On Mon, May 17, 2021 at 01:11:25PM -0700, Mike Kravetz wrote:
> On 5/16/21 9:54 PM, Naoya Horiguchi wrote:
> > From: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>
> >
> > When hugetlb page fault (under overcommitting situation) and
> > memory_failure() race, VM_BUG_ON_PAGE() is triggered by the following race:
> >
> > CPU0: CPU1:
> >
> > gather_surplus_pages()
> > page = alloc_surplus_huge_page()
> > memory_failure_hugetlb()
> > get_hwpoison_page(page)
> > __get_hwpoison_page(page)
> > get_page_unless_zero(page)
> > zero = put_page_testzero(page)
> > VM_BUG_ON_PAGE(!zero, page)
> > enqueue_huge_page(h, page)
> > put_page(page)
> >
> > __get_hwpoison_page() only checks page refcount before taking additional
> > one for memory error handling, which is wrong because there's a time
> > window where compound pages have non-zero refcount during initialization.
> >
> > So makes __get_hwpoison_page() check page status a bit more for a few
> > types of compound pages. PageSlab() check is added because otherwise
> > "non anonymous thp" path is wrongly chosen.
> >
> > Fixes: ead07f6a867b ("mm/memory-failure: introduce get_hwpoison_page() for consistent refcount handling")
> > Signed-off-by: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>
> > Reported-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>
> > Cc: stable@xxxxxxxxxxxxxxx # 5.12+
> > ---
> > ChangeLog v4:
> > - all hugetlb related check in hugetlb_lock,
> > - fix build error with #ifdef CONFIG_HUGETLB_PAGE
> > ---
> > mm/memory-failure.c | 10 ++++++++++
> > 1 file changed, 10 insertions(+)
> >
> > diff --git v5.12/mm/memory-failure.c v5.12_patched/mm/memory-failure.c
> > index a3659619d293..761f982b6d7b 100644
> > --- v5.12/mm/memory-failure.c
> > +++ v5.12_patched/mm/memory-failure.c
> > @@ -1094,6 +1094,16 @@ static int page_action(struct page_state *ps, struct page *p,
> > static int __get_hwpoison_page(struct page *page)
> > {
> > struct page *head = compound_head(page);
> > + int ret = 0;
> > +
> > +#ifdef CONFIG_HUGETLB_PAGE
> > + spin_lock(&hugetlb_lock);
> > + if (PageHuge(head) && (HPageFreed(head) || HPageMigratable(head)))
> > + ret = get_page_unless_zero(head);
> > + spin_unlock(&hugetlb_lock);
> > + if (ret > 0)
> > + return ret;
> > +#endif
>
> I must be missing something.
>
> The above code makes sure the page is not in one of these transitive
> hugetlb states as mentioned in the commit message. It only attempts
> to take a reference on the page if it is not in one of these states.
>
> However, if it is in such a transitive state (!HPageFreed(head) &&
> !HPageMigratable(head)) we will fall through and execute the code:
>
> if (get_page_unless_zero(head)) {
> if (head == compound_head(page))
> return 1;
>
> So, it seems like we will always do a get_page_unless_zero() for
> PageHuge() pages?
Right, no need to fall through in such a case.
>
> Also, due to commit db71ef79b59b ("hugetlb: make free_huge_page irq
> safe") you need to make sure interrupts are disabled when taking
> hugetlb_lock.
Thanks, I'll rebase to latest mainline and comply with new semantics.
- Naoya
