Re: [PATCH v2 1/2] f2fs: complete dropbehind write bios in safe task context

[Wenjie Qi]

  Hi Barry,

  bio_in_atomic() is not in the current f2fs dev-test base yet, so v2
  carries the same predicate locally. I will replace it with
  bio_in_atomic() once the block pull is merged into the base.

  I also checked Alexandre's v3 series. Its deferred path is currently
  specific to swap-cache folios. folio_end_writeback() detects a
  dropbehind swap-cache folio and calls swap_writeback_dropbehind_folio(),
  while the per-CPU lists and swap_dropbehind_wq live in mm/swap_state.c.
  The non-swap path still calls folio_end_dropbehind() directly, so the
  series does not yet provide a common deferred path for file-backed
  folios.

  This F2FS patch reuses the existing sbi->wq rather than allocating a new
  workqueue, although the completion deferral remains F2FS-specific.

  Regards,
  Wenjie

On Thu, Aug 20, 2026 at 7:45 PM Barry Song <baohua@xxxxxxxxxx> wrote:
>
> On Thu, Aug 20, 2026 at 3:21 PM Wenjie Qi <qwjhust@xxxxxxxxx> wrote:
> >
> > Buffered RWF_DONTCACHE writes invalidate dropbehind folios from writeback
> > completion. Keep normal and dropbehind folios in separate write bios, and
> > defer a dropbehind bio to sbi->wq unless completion runs in
> > preemptible task context.
> >
> > Use the same eligibility conditions as the proposed bio_in_atomic() helper.
> > Unlike !in_task(), this also covers disabled preemption and preemptible RCU
> > read-side critical sections.
> >
> > Keep the existing large-ATC deferral unchanged.
> >
> > Link: https://lore.kernel.org/linux-mm/20260730-blk-dontcache-v7-0-3e8e6850068d@xxxxxxxxxxxx/
> > Signed-off-by: Wenjie Qi <qiwenjie@xxxxxxxxxx>
> > ---
> >  fs/f2fs/data.c | 50 ++++++++++++++++++++++++++++++++++++++++++--------
> >  1 file changed, 42 insertions(+), 8 deletions(-)
> >
> > diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> > index 6ae0eb37d20..774a3e2d8e3 100644
> > --- a/fs/f2fs/data.c
> > +++ b/fs/f2fs/data.c
> > @@ -21,6 +21,7 @@
> >  #include <linux/fiemap.h>
> >  #include <linux/iomap.h>
> >  #include <linux/fserror.h>
> > +#include <linux/rcupdate.h>
> >
> >  #include "f2fs.h"
> >  #include "node.h"
> > @@ -43,9 +44,29 @@ struct f2fs_folio_state {
> >
> >  struct f2fs_bio {
> >         struct work_struct work;
> > +       bool dropbehind;
> >         struct bio bio;
> >  };
> >
> > +static struct f2fs_bio *to_f2fs_bio(struct bio *bio)
> > +{
> > +       return container_of(bio, struct f2fs_bio, bio);
> > +}
> > +
> > +/* Keep in sync with the proposed block-layer bio_in_atomic(). */
> > +static bool f2fs_bio_in_atomic(void)
> > +{
> > +#ifdef CONFIG_PREEMPTION
> > +       if (rcu_preempt_depth())
> > +               return true;
> > +#endif
> > +#ifndef CONFIG_PREEMPT_COUNT
> > +       return true;
> > +#else
> > +       return !preemptible();
> > +#endif
> > +}
> > +
>
> Hi Wenjie,
>
> I see that this [1] has already been included in the pull request [2],
> so perhaps you can use bio_in_atomic() directly once it lands.
>
> Also, I feel that something is quite odd with mm/filemap.c.
> It shouldn't require everyone to reinvent their own workqueues, as I
> mentioned here [3]. I also noticed that Alexandre is adding a workqueue
> in filemap.c[4], which might be extended to file page cache handling as
> well in the future.
>
> [1] https://lore.kernel.org/all/20260730-blk-dontcache-v7-1-3e8e6850068d@xxxxxxxxxxxx/
> [2] https://lore.kernel.org/all/8787a176-6f4b-47fc-a309-867acf9fbdff@xxxxxxxxx/
> [3] https://lore.kernel.org/all/CAGsJ_4wUvLoKGafR3U-ji1kUUaz6K_QdN6QxMT-apSaO5tXp0A@xxxxxxxxxxxxxx/
> [4] https://lore.kernel.org/all/20260818163221.589352-3-alex@xxxxxxxx/
>
> Thanks
> Barry