I see. The old patch assumed compressed writeback should follow the
  F2FS_WB_CP_DATA path because the current code accounts it that way.

  After rechecking it, I agree that stopping checkpoint from this path
is not
  the right fix.

  I sent a separate patch to fix the accounting side instead:

  f2fs: account compressed writeback by raw folio type

  It derives the writeback count type from the raw pagecache folio, so normal
  compressed writeback stays in F2FS_WB_DATA, while gcing/CP-guaranteed raw
  folios still use F2FS_WB_CP_DATA.





On Tue, Jun 16, 2026 at 4:24 AM Jaegeuk Kim <[email protected]> wrote:
>
> On 05/25, Wenjie Qi wrote:
> > Compressed data writes are accounted as F2FS_WB_CP_DATA because they
> > write compressed pages through fio->compressed_page.  Their end_io path
> > should therefore have the same checkpoint-stop behavior as ordinary
> > F2FS_WB_CP_DATA writes.
> >
> > However, f2fs_compress_write_end_io() only records -EIO in the inode
> > mapping when the bio fails.  The filesystem can keep checkpointing after
> > that failure, so a later checkpoint may persist metadata that points to
> > compressed data blocks whose writeback failed.
> >
> > Stop checkpointing with STOP_CP_REASON_WRITE_FAIL for failed compressed
> > F2FS_WB_CP_DATA writes, matching the ordinary data write end_io path.
> >
> > Fixes: 4c8ff7095bef ("f2fs: support data compression")
> > Cc: [email protected]
> > Signed-off-by: Wenjie Qi <[email protected]>
> > ---
> >  fs/f2fs/compress.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c
> > index caf522d667d6..9b1501004456 100644
> > --- a/fs/f2fs/compress.c
> > +++ b/fs/f2fs/compress.c
> > @@ -1488,8 +1488,11 @@ void f2fs_compress_write_end_io(struct bio *bio, 
> > struct folio *folio)
> >                               f2fs_is_compressed_page(folio));
> >       int i;
> >
> > -     if (unlikely(bio->bi_status != BLK_STS_OK))
> > +     if (unlikely(bio->bi_status != BLK_STS_OK)) {
> >               mapping_set_error(cic->inode->i_mapping, -EIO);
> > +             if (type == F2FS_WB_CP_DATA)
>
> It turns out the type should not be F2FS_WB_CP_DATA.
>
> > +                     f2fs_stop_checkpoint(sbi, true, 
> > STOP_CP_REASON_WRITE_FAIL);
> > +     }
> >
> >       f2fs_compress_free_page(page);
> >
> > --
> > 2.43.0
> >
> >
> >
> > _______________________________________________
> > Linux-f2fs-devel mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel


_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Reply via email to