If background GC is running very frequently by gc_urgent, the next GC
can see unwritten data generated by the previous GC, which incurs to write
stale corrupted data in disk.
Cc: <[email protected]>
Signed-off-by: Jaegeuk Kim <[email protected]>
---
fs/f2fs/gc.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index a07241fb8537..49fea136a3f3 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -658,6 +658,9 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
fio.page = page;
fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
+ /* wait writeback before reading out */
+ f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
+
fio.encrypted_page = f2fs_pagecache_get_page(META_MAPPING(sbi),
dn.data_blkaddr,
FGP_LOCK | FGP_CREAT, GFP_NOFS);
@@ -755,6 +758,9 @@ static int move_data_block(struct inode *inode, block_t bidx,
fio.page = page;
fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
+ /* wait writeback before reading out */
+ f2fs_wait_on_block_writeback(inode, fio.old_blkaddr);
+
if (lfs_mode)
down_write(&fio.sbi->io_order_lock);
--
2.19.0.605.g01d371f741-goog
Let me drop this, since I missed there was a patch to introdcued this. :)
On 11/06, Jaegeuk Kim wrote:
> If background GC is running very frequently by gc_urgent, the next GC
> can see unwritten data generated by the previous GC, which incurs to write
> stale corrupted data in disk.
>
> Cc: <[email protected]>
> Signed-off-by: Jaegeuk Kim <[email protected]>
> ---
> fs/f2fs/gc.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> index a07241fb8537..49fea136a3f3 100644
> --- a/fs/f2fs/gc.c
> +++ b/fs/f2fs/gc.c
> @@ -658,6 +658,9 @@ static int ra_data_block(struct inode *inode, pgoff_t index)
> fio.page = page;
> fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
>
> + /* wait writeback before reading out */
> + f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
> +
> fio.encrypted_page = f2fs_pagecache_get_page(META_MAPPING(sbi),
> dn.data_blkaddr,
> FGP_LOCK | FGP_CREAT, GFP_NOFS);
> @@ -755,6 +758,9 @@ static int move_data_block(struct inode *inode, block_t bidx,
> fio.page = page;
> fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
>
> + /* wait writeback before reading out */
> + f2fs_wait_on_block_writeback(inode, fio.old_blkaddr);
> +
> if (lfs_mode)
> down_write(&fio.sbi->io_order_lock);
>
> --
> 2.19.0.605.g01d371f741-goog