[PATCH v4] F2FS: invalidate META_MAPPING before IPU/DIO write

[PATCH v4] F2FS: invalidate META_MAPPING before IPU/DIO write

[Hyeong-Jun Kim]

Encrypted pages during GC are read and cached in META_MAPPING.
However, due to cached pages in META_MAPPING, there is an issue where
newly written pages are lost by IPU or DIO writes.

Thread A - f2fs_gc()            Thread B
/* phase 3 */
down_write(i_gc_rwsem)
ra_data_block()       ---- (a)
up_write(i_gc_rwsem)
                                f2fs_direct_IO() :
                                 - down_read(i_gc_rwsem)
                                 - __blockdev_direct_io()
                                 - get_data_block_dio_write()
                                 - f2fs_dio_submit_bio()  ---- (b)
                                 - up_read(i_gc_rwsem)
/* phase 4 */
down_write(i_gc_rwsem)
move_data_block()     ---- (c)
up_write(i_gc_rwsem)

(a) In phase 3 of f2fs_gc(), up-to-date page is read from storage and
    cached in META_MAPPING.
(b) In thread B, writing new data by IPU or DIO write on same blkaddr as
    read in (a). cached page in META_MAPPING become out-dated.
(c) In phase 4 of f2fs_gc(), out-dated page in META_MAPPING is copied to
    new blkaddr. In conclusion, the newly written data in (b) is lost.

To address this issue, invalidating pages in META_MAPPING before IPU or
DIO write.

Fixes: 6aa58d8ad20a ("f2fs: readahead encrypted block during GC")
Signed-off-by: Hyeong-Jun Kim <hj514.kim@samsung.com>

---
v4:
 - Fix change log in patch description
v3:
 - Use sbi instead of fio->sbi in f2fs_inplace_write_data()
v2:
 - Update patch description with race condition
---
 fs/f2fs/data.c    | 2 ++
 fs/f2fs/segment.c | 3 +++
 2 files changed, 5 insertions(+)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 74e1a350c1d8..9f754aaef558 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -1708,6 +1708,8 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
 		 */
 		f2fs_wait_on_block_writeback_range(inode,
 						map->m_pblk, map->m_len);
+		invalidate_mapping_pages(META_MAPPING(sbi),
+						map->m_pblk, map->m_pblk);
 
 		if (map->m_multidev_dio) {
 			block_t blk_addr = map->m_pblk;
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 526423fe84ce..df9ed75f0b7a 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -3652,6 +3652,9 @@ int f2fs_inplace_write_data(struct f2fs_io_info *fio)
 		goto drop_bio;
 	}
 
+	invalidate_mapping_pages(META_MAPPING(sbi),
+				fio->new_blkaddr, fio->new_blkaddr);
+
 	stat_inc_inplace_blocks(fio->sbi);
 
 	if (fio->bio && !(SM_I(sbi)->ipu_policy & (1 << F2FS_IPU_NOCACHE)))
-- 
2.25.1

Re: [PATCH v4] F2FS: invalidate META_MAPPING before IPU/DIO write

[Chao Yu]

On 2021/11/2 15:10, Hyeong-Jun Kim wrote:
> Encrypted pages during GC are read and cached in META_MAPPING.
> However, due to cached pages in META_MAPPING, there is an issue where
> newly written pages are lost by IPU or DIO writes.
> 
> Thread A - f2fs_gc()            Thread B
> /* phase 3 */
> down_write(i_gc_rwsem)
> ra_data_block()       ---- (a)
> up_write(i_gc_rwsem)
>                                  f2fs_direct_IO() :
>                                   - down_read(i_gc_rwsem)
>                                   - __blockdev_direct_io()
>                                   - get_data_block_dio_write()
>                                   - f2fs_dio_submit_bio()  ---- (b)
>                                   - up_read(i_gc_rwsem)
> /* phase 4 */
> down_write(i_gc_rwsem)
> move_data_block()     ---- (c)
> up_write(i_gc_rwsem)
> 
> (a) In phase 3 of f2fs_gc(), up-to-date page is read from storage and
>      cached in META_MAPPING.
> (b) In thread B, writing new data by IPU or DIO write on same blkaddr as
>      read in (a). cached page in META_MAPPING become out-dated.
> (c) In phase 4 of f2fs_gc(), out-dated page in META_MAPPING is copied to
>      new blkaddr. In conclusion, the newly written data in (b) is lost.
> 
> To address this issue, invalidating pages in META_MAPPING before IPU or
> DIO write.
> 
> Fixes: 6aa58d8ad20a ("f2fs: readahead encrypted block during GC")
> Signed-off-by: Hyeong-Jun Kim <hj514.kim@samsung.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,