY_EVERYTHING, &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6423)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6424) err = btrfs_add_nondir(trans, BTRFS_I(dir), dentry, BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6425) 1, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6427) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6428) drop_inode = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6429) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6430) struct dentry *parent = dentry->d_parent;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6431)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6432) err = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6433) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6434) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6435) if (inode->i_nlink == 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6436) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6437) * If new hard link count is 1, it's a file created
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6438) * with open(2) O_TMPFILE flag.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6439) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6440) err = btrfs_orphan_del(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6441) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6442) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6443) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6444) d_instantiate(dentry, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6445) btrfs_log_new_name(trans, BTRFS_I(inode), NULL, parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6446) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6447)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6448) fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6449) if (trans)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6450) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6451) if (drop_inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6452) inode_dec_link_count(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6453) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6454) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6455) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6456) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6457) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6458)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6459) static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6460) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6461) struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6462) struct inode *inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6463) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6464) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6465) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6466) u64 objectid = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6467) u64 index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6468)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6469) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6470) * 2 items for inode and ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6471) * 2 items for dir items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6472) * 1 for xattr if selinux is on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6473) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6474) trans = btrfs_start_transaction(root, 5);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6475) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6476) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6477)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6478) err = btrfs_find_free_ino(root, &objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6479) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6480) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6481)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6482) inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6483) dentry->d_name.len, btrfs_ino(BTRFS_I(dir)), objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6484) S_IFDIR | mode, &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6485) if (IS_ERR(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6486) err = PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6487) inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6488) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6489) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6490)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6491) /* these must be set before we unlock the inode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6492) inode->i_op = &btrfs_dir_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6493) inode->i_fop = &btrfs_dir_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6494)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6495) err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6496) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6497) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6498)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6499) btrfs_i_size_write(BTRFS_I(inode), 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6500) err = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6501) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6502) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6503)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6504) err = btrfs_add_link(trans, BTRFS_I(dir), BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6505) dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6506) dentry->d_name.len, 0, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6507) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6508) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6509)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6510) d_instantiate_new(dentry, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6512) out_fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6513) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6514) if (err && inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6515) inode_dec_link_count(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6516) discard_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6517) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6518) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6519) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6520) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6521)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6522) static noinline int uncompress_inline(struct btrfs_path *path,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6523) struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6524) size_t pg_offset, u64 extent_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6525) struct btrfs_file_extent_item *item)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6526) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6527) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6528) struct extent_buffer *leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6529) char *tmp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6530) size_t max_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6531) unsigned long inline_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6532) unsigned long ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6533) int compress_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6534)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6535) WARN_ON(pg_offset != 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6536) compress_type = btrfs_file_extent_compression(leaf, item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6537) max_size = btrfs_file_extent_ram_bytes(leaf, item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6538) inline_size = btrfs_file_extent_inline_item_len(leaf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6539) btrfs_item_nr(path->slots[0]));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6540) tmp = kmalloc(inline_size, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6541) if (!tmp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6542) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6543) ptr = btrfs_file_extent_inline_start(item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6544)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6545) read_extent_buffer(leaf, tmp, ptr, inline_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6546)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6547) max_size = min_t(unsigned long, PAGE_SIZE, max_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6548) ret = btrfs_decompress(compress_type, tmp, page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6549) extent_offset, inline_size, max_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6550)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6551) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6552) * decompression code contains a memset to fill in any space between the end
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6553) * of the uncompressed data and the end of max_size in case the decompressed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6554) * data ends up shorter than ram_bytes. That doesn't cover the hole between
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6555) * the end of an inline extent and the beginning of the next block, so we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6556) * cover that region here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6557) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6558)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6559) if (max_size + pg_offset < PAGE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6560) char *map = kmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6561) memset(map + pg_offset + max_size, 0, PAGE_SIZE - max_size - pg_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6562) kunmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6563) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6564) kfree(tmp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6565) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6567)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6568) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6569) * btrfs_get_extent - Lookup the first extent overlapping a range in a file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6570) * @inode: file to search in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6571) * @page: page to read extent data into if the extent is inline
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6572) * @pg_offset: offset into @page to copy to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6573) * @start: file offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6574) * @len: length of range starting at @start
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6575) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6576) * This returns the first &struct extent_map which overlaps with the given
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6577) * range, reading it from the B-tree and caching it if necessary. Note that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6578) * there may be more extents which overlap the given range after the returned
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6579) * extent_map.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6580) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6581) * If @page is not NULL and the extent is inline, this also reads the extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6582) * data directly into the page and marks the extent up to date in the io_tree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6583) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6584) * Return: ERR_PTR on error, non-NULL extent_map on success.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6585) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6586) struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6587) struct page *page, size_t pg_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6588) u64 start, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6589) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6590) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6591) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6592) u64 extent_start = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6593) u64 extent_end = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6594) u64 objectid = btrfs_ino(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6595) int extent_type = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6596) struct btrfs_path *path = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6597) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6598) struct btrfs_file_extent_item *item;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6599) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6600) struct btrfs_key found_key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6601) struct extent_map *em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6602) struct extent_map_tree *em_tree = &inode->extent_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6603) struct extent_io_tree *io_tree = &inode->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6605) read_lock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6606) em = lookup_extent_mapping(em_tree, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6607) read_unlock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6608)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6609) if (em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6610) if (em->start > start || em->start + em->len <= start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6611) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6612) else if (em->block_start == EXTENT_MAP_INLINE && page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6613) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6614) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6615) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6617) em = alloc_extent_map();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6618) if (!em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6619) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6620) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6621) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6622) em->start = EXTENT_MAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6623) em->orig_start = EXTENT_MAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6624) em->len = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6625) em->block_len = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6626)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6627) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6628) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6629) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6630) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6631) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6632)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6633) /* Chances are we'll be called again, so go ahead and do readahead */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6634) path->reada = READA_FORWARD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6635)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6636) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6637) * Unless we're going to uncompress the inline extent, no sleep would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6638) * happen.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6639) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6640) path->leave_spinning = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6641)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6642) path->recurse = btrfs_is_free_space_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6643)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6644) ret = btrfs_lookup_file_extent(NULL, root, path, objectid, start, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6645) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6646) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6647) } else if (ret > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6648) if (path->slots[0] == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6649) goto not_found;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6650) path->slots[0]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6651) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6652) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6653)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6654) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6655) item = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6656) struct btrfs_file_extent_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6657) btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6658) if (found_key.objectid != objectid ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6659) found_key.type != BTRFS_EXTENT_DATA_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6660) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6661) * If we backup past the first extent we want to move forward
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6662) * and see if there is an extent in front of us, otherwise we'll
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6663) * say there is a hole for our whole search range which can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6664) * cause problems.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6665) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6666) extent_end = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6667) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6668) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6669)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6670) extent_type = btrfs_file_extent_type(leaf, item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6671) extent_start = found_key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6672) extent_end = btrfs_file_extent_end(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6673) if (extent_type == BTRFS_FILE_EXTENT_REG ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6674) extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6675) /* Only regular file could have regular/prealloc extent */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6676) if (!S_ISREG(inode->vfs_inode.i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6677) ret = -EUCLEAN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6678) btrfs_crit(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6679) "regular/prealloc extent found for non-regular inode %llu",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6680) btrfs_ino(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6681) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6682) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6683) trace_btrfs_get_extent_show_fi_regular(inode, leaf, item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6684) extent_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6685) } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6686) trace_btrfs_get_extent_show_fi_inline(inode, leaf, item,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6687) path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6688) extent_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6689) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6690) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6691) if (start >= extent_end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6692) path->slots[0]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6693) if (path->slots[0] >= btrfs_header_nritems(leaf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6694) ret = btrfs_next_leaf(root, path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6695) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6696) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6697) else if (ret > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6698) goto not_found;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6699)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6700) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6701) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6702) btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6703) if (found_key.objectid != objectid ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6704) found_key.type != BTRFS_EXTENT_DATA_KEY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6705) goto not_found;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6706) if (start + len <= found_key.offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6707) goto not_found;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6708) if (start > found_key.offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6709) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6710)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6711) /* New extent overlaps with existing one */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6712) em->start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6713) em->orig_start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6714) em->len = found_key.offset - start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6715) em->block_start = EXTENT_MAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6716) goto insert;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6717) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6718)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6719) btrfs_extent_item_to_extent_map(inode, path, item, !page, em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6720)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6721) if (extent_type == BTRFS_FILE_EXTENT_REG ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6722) extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6723) goto insert;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6724) } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6725) unsigned long ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6726) char *map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6727) size_t size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6728) size_t extent_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6729) size_t copy_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6731) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6732) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6733)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6734) size = btrfs_file_extent_ram_bytes(leaf, item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6735) extent_offset = page_offset(page) + pg_offset - extent_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6736) copy_size = min_t(u64, PAGE_SIZE - pg_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6737) size - extent_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6738) em->start = extent_start + extent_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6739) em->len = ALIGN(copy_size, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6740) em->orig_block_len = em->len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6741) em->orig_start = em->start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6742) ptr = btrfs_file_extent_inline_start(item) + extent_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6743)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6744) btrfs_set_path_blocking(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6745) if (!PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6746) if (btrfs_file_extent_compression(leaf, item) !=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6747) BTRFS_COMPRESS_NONE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6748) ret = uncompress_inline(path, page, pg_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6749) extent_offset, item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6750) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6751) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6752) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6753) map = kmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6754) read_extent_buffer(leaf, map + pg_offset, ptr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6755) copy_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6756) if (pg_offset + copy_size < PAGE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6757) memset(map + pg_offset + copy_size, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6758) PAGE_SIZE - pg_offset -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6759) copy_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6760) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6761) kunmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6762) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6763) flush_dcache_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6764) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6765) set_extent_uptodate(io_tree, em->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6766) extent_map_end(em) - 1, NULL, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6767) goto insert;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6768) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6769) not_found:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6770) em->start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6771) em->orig_start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6772) em->len = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6773) em->block_start = EXTENT_MAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6774) insert:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6775) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6776) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6777) if (em->start > start || extent_map_end(em) <= start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6778) btrfs_err(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6779) "bad extent! em: [%llu %llu] passed [%llu %llu]",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6780) em->start, em->len, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6781) ret = -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6782) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6783) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6784)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6785) write_lock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6786) ret = btrfs_add_extent_mapping(fs_info, em_tree, &em, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6787) write_unlock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6788) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6789) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6791) trace_btrfs_get_extent(root, inode, em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6792)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6793) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6794) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6795) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6796) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6797) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6798) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6799)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6800) struct extent_map *btrfs_get_extent_fiemap(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6801) u64 start, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6802) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6803) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6804) struct extent_map *hole_em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6805) u64 delalloc_start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6806) u64 end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6807) u64 delalloc_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6808) u64 delalloc_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6809) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6810)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6811) em = btrfs_get_extent(inode, NULL, 0, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6812) if (IS_ERR(em))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6813) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6814) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6815) * If our em maps to:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6816) * - a hole or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6817) * - a pre-alloc extent,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6818) * there might actually be delalloc bytes behind it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6819) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6820) if (em->block_start != EXTENT_MAP_HOLE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6821) !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6822) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6823) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6824) hole_em = em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6825)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6826) /* check to see if we've wrapped (len == -1 or similar) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6827) end = start + len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6828) if (end < start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6829) end = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6830) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6831) end -= 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6832)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6833) em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6834)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6835) /* ok, we didn't find anything, lets look for delalloc */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6836) delalloc_len = count_range_bits(&inode->io_tree, &delalloc_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6837) end, len, EXTENT_DELALLOC, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6838) delalloc_end = delalloc_start + delalloc_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6839) if (delalloc_end < delalloc_start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6840) delalloc_end = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6841)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6842) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6843) * We didn't find anything useful, return the original results from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6844) * get_extent()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6845) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6846) if (delalloc_start > end || delalloc_end <= start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6847) em = hole_em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6848) hole_em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6849) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6850) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6851)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6852) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6853) * Adjust the delalloc_start to make sure it doesn't go backwards from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6854) * the start they passed in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6855) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6856) delalloc_start = max(start, delalloc_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6857) delalloc_len = delalloc_end - delalloc_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6859) if (delalloc_len > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6860) u64 hole_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6861) u64 hole_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6862) const u64 hole_end = extent_map_end(hole_em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6863)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6864) em = alloc_extent_map();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6865) if (!em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6866) err = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6867) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6868) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6870) ASSERT(hole_em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6871) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6872) * When btrfs_get_extent can't find anything it returns one
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6873) * huge hole
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6874) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6875) * Make sure what it found really fits our range, and adjust to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6876) * make sure it is based on the start from the caller
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6877) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6878) if (hole_end <= start || hole_em->start > end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6879) free_extent_map(hole_em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6880) hole_em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6881) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6882) hole_start = max(hole_em->start, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6883) hole_len = hole_end - hole_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6884) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6885)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6886) if (hole_em && delalloc_start > hole_start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6887) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6888) * Our hole starts before our delalloc, so we have to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6889) * return just the parts of the hole that go until the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6890) * delalloc starts
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6891) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6892) em->len = min(hole_len, delalloc_start - hole_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6893) em->start = hole_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6894) em->orig_start = hole_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6895) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6896) * Don't adjust block start at all, it is fixed at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6897) * EXTENT_MAP_HOLE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6898) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6899) em->block_start = hole_em->block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6900) em->block_len = hole_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6901) if (test_bit(EXTENT_FLAG_PREALLOC, &hole_em->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6902) set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6903) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6904) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6905) * Hole is out of passed range or it starts after
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6906) * delalloc range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6907) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6908) em->start = delalloc_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6909) em->len = delalloc_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6910) em->orig_start = delalloc_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6911) em->block_start = EXTENT_MAP_DELALLOC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6912) em->block_len = delalloc_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6913) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6914) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6915) return hole_em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6916) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6917) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6919) free_extent_map(hole_em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6920) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6921) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6922) return ERR_PTR(err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6923) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6924) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6925) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6926)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6927) static struct extent_map *btrfs_create_dio_extent(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6928) const u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6929) const u64 len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6930) const u64 orig_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6931) const u64 block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6932) const u64 block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6933) const u64 orig_block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6934) const u64 ram_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6935) const int type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6936) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6937) struct extent_map *em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6938) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6939)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6940) if (type != BTRFS_ORDERED_NOCOW) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6941) em = create_io_em(inode, start, len, orig_start, block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6942) block_len, orig_block_len, ram_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6943) BTRFS_COMPRESS_NONE, /* compress_type */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6944) type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6945) if (IS_ERR(em))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6946) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6947) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6948) ret = btrfs_add_ordered_extent_dio(inode, start, block_start, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6949) block_len, type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6950) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6951) if (em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6952) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6953) btrfs_drop_extent_cache(inode, start, start + len - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6954) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6955) em = ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6956) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6957) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6958)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6959) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6960) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6962) static struct extent_map *btrfs_new_extent_direct(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6963) u64 start, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6964) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6965) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6966) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6967) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6968) struct btrfs_key ins;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6969) u64 alloc_hint;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6970) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6972) alloc_hint = get_extent_allocation_hint(inode, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6973) ret = btrfs_reserve_extent(root, len, len, fs_info->sectorsize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6974) 0, alloc_hint, &ins, 1, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6975) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6976) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6977)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6978) em = btrfs_create_dio_extent(inode, start, ins.offset, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6979) ins.objectid, ins.offset, ins.offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6980) ins.offset, BTRFS_ORDERED_REGULAR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6981) btrfs_dec_block_group_reservations(fs_info, ins.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6982) if (IS_ERR(em))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6983) btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6984) 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6985)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6986) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6987) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6989) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6990) * Check if we can do nocow write into the range [@offset, @offset + @len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6991) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6992) * @offset: File offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6993) * @len: The length to write, will be updated to the nocow writeable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6994) * range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6995) * @orig_start: (optional) Return the original file offset of the file extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6996) * @orig_len: (optional) Return the original on-disk length of the file extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6997) * @ram_bytes: (optional) Return the ram_bytes of the file extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6998) * @strict: if true, omit optimizations that might force us into unnecessary
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6999) * cow. e.g., don't trust generation number.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7000) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7001) * This function will flush ordered extents in the range to ensure proper
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7002) * nocow checks for (nowait == false) case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7003) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7004) * Return:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7005) * >0 and update @len if we can do nocow write
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7006) * 0 if we can't do nocow write
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7007) * <0 if error happened
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7008) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7009) * NOTE: This only checks the file extents, caller is responsible to wait for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7010) * any ordered extents.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7011) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7012) noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7013) u64 *orig_start, u64 *orig_block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7014) u64 *ram_bytes, bool strict)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7015) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7016) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7017) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7018) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7019) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7020) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7021) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7022) struct btrfs_file_extent_item *fi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7023) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7024) u64 disk_bytenr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7025) u64 backref_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7026) u64 extent_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7027) u64 num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7028) int slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7029) int found_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7030) bool nocow = (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7031)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7032) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7033) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7034) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7035)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7036) ret = btrfs_lookup_file_extent(NULL, root, path,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7037) btrfs_ino(BTRFS_I(inode)), offset, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7038) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7039) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7040)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7041) slot = path->slots[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7042) if (ret == 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7043) if (slot == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7044) /* can't find the item, must cow */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7045) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7046) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7047) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7048) slot--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7049) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7050) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7051) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7052) btrfs_item_key_to_cpu(leaf, &key, slot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7053) if (key.objectid != btrfs_ino(BTRFS_I(inode)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7054) key.type != BTRFS_EXTENT_DATA_KEY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7055) /* not our file or wrong item type, must cow */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7056) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7057) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7058)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7059) if (key.offset > offset) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7060) /* Wrong offset, must cow */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7061) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7062) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7064) fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7065) found_type = btrfs_file_extent_type(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7066) if (found_type != BTRFS_FILE_EXTENT_REG &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7067) found_type != BTRFS_FILE_EXTENT_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7068) /* not a regular extent, must cow */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7069) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7070) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7071)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7072) if (!nocow && found_type == BTRFS_FILE_EXTENT_REG)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7073) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7074)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7075) extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7076) if (extent_end <= offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7077) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7078)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7079) disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7080) if (disk_bytenr == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7081) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7082)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7083) if (btrfs_file_extent_compression(leaf, fi) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7084) btrfs_file_extent_encryption(leaf, fi) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7085) btrfs_file_extent_other_encoding(leaf, fi))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7086) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7087)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7088) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7089) * Do the same check as in btrfs_cross_ref_exist but without the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7090) * unnecessary search.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7091) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7092) if (!strict &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7093) (btrfs_file_extent_generation(leaf, fi) <=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7094) btrfs_root_last_snapshot(&root->root_item)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7095) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7097) backref_offset = btrfs_file_extent_offset(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7098)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7099) if (orig_start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7100) *orig_start = key.offset - backref_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7101) *orig_block_len = btrfs_file_extent_disk_num_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7102) *ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7103) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7104)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7105) if (btrfs_extent_readonly(fs_info, disk_bytenr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7106) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7108) num_bytes = min(offset + *len, extent_end) - offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7109) if (!nocow && found_type == BTRFS_FILE_EXTENT_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7110) u64 range_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7111)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7112) range_end = round_up(offset + num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7113) root->fs_info->sectorsize) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7114) ret = test_range_bit(io_tree, offset, range_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7115) EXTENT_DELALLOC, 0, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7116) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7117) ret = -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7118) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7119) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7120) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7122) btrfs_release_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7123)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7124) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7125) * look for other files referencing this extent, if we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7126) * find any we must cow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7127) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7128)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7129) ret = btrfs_cross_ref_exist(root, btrfs_ino(BTRFS_I(inode)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7130) key.offset - backref_offset, disk_bytenr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7131) strict);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7132) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7133) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7134) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7135) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7136)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7137) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7138) * adjust disk_bytenr and num_bytes to cover just the bytes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7139) * in this extent we are about to write. If there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7140) * are any csums in that range we have to cow in order
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7141) * to keep the csums correct
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7142) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7143) disk_bytenr += backref_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7144) disk_bytenr += offset - key.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7145) if (csum_exist_in_range(fs_info, disk_bytenr, num_bytes))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7146) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7147) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7148) * all of the above have passed, it is safe to overwrite this extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7149) * without cow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7150) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7151) *len = num_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7152) ret = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7153) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7154) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7155) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7156) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7158) static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7159) struct extent_state **cached_state, bool writing)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7160) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7161) struct btrfs_ordered_extent *ordered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7162) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7164) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7165) lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7166) cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7167) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7168) * We're concerned with the entire range that we're going to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7169) * doing DIO to, so we need to make sure there's no ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7170) * extents in this range.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7171) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7172) ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), lockstart,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7173) lockend - lockstart + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7174)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7175) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7176) * We need to make sure there are no buffered pages in this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7177) * range either, we could have raced between the invalidate in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7178) * generic_file_direct_write and locking the extent. The
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7179) * invalidate needs to happen so that reads after a write do not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7180) * get stale data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7181) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7182) if (!ordered &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7183) (!writing || !filemap_range_has_page(inode->i_mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7184) lockstart, lockend)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7185) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7186)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7187) unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7188) cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7189)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7190) if (ordered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7191) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7192) * If we are doing a DIO read and the ordered extent we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7193) * found is for a buffered write, we can not wait for it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7194) * to complete and retry, because if we do so we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7195) * deadlock with concurrent buffered writes on page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7196) * locks. This happens only if our DIO read covers more
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7197) * than one extent map, if at this point has already
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7198) * created an ordered extent for a previous extent map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7199) * and locked its range in the inode's io tree, and a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7200) * concurrent write against that previous extent map's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7201) * range and this range started (we unlock the ranges
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7202) * in the io tree only when the bios complete and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7203) * buffered writes always lock pages before attempting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7204) * to lock range in the io tree).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7205) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7206) if (writing ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7207) test_bit(BTRFS_ORDERED_DIRECT, &ordered->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7208) btrfs_start_ordered_extent(ordered, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7209) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7210) ret = -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7211) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7212) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7213) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7214) * We could trigger writeback for this range (and wait
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7215) * for it to complete) and then invalidate the pages for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7216) * this range (through invalidate_inode_pages2_range()),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7217) * but that can lead us to a deadlock with a concurrent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7218) * call to readahead (a buffered read or a defrag call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7219) * triggered a readahead) on a page lock due to an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7220) * ordered dio extent we created before but did not have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7221) * yet a corresponding bio submitted (whence it can not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7222) * complete), which makes readahead wait for that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7223) * ordered extent to complete while holding a lock on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7224) * that page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7225) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7226) ret = -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7227) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7228)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7229) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7230) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7231)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7232) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7233) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7235) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7236) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7237)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7238) /* The callers of this must take lock_extent() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7239) static struct extent_map *create_io_em(struct btrfs_inode *inode, u64 start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7240) u64 len, u64 orig_start, u64 block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7241) u64 block_len, u64 orig_block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7242) u64 ram_bytes, int compress_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7243) int type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7244) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7245) struct extent_map_tree *em_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7246) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7247) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7248)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7249) ASSERT(type == BTRFS_ORDERED_PREALLOC ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7250) type == BTRFS_ORDERED_COMPRESSED ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7251) type == BTRFS_ORDERED_NOCOW ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7252) type == BTRFS_ORDERED_REGULAR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7254) em_tree = &inode->extent_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7255) em = alloc_extent_map();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7256) if (!em)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7257) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7259) em->start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7260) em->orig_start = orig_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7261) em->len = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7262) em->block_len = block_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7263) em->block_start = block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7264) em->orig_block_len = orig_block_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7265) em->ram_bytes = ram_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7266) em->generation = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7267) set_bit(EXTENT_FLAG_PINNED, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7268) if (type == BTRFS_ORDERED_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7269) set_bit(EXTENT_FLAG_FILLING, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7270) } else if (type == BTRFS_ORDERED_COMPRESSED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7271) set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7272) em->compress_type = compress_type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7273) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7274)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7275) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7276) btrfs_drop_extent_cache(inode, em->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7277) em->start + em->len - 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7278) write_lock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7279) ret = add_extent_mapping(em_tree, em, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7280) write_unlock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7281) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7282) * The caller has taken lock_extent(), who could race with us
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7283) * to add em?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7284) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7285) } while (ret == -EEXIST);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7286)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7287) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7288) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7289) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7290) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7292) /* em got 2 refs now, callers needs to do free_extent_map once. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7293) return em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7294) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7295)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7297) static int btrfs_get_blocks_direct_write(struct extent_map **map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7298) struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7299) struct btrfs_dio_data *dio_data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7300) u64 start, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7301) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7302) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7303) struct extent_map *em = *map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7304) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7305)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7306) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7307) * We don't allocate a new extent in the following cases
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7308) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7309) * 1) The inode is marked as NODATACOW. In this case we'll just use the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7310) * existing extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7311) * 2) The extent is marked as PREALLOC. We're good to go here and can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7312) * just use the extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7313) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7314) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7315) if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7316) ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7317) em->block_start != EXTENT_MAP_HOLE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7318) int type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7319) u64 block_start, orig_start, orig_block_len, ram_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7320)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7321) if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7322) type = BTRFS_ORDERED_PREALLOC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7323) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7324) type = BTRFS_ORDERED_NOCOW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7325) len = min(len, em->len - (start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7326) block_start = em->block_start + (start - em->start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7327)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7328) if (can_nocow_extent(inode, start, &len, &orig_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7329) &orig_block_len, &ram_bytes, false) == 1 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7330) btrfs_inc_nocow_writers(fs_info, block_start)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7331) struct extent_map *em2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7332)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7333) em2 = btrfs_create_dio_extent(BTRFS_I(inode), start, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7334) orig_start, block_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7335) len, orig_block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7336) ram_bytes, type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7337) btrfs_dec_nocow_writers(fs_info, block_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7338) if (type == BTRFS_ORDERED_PREALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7339) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7340) *map = em = em2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7341) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7342)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7343) if (em2 && IS_ERR(em2)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7344) ret = PTR_ERR(em2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7345) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7346) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7347) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7348) * For inode marked NODATACOW or extent marked PREALLOC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7349) * use the existing or preallocated extent, so does not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7350) * need to adjust btrfs_space_info's bytes_may_use.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7351) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7352) btrfs_free_reserved_data_space_noquota(fs_info, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7353) goto skip_cow;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7354) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7355) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7357) /* this will cow the extent */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7358) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7359) *map = em = btrfs_new_extent_direct(BTRFS_I(inode), start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7360) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7361) ret = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7362) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7363) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7365) len = min(len, em->len - (start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7366)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7367) skip_cow:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7368) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7369) * Need to update the i_size under the extent lock so buffered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7370) * readers will get the updated i_size when we unlock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7371) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7372) if (start + len > i_size_read(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7373) i_size_write(inode, start + len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7375) dio_data->reserve -= len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7376) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7377) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7378) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7380) static int btrfs_dio_iomap_begin(struct inode *inode, loff_t start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7381) loff_t length, unsigned int flags, struct iomap *iomap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7382) struct iomap *srcmap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7383) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7384) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7385) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7386) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7387) struct btrfs_dio_data *dio_data = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7388) u64 lockstart, lockend;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7389) const bool write = !!(flags & IOMAP_WRITE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7390) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7391) u64 len = length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7392) bool unlock_extents = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7393) bool sync = (current->journal_info == BTRFS_DIO_SYNC_STUB);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7394)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7395) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7396) * We used current->journal_info here to see if we were sync, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7397) * there's a lot of tests in the enospc machinery to not do flushing if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7398) * we have a journal_info set, so we need to clear this out and re-set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7399) * it in iomap_end.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7400) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7401) ASSERT(current->journal_info == NULL ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7402) current->journal_info == BTRFS_DIO_SYNC_STUB);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7403) current->journal_info = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7404)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7405) if (!write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7406) len = min_t(u64, len, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7408) lockstart = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7409) lockend = start + len - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7410)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7411) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7412) * The generic stuff only does filemap_write_and_wait_range, which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7413) * isn't enough if we've written compressed pages to this area, so we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7414) * need to flush the dirty pages again to make absolutely sure that any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7415) * outstanding dirty pages are on disk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7416) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7417) if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7418) &BTRFS_I(inode)->runtime_flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7419) ret = filemap_fdatawrite_range(inode->i_mapping, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7420) start + length - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7421) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7422) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7423) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7424)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7425) dio_data = kzalloc(sizeof(*dio_data), GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7426) if (!dio_data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7427) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7428)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7429) dio_data->sync = sync;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7430) dio_data->length = length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7431) if (write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7432) dio_data->reserve = round_up(length, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7433) ret = btrfs_delalloc_reserve_space(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7434) &dio_data->data_reserved,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7435) start, dio_data->reserve);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7436) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7437) extent_changeset_free(dio_data->data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7438) kfree(dio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7439) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7440) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7441) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7442) iomap->private = dio_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7444)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7445) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7446) * If this errors out it's because we couldn't invalidate pagecache for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7447) * this range and we need to fallback to buffered.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7448) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7449) if (lock_extent_direct(inode, lockstart, lockend, &cached_state, write)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7450) ret = -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7451) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7452) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7454) em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7455) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7456) ret = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7457) goto unlock_err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7458) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7460) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7461) * Ok for INLINE and COMPRESSED extents we need to fallback on buffered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7462) * io. INLINE is special, and we could probably kludge it in here, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7463) * it's still buffered so for safety lets just fall back to the generic
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7464) * buffered path.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7465) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7466) * For COMPRESSED we _have_ to read the entire extent in so we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7467) * decompress it, so there will be buffering required no matter what we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7468) * do, so go ahead and fallback to buffered.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7469) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7470) * We return -ENOTBLK because that's what makes DIO go ahead and go back
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7471) * to buffered IO. Don't blame me, this is the price we pay for using
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7472) * the generic code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7473) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7474) if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7475) em->block_start == EXTENT_MAP_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7476) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7477) ret = -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7478) goto unlock_err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7479) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7480)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7481) len = min(len, em->len - (start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7482) if (write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7483) ret = btrfs_get_blocks_direct_write(&em, inode, dio_data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7484) start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7485) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7486) goto unlock_err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7487) unlock_extents = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7488) /* Recalc len in case the new em is smaller than requested */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7489) len = min(len, em->len - (start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7490) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7491) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7492) * We need to unlock only the end area that we aren't using.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7493) * The rest is going to be unlocked by the endio routine.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7494) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7495) lockstart = start + len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7496) if (lockstart < lockend)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7497) unlock_extents = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7498) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7499)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7500) if (unlock_extents)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7501) unlock_extent_cached(&BTRFS_I(inode)->io_tree,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7502) lockstart, lockend, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7503) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7504) free_extent_state(cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7505)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7506) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7507) * Translate extent map information to iomap.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7508) * We trim the extents (and move the addr) even though iomap code does
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7509) * that, since we have locked only the parts we are performing I/O in.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7510) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7511) if ((em->block_start == EXTENT_MAP_HOLE) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7512) (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) && !write)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7513) iomap->addr = IOMAP_NULL_ADDR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7514) iomap->type = IOMAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7515) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7516) iomap->addr = em->block_start + (start - em->start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7517) iomap->type = IOMAP_MAPPED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7518) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7519) iomap->offset = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7520) iomap->bdev = fs_info->fs_devices->latest_bdev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7521) iomap->length = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7522)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7523) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7525) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7527) unlock_err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7528) unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7529) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7530) err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7531) if (dio_data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7532) btrfs_delalloc_release_space(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7533) dio_data->data_reserved, start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7534) dio_data->reserve, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7535) btrfs_delalloc_release_extents(BTRFS_I(inode), dio_data->reserve);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7536) extent_changeset_free(dio_data->data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7537) kfree(dio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7538) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7539) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7540) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7542) static int btrfs_dio_iomap_end(struct inode *inode, loff_t pos, loff_t length,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7543) ssize_t written, unsigned int flags, struct iomap *iomap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7544) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7545) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7546) struct btrfs_dio_data *dio_data = iomap->private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7547) size_t submitted = dio_data->submitted;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7548) const bool write = !!(flags & IOMAP_WRITE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7549)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7550) if (!write && (iomap->type == IOMAP_HOLE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7551) /* If reading from a hole, unlock and return */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7552) unlock_extent(&BTRFS_I(inode)->io_tree, pos, pos + length - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7553) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7554) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7555)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7556) if (submitted < length) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7557) pos += submitted;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7558) length -= submitted;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7559) if (write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7560) __endio_write_update_ordered(BTRFS_I(inode), pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7561) length, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7562) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7563) unlock_extent(&BTRFS_I(inode)->io_tree, pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7564) pos + length - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7565) ret = -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7567)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7568) if (write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7569) if (dio_data->reserve)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7570) btrfs_delalloc_release_space(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7571) dio_data->data_reserved, pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7572) dio_data->reserve, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7573) btrfs_delalloc_release_extents(BTRFS_I(inode), dio_data->length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7574) extent_changeset_free(dio_data->data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7575) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7576) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7577) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7578) * We're all done, we can re-set the current->journal_info now safely
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7579) * for our endio.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7580) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7581) if (dio_data->sync) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7582) ASSERT(current->journal_info == NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7583) current->journal_info = BTRFS_DIO_SYNC_STUB;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7584) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7585) kfree(dio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7586) iomap->private = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7587)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7588) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7589) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7591) static void btrfs_dio_private_put(struct btrfs_dio_private *dip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7592) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7593) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7594) * This implies a barrier so that stores to dio_bio->bi_status before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7595) * this and loads of dio_bio->bi_status after this are fully ordered.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7596) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7597) if (!refcount_dec_and_test(&dip->refs))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7598) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7599)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7600) if (bio_op(dip->dio_bio) == REQ_OP_WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7601) __endio_write_update_ordered(BTRFS_I(dip->inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7602) dip->logical_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7603) dip->bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7604) !dip->dio_bio->bi_status);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7605) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7606) unlock_extent(&BTRFS_I(dip->inode)->io_tree,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7607) dip->logical_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7608) dip->logical_offset + dip->bytes - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7609) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7610)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7611) bio_endio(dip->dio_bio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7612) kfree(dip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7613) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7614)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7615) static blk_status_t submit_dio_repair_bio(struct inode *inode, struct bio *bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7616) int mirror_num,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7617) unsigned long bio_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7618) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7619) struct btrfs_dio_private *dip = bio->bi_private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7620) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7621) blk_status_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7622)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7623) BUG_ON(bio_op(bio) == REQ_OP_WRITE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7625) ret = btrfs_bio_wq_end_io(fs_info, bio, BTRFS_WQ_ENDIO_DATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7626) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7627) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7629) refcount_inc(&dip->refs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7630) ret = btrfs_map_bio(fs_info, bio, mirror_num);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7631) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7632) refcount_dec(&dip->refs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7633) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7634) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7635)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7636) static blk_status_t btrfs_check_read_dio_bio(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7637) struct btrfs_io_bio *io_bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7638) const bool uptodate)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7639) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7640) struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7641) const u32 sectorsize = fs_info->sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7642) struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7643) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7644) const bool csum = !(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7645) struct bio_vec bvec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7646) struct bvec_iter iter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7647) u64 start = io_bio->logical;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7648) int icsum = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7649) blk_status_t err = BLK_STS_OK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7651) __bio_for_each_segment(bvec, &io_bio->bio, iter, io_bio->iter) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7652) unsigned int i, nr_sectors, pgoff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7653)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7654) nr_sectors = BTRFS_BYTES_TO_BLKS(fs_info, bvec.bv_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7655) pgoff = bvec.bv_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7656) for (i = 0; i < nr_sectors; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7657) ASSERT(pgoff < PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7658) if (uptodate &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7659) (!csum || !check_data_csum(inode, io_bio, icsum,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7660) bvec.bv_page, pgoff,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7661) start, sectorsize))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7662) clean_io_failure(fs_info, failure_tree, io_tree,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7663) start, bvec.bv_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7664) btrfs_ino(BTRFS_I(inode)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7665) pgoff);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7666) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7667) blk_status_t status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7668)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7669) status = btrfs_submit_read_repair(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7670) &io_bio->bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7671) start - io_bio->logical,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7672) bvec.bv_page, pgoff,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7673) start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7674) start + sectorsize - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7675) io_bio->mirror_num,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7676) submit_dio_repair_bio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7677) if (status)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7678) err = status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7679) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7680) start += sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7681) icsum++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7682) pgoff += sectorsize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7683) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7684) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7685) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7686) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7687)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7688) static void __endio_write_update_ordered(struct btrfs_inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7689) const u64 offset, const u64 bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7690) const bool uptodate)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7691) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7692) struct btrfs_fs_info *fs_info = inode->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7693) struct btrfs_ordered_extent *ordered = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7694) struct btrfs_workqueue *wq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7695) u64 ordered_offset = offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7696) u64 ordered_bytes = bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7697) u64 last_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7698)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7699) if (btrfs_is_free_space_inode(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7700) wq = fs_info->endio_freespace_worker;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7701) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7702) wq = fs_info->endio_write_workers;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7703)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7704) while (ordered_offset < offset + bytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7705) last_offset = ordered_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7706) if (btrfs_dec_test_first_ordered_pending(inode, &ordered,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7707) &ordered_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7708) ordered_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7709) uptodate)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7710) btrfs_init_work(&ordered->work, finish_ordered_fn, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7711) NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7712) btrfs_queue_work(wq, &ordered->work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7713) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7714) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7715) * If btrfs_dec_test_ordered_pending does not find any ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7716) * extent in the range, we can exit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7717) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7718) if (ordered_offset == last_offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7719) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7720) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7721) * Our bio might span multiple ordered extents. In this case
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7722) * we keep going until we have accounted the whole dio.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7723) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7724) if (ordered_offset < offset + bytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7725) ordered_bytes = offset + bytes - ordered_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7726) ordered = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7727) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7728) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7731) static blk_status_t btrfs_submit_bio_start_direct_io(void *private_data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7732) struct bio *bio, u64 offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7733) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7734) struct inode *inode = private_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7736) return btrfs_csum_one_bio(BTRFS_I(inode), bio, offset, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7737) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7738)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7739) static void btrfs_end_dio_bio(struct bio *bio)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7740) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7741) struct btrfs_dio_private *dip = bio->bi_private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7742) blk_status_t err = bio->bi_status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7743)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7744) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7745) btrfs_warn(BTRFS_I(dip->inode)->root->fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7746) "direct IO failed ino %llu rw %d,%u sector %#Lx len %u err no %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7747) btrfs_ino(BTRFS_I(dip->inode)), bio_op(bio),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7748) bio->bi_opf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7749) (unsigned long long)bio->bi_iter.bi_sector,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7750) bio->bi_iter.bi_size, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7751)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7752) if (bio_op(bio) == REQ_OP_READ) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7753) err = btrfs_check_read_dio_bio(dip->inode, btrfs_io_bio(bio),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7754) !err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7755) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7756)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7757) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7758) dip->dio_bio->bi_status = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7759)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7760) bio_put(bio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7761) btrfs_dio_private_put(dip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7762) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7763)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7764) static inline blk_status_t btrfs_submit_dio_bio(struct bio *bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7765) struct inode *inode, u64 file_offset, int async_submit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7766) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7767) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7768) struct btrfs_dio_private *dip = bio->bi_private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7769) bool write = bio_op(bio) == REQ_OP_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7770) blk_status_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7771)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7772) /* Check btrfs_submit_bio_hook() for rules about async submit. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7773) if (async_submit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7774) async_submit = !atomic_read(&BTRFS_I(inode)->sync_writers);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7775)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7776) if (!write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7777) ret = btrfs_bio_wq_end_io(fs_info, bio, BTRFS_WQ_ENDIO_DATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7778) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7779) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7780) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7781)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7782) if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7783) goto map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7784)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7785) if (write && async_submit) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7786) ret = btrfs_wq_submit_bio(fs_info, bio, 0, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7787) file_offset, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7788) btrfs_submit_bio_start_direct_io);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7789) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7790) } else if (write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7791) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7792) * If we aren't doing async submit, calculate the csum of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7793) * bio now.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7794) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7795) ret = btrfs_csum_one_bio(BTRFS_I(inode), bio, file_offset, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7796) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7797) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7798) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7799) u64 csum_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7801) csum_offset = file_offset - dip->logical_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7802) csum_offset >>= inode->i_sb->s_blocksize_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7803) csum_offset *= btrfs_super_csum_size(fs_info->super_copy);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7804) btrfs_io_bio(bio)->csum = dip->csums + csum_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7805) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7806) map:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7807) ret = btrfs_map_bio(fs_info, bio, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7808) err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7809) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7810) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7811)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7812) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7813) * If this succeeds, the btrfs_dio_private is responsible for cleaning up locked
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7814) * or ordered extents whether or not we submit any bios.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7815) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7816) static struct btrfs_dio_private *btrfs_create_dio_private(struct bio *dio_bio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7817) struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7818) loff_t file_offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7819) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7820) const bool write = (bio_op(dio_bio) == REQ_OP_WRITE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7821) const bool csum = !(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7822) size_t dip_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7823) struct btrfs_dio_private *dip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7825) dip_size = sizeof(*dip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7826) if (!write && csum) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7827) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7828) const u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7829) size_t nblocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7830)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7831) nblocks = dio_bio->bi_iter.bi_size >> inode->i_sb->s_blocksize_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7832) dip_size += csum_size * nblocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7833) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7834)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7835) dip = kzalloc(dip_size, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7836) if (!dip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7837) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7838)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7839) dip->inode = inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7840) dip->logical_offset = file_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7841) dip->bytes = dio_bio->bi_iter.bi_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7842) dip->disk_bytenr = (u64)dio_bio->bi_iter.bi_sector << 9;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7843) dip->dio_bio = dio_bio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7844) refcount_set(&dip->refs, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7845) return dip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7846) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7847)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7848) static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7849) struct bio *dio_bio, loff_t file_offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7850) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7851) const bool write = (bio_op(dio_bio) == REQ_OP_WRITE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7852) const bool csum = !(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7853) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7854) const bool raid56 = (btrfs_data_alloc_profile(fs_info) &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7855) BTRFS_BLOCK_GROUP_RAID56_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7856) struct btrfs_dio_private *dip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7857) struct bio *bio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7858) u64 start_sector;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7859) int async_submit = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7860) u64 submit_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7861) int clone_offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7862) int clone_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7863) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7864) blk_status_t status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7865) struct btrfs_io_geometry geom;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7866) struct btrfs_dio_data *dio_data = iomap->private;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7867)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7868) dip = btrfs_create_dio_private(dio_bio, inode, file_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7869) if (!dip) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7870) if (!write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7871) unlock_extent(&BTRFS_I(inode)->io_tree, file_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7872) file_offset + dio_bio->bi_iter.bi_size - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7874) dio_bio->bi_status = BLK_STS_RESOURCE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7875) bio_endio(dio_bio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7876) return BLK_QC_T_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7877) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7878)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7879) if (!write && csum) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7880) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7881) * Load the csums up front to reduce csum tree searches and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7882) * contention when submitting bios.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7883) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7884) status = btrfs_lookup_bio_sums(inode, dio_bio, file_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7885) dip->csums);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7886) if (status != BLK_STS_OK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7887) goto out_err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7888) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7889)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7890) start_sector = dio_bio->bi_iter.bi_sector;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7891) submit_len = dio_bio->bi_iter.bi_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7892)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7893) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7894) ret = btrfs_get_io_geometry(fs_info, btrfs_op(dio_bio),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7895) start_sector << 9, submit_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7896) &geom);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7897) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7898) status = errno_to_blk_status(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7899) goto out_err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7900) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7901) ASSERT(geom.len <= INT_MAX);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7902)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7903) clone_len = min_t(int, submit_len, geom.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7904)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7905) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7906) * This will never fail as it's passing GPF_NOFS and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7907) * the allocation is backed by btrfs_bioset.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7908) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7909) bio = btrfs_bio_clone_partial(dio_bio, clone_offset, clone_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7910) bio->bi_private = dip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7911) bio->bi_end_io = btrfs_end_dio_bio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7912) btrfs_io_bio(bio)->logical = file_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7913)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7914) ASSERT(submit_len >= clone_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7915) submit_len -= clone_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7916)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7917) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7918) * Increase the count before we submit the bio so we know
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7919) * the end IO handler won't happen before we increase the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7920) * count. Otherwise, the dip might get freed before we're
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7921) * done setting it up.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7922) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7923) * We transfer the initial reference to the last bio, so we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7924) * don't need to increment the reference count for the last one.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7925) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7926) if (submit_len > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7927) refcount_inc(&dip->refs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7928) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7929) * If we are submitting more than one bio, submit them
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7930) * all asynchronously. The exception is RAID 5 or 6, as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7931) * asynchronous checksums make it difficult to collect
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7932) * full stripe writes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7933) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7934) if (!raid56)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7935) async_submit = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7936) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7937)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7938) status = btrfs_submit_dio_bio(bio, inode, file_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7939) async_submit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7940) if (status) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7941) bio_put(bio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7942) if (submit_len > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7943) refcount_dec(&dip->refs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7944) goto out_err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7945) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7946)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7947) dio_data->submitted += clone_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7948) clone_offset += clone_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7949) start_sector += clone_len >> 9;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7950) file_offset += clone_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7951) } while (submit_len > 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7952) return BLK_QC_T_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7953)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7954) out_err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7955) dip->dio_bio->bi_status = status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7956) btrfs_dio_private_put(dip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7957) return BLK_QC_T_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7958) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7959)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7960) static ssize_t check_direct_IO(struct btrfs_fs_info *fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7961) const struct iov_iter *iter, loff_t offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7962) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7963) int seg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7964) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7965) unsigned int blocksize_mask = fs_info->sectorsize - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7966) ssize_t retval = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7968) if (offset & blocksize_mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7969) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7970)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7971) if (iov_iter_alignment(iter) & blocksize_mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7972) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7973)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7974) /* If this is a write we don't need to check anymore */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7975) if (iov_iter_rw(iter) != READ || !iter_is_iovec(iter))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7976) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7977) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7978) * Check to make sure we don't have duplicate iov_base's in this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7979) * iovec, if so return EINVAL, otherwise we'll get csum errors
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7980) * when reading back.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7981) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7982) for (seg = 0; seg < iter->nr_segs; seg++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7983) for (i = seg + 1; i < iter->nr_segs; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7984) if (iter->iov[seg].iov_base == iter->iov[i].iov_base)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7985) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7986) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7987) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7988) retval = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7989) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7990) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7991) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7992)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7993) static inline int btrfs_maybe_fsync_end_io(struct kiocb *iocb, ssize_t size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7994) int error, unsigned flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7995) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7996) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7997) * Now if we're still in the context of our submitter we know we can't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7998) * safely run generic_write_sync(), so clear our flag here so that the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7999) * caller knows to follow up with a sync.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8000) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8001) if (current->journal_info == BTRFS_DIO_SYNC_STUB) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8002) current->journal_info = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8003) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8004) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8005)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8006) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8007) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8008)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8009) if (size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8010) iocb->ki_flags |= IOCB_DSYNC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8011) return generic_write_sync(iocb, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8012) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8013)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8014) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8015) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8016)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8017) static const struct iomap_ops btrfs_dio_iomap_ops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8018) .iomap_begin = btrfs_dio_iomap_begin,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8019) .iomap_end = btrfs_dio_iomap_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8020) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8022) static const struct iomap_dio_ops btrfs_dio_ops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8023) .submit_io = btrfs_submit_direct,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8024) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8025)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8026) static const struct iomap_dio_ops btrfs_sync_dops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8027) .submit_io = btrfs_submit_direct,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8028) .end_io = btrfs_maybe_fsync_end_io,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8029) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8030)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8031) ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8032) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8033) struct file *file = iocb->ki_filp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8034) struct inode *inode = file->f_mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8035) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8036) struct extent_changeset *data_reserved = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8037) loff_t offset = iocb->ki_pos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8038) size_t count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8039) bool relock = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8040) ssize_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8041)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8042) if (check_direct_IO(fs_info, iter, offset)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8043) ASSERT(current->journal_info == NULL ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8044) current->journal_info == BTRFS_DIO_SYNC_STUB);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8045) current->journal_info = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8046) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8047) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8048)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8049) count = iov_iter_count(iter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8050) if (iov_iter_rw(iter) == WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8051) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8052) * If the write DIO is beyond the EOF, we need update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8053) * the isize, but it is protected by i_mutex. So we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8054) * not unlock the i_mutex at this case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8055) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8056) if (offset + count <= inode->i_size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8057) inode_unlock(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8058) relock = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8059) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8060) down_read(&BTRFS_I(inode)->dio_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8061) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8062)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8063) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8064) * We have are actually a sync iocb, so we need our fancy endio to know
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8065) * if we need to sync.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8066) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8067) if (current->journal_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8068) ret = iomap_dio_rw(iocb, iter, &btrfs_dio_iomap_ops,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8069) &btrfs_sync_dops, is_sync_kiocb(iocb));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8070) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8071) ret = iomap_dio_rw(iocb, iter, &btrfs_dio_iomap_ops,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8072) &btrfs_dio_ops, is_sync_kiocb(iocb));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8073)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8074) if (ret == -ENOTBLK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8075) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8076)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8077) if (iov_iter_rw(iter) == WRITE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8078) up_read(&BTRFS_I(inode)->dio_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8079)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8080) if (relock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8081) inode_lock(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8082)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8083) extent_changeset_free(data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8084) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8085) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8086)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8087) static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8088) u64 start, u64 len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8089) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8090) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8091)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8092) ret = fiemap_prep(inode, fieinfo, start, &len, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8093) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8094) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8095)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8096) return extent_fiemap(BTRFS_I(inode), fieinfo, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8097) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8098)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8099) int btrfs_readpage(struct file *file, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8100) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8101) struct btrfs_inode *inode = BTRFS_I(page->mapping->host);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8102) u64 start = page_offset(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8103) u64 end = start + PAGE_SIZE - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8104) unsigned long bio_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8105) struct bio *bio = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8106) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8108) btrfs_lock_and_flush_ordered_range(inode, start, end, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8109)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8110) ret = btrfs_do_readpage(page, NULL, &bio, &bio_flags, 0, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8111) if (bio)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8112) ret = submit_one_bio(bio, 0, bio_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8113) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8114) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8116) static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8117) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8118) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8119) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8120)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8121) if (current->flags & PF_MEMALLOC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8122) redirty_page_for_writepage(wbc, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8123) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8124) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8125) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8126)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8127) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8128) * If we are under memory pressure we will call this directly from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8129) * VM, we need to make sure we have the inode referenced for the ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8130) * extent. If not just return like we didn't do anything.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8131) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8132) if (!igrab(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8133) redirty_page_for_writepage(wbc, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8134) return AOP_WRITEPAGE_ACTIVATE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8135) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8136) ret = extent_write_full_page(page, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8137) btrfs_add_delayed_iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8138) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8139) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8141) static int btrfs_writepages(struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8142) struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8143) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8144) return extent_writepages(mapping, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8145) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8147) static void btrfs_readahead(struct readahead_control *rac)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8148) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8149) extent_readahead(rac);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8150) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8152) static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8153) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8154) int ret = try_release_extent_mapping(page, gfp_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8155) if (ret == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8156) detach_page_private(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8157) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8158) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8160) static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8161) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8162) if (PageWriteback(page) || PageDirty(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8163) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8164) return __btrfs_releasepage(page, gfp_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8165) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8166)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8167) #ifdef CONFIG_MIGRATION
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8168) static int btrfs_migratepage(struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8169) struct page *newpage, struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8170) enum migrate_mode mode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8171) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8172) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8173)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8174) ret = migrate_page_move_mapping(mapping, newpage, page, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8175) if (ret != MIGRATEPAGE_SUCCESS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8176) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8177)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8178) if (page_has_private(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8179) attach_page_private(newpage, detach_page_private(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8181) if (PagePrivate2(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8182) ClearPagePrivate2(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8183) SetPagePrivate2(newpage);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8184) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8185)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8186) if (mode != MIGRATE_SYNC_NO_COPY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8187) migrate_page_copy(newpage, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8188) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8189) migrate_page_states(newpage, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8190) return MIGRATEPAGE_SUCCESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8191) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8192) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8194) static void btrfs_invalidatepage(struct page *page, unsigned int offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8195) unsigned int length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8196) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8197) struct btrfs_inode *inode = BTRFS_I(page->mapping->host);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8198) struct extent_io_tree *tree = &inode->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8199) struct btrfs_ordered_extent *ordered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8200) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8201) u64 page_start = page_offset(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8202) u64 page_end = page_start + PAGE_SIZE - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8203) u64 start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8204) u64 end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8205) int inode_evicting = inode->vfs_inode.i_state & I_FREEING;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8206)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8207) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8208) * we have the page locked, so new writeback can't start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8209) * and the dirty bit won't be cleared while we are here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8210) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8211) * Wait for IO on this page so that we can safely clear
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8212) * the PagePrivate2 bit and do ordered accounting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8213) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8214) wait_on_page_writeback(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8215)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8216) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8217) * For subpage case, we have call sites like
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8218) * btrfs_punch_hole_lock_range() which passes range not aligned to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8219) * sectorsize.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8220) * If the range doesn't cover the full page, we don't need to and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8221) * shouldn't clear page extent mapped, as page->private can still
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8222) * record subpage dirty bits for other part of the range.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8223) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8224) * For cases that can invalidate the full even the range doesn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8225) * cover the full page, like invalidating the last page, we're
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8226) * still safe to wait for ordered extent to finish.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8227) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8228) if (!(offset == 0 && length == PAGE_SIZE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8229) btrfs_releasepage(page, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8230) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8231) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8232)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8233) if (!inode_evicting)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8234) lock_extent_bits(tree, page_start, page_end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8235)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8236) start = page_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8237) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8238) ordered = btrfs_lookup_ordered_range(inode, start, page_end - start + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8239) if (ordered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8240) end = min(page_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8241) ordered->file_offset + ordered->num_bytes - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8242) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8243) * IO on this page will never be started, so we need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8244) * to account for any ordered extents now
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8245) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8246) if (!inode_evicting)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8247) clear_extent_bit(tree, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8248) EXTENT_DELALLOC | EXTENT_DELALLOC_NEW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8249) EXTENT_LOCKED | EXTENT_DO_ACCOUNTING |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8250) EXTENT_DEFRAG, 1, 0, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8251) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8252) * whoever cleared the private bit is responsible
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8253) * for the finish_ordered_io
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8254) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8255) if (TestClearPagePrivate2(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8256) struct btrfs_ordered_inode_tree *tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8257) u64 new_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8259) tree = &inode->ordered_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8260)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8261) spin_lock_irq(&tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8262) set_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8263) new_len = start - ordered->file_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8264) if (new_len < ordered->truncated_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8265) ordered->truncated_len = new_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8266) spin_unlock_irq(&tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8267)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8268) if (btrfs_dec_test_ordered_pending(inode, &ordered,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8269) start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8270) end - start + 1, 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8271) btrfs_finish_ordered_io(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8272) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8273) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8274) if (!inode_evicting) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8275) cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8276) lock_extent_bits(tree, start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8277) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8278) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8279)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8280) start = end + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8281) if (start < page_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8282) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8283) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8284)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8285) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8286) * Qgroup reserved space handler
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8287) * Page here will be either
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8288) * 1) Already written to disk or ordered extent already submitted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8289) * Then its QGROUP_RESERVED bit in io_tree is already cleaned.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8290) * Qgroup will be handled by its qgroup_record then.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8291) * btrfs_qgroup_free_data() call will do nothing here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8292) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8293) * 2) Not written to disk yet
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8294) * Then btrfs_qgroup_free_data() call will clear the QGROUP_RESERVED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8295) * bit of its io_tree, and free the qgroup reserved data space.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8296) * Since the IO will never happen for this page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8297) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8298) btrfs_qgroup_free_data(inode, NULL, page_start, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8299) if (!inode_evicting) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8300) clear_extent_bit(tree, page_start, page_end, EXTENT_LOCKED |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8301) EXTENT_DELALLOC | EXTENT_DELALLOC_NEW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8302) EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 1, 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8303) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8304)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8305) __btrfs_releasepage(page, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8306) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8307)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8308) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8309) detach_page_private(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8312) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8313) * btrfs_page_mkwrite() is not allowed to change the file size as it gets
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8314) * called from a page fault handler when a page is first dirtied. Hence we must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8315) * be careful to check for EOF conditions here. We set the page up correctly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8316) * for a written page which means we get ENOSPC checking when writing into
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8317) * holes and correct delalloc and unwritten extent mapping on filesystems that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8318) * support these features.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8319) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8320) * We are not allowed to take the i_mutex here so we have to play games to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8321) * protect against truncate races as the page could now be beyond EOF. Because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8322) * truncate_setsize() writes the inode size before removing pages, once we have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8323) * the page lock we can determine safely if the page is beyond EOF. If it is not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8324) * beyond EOF, then the page is guaranteed safe against truncation until we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8325) * unlock the page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8326) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8327) vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8328) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8329) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8330) struct inode *inode = file_inode(vmf->vma->vm_file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8331) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8332) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8333) struct btrfs_ordered_extent *ordered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8334) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8335) struct extent_changeset *data_reserved = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8336) char *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8337) unsigned long zero_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8338) loff_t size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8339) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8340) int ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8341) int reserved = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8342) u64 reserved_space;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8343) u64 page_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8344) u64 page_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8345) u64 end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8346)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8347) reserved_space = PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8349) sb_start_pagefault(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8350) page_start = page_offset(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8351) page_end = page_start + PAGE_SIZE - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8352) end = page_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8353)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8354) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8355) * Reserving delalloc space after obtaining the page lock can lead to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8356) * deadlock. For example, if a dirty page is locked by this function
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8357) * and the call to btrfs_delalloc_reserve_space() ends up triggering
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8358) * dirty page write out, then the btrfs_writepage() function could
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8359) * end up waiting indefinitely to get a lock on the page currently
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8360) * being processed by btrfs_page_mkwrite() function.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8361) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8362) ret2 = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8363) page_start, reserved_space);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8364) if (!ret2) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8365) ret2 = file_update_time(vmf->vma->vm_file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8366) reserved = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8367) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8368) if (ret2) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8369) ret = vmf_error(ret2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8370) if (reserved)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8371) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8372) goto out_noreserve;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8373) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8375) ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8376) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8377) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8378) size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8380) if ((page->mapping != inode->i_mapping) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8381) (page_start >= size)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8382) /* page got truncated out from underneath us */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8383) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8384) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8385) wait_on_page_writeback(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8387) lock_extent_bits(io_tree, page_start, page_end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8388) set_page_extent_mapped(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8389)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8390) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8391) * we can't set the delalloc bits if there are pending ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8392) * extents. Drop our locks and wait for them to finish
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8393) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8394) ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), page_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8395) PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8396) if (ordered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8397) unlock_extent_cached(io_tree, page_start, page_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8398) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8399) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8400) btrfs_start_ordered_extent(ordered, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8401) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8402) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8403) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8404)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8405) if (page->index == ((size - 1) >> PAGE_SHIFT)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8406) reserved_space = round_up(size - page_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8407) fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8408) if (reserved_space < PAGE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8409) end = page_start + reserved_space - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8410) btrfs_delalloc_release_space(BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8411) data_reserved, page_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8412) PAGE_SIZE - reserved_space, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8413) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8414) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8415)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8416) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8417) * page_mkwrite gets called when the page is firstly dirtied after it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8418) * faulted in, but write(2) could also dirty a page and set delalloc
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8419) * bits, thus in this case for space account reason, we still need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8420) * clear any delalloc bits within this page range since we have to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8421) * reserve data&meta space before lock_page() (see above comments).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8422) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8423) clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8424) EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8425) EXTENT_DEFRAG, 0, 0, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8427) ret2 = btrfs_set_extent_delalloc(BTRFS_I(inode), page_start, end, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8428) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8429) if (ret2) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8430) unlock_extent_cached(io_tree, page_start, page_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8431) &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8432) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8433) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8434) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8436) /* page is wholly or partially inside EOF */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8437) if (page_start + PAGE_SIZE > size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8438) zero_start = offset_in_page(size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8439) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8440) zero_start = PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8441)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8442) if (zero_start != PAGE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8443) kaddr = kmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8444) memset(kaddr + zero_start, 0, PAGE_SIZE - zero_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8445) flush_dcache_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8446) kunmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8447) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8448) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8449) set_page_dirty(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8450) SetPageUptodate(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8451)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8452) btrfs_set_inode_last_sub_trans(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8454) unlock_extent_cached(io_tree, page_start, page_end, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8455)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8456) btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8457) sb_end_pagefault(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8458) extent_changeset_free(data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8459) return VM_FAULT_LOCKED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8460)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8461) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8462) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8463) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8464) btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8465) btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved, page_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8466) reserved_space, (ret != 0));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8467) out_noreserve:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8468) sb_end_pagefault(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8469) extent_changeset_free(data_reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8470) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8471) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8472)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8473) static int btrfs_truncate(struct inode *inode, bool skip_writeback)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8474) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8475) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8476) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8477) struct btrfs_block_rsv *rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8478) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8479) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8480) u64 mask = fs_info->sectorsize - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8481) u64 min_size = btrfs_calc_metadata_size(fs_info, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8482)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8483) if (!skip_writeback) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8484) ret = btrfs_wait_ordered_range(inode, inode->i_size & (~mask),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8485) (u64)-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8486) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8487) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8488) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8489)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8490) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8491) * Yes ladies and gentlemen, this is indeed ugly. We have a couple of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8492) * things going on here:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8493) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8494) * 1) We need to reserve space to update our inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8495) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8496) * 2) We need to have something to cache all the space that is going to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8497) * be free'd up by the truncate operation, but also have some slack
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8498) * space reserved in case it uses space during the truncate (thank you
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8499) * very much snapshotting).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8500) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8501) * And we need these to be separate. The fact is we can use a lot of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8502) * space doing the truncate, and we have no earthly idea how much space
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8503) * we will use, so we need the truncate reservation to be separate so it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8504) * doesn't end up using space reserved for updating the inode. We also
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8505) * need to be able to stop the transaction and start a new one, which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8506) * means we need to be able to update the inode several times, and we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8507) * have no idea of knowing how many times that will be, so we can't just
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8508) * reserve 1 item for the entirety of the operation, so that has to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8509) * done separately as well.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8510) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8511) * So that leaves us with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8512) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8513) * 1) rsv - for the truncate reservation, which we will steal from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8514) * transaction reservation.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8515) * 2) fs_info->trans_block_rsv - this will have 1 items worth left for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8516) * updating the inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8517) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8518) rsv = btrfs_alloc_block_rsv(fs_info, BTRFS_BLOCK_RSV_TEMP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8519) if (!rsv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8520) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8521) rsv->size = min_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8522) rsv->failfast = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8524) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8525) * 1 for the truncate slack space
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8526) * 1 for updating the inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8527) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8528) trans = btrfs_start_transaction(root, 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8529) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8530) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8531) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8532) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8533)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8534) /* Migrate the slack space for the truncate to our reserve */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8535) ret = btrfs_block_rsv_migrate(&fs_info->trans_block_rsv, rsv,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8536) min_size, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8537) BUG_ON(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8538)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8539) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8540) * So if we truncate and then write and fsync we normally would just
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8541) * write the extents that changed, which is a problem if we need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8542) * first truncate that entire inode. So set this flag so we write out
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8543) * all of the extents in the inode to the sync log so we're completely
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8544) * safe.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8545) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8546) set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8547) trans->block_rsv = rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8548)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8549) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8550) ret = btrfs_truncate_inode_items(trans, root, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8551) inode->i_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8552) BTRFS_EXTENT_DATA_KEY);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8553) trans->block_rsv = &fs_info->trans_block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8554) if (ret != -ENOSPC && ret != -EAGAIN)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8555) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8556)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8557) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8558) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8559) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8560)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8561) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8562) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8564) trans = btrfs_start_transaction(root, 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8565) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8566) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8567) trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8568) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8569) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8570)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8571) btrfs_block_rsv_release(fs_info, rsv, -1, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8572) ret = btrfs_block_rsv_migrate(&fs_info->trans_block_rsv,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8573) rsv, min_size, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8574) BUG_ON(ret); /* shouldn't happen */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8575) trans->block_rsv = rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8576) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8577)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8578) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8579) * We can't call btrfs_truncate_block inside a trans handle as we could
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8580) * deadlock with freeze, if we got NEED_TRUNCATE_BLOCK then we know
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8581) * we've truncated everything except the last little bit, and can do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8582) * btrfs_truncate_block and then update the disk_i_size.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8583) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8584) if (ret == NEED_TRUNCATE_BLOCK) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8585) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8586) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8587)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8588) ret = btrfs_truncate_block(inode, inode->i_size, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8589) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8590) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8591) trans = btrfs_start_transaction(root, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8592) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8593) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8594) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8595) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8596) btrfs_inode_safe_disk_i_size_write(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8597) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8598)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8599) if (trans) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8600) int ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8601)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8602) trans->block_rsv = &fs_info->trans_block_rsv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8603) ret2 = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8604) if (ret2 && !ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8605) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8606)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8607) ret2 = btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8608) if (ret2 && !ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8609) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8610) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8611) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8612) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8613) btrfs_free_block_rsv(fs_info, rsv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8614)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8615) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8618) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8619) * create a new subvolume directory/inode (helper for the ioctl).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8620) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8621) int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8622) struct btrfs_root *new_root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8623) struct btrfs_root *parent_root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8624) u64 new_dirid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8625) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8626) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8627) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8628) u64 index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8629)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8630) inode = btrfs_new_inode(trans, new_root, NULL, "..", 2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8631) new_dirid, new_dirid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8632) S_IFDIR | (~current_umask() & S_IRWXUGO),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8633) &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8634) if (IS_ERR(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8635) return PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8636) inode->i_op = &btrfs_dir_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8637) inode->i_fop = &btrfs_dir_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8638)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8639) set_nlink(inode, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8640) btrfs_i_size_write(BTRFS_I(inode), 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8641) unlock_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8643) err = btrfs_subvol_inherit_props(trans, new_root, parent_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8644) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8645) btrfs_err(new_root->fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8646) "error inheriting subvolume %llu properties: %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8647) new_root->root_key.objectid, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8649) err = btrfs_update_inode(trans, new_root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8651) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8652) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8653) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8654)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8655) struct inode *btrfs_alloc_inode(struct super_block *sb)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8656) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8657) struct btrfs_fs_info *fs_info = btrfs_sb(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8658) struct btrfs_inode *ei;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8659) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8660)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8661) ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8662) if (!ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8663) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8664)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8665) ei->root = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8666) ei->generation = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8667) ei->last_trans = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8668) ei->last_sub_trans = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8669) ei->logged_trans = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8670) ei->delalloc_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8671) ei->new_delalloc_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8672) ei->defrag_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8673) ei->disk_i_size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8674) ei->flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8675) ei->csum_bytes = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8676) ei->index_cnt = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8677) ei->dir_index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8678) ei->last_unlink_trans = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8679) ei->last_reflink_trans = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8680) ei->last_log_commit = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8681)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8682) spin_lock_init(&ei->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8683) ei->outstanding_extents = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8684) if (sb->s_magic != BTRFS_TEST_MAGIC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8685) btrfs_init_metadata_block_rsv(fs_info, &ei->block_rsv,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8686) BTRFS_BLOCK_RSV_DELALLOC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8687) ei->runtime_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8688) ei->prop_compress = BTRFS_COMPRESS_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8689) ei->defrag_compress = BTRFS_COMPRESS_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8690)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8691) ei->delayed_node = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8692)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8693) ei->i_otime.tv_sec = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8694) ei->i_otime.tv_nsec = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8695)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8696) inode = &ei->vfs_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8697) extent_map_tree_init(&ei->extent_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8698) extent_io_tree_init(fs_info, &ei->io_tree, IO_TREE_INODE_IO, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8699) extent_io_tree_init(fs_info, &ei->io_failure_tree,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8700) IO_TREE_INODE_IO_FAILURE, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8701) extent_io_tree_init(fs_info, &ei->file_extent_tree,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8702) IO_TREE_INODE_FILE_EXTENT, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8703) ei->io_tree.track_uptodate = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8704) ei->io_failure_tree.track_uptodate = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8705) atomic_set(&ei->sync_writers, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8706) mutex_init(&ei->log_mutex);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8707) btrfs_ordered_inode_tree_init(&ei->ordered_tree);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8708) INIT_LIST_HEAD(&ei->delalloc_inodes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8709) INIT_LIST_HEAD(&ei->delayed_iput);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8710) RB_CLEAR_NODE(&ei->rb_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8711) init_rwsem(&ei->dio_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8712)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8713) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8714) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8715)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8716) #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8717) void btrfs_test_destroy_inode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8718) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8719) btrfs_drop_extent_cache(BTRFS_I(inode), 0, (u64)-1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8720) kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8721) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8722) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8724) void btrfs_free_inode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8725) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8726) kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8727) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8728)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8729) void btrfs_destroy_inode(struct inode *vfs_inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8730) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8731) struct btrfs_ordered_extent *ordered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8732) struct btrfs_inode *inode = BTRFS_I(vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8733) struct btrfs_root *root = inode->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8735) WARN_ON(!hlist_empty(&vfs_inode->i_dentry));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8736) WARN_ON(vfs_inode->i_data.nrpages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8737) WARN_ON(inode->block_rsv.reserved);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8738) WARN_ON(inode->block_rsv.size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8739) WARN_ON(inode->outstanding_extents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8740) WARN_ON(inode->delalloc_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8741) WARN_ON(inode->new_delalloc_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8742) WARN_ON(inode->csum_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8743) WARN_ON(inode->defrag_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8744)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8745) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8746) * This can happen where we create an inode, but somebody else also
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8747) * created the same inode and we need to destroy the one we already
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8748) * created.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8749) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8750) if (!root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8751) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8752)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8753) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8754) ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8755) if (!ordered)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8756) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8757) else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8758) btrfs_err(root->fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8759) "found ordered extent %llu %llu on inode cleanup",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8760) ordered->file_offset, ordered->num_bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8761) btrfs_remove_ordered_extent(inode, ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8762) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8763) btrfs_put_ordered_extent(ordered);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8764) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8765) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8766) btrfs_qgroup_check_reserved_leak(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8767) inode_tree_del(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8768) btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8769) btrfs_inode_clear_file_extent_range(inode, 0, (u64)-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8770) btrfs_put_root(inode->root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8771) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8772)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8773) int btrfs_drop_inode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8774) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8775) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8776)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8777) if (root == NULL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8778) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8779)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8780) /* the snap/subvol tree is on deleting */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8781) if (btrfs_root_refs(&root->root_item) == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8782) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8783) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8784) return generic_drop_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8785) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8786)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8787) static void init_once(void *foo)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8788) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8789) struct btrfs_inode *ei = (struct btrfs_inode *) foo;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8791) inode_init_once(&ei->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8792) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8793)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8794) void __cold btrfs_destroy_cachep(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8795) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8796) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8797) * Make sure all delayed rcu free inodes are flushed before we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8798) * destroy cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8799) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8800) rcu_barrier();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8801) kmem_cache_destroy(btrfs_inode_cachep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8802) kmem_cache_destroy(btrfs_trans_handle_cachep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8803) kmem_cache_destroy(btrfs_path_cachep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8804) kmem_cache_destroy(btrfs_free_space_cachep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8805) kmem_cache_destroy(btrfs_free_space_bitmap_cachep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8806) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8807)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8808) int __init btrfs_init_cachep(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8809) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8810) btrfs_inode_cachep = kmem_cache_create("btrfs_inode",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8811) sizeof(struct btrfs_inode), 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8812) SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD | SLAB_ACCOUNT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8813) init_once);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8814) if (!btrfs_inode_cachep)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8815) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8816)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8817) btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8818) sizeof(struct btrfs_trans_handle), 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8819) SLAB_TEMPORARY | SLAB_MEM_SPREAD, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8820) if (!btrfs_trans_handle_cachep)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8821) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8822)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8823) btrfs_path_cachep = kmem_cache_create("btrfs_path",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8824) sizeof(struct btrfs_path), 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8825) SLAB_MEM_SPREAD, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8826) if (!btrfs_path_cachep)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8827) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8828)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8829) btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8830) sizeof(struct btrfs_free_space), 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8831) SLAB_MEM_SPREAD, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8832) if (!btrfs_free_space_cachep)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8833) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8834)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8835) btrfs_free_space_bitmap_cachep = kmem_cache_create("btrfs_free_space_bitmap",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8836) PAGE_SIZE, PAGE_SIZE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8837) SLAB_MEM_SPREAD, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8838) if (!btrfs_free_space_bitmap_cachep)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8839) goto fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8840)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8841) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8842) fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8843) btrfs_destroy_cachep();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8844) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8845) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8847) static int btrfs_getattr(const struct path *path, struct kstat *stat,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8848) u32 request_mask, unsigned int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8849) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8850) u64 delalloc_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8851) struct inode *inode = d_inode(path->dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8852) u32 blocksize = inode->i_sb->s_blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8853) u32 bi_flags = BTRFS_I(inode)->flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8854)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8855) stat->result_mask |= STATX_BTIME;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8856) stat->btime.tv_sec = BTRFS_I(inode)->i_otime.tv_sec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8857) stat->btime.tv_nsec = BTRFS_I(inode)->i_otime.tv_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8858) if (bi_flags & BTRFS_INODE_APPEND)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8859) stat->attributes |= STATX_ATTR_APPEND;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8860) if (bi_flags & BTRFS_INODE_COMPRESS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8861) stat->attributes |= STATX_ATTR_COMPRESSED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8862) if (bi_flags & BTRFS_INODE_IMMUTABLE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8863) stat->attributes |= STATX_ATTR_IMMUTABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8864) if (bi_flags & BTRFS_INODE_NODUMP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8865) stat->attributes |= STATX_ATTR_NODUMP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8866)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8867) stat->attributes_mask |= (STATX_ATTR_APPEND |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8868) STATX_ATTR_COMPRESSED |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8869) STATX_ATTR_IMMUTABLE |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8870) STATX_ATTR_NODUMP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8872) generic_fillattr(inode, stat);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8873) stat->dev = BTRFS_I(inode)->root->anon_dev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8875) spin_lock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8876) delalloc_bytes = BTRFS_I(inode)->new_delalloc_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8877) spin_unlock(&BTRFS_I(inode)->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8878) stat->blocks = (ALIGN(inode_get_bytes(inode), blocksize) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8879) ALIGN(delalloc_bytes, blocksize)) >> 9;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8880) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8882)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8883) static int btrfs_rename_exchange(struct inode *old_dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8884) struct dentry *old_dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8885) struct inode *new_dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8886) struct dentry *new_dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8887) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8888) struct btrfs_fs_info *fs_info = btrfs_sb(old_dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8889) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8890) struct btrfs_root *root = BTRFS_I(old_dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8891) struct btrfs_root *dest = BTRFS_I(new_dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8892) struct inode *new_inode = new_dentry->d_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8893) struct inode *old_inode = old_dentry->d_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8894) struct timespec64 ctime = current_time(old_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8895) u64 old_ino = btrfs_ino(BTRFS_I(old_inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8896) u64 new_ino = btrfs_ino(BTRFS_I(new_inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8897) u64 old_idx = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8898) u64 new_idx = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8899) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8900) int ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8901) bool root_log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8902) bool dest_log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8903) bool need_abort = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8904)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8905) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8906) * For non-subvolumes allow exchange only within one subvolume, in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8907) * same inode namespace. Two subvolumes (represented as directory) can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8908) * be exchanged as they're a logical link and have a fixed inode number.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8909) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8910) if (root != dest &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8911) (old_ino != BTRFS_FIRST_FREE_OBJECTID ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8912) new_ino != BTRFS_FIRST_FREE_OBJECTID))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8913) return -EXDEV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8914)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8915) /* close the race window with snapshot create/destroy ioctl */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8916) if (old_ino == BTRFS_FIRST_FREE_OBJECTID ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8917) new_ino == BTRFS_FIRST_FREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8918) down_read(&fs_info->subvol_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8919)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8920) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8921) * We want to reserve the absolute worst case amount of items. So if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8922) * both inodes are subvols and we need to unlink them then that would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8923) * require 4 item modifications, but if they are both normal inodes it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8924) * would require 5 item modifications, so we'll assume their normal
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8925) * inodes. So 5 * 2 is 10, plus 2 for the new links, so 12 total items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8926) * should cover the worst case number of items we'll modify.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8927) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8928) trans = btrfs_start_transaction(root, 12);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8929) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8930) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8931) goto out_notrans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8932) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8933)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8934) if (dest != root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8935) btrfs_record_root_in_trans(trans, dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8936)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8937) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8938) * We need to find a free sequence number both in the source and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8939) * in the destination directory for the exchange.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8940) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8941) ret = btrfs_set_inode_index(BTRFS_I(new_dir), &old_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8942) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8943) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8944) ret = btrfs_set_inode_index(BTRFS_I(old_dir), &new_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8945) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8946) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8948) BTRFS_I(old_inode)->dir_index = 0ULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8949) BTRFS_I(new_inode)->dir_index = 0ULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8950)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8951) /* Reference for the source. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8952) if (old_ino == BTRFS_FIRST_FREE_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8953) /* force full log commit if subvolume involved. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8954) btrfs_set_log_full_commit(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8955) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8956) btrfs_pin_log_trans(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8957) root_log_pinned = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8958) ret = btrfs_insert_inode_ref(trans, dest,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8959) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8960) new_dentry->d_name.len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8961) old_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8962) btrfs_ino(BTRFS_I(new_dir)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8963) old_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8964) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8965) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8966) need_abort = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8967) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8968)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8969) /* And now for the dest. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8970) if (new_ino == BTRFS_FIRST_FREE_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8971) /* force full log commit if subvolume involved. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8972) btrfs_set_log_full_commit(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8973) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8974) btrfs_pin_log_trans(dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8975) dest_log_pinned = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8976) ret = btrfs_insert_inode_ref(trans, root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8977) old_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8978) old_dentry->d_name.len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8979) new_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8980) btrfs_ino(BTRFS_I(old_dir)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8981) new_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8982) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8983) if (need_abort)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8984) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8985) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8986) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8987) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8989) /* Update inode version and ctime/mtime. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8990) inode_inc_iversion(old_dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8991) inode_inc_iversion(new_dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8992) inode_inc_iversion(old_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8993) inode_inc_iversion(new_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8994) old_dir->i_ctime = old_dir->i_mtime = ctime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8995) new_dir->i_ctime = new_dir->i_mtime = ctime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8996) old_inode->i_ctime = ctime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8997) new_inode->i_ctime = ctime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8998)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8999) if (old_dentry->d_parent != new_dentry->d_parent) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9000) btrfs_record_unlink_dir(trans, BTRFS_I(old_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9001) BTRFS_I(old_inode), 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9002) btrfs_record_unlink_dir(trans, BTRFS_I(new_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9003) BTRFS_I(new_inode), 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9004) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9005)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9006) /* src is a subvolume */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9007) if (old_ino == BTRFS_FIRST_FREE_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9008) ret = btrfs_unlink_subvol(trans, old_dir, old_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9009) } else { /* src is an inode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9010) ret = __btrfs_unlink_inode(trans, root, BTRFS_I(old_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9011) BTRFS_I(old_dentry->d_inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9012) old_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9013) old_dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9014) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9015) ret = btrfs_update_inode(trans, root, old_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9016) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9017) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9018) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9019) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9020) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9022) /* dest is a subvolume */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9023) if (new_ino == BTRFS_FIRST_FREE_OBJECTID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9024) ret = btrfs_unlink_subvol(trans, new_dir, new_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9025) } else { /* dest is an inode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9026) ret = __btrfs_unlink_inode(trans, dest, BTRFS_I(new_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9027) BTRFS_I(new_dentry->d_inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9028) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9029) new_dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9030) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9031) ret = btrfs_update_inode(trans, dest, new_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9032) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9033) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9034) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9035) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9036) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9038) ret = btrfs_add_link(trans, BTRFS_I(new_dir), BTRFS_I(old_inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9039) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9040) new_dentry->d_name.len, 0, old_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9041) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9042) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9043) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9044) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9045)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9046) ret = btrfs_add_link(trans, BTRFS_I(old_dir), BTRFS_I(new_inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9047) old_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9048) old_dentry->d_name.len, 0, new_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9049) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9050) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9051) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9052) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9054) if (old_inode->i_nlink == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9055) BTRFS_I(old_inode)->dir_index = old_idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9056) if (new_inode->i_nlink == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9057) BTRFS_I(new_inode)->dir_index = new_idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9058)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9059) if (root_log_pinned) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9060) btrfs_log_new_name(trans, BTRFS_I(old_inode), BTRFS_I(old_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9061) new_dentry->d_parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9062) btrfs_end_log_trans(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9063) root_log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9064) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9065) if (dest_log_pinned) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9066) btrfs_log_new_name(trans, BTRFS_I(new_inode), BTRFS_I(new_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9067) old_dentry->d_parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9068) btrfs_end_log_trans(dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9069) dest_log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9070) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9071) out_fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9072) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9073) * If we have pinned a log and an error happened, we unpin tasks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9074) * trying to sync the log and force them to fallback to a transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9075) * commit if the log currently contains any of the inodes involved in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9076) * this rename operation (to ensure we do not persist a log with an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9077) * inconsistent state for any of these inodes or leading to any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9078) * inconsistencies when replayed). If the transaction was aborted, the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9079) * abortion reason is propagated to userspace when attempting to commit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9080) * the transaction. If the log does not contain any of these inodes, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9081) * allow the tasks to sync it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9082) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9083) if (ret && (root_log_pinned || dest_log_pinned)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9084) if (btrfs_inode_in_log(BTRFS_I(old_dir), fs_info->generation) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9085) btrfs_inode_in_log(BTRFS_I(new_dir), fs_info->generation) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9086) btrfs_inode_in_log(BTRFS_I(old_inode), fs_info->generation) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9087) (new_inode &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9088) btrfs_inode_in_log(BTRFS_I(new_inode), fs_info->generation)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9089) btrfs_set_log_full_commit(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9090)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9091) if (root_log_pinned) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9092) btrfs_end_log_trans(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9093) root_log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9094) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9095) if (dest_log_pinned) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9096) btrfs_end_log_trans(dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9097) dest_log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9098) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9099) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9100) ret2 = btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9101) ret = ret ? ret : ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9102) out_notrans:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9103) if (new_ino == BTRFS_FIRST_FREE_OBJECTID ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9104) old_ino == BTRFS_FIRST_FREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9105) up_read(&fs_info->subvol_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9106)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9107) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9108) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9109)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9110) static int btrfs_whiteout_for_rename(struct btrfs_trans_handle *trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9111) struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9112) struct inode *dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9113) struct dentry *dentry)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9114) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9115) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9116) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9117) u64 objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9118) u64 index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9119)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9120) ret = btrfs_find_free_ino(root, &objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9121) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9122) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9123)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9124) inode = btrfs_new_inode(trans, root, dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9125) dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9126) dentry->d_name.len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9127) btrfs_ino(BTRFS_I(dir)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9128) objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9129) S_IFCHR | WHITEOUT_MODE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9130) &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9132) if (IS_ERR(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9133) ret = PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9134) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9135) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9136)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9137) inode->i_op = &btrfs_special_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9138) init_special_inode(inode, inode->i_mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9139) WHITEOUT_DEV);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9141) ret = btrfs_init_inode_security(trans, inode, dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9142) &dentry->d_name);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9143) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9144) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9145)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9146) ret = btrfs_add_nondir(trans, BTRFS_I(dir), dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9147) BTRFS_I(inode), 0, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9148) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9149) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9150)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9151) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9152) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9153) unlock_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9154) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9155) inode_dec_link_count(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9156) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9158) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9159) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9160)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9161) static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9162) struct inode *new_dir, struct dentry *new_dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9163) unsigned int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9164) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9165) struct btrfs_fs_info *fs_info = btrfs_sb(old_dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9166) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9167) unsigned int trans_num_items;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9168) struct btrfs_root *root = BTRFS_I(old_dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9169) struct btrfs_root *dest = BTRFS_I(new_dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9170) struct inode *new_inode = d_inode(new_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9171) struct inode *old_inode = d_inode(old_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9172) u64 index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9173) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9174) int ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9175) u64 old_ino = btrfs_ino(BTRFS_I(old_inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9176) bool log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9177)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9178) if (btrfs_ino(BTRFS_I(new_dir)) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9179) return -EPERM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9181) /* we only allow rename subvolume link between subvolumes */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9182) if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9183) return -EXDEV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9184)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9185) if (old_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9186) (new_inode && btrfs_ino(BTRFS_I(new_inode)) == BTRFS_FIRST_FREE_OBJECTID))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9187) return -ENOTEMPTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9189) if (S_ISDIR(old_inode->i_mode) && new_inode &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9190) new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9191) return -ENOTEMPTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9192)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9194) /* check for collisions, even if the name isn't there */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9195) ret = btrfs_check_dir_item_collision(dest, new_dir->i_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9196) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9197) new_dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9198)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9199) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9200) if (ret == -EEXIST) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9201) /* we shouldn't get
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9202) * eexist without a new_inode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9203) if (WARN_ON(!new_inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9204) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9205) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9206) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9207) /* maybe -EOVERFLOW */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9208) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9209) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9210) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9211) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9212)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9213) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9214) * we're using rename to replace one file with another. Start IO on it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9215) * now so we don't add too much work to the end of the transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9216) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9217) if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9218) filemap_flush(old_inode->i_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9219)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9220) /* close the racy window with snapshot create/destroy ioctl */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9221) if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9222) down_read(&fs_info->subvol_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9223) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9224) * We want to reserve the absolute worst case amount of items. So if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9225) * both inodes are subvols and we need to unlink them then that would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9226) * require 4 item modifications, but if they are both normal inodes it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9227) * would require 5 item modifications, so we'll assume they are normal
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9228) * inodes. So 5 * 2 is 10, plus 1 for the new link, so 11 total items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9229) * should cover the worst case number of items we'll modify.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9230) * If our rename has the whiteout flag, we need more 5 units for the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9231) * new inode (1 inode item, 1 inode ref, 2 dir items and 1 xattr item
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9232) * when selinux is enabled).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9233) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9234) trans_num_items = 11;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9235) if (flags & RENAME_WHITEOUT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9236) trans_num_items += 5;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9237) trans = btrfs_start_transaction(root, trans_num_items);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9238) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9239) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9240) goto out_notrans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9241) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9242)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9243) if (dest != root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9244) btrfs_record_root_in_trans(trans, dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9246) ret = btrfs_set_inode_index(BTRFS_I(new_dir), &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9247) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9248) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9249)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9250) BTRFS_I(old_inode)->dir_index = 0ULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9251) if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9252) /* force full log commit if subvolume involved. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9253) btrfs_set_log_full_commit(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9254) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9255) btrfs_pin_log_trans(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9256) log_pinned = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9257) ret = btrfs_insert_inode_ref(trans, dest,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9258) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9259) new_dentry->d_name.len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9260) old_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9261) btrfs_ino(BTRFS_I(new_dir)), index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9262) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9263) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9264) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9265)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9266) inode_inc_iversion(old_dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9267) inode_inc_iversion(new_dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9268) inode_inc_iversion(old_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9269) old_dir->i_ctime = old_dir->i_mtime =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9270) new_dir->i_ctime = new_dir->i_mtime =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9271) old_inode->i_ctime = current_time(old_dir);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9272)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9273) if (old_dentry->d_parent != new_dentry->d_parent)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9274) btrfs_record_unlink_dir(trans, BTRFS_I(old_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9275) BTRFS_I(old_inode), 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9276)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9277) if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9278) ret = btrfs_unlink_subvol(trans, old_dir, old_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9279) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9280) ret = __btrfs_unlink_inode(trans, root, BTRFS_I(old_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9281) BTRFS_I(d_inode(old_dentry)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9282) old_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9283) old_dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9284) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9285) ret = btrfs_update_inode(trans, root, old_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9286) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9287) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9288) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9289) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9290) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9292) if (new_inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9293) inode_inc_iversion(new_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9294) new_inode->i_ctime = current_time(new_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9295) if (unlikely(btrfs_ino(BTRFS_I(new_inode)) ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9296) BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9297) ret = btrfs_unlink_subvol(trans, new_dir, new_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9298) BUG_ON(new_inode->i_nlink == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9299) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9300) ret = btrfs_unlink_inode(trans, dest, BTRFS_I(new_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9301) BTRFS_I(d_inode(new_dentry)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9302) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9303) new_dentry->d_name.len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9304) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9305) if (!ret && new_inode->i_nlink == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9306) ret = btrfs_orphan_add(trans,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9307) BTRFS_I(d_inode(new_dentry)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9308) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9309) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9310) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9311) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9312) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9313)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9314) ret = btrfs_add_link(trans, BTRFS_I(new_dir), BTRFS_I(old_inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9315) new_dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9316) new_dentry->d_name.len, 0, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9317) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9318) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9319) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9320) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9322) if (old_inode->i_nlink == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9323) BTRFS_I(old_inode)->dir_index = index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9324)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9325) if (log_pinned) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9326) btrfs_log_new_name(trans, BTRFS_I(old_inode), BTRFS_I(old_dir),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9327) new_dentry->d_parent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9328) btrfs_end_log_trans(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9329) log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9330) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9331)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9332) if (flags & RENAME_WHITEOUT) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9333) ret = btrfs_whiteout_for_rename(trans, root, old_dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9334) old_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9336) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9337) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9338) goto out_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9339) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9340) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9341) out_fail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9342) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9343) * If we have pinned the log and an error happened, we unpin tasks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9344) * trying to sync the log and force them to fallback to a transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9345) * commit if the log currently contains any of the inodes involved in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9346) * this rename operation (to ensure we do not persist a log with an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9347) * inconsistent state for any of these inodes or leading to any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9348) * inconsistencies when replayed). If the transaction was aborted, the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9349) * abortion reason is propagated to userspace when attempting to commit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9350) * the transaction. If the log does not contain any of these inodes, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9351) * allow the tasks to sync it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9352) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9353) if (ret && log_pinned) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9354) if (btrfs_inode_in_log(BTRFS_I(old_dir), fs_info->generation) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9355) btrfs_inode_in_log(BTRFS_I(new_dir), fs_info->generation) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9356) btrfs_inode_in_log(BTRFS_I(old_inode), fs_info->generation) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9357) (new_inode &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9358) btrfs_inode_in_log(BTRFS_I(new_inode), fs_info->generation)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9359) btrfs_set_log_full_commit(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9360)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9361) btrfs_end_log_trans(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9362) log_pinned = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9363) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9364) ret2 = btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9365) ret = ret ? ret : ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9366) out_notrans:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9367) if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9368) up_read(&fs_info->subvol_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9369)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9370) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9371) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9373) static int btrfs_rename2(struct inode *old_dir, struct dentry *old_dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9374) struct inode *new_dir, struct dentry *new_dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9375) unsigned int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9376) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9377) if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9378) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9380) if (flags & RENAME_EXCHANGE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9381) return btrfs_rename_exchange(old_dir, old_dentry, new_dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9382) new_dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9384) return btrfs_rename(old_dir, old_dentry, new_dir, new_dentry, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9385) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9387) struct btrfs_delalloc_work {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9388) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9389) struct completion completion;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9390) struct list_head list;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9391) struct btrfs_work work;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9392) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9393)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9394) static void btrfs_run_delalloc_work(struct btrfs_work *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9395) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9396) struct btrfs_delalloc_work *delalloc_work;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9397) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9399) delalloc_work = container_of(work, struct btrfs_delalloc_work,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9400) work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9401) inode = delalloc_work->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9402) filemap_flush(inode->i_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9403) if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9404) &BTRFS_I(inode)->runtime_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9405) filemap_flush(inode->i_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9406)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9407) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9408) complete(&delalloc_work->completion);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9409) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9410)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9411) static struct btrfs_delalloc_work *btrfs_alloc_delalloc_work(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9412) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9413) struct btrfs_delalloc_work *work;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9414)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9415) work = kmalloc(sizeof(*work), GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9416) if (!work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9417) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9419) init_completion(&work->completion);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9420) INIT_LIST_HEAD(&work->list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9421) work->inode = inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9422) btrfs_init_work(&work->work, btrfs_run_delalloc_work, NULL, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9423)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9424) return work;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9425) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9427) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9428) * some fairly slow code that needs optimization. This walks the list
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9429) * of all the inodes with pending delalloc and forces them to disk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9430) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9431) static int start_delalloc_inodes(struct btrfs_root *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9432) struct writeback_control *wbc, bool snapshot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9433) bool in_reclaim_context)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9434) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9435) struct btrfs_inode *binode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9436) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9437) struct btrfs_delalloc_work *work, *next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9438) struct list_head works;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9439) struct list_head splice;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9440) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9441) bool full_flush = wbc->nr_to_write == LONG_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9442)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9443) INIT_LIST_HEAD(&works);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9444) INIT_LIST_HEAD(&splice);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9445)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9446) mutex_lock(&root->delalloc_mutex);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9447) spin_lock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9448) list_splice_init(&root->delalloc_inodes, &splice);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9449) while (!list_empty(&splice)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9450) binode = list_entry(splice.next, struct btrfs_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9451) delalloc_inodes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9452)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9453) list_move_tail(&binode->delalloc_inodes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9454) &root->delalloc_inodes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9455)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9456) if (in_reclaim_context &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9457) test_bit(BTRFS_INODE_NO_DELALLOC_FLUSH, &binode->runtime_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9458) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9460) inode = igrab(&binode->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9461) if (!inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9462) cond_resched_lock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9463) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9464) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9465) spin_unlock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9466)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9467) if (snapshot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9468) set_bit(BTRFS_INODE_SNAPSHOT_FLUSH,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9469) &binode->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9470) if (full_flush) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9471) work = btrfs_alloc_delalloc_work(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9472) if (!work) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9473) iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9474) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9475) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9476) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9477) list_add_tail(&work->list, &works);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9478) btrfs_queue_work(root->fs_info->flush_workers,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9479) &work->work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9480) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9481) ret = sync_inode(inode, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9482) if (!ret &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9483) test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9484) &BTRFS_I(inode)->runtime_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9485) ret = sync_inode(inode, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9486) btrfs_add_delayed_iput(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9487) if (ret || wbc->nr_to_write <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9488) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9489) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9490) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9491) spin_lock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9492) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9493) spin_unlock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9494)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9495) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9496) list_for_each_entry_safe(work, next, &works, list) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9497) list_del_init(&work->list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9498) wait_for_completion(&work->completion);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9499) kfree(work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9500) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9501)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9502) if (!list_empty(&splice)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9503) spin_lock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9504) list_splice_tail(&splice, &root->delalloc_inodes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9505) spin_unlock(&root->delalloc_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9506) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9507) mutex_unlock(&root->delalloc_mutex);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9508) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9509) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9510)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9511) int btrfs_start_delalloc_snapshot(struct btrfs_root *root)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9512) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9513) struct writeback_control wbc = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9514) .nr_to_write = LONG_MAX,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9515) .sync_mode = WB_SYNC_NONE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9516) .range_start = 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9517) .range_end = LLONG_MAX,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9518) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9519) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9521) if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9522) return -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9524) return start_delalloc_inodes(root, &wbc, true, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9525) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9527) int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, u64 nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9528) bool in_reclaim_context)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9529) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9530) struct writeback_control wbc = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9531) .nr_to_write = (nr == U64_MAX) ? LONG_MAX : (unsigned long)nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9532) .sync_mode = WB_SYNC_NONE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9533) .range_start = 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9534) .range_end = LLONG_MAX,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9535) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9536) struct btrfs_root *root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9537) struct list_head splice;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9538) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9540) if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9541) return -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9542)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9543) INIT_LIST_HEAD(&splice);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9544)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9545) mutex_lock(&fs_info->delalloc_root_mutex);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9546) spin_lock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9547) list_splice_init(&fs_info->delalloc_roots, &splice);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9548) while (!list_empty(&splice) && nr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9549) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9550) * Reset nr_to_write here so we know that we're doing a full
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9551) * flush.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9552) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9553) if (nr == U64_MAX)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9554) wbc.nr_to_write = LONG_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9555)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9556) root = list_first_entry(&splice, struct btrfs_root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9557) delalloc_root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9558) root = btrfs_grab_root(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9559) BUG_ON(!root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9560) list_move_tail(&root->delalloc_root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9561) &fs_info->delalloc_roots);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9562) spin_unlock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9564) ret = start_delalloc_inodes(root, &wbc, false, in_reclaim_context);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9565) btrfs_put_root(root);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9566) if (ret < 0 || wbc.nr_to_write <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9567) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9568) spin_lock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9569) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9570) spin_unlock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9571)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9572) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9573) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9574) if (!list_empty(&splice)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9575) spin_lock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9576) list_splice_tail(&splice, &fs_info->delalloc_roots);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9577) spin_unlock(&fs_info->delalloc_root_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9578) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9579) mutex_unlock(&fs_info->delalloc_root_mutex);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9580) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9581) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9582)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9583) static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9584) const char *symname)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9585) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9586) struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9587) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9588) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9589) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9590) struct btrfs_key key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9591) struct inode *inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9592) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9593) u64 objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9594) u64 index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9595) int name_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9596) int datasize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9597) unsigned long ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9598) struct btrfs_file_extent_item *ei;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9599) struct extent_buffer *leaf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9600)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9601) name_len = strlen(symname);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9602) if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(fs_info))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9603) return -ENAMETOOLONG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9605) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9606) * 2 items for inode item and ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9607) * 2 items for dir items
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9608) * 1 item for updating parent inode item
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9609) * 1 item for the inline extent item
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9610) * 1 item for xattr if selinux is on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9611) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9612) trans = btrfs_start_transaction(root, 7);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9613) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9614) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9615)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9616) err = btrfs_find_free_ino(root, &objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9617) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9618) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9619)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9620) inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9621) dentry->d_name.len, btrfs_ino(BTRFS_I(dir)),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9622) objectid, S_IFLNK|S_IRWXUGO, &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9623) if (IS_ERR(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9624) err = PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9625) inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9626) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9627) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9629) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9630) * If the active LSM wants to access the inode during
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9631) * d_instantiate it needs these. Smack checks to see
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9632) * if the filesystem supports xattrs by looking at the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9633) * ops vector.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9634) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9635) inode->i_fop = &btrfs_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9636) inode->i_op = &btrfs_file_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9637) inode->i_mapping->a_ops = &btrfs_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9638)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9639) err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9640) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9641) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9643) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9644) if (!path) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9645) err = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9646) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9647) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9648) key.objectid = btrfs_ino(BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9649) key.offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9650) key.type = BTRFS_EXTENT_DATA_KEY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9651) datasize = btrfs_file_extent_calc_inline_size(name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9652) err = btrfs_insert_empty_item(trans, root, path, &key,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9653) datasize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9654) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9655) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9656) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9657) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9658) leaf = path->nodes[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9659) ei = btrfs_item_ptr(leaf, path->slots[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9660) struct btrfs_file_extent_item);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9661) btrfs_set_file_extent_generation(leaf, ei, trans->transid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9662) btrfs_set_file_extent_type(leaf, ei,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9663) BTRFS_FILE_EXTENT_INLINE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9664) btrfs_set_file_extent_encryption(leaf, ei, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9665) btrfs_set_file_extent_compression(leaf, ei, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9666) btrfs_set_file_extent_other_encoding(leaf, ei, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9667) btrfs_set_file_extent_ram_bytes(leaf, ei, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9668)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9669) ptr = btrfs_file_extent_inline_start(ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9670) write_extent_buffer(leaf, symname, ptr, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9671) btrfs_mark_buffer_dirty(leaf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9672) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9673)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9674) inode->i_op = &btrfs_symlink_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9675) inode_nohighmem(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9676) inode_set_bytes(inode, name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9677) btrfs_i_size_write(BTRFS_I(inode), name_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9678) err = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9679) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9680) * Last step, add directory indexes for our symlink inode. This is the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9681) * last step to avoid extra cleanup of these indexes if an error happens
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9682) * elsewhere above.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9683) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9684) if (!err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9685) err = btrfs_add_nondir(trans, BTRFS_I(dir), dentry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9686) BTRFS_I(inode), 0, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9687) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9688) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9689)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9690) d_instantiate_new(dentry, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9691)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9692) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9693) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9694) if (err && inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9695) inode_dec_link_count(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9696) discard_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9697) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9698) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9699) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9700) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9702) static struct btrfs_trans_handle *insert_prealloc_file_extent(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9703) struct btrfs_trans_handle *trans_in,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9704) struct inode *inode, struct btrfs_key *ins,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9705) u64 file_offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9706) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9707) struct btrfs_file_extent_item stack_fi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9708) struct btrfs_replace_extent_info extent_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9709) struct btrfs_trans_handle *trans = trans_in;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9710) struct btrfs_path *path;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9711) u64 start = ins->objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9712) u64 len = ins->offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9713) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9714)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9715) memset(&stack_fi, 0, sizeof(stack_fi));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9716)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9717) btrfs_set_stack_file_extent_type(&stack_fi, BTRFS_FILE_EXTENT_PREALLOC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9718) btrfs_set_stack_file_extent_disk_bytenr(&stack_fi, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9719) btrfs_set_stack_file_extent_disk_num_bytes(&stack_fi, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9720) btrfs_set_stack_file_extent_num_bytes(&stack_fi, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9721) btrfs_set_stack_file_extent_ram_bytes(&stack_fi, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9722) btrfs_set_stack_file_extent_compression(&stack_fi, BTRFS_COMPRESS_NONE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9723) /* Encryption and other encoding is reserved and all 0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9724)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9725) ret = btrfs_qgroup_release_data(BTRFS_I(inode), file_offset, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9726) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9727) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9728)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9729) if (trans) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9730) ret = insert_reserved_file_extent(trans, BTRFS_I(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9731) file_offset, &stack_fi, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9732) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9733) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9734) return trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9735) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9737) extent_info.disk_offset = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9738) extent_info.disk_len = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9739) extent_info.data_offset = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9740) extent_info.data_len = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9741) extent_info.file_offset = file_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9742) extent_info.extent_buf = (char *)&stack_fi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9743) extent_info.is_new_extent = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9744) extent_info.qgroup_reserved = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9745) extent_info.insertions = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9746)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9747) path = btrfs_alloc_path();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9748) if (!path)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9749) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9750)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9751) ret = btrfs_replace_file_extents(inode, path, file_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9752) file_offset + len - 1, &extent_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9753) &trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9754) btrfs_free_path(path);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9755) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9756) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9757)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9758) return trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9759) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9760)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9761) static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9762) u64 start, u64 num_bytes, u64 min_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9763) loff_t actual_len, u64 *alloc_hint,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9764) struct btrfs_trans_handle *trans)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9765) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9766) struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9767) struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9768) struct extent_map *em;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9769) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9770) struct btrfs_key ins;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9771) u64 cur_offset = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9772) u64 clear_offset = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9773) u64 i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9774) u64 cur_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9775) u64 last_alloc = (u64)-1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9776) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9777) bool own_trans = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9778) u64 end = start + num_bytes - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9779)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9780) if (trans)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9781) own_trans = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9782) while (num_bytes > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9783) cur_bytes = min_t(u64, num_bytes, SZ_256M);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9784) cur_bytes = max(cur_bytes, min_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9785) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9786) * If we are severely fragmented we could end up with really
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9787) * small allocations, so if the allocator is returning small
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9788) * chunks lets make its job easier by only searching for those
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9789) * sized chunks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9790) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9791) cur_bytes = min(cur_bytes, last_alloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9792) ret = btrfs_reserve_extent(root, cur_bytes, cur_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9793) min_size, 0, *alloc_hint, &ins, 1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9794) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9795) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9796)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9797) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9798) * We've reserved this space, and thus converted it from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9799) * ->bytes_may_use to ->bytes_reserved. Any error that happens
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9800) * from here on out we will only need to clear our reservation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9801) * for the remaining unreserved area, so advance our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9802) * clear_offset by our extent size.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9803) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9804) clear_offset += ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9806) last_alloc = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9807) trans = insert_prealloc_file_extent(trans, inode, &ins, cur_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9808) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9809) * Now that we inserted the prealloc extent we can finally
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9810) * decrement the number of reservations in the block group.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9811) * If we did it before, we could race with relocation and have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9812) * relocation miss the reserved extent, making it fail later.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9813) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9814) btrfs_dec_block_group_reservations(fs_info, ins.objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9815) if (IS_ERR(trans)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9816) ret = PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9817) btrfs_free_reserved_extent(fs_info, ins.objectid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9818) ins.offset, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9819) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9820) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9821)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9822) btrfs_drop_extent_cache(BTRFS_I(inode), cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9823) cur_offset + ins.offset -1, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9825) em = alloc_extent_map();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9826) if (!em) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9827) set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9828) &BTRFS_I(inode)->runtime_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9829) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9830) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9831)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9832) em->start = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9833) em->orig_start = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9834) em->len = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9835) em->block_start = ins.objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9836) em->block_len = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9837) em->orig_block_len = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9838) em->ram_bytes = ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9839) set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9840) em->generation = trans->transid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9841)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9842) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9843) write_lock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9844) ret = add_extent_mapping(em_tree, em, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9845) write_unlock(&em_tree->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9846) if (ret != -EEXIST)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9847) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9848) btrfs_drop_extent_cache(BTRFS_I(inode), cur_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9849) cur_offset + ins.offset - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9850) 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9851) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9852) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9853) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9854) num_bytes -= ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9855) cur_offset += ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9856) *alloc_hint = ins.objectid + ins.offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9857)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9858) inode_inc_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9859) inode->i_ctime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9860) BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9861) if (!(mode & FALLOC_FL_KEEP_SIZE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9862) (actual_len > inode->i_size) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9863) (cur_offset > inode->i_size)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9864) if (cur_offset > actual_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9865) i_size = actual_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9866) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9867) i_size = cur_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9868) i_size_write(inode, i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9869) btrfs_inode_safe_disk_i_size_write(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9870) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9872) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9873)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9874) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9875) btrfs_abort_transaction(trans, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9876) if (own_trans)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9877) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9878) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9879) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9880)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9881) if (own_trans) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9882) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9883) trans = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9884) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9885) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9886) if (clear_offset < end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9887) btrfs_free_reserved_data_space(BTRFS_I(inode), NULL, clear_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9888) end - clear_offset + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9889) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9890) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9891)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9892) int btrfs_prealloc_file_range(struct inode *inode, int mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9893) u64 start, u64 num_bytes, u64 min_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9894) loff_t actual_len, u64 *alloc_hint)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9895) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9896) return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9897) min_size, actual_len, alloc_hint,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9898) NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9899) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9900)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9901) int btrfs_prealloc_file_range_trans(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9902) struct btrfs_trans_handle *trans, int mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9903) u64 start, u64 num_bytes, u64 min_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9904) loff_t actual_len, u64 *alloc_hint)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9905) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9906) return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9907) min_size, actual_len, alloc_hint, trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9908) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9909)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9910) static int btrfs_set_page_dirty(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9911) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9912) return __set_page_dirty_nobuffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9913) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9914)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9915) static int btrfs_permission(struct inode *inode, int mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9916) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9917) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9918) umode_t mode = inode->i_mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9919)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9920) if (mask & MAY_WRITE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9921) (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9922) if (btrfs_root_readonly(root))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9923) return -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9924) if (BTRFS_I(inode)->flags & BTRFS_INODE_READONLY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9925) return -EACCES;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9926) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9927) return generic_permission(inode, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9928) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9929)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9930) static int btrfs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9931) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9932) struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9933) struct btrfs_trans_handle *trans;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9934) struct btrfs_root *root = BTRFS_I(dir)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9935) struct inode *inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9936) u64 objectid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9937) u64 index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9938) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9939)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9940) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9941) * 5 units required for adding orphan entry
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9942) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9943) trans = btrfs_start_transaction(root, 5);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9944) if (IS_ERR(trans))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9945) return PTR_ERR(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9946)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9947) ret = btrfs_find_free_ino(root, &objectid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9948) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9949) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9950)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9951) inode = btrfs_new_inode(trans, root, dir, NULL, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9952) btrfs_ino(BTRFS_I(dir)), objectid, mode, &index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9953) if (IS_ERR(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9954) ret = PTR_ERR(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9955) inode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9956) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9957) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9958)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9959) inode->i_fop = &btrfs_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9960) inode->i_op = &btrfs_file_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9962) inode->i_mapping->a_ops = &btrfs_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9963)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9964) ret = btrfs_init_inode_security(trans, inode, dir, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9965) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9966) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9968) ret = btrfs_update_inode(trans, root, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9969) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9970) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9971) ret = btrfs_orphan_add(trans, BTRFS_I(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9972) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9973) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9974)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9975) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9976) * We set number of links to 0 in btrfs_new_inode(), and here we set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9977) * it to 1 because d_tmpfile() will issue a warning if the count is 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9978) * through:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9979) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9980) * d_tmpfile() -> inode_dec_link_count() -> drop_nlink()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9981) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9982) set_nlink(inode, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9983) d_tmpfile(dentry, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9984) unlock_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9985) mark_inode_dirty(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9986) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9987) btrfs_end_transaction(trans);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9988) if (ret && inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9989) discard_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9990) btrfs_btree_balance_dirty(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9991) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9992) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9994) void btrfs_set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9995) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9996) struct inode *inode = tree->private_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9997) unsigned long index = start >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9998) unsigned long end_index = end >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9999) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10000)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10001) while (index <= end_index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10002) page = find_get_page(inode->i_mapping, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10003) ASSERT(page); /* Pages should be in the extent_io_tree */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10004) set_page_writeback(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10005) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10006) index++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10007) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10008) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10009)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10010) #ifdef CONFIG_SWAP
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10011) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10012) * Add an entry indicating a block group or device which is pinned by a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10013) * swapfile. Returns 0 on success, 1 if there is already an entry for it, or a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10014) * negative errno on failure.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10015) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10016) static int btrfs_add_swapfile_pin(struct inode *inode, void *ptr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10017) bool is_block_group)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10018) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10019) struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10020) struct btrfs_swapfile_pin *sp, *entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10021) struct rb_node **p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10022) struct rb_node *parent = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10024) sp = kmalloc(sizeof(*sp), GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10025) if (!sp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10026) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10027) sp->ptr = ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10028) sp->inode = inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10029) sp->is_block_group = is_block_group;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10030) sp->bg_extent_count = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10031)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10032) spin_lock(&fs_info->swapfile_pins_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10033) p = &fs_info->swapfile_pins.rb_node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10034) while (*p) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10035) parent = *p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10036) entry = rb_entry(parent, struct btrfs_swapfile_pin, node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10037) if (sp->ptr < entry->ptr ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10038) (sp->ptr == entry->ptr && sp->inode < entry->inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10039) p = &(*p)->rb_left;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10040) } else if (sp->ptr > entry->ptr ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10041) (sp->ptr == entry->ptr && sp->inode > entry->inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10042) p = &(*p)->rb_right;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10043) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10044) if (is_block_group)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10045) entry->bg_extent_count++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10046) spin_unlock(&fs_info->swapfile_pins_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10047) kfree(sp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10048) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10049) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10050) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10051) rb_link_node(&sp->node, parent, p);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10052) rb_insert_color(&sp->node, &fs_info->swapfile_pins);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10053) spin_unlock(&fs_info->swapfile_pins_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10054) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10055) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10056)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10057) /* Free all of the entries pinned by this swapfile. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10058) static void btrfs_free_swapfile_pins(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10059) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10060) struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10061) struct btrfs_swapfile_pin *sp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10062) struct rb_node *node, *next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10064) spin_lock(&fs_info->swapfile_pins_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10065) node = rb_first(&fs_info->swapfile_pins);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10066) while (node) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10067) next = rb_next(node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10068) sp = rb_entry(node, struct btrfs_swapfile_pin, node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10069) if (sp->inode == inode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10070) rb_erase(&sp->node, &fs_info->swapfile_pins);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10071) if (sp->is_block_group) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10072) btrfs_dec_block_group_swap_extents(sp->ptr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10073) sp->bg_extent_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10074) btrfs_put_block_group(sp->ptr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10075) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10076) kfree(sp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10077) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10078) node = next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10079) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10080) spin_unlock(&fs_info->swapfile_pins_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10081) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10082)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10083) struct btrfs_swap_info {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10084) u64 start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10085) u64 block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10086) u64 block_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10087) u64 lowest_ppage;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10088) u64 highest_ppage;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10089) unsigned long nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10090) int nr_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10091) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10092)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10093) static int btrfs_add_swap_extent(struct swap_info_struct *sis,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10094) struct btrfs_swap_info *bsi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10095) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10096) unsigned long nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10097) unsigned long max_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10098) u64 first_ppage, first_ppage_reported, next_ppage;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10099) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10100)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10101) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10102) * Our swapfile may have had its size extended after the swap header was
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10103) * written. In that case activating the swapfile should not go beyond
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10104) * the max size set in the swap header.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10105) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10106) if (bsi->nr_pages >= sis->max)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10107) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10108)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10109) max_pages = sis->max - bsi->nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10110) first_ppage = ALIGN(bsi->block_start, PAGE_SIZE) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10111) next_ppage = ALIGN_DOWN(bsi->block_start + bsi->block_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10112) PAGE_SIZE) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10114) if (first_ppage >= next_ppage)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10115) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10116) nr_pages = next_ppage - first_ppage;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10117) nr_pages = min(nr_pages, max_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10119) first_ppage_reported = first_ppage;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10120) if (bsi->start == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10121) first_ppage_reported++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10122) if (bsi->lowest_ppage > first_ppage_reported)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10123) bsi->lowest_ppage = first_ppage_reported;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10124) if (bsi->highest_ppage < (next_ppage - 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10125) bsi->highest_ppage = next_ppage - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10126)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10127) ret = add_swap_extent(sis, bsi->nr_pages, nr_pages, first_ppage);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10128) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10129) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10130) bsi->nr_extents += ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10131) bsi->nr_pages += nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10132) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10133) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10134)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10135) static void btrfs_swap_deactivate(struct file *file)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10136) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10137) struct inode *inode = file_inode(file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10139) btrfs_free_swapfile_pins(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10140) atomic_dec(&BTRFS_I(inode)->root->nr_swapfiles);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10141) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10142)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10143) static int btrfs_swap_activate(struct swap_info_struct *sis, struct file *file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10144) sector_t *span)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10145) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10146) struct inode *inode = file_inode(file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10147) struct btrfs_root *root = BTRFS_I(inode)->root;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10148) struct btrfs_fs_info *fs_info = root->fs_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10149) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10150) struct extent_state *cached_state = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10151) struct extent_map *em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10152) struct btrfs_device *device = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10153) struct btrfs_swap_info bsi = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10154) .lowest_ppage = (sector_t)-1ULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10155) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10156) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10157) u64 isize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10158) u64 start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10160) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10161) * If the swap file was just created, make sure delalloc is done. If the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10162) * file changes again after this, the user is doing something stupid and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10163) * we don't really care.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10164) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10165) ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10166) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10167) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10168)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10169) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10170) * The inode is locked, so these flags won't change after we check them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10171) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10172) if (BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10173) btrfs_warn(fs_info, "swapfile must not be compressed");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10174) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10175) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10176) if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10177) btrfs_warn(fs_info, "swapfile must not be copy-on-write");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10178) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10179) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10180) if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10181) btrfs_warn(fs_info, "swapfile must not be checksummed");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10182) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10183) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10184)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10185) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10186) * Balance or device remove/replace/resize can move stuff around from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10187) * under us. The exclop protection makes sure they aren't running/won't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10188) * run concurrently while we are mapping the swap extents, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10189) * fs_info->swapfile_pins prevents them from running while the swap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10190) * file is active and moving the extents. Note that this also prevents
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10191) * a concurrent device add which isn't actually necessary, but it's not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10192) * really worth the trouble to allow it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10193) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10194) if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_SWAP_ACTIVATE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10195) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10196) "cannot activate swapfile while exclusive operation is running");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10197) return -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10198) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10200) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10201) * Prevent snapshot creation while we are activating the swap file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10202) * We do not want to race with snapshot creation. If snapshot creation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10203) * already started before we bumped nr_swapfiles from 0 to 1 and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10204) * completes before the first write into the swap file after it is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10205) * activated, than that write would fallback to COW.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10206) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10207) if (!btrfs_drew_try_write_lock(&root->snapshot_lock)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10208) btrfs_exclop_finish(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10209) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10210) "cannot activate swapfile because snapshot creation is in progress");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10211) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10212) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10213) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10214) * Snapshots can create extents which require COW even if NODATACOW is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10215) * set. We use this counter to prevent snapshots. We must increment it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10216) * before walking the extents because we don't want a concurrent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10217) * snapshot to run after we've already checked the extents.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10218) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10219) atomic_inc(&root->nr_swapfiles);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10220)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10221) isize = ALIGN_DOWN(inode->i_size, fs_info->sectorsize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10222)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10223) lock_extent_bits(io_tree, 0, isize - 1, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10224) start = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10225) while (start < isize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10226) u64 logical_block_start, physical_block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10227) struct btrfs_block_group *bg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10228) u64 len = isize - start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10229)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10230) em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10231) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10232) ret = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10233) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10234) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10235)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10236) if (em->block_start == EXTENT_MAP_HOLE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10237) btrfs_warn(fs_info, "swapfile must not have holes");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10238) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10239) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10240) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10241) if (em->block_start == EXTENT_MAP_INLINE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10242) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10243) * It's unlikely we'll ever actually find ourselves
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10244) * here, as a file small enough to fit inline won't be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10245) * big enough to store more than the swap header, but in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10246) * case something changes in the future, let's catch it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10247) * here rather than later.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10248) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10249) btrfs_warn(fs_info, "swapfile must not be inline");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10250) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10251) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10252) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10253) if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10254) btrfs_warn(fs_info, "swapfile must not be compressed");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10255) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10256) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10257) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10259) logical_block_start = em->block_start + (start - em->start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10260) len = min(len, em->len - (start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10261) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10262) em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10264) ret = can_nocow_extent(inode, start, &len, NULL, NULL, NULL, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10265) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10266) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10267) } else if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10268) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10269) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10270) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10271) "swapfile must not be copy-on-write");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10272) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10273) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10274) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10275)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10276) em = btrfs_get_chunk_map(fs_info, logical_block_start, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10277) if (IS_ERR(em)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10278) ret = PTR_ERR(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10279) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10280) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10282) if (em->map_lookup->type & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10283) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10284) "swapfile must have single data profile");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10285) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10286) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10287) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10288)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10289) if (device == NULL) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10290) device = em->map_lookup->stripes[0].dev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10291) ret = btrfs_add_swapfile_pin(inode, device, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10292) if (ret == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10293) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10294) else if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10295) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10296) } else if (device != em->map_lookup->stripes[0].dev) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10297) btrfs_warn(fs_info, "swapfile must be on one device");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10298) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10299) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10300) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10302) physical_block_start = (em->map_lookup->stripes[0].physical +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10303) (logical_block_start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10304) len = min(len, em->len - (logical_block_start - em->start));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10305) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10306) em = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10307)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10308) bg = btrfs_lookup_block_group(fs_info, logical_block_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10309) if (!bg) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10310) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10311) "could not find block group containing swapfile");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10312) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10313) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10314) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10315)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10316) if (!btrfs_inc_block_group_swap_extents(bg)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10317) btrfs_warn(fs_info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10318) "block group for swapfile at %llu is read-only%s",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10319) bg->start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10320) atomic_read(&fs_info->scrubs_running) ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10321) " (scrub running)" : "");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10322) btrfs_put_block_group(bg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10323) ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10324) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10325) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10326)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10327) ret = btrfs_add_swapfile_pin(inode, bg, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10328) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10329) btrfs_put_block_group(bg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10330) if (ret == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10331) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10332) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10333) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10334) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10336) if (bsi.block_len &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10337) bsi.block_start + bsi.block_len == physical_block_start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10338) bsi.block_len += len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10339) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10340) if (bsi.block_len) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10341) ret = btrfs_add_swap_extent(sis, &bsi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10342) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10343) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10344) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10345) bsi.start = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10346) bsi.block_start = physical_block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10347) bsi.block_len = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10348) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10350) start += len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10351) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10352)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10353) if (bsi.block_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10354) ret = btrfs_add_swap_extent(sis, &bsi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10356) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10357) if (!IS_ERR_OR_NULL(em))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10358) free_extent_map(em);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10359)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10360) unlock_extent_cached(io_tree, 0, isize - 1, &cached_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10361)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10362) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10363) btrfs_swap_deactivate(file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10365) btrfs_drew_write_unlock(&root->snapshot_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10366)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10367) btrfs_exclop_finish(fs_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10368)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10369) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10370) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10371)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10372) if (device)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10373) sis->bdev = device->bdev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10374) *span = bsi.highest_ppage - bsi.lowest_ppage + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10375) sis->max = bsi.nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10376) sis->pages = bsi.nr_pages - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10377) sis->highest_bit = bsi.nr_pages - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10378) return bsi.nr_extents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10379) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10380) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10381) static void btrfs_swap_deactivate(struct file *file)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10382) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10383) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10384)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10385) static int btrfs_swap_activate(struct swap_info_struct *sis, struct file *file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10386) sector_t *span)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10387) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10388) return -EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10389) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10390) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10392) static const struct inode_operations btrfs_dir_inode_operations = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10393) .getattr = btrfs_getattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10394) .lookup = btrfs_lookup,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10395) .create = btrfs_create,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10396) .unlink = btrfs_unlink,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10397) .link = btrfs_link,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10398) .mkdir = btrfs_mkdir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10399) .rmdir = btrfs_rmdir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10400) .rename = btrfs_rename2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10401) .symlink = btrfs_symlink,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10402) .setattr = btrfs_setattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10403) .mknod = btrfs_mknod,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10404) .listxattr = btrfs_listxattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10405) .permission = btrfs_permission,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10406) .get_acl = btrfs_get_acl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10407) .set_acl = btrfs_set_acl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10408) .update_time = btrfs_update_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10409) .tmpfile = btrfs_tmpfile,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10410) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10411)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10412) static const struct file_operations btrfs_dir_file_operations = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10413) .llseek = generic_file_llseek,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10414) .read = generic_read_dir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10415) .iterate_shared = btrfs_real_readdir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10416) .open = btrfs_opendir,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10417) .unlocked_ioctl = btrfs_ioctl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10418) #ifdef CONFIG_COMPAT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10419) .compat_ioctl = btrfs_compat_ioctl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10420) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10421) .release = btrfs_release_file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10422) .fsync = btrfs_sync_file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10423) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10424)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10425) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10426) * btrfs doesn't support the bmap operation because swapfiles
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10427) * use bmap to make a mapping of extents in the file. They assume
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10428) * these extents won't change over the life of the file and they
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10429) * use the bmap result to do IO directly to the drive.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10430) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10431) * the btrfs bmap call would return logical addresses that aren't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10432) * suitable for IO and they also will change frequently as COW
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10433) * operations happen. So, swapfile + btrfs == corruption.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10434) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10435) * For now we're avoiding this by dropping bmap.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10436) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10437) static const struct address_space_operations btrfs_aops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10438) .readpage = btrfs_readpage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10439) .writepage = btrfs_writepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10440) .writepages = btrfs_writepages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10441) .readahead = btrfs_readahead,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10442) .direct_IO = noop_direct_IO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10443) .invalidatepage = btrfs_invalidatepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10444) .releasepage = btrfs_releasepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10445) #ifdef CONFIG_MIGRATION
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10446) .migratepage = btrfs_migratepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10447) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10448) .set_page_dirty = btrfs_set_page_dirty,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10449) .error_remove_page = generic_error_remove_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10450) .swap_activate = btrfs_swap_activate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10451) .swap_deactivate = btrfs_swap_deactivate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10452) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10454) static const struct inode_operations btrfs_file_inode_operations = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10455) .getattr = btrfs_getattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10456) .setattr = btrfs_setattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10457) .listxattr = btrfs_listxattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10458) .permission = btrfs_permission,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10459) .fiemap = btrfs_fiemap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10460) .get_acl = btrfs_get_acl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10461) .set_acl = btrfs_set_acl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10462) .update_time = btrfs_update_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10463) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10464) static const struct inode_operations btrfs_special_inode_operations = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10465) .getattr = btrfs_getattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10466) .setattr = btrfs_setattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10467) .permission = btrfs_permission,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10468) .listxattr = btrfs_listxattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10469) .get_acl = btrfs_get_acl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10470) .set_acl = btrfs_set_acl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10471) .update_time = btrfs_update_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10472) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10473) static const struct inode_operations btrfs_symlink_inode_operations = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10474) .get_link = page_get_link,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10475) .getattr = btrfs_getattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10476) .setattr = btrfs_setattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10477) .permission = btrfs_permission,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10478) .listxattr = btrfs_listxattr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10479) .update_time = btrfs_update_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10480) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10481)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10482) const struct dentry_operations btrfs_dentry_operations = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10483) .d_delete = btrfs_dentry_delete,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10484) };
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags