releases/5.15.153/f2fs-invalidate-meta_mapping-before-ipu-dio-write.patch

   1 From e8859acb23a33ce1eeee3c6cf6c15038198aa75a Mon Sep 17 00:00:00 2001
   2 From: Sasha Levin <sashal@kernel.org>
   3 Date: Tue, 2 Nov 2021 16:10:02 +0900
   4 Subject: f2fs: invalidate META_MAPPING before IPU/DIO write
   5
   6 From: Hyeong-Jun Kim <hj514.kim@samsung.com>
   7
   8 [ Upstream commit e3b49ea36802053f312013fd4ccb6e59920a9f76 ]
   9
  10 Encrypted pages during GC are read and cached in META_MAPPING.
  11 However, due to cached pages in META_MAPPING, there is an issue where
  12 newly written pages are lost by IPU or DIO writes.
  13
  14 Thread A - f2fs_gc()            Thread B
  15 /* phase 3 */
  16 down_write(i_gc_rwsem)
  17 ra_data_block()       ---- (a)
  18 up_write(i_gc_rwsem)
  19                                 f2fs_direct_IO() :
  20                                  - down_read(i_gc_rwsem)
  21                                  - __blockdev_direct_io()
  22                                  - get_data_block_dio_write()
  23                                  - f2fs_dio_submit_bio()  ---- (b)
  24                                  - up_read(i_gc_rwsem)
  25 /* phase 4 */
  26 down_write(i_gc_rwsem)
  27 move_data_block()     ---- (c)
  28 up_write(i_gc_rwsem)
  29
  30 (a) In phase 3 of f2fs_gc(), up-to-date page is read from storage and
  31     cached in META_MAPPING.
  32 (b) In thread B, writing new data by IPU or DIO write on same blkaddr as
  33     read in (a). cached page in META_MAPPING become out-dated.
  34 (c) In phase 4 of f2fs_gc(), out-dated page in META_MAPPING is copied to
  35     new blkaddr. In conclusion, the newly written data in (b) is lost.
  36
  37 To address this issue, invalidating pages in META_MAPPING before IPU or
  38 DIO write.
  39
  40 Fixes: 6aa58d8ad20a ("f2fs: readahead encrypted block during GC")
  41 Signed-off-by: Hyeong-Jun Kim <hj514.kim@samsung.com>
  42 Reviewed-by: Chao Yu <chao@kernel.org>
  43 Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
  44 Stable-dep-of: fd244524c2cf ("f2fs: compress: fix to cover normal cluster write with cp_rwsem")
  45 Signed-off-by: Sasha Levin <sashal@kernel.org>
  46 ---
  47  fs/f2fs/data.c    | 2 ++
  48  fs/f2fs/segment.c | 3 +++
  49  2 files changed, 5 insertions(+)
  50
  51 diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
  52 index e005f97fd273e..25dafd1261d71 100644
  53 --- a/fs/f2fs/data.c
  54 +++ b/fs/f2fs/data.c
  55 @@ -1711,6 +1711,8 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
  56                  */
  57                 f2fs_wait_on_block_writeback_range(inode,
  58                                                 map->m_pblk, map->m_len);
  59 +               invalidate_mapping_pages(META_MAPPING(sbi),
  60 +                                               map->m_pblk, map->m_pblk);
  61
  62                 if (map->m_multidev_dio) {
  63                         block_t blk_addr = map->m_pblk;
  64 diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
  65 index 25d53617a50e6..5eca50e50e16b 100644
  66 --- a/fs/f2fs/segment.c
  67 +++ b/fs/f2fs/segment.c
  68 @@ -3622,6 +3622,9 @@ int f2fs_inplace_write_data(struct f2fs_io_info *fio)
  69                 goto drop_bio;
  70         }
  71
  72 +       invalidate_mapping_pages(META_MAPPING(sbi),
  73 +                               fio->new_blkaddr, fio->new_blkaddr);
  74 +
  75         stat_inc_inplace_blocks(fio->sbi);
  76
  77         if (fio->bio && !(SM_I(sbi)->ipu_policy & (1 << F2FS_IPU_NOCACHE)))
  78 --
  79 2.43.0
  80