On Mon, Feb 24, 2025 at 10:53:06PM -0800, Andrew Morton wrote: > On Mon, 24 Feb 2025 21:20:02 -0500 Mike Snitzer <snitzer@xxxxxxxxxx> wrote: > > > Add PF_KCOMPACTD flag and current_is_kcompactd() helper to check for > > it so nfs_release_folio() can skip calling nfs_wb_folio() from > > kcompactd. > > > > --- a/mm/compaction.c~a > +++ a/mm/compaction.c > @@ -3182,7 +3182,7 @@ static int kcompactd(void *p) > long default_timeout = msecs_to_jiffies(HPAGE_FRAG_CHECK_INTERVAL_MSEC); > long timeout = default_timeout; > > - tsk->flags |= PF_KCOMPACTD; > + current->flags |= PF_KCOMPACTD; > set_freezable(); > > pgdat->kcompactd_max_order = 0; > @@ -3239,7 +3239,7 @@ static int kcompactd(void *p) > pgdat->proactive_compact_trigger = false; > } > > - tsk->flags &= ~PF_KCOMPACTD; > + current->flags &= ~PF_KCOMPACTD; > > return 0; > } > > I am of course concerned about how well tested this was! Understandable, the patch was created and tested against v6.12.16. Sincere apologies, I didn't rebase before submitting. I'll rebase, retest and then send v3 today. Thanks, Mike
