^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) * linux/fs/ext4/inode.c
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) * Copyright (C) 1992, 1993, 1994, 1995
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) * Remy Card (card@masi.ibp.fr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) * Laboratoire MASI - Institut Blaise Pascal
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) * Universite Pierre et Marie Curie (Paris VI)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) * from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) * linux/fs/minix/inode.c
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) * Copyright (C) 1991, 1992 Linus Torvalds
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) * 64-bit file support on 64-bit platforms by Jakub Jelinek
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) * (jj@sunsite.ms.mff.cuni.cz)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) * Assorted race fixes, rewrite of ext4_get_block() by Al Viro, 2000
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) #include <linux/fs.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) #include <linux/time.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) #include <linux/highuid.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) #include <linux/pagemap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) #include <linux/dax.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) #include <linux/quotaops.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) #include <linux/string.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) #include <linux/buffer_head.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) #include <linux/writeback.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) #include <linux/pagevec.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) #include <linux/mpage.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) #include <linux/namei.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) #include <linux/uio.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) #include <linux/bio.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) #include <linux/workqueue.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) #include <linux/kernel.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) #include <linux/printk.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) #include <linux/slab.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) #include <linux/bitops.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) #include <linux/iomap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) #include <linux/iversion.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) #include "ext4_jbd2.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) #include "xattr.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) #include "acl.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) #include "truncate.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) #include <trace/events/ext4.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) #include <trace/events/android_fs.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) static __u32 ext4_inode_csum(struct inode *inode, struct ext4_inode *raw,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) __u32 csum;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) __u16 dummy_csum = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) int offset = offsetof(struct ext4_inode, i_checksum_lo);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) unsigned int csum_size = sizeof(dummy_csum);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)raw, offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) csum = ext4_chksum(sbi, csum, (__u8 *)&dummy_csum, csum_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) offset += csum_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) csum = ext4_chksum(sbi, csum, (__u8 *)raw + offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) EXT4_GOOD_OLD_INODE_SIZE - offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) offset = offsetof(struct ext4_inode, i_checksum_hi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) csum = ext4_chksum(sbi, csum, (__u8 *)raw +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) EXT4_GOOD_OLD_INODE_SIZE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) offset - EXT4_GOOD_OLD_INODE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) if (EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) csum = ext4_chksum(sbi, csum, (__u8 *)&dummy_csum,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) csum_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) offset += csum_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) csum = ext4_chksum(sbi, csum, (__u8 *)raw + offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) EXT4_INODE_SIZE(inode->i_sb) - offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) return csum;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) static int ext4_inode_csum_verify(struct inode *inode, struct ext4_inode *raw,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) __u32 provided, calculated;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) cpu_to_le32(EXT4_OS_LINUX) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) !ext4_has_metadata_csum(inode->i_sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) provided = le16_to_cpu(raw->i_checksum_lo);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) calculated = ext4_inode_csum(inode, raw, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97) EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) provided |= ((__u32)le16_to_cpu(raw->i_checksum_hi)) << 16;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) calculated &= 0xFFFF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) return provided == calculated;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) __u32 csum;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) cpu_to_le32(EXT4_OS_LINUX) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) !ext4_has_metadata_csum(inode->i_sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) csum = ext4_inode_csum(inode, raw, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) raw->i_checksum_lo = cpu_to_le16(csum & 0xFFFF);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) raw->i_checksum_hi = cpu_to_le16(csum >> 16);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) static inline int ext4_begin_ordered_truncate(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) loff_t new_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) trace_ext4_begin_ordered_truncate(inode, new_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) * If jinode is zero, then we never opened the file for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) * writing, so there's no need to call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) * jbd2_journal_begin_ordered_truncate() since there's no
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) * outstanding writes we need to flush.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) if (!EXT4_I(inode)->jinode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) return jbd2_journal_begin_ordered_truncate(EXT4_JOURNAL(inode),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) EXT4_I(inode)->jinode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) new_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) static void ext4_invalidatepage(struct page *page, unsigned int offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) unsigned int length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) static int __ext4_journalled_writepage(struct page *page, unsigned int len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) static int ext4_meta_trans_blocks(struct inode *inode, int lblocks,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) int pextents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) * Test whether an inode is a fast symlink.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) * A fast symlink has its symlink data stored in ext4_inode_info->i_data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) int ext4_inode_is_fast_symlink(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) if (!(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) int ea_blocks = EXT4_I(inode)->i_file_acl ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) EXT4_CLUSTER_SIZE(inode->i_sb) >> 9 : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) return (S_ISLNK(inode->i_mode) && inode->i_blocks - ea_blocks == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) return S_ISLNK(inode->i_mode) && inode->i_size &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) (inode->i_size < EXT4_N_BLOCKS * 4);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) * Called at the last iput() if i_nlink is zero.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) void ext4_evict_inode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) * Credits for final inode cleanup and freeing:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) * sb + inode (ext4_orphan_del()), block bitmap, group descriptor
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175) * (xattr block freeing), bitmap, group descriptor (inode freeing)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) int extra_credits = 6;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178) struct ext4_xattr_inode_array *ea_inode_array = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) bool freeze_protected = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181) trace_ext4_evict_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) if (inode->i_nlink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) * When journalling data dirty buffers are tracked only in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) * journal. So although mm thinks everything is clean and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) * ready for reaping the inode might still have some pages to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) * write in the running transaction or waiting to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) * checkpointed. Thus calling jbd2_journal_invalidatepage()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) * (via truncate_inode_pages()) to discard these buffers can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) * cause data loss. Also even if we did not discard these
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) * buffers, we would have no way to find them after the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) * is reaped and thus user could see stale data if he tries to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194) * read them before the transaction is checkpointed. So be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) * careful and force everything to disk here... We use
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) * ei->i_datasync_tid to store the newest transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) * containing inode's data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) * Note that directories do not have this problem because they
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) * don't use page cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) if (inode->i_ino != EXT4_JOURNAL_INO &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) ext4_should_journal_data(inode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode)) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) inode->i_data.nrpages) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) tid_t commit_tid = EXT4_I(inode)->i_datasync_tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) jbd2_complete_transaction(journal, commit_tid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) filemap_write_and_wait(&inode->i_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) truncate_inode_pages_final(&inode->i_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217) if (is_bad_inode(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219) dquot_initialize(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) if (ext4_should_order_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) ext4_begin_ordered_truncate(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223) truncate_inode_pages_final(&inode->i_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) * For inodes with journalled data, transaction commit could have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) * dirtied the inode. Flush worker is ignoring it because of I_FREEING
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) * flag but we still need to remove the inode from the writeback lists.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) if (!list_empty_careful(&inode->i_io_list)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231) WARN_ON_ONCE(!ext4_should_journal_data(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232) inode_io_list_del(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236) * Protect us against freezing - iput() caller didn't have to have any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) * protection against it. When we are in a running transaction though,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238) * we are already protected against freezing and we cannot grab further
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239) * protection due to lock ordering constraints.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) if (!ext4_journal_current_handle()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) sb_start_intwrite(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) freeze_protected = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) if (!IS_NOQUOTA(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247) extra_credits += EXT4_MAXQUOTAS_DEL_BLOCKS(inode->i_sb);
^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) * Block bitmap, group descriptor, and inode are accounted in both
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251) * ext4_blocks_for_truncate() and extra_credits. So subtract 3.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253) handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) ext4_blocks_for_truncate(inode) + extra_credits - 3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) ext4_std_error(inode->i_sb, PTR_ERR(handle));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258) * If we're going to skip the normal cleanup, we still need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) * make sure that the in-core orphan linked list is properly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260) * cleaned up.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) ext4_orphan_del(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) if (freeze_protected)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264) sb_end_intwrite(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) if (IS_SYNC(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) ext4_handle_sync(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) * Set inode->i_size to 0 before calling ext4_truncate(). We need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 273) * special handling of symlinks here because i_size is used to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 274) * determine whether ext4_inode_info->i_data contains symlink data or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 275) * block mappings. Setting i_size to 0 will remove its fast symlink
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 276) * status. Erase i_data so that it becomes a valid empty block map.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 277) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 278) if (ext4_inode_is_fast_symlink(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 279) memset(EXT4_I(inode)->i_data, 0, sizeof(EXT4_I(inode)->i_data));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 280) inode->i_size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 281) err = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 282) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 283) ext4_warning(inode->i_sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 284) "couldn't mark inode dirty (err %d)", err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 285) goto stop_handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 286) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 287) if (inode->i_blocks) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 288) err = ext4_truncate(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 289) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 290) ext4_error_err(inode->i_sb, -err,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 291) "couldn't truncate inode %lu (err %d)",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 292) inode->i_ino, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 293) goto stop_handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 294) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 295) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 297) /* Remove xattr references. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 298) err = ext4_xattr_delete_inode(handle, inode, &ea_inode_array,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 299) extra_credits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 300) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 301) ext4_warning(inode->i_sb, "xattr delete (err %d)", err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 302) stop_handle:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 303) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 304) ext4_orphan_del(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 305) if (freeze_protected)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 306) sb_end_intwrite(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 307) ext4_xattr_inode_array_free(ea_inode_array);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 308) goto no_delete;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 309) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 310)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 311) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 312) * Kill off the orphan record which ext4_truncate created.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 313) * AKPM: I think this can be inside the above `if'.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 314) * Note that ext4_orphan_del() has to be able to cope with the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 315) * deletion of a non-existent orphan - this is because we don't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 316) * know if ext4_truncate() actually created an orphan record.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 317) * (Well, we could do this if we need to, but heck - it works)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 318) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 319) ext4_orphan_del(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 320) EXT4_I(inode)->i_dtime = (__u32)ktime_get_real_seconds();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 322) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 323) * One subtle ordering requirement: if anything has gone wrong
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 324) * (transaction abort, IO errors, whatever), then we can still
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 325) * do these next steps (the fs will already have been marked as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 326) * having errors), but we can't free the inode if the mark_dirty
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 327) * fails.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 328) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 329) if (ext4_mark_inode_dirty(handle, inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 330) /* If that failed, just do the required in-core inode clear. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 331) ext4_clear_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 332) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 333) ext4_free_inode(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 334) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 335) if (freeze_protected)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 336) sb_end_intwrite(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 337) ext4_xattr_inode_array_free(ea_inode_array);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 338) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 339) no_delete:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 340) if (!list_empty(&EXT4_I(inode)->i_fc_list))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 341) ext4_fc_mark_ineligible(inode->i_sb, EXT4_FC_REASON_NOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 342) ext4_clear_inode(inode); /* We must guarantee clearing of inode... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 343) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 344)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 345) #ifdef CONFIG_QUOTA
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 346) qsize_t *ext4_get_reserved_space(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 347) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 348) return &EXT4_I(inode)->i_reserved_quota;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 349) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 350) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 351)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 352) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 353) * Called with i_data_sem down, which is important since we can call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 354) * ext4_discard_preallocations() from here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 355) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 356) void ext4_da_update_reserve_space(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 357) int used, int quota_claim)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 358) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 359) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 360) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 361)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 362) spin_lock(&ei->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 363) trace_ext4_da_update_reserve_space(inode, used, quota_claim);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 364) if (unlikely(used > ei->i_reserved_data_blocks)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 365) ext4_warning(inode->i_sb, "%s: ino %lu, used %d "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 366) "with only %d reserved data blocks",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 367) __func__, inode->i_ino, used,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 368) ei->i_reserved_data_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 369) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 370) used = ei->i_reserved_data_blocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 371) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 373) /* Update per-inode reservations */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 374) ei->i_reserved_data_blocks -= used;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 375) percpu_counter_sub(&sbi->s_dirtyclusters_counter, used);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 376)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 377) spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 379) /* Update quota subsystem for data blocks */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 380) if (quota_claim)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 381) dquot_claim_block(inode, EXT4_C2B(sbi, used));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 382) else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 383) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 384) * We did fallocate with an offset that is already delayed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 385) * allocated. So on delayed allocated writeback we should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 386) * not re-claim the quota for fallocated blocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 387) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 388) dquot_release_reservation_block(inode, EXT4_C2B(sbi, used));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 389) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 391) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 392) * If we have done all the pending block allocations and if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 393) * there aren't any writers on the inode, we can discard the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 394) * inode's preallocations.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 395) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 396) if ((ei->i_reserved_data_blocks == 0) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 397) !inode_is_open_for_write(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 398) ext4_discard_preallocations(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 399) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 400)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 401) static int __check_block_validity(struct inode *inode, const char *func,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 402) unsigned int line,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 403) struct ext4_map_blocks *map)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 404) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 405) if (ext4_has_feature_journal(inode->i_sb) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 406) (inode->i_ino ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 407) le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_journal_inum)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 408) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 409) if (!ext4_inode_block_valid(inode, map->m_pblk, map->m_len)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 410) ext4_error_inode(inode, func, line, map->m_pblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 411) "lblock %lu mapped to illegal pblock %llu "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 412) "(length %d)", (unsigned long) map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 413) map->m_pblk, map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 414) return -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 415) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 416) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 417) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 419) int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk, ext4_fsblk_t pblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 420) ext4_lblk_t len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 421) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 422) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 423)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 424) if (IS_ENCRYPTED(inode) && S_ISREG(inode->i_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 425) return fscrypt_zeroout_range(inode, lblk, pblk, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 427) ret = sb_issue_zeroout(inode->i_sb, pblk, len, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 428) if (ret > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 429) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 430)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 431) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 432) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 433)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 434) #define check_block_validity(inode, map) \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 435) __check_block_validity((inode), __func__, __LINE__, (map))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 436)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 437) #ifdef ES_AGGRESSIVE_TEST
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 438) static void ext4_map_blocks_es_recheck(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 439) struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 440) struct ext4_map_blocks *es_map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 441) struct ext4_map_blocks *map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 442) int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 443) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 444) int retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 445)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 446) map->m_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 447) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 448) * There is a race window that the result is not the same.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 449) * e.g. xfstests #223 when dioread_nolock enables. The reason
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 450) * is that we lookup a block mapping in extent status tree with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 451) * out taking i_data_sem. So at the time the unwritten extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 452) * could be converted.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 453) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 454) down_read(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 455) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 456) retval = ext4_ext_map_blocks(handle, inode, map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 457) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 458) retval = ext4_ind_map_blocks(handle, inode, map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 459) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 460) up_read((&EXT4_I(inode)->i_data_sem));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 461)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 462) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 463) * We don't check m_len because extent will be collpased in status
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 464) * tree. So the m_len might not equal.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 465) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 466) if (es_map->m_lblk != map->m_lblk ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 467) es_map->m_flags != map->m_flags ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 468) es_map->m_pblk != map->m_pblk) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 469) printk("ES cache assertion failed for inode: %lu "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 470) "es_cached ex [%d/%d/%llu/%x] != "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 471) "found ex [%d/%d/%llu/%x] retval %d flags %x\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 472) inode->i_ino, es_map->m_lblk, es_map->m_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 473) es_map->m_pblk, es_map->m_flags, map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 474) map->m_len, map->m_pblk, map->m_flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 475) retval, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 476) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 477) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 478) #endif /* ES_AGGRESSIVE_TEST */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 479)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 480) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 481) * The ext4_map_blocks() function tries to look up the requested blocks,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 482) * and returns if the blocks are already mapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 483) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 484) * Otherwise it takes the write lock of the i_data_sem and allocate blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 485) * and store the allocated blocks in the result buffer head and mark it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 486) * mapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 487) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 488) * If file type is extents based, it will call ext4_ext_map_blocks(),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 489) * Otherwise, call with ext4_ind_map_blocks() to handle indirect mapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 490) * based files
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 491) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 492) * On success, it returns the number of blocks being mapped or allocated. if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 493) * create==0 and the blocks are pre-allocated and unwritten, the resulting @map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 494) * is marked as unwritten. If the create == 1, it will mark @map as mapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 495) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 496) * It returns 0 if plain look up failed (blocks have not been allocated), in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 497) * that case, @map is returned as unmapped but we still do fill map->m_len to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 498) * indicate the length of a hole starting at map->m_lblk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 499) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 500) * It returns the error in case of allocation failure.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 501) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 502) int ext4_map_blocks(handle_t *handle, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 503) struct ext4_map_blocks *map, int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 504) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 505) struct extent_status es;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 506) int retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 507) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 508) #ifdef ES_AGGRESSIVE_TEST
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 509) struct ext4_map_blocks orig_map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 510)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 511) memcpy(&orig_map, map, sizeof(*map));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 512) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 513)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 514) map->m_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 515) ext_debug(inode, "flag 0x%x, max_blocks %u, logical block %lu\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 516) flags, map->m_len, (unsigned long) map->m_lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 517)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 518) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 519) * ext4_map_blocks returns an int, and m_len is an unsigned int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 520) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 521) if (unlikely(map->m_len > INT_MAX))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 522) map->m_len = INT_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 524) /* We can handle the block number less than EXT_MAX_BLOCKS */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 525) if (unlikely(map->m_lblk >= EXT_MAX_BLOCKS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 526) return -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 527)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 528) /* Lookup extent status tree firstly */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 529) if (!(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 530) ext4_es_lookup_extent(inode, map->m_lblk, NULL, &es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 531) if (ext4_es_is_written(&es) || ext4_es_is_unwritten(&es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 532) map->m_pblk = ext4_es_pblock(&es) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 533) map->m_lblk - es.es_lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 534) map->m_flags |= ext4_es_is_written(&es) ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 535) EXT4_MAP_MAPPED : EXT4_MAP_UNWRITTEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 536) retval = es.es_len - (map->m_lblk - es.es_lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 537) if (retval > map->m_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 538) retval = map->m_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 539) map->m_len = retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 540) } else if (ext4_es_is_delayed(&es) || ext4_es_is_hole(&es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 541) map->m_pblk = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 542) retval = es.es_len - (map->m_lblk - es.es_lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 543) if (retval > map->m_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 544) retval = map->m_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 545) map->m_len = retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 546) retval = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 547) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 548) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 549) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 550) #ifdef ES_AGGRESSIVE_TEST
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 551) ext4_map_blocks_es_recheck(handle, inode, map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 552) &orig_map, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 553) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 554) goto found;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 555) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 556)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 557) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 558) * Try to see if we can get the block without requesting a new
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 559) * file system block.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 560) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 561) down_read(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 562) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 563) retval = ext4_ext_map_blocks(handle, inode, map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 564) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 565) retval = ext4_ind_map_blocks(handle, inode, map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 567) if (retval > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 568) unsigned int status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 569)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 570) if (unlikely(retval != map->m_len)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 571) ext4_warning(inode->i_sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 572) "ES len assertion failed for inode "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 573) "%lu: retval %d != map->m_len %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 574) inode->i_ino, retval, map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 575) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 576) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 577)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 578) status = map->m_flags & EXT4_MAP_UNWRITTEN ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 579) EXTENT_STATUS_UNWRITTEN : EXTENT_STATUS_WRITTEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 580) if (!(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 581) !(status & EXTENT_STATUS_WRITTEN) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 582) ext4_es_scan_range(inode, &ext4_es_is_delayed, map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 583) map->m_lblk + map->m_len - 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 584) status |= EXTENT_STATUS_DELAYED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 585) ret = ext4_es_insert_extent(inode, map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 586) map->m_len, map->m_pblk, status);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 587) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 588) retval = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 589) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 590) up_read((&EXT4_I(inode)->i_data_sem));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 591)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 592) found:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 593) if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 594) ret = check_block_validity(inode, map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 595) if (ret != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 596) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 597) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 598)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 599) /* If it is only a block(s) look up */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 600) if ((flags & EXT4_GET_BLOCKS_CREATE) == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 601) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 602)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 603) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 604) * Returns if the blocks have already allocated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 605) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 606) * Note that if blocks have been preallocated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 607) * ext4_ext_get_block() returns the create = 0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 608) * with buffer head unmapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 609) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 610) if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 611) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 612) * If we need to convert extent to unwritten
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 613) * we continue and do the actual work in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 614) * ext4_ext_map_blocks()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 615) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 616) if (!(flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 617) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 618)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 619) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 620) * Here we clear m_flags because after allocating an new extent,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 621) * it will be set again.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 622) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 623) map->m_flags &= ~EXT4_MAP_FLAGS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 625) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 626) * New blocks allocate and/or writing to unwritten extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 627) * will possibly result in updating i_data, so we take
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 628) * the write lock of i_data_sem, and call get_block()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 629) * with create == 1 flag.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 630) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 631) down_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 632)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 633) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 634) * We need to check for EXT4 here because migrate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 635) * could have changed the inode type in between
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 636) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 637) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 638) retval = ext4_ext_map_blocks(handle, inode, map, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 639) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 640) retval = ext4_ind_map_blocks(handle, inode, map, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 641)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 642) if (retval > 0 && map->m_flags & EXT4_MAP_NEW) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 643) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 644) * We allocated new blocks which will result in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 645) * i_data's format changing. Force the migrate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 646) * to fail by clearing migrate flags
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 647) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 648) ext4_clear_inode_state(inode, EXT4_STATE_EXT_MIGRATE);
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 652) * Update reserved blocks/metadata blocks after successful
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 653) * block allocation which had been deferred till now. We don't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 654) * support fallocate for non extent files. So we can update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 655) * reserve space here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 656) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 657) if ((retval > 0) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 658) (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 659) ext4_da_update_reserve_space(inode, retval, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 660) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 661)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 662) if (retval > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 663) unsigned int status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 664)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 665) if (unlikely(retval != map->m_len)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 666) ext4_warning(inode->i_sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 667) "ES len assertion failed for inode "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 668) "%lu: retval %d != map->m_len %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 669) inode->i_ino, retval, map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 670) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 671) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 672)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 673) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 674) * We have to zeroout blocks before inserting them into extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 675) * status tree. Otherwise someone could look them up there and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 676) * use them before they are really zeroed. We also have to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 677) * unmap metadata before zeroing as otherwise writeback can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 678) * overwrite zeros with stale data from block device.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 679) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 680) if (flags & EXT4_GET_BLOCKS_ZERO &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 681) map->m_flags & EXT4_MAP_MAPPED &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 682) map->m_flags & EXT4_MAP_NEW) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 683) ret = ext4_issue_zeroout(inode, map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 684) map->m_pblk, map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 685) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 686) retval = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 687) goto out_sem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 688) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 689) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 690)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 691) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 692) * If the extent has been zeroed out, we don't need to update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 693) * extent status tree.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 694) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 695) if ((flags & EXT4_GET_BLOCKS_PRE_IO) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 696) ext4_es_lookup_extent(inode, map->m_lblk, NULL, &es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 697) if (ext4_es_is_written(&es))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 698) goto out_sem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 699) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 700) status = map->m_flags & EXT4_MAP_UNWRITTEN ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 701) EXTENT_STATUS_UNWRITTEN : EXTENT_STATUS_WRITTEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 702) if (!(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 703) !(status & EXTENT_STATUS_WRITTEN) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 704) ext4_es_scan_range(inode, &ext4_es_is_delayed, map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 705) map->m_lblk + map->m_len - 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 706) status |= EXTENT_STATUS_DELAYED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 707) ret = ext4_es_insert_extent(inode, map->m_lblk, map->m_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 708) map->m_pblk, status);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 709) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 710) retval = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 711) goto out_sem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 712) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 713) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 714)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 715) out_sem:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 716) up_write((&EXT4_I(inode)->i_data_sem));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 717) if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 718) ret = check_block_validity(inode, map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 719) if (ret != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 720) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 721)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 722) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 723) * Inodes with freshly allocated blocks where contents will be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 724) * visible after transaction commit must be on transaction's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 725) * ordered data list.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 726) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 727) if (map->m_flags & EXT4_MAP_NEW &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 728) !(map->m_flags & EXT4_MAP_UNWRITTEN) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 729) !(flags & EXT4_GET_BLOCKS_ZERO) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 730) !ext4_is_quota_file(inode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 731) ext4_should_order_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 732) loff_t start_byte =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 733) (loff_t)map->m_lblk << inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 734) loff_t length = (loff_t)map->m_len << inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 736) if (flags & EXT4_GET_BLOCKS_IO_SUBMIT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 737) ret = ext4_jbd2_inode_add_wait(handle, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 738) start_byte, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 739) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 740) ret = ext4_jbd2_inode_add_write(handle, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 741) start_byte, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 742) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 743) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 744) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 745) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 746) if (retval > 0 && (map->m_flags & EXT4_MAP_UNWRITTEN ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 747) map->m_flags & EXT4_MAP_MAPPED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 748) ext4_fc_track_range(handle, inode, map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 749) map->m_lblk + map->m_len - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 750) if (retval < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 751) ext_debug(inode, "failed with err %d\n", retval);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 752) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 753) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 755) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 756) * Update EXT4_MAP_FLAGS in bh->b_state. For buffer heads attached to pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 757) * we have to be careful as someone else may be manipulating b_state as well.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 758) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 759) static void ext4_update_bh_state(struct buffer_head *bh, unsigned long flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 760) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 761) unsigned long old_state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 762) unsigned long new_state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 763)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 764) flags &= EXT4_MAP_FLAGS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 765)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 766) /* Dummy buffer_head? Set non-atomically. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 767) if (!bh->b_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 768) bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 769) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 770) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 771) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 772) * Someone else may be modifying b_state. Be careful! This is ugly but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 773) * once we get rid of using bh as a container for mapping information
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 774) * to pass to / from get_block functions, this can go away.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 775) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 776) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 777) old_state = READ_ONCE(bh->b_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 778) new_state = (old_state & ~EXT4_MAP_FLAGS) | flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 779) } while (unlikely(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 780) cmpxchg(&bh->b_state, old_state, new_state) != old_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 781) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 782)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 783) static int _ext4_get_block(struct inode *inode, sector_t iblock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 784) struct buffer_head *bh, int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 785) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 786) struct ext4_map_blocks map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 787) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 788)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 789) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 790) return -ERANGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 791)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 792) map.m_lblk = iblock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 793) map.m_len = bh->b_size >> inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 794)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 795) ret = ext4_map_blocks(ext4_journal_current_handle(), inode, &map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 796) flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 797) if (ret > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 798) map_bh(bh, inode->i_sb, map.m_pblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 799) ext4_update_bh_state(bh, map.m_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 800) bh->b_size = inode->i_sb->s_blocksize * map.m_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 801) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 802) } else if (ret == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 803) /* hole case, need to fill in bh->b_size */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 804) bh->b_size = inode->i_sb->s_blocksize * map.m_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 805) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 806) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 807) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 809) int ext4_get_block(struct inode *inode, sector_t iblock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 810) struct buffer_head *bh, int create)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 811) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 812) return _ext4_get_block(inode, iblock, bh,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 813) create ? EXT4_GET_BLOCKS_CREATE : 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 814) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 815)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 816) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 817) * Get block function used when preparing for buffered write if we require
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 818) * creating an unwritten extent if blocks haven't been allocated. The extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 819) * will be converted to written after the IO is complete.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 820) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 821) int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 822) struct buffer_head *bh_result, int create)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 823) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 824) ext4_debug("ext4_get_block_unwritten: inode %lu, create flag %d\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 825) inode->i_ino, create);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 826) return _ext4_get_block(inode, iblock, bh_result,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 827) EXT4_GET_BLOCKS_IO_CREATE_EXT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 828) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 830) /* Maximum number of blocks we map for direct IO at once. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 831) #define DIO_MAX_BLOCKS 4096
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 832)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 833) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 834) * `handle' can be NULL if create is zero
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 835) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 836) struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 837) ext4_lblk_t block, int map_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 838) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 839) struct ext4_map_blocks map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 840) struct buffer_head *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 841) int create = map_flags & EXT4_GET_BLOCKS_CREATE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 842) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 843)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 844) J_ASSERT((EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 845) || handle != NULL || create == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 847) map.m_lblk = block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 848) map.m_len = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 849) err = ext4_map_blocks(handle, inode, &map, map_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 850)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 851) if (err == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 852) return create ? ERR_PTR(-ENOSPC) : NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 853) if (err < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 854) return ERR_PTR(err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 856) bh = sb_getblk(inode->i_sb, map.m_pblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 857) if (unlikely(!bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 858) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 859) if (map.m_flags & EXT4_MAP_NEW) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 860) J_ASSERT(create != 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 861) J_ASSERT((EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 862) || (handle != NULL));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 863)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 864) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 865) * Now that we do not always journal data, we should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 866) * keep in mind whether this should always journal the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 867) * new buffer as metadata. For now, regular file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 868) * writes use ext4_get_block instead, so it's not a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 869) * problem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 870) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 871) lock_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 872) BUFFER_TRACE(bh, "call get_create_access");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 873) err = ext4_journal_get_create_access(handle, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 874) if (unlikely(err)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 875) unlock_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 876) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 877) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 878) if (!buffer_uptodate(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 879) memset(bh->b_data, 0, inode->i_sb->s_blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 880) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 882) unlock_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 883) BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 884) err = ext4_handle_dirty_metadata(handle, inode, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 885) if (unlikely(err))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 886) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 887) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 888) BUFFER_TRACE(bh, "not a new buffer");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 889) return bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 890) errout:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 891) brelse(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 892) return ERR_PTR(err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 893) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 894)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 895) struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 896) ext4_lblk_t block, int map_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 897) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 898) struct buffer_head *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 899) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 900)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 901) bh = ext4_getblk(handle, inode, block, map_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 902) if (IS_ERR(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 903) return bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 904) if (!bh || ext4_buffer_uptodate(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 905) return bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 906)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 907) ret = ext4_read_bh_lock(bh, REQ_META | REQ_PRIO, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 908) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 909) put_bh(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 910) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 911) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 912) return bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 913) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 914)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 915) /* Read a contiguous batch of blocks. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 916) int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 917) bool wait, struct buffer_head **bhs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 918) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 919) int i, err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 920)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 921) for (i = 0; i < bh_count; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 922) bhs[i] = ext4_getblk(NULL, inode, block + i, 0 /* map_flags */);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 923) if (IS_ERR(bhs[i])) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 924) err = PTR_ERR(bhs[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 925) bh_count = i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 926) goto out_brelse;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 927) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 928) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 929)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 930) for (i = 0; i < bh_count; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 931) /* Note that NULL bhs[i] is valid because of holes. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 932) if (bhs[i] && !ext4_buffer_uptodate(bhs[i]))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 933) ext4_read_bh_lock(bhs[i], REQ_META | REQ_PRIO, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 934)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 935) if (!wait)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 936) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 937)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 938) for (i = 0; i < bh_count; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 939) if (bhs[i])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 940) wait_on_buffer(bhs[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 941)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 942) for (i = 0; i < bh_count; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 943) if (bhs[i] && !buffer_uptodate(bhs[i])) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 944) err = -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 945) goto out_brelse;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 946) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 947) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 948) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 949)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 950) out_brelse:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 951) for (i = 0; i < bh_count; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 952) brelse(bhs[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 953) bhs[i] = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 954) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 955) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 956) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 957)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 958) int ext4_walk_page_buffers(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 959) struct buffer_head *head,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 960) unsigned from,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 961) unsigned to,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 962) int *partial,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 963) int (*fn)(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 964) struct buffer_head *bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 965) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 966) struct buffer_head *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 967) unsigned block_start, block_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 968) unsigned blocksize = head->b_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 969) int err, ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 970) struct buffer_head *next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 972) for (bh = head, block_start = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 973) ret == 0 && (bh != head || !block_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 974) block_start = block_end, bh = next) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 975) next = bh->b_this_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 976) block_end = block_start + blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 977) if (block_end <= from || block_start >= to) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 978) if (partial && !buffer_uptodate(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 979) *partial = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 980) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 981) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 982) err = (*fn)(handle, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 983) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 984) ret = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 985) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 986) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 987) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 989) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 990) * To preserve ordering, it is essential that the hole instantiation and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 991) * the data write be encapsulated in a single transaction. We cannot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 992) * close off a transaction and start a new one between the ext4_get_block()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 993) * and the commit_write(). So doing the jbd2_journal_start at the start of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 994) * prepare_write() is the right place.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 995) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 996) * Also, this function can nest inside ext4_writepage(). In that case, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 997) * *know* that ext4_writepage() has generated enough buffer credits to do the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 998) * whole page. So we won't block on the journal in that case, which is good,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 999) * because the caller may be PF_MEMALLOC.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1000) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1001) * By accident, ext4 can be reentered when a transaction is open via
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1002) * quota file writes. If we were to commit the transaction while thus
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1003) * reentered, there can be a deadlock - we would be holding a quota
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1004) * lock, and the commit would never complete if another thread had a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1005) * transaction open and was blocking on the quota lock - a ranking
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1006) * violation.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1007) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1008) * So what we do is to rely on the fact that jbd2_journal_stop/journal_start
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1009) * will _not_ run commit under these circumstances because handle->h_ref
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1010) * is elevated. We'll still have enough credits for the tiny quotafile
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1011) * write.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1012) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1013) int do_journal_get_write_access(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1014) struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1015) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1016) int dirty = buffer_dirty(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1017) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1018)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1019) if (!buffer_mapped(bh) || buffer_freed(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1020) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1021) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1022) * __block_write_begin() could have dirtied some buffers. Clean
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1023) * the dirty bit as jbd2_journal_get_write_access() could complain
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1024) * otherwise about fs integrity issues. Setting of the dirty bit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1025) * by __block_write_begin() isn't a real problem here as we clear
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1026) * the bit before releasing a page lock and thus writeback cannot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1027) * ever write the buffer.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1028) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1029) if (dirty)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1030) clear_buffer_dirty(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1031) BUFFER_TRACE(bh, "get write access");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1032) ret = ext4_journal_get_write_access(handle, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1033) if (!ret && dirty)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1034) ret = ext4_handle_dirty_metadata(handle, NULL, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1035) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1036) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1038) #ifdef CONFIG_FS_ENCRYPTION
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1039) static int ext4_block_write_begin(struct page *page, loff_t pos, unsigned len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1040) get_block_t *get_block)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1041) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1042) unsigned from = pos & (PAGE_SIZE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1043) unsigned to = from + len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1044) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1045) unsigned block_start, block_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1046) sector_t block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1047) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1048) unsigned blocksize = inode->i_sb->s_blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1049) unsigned bbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1050) struct buffer_head *bh, *head, *wait[2];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1051) int nr_wait = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1052) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1054) BUG_ON(!PageLocked(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1055) BUG_ON(from > PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1056) BUG_ON(to > PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1057) BUG_ON(from > to);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1058)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1059) if (!page_has_buffers(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1060) create_empty_buffers(page, blocksize, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1061) head = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1062) bbits = ilog2(blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1063) block = (sector_t)page->index << (PAGE_SHIFT - bbits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1064)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1065) for (bh = head, block_start = 0; bh != head || !block_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1066) block++, block_start = block_end, bh = bh->b_this_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1067) block_end = block_start + blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1068) if (block_end <= from || block_start >= to) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1069) if (PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1070) if (!buffer_uptodate(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1071) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1072) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1073) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1074) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1075) if (buffer_new(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1076) clear_buffer_new(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1077) if (!buffer_mapped(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1078) WARN_ON(bh->b_size != blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1079) err = get_block(inode, block, bh, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1080) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1081) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1082) if (buffer_new(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1083) if (PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1084) clear_buffer_new(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1085) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1086) mark_buffer_dirty(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1087) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1088) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1089) if (block_end > to || block_start < from)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1090) zero_user_segments(page, to, block_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1091) block_start, from);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1092) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1093) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1094) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1095) if (PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1096) if (!buffer_uptodate(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1097) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1098) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1099) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1100) if (!buffer_uptodate(bh) && !buffer_delay(bh) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1101) !buffer_unwritten(bh) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1102) (block_start < from || block_end > to)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1103) ext4_read_bh_lock(bh, 0, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1104) wait[nr_wait++] = bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1105) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1106) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1107) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1108) * If we issued read requests, let them complete.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1109) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1110) for (i = 0; i < nr_wait; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1111) wait_on_buffer(wait[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1112) if (!buffer_uptodate(wait[i]))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1113) err = -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1114) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1115) if (unlikely(err)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1116) page_zero_new_buffers(page, from, to);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1117) } else if (fscrypt_inode_uses_fs_layer_crypto(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1118) for (i = 0; i < nr_wait; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1119) int err2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1120)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1121) err2 = fscrypt_decrypt_pagecache_blocks(page, blocksize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1122) bh_offset(wait[i]));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1123) if (err2) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1124) clear_buffer_uptodate(wait[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1125) err = err2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1126) }
^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)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1130) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1131) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1132) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1133)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1134) static int ext4_write_begin(struct file *file, struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1135) loff_t pos, unsigned len, unsigned flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1136) struct page **pagep, void **fsdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1137) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1138) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1139) int ret, needed_blocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1140) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1141) int retries = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1142) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1143) pgoff_t index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1144) unsigned from, to;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1145)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1146) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1147) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1148)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1149) if (trace_android_fs_datawrite_start_enabled()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1150) char *path, pathbuf[MAX_TRACE_PATHBUF_LEN];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1152) path = android_fstrace_get_pathname(pathbuf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1153) MAX_TRACE_PATHBUF_LEN,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1154) inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1155) trace_android_fs_datawrite_start(inode, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1156) current->pid, path,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1157) current->comm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1158) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1159) trace_ext4_write_begin(inode, pos, len, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1160) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1161) * Reserve one block more for addition to orphan list in case
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1162) * we allocate blocks but write fails for some reason
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1163) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1164) needed_blocks = ext4_writepage_trans_blocks(inode) + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1165) index = pos >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1166) from = pos & (PAGE_SIZE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1167) to = from + len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1168)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1169) if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1170) ret = ext4_try_to_write_inline_data(mapping, inode, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1171) flags, pagep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1172) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1173) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1174) if (ret == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1175) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1176) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1177)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1178) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1179) * grab_cache_page_write_begin() can take a long time if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1180) * system is thrashing due to memory pressure, or if the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1181) * is being written back. So grab it first before we start
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1182) * the transaction handle. This also allows us to allocate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1183) * the page (if needed) without using GFP_NOFS.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1184) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1185) retry_grab:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1186) page = grab_cache_page_write_begin(mapping, index, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1187) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1188) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1189) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1190)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1191) retry_journal:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1192) handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE, needed_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1193) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1194) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1195) return PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1196) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1197)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1198) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1199) if (page->mapping != mapping) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1200) /* The page got truncated from under us */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1201) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1202) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1203) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1204) goto retry_grab;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1205) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1206) /* In case writeback began while the page was unlocked */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1207) wait_for_stable_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1209) #ifdef CONFIG_FS_ENCRYPTION
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1210) if (ext4_should_dioread_nolock(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1211) ret = ext4_block_write_begin(page, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1212) ext4_get_block_unwritten);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1213) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1214) ret = ext4_block_write_begin(page, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1215) ext4_get_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1216) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1217) if (ext4_should_dioread_nolock(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1218) ret = __block_write_begin(page, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1219) ext4_get_block_unwritten);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1220) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1221) ret = __block_write_begin(page, pos, len, ext4_get_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1222) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1223) if (!ret && ext4_should_journal_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1224) ret = ext4_walk_page_buffers(handle, page_buffers(page),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1225) from, to, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1226) do_journal_get_write_access);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1227) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1228)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1229) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1230) bool extended = (pos + len > inode->i_size) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1231) !ext4_verity_in_progress(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1232)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1233) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1234) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1235) * __block_write_begin may have instantiated a few blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1236) * outside i_size. Trim these off again. Don't need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1237) * i_size_read because we hold i_mutex.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1238) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1239) * Add inode to orphan list in case we crash before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1240) * truncate finishes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1241) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1242) if (extended && ext4_can_truncate(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1243) ext4_orphan_add(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1244)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1245) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1246) if (extended) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1247) ext4_truncate_failed_write(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1248) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1249) * If truncate failed early the inode might
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1250) * still be on the orphan list; we need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1251) * make sure the inode is removed from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1252) * orphan list in that case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1253) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1254) if (inode->i_nlink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1255) ext4_orphan_del(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1256) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1257)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1258) if (ret == -ENOSPC &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1259) ext4_should_retry_alloc(inode->i_sb, &retries))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1260) goto retry_journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1261) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1262) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1263) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1264) *pagep = page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1265) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1266) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1267)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1268) /* For write_end() in data=journal mode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1269) static int write_end_fn(handle_t *handle, struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1270) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1271) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1272) if (!buffer_mapped(bh) || buffer_freed(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1273) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1274) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1275) ret = ext4_handle_dirty_metadata(handle, NULL, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1276) clear_buffer_meta(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1277) clear_buffer_prio(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1278) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1279) }
^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) * We need to pick up the new inode size which generic_commit_write gave us
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1283) * `file' can be NULL - eg, when called from page_symlink().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1284) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1285) * ext4 never places buffers on inode->i_mapping->private_list. metadata
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1286) * buffers are managed internally.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1287) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1288) static int ext4_write_end(struct file *file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1289) struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1290) loff_t pos, unsigned len, unsigned copied,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1291) struct page *page, void *fsdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1292) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1293) handle_t *handle = ext4_journal_current_handle();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1294) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1295) loff_t old_size = inode->i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1296) int ret = 0, ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1297) int i_size_changed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1298) int inline_data = ext4_has_inline_data(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1299) bool verity = ext4_verity_in_progress(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1300)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1301) trace_android_fs_datawrite_end(inode, pos, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1302) trace_ext4_write_end(inode, pos, len, copied);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1303) if (inline_data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1304) ret = ext4_write_inline_data_end(inode, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1305) copied, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1306) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1307) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1308) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1309) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1311) copied = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1312) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1313) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1314) copied = block_write_end(file, mapping, pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1315) len, copied, page, fsdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1316) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1317) * it's important to update i_size while still holding page lock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1318) * page writeout could otherwise come in and zero beyond i_size.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1319) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1320) * If FS_IOC_ENABLE_VERITY is running on this inode, then Merkle tree
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1321) * blocks are being written past EOF, so skip the i_size update.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1322) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1323) if (!verity)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1324) i_size_changed = ext4_update_inode_size(inode, pos + copied);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1325) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1326) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1327)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1328) if (old_size < pos && !verity)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1329) pagecache_isize_extended(inode, old_size, pos);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1330) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1331) * Don't mark the inode dirty under page lock. First, it unnecessarily
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1332) * makes the holding time of page lock longer. Second, it forces lock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1333) * ordering of page lock and transaction start for journaling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1334) * filesystems.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1335) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1336) if (i_size_changed || inline_data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1337) ret = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1338)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1339) errout:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1340) if (pos + len > inode->i_size && !verity && ext4_can_truncate(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1341) /* if we have allocated more blocks and copied
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1342) * less. We will have blocks allocated outside
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1343) * inode->i_size. So truncate them
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1344) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1345) ext4_orphan_add(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1346)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1347) ret2 = ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1348) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1349) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1350)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1351) if (pos + len > inode->i_size && !verity) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1352) ext4_truncate_failed_write(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1353) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1354) * If truncate failed early the inode might still be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1355) * on the orphan list; we need to make sure the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1356) * is removed from the orphan list in that case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1357) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1358) if (inode->i_nlink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1359) ext4_orphan_del(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1360) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1361)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1362) return ret ? ret : copied;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1363) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1365) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1366) * This is a private version of page_zero_new_buffers() which doesn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1367) * set the buffer to be dirty, since in data=journalled mode we need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1368) * to call ext4_handle_dirty_metadata() instead.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1369) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1370) static void ext4_journalled_zero_new_buffers(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1371) struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1372) unsigned from, unsigned to)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1373) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1374) unsigned int block_start = 0, block_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1375) struct buffer_head *head, *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1376)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1377) bh = head = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1378) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1379) block_end = block_start + bh->b_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1380) if (buffer_new(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1381) if (block_end > from && block_start < to) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1382) if (!PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1383) unsigned start, size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1384)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1385) start = max(from, block_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1386) size = min(to, block_end) - start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1387)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1388) zero_user(page, start, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1389) write_end_fn(handle, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1390) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1391) clear_buffer_new(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1392) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1393) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1394) block_start = block_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1395) bh = bh->b_this_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1396) } while (bh != head);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1397) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1399) static int ext4_journalled_write_end(struct file *file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1400) struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1401) loff_t pos, unsigned len, unsigned copied,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1402) struct page *page, void *fsdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1403) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1404) handle_t *handle = ext4_journal_current_handle();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1405) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1406) loff_t old_size = inode->i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1407) int ret = 0, ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1408) int partial = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1409) unsigned from, to;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1410) int size_changed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1411) int inline_data = ext4_has_inline_data(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1412) bool verity = ext4_verity_in_progress(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1413)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1414) trace_android_fs_datawrite_end(inode, pos, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1415) trace_ext4_journalled_write_end(inode, pos, len, copied);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1416) from = pos & (PAGE_SIZE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1417) to = from + len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1419) BUG_ON(!ext4_handle_valid(handle));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1420)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1421) if (inline_data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1422) ret = ext4_write_inline_data_end(inode, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1423) copied, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1424) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1425) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1426) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1427) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1428) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1429) copied = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1430) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1431) } else if (unlikely(copied < len) && !PageUptodate(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1432) copied = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1433) ext4_journalled_zero_new_buffers(handle, page, from, to);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1434) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1435) if (unlikely(copied < len))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1436) ext4_journalled_zero_new_buffers(handle, page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1437) from + copied, to);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1438) ret = ext4_walk_page_buffers(handle, page_buffers(page), from,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1439) from + copied, &partial,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1440) write_end_fn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1441) if (!partial)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1442) SetPageUptodate(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1443) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1444) if (!verity)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1445) size_changed = ext4_update_inode_size(inode, pos + copied);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1446) ext4_set_inode_state(inode, EXT4_STATE_JDATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1447) EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1448) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1449) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1450)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1451) if (old_size < pos && !verity)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1452) pagecache_isize_extended(inode, old_size, pos);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1454) if (size_changed || inline_data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1455) ret2 = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1456) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1457) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1458) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1460) errout:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1461) if (pos + len > inode->i_size && !verity && ext4_can_truncate(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1462) /* if we have allocated more blocks and copied
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1463) * less. We will have blocks allocated outside
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1464) * inode->i_size. So truncate them
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1465) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1466) ext4_orphan_add(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1467)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1468) ret2 = ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1469) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1470) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1471) if (pos + len > inode->i_size && !verity) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1472) ext4_truncate_failed_write(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1473) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1474) * If truncate failed early the inode might still be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1475) * on the orphan list; we need to make sure the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1476) * is removed from the orphan list in that case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1477) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1478) if (inode->i_nlink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1479) ext4_orphan_del(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1480) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1481)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1482) return ret ? ret : copied;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1483) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1484)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1485) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1486) * Reserve space for a single cluster
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1487) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1488) static int ext4_da_reserve_space(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1489) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1490) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1491) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1492) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1493)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1494) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1495) * We will charge metadata quota at writeout time; this saves
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1496) * us from metadata over-estimation, though we may go over by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1497) * a small amount in the end. Here we just reserve for data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1498) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1499) ret = dquot_reserve_block(inode, EXT4_C2B(sbi, 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1500) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1501) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1502)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1503) spin_lock(&ei->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1504) if (ext4_claim_free_clusters(sbi, 1, 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1505) spin_unlock(&ei->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1506) dquot_release_reservation_block(inode, EXT4_C2B(sbi, 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1507) return -ENOSPC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1508) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1509) ei->i_reserved_data_blocks++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1510) trace_ext4_da_reserve_space(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1511) spin_unlock(&ei->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1513) return 0; /* success */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1514) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1515)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1516) void ext4_da_release_space(struct inode *inode, int to_free)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1517) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1518) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1519) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1521) if (!to_free)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1522) return; /* Nothing to release, exit */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1524) spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1525)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1526) trace_ext4_da_release_space(inode, to_free);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1527) if (unlikely(to_free > ei->i_reserved_data_blocks)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1528) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1529) * if there aren't enough reserved blocks, then the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1530) * counter is messed up somewhere. Since this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1531) * function is called from invalidate page, it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1532) * harmless to return without any action.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1533) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1534) ext4_warning(inode->i_sb, "ext4_da_release_space: "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1535) "ino %lu, to_free %d with only %d reserved "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1536) "data blocks", inode->i_ino, to_free,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1537) ei->i_reserved_data_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1538) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1539) to_free = ei->i_reserved_data_blocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1540) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1541) ei->i_reserved_data_blocks -= to_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1542)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1543) /* update fs dirty data blocks counter */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1544) percpu_counter_sub(&sbi->s_dirtyclusters_counter, to_free);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1546) spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1547)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1548) dquot_release_reservation_block(inode, EXT4_C2B(sbi, to_free));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1549) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1550)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1551) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1552) * Delayed allocation stuff
^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) struct mpage_da_data {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1556) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1557) struct writeback_control *wbc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1558)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1559) pgoff_t first_page; /* The first page to write */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1560) pgoff_t next_page; /* Current page to examine */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1561) pgoff_t last_page; /* Last page to examine */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1562) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1563) * Extent to map - this can be after first_page because that can be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1564) * fully mapped. We somewhat abuse m_flags to store whether the extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1565) * is delalloc or unwritten.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1566) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1567) struct ext4_map_blocks map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1568) struct ext4_io_submit io_submit; /* IO submission data */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1569) unsigned int do_map:1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1570) unsigned int scanned_until_end:1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1571) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1572)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1573) static void mpage_release_unused_pages(struct mpage_da_data *mpd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1574) bool invalidate)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1575) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1576) int nr_pages, i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1577) pgoff_t index, end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1578) struct pagevec pvec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1579) struct inode *inode = mpd->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1580) struct address_space *mapping = inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1581)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1582) /* This is necessary when next_page == 0. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1583) if (mpd->first_page >= mpd->next_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1584) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1585)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1586) mpd->scanned_until_end = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1587) index = mpd->first_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1588) end = mpd->next_page - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1589) if (invalidate) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1590) ext4_lblk_t start, last;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1591) start = index << (PAGE_SHIFT - inode->i_blkbits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1592) last = end << (PAGE_SHIFT - inode->i_blkbits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1593) ext4_es_remove_extent(inode, start, last - start + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1594) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1595)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1596) pagevec_init(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1597) while (index <= end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1598) nr_pages = pagevec_lookup_range(&pvec, mapping, &index, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1599) if (nr_pages == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1600) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1601) for (i = 0; i < nr_pages; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1602) struct page *page = pvec.pages[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1603)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1604) BUG_ON(!PageLocked(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1605) BUG_ON(PageWriteback(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1606) if (invalidate) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1607) if (page_mapped(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1608) clear_page_dirty_for_io(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1609) block_invalidatepage(page, 0, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1610) ClearPageUptodate(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1611) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1612) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1613) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1614) pagevec_release(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1615) }
^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) static void ext4_print_free_blocks(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1619) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1620) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1621) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1622) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1623)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1624) ext4_msg(sb, KERN_CRIT, "Total free blocks count %lld",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1625) EXT4_C2B(EXT4_SB(inode->i_sb),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1626) ext4_count_free_clusters(sb)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1627) ext4_msg(sb, KERN_CRIT, "Free/Dirty block details");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1628) ext4_msg(sb, KERN_CRIT, "free_blocks=%lld",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1629) (long long) EXT4_C2B(EXT4_SB(sb),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1630) percpu_counter_sum(&sbi->s_freeclusters_counter)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1631) ext4_msg(sb, KERN_CRIT, "dirty_blocks=%lld",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1632) (long long) EXT4_C2B(EXT4_SB(sb),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1633) percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1634) ext4_msg(sb, KERN_CRIT, "Block reservation details");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1635) ext4_msg(sb, KERN_CRIT, "i_reserved_data_blocks=%u",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1636) ei->i_reserved_data_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1637) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1638) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1639)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1640) static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1641) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1642) return (buffer_delay(bh) || buffer_unwritten(bh)) && buffer_dirty(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1643) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1644)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1645) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1646) * ext4_insert_delayed_block - adds a delayed block to the extents status
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1647) * tree, incrementing the reserved cluster/block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1648) * count or making a pending reservation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1649) * where needed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1650) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1651) * @inode - file containing the newly added block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1652) * @lblk - logical block to be added
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1653) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1654) * Returns 0 on success, negative error code on failure.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1655) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1656) static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1657) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1658) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1659) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1660) bool allocated = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1661) bool reserved = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1662)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1663) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1664) * If the cluster containing lblk is shared with a delayed,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1665) * written, or unwritten extent in a bigalloc file system, it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1666) * already been accounted for and does not need to be reserved.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1667) * A pending reservation must be made for the cluster if it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1668) * shared with a written or unwritten extent and doesn't already
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1669) * have one. Written and unwritten extents can be purged from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1670) * extents status tree if the system is under memory pressure, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1671) * it's necessary to examine the extent tree if a search of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1672) * extents status tree doesn't get a match.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1673) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1674) if (sbi->s_cluster_ratio == 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1675) ret = ext4_da_reserve_space(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1676) if (ret != 0) /* ENOSPC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1677) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1678) reserved = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1679) } else { /* bigalloc */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1680) if (!ext4_es_scan_clu(inode, &ext4_es_is_delonly, lblk)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1681) if (!ext4_es_scan_clu(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1682) &ext4_es_is_mapped, lblk)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1683) ret = ext4_clu_mapped(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1684) EXT4_B2C(sbi, lblk));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1685) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1686) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1687) if (ret == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1688) ret = ext4_da_reserve_space(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1689) if (ret != 0) /* ENOSPC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1690) goto errout;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1691) reserved = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1692) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1693) allocated = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1694) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1695) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1696) allocated = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1697) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1698) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1699) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1700)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1701) ret = ext4_es_insert_delayed_block(inode, lblk, allocated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1702) if (ret && reserved)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1703) ext4_da_release_space(inode, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1704)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1705) errout:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1706) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1707) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1708)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1709) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1710) * This function is grabs code from the very beginning of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1711) * ext4_map_blocks, but assumes that the caller is from delayed write
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1712) * time. This function looks up the requested blocks and sets the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1713) * buffer delay bit under the protection of i_data_sem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1714) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1715) static int ext4_da_map_blocks(struct inode *inode, sector_t iblock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1716) struct ext4_map_blocks *map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1717) struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1718) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1719) struct extent_status es;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1720) int retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1721) sector_t invalid_block = ~((sector_t) 0xffff);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1722) #ifdef ES_AGGRESSIVE_TEST
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1723) struct ext4_map_blocks orig_map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1724)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1725) memcpy(&orig_map, map, sizeof(*map));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1726) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1727)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1728) if (invalid_block < ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1729) invalid_block = ~0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1731) map->m_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1732) ext_debug(inode, "max_blocks %u, logical block %lu\n", map->m_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1733) (unsigned long) map->m_lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1735) /* Lookup extent status tree firstly */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1736) if (ext4_es_lookup_extent(inode, iblock, NULL, &es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1737) if (ext4_es_is_hole(&es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1738) retval = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1739) down_read(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1740) goto add_delayed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1741) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1742)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1743) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1744) * Delayed extent could be allocated by fallocate.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1745) * So we need to check it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1746) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1747) if (ext4_es_is_delayed(&es) && !ext4_es_is_unwritten(&es)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1748) map_bh(bh, inode->i_sb, invalid_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1749) set_buffer_new(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1750) set_buffer_delay(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1751) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1752) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1753)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1754) map->m_pblk = ext4_es_pblock(&es) + iblock - es.es_lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1755) retval = es.es_len - (iblock - es.es_lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1756) if (retval > map->m_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1757) retval = map->m_len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1758) map->m_len = retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1759) if (ext4_es_is_written(&es))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1760) map->m_flags |= EXT4_MAP_MAPPED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1761) else if (ext4_es_is_unwritten(&es))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1762) map->m_flags |= EXT4_MAP_UNWRITTEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1763) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1764) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1765)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1766) #ifdef ES_AGGRESSIVE_TEST
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1767) ext4_map_blocks_es_recheck(NULL, inode, map, &orig_map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1768) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1769) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1770) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1771)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1772) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1773) * Try to see if we can get the block without requesting a new
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1774) * file system block.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1775) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1776) down_read(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1777) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1778) retval = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1779) else if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1780) retval = ext4_ext_map_blocks(NULL, inode, map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1781) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1782) retval = ext4_ind_map_blocks(NULL, inode, map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1783)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1784) add_delayed:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1785) if (retval == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1786) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1787)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1788) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1789) * XXX: __block_prepare_write() unmaps passed block,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1790) * is it OK?
^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) ret = ext4_insert_delayed_block(inode, map->m_lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1794) if (ret != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1795) retval = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1796) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1797) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1798)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1799) map_bh(bh, inode->i_sb, invalid_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1800) set_buffer_new(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1801) set_buffer_delay(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1802) } else if (retval > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1803) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1804) unsigned int status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1806) if (unlikely(retval != map->m_len)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1807) ext4_warning(inode->i_sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1808) "ES len assertion failed for inode "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1809) "%lu: retval %d != map->m_len %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1810) inode->i_ino, retval, map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1811) WARN_ON(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1812) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1813)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1814) status = map->m_flags & EXT4_MAP_UNWRITTEN ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1815) EXTENT_STATUS_UNWRITTEN : EXTENT_STATUS_WRITTEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1816) ret = ext4_es_insert_extent(inode, map->m_lblk, map->m_len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1817) map->m_pblk, status);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1818) if (ret != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1819) retval = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1820) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1821)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1822) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1823) up_read((&EXT4_I(inode)->i_data_sem));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1825) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1826) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1827)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1828) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1829) * This is a special get_block_t callback which is used by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1830) * ext4_da_write_begin(). It will either return mapped block or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1831) * reserve space for a single block.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1832) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1833) * For delayed buffer_head we have BH_Mapped, BH_New, BH_Delay set.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1834) * We also have b_blocknr = -1 and b_bdev initialized properly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1835) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1836) * For unwritten buffer_head we have BH_Mapped, BH_New, BH_Unwritten set.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1837) * We also have b_blocknr = physicalblock mapping unwritten extent and b_bdev
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1838) * initialized properly.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1839) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1840) int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1841) struct buffer_head *bh, int create)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1842) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1843) struct ext4_map_blocks map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1844) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1846) BUG_ON(create == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1847) BUG_ON(bh->b_size != inode->i_sb->s_blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1849) map.m_lblk = iblock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1850) map.m_len = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1851)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1852) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1853) * first, we need to know whether the block is allocated already
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1854) * preallocated blocks are unmapped but should treated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1855) * the same as allocated blocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1856) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1857) ret = ext4_da_map_blocks(inode, iblock, &map, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1858) if (ret <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1859) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1860)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1861) map_bh(bh, inode->i_sb, map.m_pblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1862) ext4_update_bh_state(bh, map.m_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1863)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1864) if (buffer_unwritten(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1865) /* A delayed write to unwritten bh should be marked
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1866) * new and mapped. Mapped ensures that we don't do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1867) * get_block multiple times when we write to the same
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1868) * offset and new ensures that we do proper zero out
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1869) * for partial write.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1870) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1871) set_buffer_new(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1872) set_buffer_mapped(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1874) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1875) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1876)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1877) static int bget_one(handle_t *handle, struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1878) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1879) get_bh(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1880) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1882)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1883) static int bput_one(handle_t *handle, struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1884) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1885) put_bh(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1886) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1887) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1888)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1889) static int __ext4_journalled_writepage(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1890) unsigned int len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1891) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1892) struct address_space *mapping = page->mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1893) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1894) struct buffer_head *page_bufs = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1895) handle_t *handle = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1896) int ret = 0, err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1897) int inline_data = ext4_has_inline_data(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1898) struct buffer_head *inode_bh = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1899)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1900) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1901)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1902) if (inline_data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1903) BUG_ON(page->index != 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1904) BUG_ON(len > ext4_get_max_inline_size(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1905) inode_bh = ext4_journalled_write_inline_data(inode, len, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1906) if (inode_bh == NULL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1907) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1908) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1909) page_bufs = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1910) if (!page_bufs) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1911) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1912) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1913) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1914) ext4_walk_page_buffers(handle, page_bufs, 0, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1915) NULL, bget_one);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1916) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1917) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1918) * We need to release the page lock before we start the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1919) * journal, so grab a reference so the page won't disappear
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1920) * out from under us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1921) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1922) get_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1923) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1924)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1925) handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1926) ext4_writepage_trans_blocks(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1927) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1928) ret = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1929) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1930) goto out_no_pagelock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1931) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1932) BUG_ON(!ext4_handle_valid(handle));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1933)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1934) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1935) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1936) if (page->mapping != mapping) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1937) /* The page got truncated from under us */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1938) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1939) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1940) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1941) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1943) if (inline_data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1944) ret = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1945) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1946) ret = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1947) do_journal_get_write_access);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1948)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1949) err = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1950) write_end_fn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1951) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1952) if (ret == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1953) ret = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1954) err = ext4_jbd2_inode_add_write(handle, inode, page_offset(page), len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1955) if (ret == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1956) ret = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1957) EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1958) err = ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1959) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1960) ret = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1962) ext4_set_inode_state(inode, EXT4_STATE_JDATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1963) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1964) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1965) out_no_pagelock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1966) if (!inline_data && page_bufs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1967) ext4_walk_page_buffers(NULL, page_bufs, 0, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1968) NULL, bput_one);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1969) brelse(inode_bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1970) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1971) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1972)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1973) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1974) * Note that we don't need to start a transaction unless we're journaling data
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1975) * because we should have holes filled from ext4_page_mkwrite(). We even don't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1976) * need to file the inode to the transaction's list in ordered mode because if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1977) * we are writing back data added by write(), the inode is already there and if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1978) * we are writing back data modified via mmap(), no one guarantees in which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1979) * transaction the data will hit the disk. In case we are journaling data, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1980) * cannot start transaction directly because transaction start ranks above page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1981) * lock so we have to do some magic.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1982) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1983) * This function can get called via...
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1984) * - ext4_writepages after taking page lock (have journal handle)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1985) * - journal_submit_inode_data_buffers (no journal handle)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1986) * - shrink_page_list via the kswapd/direct reclaim (no journal handle)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1987) * - grab_page_cache when doing write_begin (have journal handle)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1988) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1989) * We don't do any block allocation in this function. If we have page with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1990) * multiple blocks we need to write those buffer_heads that are mapped. This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1991) * is important for mmaped based write. So if we do with blocksize 1K
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1992) * truncate(f, 1024);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1993) * a = mmap(f, 0, 4096);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1994) * a[0] = 'a';
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1995) * truncate(f, 4096);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1996) * we have in the page first buffer_head mapped via page_mkwrite call back
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1997) * but other buffer_heads would be unmapped but dirty (dirty done via the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1998) * do_wp_page). So writepage should write the first block. If we modify
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1999) * the mmap area beyond 1024 we will again get a page_fault and the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2000) * page_mkwrite callback will do the block allocation and mark the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2001) * buffer_heads mapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2002) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2003) * We redirty the page if we have any buffer_heads that is either delay or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2004) * unwritten in the page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2005) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2006) * We can get recursively called as show below.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2007) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2008) * ext4_writepage() -> kmalloc() -> __alloc_pages() -> page_launder() ->
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2009) * ext4_writepage()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2010) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2011) * But since we don't do any block allocation we should not deadlock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2012) * Page also have the dirty flag cleared so we don't get recurive page_lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2013) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2014) static int ext4_writepage(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2015) struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2016) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2017) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2018) loff_t size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2019) unsigned int len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2020) struct buffer_head *page_bufs = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2021) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2022) struct ext4_io_submit io_submit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2023) bool keep_towrite = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2024)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2025) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2026) inode->i_mapping->a_ops->invalidatepage(page, 0, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2027) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2028) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2029) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2030)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2031) trace_ext4_writepage(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2032) size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2033) if (page->index == size >> PAGE_SHIFT &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2034) !ext4_verity_in_progress(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2035) len = size & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2036) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2037) len = PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2038)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2039) /* Should never happen but for bugs in other kernel subsystems */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2040) if (!page_has_buffers(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2041) ext4_warning_inode(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2042) "page %lu does not have buffers attached", page->index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2043) ClearPageDirty(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2044) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2045) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2046) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2047)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2048) page_bufs = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2049) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2050) * We cannot do block allocation or other extent handling in this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2051) * function. If there are buffers needing that, we have to redirty
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2052) * the page. But we may reach here when we do a journal commit via
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2053) * journal_submit_inode_data_buffers() and in that case we must write
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2054) * allocated buffers to achieve data=ordered mode guarantees.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2055) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2056) * Also, if there is only one buffer per page (the fs block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2057) * size == the page size), if one buffer needs block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2058) * allocation or needs to modify the extent tree to clear the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2059) * unwritten flag, we know that the page can't be written at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2060) * all, so we might as well refuse the write immediately.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2061) * Unfortunately if the block size != page size, we can't as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2062) * easily detect this case using ext4_walk_page_buffers(), but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2063) * for the extremely common case, this is an optimization that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2064) * skips a useless round trip through ext4_bio_write_page().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2065) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2066) if (ext4_walk_page_buffers(NULL, page_bufs, 0, len, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2067) ext4_bh_delay_or_unwritten)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2068) redirty_page_for_writepage(wbc, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2069) if ((current->flags & PF_MEMALLOC) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2070) (inode->i_sb->s_blocksize == PAGE_SIZE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2071) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2072) * For memory cleaning there's no point in writing only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2073) * some buffers. So just bail out. Warn if we came here
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2074) * from direct reclaim.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2075) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2076) WARN_ON_ONCE((current->flags & (PF_MEMALLOC|PF_KSWAPD))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2077) == PF_MEMALLOC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2078) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2079) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2080) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2081) keep_towrite = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2082) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2084) if (PageChecked(page) && ext4_should_journal_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2085) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2086) * It's mmapped pagecache. Add buffers and journal it. There
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2087) * doesn't seem much point in redirtying the page here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2088) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2089) return __ext4_journalled_writepage(page, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2090)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2091) ext4_io_submit_init(&io_submit, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2092) io_submit.io_end = ext4_init_io_end(inode, GFP_NOFS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2093) if (!io_submit.io_end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2094) redirty_page_for_writepage(wbc, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2095) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2096) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2097) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2098) ret = ext4_bio_write_page(&io_submit, page, len, wbc, keep_towrite);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2099) ext4_io_submit(&io_submit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2100) /* Drop io_end reference we got from init */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2101) ext4_put_io_end_defer(io_submit.io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2102) return ret;
^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) static int mpage_submit_page(struct mpage_da_data *mpd, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2106) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2107) int len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2108) loff_t size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2109) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2110)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2111) BUG_ON(page->index != mpd->first_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2112) clear_page_dirty_for_io(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2113) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2114) * We have to be very careful here! Nothing protects writeback path
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2115) * against i_size changes and the page can be writeably mapped into
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2116) * page tables. So an application can be growing i_size and writing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2117) * data through mmap while writeback runs. clear_page_dirty_for_io()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2118) * write-protects our page in page tables and the page cannot get
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2119) * written to again until we release page lock. So only after
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2120) * clear_page_dirty_for_io() we are safe to sample i_size for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2121) * ext4_bio_write_page() to zero-out tail of the written page. We rely
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2122) * on the barrier provided by TestClearPageDirty in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2123) * clear_page_dirty_for_io() to make sure i_size is really sampled only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2124) * after page tables are updated.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2125) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2126) size = i_size_read(mpd->inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2127) if (page->index == size >> PAGE_SHIFT &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2128) !ext4_verity_in_progress(mpd->inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2129) len = size & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2130) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2131) len = PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2132) err = ext4_bio_write_page(&mpd->io_submit, page, len, mpd->wbc, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2133) if (!err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2134) mpd->wbc->nr_to_write--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2135) mpd->first_page++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2136)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2137) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2138) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2140) #define BH_FLAGS (BIT(BH_Unwritten) | BIT(BH_Delay))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2141)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2142) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2143) * mballoc gives us at most this number of blocks...
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2144) * XXX: That seems to be only a limitation of ext4_mb_normalize_request().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2145) * The rest of mballoc seems to handle chunks up to full group size.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2146) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2147) #define MAX_WRITEPAGES_EXTENT_LEN 2048
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2148)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2149) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2150) * mpage_add_bh_to_extent - try to add bh to extent of blocks to map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2151) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2152) * @mpd - extent of blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2153) * @lblk - logical number of the block in the file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2154) * @bh - buffer head we want to add to the extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2155) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2156) * The function is used to collect contig. blocks in the same state. If the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2157) * buffer doesn't require mapping for writeback and we haven't started the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2158) * extent of buffers to map yet, the function returns 'true' immediately - the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2159) * caller can write the buffer right away. Otherwise the function returns true
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2160) * if the block has been added to the extent, false if the block couldn't be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2161) * added.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2162) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2163) static bool mpage_add_bh_to_extent(struct mpage_da_data *mpd, ext4_lblk_t lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2164) struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2165) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2166) struct ext4_map_blocks *map = &mpd->map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2168) /* Buffer that doesn't need mapping for writeback? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2169) if (!buffer_dirty(bh) || !buffer_mapped(bh) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2170) (!buffer_delay(bh) && !buffer_unwritten(bh))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2171) /* So far no extent to map => we write the buffer right away */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2172) if (map->m_len == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2173) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2174) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2175) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2176)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2177) /* First block in the extent? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2178) if (map->m_len == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2179) /* We cannot map unless handle is started... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2180) if (!mpd->do_map)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2181) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2182) map->m_lblk = lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2183) map->m_len = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2184) map->m_flags = bh->b_state & BH_FLAGS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2185) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2186) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2187)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2188) /* Don't go larger than mballoc is willing to allocate */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2189) if (map->m_len >= MAX_WRITEPAGES_EXTENT_LEN)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2190) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2192) /* Can we merge the block to our big extent? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2193) if (lblk == map->m_lblk + map->m_len &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2194) (bh->b_state & BH_FLAGS) == map->m_flags) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2195) map->m_len++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2196) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2197) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2198) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2199) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2200)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2201) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2202) * mpage_process_page_bufs - submit page buffers for IO or add them to extent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2203) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2204) * @mpd - extent of blocks for mapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2205) * @head - the first buffer in the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2206) * @bh - buffer we should start processing from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2207) * @lblk - logical number of the block in the file corresponding to @bh
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2208) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2209) * Walk through page buffers from @bh upto @head (exclusive) and either submit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2210) * the page for IO if all buffers in this page were mapped and there's no
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2211) * accumulated extent of buffers to map or add buffers in the page to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2212) * extent of buffers to map. The function returns 1 if the caller can continue
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2213) * by processing the next page, 0 if it should stop adding buffers to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2214) * extent to map because we cannot extend it anymore. It can also return value
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2215) * < 0 in case of error during IO submission.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2216) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2217) static int mpage_process_page_bufs(struct mpage_da_data *mpd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2218) struct buffer_head *head,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2219) struct buffer_head *bh,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2220) ext4_lblk_t lblk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2221) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2222) struct inode *inode = mpd->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2223) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2224) ext4_lblk_t blocks = (i_size_read(inode) + i_blocksize(inode) - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2225) >> inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2226)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2227) if (ext4_verity_in_progress(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2228) blocks = EXT_MAX_BLOCKS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2229)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2230) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2231) BUG_ON(buffer_locked(bh));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2232)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2233) if (lblk >= blocks || !mpage_add_bh_to_extent(mpd, lblk, bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2234) /* Found extent to map? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2235) if (mpd->map.m_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2236) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2237) /* Buffer needs mapping and handle is not started? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2238) if (!mpd->do_map)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2239) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2240) /* Everything mapped so far and we hit EOF */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2241) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2242) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2243) } while (lblk++, (bh = bh->b_this_page) != head);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2244) /* So far everything mapped? Submit the page for IO. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2245) if (mpd->map.m_len == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2246) err = mpage_submit_page(mpd, head->b_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2247) if (err < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2248) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2249) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2250) if (lblk >= blocks) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2251) mpd->scanned_until_end = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2252) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2253) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2254) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2255) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2256)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2257) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2258) * mpage_process_page - update page buffers corresponding to changed extent and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2259) * may submit fully mapped page for IO
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2260) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2261) * @mpd - description of extent to map, on return next extent to map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2262) * @m_lblk - logical block mapping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2263) * @m_pblk - corresponding physical mapping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2264) * @map_bh - determines on return whether this page requires any further
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2265) * mapping or not.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2266) * Scan given page buffers corresponding to changed extent and update buffer
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2267) * state according to new extent state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2268) * We map delalloc buffers to their physical location, clear unwritten bits.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2269) * If the given page is not fully mapped, we update @map to the next extent in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2270) * the given page that needs mapping & return @map_bh as true.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2271) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2272) static int mpage_process_page(struct mpage_da_data *mpd, struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2273) ext4_lblk_t *m_lblk, ext4_fsblk_t *m_pblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2274) bool *map_bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2275) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2276) struct buffer_head *head, *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2277) ext4_io_end_t *io_end = mpd->io_submit.io_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2278) ext4_lblk_t lblk = *m_lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2279) ext4_fsblk_t pblock = *m_pblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2280) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2281) int blkbits = mpd->inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2282) ssize_t io_end_size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2283) struct ext4_io_end_vec *io_end_vec = ext4_last_io_end_vec(io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2284)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2285) bh = head = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2286) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2287) if (lblk < mpd->map.m_lblk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2288) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2289) if (lblk >= mpd->map.m_lblk + mpd->map.m_len) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2290) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2291) * Buffer after end of mapped extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2292) * Find next buffer in the page to map.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2293) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2294) mpd->map.m_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2295) mpd->map.m_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2296) io_end_vec->size += io_end_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2297) io_end_size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2298)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2299) err = mpage_process_page_bufs(mpd, head, bh, lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2300) if (err > 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2301) err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2302) if (!err && mpd->map.m_len && mpd->map.m_lblk > lblk) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2303) io_end_vec = ext4_alloc_io_end_vec(io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2304) if (IS_ERR(io_end_vec)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2305) err = PTR_ERR(io_end_vec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2306) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2307) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2308) io_end_vec->offset = (loff_t)mpd->map.m_lblk << blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2309) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2310) *map_bh = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2311) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2312) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2313) if (buffer_delay(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2314) clear_buffer_delay(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2315) bh->b_blocknr = pblock++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2316) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2317) clear_buffer_unwritten(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2318) io_end_size += (1 << blkbits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2319) } while (lblk++, (bh = bh->b_this_page) != head);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2320)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2321) io_end_vec->size += io_end_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2322) io_end_size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2323) *map_bh = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2324) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2325) *m_lblk = lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2326) *m_pblk = pblock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2327) return err;
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2331) * mpage_map_buffers - update buffers corresponding to changed extent and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2332) * submit fully mapped pages for IO
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2333) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2334) * @mpd - description of extent to map, on return next extent to map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2335) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2336) * Scan buffers corresponding to changed extent (we expect corresponding pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2337) * to be already locked) and update buffer state according to new extent state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2338) * We map delalloc buffers to their physical location, clear unwritten bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2339) * and mark buffers as uninit when we perform writes to unwritten extents
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2340) * and do extent conversion after IO is finished. If the last page is not fully
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2341) * mapped, we update @map to the next extent in the last page that needs
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2342) * mapping. Otherwise we submit the page for IO.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2343) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2344) static int mpage_map_and_submit_buffers(struct mpage_da_data *mpd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2345) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2346) struct pagevec pvec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2347) int nr_pages, i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2348) struct inode *inode = mpd->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2349) int bpp_bits = PAGE_SHIFT - inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2350) pgoff_t start, end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2351) ext4_lblk_t lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2352) ext4_fsblk_t pblock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2353) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2354) bool map_bh = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2356) start = mpd->map.m_lblk >> bpp_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2357) end = (mpd->map.m_lblk + mpd->map.m_len - 1) >> bpp_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2358) lblk = start << bpp_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2359) pblock = mpd->map.m_pblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2360)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2361) pagevec_init(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2362) while (start <= end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2363) nr_pages = pagevec_lookup_range(&pvec, inode->i_mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2364) &start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2365) if (nr_pages == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2366) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2367) for (i = 0; i < nr_pages; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2368) struct page *page = pvec.pages[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2369)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2370) err = mpage_process_page(mpd, page, &lblk, &pblock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2371) &map_bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2372) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2373) * If map_bh is true, means page may require further bh
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2374) * mapping, or maybe the page was submitted for IO.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2375) * So we return to call further extent mapping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2376) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2377) if (err < 0 || map_bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2378) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2379) /* Page fully mapped - let IO run! */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2380) err = mpage_submit_page(mpd, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2381) if (err < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2382) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2383) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2384) pagevec_release(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2385) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2386) /* Extent fully mapped and matches with page boundary. We are done. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2387) mpd->map.m_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2388) mpd->map.m_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2389) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2390) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2391) pagevec_release(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2392) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2393) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2394)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2395) static int mpage_map_one_extent(handle_t *handle, struct mpage_da_data *mpd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2396) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2397) struct inode *inode = mpd->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2398) struct ext4_map_blocks *map = &mpd->map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2399) int get_blocks_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2400) int err, dioread_nolock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2401)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2402) trace_ext4_da_write_pages_extent(inode, map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2403) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2404) * Call ext4_map_blocks() to allocate any delayed allocation blocks, or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2405) * to convert an unwritten extent to be initialized (in the case
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2406) * where we have written into one or more preallocated blocks). It is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2407) * possible that we're going to need more metadata blocks than
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2408) * previously reserved. However we must not fail because we're in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2409) * writeback and there is nothing we can do about it so it might result
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2410) * in data loss. So use reserved blocks to allocate metadata if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2411) * possible.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2412) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2413) * We pass in the magic EXT4_GET_BLOCKS_DELALLOC_RESERVE if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2414) * the blocks in question are delalloc blocks. This indicates
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2415) * that the blocks and quotas has already been checked when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2416) * the data was copied into the page cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2417) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2418) get_blocks_flags = EXT4_GET_BLOCKS_CREATE |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2419) EXT4_GET_BLOCKS_METADATA_NOFAIL |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2420) EXT4_GET_BLOCKS_IO_SUBMIT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2421) dioread_nolock = ext4_should_dioread_nolock(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2422) if (dioread_nolock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2423) get_blocks_flags |= EXT4_GET_BLOCKS_IO_CREATE_EXT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2424) if (map->m_flags & BIT(BH_Delay))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2425) get_blocks_flags |= EXT4_GET_BLOCKS_DELALLOC_RESERVE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2427) err = ext4_map_blocks(handle, inode, map, get_blocks_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2428) if (err < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2429) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2430) if (dioread_nolock && (map->m_flags & EXT4_MAP_UNWRITTEN)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2431) if (!mpd->io_submit.io_end->handle &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2432) ext4_handle_valid(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2433) mpd->io_submit.io_end->handle = handle->h_rsv_handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2434) handle->h_rsv_handle = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2435) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2436) ext4_set_io_unwritten_flag(inode, mpd->io_submit.io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2437) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2438)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2439) BUG_ON(map->m_len == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2440) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2441) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2442)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2443) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2444) * mpage_map_and_submit_extent - map extent starting at mpd->lblk of length
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2445) * mpd->len and submit pages underlying it for IO
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2446) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2447) * @handle - handle for journal operations
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2448) * @mpd - extent to map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2449) * @give_up_on_write - we set this to true iff there is a fatal error and there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2450) * is no hope of writing the data. The caller should discard
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2451) * dirty pages to avoid infinite loops.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2452) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2453) * The function maps extent starting at mpd->lblk of length mpd->len. If it is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2454) * delayed, blocks are allocated, if it is unwritten, we may need to convert
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2455) * them to initialized or split the described range from larger unwritten
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2456) * extent. Note that we need not map all the described range since allocation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2457) * can return less blocks or the range is covered by more unwritten extents. We
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2458) * cannot map more because we are limited by reserved transaction credits. On
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2459) * the other hand we always make sure that the last touched page is fully
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2460) * mapped so that it can be written out (and thus forward progress is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2461) * guaranteed). After mapping we submit all mapped pages for IO.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2462) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2463) static int mpage_map_and_submit_extent(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2464) struct mpage_da_data *mpd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2465) bool *give_up_on_write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2466) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2467) struct inode *inode = mpd->inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2468) struct ext4_map_blocks *map = &mpd->map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2469) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2470) loff_t disksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2471) int progress = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2472) ext4_io_end_t *io_end = mpd->io_submit.io_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2473) struct ext4_io_end_vec *io_end_vec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2474)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2475) io_end_vec = ext4_alloc_io_end_vec(io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2476) if (IS_ERR(io_end_vec))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2477) return PTR_ERR(io_end_vec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2478) io_end_vec->offset = ((loff_t)map->m_lblk) << inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2479) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2480) err = mpage_map_one_extent(handle, mpd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2481) if (err < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2482) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2483)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2484) if (ext4_forced_shutdown(EXT4_SB(sb)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2485) ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2486) goto invalidate_dirty_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2487) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2488) * Let the uper layers retry transient errors.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2489) * In the case of ENOSPC, if ext4_count_free_blocks()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2490) * is non-zero, a commit should free up blocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2491) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2492) if ((err == -ENOMEM) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2493) (err == -ENOSPC && ext4_count_free_clusters(sb))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2494) if (progress)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2495) goto update_disksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2496) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2497) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2498) ext4_msg(sb, KERN_CRIT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2499) "Delayed block allocation failed for "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2500) "inode %lu at logical offset %llu with"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2501) " max blocks %u with error %d",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2502) inode->i_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2503) (unsigned long long)map->m_lblk,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2504) (unsigned)map->m_len, -err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2505) ext4_msg(sb, KERN_CRIT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2506) "This should not happen!! Data will "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2507) "be lost\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2508) if (err == -ENOSPC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2509) ext4_print_free_blocks(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2510) invalidate_dirty_pages:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2511) *give_up_on_write = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2512) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2513) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2514) progress = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2515) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2516) * Update buffer state, submit mapped pages, and get us new
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2517) * extent to map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2518) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2519) err = mpage_map_and_submit_buffers(mpd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2520) if (err < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2521) goto update_disksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2522) } while (map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2524) update_disksize:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2525) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2526) * Update on-disk size after IO is submitted. Races with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2527) * truncate are avoided by checking i_size under i_data_sem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2528) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2529) disksize = ((loff_t)mpd->first_page) << PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2530) if (disksize > READ_ONCE(EXT4_I(inode)->i_disksize)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2531) int err2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2532) loff_t i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2533)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2534) down_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2535) i_size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2536) if (disksize > i_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2537) disksize = i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2538) if (disksize > EXT4_I(inode)->i_disksize)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2539) EXT4_I(inode)->i_disksize = disksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2540) up_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2541) err2 = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2542) if (err2) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2543) ext4_error_err(inode->i_sb, -err2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2544) "Failed to mark inode %lu dirty",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2545) inode->i_ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2546) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2547) if (!err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2548) err = err2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2549) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2550) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2551) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2552)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2553) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2554) * Calculate the total number of credits to reserve for one writepages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2555) * iteration. This is called from ext4_writepages(). We map an extent of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2556) * up to MAX_WRITEPAGES_EXTENT_LEN blocks and then we go on and finish mapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2557) * the last partial page. So in total we can map MAX_WRITEPAGES_EXTENT_LEN +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2558) * bpp - 1 blocks in bpp different extents.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2559) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2560) static int ext4_da_writepages_trans_blocks(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2561) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2562) int bpp = ext4_journal_blocks_per_page(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2564) return ext4_meta_trans_blocks(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2565) MAX_WRITEPAGES_EXTENT_LEN + bpp - 1, bpp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2567)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2568) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2569) * mpage_prepare_extent_to_map - find & lock contiguous range of dirty pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2570) * and underlying extent to map
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2571) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2572) * @mpd - where to look for pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2573) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2574) * Walk dirty pages in the mapping. If they are fully mapped, submit them for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2575) * IO immediately. When we find a page which isn't mapped we start accumulating
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2576) * extent of buffers underlying these pages that needs mapping (formed by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2577) * either delayed or unwritten buffers). We also lock the pages containing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2578) * these buffers. The extent found is returned in @mpd structure (starting at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2579) * mpd->lblk with length mpd->len blocks).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2580) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2581) * Note that this function can attach bios to one io_end structure which are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2582) * neither logically nor physically contiguous. Although it may seem as an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2583) * unnecessary complication, it is actually inevitable in blocksize < pagesize
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2584) * case as we need to track IO to all buffers underlying a page in one io_end.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2585) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2586) static int mpage_prepare_extent_to_map(struct mpage_da_data *mpd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2587) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2588) struct address_space *mapping = mpd->inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2589) struct pagevec pvec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2590) unsigned int nr_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2591) long left = mpd->wbc->nr_to_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2592) pgoff_t index = mpd->first_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2593) pgoff_t end = mpd->last_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2594) xa_mark_t tag;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2595) int i, err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2596) int blkbits = mpd->inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2597) ext4_lblk_t lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2598) struct buffer_head *head;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2599)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2600) if (mpd->wbc->sync_mode == WB_SYNC_ALL || mpd->wbc->tagged_writepages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2601) tag = PAGECACHE_TAG_TOWRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2602) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2603) tag = PAGECACHE_TAG_DIRTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2605) pagevec_init(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2606) mpd->map.m_len = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2607) mpd->next_page = index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2608) while (index <= end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2609) nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2610) tag);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2611) if (nr_pages == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2612) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2613)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2614) for (i = 0; i < nr_pages; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2615) struct page *page = pvec.pages[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2616)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2617) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2618) * Accumulated enough dirty pages? This doesn't apply
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2619) * to WB_SYNC_ALL mode. For integrity sync we have to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2620) * keep going because someone may be concurrently
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2621) * dirtying pages, and we might have synced a lot of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2622) * newly appeared dirty pages, but have not synced all
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2623) * of the old dirty pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2624) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2625) if (mpd->wbc->sync_mode == WB_SYNC_NONE && left <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2626) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2627)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2628) /* If we can't merge this page, we are done. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2629) if (mpd->map.m_len > 0 && mpd->next_page != page->index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2630) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2631)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2632) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2633) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2634) * If the page is no longer dirty, or its mapping no
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2635) * longer corresponds to inode we are writing (which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2636) * means it has been truncated or invalidated), or the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2637) * page is already under writeback and we are not doing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2638) * a data integrity writeback, skip the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2639) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2640) if (!PageDirty(page) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2641) (PageWriteback(page) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2642) (mpd->wbc->sync_mode == WB_SYNC_NONE)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2643) unlikely(page->mapping != mapping)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2644) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2645) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2646) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2647)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2648) wait_on_page_writeback(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2649) BUG_ON(PageWriteback(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2651) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2652) * Should never happen but for buggy code in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2653) * other subsystems that call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2654) * set_page_dirty() without properly warning
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2655) * the file system first. See [1] for more
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2656) * information.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2657) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2658) * [1] https://lore.kernel.org/linux-mm/20180103100430.GE4911@quack2.suse.cz
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2659) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2660) if (!page_has_buffers(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2661) ext4_warning_inode(mpd->inode, "page %lu does not have buffers attached", page->index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2662) ClearPageDirty(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2663) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2664) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2665) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2666)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2667) if (mpd->map.m_len == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2668) mpd->first_page = page->index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2669) mpd->next_page = page->index + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2670) /* Add all dirty buffers to mpd */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2671) lblk = ((ext4_lblk_t)page->index) <<
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2672) (PAGE_SHIFT - blkbits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2673) head = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2674) err = mpage_process_page_bufs(mpd, head, head, lblk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2675) if (err <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2676) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2677) err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2678) left--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2679) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2680) pagevec_release(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2681) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2682) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2683) mpd->scanned_until_end = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2684) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2685) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2686) pagevec_release(&pvec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2687) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2688) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2689)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2690) static int ext4_writepages(struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2691) struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2692) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2693) pgoff_t writeback_index = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2694) long nr_to_write = wbc->nr_to_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2695) int range_whole = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2696) int cycled = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2697) handle_t *handle = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2698) struct mpage_da_data mpd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2699) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2700) int needed_blocks, rsv_blocks = 0, ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2701) struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2702) struct blk_plug plug;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2703) bool give_up_on_write = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2704)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2705) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2706) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2707)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2708) percpu_down_read(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2709) trace_ext4_writepages(inode, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2710)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2711) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2712) * No pages to write? This is mainly a kludge to avoid starting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2713) * a transaction for special inodes like journal inode on last iput()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2714) * because that could violate lock ordering on umount
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2715) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2716) if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2717) goto out_writepages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2718)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2719) if (ext4_should_journal_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2720) ret = generic_writepages(mapping, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2721) goto out_writepages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2722) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2724) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2725) * If the filesystem has aborted, it is read-only, so return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2726) * right away instead of dumping stack traces later on that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2727) * will obscure the real source of the problem. We test
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2728) * EXT4_MF_FS_ABORTED instead of sb->s_flag's SB_RDONLY because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2729) * the latter could be true if the filesystem is mounted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2730) * read-only, and in that case, ext4_writepages should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2731) * *never* be called, so if that ever happens, we would want
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2732) * the stack trace.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2733) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2734) if (unlikely(ext4_forced_shutdown(EXT4_SB(mapping->host->i_sb)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2735) ext4_test_mount_flag(inode->i_sb, EXT4_MF_FS_ABORTED))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2736) ret = -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2737) goto out_writepages;
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2741) * If we have inline data and arrive here, it means that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2742) * we will soon create the block for the 1st page, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2743) * we'd better clear the inline data here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2744) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2745) if (ext4_has_inline_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2746) /* Just inode will be modified... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2747) handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2748) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2749) ret = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2750) goto out_writepages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2751) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2752) BUG_ON(ext4_test_inode_state(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2753) EXT4_STATE_MAY_INLINE_DATA));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2754) ext4_destroy_inline_data(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2755) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2756) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2757)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2758) if (ext4_should_dioread_nolock(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2759) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2760) * We may need to convert up to one extent per block in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2761) * the page and we may dirty the inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2762) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2763) rsv_blocks = 1 + ext4_chunk_trans_blocks(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2764) PAGE_SIZE >> inode->i_blkbits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2765) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2766)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2767) if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2768) range_whole = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2769)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2770) if (wbc->range_cyclic) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2771) writeback_index = mapping->writeback_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2772) if (writeback_index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2773) cycled = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2774) mpd.first_page = writeback_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2775) mpd.last_page = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2776) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2777) mpd.first_page = wbc->range_start >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2778) mpd.last_page = wbc->range_end >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2779) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2780)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2781) mpd.inode = inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2782) mpd.wbc = wbc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2783) ext4_io_submit_init(&mpd.io_submit, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2784) retry:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2785) if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2786) tag_pages_for_writeback(mapping, mpd.first_page, mpd.last_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2787) blk_start_plug(&plug);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2788)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2789) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2790) * First writeback pages that don't need mapping - we can avoid
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2791) * starting a transaction unnecessarily and also avoid being blocked
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2792) * in the block layer on device congestion while having transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2793) * started.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2794) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2795) mpd.do_map = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2796) mpd.scanned_until_end = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2797) mpd.io_submit.io_end = ext4_init_io_end(inode, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2798) if (!mpd.io_submit.io_end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2799) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2800) goto unplug;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2801) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2802) ret = mpage_prepare_extent_to_map(&mpd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2803) /* Unlock pages we didn't use */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2804) mpage_release_unused_pages(&mpd, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2805) /* Submit prepared bio */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2806) ext4_io_submit(&mpd.io_submit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2807) ext4_put_io_end_defer(mpd.io_submit.io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2808) mpd.io_submit.io_end = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2809) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2810) goto unplug;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2811)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2812) while (!mpd.scanned_until_end && wbc->nr_to_write > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2813) /* For each extent of pages we use new io_end */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2814) mpd.io_submit.io_end = ext4_init_io_end(inode, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2815) if (!mpd.io_submit.io_end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2816) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2817) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2818) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2819)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2820) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2821) * We have two constraints: We find one extent to map and we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2822) * must always write out whole page (makes a difference when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2823) * blocksize < pagesize) so that we don't block on IO when we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2824) * try to write out the rest of the page. Journalled mode is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2825) * not supported by delalloc.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2826) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2827) BUG_ON(ext4_should_journal_data(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2828) needed_blocks = ext4_da_writepages_trans_blocks(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2830) /* start a new transaction */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2831) handle = ext4_journal_start_with_reserve(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2832) EXT4_HT_WRITE_PAGE, needed_blocks, rsv_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2833) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2834) ret = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2835) ext4_msg(inode->i_sb, KERN_CRIT, "%s: jbd2_start: "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2836) "%ld pages, ino %lu; err %d", __func__,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2837) wbc->nr_to_write, inode->i_ino, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2838) /* Release allocated io_end */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2839) ext4_put_io_end(mpd.io_submit.io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2840) mpd.io_submit.io_end = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2841) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2842) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2843) mpd.do_map = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2844)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2845) trace_ext4_da_write_pages(inode, mpd.first_page, mpd.wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2846) ret = mpage_prepare_extent_to_map(&mpd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2847) if (!ret && mpd.map.m_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2848) ret = mpage_map_and_submit_extent(handle, &mpd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2849) &give_up_on_write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2850) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2851) * Caution: If the handle is synchronous,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2852) * ext4_journal_stop() can wait for transaction commit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2853) * to finish which may depend on writeback of pages to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2854) * complete or on page lock to be released. In that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2855) * case, we have to wait until after we have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2856) * submitted all the IO, released page locks we hold,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2857) * and dropped io_end reference (for extent conversion
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2858) * to be able to complete) before stopping the handle.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2859) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2860) if (!ext4_handle_valid(handle) || handle->h_sync == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2861) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2862) handle = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2863) mpd.do_map = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2864) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2865) /* Unlock pages we didn't use */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2866) mpage_release_unused_pages(&mpd, give_up_on_write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2867) /* Submit prepared bio */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2868) ext4_io_submit(&mpd.io_submit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2870) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2871) * Drop our io_end reference we got from init. We have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2872) * to be careful and use deferred io_end finishing if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2873) * we are still holding the transaction as we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2874) * release the last reference to io_end which may end
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2875) * up doing unwritten extent conversion.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2876) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2877) if (handle) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2878) ext4_put_io_end_defer(mpd.io_submit.io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2879) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2880) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2881) ext4_put_io_end(mpd.io_submit.io_end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2882) mpd.io_submit.io_end = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2883)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2884) if (ret == -ENOSPC && sbi->s_journal) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2885) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2886) * Commit the transaction which would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2887) * free blocks released in the transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2888) * and try again
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2889) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2890) jbd2_journal_force_commit_nested(sbi->s_journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2891) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2892) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2893) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2894) /* Fatal error - ENOMEM, EIO... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2895) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2896) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2897) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2898) unplug:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2899) blk_finish_plug(&plug);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2900) if (!ret && !cycled && wbc->nr_to_write > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2901) cycled = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2902) mpd.last_page = writeback_index - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2903) mpd.first_page = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2904) goto retry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2905) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2906)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2907) /* Update index */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2908) if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2909) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2910) * Set the writeback_index so that range_cyclic
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2911) * mode will write it back later
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2912) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2913) mapping->writeback_index = mpd.first_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2914)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2915) out_writepages:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2916) trace_ext4_writepages_result(inode, wbc, ret,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2917) nr_to_write - wbc->nr_to_write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2918) percpu_up_read(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2919) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2920) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2921)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2922) static int ext4_dax_writepages(struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2923) struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2924) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2925) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2926) long nr_to_write = wbc->nr_to_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2927) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2928) struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2929)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2930) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2931) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2932)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2933) percpu_down_read(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2934) trace_ext4_writepages(inode, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2935)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2936) ret = dax_writeback_mapping_range(mapping, sbi->s_daxdev, wbc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2937) trace_ext4_writepages_result(inode, wbc, ret,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2938) nr_to_write - wbc->nr_to_write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2939) percpu_up_read(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2940) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2941) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2943) static int ext4_nonda_switch(struct super_block *sb)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2944) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2945) s64 free_clusters, dirty_clusters;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2946) struct ext4_sb_info *sbi = EXT4_SB(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2948) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2949) * switch to non delalloc mode if we are running low
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2950) * on free block. The free block accounting via percpu
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2951) * counters can get slightly wrong with percpu_counter_batch getting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2952) * accumulated on each CPU without updating global counters
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2953) * Delalloc need an accurate free block accounting. So switch
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2954) * to non delalloc when we are near to error range.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2955) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2956) free_clusters =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2957) percpu_counter_read_positive(&sbi->s_freeclusters_counter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2958) dirty_clusters =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2959) percpu_counter_read_positive(&sbi->s_dirtyclusters_counter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2960) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2961) * Start pushing delalloc when 1/2 of free blocks are dirty.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2962) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2963) if (dirty_clusters && (free_clusters < 2 * dirty_clusters))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2964) try_to_writeback_inodes_sb(sb, WB_REASON_FS_FREE_SPACE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2965)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2966) if (2 * free_clusters < 3 * dirty_clusters ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2967) free_clusters < (dirty_clusters + EXT4_FREECLUSTERS_WATERMARK)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2968) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2969) * free block count is less than 150% of dirty blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2970) * or free blocks is less than watermark
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2971) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2972) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2973) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2974) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2975) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2977) /* We always reserve for an inode update; the superblock could be there too */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2978) static int ext4_da_write_credits(struct inode *inode, loff_t pos, unsigned len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2979) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2980) if (likely(ext4_has_feature_large_file(inode->i_sb)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2981) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2982)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2983) if (pos + len <= 0x7fffffffULL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2984) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2985)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2986) /* We might need to update the superblock to set LARGE_FILE */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2987) return 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2988) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2989)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2990) static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2991) loff_t pos, unsigned len, unsigned flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2992) struct page **pagep, void **fsdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2993) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2994) int ret, retries = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2995) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2996) pgoff_t index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2997) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2998) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2999)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3000) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3001) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3002)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3003) index = pos >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3004)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3005) if (ext4_nonda_switch(inode->i_sb) || S_ISLNK(inode->i_mode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3006) ext4_verity_in_progress(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3007) *fsdata = (void *)FALL_BACK_TO_NONDELALLOC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3008) return ext4_write_begin(file, mapping, pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3009) len, flags, pagep, fsdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3010) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3011) *fsdata = (void *)0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3012) if (trace_android_fs_datawrite_start_enabled()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3013) char *path, pathbuf[MAX_TRACE_PATHBUF_LEN];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3015) path = android_fstrace_get_pathname(pathbuf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3016) MAX_TRACE_PATHBUF_LEN,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3017) inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3018) trace_android_fs_datawrite_start(inode, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3019) current->pid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3020) path, current->comm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3021) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3022) trace_ext4_da_write_begin(inode, pos, len, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3024) if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3025) ret = ext4_da_write_inline_data_begin(mapping, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3026) pos, len, flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3027) pagep, fsdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3028) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3029) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3030) if (ret == 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3031) return 0;
^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) * grab_cache_page_write_begin() can take a long time if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3036) * system is thrashing due to memory pressure, or if the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3037) * is being written back. So grab it first before we start
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3038) * the transaction handle. This also allows us to allocate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3039) * the page (if needed) without using GFP_NOFS.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3040) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3041) retry_grab:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3042) page = grab_cache_page_write_begin(mapping, index, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3043) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3044) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3045) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3046)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3047) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3048) * With delayed allocation, we don't log the i_disksize update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3049) * if there is delayed block allocation. But we still need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3050) * to journalling the i_disksize update if writes to the end
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3051) * of file which has an already mapped buffer.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3052) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3053) retry_journal:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3054) handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3055) ext4_da_write_credits(inode, pos, len));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3056) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3057) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3058) return PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3059) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3060)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3061) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3062) if (page->mapping != mapping) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3063) /* The page got truncated from under us */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3064) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3065) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3066) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3067) goto retry_grab;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3068) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3069) /* In case writeback began while the page was unlocked */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3070) wait_for_stable_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3071)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3072) #ifdef CONFIG_FS_ENCRYPTION
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3073) ret = ext4_block_write_begin(page, pos, len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3074) ext4_da_get_block_prep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3075) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3076) ret = __block_write_begin(page, pos, len, ext4_da_get_block_prep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3077) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3078) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3079) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3080) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3081) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3082) * block_write_begin may have instantiated a few blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3083) * outside i_size. Trim these off again. Don't need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3084) * i_size_read because we hold i_mutex.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3085) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3086) if (pos + len > inode->i_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3087) ext4_truncate_failed_write(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3088)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3089) if (ret == -ENOSPC &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3090) ext4_should_retry_alloc(inode->i_sb, &retries))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3091) goto retry_journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3092)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3093) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3094) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3095) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3097) *pagep = page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3098) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3099) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3100)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3101) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3102) * Check if we should update i_disksize
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3103) * when write to the end of file but not require block allocation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3104) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3105) static int ext4_da_should_update_i_disksize(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3106) unsigned long offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3107) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3108) struct buffer_head *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3109) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3110) unsigned int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3111) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3113) bh = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3114) idx = offset >> inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3116) for (i = 0; i < idx; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3117) bh = bh->b_this_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3119) if (!buffer_mapped(bh) || (buffer_delay(bh)) || buffer_unwritten(bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3120) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3121) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3122) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3123)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3124) static int ext4_da_write_end(struct file *file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3125) struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3126) loff_t pos, unsigned len, unsigned copied,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3127) struct page *page, void *fsdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3128) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3129) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3130) int ret = 0, ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3131) handle_t *handle = ext4_journal_current_handle();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3132) loff_t new_i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3133) unsigned long start, end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3134) int write_mode = (int)(unsigned long)fsdata;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3135)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3136) if (write_mode == FALL_BACK_TO_NONDELALLOC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3137) return ext4_write_end(file, mapping, pos,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3138) len, copied, page, fsdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3140) trace_android_fs_datawrite_end(inode, pos, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3141) trace_ext4_da_write_end(inode, pos, len, copied);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3142) start = pos & (PAGE_SIZE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3143) end = start + copied - 1;
^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) * Since we are holding inode lock, we are sure i_disksize <=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3147) * i_size. We also know that if i_disksize < i_size, there are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3148) * delalloc writes pending in the range upto i_size. If the end of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3149) * the current write is <= i_size, there's no need to touch
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3150) * i_disksize since writeback will push i_disksize upto i_size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3151) * eventually. If the end of the current write is > i_size and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3152) * inside an allocated block (ext4_da_should_update_i_disksize()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3153) * check), we need to update i_disksize here as neither
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3154) * ext4_writepage() nor certain ext4_writepages() paths not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3155) * allocating blocks update i_disksize.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3156) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3157) * Note that we defer inode dirtying to generic_write_end() /
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3158) * ext4_da_write_inline_data_end().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3159) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3160) new_i_size = pos + copied;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3161) if (copied && new_i_size > inode->i_size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3162) if (ext4_has_inline_data(inode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3163) ext4_da_should_update_i_disksize(page, end))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3164) ext4_update_i_disksize(inode, new_i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3165) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3166)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3167) if (write_mode != CONVERT_INLINE_DATA &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3168) ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3169) ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3170) ret = ext4_da_write_inline_data_end(inode, pos, len, copied,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3171) page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3172) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3173) ret = generic_write_end(file, mapping, pos, len, copied,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3174) page, fsdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3176) copied = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3177) ret2 = ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3178) if (unlikely(ret2 && !ret))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3179) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3181) return ret ? ret : copied;
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3185) * Force all delayed allocation blocks to be allocated for a given inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3186) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3187) int ext4_alloc_da_blocks(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3188) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3189) trace_ext4_alloc_da_blocks(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3190)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3191) if (!EXT4_I(inode)->i_reserved_data_blocks)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3192) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3194) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3195) * We do something simple for now. The filemap_flush() will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3196) * also start triggering a write of the data blocks, which is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3197) * not strictly speaking necessary (and for users of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3198) * laptop_mode, not even desirable). However, to do otherwise
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3199) * would require replicating code paths in:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3200) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3201) * ext4_writepages() ->
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3202) * write_cache_pages() ---> (via passed in callback function)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3203) * __mpage_da_writepage() -->
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3204) * mpage_add_bh_to_extent()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3205) * mpage_da_map_blocks()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3206) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3207) * The problem is that write_cache_pages(), located in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3208) * mm/page-writeback.c, marks pages clean in preparation for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3209) * doing I/O, which is not desirable if we're not planning on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3210) * doing I/O at all.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3211) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3212) * We could call write_cache_pages(), and then redirty all of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3213) * the pages by calling redirty_page_for_writepage() but that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3214) * would be ugly in the extreme. So instead we would need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3215) * replicate parts of the code in the above functions,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3216) * simplifying them because we wouldn't actually intend to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3217) * write out the pages, but rather only collect contiguous
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3218) * logical block extents, call the multi-block allocator, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3219) * then update the buffer heads with the block allocations.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3220) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3221) * For now, though, we'll cheat by calling filemap_flush(),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3222) * which will map the blocks, and start the I/O, but not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3223) * actually wait for the I/O to complete.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3224) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3225) return filemap_flush(inode->i_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3226) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3227)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3228) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3229) * bmap() is special. It gets used by applications such as lilo and by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3230) * the swapper to find the on-disk block of a specific piece of data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3231) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3232) * Naturally, this is dangerous if the block concerned is still in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3233) * journal. If somebody makes a swapfile on an ext4 data-journaling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3234) * filesystem and enables swap, then they may get a nasty shock when the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3235) * data getting swapped to that swapfile suddenly gets overwritten by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3236) * the original zero's written out previously to the journal and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3237) * awaiting writeback in the kernel's buffer cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3238) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3239) * So, if we see any bmap calls here on a modified, data-journaled file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3240) * take extra steps to flush any blocks which might be in the cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3241) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3242) static sector_t ext4_bmap(struct address_space *mapping, sector_t block)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3243) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3244) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3245) journal_t *journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3246) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3247)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3248) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3249) * We can get here for an inline file via the FIBMAP ioctl
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3250) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3251) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3252) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3254) if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3255) test_opt(inode->i_sb, DELALLOC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3256) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3257) * With delalloc we want to sync the file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3258) * so that we can make sure we allocate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3259) * blocks for file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3260) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3261) filemap_write_and_wait(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3262) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3264) if (EXT4_JOURNAL(inode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3265) ext4_test_inode_state(inode, EXT4_STATE_JDATA)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3266) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3267) * This is a REALLY heavyweight approach, but the use of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3268) * bmap on dirty files is expected to be extremely rare:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3269) * only if we run lilo or swapon on a freshly made file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3270) * do we expect this to happen.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3271) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3272) * (bmap requires CAP_SYS_RAWIO so this does not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3273) * represent an unprivileged user DOS attack --- we'd be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3274) * in trouble if mortal users could trigger this path at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3275) * will.)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3276) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3277) * NB. EXT4_STATE_JDATA is not set on files other than
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3278) * regular files. If somebody wants to bmap a directory
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3279) * or symlink and gets confused because the buffer
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3280) * hasn't yet been flushed to disk, they deserve
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3281) * everything they get.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3282) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3283)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3284) ext4_clear_inode_state(inode, EXT4_STATE_JDATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3285) journal = EXT4_JOURNAL(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3286) jbd2_journal_lock_updates(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3287) err = jbd2_journal_flush(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3288) jbd2_journal_unlock_updates(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3289)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3290) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3291) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3292) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3293)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3294) return iomap_bmap(mapping, block, &ext4_iomap_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3295) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3297) static int ext4_readpage(struct file *file, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3298) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3299) int ret = -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3300) struct inode *inode = page->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3302) trace_ext4_readpage(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3303)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3304) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3305) ret = ext4_readpage_inline(inode, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3306)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3307) if (ret == -EAGAIN)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3308) return ext4_mpage_readpages(inode, NULL, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3309)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3310) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3311) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3312)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3313) static void ext4_readahead(struct readahead_control *rac)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3314) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3315) struct inode *inode = rac->mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3316)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3317) /* If the file has inline data, no need to do readahead. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3318) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3319) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3320)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3321) ext4_mpage_readpages(inode, rac, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3322) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3323)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3324) static void ext4_invalidatepage(struct page *page, unsigned int offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3325) unsigned int length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3326) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3327) trace_ext4_invalidatepage(page, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3329) /* No journalling happens on data buffers when this function is used */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3330) WARN_ON(page_has_buffers(page) && buffer_jbd(page_buffers(page)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3331)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3332) block_invalidatepage(page, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3333) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3335) static int __ext4_journalled_invalidatepage(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3336) unsigned int offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3337) unsigned int length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3338) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3339) journal_t *journal = EXT4_JOURNAL(page->mapping->host);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3340)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3341) trace_ext4_journalled_invalidatepage(page, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3342)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3343) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3344) * If it's a full truncate we just forget about the pending dirtying
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3345) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3346) if (offset == 0 && length == PAGE_SIZE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3347) ClearPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3349) return jbd2_journal_invalidatepage(journal, page, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3350) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3351)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3352) /* Wrapper for aops... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3353) static void ext4_journalled_invalidatepage(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3354) unsigned int offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3355) unsigned int length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3356) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3357) WARN_ON(__ext4_journalled_invalidatepage(page, offset, length) < 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3358) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3359)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3360) static int ext4_releasepage(struct page *page, gfp_t wait)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3361) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3362) journal_t *journal = EXT4_JOURNAL(page->mapping->host);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3364) trace_ext4_releasepage(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3365)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3366) /* Page has dirty journalled data -> cannot release */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3367) if (PageChecked(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3368) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3369) if (journal)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3370) return jbd2_journal_try_to_free_buffers(journal, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3371) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3372) return try_to_free_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3373) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3375) static bool ext4_inode_datasync_dirty(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3376) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3377) journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3379) if (journal) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3380) if (jbd2_transaction_committed(journal,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3381) EXT4_I(inode)->i_datasync_tid))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3382) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3383) if (test_opt2(inode->i_sb, JOURNAL_FAST_COMMIT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3384) return !list_empty(&EXT4_I(inode)->i_fc_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3385) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3386) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3387)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3388) /* Any metadata buffers to write? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3389) if (!list_empty(&inode->i_mapping->private_list))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3390) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3391) return inode->i_state & I_DIRTY_DATASYNC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3392) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3393)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3394) static void ext4_set_iomap(struct inode *inode, struct iomap *iomap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3395) struct ext4_map_blocks *map, loff_t offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3396) loff_t length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3397) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3398) u8 blkbits = inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3399)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3400) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3401) * Writes that span EOF might trigger an I/O size update on completion,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3402) * so consider them to be dirty for the purpose of O_DSYNC, even if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3403) * there is no other metadata changes being made or are pending.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3404) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3405) iomap->flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3406) if (ext4_inode_datasync_dirty(inode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3407) offset + length > i_size_read(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3408) iomap->flags |= IOMAP_F_DIRTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3409)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3410) if (map->m_flags & EXT4_MAP_NEW)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3411) iomap->flags |= IOMAP_F_NEW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3412)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3413) iomap->bdev = inode->i_sb->s_bdev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3414) iomap->dax_dev = EXT4_SB(inode->i_sb)->s_daxdev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3415) iomap->offset = (u64) map->m_lblk << blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3416) iomap->length = (u64) map->m_len << blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3417)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3418) if ((map->m_flags & EXT4_MAP_MAPPED) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3419) !ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3420) iomap->flags |= IOMAP_F_MERGED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3421)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3422) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3423) * Flags passed to ext4_map_blocks() for direct I/O writes can result
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3424) * in m_flags having both EXT4_MAP_MAPPED and EXT4_MAP_UNWRITTEN bits
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3425) * set. In order for any allocated unwritten extents to be converted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3426) * into written extents correctly within the ->end_io() handler, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3427) * need to ensure that the iomap->type is set appropriately. Hence, the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3428) * reason why we need to check whether the EXT4_MAP_UNWRITTEN bit has
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3429) * been set first.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3430) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3431) if (map->m_flags & EXT4_MAP_UNWRITTEN) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3432) iomap->type = IOMAP_UNWRITTEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3433) iomap->addr = (u64) map->m_pblk << blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3434) } else if (map->m_flags & EXT4_MAP_MAPPED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3435) iomap->type = IOMAP_MAPPED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3436) iomap->addr = (u64) map->m_pblk << blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3437) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3438) iomap->type = IOMAP_HOLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3439) iomap->addr = IOMAP_NULL_ADDR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3440) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3441) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3442)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3443) static int ext4_iomap_alloc(struct inode *inode, struct ext4_map_blocks *map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3444) unsigned int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3445) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3446) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3447) u8 blkbits = inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3448) int ret, dio_credits, m_flags = 0, retries = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3449)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3450) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3451) * Trim the mapping request to the maximum value that we can map at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3452) * once for direct I/O.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3453) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3454) if (map->m_len > DIO_MAX_BLOCKS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3455) map->m_len = DIO_MAX_BLOCKS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3456) dio_credits = ext4_chunk_trans_blocks(inode, map->m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3458) retry:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3459) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3460) * Either we allocate blocks and then don't get an unwritten extent, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3461) * in that case we have reserved enough credits. Or, the blocks are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3462) * already allocated and unwritten. In that case, the extent conversion
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3463) * fits into the credits as well.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3464) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3465) handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS, dio_credits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3466) if (IS_ERR(handle))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3467) return PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3468)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3469) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3470) * DAX and direct I/O are the only two operations that are currently
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3471) * supported with IOMAP_WRITE.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3472) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3473) WARN_ON(!IS_DAX(inode) && !(flags & IOMAP_DIRECT));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3474) if (IS_DAX(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3475) m_flags = EXT4_GET_BLOCKS_CREATE_ZERO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3476) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3477) * We use i_size instead of i_disksize here because delalloc writeback
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3478) * can complete at any point during the I/O and subsequently push the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3479) * i_disksize out to i_size. This could be beyond where direct I/O is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3480) * happening and thus expose allocated blocks to direct I/O reads.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3481) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3482) else if (((loff_t)map->m_lblk << blkbits) >= i_size_read(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3483) m_flags = EXT4_GET_BLOCKS_CREATE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3484) else if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3485) m_flags = EXT4_GET_BLOCKS_IO_CREATE_EXT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3486)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3487) ret = ext4_map_blocks(handle, inode, map, m_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3489) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3490) * We cannot fill holes in indirect tree based inodes as that could
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3491) * expose stale data in the case of a crash. Use the magic error code
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3492) * to fallback to buffered I/O.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3493) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3494) if (!m_flags && !ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3495) ret = -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3496)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3497) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3498) if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3499) goto retry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3500)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3501) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3502) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3503)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3504)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3505) static int ext4_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3506) unsigned flags, struct iomap *iomap, struct iomap *srcmap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3507) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3508) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3509) struct ext4_map_blocks map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3510) u8 blkbits = inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3512) if ((offset >> blkbits) > EXT4_MAX_LOGICAL_BLOCK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3513) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3514)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3515) if (WARN_ON_ONCE(ext4_has_inline_data(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3516) return -ERANGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3517)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3518) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3519) * Calculate the first and last logical blocks respectively.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3520) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3521) map.m_lblk = offset >> blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3522) map.m_len = min_t(loff_t, (offset + length - 1) >> blkbits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3523) EXT4_MAX_LOGICAL_BLOCK) - map.m_lblk + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3525) if (flags & IOMAP_WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3526) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3527) * We check here if the blocks are already allocated, then we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3528) * don't need to start a journal txn and we can directly return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3529) * the mapping information. This could boost performance
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3530) * especially in multi-threaded overwrite requests.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3531) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3532) if (offset + length <= i_size_read(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3533) ret = ext4_map_blocks(NULL, inode, &map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3534) if (ret > 0 && (map.m_flags & EXT4_MAP_MAPPED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3535) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3536) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3537) ret = ext4_iomap_alloc(inode, &map, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3538) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3539) ret = ext4_map_blocks(NULL, inode, &map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3540) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3542) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3543) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3544) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3546) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3547) * When inline encryption is enabled, sometimes I/O to an encrypted file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3548) * has to be broken up to guarantee DUN contiguity. Handle this by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3549) * limiting the length of the mapping returned.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3550) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3551) map.m_len = fscrypt_limit_io_blocks(inode, map.m_lblk, map.m_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3552)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3553) ext4_set_iomap(inode, iomap, &map, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3554)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3555) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3556) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3557)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3558) static int ext4_iomap_overwrite_begin(struct inode *inode, loff_t offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3559) loff_t length, unsigned flags, struct iomap *iomap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3560) struct iomap *srcmap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3561) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3562) int ret;
^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) * Even for writes we don't need to allocate blocks, so just pretend
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3566) * we are reading to save overhead of starting a transaction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3567) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3568) flags &= ~IOMAP_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3569) ret = ext4_iomap_begin(inode, offset, length, flags, iomap, srcmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3570) WARN_ON_ONCE(iomap->type != IOMAP_MAPPED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3571) return 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) static int ext4_iomap_end(struct inode *inode, loff_t offset, loff_t length,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3575) ssize_t written, unsigned flags, struct iomap *iomap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3576) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3577) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3578) * Check to see whether an error occurred while writing out the data to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3579) * the allocated blocks. If so, return the magic error code so that we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3580) * fallback to buffered I/O and attempt to complete the remainder of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3581) * the I/O. Any blocks that may have been allocated in preparation for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3582) * the direct I/O will be reused during buffered I/O.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3583) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3584) if (flags & (IOMAP_WRITE | IOMAP_DIRECT) && written == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3585) return -ENOTBLK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3586)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3587) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3588) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3589)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3590) const struct iomap_ops ext4_iomap_ops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3591) .iomap_begin = ext4_iomap_begin,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3592) .iomap_end = ext4_iomap_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3593) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3594)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3595) const struct iomap_ops ext4_iomap_overwrite_ops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3596) .iomap_begin = ext4_iomap_overwrite_begin,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3597) .iomap_end = ext4_iomap_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3598) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3599)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3600) static bool ext4_iomap_is_delalloc(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3601) struct ext4_map_blocks *map)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3602) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3603) struct extent_status es;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3604) ext4_lblk_t offset = 0, end = map->m_lblk + map->m_len - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3605)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3606) ext4_es_find_extent_range(inode, &ext4_es_is_delayed,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3607) map->m_lblk, end, &es);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3608)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3609) if (!es.es_len || es.es_lblk > end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3610) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3611)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3612) if (es.es_lblk > map->m_lblk) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3613) map->m_len = es.es_lblk - map->m_lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3614) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3615) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3616)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3617) offset = map->m_lblk - es.es_lblk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3618) map->m_len = es.es_len - offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3619)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3620) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3621) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3622)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3623) static int ext4_iomap_begin_report(struct inode *inode, loff_t offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3624) loff_t length, unsigned int flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3625) struct iomap *iomap, struct iomap *srcmap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3626) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3627) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3628) bool delalloc = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3629) struct ext4_map_blocks map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3630) u8 blkbits = inode->i_blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3631)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3632) if ((offset >> blkbits) > EXT4_MAX_LOGICAL_BLOCK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3633) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3634)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3635) if (ext4_has_inline_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3636) ret = ext4_inline_data_iomap(inode, iomap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3637) if (ret != -EAGAIN) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3638) if (ret == 0 && offset >= iomap->length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3639) ret = -ENOENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3640) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3641) }
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3645) * Calculate the first and last logical block respectively.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3646) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3647) map.m_lblk = offset >> blkbits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3648) map.m_len = min_t(loff_t, (offset + length - 1) >> blkbits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3649) EXT4_MAX_LOGICAL_BLOCK) - map.m_lblk + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3651) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3652) * Fiemap callers may call for offset beyond s_bitmap_maxbytes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3653) * So handle it here itself instead of querying ext4_map_blocks().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3654) * Since ext4_map_blocks() will warn about it and will return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3655) * -EIO error.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3656) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3657) if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3658) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3659)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3660) if (offset >= sbi->s_bitmap_maxbytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3661) map.m_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3662) goto set_iomap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3663) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3664) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3665)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3666) ret = ext4_map_blocks(NULL, inode, &map, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3667) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3668) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3669) if (ret == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3670) delalloc = ext4_iomap_is_delalloc(inode, &map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3671)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3672) set_iomap:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3673) ext4_set_iomap(inode, iomap, &map, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3674) if (delalloc && iomap->type == IOMAP_HOLE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3675) iomap->type = IOMAP_DELALLOC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3676)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3677) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3678) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3679)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3680) const struct iomap_ops ext4_iomap_report_ops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3681) .iomap_begin = ext4_iomap_begin_report,
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3685) * Pages can be marked dirty completely asynchronously from ext4's journalling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3686) * activity. By filemap_sync_pte(), try_to_unmap_one(), etc. We cannot do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3687) * much here because ->set_page_dirty is called under VFS locks. The page is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3688) * not necessarily locked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3689) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3690) * We cannot just dirty the page and leave attached buffers clean, because the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3691) * buffers' dirty state is "definitive". We cannot just set the buffers dirty
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3692) * or jbddirty because all the journalling code will explode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3693) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3694) * So what we do is to mark the page "pending dirty" and next time writepage
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3695) * is called, propagate that into the buffers appropriately.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3696) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3697) static int ext4_journalled_set_page_dirty(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3698) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3699) SetPageChecked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3700) return __set_page_dirty_nobuffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3701) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3702)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3703) static int ext4_set_page_dirty(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3704) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3705) WARN_ON_ONCE(!PageLocked(page) && !PageDirty(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3706) WARN_ON_ONCE(!page_has_buffers(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3707) return __set_page_dirty_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3708) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3709)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3710) static int ext4_iomap_swap_activate(struct swap_info_struct *sis,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3711) struct file *file, sector_t *span)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3712) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3713) return iomap_swapfile_activate(sis, file, span,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3714) &ext4_iomap_report_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3715) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3716)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3717) static const struct address_space_operations ext4_aops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3718) .readpage = ext4_readpage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3719) .readahead = ext4_readahead,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3720) .writepage = ext4_writepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3721) .writepages = ext4_writepages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3722) .write_begin = ext4_write_begin,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3723) .write_end = ext4_write_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3724) .set_page_dirty = ext4_set_page_dirty,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3725) .bmap = ext4_bmap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3726) .invalidatepage = ext4_invalidatepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3727) .releasepage = ext4_releasepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3728) .direct_IO = noop_direct_IO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3729) .migratepage = buffer_migrate_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3730) .is_partially_uptodate = block_is_partially_uptodate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3731) .error_remove_page = generic_error_remove_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3732) .swap_activate = ext4_iomap_swap_activate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3733) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3735) static const struct address_space_operations ext4_journalled_aops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3736) .readpage = ext4_readpage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3737) .readahead = ext4_readahead,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3738) .writepage = ext4_writepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3739) .writepages = ext4_writepages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3740) .write_begin = ext4_write_begin,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3741) .write_end = ext4_journalled_write_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3742) .set_page_dirty = ext4_journalled_set_page_dirty,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3743) .bmap = ext4_bmap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3744) .invalidatepage = ext4_journalled_invalidatepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3745) .releasepage = ext4_releasepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3746) .direct_IO = noop_direct_IO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3747) .is_partially_uptodate = block_is_partially_uptodate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3748) .error_remove_page = generic_error_remove_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3749) .swap_activate = ext4_iomap_swap_activate,
^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 const struct address_space_operations ext4_da_aops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3753) .readpage = ext4_readpage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3754) .readahead = ext4_readahead,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3755) .writepage = ext4_writepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3756) .writepages = ext4_writepages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3757) .write_begin = ext4_da_write_begin,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3758) .write_end = ext4_da_write_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3759) .set_page_dirty = ext4_set_page_dirty,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3760) .bmap = ext4_bmap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3761) .invalidatepage = ext4_invalidatepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3762) .releasepage = ext4_releasepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3763) .direct_IO = noop_direct_IO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3764) .migratepage = buffer_migrate_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3765) .is_partially_uptodate = block_is_partially_uptodate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3766) .error_remove_page = generic_error_remove_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3767) .swap_activate = ext4_iomap_swap_activate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3768) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3769)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3770) static const struct address_space_operations ext4_dax_aops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3771) .writepages = ext4_dax_writepages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3772) .direct_IO = noop_direct_IO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3773) .set_page_dirty = noop_set_page_dirty,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3774) .bmap = ext4_bmap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3775) .invalidatepage = noop_invalidatepage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3776) .swap_activate = ext4_iomap_swap_activate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3777) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3778)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3779) void ext4_set_aops(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3780) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3781) switch (ext4_inode_journal_mode(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3782) case EXT4_INODE_ORDERED_DATA_MODE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3783) case EXT4_INODE_WRITEBACK_DATA_MODE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3784) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3785) case EXT4_INODE_JOURNAL_DATA_MODE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3786) inode->i_mapping->a_ops = &ext4_journalled_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3787) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3788) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3789) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3790) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3791) if (IS_DAX(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3792) inode->i_mapping->a_ops = &ext4_dax_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3793) else if (test_opt(inode->i_sb, DELALLOC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3794) inode->i_mapping->a_ops = &ext4_da_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3795) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3796) inode->i_mapping->a_ops = &ext4_aops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3797) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3798)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3799) static int __ext4_block_zero_page_range(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3800) struct address_space *mapping, loff_t from, loff_t length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3801) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3802) ext4_fsblk_t index = from >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3803) unsigned offset = from & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3804) unsigned blocksize, pos;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3805) ext4_lblk_t iblock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3806) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3807) struct buffer_head *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3808) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3809) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3810)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3811) page = find_or_create_page(mapping, from >> PAGE_SHIFT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3812) mapping_gfp_constraint(mapping, ~__GFP_FS));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3813) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3814) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3815)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3816) blocksize = inode->i_sb->s_blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3817)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3818) iblock = index << (PAGE_SHIFT - inode->i_sb->s_blocksize_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3819)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3820) if (!page_has_buffers(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3821) create_empty_buffers(page, blocksize, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3822)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3823) /* Find the buffer that contains "offset" */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3824) bh = page_buffers(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3825) pos = blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3826) while (offset >= pos) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3827) bh = bh->b_this_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3828) iblock++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3829) pos += blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3830) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3831) if (buffer_freed(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3832) BUFFER_TRACE(bh, "freed: skip");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3833) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3834) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3835) if (!buffer_mapped(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3836) BUFFER_TRACE(bh, "unmapped");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3837) ext4_get_block(inode, iblock, bh, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3838) /* unmapped? It's a hole - nothing to do */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3839) if (!buffer_mapped(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3840) BUFFER_TRACE(bh, "still unmapped");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3841) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3842) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3843) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3844)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3845) /* Ok, it's mapped. Make sure it's up-to-date */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3846) if (PageUptodate(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3847) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3849) if (!buffer_uptodate(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3850) err = ext4_read_bh_lock(bh, 0, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3851) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3852) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3853) if (fscrypt_inode_uses_fs_layer_crypto(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3854) /* We expect the key to be set. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3855) BUG_ON(!fscrypt_has_encryption_key(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3856) err = fscrypt_decrypt_pagecache_blocks(page, blocksize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3857) bh_offset(bh));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3858) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3859) clear_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3860) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3861) }
^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) if (ext4_should_journal_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3865) BUFFER_TRACE(bh, "get write access");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3866) err = ext4_journal_get_write_access(handle, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3867) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3868) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3869) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3870) zero_user(page, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3871) BUFFER_TRACE(bh, "zeroed end of block");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3872)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3873) if (ext4_should_journal_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3874) err = ext4_handle_dirty_metadata(handle, inode, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3875) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3876) err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3877) mark_buffer_dirty(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3878) if (ext4_should_order_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3879) err = ext4_jbd2_inode_add_write(handle, inode, from,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3880) length);
^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) unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3884) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3885) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3886) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3887) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3888)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3889) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3890) * ext4_block_zero_page_range() zeros out a mapping of length 'length'
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3891) * starting from file offset 'from'. The range to be zero'd must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3892) * be contained with in one block. If the specified range exceeds
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3893) * the end of the block it will be shortened to end of the block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3894) * that cooresponds to 'from'
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3895) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3896) static int ext4_block_zero_page_range(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3897) struct address_space *mapping, loff_t from, loff_t length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3898) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3899) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3900) unsigned offset = from & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3901) unsigned blocksize = inode->i_sb->s_blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3902) unsigned max = blocksize - (offset & (blocksize - 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3903)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3904) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3905) * correct length if it does not fall between
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3906) * 'from' and the end of the block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3907) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3908) if (length > max || length < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3909) length = max;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3910)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3911) if (IS_DAX(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3912) return iomap_zero_range(inode, from, length, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3913) &ext4_iomap_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3914) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3915) return __ext4_block_zero_page_range(handle, mapping, from, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3916) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3918) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3919) * ext4_block_truncate_page() zeroes out a mapping from file offset `from'
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3920) * up to the end of the block which corresponds to `from'.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3921) * This required during truncate. We need to physically zero the tail end
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3922) * of that block so it doesn't yield old data if the file is later grown.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3923) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3924) static int ext4_block_truncate_page(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3925) struct address_space *mapping, loff_t from)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3926) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3927) unsigned offset = from & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3928) unsigned length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3929) unsigned blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3930) struct inode *inode = mapping->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3932) /* If we are processing an encrypted inode during orphan list handling */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3933) if (IS_ENCRYPTED(inode) && !fscrypt_has_encryption_key(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3934) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3935)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3936) blocksize = inode->i_sb->s_blocksize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3937) length = blocksize - (offset & (blocksize - 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3938)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3939) return ext4_block_zero_page_range(handle, mapping, from, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3940) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3941)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3942) int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3943) loff_t lstart, loff_t length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3944) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3945) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3946) struct address_space *mapping = inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3947) unsigned partial_start, partial_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3948) ext4_fsblk_t start, end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3949) loff_t byte_end = (lstart + length - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3950) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3951)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3952) partial_start = lstart & (sb->s_blocksize - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3953) partial_end = byte_end & (sb->s_blocksize - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3954)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3955) start = lstart >> sb->s_blocksize_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3956) end = byte_end >> sb->s_blocksize_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3957)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3958) /* Handle partial zero within the single block */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3959) if (start == end &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3960) (partial_start || (partial_end != sb->s_blocksize - 1))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3961) err = ext4_block_zero_page_range(handle, mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3962) lstart, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3963) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3964) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3965) /* Handle partial zero out on the start of the range */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3966) if (partial_start) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3967) err = ext4_block_zero_page_range(handle, mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3968) lstart, sb->s_blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3969) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3970) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3971) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3972) /* Handle partial zero out on the end of the range */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3973) if (partial_end != sb->s_blocksize - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3974) err = ext4_block_zero_page_range(handle, mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3975) byte_end - partial_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3976) partial_end + 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3977) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3978) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3979)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3980) int ext4_can_truncate(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3981) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3982) if (S_ISREG(inode->i_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3983) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3984) if (S_ISDIR(inode->i_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3985) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3986) if (S_ISLNK(inode->i_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3987) return !ext4_inode_is_fast_symlink(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3988) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3989) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3990)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3991) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3992) * We have to make sure i_disksize gets properly updated before we truncate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3993) * page cache due to hole punching or zero range. Otherwise i_disksize update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3994) * can get lost as it may have been postponed to submission of writeback but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3995) * that will never happen after we truncate page cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3996) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3997) int ext4_update_disksize_before_punch(struct inode *inode, loff_t offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3998) loff_t len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3999) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4000) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4001) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4002)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4003) loff_t size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4004)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4005) WARN_ON(!inode_is_locked(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4006) if (offset > size || offset + len < size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4007) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4008)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4009) if (EXT4_I(inode)->i_disksize >= size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4010) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4011)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4012) handle = ext4_journal_start(inode, EXT4_HT_MISC, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4013) if (IS_ERR(handle))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4014) return PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4015) ext4_update_i_disksize(inode, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4016) ret = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4017) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4018)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4019) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4020) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4022) static void ext4_wait_dax_page(struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4023) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4024) up_write(&ei->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4025) schedule();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4026) down_write(&ei->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4027) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4028)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4029) int ext4_break_layouts(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4030) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4031) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4032) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4033) int error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4034)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4035) if (WARN_ON_ONCE(!rwsem_is_locked(&ei->i_mmap_sem)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4036) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4038) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4039) page = dax_layout_busy_page(inode->i_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4040) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4041) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4042)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4043) error = ___wait_var_event(&page->_refcount,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4044) atomic_read(&page->_refcount) == 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4045) TASK_INTERRUPTIBLE, 0, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4046) ext4_wait_dax_page(ei));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4047) } while (error == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4048)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4049) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4050) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4051)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4052) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4053) * ext4_punch_hole: punches a hole in a file by releasing the blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4054) * associated with the given offset and length
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4055) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4056) * @inode: File inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4057) * @offset: The offset where the hole will begin
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4058) * @len: The length of the hole
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4059) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4060) * Returns: 0 on success or negative on failure
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4061) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4062)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4063) int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4064) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4065) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4066) ext4_lblk_t first_block, stop_block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4067) struct address_space *mapping = inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4068) loff_t first_block_offset, last_block_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4069) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4070) unsigned int credits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4071) int ret = 0, ret2 = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4072)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4073) trace_ext4_punch_hole(inode, offset, length, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4074)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4075) ext4_clear_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4076) if (ext4_has_inline_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4077) down_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4078) ret = ext4_convert_inline_data(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4079) up_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4080) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4081) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4082) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4084) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4085) * Write out all dirty pages to avoid race conditions
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4086) * Then release them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4087) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4088) if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4089) ret = filemap_write_and_wait_range(mapping, offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4090) offset + length - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4091) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4092) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4093) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4094)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4095) inode_lock(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4097) /* No need to punch hole beyond i_size */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4098) if (offset >= inode->i_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4099) goto out_mutex;
^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) * If the hole extends beyond i_size, set the hole
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4103) * to end after the page that contains i_size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4104) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4105) if (offset + length > inode->i_size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4106) length = inode->i_size +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4107) PAGE_SIZE - (inode->i_size & (PAGE_SIZE - 1)) -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4108) offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4109) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4110)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4111) if (offset & (sb->s_blocksize - 1) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4112) (offset + length) & (sb->s_blocksize - 1)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4113) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4114) * Attach jinode to inode for jbd2 if we do any zeroing of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4115) * partial block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4116) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4117) ret = ext4_inode_attach_jinode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4118) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4119) goto out_mutex;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4120)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4121) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4122)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4123) /* Wait all existing dio workers, newcomers will block on i_mutex */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4124) inode_dio_wait(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4125)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4126) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4127) * Prevent page faults from reinstantiating pages we have released from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4128) * page cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4129) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4130) down_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4132) ret = ext4_break_layouts(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4133) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4134) goto out_dio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4135)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4136) first_block_offset = round_up(offset, sb->s_blocksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4137) last_block_offset = round_down((offset + length), sb->s_blocksize) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4139) /* Now release the pages and zero block aligned part of pages*/
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4140) if (last_block_offset > first_block_offset) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4141) ret = ext4_update_disksize_before_punch(inode, offset, length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4142) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4143) goto out_dio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4144) truncate_pagecache_range(inode, first_block_offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4145) last_block_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4146) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4147)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4148) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4149) credits = ext4_writepage_trans_blocks(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4150) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4151) credits = ext4_blocks_for_truncate(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4152) handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4153) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4154) ret = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4155) ext4_std_error(sb, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4156) goto out_dio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4157) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4158)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4159) ret = ext4_zero_partial_blocks(handle, inode, offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4160) length);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4161) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4162) goto out_stop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4164) first_block = (offset + sb->s_blocksize - 1) >>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4165) EXT4_BLOCK_SIZE_BITS(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4166) stop_block = (offset + length) >> EXT4_BLOCK_SIZE_BITS(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4168) /* If there are blocks to remove, do it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4169) if (stop_block > first_block) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4170)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4171) down_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4172) ext4_discard_preallocations(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4173)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4174) ret = ext4_es_remove_extent(inode, first_block,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4175) stop_block - first_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4176) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4177) up_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4178) goto out_stop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4179) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4181) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4182) ret = ext4_ext_remove_space(inode, first_block,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4183) stop_block - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4184) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4185) ret = ext4_ind_remove_space(handle, inode, first_block,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4186) stop_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4187)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4188) up_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4189) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4190) ext4_fc_track_range(handle, inode, first_block, stop_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4191) if (IS_SYNC(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4192) ext4_handle_sync(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4194) inode->i_mtime = inode->i_ctime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4195) ret2 = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4196) if (unlikely(ret2))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4197) ret = ret2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4198) if (ret >= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4199) ext4_update_inode_fsync_trans(handle, inode, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4200) out_stop:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4201) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4202) out_dio:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4203) up_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4204) out_mutex:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4205) inode_unlock(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4206) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4207) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4209) int ext4_inode_attach_jinode(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4210) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4211) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4212) struct jbd2_inode *jinode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4213)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4214) if (ei->jinode || !EXT4_SB(inode->i_sb)->s_journal)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4215) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4216)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4217) jinode = jbd2_alloc_inode(GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4218) spin_lock(&inode->i_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4219) if (!ei->jinode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4220) if (!jinode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4221) spin_unlock(&inode->i_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4222) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4223) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4224) ei->jinode = jinode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4225) jbd2_journal_init_jbd_inode(ei->jinode, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4226) jinode = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4227) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4228) spin_unlock(&inode->i_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4229) if (unlikely(jinode != NULL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4230) jbd2_free_inode(jinode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4231) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4232) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4233)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4234) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4235) * ext4_truncate()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4236) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4237) * We block out ext4_get_block() block instantiations across the entire
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4238) * transaction, and VFS/VM ensures that ext4_truncate() cannot run
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4239) * simultaneously on behalf of the same inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4240) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4241) * As we work through the truncate and commit bits of it to the journal there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4242) * is one core, guiding principle: the file's tree must always be consistent on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4243) * disk. We must be able to restart the truncate after a crash.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4244) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4245) * The file's tree may be transiently inconsistent in memory (although it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4246) * probably isn't), but whenever we close off and commit a journal transaction,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4247) * the contents of (the filesystem + the journal) must be consistent and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4248) * restartable. It's pretty simple, really: bottom up, right to left (although
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4249) * left-to-right works OK too).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4250) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4251) * Note that at recovery time, journal replay occurs *before* the restart of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4252) * truncate against the orphan inode list.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4253) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4254) * The committed inode has the new, desired i_size (which is the same as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4255) * i_disksize in this case). After a crash, ext4_orphan_cleanup() will see
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4256) * that this inode's truncate did not complete and it will again call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4257) * ext4_truncate() to have another go. So there will be instantiated blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4258) * to the right of the truncation point in a crashed ext4 filesystem. But
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4259) * that's fine - as long as they are linked from the inode, the post-crash
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4260) * ext4_truncate() run will find them and release them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4261) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4262) int ext4_truncate(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4263) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4264) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4265) unsigned int credits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4266) int err = 0, err2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4267) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4268) struct address_space *mapping = inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4269)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4270) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4271) * There is a possibility that we're either freeing the inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4272) * or it's a completely new inode. In those cases we might not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4273) * have i_mutex locked because it's not necessary.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4274) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4275) if (!(inode->i_state & (I_NEW|I_FREEING)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4276) WARN_ON(!inode_is_locked(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4277) trace_ext4_truncate_enter(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4278)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4279) if (!ext4_can_truncate(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4280) goto out_trace;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4282) if (inode->i_size == 0 && !test_opt(inode->i_sb, NO_AUTO_DA_ALLOC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4283) ext4_set_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4284)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4285) if (ext4_has_inline_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4286) int has_inline = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4287)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4288) err = ext4_inline_data_truncate(inode, &has_inline);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4289) if (err || has_inline)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4290) goto out_trace;
^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) /* If we zero-out tail of the page, we have to create jinode for jbd2 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4294) if (inode->i_size & (inode->i_sb->s_blocksize - 1)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4295) if (ext4_inode_attach_jinode(inode) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4296) goto out_trace;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4297) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4298)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4299) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4300) credits = ext4_writepage_trans_blocks(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4301) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4302) credits = ext4_blocks_for_truncate(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4303)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4304) handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4305) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4306) err = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4307) goto out_trace;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4308) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4309)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4310) if (inode->i_size & (inode->i_sb->s_blocksize - 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4311) ext4_block_truncate_page(handle, mapping, inode->i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4312)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4313) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4314) * We add the inode to the orphan list, so that if this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4315) * truncate spans multiple transactions, and we crash, we will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4316) * resume the truncate when the filesystem recovers. It also
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4317) * marks the inode dirty, to catch the new size.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4318) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4319) * Implication: the file must always be in a sane, consistent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4320) * truncatable state while each transaction commits.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4321) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4322) err = ext4_orphan_add(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4323) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4324) goto out_stop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4325)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4326) down_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4327)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4328) ext4_discard_preallocations(inode, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4330) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4331) err = ext4_ext_truncate(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4332) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4333) ext4_ind_truncate(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4335) up_write(&ei->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4336) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4337) goto out_stop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4338)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4339) if (IS_SYNC(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4340) ext4_handle_sync(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4341)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4342) out_stop:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4343) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4344) * If this was a simple ftruncate() and the file will remain alive,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4345) * then we need to clear up the orphan record which we created above.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4346) * However, if this was a real unlink then we were called by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4347) * ext4_evict_inode(), and we allow that function to clean up the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4348) * orphan info for us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4349) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4350) if (inode->i_nlink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4351) ext4_orphan_del(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4352)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4353) inode->i_mtime = inode->i_ctime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4354) err2 = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4355) if (unlikely(err2 && !err))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4356) err = err2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4357) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4358)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4359) out_trace:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4360) trace_ext4_truncate_exit(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4361) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4362) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4364) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4365) * ext4_get_inode_loc returns with an extra refcount against the inode's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4366) * underlying buffer_head on success. If 'in_mem' is true, we have all
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4367) * data in memory that is needed to recreate the on-disk version of this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4368) * inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4369) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4370) static int __ext4_get_inode_loc(struct super_block *sb, unsigned long ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4371) struct ext4_iloc *iloc, int in_mem,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4372) ext4_fsblk_t *ret_block)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4373) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4374) struct ext4_group_desc *gdp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4375) struct buffer_head *bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4376) ext4_fsblk_t block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4377) struct blk_plug plug;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4378) int inodes_per_block, inode_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4380) iloc->bh = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4381) if (ino < EXT4_ROOT_INO ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4382) ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4383) return -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4384)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4385) iloc->block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4386) gdp = ext4_get_group_desc(sb, iloc->block_group, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4387) if (!gdp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4388) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4389)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4390) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4391) * Figure out the offset within the block group inode table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4392) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4393) inodes_per_block = EXT4_SB(sb)->s_inodes_per_block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4394) inode_offset = ((ino - 1) %
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4395) EXT4_INODES_PER_GROUP(sb));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4396) block = ext4_inode_table(sb, gdp) + (inode_offset / inodes_per_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4397) iloc->offset = (inode_offset % inodes_per_block) * EXT4_INODE_SIZE(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4399) bh = sb_getblk(sb, block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4400) if (unlikely(!bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4401) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4402) if (ext4_simulate_fail(sb, EXT4_SIM_INODE_EIO))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4403) goto simulate_eio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4404) if (!buffer_uptodate(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4405) lock_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4406)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4407) if (ext4_buffer_uptodate(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4408) /* someone brought it uptodate while we waited */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4409) unlock_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4410) goto has_buffer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4411) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4412)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4413) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4414) * If we have all information of the inode in memory and this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4415) * is the only valid inode in the block, we need not read the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4416) * block.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4417) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4418) if (in_mem) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4419) struct buffer_head *bitmap_bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4420) int i, start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4421)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4422) start = inode_offset & ~(inodes_per_block - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4423)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4424) /* Is the inode bitmap in cache? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4425) bitmap_bh = sb_getblk(sb, ext4_inode_bitmap(sb, gdp));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4426) if (unlikely(!bitmap_bh))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4427) goto make_io;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4428)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4429) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4430) * If the inode bitmap isn't in cache then the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4431) * optimisation may end up performing two reads instead
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4432) * of one, so skip it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4433) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4434) if (!buffer_uptodate(bitmap_bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4435) brelse(bitmap_bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4436) goto make_io;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4437) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4438) for (i = start; i < start + inodes_per_block; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4439) if (i == inode_offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4440) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4441) if (ext4_test_bit(i, bitmap_bh->b_data))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4442) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4443) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4444) brelse(bitmap_bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4445) if (i == start + inodes_per_block) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4446) /* all other inodes are free, so skip I/O */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4447) memset(bh->b_data, 0, bh->b_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4448) set_buffer_uptodate(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4449) unlock_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4450) goto has_buffer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4451) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4452) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4454) make_io:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4455) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4456) * If we need to do any I/O, try to pre-readahead extra
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4457) * blocks from the inode table.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4458) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4459) blk_start_plug(&plug);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4460) if (EXT4_SB(sb)->s_inode_readahead_blks) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4461) ext4_fsblk_t b, end, table;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4462) unsigned num;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4463) __u32 ra_blks = EXT4_SB(sb)->s_inode_readahead_blks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4464)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4465) table = ext4_inode_table(sb, gdp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4466) /* s_inode_readahead_blks is always a power of 2 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4467) b = block & ~((ext4_fsblk_t) ra_blks - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4468) if (table > b)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4469) b = table;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4470) end = b + ra_blks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4471) num = EXT4_INODES_PER_GROUP(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4472) if (ext4_has_group_desc_csum(sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4473) num -= ext4_itable_unused_count(sb, gdp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4474) table += num / inodes_per_block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4475) if (end > table)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4476) end = table;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4477) while (b <= end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4478) ext4_sb_breadahead_unmovable(sb, b++);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4479) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4480)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4481) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4482) * There are other valid inodes in the buffer, this inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4483) * has in-inode xattrs, or we don't have this inode in memory.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4484) * Read the block from disk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4485) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4486) trace_ext4_load_inode(sb, ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4487) ext4_read_bh_nowait(bh, REQ_META | REQ_PRIO, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4488) blk_finish_plug(&plug);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4489) wait_on_buffer(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4490) if (!buffer_uptodate(bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4491) simulate_eio:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4492) if (ret_block)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4493) *ret_block = block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4494) brelse(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4495) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4496) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4497) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4498) has_buffer:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4499) iloc->bh = bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4500) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4501) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4502)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4503) static int __ext4_get_inode_loc_noinmem(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4504) struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4505) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4506) ext4_fsblk_t err_blk = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4507) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4508)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4509) ret = __ext4_get_inode_loc(inode->i_sb, inode->i_ino, iloc, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4510) &err_blk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4512) if (ret == -EIO)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4513) ext4_error_inode_block(inode, err_blk, EIO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4514) "unable to read itable block");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4515)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4516) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4517) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4518)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4519) int ext4_get_inode_loc(struct inode *inode, struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4520) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4521) ext4_fsblk_t err_blk = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4522) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4524) /* We have all inode data except xattrs in memory here. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4525) ret = __ext4_get_inode_loc(inode->i_sb, inode->i_ino, iloc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4526) !ext4_test_inode_state(inode, EXT4_STATE_XATTR), &err_blk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4527)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4528) if (ret == -EIO)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4529) ext4_error_inode_block(inode, err_blk, EIO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4530) "unable to read itable block");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4531)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4532) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4533) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4534)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4536) int ext4_get_fc_inode_loc(struct super_block *sb, unsigned long ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4537) struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4538) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4539) return __ext4_get_inode_loc(sb, ino, iloc, 0, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4540) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4542) static bool ext4_should_enable_dax(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4543) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4544) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4546) if (test_opt2(inode->i_sb, DAX_NEVER))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4547) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4548) if (!S_ISREG(inode->i_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4549) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4550) if (ext4_should_journal_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4551) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4552) if (ext4_has_inline_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4553) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4554) if (ext4_test_inode_flag(inode, EXT4_INODE_ENCRYPT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4555) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4556) if (ext4_test_inode_flag(inode, EXT4_INODE_VERITY))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4557) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4558) if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4559) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4560) if (test_opt(inode->i_sb, DAX_ALWAYS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4561) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4562)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4563) return ext4_test_inode_flag(inode, EXT4_INODE_DAX);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4564) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4565)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4566) void ext4_set_inode_flags(struct inode *inode, bool init)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4567) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4568) unsigned int flags = EXT4_I(inode)->i_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4569) unsigned int new_fl = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4570)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4571) WARN_ON_ONCE(IS_DAX(inode) && init);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4572)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4573) if (flags & EXT4_SYNC_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4574) new_fl |= S_SYNC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4575) if (flags & EXT4_APPEND_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4576) new_fl |= S_APPEND;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4577) if (flags & EXT4_IMMUTABLE_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4578) new_fl |= S_IMMUTABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4579) if (flags & EXT4_NOATIME_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4580) new_fl |= S_NOATIME;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4581) if (flags & EXT4_DIRSYNC_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4582) new_fl |= S_DIRSYNC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4583)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4584) /* Because of the way inode_set_flags() works we must preserve S_DAX
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4585) * here if already set. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4586) new_fl |= (inode->i_flags & S_DAX);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4587) if (init && ext4_should_enable_dax(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4588) new_fl |= S_DAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4589)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4590) if (flags & EXT4_ENCRYPT_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4591) new_fl |= S_ENCRYPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4592) if (flags & EXT4_CASEFOLD_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4593) new_fl |= S_CASEFOLD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4594) if (flags & EXT4_VERITY_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4595) new_fl |= S_VERITY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4596) inode_set_flags(inode, new_fl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4597) S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|S_DAX|
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4598) S_ENCRYPTED|S_CASEFOLD|S_VERITY);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4599) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4600)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4601) static blkcnt_t ext4_inode_blocks(struct ext4_inode *raw_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4602) struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4603) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4604) blkcnt_t i_blocks ;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4605) struct inode *inode = &(ei->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4606) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4607)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4608) if (ext4_has_feature_huge_file(sb)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4609) /* we are using combined 48 bit field */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4610) i_blocks = ((u64)le16_to_cpu(raw_inode->i_blocks_high)) << 32 |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4611) le32_to_cpu(raw_inode->i_blocks_lo);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4612) if (ext4_test_inode_flag(inode, EXT4_INODE_HUGE_FILE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4613) /* i_blocks represent file system block size */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4614) return i_blocks << (inode->i_blkbits - 9);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4615) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4616) return i_blocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4617) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4618) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4619) return le32_to_cpu(raw_inode->i_blocks_lo);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4620) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4621) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4622)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4623) static inline int ext4_iget_extra_inode(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4624) struct ext4_inode *raw_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4625) struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4626) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4627) __le32 *magic = (void *)raw_inode +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4628) EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4629)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4630) if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize + sizeof(__le32) <=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4631) EXT4_INODE_SIZE(inode->i_sb) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4632) *magic == cpu_to_le32(EXT4_XATTR_MAGIC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4633) ext4_set_inode_state(inode, EXT4_STATE_XATTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4634) return ext4_find_inline_data_nolock(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4635) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4636) EXT4_I(inode)->i_inline_off = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4637) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4638) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4639)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4640) int ext4_get_projid(struct inode *inode, kprojid_t *projid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4641) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4642) if (!ext4_has_feature_project(inode->i_sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4643) return -EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4644) *projid = EXT4_I(inode)->i_projid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4645) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4646) }
^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) * ext4 has self-managed i_version for ea inodes, it stores the lower 32bit of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4650) * refcount in i_version, so use raw values if inode has EXT4_EA_INODE_FL flag
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4651) * set.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4652) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4653) static inline void ext4_inode_set_iversion_queried(struct inode *inode, u64 val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4654) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4655) if (unlikely(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4656) inode_set_iversion_raw(inode, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4657) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4658) inode_set_iversion_queried(inode, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4659) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4660) static inline u64 ext4_inode_peek_iversion(const struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4661) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4662) if (unlikely(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4663) return inode_peek_iversion_raw(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4664) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4665) return inode_peek_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4666) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4667)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4668) struct inode *__ext4_iget(struct super_block *sb, unsigned long ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4669) ext4_iget_flags flags, const char *function,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4670) unsigned int line)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4671) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4672) struct ext4_iloc iloc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4673) struct ext4_inode *raw_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4674) struct ext4_inode_info *ei;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4675) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4676) journal_t *journal = EXT4_SB(sb)->s_journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4677) long ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4678) loff_t size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4679) int block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4680) uid_t i_uid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4681) gid_t i_gid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4682) projid_t i_projid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4683)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4684) if ((!(flags & EXT4_IGET_SPECIAL) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4685) (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4686) (ino < EXT4_ROOT_INO) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4687) (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4688) if (flags & EXT4_IGET_HANDLE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4689) return ERR_PTR(-ESTALE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4690) __ext4_error(sb, function, line, EFSCORRUPTED, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4691) "inode #%lu: comm %s: iget: illegal inode #",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4692) ino, current->comm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4693) return ERR_PTR(-EFSCORRUPTED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4694) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4695)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4696) inode = iget_locked(sb, ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4697) if (!inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4698) return ERR_PTR(-ENOMEM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4699) if (!(inode->i_state & I_NEW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4700) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4702) ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4703) iloc.bh = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4704)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4705) ret = __ext4_get_inode_loc_noinmem(inode, &iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4706) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4707) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4708) raw_inode = ext4_raw_inode(&iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4709)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4710) if ((ino == EXT4_ROOT_INO) && (raw_inode->i_links_count == 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4711) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4712) "iget: root inode unallocated");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4713) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4714) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4715) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4716)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4717) if ((flags & EXT4_IGET_HANDLE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4718) (raw_inode->i_links_count == 0) && (raw_inode->i_mode == 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4719) ret = -ESTALE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4720) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4721) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4722)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4723) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4724) ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4725) if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize >
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4726) EXT4_INODE_SIZE(inode->i_sb) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4727) (ei->i_extra_isize & 3)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4728) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4729) "iget: bad extra_isize %u "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4730) "(inode size %u)",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4731) ei->i_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4732) EXT4_INODE_SIZE(inode->i_sb));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4733) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4734) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4735) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4736) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4737) ei->i_extra_isize = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4738)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4739) /* Precompute checksum seed for inode metadata */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4740) if (ext4_has_metadata_csum(sb)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4741) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4742) __u32 csum;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4743) __le32 inum = cpu_to_le32(inode->i_ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4744) __le32 gen = raw_inode->i_generation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4745) csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4746) sizeof(inum));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4747) ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4748) sizeof(gen));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4749) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4750)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4751) if ((!ext4_inode_csum_verify(inode, raw_inode, ei) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4752) ext4_simulate_fail(sb, EXT4_SIM_INODE_CRC)) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4753) (!(EXT4_SB(sb)->s_mount_state & EXT4_FC_REPLAY))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4754) ext4_error_inode_err(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4755) EFSBADCRC, "iget: checksum invalid");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4756) ret = -EFSBADCRC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4757) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4758) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4759)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4760) inode->i_mode = le16_to_cpu(raw_inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4761) i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4762) i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4763) if (ext4_has_feature_project(sb) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4764) EXT4_INODE_SIZE(sb) > EXT4_GOOD_OLD_INODE_SIZE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4765) EXT4_FITS_IN_INODE(raw_inode, ei, i_projid))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4766) i_projid = (projid_t)le32_to_cpu(raw_inode->i_projid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4767) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4768) i_projid = EXT4_DEF_PROJID;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4769)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4770) if (!(test_opt(inode->i_sb, NO_UID32))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4771) i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4772) i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4773) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4774) i_uid_write(inode, i_uid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4775) i_gid_write(inode, i_gid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4776) ei->i_projid = make_kprojid(&init_user_ns, i_projid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4777) set_nlink(inode, le16_to_cpu(raw_inode->i_links_count));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4778)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4779) ext4_clear_state_flags(ei); /* Only relevant on 32-bit archs */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4780) ei->i_inline_off = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4781) ei->i_dir_start_lookup = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4782) ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4783) /* We now have enough fields to check if the inode was active or not.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4784) * This is needed because nfsd might try to access dead inodes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4785) * the test is that same one that e2fsck uses
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4786) * NeilBrown 1999oct15
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4787) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4788) if (inode->i_nlink == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4789) if ((inode->i_mode == 0 ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4790) !(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4791) ino != EXT4_BOOT_LOADER_INO) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4792) /* this inode is deleted */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4793) ret = -ESTALE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4794) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4795) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4796) /* The only unlinked inodes we let through here have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4797) * valid i_mode and are being read by the orphan
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4798) * recovery code: that's fine, we're about to complete
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4799) * the process of deleting those.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4800) * OR it is the EXT4_BOOT_LOADER_INO which is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4801) * not initialized on a new filesystem. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4802) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4803) ei->i_flags = le32_to_cpu(raw_inode->i_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4804) ext4_set_inode_flags(inode, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4805) inode->i_blocks = ext4_inode_blocks(raw_inode, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4806) ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl_lo);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4807) if (ext4_has_feature_64bit(sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4808) ei->i_file_acl |=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4809) ((__u64)le16_to_cpu(raw_inode->i_file_acl_high)) << 32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4810) inode->i_size = ext4_isize(sb, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4811) if ((size = i_size_read(inode)) < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4812) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4813) "iget: bad i_size value: %lld", size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4814) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4815) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4816) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4817) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4818) * If dir_index is not enabled but there's dir with INDEX flag set,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4819) * we'd normally treat htree data as empty space. But with metadata
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4820) * checksumming that corrupts checksums so forbid that.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4821) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4822) if (!ext4_has_feature_dir_index(sb) && ext4_has_metadata_csum(sb) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4823) ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4824) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4825) "iget: Dir with htree data on filesystem without dir_index feature.");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4826) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4827) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4828) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4829) ei->i_disksize = inode->i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4830) #ifdef CONFIG_QUOTA
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4831) ei->i_reserved_quota = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4832) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4833) inode->i_generation = le32_to_cpu(raw_inode->i_generation);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4834) ei->i_block_group = iloc.block_group;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4835) ei->i_last_alloc_group = ~0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4836) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4837) * NOTE! The in-memory inode i_data array is in little-endian order
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4838) * even on big-endian machines: we do NOT byteswap the block numbers!
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4839) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4840) for (block = 0; block < EXT4_N_BLOCKS; block++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4841) ei->i_data[block] = raw_inode->i_block[block];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4842) INIT_LIST_HEAD(&ei->i_orphan);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4843) ext4_fc_init_inode(&ei->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4844)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4845) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4846) * Set transaction id's of transactions that have to be committed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4847) * to finish f[data]sync. We set them to currently running transaction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4848) * as we cannot be sure that the inode or some of its metadata isn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4849) * part of the transaction - the inode could have been reclaimed and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4850) * now it is reread from disk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4851) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4852) if (journal) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4853) transaction_t *transaction;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4854) tid_t tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4856) read_lock(&journal->j_state_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4857) if (journal->j_running_transaction)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4858) transaction = journal->j_running_transaction;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4859) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4860) transaction = journal->j_committing_transaction;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4861) if (transaction)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4862) tid = transaction->t_tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4863) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4864) tid = journal->j_commit_sequence;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4865) read_unlock(&journal->j_state_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4866) ei->i_sync_tid = tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4867) ei->i_datasync_tid = tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4868) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4870) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4871) if (ei->i_extra_isize == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4872) /* The extra space is currently unused. Use it. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4873) BUILD_BUG_ON(sizeof(struct ext4_inode) & 3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4874) ei->i_extra_isize = sizeof(struct ext4_inode) -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4875) EXT4_GOOD_OLD_INODE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4876) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4877) ret = ext4_iget_extra_inode(inode, raw_inode, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4878) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4879) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4880) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4882)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4883) EXT4_INODE_GET_XTIME(i_ctime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4884) EXT4_INODE_GET_XTIME(i_mtime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4885) EXT4_INODE_GET_XTIME(i_atime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4886) EXT4_EINODE_GET_XTIME(i_crtime, ei, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4887)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4888) if (likely(!test_opt2(inode->i_sb, HURD_COMPAT))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4889) u64 ivers = le32_to_cpu(raw_inode->i_disk_version);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4890)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4891) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4892) if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4893) ivers |=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4894) (__u64)(le32_to_cpu(raw_inode->i_version_hi)) << 32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4895) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4896) ext4_inode_set_iversion_queried(inode, ivers);
^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) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4900) if (ei->i_file_acl &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4901) !ext4_inode_block_valid(inode, ei->i_file_acl, 1)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4902) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4903) "iget: bad extended attribute block %llu",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4904) ei->i_file_acl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4905) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4906) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4907) } else if (!ext4_has_inline_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4908) /* validate the block references in the inode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4909) if (!(EXT4_SB(sb)->s_mount_state & EXT4_FC_REPLAY) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4910) (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4911) (S_ISLNK(inode->i_mode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4912) !ext4_inode_is_fast_symlink(inode)))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4913) if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4914) ret = ext4_ext_check_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4915) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4916) ret = ext4_ind_check_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4917) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4918) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4919) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4920) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4921)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4922) if (S_ISREG(inode->i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4923) inode->i_op = &ext4_file_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4924) inode->i_fop = &ext4_file_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4925) ext4_set_aops(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4926) } else if (S_ISDIR(inode->i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4927) inode->i_op = &ext4_dir_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4928) inode->i_fop = &ext4_dir_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4929) } else if (S_ISLNK(inode->i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4930) /* VFS does not allow setting these so must be corruption */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4931) if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4932) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4933) "iget: immutable or append flags "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4934) "not allowed on symlinks");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4935) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4936) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4937) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4938) if (IS_ENCRYPTED(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4939) inode->i_op = &ext4_encrypted_symlink_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4940) ext4_set_aops(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4941) } else if (ext4_inode_is_fast_symlink(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4942) inode->i_link = (char *)ei->i_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4943) inode->i_op = &ext4_fast_symlink_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4944) nd_terminate_link(ei->i_data, inode->i_size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4945) sizeof(ei->i_data) - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4946) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4947) inode->i_op = &ext4_symlink_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4948) ext4_set_aops(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4949) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4950) inode_nohighmem(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4951) } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4952) S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4953) inode->i_op = &ext4_special_inode_operations;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4954) if (raw_inode->i_block[0])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4955) init_special_inode(inode, inode->i_mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4956) old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4957) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4958) init_special_inode(inode, inode->i_mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4959) new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4960) } else if (ino == EXT4_BOOT_LOADER_INO) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4961) make_bad_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4962) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4963) ret = -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4964) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4965) "iget: bogus i_mode (%o)", inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4966) goto bad_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4967) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4968) if (IS_CASEFOLDED(inode) && !ext4_has_feature_casefold(inode->i_sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4969) ext4_error_inode(inode, function, line, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4970) "casefold flag without casefold feature");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4971) brelse(iloc.bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4972)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4973) unlock_new_inode(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4974) return inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4975)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4976) bad_inode:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4977) brelse(iloc.bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4978) iget_failed(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4979) return ERR_PTR(ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4980) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4981)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4982) static int ext4_inode_blocks_set(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4983) struct ext4_inode *raw_inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4984) struct ext4_inode_info *ei)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4985) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4986) struct inode *inode = &(ei->vfs_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4987) u64 i_blocks = READ_ONCE(inode->i_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4988) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4989)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4990) if (i_blocks <= ~0U) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4991) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4992) * i_blocks can be represented in a 32 bit variable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4993) * as multiple of 512 bytes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4994) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4995) raw_inode->i_blocks_lo = cpu_to_le32(i_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4996) raw_inode->i_blocks_high = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4997) ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4998) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4999) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5000) if (!ext4_has_feature_huge_file(sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5001) return -EFBIG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5002)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5003) if (i_blocks <= 0xffffffffffffULL) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5004) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5005) * i_blocks can be represented in a 48 bit variable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5006) * as multiple of 512 bytes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5007) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5008) raw_inode->i_blocks_lo = cpu_to_le32(i_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5009) raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5010) ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5011) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5012) ext4_set_inode_flag(inode, EXT4_INODE_HUGE_FILE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5013) /* i_block is stored in file system block size */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5014) i_blocks = i_blocks >> (inode->i_blkbits - 9);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5015) raw_inode->i_blocks_lo = cpu_to_le32(i_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5016) raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5017) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5018) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5019) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5020)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5021) static void __ext4_update_other_inode_time(struct super_block *sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5022) unsigned long orig_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5023) unsigned long ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5024) struct ext4_inode *raw_inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5025) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5026) struct inode *inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5027)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5028) inode = find_inode_by_ino_rcu(sb, ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5029) if (!inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5030) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5031)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5032) if ((inode->i_state & (I_FREEING | I_WILL_FREE | I_NEW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5033) I_DIRTY_INODE)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5034) ((inode->i_state & I_DIRTY_TIME) == 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5035) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5036)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5037) spin_lock(&inode->i_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5038) if (((inode->i_state & (I_FREEING | I_WILL_FREE | I_NEW |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5039) I_DIRTY_INODE)) == 0) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5040) (inode->i_state & I_DIRTY_TIME)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5041) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5042)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5043) inode->i_state &= ~I_DIRTY_TIME;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5044) spin_unlock(&inode->i_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5045)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5046) spin_lock(&ei->i_raw_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5047) EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5048) EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5049) EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5050) ext4_inode_csum_set(inode, raw_inode, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5051) spin_unlock(&ei->i_raw_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5052) trace_ext4_other_inode_update_time(inode, orig_ino);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5053) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5054) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5055) spin_unlock(&inode->i_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5056) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5057)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5058) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5059) * Opportunistically update the other time fields for other inodes in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5060) * the same inode table block.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5061) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5062) static void ext4_update_other_inodes_time(struct super_block *sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5063) unsigned long orig_ino, char *buf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5064) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5065) unsigned long ino;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5066) int i, inodes_per_block = EXT4_SB(sb)->s_inodes_per_block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5067) int inode_size = EXT4_INODE_SIZE(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5068)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5069) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5070) * Calculate the first inode in the inode table block. Inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5071) * numbers are one-based. That is, the first inode in a block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5072) * (assuming 4k blocks and 256 byte inodes) is (n*16 + 1).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5073) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5074) ino = ((orig_ino - 1) & ~(inodes_per_block - 1)) + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5075) rcu_read_lock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5076) for (i = 0; i < inodes_per_block; i++, ino++, buf += inode_size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5077) if (ino == orig_ino)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5078) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5079) __ext4_update_other_inode_time(sb, orig_ino, ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5080) (struct ext4_inode *)buf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5081) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5082) rcu_read_unlock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5083) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5084)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5085) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5086) * Post the struct inode info into an on-disk inode location in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5087) * buffer-cache. This gobbles the caller's reference to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5088) * buffer_head in the inode location struct.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5089) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5090) * The caller must have write access to iloc->bh.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5091) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5092) static int ext4_do_update_inode(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5093) struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5094) struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5095) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5096) struct ext4_inode *raw_inode = ext4_raw_inode(iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5097) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5098) struct buffer_head *bh = iloc->bh;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5099) struct super_block *sb = inode->i_sb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5100) int err = 0, block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5101) int need_datasync = 0, set_large_file = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5102) uid_t i_uid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5103) gid_t i_gid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5104) projid_t i_projid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5106) spin_lock(&ei->i_raw_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5108) /* For fields not tracked in the in-memory inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5109) * initialise them to zero for new inodes. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5110) if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5111) memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5113) err = ext4_inode_blocks_set(handle, raw_inode, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5114) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5115) spin_unlock(&ei->i_raw_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5116) goto out_brelse;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5117) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5119) raw_inode->i_mode = cpu_to_le16(inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5120) i_uid = i_uid_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5121) i_gid = i_gid_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5122) i_projid = from_kprojid(&init_user_ns, ei->i_projid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5123) if (!(test_opt(inode->i_sb, NO_UID32))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5124) raw_inode->i_uid_low = cpu_to_le16(low_16_bits(i_uid));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5125) raw_inode->i_gid_low = cpu_to_le16(low_16_bits(i_gid));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5126) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5127) * Fix up interoperability with old kernels. Otherwise, old inodes get
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5128) * re-used with the upper 16 bits of the uid/gid intact
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5129) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5130) if (ei->i_dtime && list_empty(&ei->i_orphan)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5131) raw_inode->i_uid_high = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5132) raw_inode->i_gid_high = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5133) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5134) raw_inode->i_uid_high =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5135) cpu_to_le16(high_16_bits(i_uid));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5136) raw_inode->i_gid_high =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5137) cpu_to_le16(high_16_bits(i_gid));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5138) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5139) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5140) raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(i_uid));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5141) raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(i_gid));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5142) raw_inode->i_uid_high = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5143) raw_inode->i_gid_high = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5144) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5145) raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5147) EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5148) EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5149) EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5150) EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5152) raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5153) raw_inode->i_flags = cpu_to_le32(ei->i_flags & 0xFFFFFFFF);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5154) if (likely(!test_opt2(inode->i_sb, HURD_COMPAT)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5155) raw_inode->i_file_acl_high =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5156) cpu_to_le16(ei->i_file_acl >> 32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5157) raw_inode->i_file_acl_lo = cpu_to_le32(ei->i_file_acl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5158) if (READ_ONCE(ei->i_disksize) != ext4_isize(inode->i_sb, raw_inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5159) ext4_isize_set(raw_inode, ei->i_disksize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5160) need_datasync = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5161) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5162) if (ei->i_disksize > 0x7fffffffULL) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5163) if (!ext4_has_feature_large_file(sb) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5164) EXT4_SB(sb)->s_es->s_rev_level ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5165) cpu_to_le32(EXT4_GOOD_OLD_REV))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5166) set_large_file = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5167) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5168) raw_inode->i_generation = cpu_to_le32(inode->i_generation);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5169) if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5170) if (old_valid_dev(inode->i_rdev)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5171) raw_inode->i_block[0] =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5172) cpu_to_le32(old_encode_dev(inode->i_rdev));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5173) raw_inode->i_block[1] = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5174) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5175) raw_inode->i_block[0] = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5176) raw_inode->i_block[1] =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5177) cpu_to_le32(new_encode_dev(inode->i_rdev));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5178) raw_inode->i_block[2] = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5179) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5180) } else if (!ext4_has_inline_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5181) for (block = 0; block < EXT4_N_BLOCKS; block++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5182) raw_inode->i_block[block] = ei->i_data[block];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5183) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5184)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5185) if (likely(!test_opt2(inode->i_sb, HURD_COMPAT))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5186) u64 ivers = ext4_inode_peek_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5187)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5188) raw_inode->i_disk_version = cpu_to_le32(ivers);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5189) if (ei->i_extra_isize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5190) if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5191) raw_inode->i_version_hi =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5192) cpu_to_le32(ivers >> 32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5193) raw_inode->i_extra_isize =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5194) cpu_to_le16(ei->i_extra_isize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5195) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5196) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5197)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5198) BUG_ON(!ext4_has_feature_project(inode->i_sb) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5199) i_projid != EXT4_DEF_PROJID);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5200)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5201) if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5202) EXT4_FITS_IN_INODE(raw_inode, ei, i_projid))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5203) raw_inode->i_projid = cpu_to_le32(i_projid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5204)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5205) ext4_inode_csum_set(inode, raw_inode, ei);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5206) spin_unlock(&ei->i_raw_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5207) if (inode->i_sb->s_flags & SB_LAZYTIME)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5208) ext4_update_other_inodes_time(inode->i_sb, inode->i_ino,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5209) bh->b_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5210)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5211) BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5212) err = ext4_handle_dirty_metadata(handle, NULL, bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5213) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5214) goto out_brelse;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5215) ext4_clear_inode_state(inode, EXT4_STATE_NEW);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5216) if (set_large_file) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5217) BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get write access");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5218) err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5219) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5220) goto out_brelse;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5221) ext4_set_feature_large_file(sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5222) ext4_handle_sync(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5223) err = ext4_handle_dirty_super(handle, sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5224) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5225) ext4_update_inode_fsync_trans(handle, inode, need_datasync);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5226) out_brelse:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5227) brelse(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5228) ext4_std_error(inode->i_sb, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5229) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5230) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5231)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5232) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5233) * ext4_write_inode()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5234) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5235) * We are called from a few places:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5236) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5237) * - Within generic_file_aio_write() -> generic_write_sync() for O_SYNC files.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5238) * Here, there will be no transaction running. We wait for any running
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5239) * transaction to commit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5240) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5241) * - Within flush work (sys_sync(), kupdate and such).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5242) * We wait on commit, if told to.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5243) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5244) * - Within iput_final() -> write_inode_now()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5245) * We wait on commit, if told to.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5246) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5247) * In all cases it is actually safe for us to return without doing anything,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5248) * because the inode has been copied into a raw inode buffer in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5249) * ext4_mark_inode_dirty(). This is a correctness thing for WB_SYNC_ALL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5250) * writeback.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5251) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5252) * Note that we are absolutely dependent upon all inode dirtiers doing the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5253) * right thing: they *must* call mark_inode_dirty() after dirtying info in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5254) * which we are interested.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5255) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5256) * It would be a bug for them to not do this. The code:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5257) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5258) * mark_inode_dirty(inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5259) * stuff();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5260) * inode->i_size = expr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5261) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5262) * is in error because write_inode() could occur while `stuff()' is running,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5263) * and the new i_size will be lost. Plus the inode will no longer be on the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5264) * superblock's dirty inode list.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5265) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5266) int ext4_write_inode(struct inode *inode, struct writeback_control *wbc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5267) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5268) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5269)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5270) if (WARN_ON_ONCE(current->flags & PF_MEMALLOC) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5271) sb_rdonly(inode->i_sb))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5272) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5273)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5274) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5275) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5276)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5277) if (EXT4_SB(inode->i_sb)->s_journal) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5278) if (ext4_journal_current_handle()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5279) jbd_debug(1, "called recursively, non-PF_MEMALLOC!\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5280) dump_stack();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5281) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5282) }
^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) * No need to force transaction in WB_SYNC_NONE mode. Also
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5286) * ext4_sync_fs() will force the commit after everything is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5287) * written.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5288) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5289) if (wbc->sync_mode != WB_SYNC_ALL || wbc->for_sync)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5290) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5292) err = ext4_fc_commit(EXT4_SB(inode->i_sb)->s_journal,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5293) EXT4_I(inode)->i_sync_tid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5294) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5295) struct ext4_iloc iloc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5297) err = __ext4_get_inode_loc_noinmem(inode, &iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5298) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5299) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5300) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5301) * sync(2) will flush the whole buffer cache. No need to do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5302) * it here separately for each inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5303) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5304) if (wbc->sync_mode == WB_SYNC_ALL && !wbc->for_sync)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5305) sync_dirty_buffer(iloc.bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5306) if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5307) ext4_error_inode_block(inode, iloc.bh->b_blocknr, EIO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5308) "IO error syncing inode");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5309) err = -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5311) brelse(iloc.bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5312) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5313) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5314) }
^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) * In data=journal mode ext4_journalled_invalidatepage() may fail to invalidate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5318) * buffers that are attached to a page stradding i_size and are undergoing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5319) * commit. In that case we have to wait for commit to finish and try again.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5320) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5321) static void ext4_wait_for_tail_page_commit(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5322) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5323) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5324) unsigned offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5325) journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5326) tid_t commit_tid = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5327) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5329) offset = inode->i_size & (PAGE_SIZE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5330) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5331) * If the page is fully truncated, we don't need to wait for any commit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5332) * (and we even should not as __ext4_journalled_invalidatepage() may
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5333) * strip all buffers from the page but keep the page dirty which can then
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5334) * confuse e.g. concurrent ext4_writepage() seeing dirty page without
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5335) * buffers). Also we don't need to wait for any commit if all buffers in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5336) * the page remain valid. This is most beneficial for the common case of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5337) * blocksize == PAGESIZE.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5338) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5339) if (!offset || offset > (PAGE_SIZE - i_blocksize(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5340) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5341) while (1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5342) page = find_lock_page(inode->i_mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5343) inode->i_size >> PAGE_SHIFT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5344) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5345) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5346) ret = __ext4_journalled_invalidatepage(page, offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5347) PAGE_SIZE - offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5348) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5349) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5350) if (ret != -EBUSY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5351) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5352) commit_tid = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5353) read_lock(&journal->j_state_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5354) if (journal->j_committing_transaction)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5355) commit_tid = journal->j_committing_transaction->t_tid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5356) read_unlock(&journal->j_state_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5357) if (commit_tid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5358) jbd2_log_wait_commit(journal, commit_tid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5359) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5360) }
^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) * ext4_setattr()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5364) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5365) * Called from notify_change.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5366) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5367) * We want to trap VFS attempts to truncate the file as soon as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5368) * possible. In particular, we want to make sure that when the VFS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5369) * shrinks i_size, we put the inode on the orphan list and modify
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5370) * i_disksize immediately, so that during the subsequent flushing of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5371) * dirty pages and freeing of disk blocks, we can guarantee that any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5372) * commit will leave the blocks being flushed in an unused state on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5373) * disk. (On recovery, the inode will get truncated and the blocks will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5374) * be freed, so we have a strong guarantee that no future commit will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5375) * leave these blocks visible to the user.)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5376) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5377) * Another thing we have to assure is that if we are in ordered mode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5378) * and inode is still attached to the committing transaction, we must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5379) * we start writeout of all the dirty pages which are being truncated.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5380) * This way we are sure that all the data written in the previous
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5381) * transaction are already on disk (truncate waits for pages under
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5382) * writeback).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5383) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5384) * Called with inode->i_mutex down.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5385) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5386) int ext4_setattr(struct dentry *dentry, struct iattr *attr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5387) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5388) struct inode *inode = d_inode(dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5389) int error, rc = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5390) int orphan = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5391) const unsigned int ia_valid = attr->ia_valid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5392)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5393) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5394) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5395)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5396) if (unlikely(IS_IMMUTABLE(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5397) return -EPERM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5399) if (unlikely(IS_APPEND(inode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5400) (ia_valid & (ATTR_MODE | ATTR_UID |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5401) ATTR_GID | ATTR_TIMES_SET))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5402) return -EPERM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5403)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5404) error = setattr_prepare(dentry, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5405) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5406) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5408) error = fscrypt_prepare_setattr(dentry, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5409) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5410) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5411)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5412) error = fsverity_prepare_setattr(dentry, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5413) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5414) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5415)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5416) if (is_quota_modification(inode, attr)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5417) error = dquot_initialize(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5418) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5419) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5420) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5421) ext4_fc_start_update(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5422) if ((ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5423) (ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5424) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5425)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5426) /* (user+group)*(old+new) structure, inode write (sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5427) * inode block, ? - but truncate inode update has it) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5428) handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5429) (EXT4_MAXQUOTAS_INIT_BLOCKS(inode->i_sb) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5430) EXT4_MAXQUOTAS_DEL_BLOCKS(inode->i_sb)) + 3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5431) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5432) error = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5433) goto err_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5434) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5436) /* dquot_transfer() calls back ext4_get_inode_usage() which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5437) * counts xattr inode references.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5438) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5439) down_read(&EXT4_I(inode)->xattr_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5440) error = dquot_transfer(inode, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5441) up_read(&EXT4_I(inode)->xattr_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5442)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5443) if (error) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5444) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5445) ext4_fc_stop_update(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5446) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5447) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5448) /* Update corresponding info in inode so that everything is in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5449) * one transaction */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5450) if (attr->ia_valid & ATTR_UID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5451) inode->i_uid = attr->ia_uid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5452) if (attr->ia_valid & ATTR_GID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5453) inode->i_gid = attr->ia_gid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5454) error = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5455) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5456) if (unlikely(error)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5457) ext4_fc_stop_update(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5458) return error;
^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)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5462) if (attr->ia_valid & ATTR_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5463) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5464) loff_t oldsize = inode->i_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5465) int shrink = (attr->ia_size < inode->i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5466)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5467) if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5468) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5469)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5470) if (attr->ia_size > sbi->s_bitmap_maxbytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5471) ext4_fc_stop_update(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5472) return -EFBIG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5473) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5474) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5475) if (!S_ISREG(inode->i_mode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5476) ext4_fc_stop_update(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5477) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5478) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5479)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5480) if (IS_I_VERSION(inode) && attr->ia_size != inode->i_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5481) inode_inc_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5482)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5483) if (shrink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5484) if (ext4_should_order_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5485) error = ext4_begin_ordered_truncate(inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5486) attr->ia_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5487) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5488) goto err_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5489) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5490) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5491) * Blocks are going to be removed from the inode. Wait
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5492) * for dio in flight.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5493) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5494) inode_dio_wait(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5495) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5496)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5497) down_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5498)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5499) rc = ext4_break_layouts(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5500) if (rc) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5501) up_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5502) goto err_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5503) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5504)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5505) if (attr->ia_size != inode->i_size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5506) handle = ext4_journal_start(inode, EXT4_HT_INODE, 3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5507) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5508) error = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5509) goto out_mmap_sem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5510) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5511) if (ext4_handle_valid(handle) && shrink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5512) error = ext4_orphan_add(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5513) orphan = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5514) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5515) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5516) * Update c/mtime on truncate up, ext4_truncate() will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5517) * update c/mtime in shrink case below
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5518) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5519) if (!shrink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5520) inode->i_mtime = current_time(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5521) inode->i_ctime = inode->i_mtime;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5522) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5524) if (shrink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5525) ext4_fc_track_range(handle, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5526) (attr->ia_size > 0 ? attr->ia_size - 1 : 0) >>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5527) inode->i_sb->s_blocksize_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5528) EXT_MAX_BLOCKS - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5529) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5530) ext4_fc_track_range(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5531) handle, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5532) (oldsize > 0 ? oldsize - 1 : oldsize) >>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5533) inode->i_sb->s_blocksize_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5534) (attr->ia_size > 0 ? attr->ia_size - 1 : 0) >>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5535) inode->i_sb->s_blocksize_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5536)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5537) down_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5538) EXT4_I(inode)->i_disksize = attr->ia_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5539) rc = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5540) if (!error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5541) error = rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5542) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5543) * We have to update i_size under i_data_sem together
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5544) * with i_disksize to avoid races with writeback code
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5545) * running ext4_wb_update_i_disksize().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5546) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5547) if (!error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5548) i_size_write(inode, attr->ia_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5549) up_write(&EXT4_I(inode)->i_data_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5550) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5551) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5552) goto out_mmap_sem;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5553) if (!shrink) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5554) pagecache_isize_extended(inode, oldsize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5555) inode->i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5556) } else if (ext4_should_journal_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5557) ext4_wait_for_tail_page_commit(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5558) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5559) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5560)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5561) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5562) * Truncate pagecache after we've waited for commit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5563) * in data=journal mode to make pages freeable.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5564) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5565) truncate_pagecache(inode, inode->i_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5566) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5567) * Call ext4_truncate() even if i_size didn't change to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5568) * truncate possible preallocated blocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5569) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5570) if (attr->ia_size <= oldsize) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5571) rc = ext4_truncate(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5572) if (rc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5573) error = rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5574) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5575) out_mmap_sem:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5576) up_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5577) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5578)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5579) if (!error) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5580) setattr_copy(inode, attr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5581) mark_inode_dirty(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5582) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5583)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5584) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5585) * If the call to ext4_truncate failed to get a transaction handle at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5586) * all, we need to clean up the in-core orphan list manually.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5587) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5588) if (orphan && inode->i_nlink)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5589) ext4_orphan_del(NULL, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5591) if (!error && (ia_valid & ATTR_MODE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5592) rc = posix_acl_chmod(inode, inode->i_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5593)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5594) err_out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5595) if (error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5596) ext4_std_error(inode->i_sb, error);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5597) if (!error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5598) error = rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5599) ext4_fc_stop_update(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5600) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5601) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5602)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5603) int ext4_getattr(const struct path *path, struct kstat *stat,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5604) u32 request_mask, unsigned int query_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5605) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5606) struct inode *inode = d_inode(path->dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5607) struct ext4_inode *raw_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5608) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5609) unsigned int flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5610)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5611) if ((request_mask & STATX_BTIME) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5612) EXT4_FITS_IN_INODE(raw_inode, ei, i_crtime)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5613) stat->result_mask |= STATX_BTIME;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5614) stat->btime.tv_sec = ei->i_crtime.tv_sec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5615) stat->btime.tv_nsec = ei->i_crtime.tv_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5618) flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5619) if (flags & EXT4_APPEND_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5620) stat->attributes |= STATX_ATTR_APPEND;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5621) if (flags & EXT4_COMPR_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5622) stat->attributes |= STATX_ATTR_COMPRESSED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5623) if (flags & EXT4_ENCRYPT_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5624) stat->attributes |= STATX_ATTR_ENCRYPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5625) if (flags & EXT4_IMMUTABLE_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5626) stat->attributes |= STATX_ATTR_IMMUTABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5627) if (flags & EXT4_NODUMP_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5628) stat->attributes |= STATX_ATTR_NODUMP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5629) if (flags & EXT4_VERITY_FL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5630) stat->attributes |= STATX_ATTR_VERITY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5631)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5632) stat->attributes_mask |= (STATX_ATTR_APPEND |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5633) STATX_ATTR_COMPRESSED |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5634) STATX_ATTR_ENCRYPTED |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5635) STATX_ATTR_IMMUTABLE |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5636) STATX_ATTR_NODUMP |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5637) STATX_ATTR_VERITY);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5638)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5639) generic_fillattr(inode, stat);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5640) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5641) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5643) int ext4_file_getattr(const struct path *path, struct kstat *stat,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5644) u32 request_mask, unsigned int query_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5645) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5646) struct inode *inode = d_inode(path->dentry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5647) u64 delalloc_blocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5649) ext4_getattr(path, stat, request_mask, query_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5651) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5652) * If there is inline data in the inode, the inode will normally not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5653) * have data blocks allocated (it may have an external xattr block).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5654) * Report at least one sector for such files, so tools like tar, rsync,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5655) * others don't incorrectly think the file is completely sparse.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5656) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5657) if (unlikely(ext4_has_inline_data(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5658) stat->blocks += (stat->size + 511) >> 9;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5659)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5660) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5661) * We can't update i_blocks if the block allocation is delayed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5662) * otherwise in the case of system crash before the real block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5663) * allocation is done, we will have i_blocks inconsistent with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5664) * on-disk file blocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5665) * We always keep i_blocks updated together with real
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5666) * allocation. But to not confuse with user, stat
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5667) * will return the blocks that include the delayed allocation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5668) * blocks for this file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5669) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5670) delalloc_blocks = EXT4_C2B(EXT4_SB(inode->i_sb),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5671) EXT4_I(inode)->i_reserved_data_blocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5672) stat->blocks += delalloc_blocks << (inode->i_sb->s_blocksize_bits - 9);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5673) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5674) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5675)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5676) static int ext4_index_trans_blocks(struct inode *inode, int lblocks,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5677) int pextents)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5678) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5679) if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5680) return ext4_ind_trans_blocks(inode, lblocks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5681) return ext4_ext_index_trans_blocks(inode, pextents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5682) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5683)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5684) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5685) * Account for index blocks, block groups bitmaps and block group
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5686) * descriptor blocks if modify datablocks and index blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5687) * worse case, the indexs blocks spread over different block groups
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5688) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5689) * If datablocks are discontiguous, they are possible to spread over
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5690) * different block groups too. If they are contiguous, with flexbg,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5691) * they could still across block group boundary.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5692) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5693) * Also account for superblock, inode, quota and xattr blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5694) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5695) static int ext4_meta_trans_blocks(struct inode *inode, int lblocks,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5696) int pextents)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5697) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5698) ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5699) int gdpblocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5700) int idxblocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5701) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5702)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5703) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5704) * How many index blocks need to touch to map @lblocks logical blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5705) * to @pextents physical extents?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5706) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5707) idxblocks = ext4_index_trans_blocks(inode, lblocks, pextents);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5708)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5709) ret = idxblocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5710)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5711) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5712) * Now let's see how many group bitmaps and group descriptors need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5713) * to account
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5714) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5715) groups = idxblocks + pextents;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5716) gdpblocks = groups;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5717) if (groups > ngroups)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5718) groups = ngroups;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5719) if (groups > EXT4_SB(inode->i_sb)->s_gdb_count)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5720) gdpblocks = EXT4_SB(inode->i_sb)->s_gdb_count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5721)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5722) /* bitmaps and block group descriptor blocks */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5723) ret += groups + gdpblocks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5724)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5725) /* Blocks for super block, inode, quota and xattr blocks */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5726) ret += EXT4_META_TRANS_BLOCKS(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5727)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5728) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5731) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5732) * Calculate the total number of credits to reserve to fit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5733) * the modification of a single pages into a single transaction,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5734) * which may include multiple chunks of block allocations.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5735) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5736) * This could be called via ext4_write_begin()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5737) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5738) * We need to consider the worse case, when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5739) * one new block per extent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5740) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5741) int ext4_writepage_trans_blocks(struct inode *inode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5742) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5743) int bpp = ext4_journal_blocks_per_page(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5744) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5745)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5746) ret = ext4_meta_trans_blocks(inode, bpp, bpp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5747)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5748) /* Account for data blocks for journalled mode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5749) if (ext4_should_journal_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5750) ret += bpp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5751) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5752) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5753)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5754) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5755) * Calculate the journal credits for a chunk of data modification.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5756) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5757) * This is called from DIO, fallocate or whoever calling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5758) * ext4_map_blocks() to map/allocate a chunk of contiguous disk blocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5759) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5760) * journal buffers for data blocks are not included here, as DIO
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5761) * and fallocate do no need to journal data buffers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5762) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5763) int ext4_chunk_trans_blocks(struct inode *inode, int nrblocks)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5764) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5765) return ext4_meta_trans_blocks(inode, nrblocks, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5766) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5767)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5768) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5769) * The caller must have previously called ext4_reserve_inode_write().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5770) * Give this, we know that the caller already has write access to iloc->bh.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5771) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5772) int ext4_mark_iloc_dirty(handle_t *handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5773) struct inode *inode, struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5774) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5775) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5776)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5777) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5778) put_bh(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5779) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5780) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5781) ext4_fc_track_inode(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5782)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5783) if (IS_I_VERSION(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5784) inode_inc_iversion(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5785)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5786) /* the do_update_inode consumes one bh->b_count */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5787) get_bh(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5788)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5789) /* ext4_do_update_inode() does jbd2_journal_dirty_metadata */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5790) err = ext4_do_update_inode(handle, inode, iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5791) put_bh(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5792) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5793) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5794)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5795) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5796) * On success, We end up with an outstanding reference count against
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5797) * iloc->bh. This _must_ be cleaned up later.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5798) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5799)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5800) int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5801) ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5802) struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5803) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5804) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5806) if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5807) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5809) err = ext4_get_inode_loc(inode, iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5810) if (!err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5811) BUFFER_TRACE(iloc->bh, "get_write_access");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5812) err = ext4_journal_get_write_access(handle, iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5813) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5814) brelse(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5815) iloc->bh = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5816) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5817) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5818) ext4_std_error(inode->i_sb, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5819) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5820) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5821)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5822) static int __ext4_expand_extra_isize(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5823) unsigned int new_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5824) struct ext4_iloc *iloc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5825) handle_t *handle, int *no_expand)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5826) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5827) struct ext4_inode *raw_inode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5828) struct ext4_xattr_ibody_header *header;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5829) unsigned int inode_size = EXT4_INODE_SIZE(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5830) struct ext4_inode_info *ei = EXT4_I(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5831) int error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5832)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5833) /* this was checked at iget time, but double check for good measure */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5834) if ((EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize > inode_size) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5835) (ei->i_extra_isize & 3)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5836) EXT4_ERROR_INODE(inode, "bad extra_isize %u (inode size %u)",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5837) ei->i_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5838) EXT4_INODE_SIZE(inode->i_sb));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5839) return -EFSCORRUPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5840) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5841) if ((new_extra_isize < ei->i_extra_isize) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5842) (new_extra_isize < 4) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5843) (new_extra_isize > inode_size - EXT4_GOOD_OLD_INODE_SIZE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5844) return -EINVAL; /* Should never happen */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5846) raw_inode = ext4_raw_inode(iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5847)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5848) header = IHDR(inode, raw_inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5849)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5850) /* No extended attributes present */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5851) if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5852) header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5853) memset((void *)raw_inode + EXT4_GOOD_OLD_INODE_SIZE +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5854) EXT4_I(inode)->i_extra_isize, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5855) new_extra_isize - EXT4_I(inode)->i_extra_isize);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5856) EXT4_I(inode)->i_extra_isize = new_extra_isize;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5857) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5858) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5859)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5860) /* try to expand with EAs present */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5861) error = ext4_expand_extra_isize_ea(inode, new_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5862) raw_inode, handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5863) if (error) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5864) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5865) * Inode size expansion failed; don't try again
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5866) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5867) *no_expand = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5868) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5870) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5871) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5872)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5873) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5874) * Expand an inode by new_extra_isize bytes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5875) * Returns 0 on success or negative error number on failure.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5876) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5877) static int ext4_try_to_expand_extra_isize(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5878) unsigned int new_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5879) struct ext4_iloc iloc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5880) handle_t *handle)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5881) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5882) int no_expand;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5883) int error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5884)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5885) if (ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5886) return -EOVERFLOW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5887)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5888) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5889) * In nojournal mode, we can immediately attempt to expand
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5890) * the inode. When journaled, we first need to obtain extra
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5891) * buffer credits since we may write into the EA block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5892) * with this same handle. If journal_extend fails, then it will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5893) * only result in a minor loss of functionality for that inode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5894) * If this is felt to be critical, then e2fsck should be run to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5895) * force a large enough s_min_extra_isize.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5896) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5897) if (ext4_journal_extend(handle,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5898) EXT4_DATA_TRANS_BLOCKS(inode->i_sb), 0) != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5899) return -ENOSPC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5900)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5901) if (ext4_write_trylock_xattr(inode, &no_expand) == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5902) return -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5903)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5904) error = __ext4_expand_extra_isize(inode, new_extra_isize, &iloc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5905) handle, &no_expand);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5906) ext4_write_unlock_xattr(inode, &no_expand);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5907)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5908) return error;
^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) int ext4_expand_extra_isize(struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5912) unsigned int new_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5913) struct ext4_iloc *iloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5914) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5915) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5916) int no_expand;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5917) int error, rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5919) if (ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5920) brelse(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5921) return -EOVERFLOW;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5922) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5924) handle = ext4_journal_start(inode, EXT4_HT_INODE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5925) EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5926) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5927) error = PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5928) brelse(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5929) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5930) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5932) ext4_write_lock_xattr(inode, &no_expand);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5933)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5934) BUFFER_TRACE(iloc->bh, "get_write_access");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5935) error = ext4_journal_get_write_access(handle, iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5936) if (error) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5937) brelse(iloc->bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5938) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5939) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5940)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5941) error = __ext4_expand_extra_isize(inode, new_extra_isize, iloc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5942) handle, &no_expand);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5943)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5944) rc = ext4_mark_iloc_dirty(handle, inode, iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5945) if (!error)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5946) error = rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5948) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5949) ext4_write_unlock_xattr(inode, &no_expand);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5950) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5951) return error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5952) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5953)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5954) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5955) * What we do here is to mark the in-core inode as clean with respect to inode
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5956) * dirtiness (it may still be data-dirty).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5957) * This means that the in-core inode may be reaped by prune_icache
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5958) * without having to perform any I/O. This is a very good thing,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5959) * because *any* task may call prune_icache - even ones which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5960) * have a transaction open against a different journal.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5961) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5962) * Is this cheating? Not really. Sure, we haven't written the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5963) * inode out, but prune_icache isn't a user-visible syncing function.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5964) * Whenever the user wants stuff synced (sys_sync, sys_msync, sys_fsync)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5965) * we start and wait on commits.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5966) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5967) int __ext4_mark_inode_dirty(handle_t *handle, struct inode *inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5968) const char *func, unsigned int line)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5969) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5970) struct ext4_iloc iloc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5971) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5972) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5973)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5974) might_sleep();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5975) trace_ext4_mark_inode_dirty(inode, _RET_IP_);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5976) err = ext4_reserve_inode_write(handle, inode, &iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5977) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5978) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5979)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5980) if (EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5981) ext4_try_to_expand_extra_isize(inode, sbi->s_want_extra_isize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5982) iloc, handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5983)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5984) err = ext4_mark_iloc_dirty(handle, inode, &iloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5985) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5986) if (unlikely(err))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5987) ext4_error_inode_err(inode, func, line, 0, err,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5988) "mark_inode_dirty error");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5989) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5990) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5991)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5992) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5993) * ext4_dirty_inode() is called from __mark_inode_dirty()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5994) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5995) * We're really interested in the case where a file is being extended.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5996) * i_size has been changed by generic_commit_write() and we thus need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5997) * to include the updated inode in the current transaction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5998) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5999) * Also, dquot_alloc_block() will always dirty the inode when blocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6000) * are allocated to the file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6001) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6002) * If the inode is marked synchronous, we don't honour that here - doing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6003) * so would cause a commit on atime updates, which we don't bother doing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6004) * We handle synchronous inodes at the highest possible level.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6005) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6006) * If only the I_DIRTY_TIME flag is set, we can skip everything. If
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6007) * I_DIRTY_TIME and I_DIRTY_SYNC is set, the only inode fields we need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6008) * to copy into the on-disk inode structure are the timestamp files.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6009) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6010) void ext4_dirty_inode(struct inode *inode, int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6011) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6012) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6013)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6014) if (flags == I_DIRTY_TIME)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6015) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6016) handle = ext4_journal_start(inode, EXT4_HT_INODE, 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6017) if (IS_ERR(handle))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6018) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6019)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6020) ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6022) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6023) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6024) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6025) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6026)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6027) int ext4_change_inode_journal_flag(struct inode *inode, int val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6028) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6029) journal_t *journal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6030) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6031) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6032) struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
^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) * We have to be very careful here: changing a data block's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6036) * journaling status dynamically is dangerous. If we write a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6037) * data block to the journal, change the status and then delete
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6038) * that block, we risk forgetting to revoke the old log record
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6039) * from the journal and so a subsequent replay can corrupt data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6040) * So, first we make sure that the journal is empty and that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6041) * nobody is changing anything.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6042) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6043)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6044) journal = EXT4_JOURNAL(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6045) if (!journal)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6046) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6047) if (is_journal_aborted(journal))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6048) return -EROFS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6049)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6050) /* Wait for all existing dio workers */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6051) inode_dio_wait(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6052)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6053) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6054) * Before flushing the journal and switching inode's aops, we have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6055) * to flush all dirty data the inode has. There can be outstanding
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6056) * delayed allocations, there can be unwritten extents created by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6057) * fallocate or buffered writes in dioread_nolock mode covered by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6058) * dirty data which can be converted only after flushing the dirty
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6059) * data (and journalled aops don't know how to handle these cases).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6060) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6061) if (val) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6062) down_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6063) err = filemap_write_and_wait(inode->i_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6064) if (err < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6065) up_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6066) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6067) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6068) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6069)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6070) percpu_down_write(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6071) jbd2_journal_lock_updates(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6072)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6073) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6074) * OK, there are no updates running now, and all cached data is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6075) * synced to disk. We are now in a completely consistent state
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6076) * which doesn't have anything in the journal, and we know that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6077) * no filesystem updates are running, so it is safe to modify
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6078) * the inode's in-core data-journaling state flag now.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6079) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6080)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6081) if (val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6082) ext4_set_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6083) else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6084) err = jbd2_journal_flush(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6085) if (err < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6086) jbd2_journal_unlock_updates(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6087) percpu_up_write(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6088) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6089) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6090) ext4_clear_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6091) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6092) ext4_set_aops(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6093)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6094) jbd2_journal_unlock_updates(journal);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6095) percpu_up_write(&sbi->s_writepages_rwsem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6097) if (val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6098) up_write(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6099)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6100) /* Finally we can mark the inode as dirty. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6102) handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6103) if (IS_ERR(handle))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6104) return PTR_ERR(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6106) ext4_fc_mark_ineligible(inode->i_sb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6107) EXT4_FC_REASON_JOURNAL_FLAG_CHANGE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6108) err = ext4_mark_inode_dirty(handle, inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6109) ext4_handle_sync(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6110) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6111) ext4_std_error(inode->i_sb, err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6113) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6114) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6116) static int ext4_bh_unmapped(handle_t *handle, struct buffer_head *bh)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6117) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6118) return !buffer_mapped(bh);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6119) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6120)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6121) vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6122) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6123) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6124) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6125) loff_t size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6126) unsigned long len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6127) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6128) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6129) struct file *file = vma->vm_file;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6130) struct inode *inode = file_inode(file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6131) struct address_space *mapping = inode->i_mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6132) handle_t *handle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6133) get_block_t *get_block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6134) int retries = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6135)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6136) if (unlikely(IS_IMMUTABLE(inode)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6137) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6139) sb_start_pagefault(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6140) file_update_time(vma->vm_file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6141)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6142) down_read(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6143)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6144) err = ext4_convert_inline_data(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6145) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6146) goto out_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) * On data journalling we skip straight to the transaction handle:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6150) * there's no delalloc; page truncated will be checked later; the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6151) * early return w/ all buffers mapped (calculates size/len) can't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6152) * be used; and there's no dioread_nolock, so only ext4_get_block.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6153) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6154) if (ext4_should_journal_data(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6155) goto retry_alloc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6156)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6157) /* Delalloc case is easy... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6158) if (test_opt(inode->i_sb, DELALLOC) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6159) !ext4_nonda_switch(inode->i_sb)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6160) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6161) err = block_page_mkwrite(vma, vmf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6162) ext4_da_get_block_prep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6163) } while (err == -ENOSPC &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6164) ext4_should_retry_alloc(inode->i_sb, &retries));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6165) goto out_ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6166) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6168) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6169) size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6170) /* Page got truncated from under us? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6171) if (page->mapping != mapping || page_offset(page) > size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6172) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6173) ret = VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6174) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6175) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6176)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6177) if (page->index == size >> PAGE_SHIFT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6178) len = size & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6179) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6180) len = PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6181) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6182) * Return if we have all the buffers mapped. This avoids the need to do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6183) * journal_start/journal_stop which can block and take a long time
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6184) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6185) * This cannot be done for data journalling, as we have to add the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6186) * inode to the transaction's list to writeprotect pages on commit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6187) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6188) if (page_has_buffers(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6189) if (!ext4_walk_page_buffers(NULL, page_buffers(page),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6190) 0, len, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6191) ext4_bh_unmapped)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6192) /* Wait so that we don't change page under IO */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6193) wait_for_stable_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6194) ret = VM_FAULT_LOCKED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6195) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6196) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6197) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6198) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6199) /* OK, we need to fill the hole... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6200) if (ext4_should_dioread_nolock(inode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6201) get_block = ext4_get_block_unwritten;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6202) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6203) get_block = ext4_get_block;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6204) retry_alloc:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6205) handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6206) ext4_writepage_trans_blocks(inode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6207) if (IS_ERR(handle)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6208) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6209) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6210) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6211) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6212) * Data journalling can't use block_page_mkwrite() because it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6213) * will set_buffer_dirty() before do_journal_get_write_access()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6214) * thus might hit warning messages for dirty metadata buffers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6215) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6216) if (!ext4_should_journal_data(inode)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6217) err = block_page_mkwrite(vma, vmf, get_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6218) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6219) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6220) size = i_size_read(inode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6221) /* Page got truncated from under us? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6222) if (page->mapping != mapping || page_offset(page) > size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6223) ret = VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6224) goto out_error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6225) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6226)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6227) if (page->index == size >> PAGE_SHIFT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6228) len = size & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6229) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6230) len = PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6231)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6232) err = __block_write_begin(page, 0, len, ext4_get_block);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6233) if (!err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6234) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6235) if (ext4_walk_page_buffers(handle, page_buffers(page),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6236) 0, len, NULL, do_journal_get_write_access))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6237) goto out_error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6238) if (ext4_walk_page_buffers(handle, page_buffers(page),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6239) 0, len, NULL, write_end_fn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6240) goto out_error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6241) if (ext4_jbd2_inode_add_write(handle, inode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6242) page_offset(page), len))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6243) goto out_error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6244) ext4_set_inode_state(inode, EXT4_STATE_JDATA);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6245) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6246) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6247) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6249) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6250) if (err == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6251) goto retry_alloc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6252) out_ret:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6253) ret = block_page_mkwrite_return(err);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6254) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6255) up_read(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6256) sb_end_pagefault(inode->i_sb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6257) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6258) out_error:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6259) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6260) ext4_journal_stop(handle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6261) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6262) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6264) vm_fault_t ext4_filemap_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6265) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6266) struct inode *inode = file_inode(vmf->vma->vm_file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6267) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6268)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6269) down_read(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6270) ret = filemap_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6271) up_read(&EXT4_I(inode)->i_mmap_sem);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6272)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6273) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6274) }
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags