Re: [PATCH v3] mm: use compare-exchange operation to set KASAN page tag

From: Andrey Konovalov
Date: Mon Jan 24 2022 - 12:25:48 EST


On Mon, Jan 24, 2022 at 6:22 PM Andrey Konovalov <andreyknvl@xxxxxxxxx> wrote:
>
> On Thu, Jan 20, 2022 at 3:02 AM Peter Collingbourne <pcc@xxxxxxxxxx> wrote:
> >
> > It has been reported that the tag setting operation on newly-allocated
> > pages can cause the page flags to be corrupted when performed
> > concurrently with other flag updates as a result of the use of
> > non-atomic operations. Fix the problem by using a compare-exchange
> > loop to update the tag.
> >
> > Signed-off-by: Peter Collingbourne <pcc@xxxxxxxxxx>
> > Link: https://linux-review.googlesource.com/id/I456b24a2b9067d93968d43b4bb3351c0cec63101
> > Fixes: 2813b9c02962 ("kasan, mm, arm64: tag non slab memory allocated via pagealloc")
> > Cc: stable@xxxxxxxxxxxxxxx
> > ---
> > v3:
> > - use try_cmpxchg() as suggested by Peter Zijlstra on another
> > patch
> >
> > v2:
> > - use READ_ONCE()
> >
> > include/linux/mm.h | 17 ++++++++++++-----
> > 1 file changed, 12 insertions(+), 5 deletions(-)
> >
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > index c768a7c81b0b..87473fe52c3f 100644
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -1531,11 +1531,18 @@ static inline u8 page_kasan_tag(const struct page *page)
> >
> > static inline void page_kasan_tag_set(struct page *page, u8 tag)
> > {
> > - if (kasan_enabled()) {
> > - tag ^= 0xff;
> > - page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
> > - page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
> > - }
> > + unsigned long old_flags, flags;
> > +
> > + if (!kasan_enabled())
> > + return;
> > +
> > + tag ^= 0xff;
> > + old_flags = READ_ONCE(page->flags);
> > + do {
> > + flags = old_flags;
> > + flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
> > + flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
> > + } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
> > }
> >
> > static inline void page_kasan_tag_reset(struct page *page)
> > --
> > 2.34.1.703.g22d0c6ccf7-goog
> >
>
> Reviewed-by: Andrey Konovalov <andreyknvl@xxxxxxxxx>
>
> FWIW, try_cmpxchg() doesn't seem to be doing annotated atomic accesses
> when accessing old_flags, so using READ_ONCE() in page_kasan_tag_set()
> seems pointless after all.

Ah, nevermind. For a second I thought that try_cmpxchg() is actually
accessing page->flags through its second argument.