Re: [PATCH 1/1] mm: remove extra check for VMA_LOCK_OFFSET when read-locking a vma

[Suren Baghdasaryan]

On Tue, Jan 7, 2025 at 4:28 AM Wei Yang <richard.weiyang@xxxxxxxxx> wrote:
>
> On Mon, Jan 06, 2025 at 07:04:15PM -0800, Suren Baghdasaryan wrote:
> >Since we limit vm_refcnt at VMA_REF_LIMIT and it's smaller than
> >VMA_LOCK_OFFSET, there is no need to check again if VMA_LOCK_OFFSET bit
> >is set. Remove the extra check and add a clarifying comment.
> >
> >Fixes: e8f32ff00a66 ("mm: replace vm_lock and detached flag with a reference count")
> >Suggested-by: Wei Yang <richard.weiyang@xxxxxxxxx>
> >Signed-off-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
>
> Reviewed-by: Wei Yang <richard.weiyang@xxxxxxxxx>

Since I have to respin v8, I'll fold this fix into the original patch.

>
> >---
> >Applies over mm-unstable
> >
> > include/linux/mm.h | 9 ++++++---
> > 1 file changed, 6 insertions(+), 3 deletions(-)
> >
> >diff --git a/include/linux/mm.h b/include/linux/mm.h
> >index 486638d22fc6..b5f262fc7dc5 100644
> >--- a/include/linux/mm.h
> >+++ b/include/linux/mm.h
> >@@ -747,7 +747,11 @@ static inline bool vma_start_read(struct mm_struct *mm, struct vm_area_struct *v
> >
> >
> >       rwsem_acquire_read(&vma->vmlock_dep_map, 0, 0, _RET_IP_);
> >-      /* Limit at VMA_REF_LIMIT to leave one count for a writer */
> >+      /*
> >+       * Limit at VMA_REF_LIMIT to leave one count for a writer.
> >+       * If VMA_LOCK_OFFSET is set, __refcount_inc_not_zero_limited() will fail
> >+       * because VMA_REF_LIMIT is less than VMA_LOCK_OFFSET.
> >+       */
> >       if (unlikely(!__refcount_inc_not_zero_limited(&vma->vm_refcnt, &oldcnt,
> >                                                     VMA_REF_LIMIT))) {
> >               rwsem_release(&vma->vmlock_dep_map, _RET_IP_);
> >@@ -766,8 +770,7 @@ static inline bool vma_start_read(struct mm_struct *mm, struct vm_area_struct *v
> >        * after it has been unlocked.
> >        * This pairs with RELEASE semantics in vma_end_write_all().
> >        */
> >-      if (unlikely(oldcnt & VMA_LOCK_OFFSET ||
> >-                   vma->vm_lock_seq == raw_read_seqcount(&mm->mm_lock_seq))) {
> >+      if (unlikely(vma->vm_lock_seq == raw_read_seqcount(&mm->mm_lock_seq))) {
> >               vma_refcount_put(vma);
> >               return false;
> >       }
> >
> >base-commit: f349e79bfbf3abfade8011797ff6d0d47b67dab7
> >--
> >2.47.1.613.gc27f4b7a9f-goog
>
> --
> Wei Yang
> Help you, Help me