Re: [v3 PATCH 2/5] mm: filemap: check if THP has hwpoisoned subpage for PMD page fault

[Peter Xu]

On Thu, Oct 07, 2021 at 02:28:35PM -0700, Yang Shi wrote:
> On Wed, Oct 6, 2021 at 4:57 PM Yang Shi <shy828301@xxxxxxxxx> wrote:
> >
> > On Wed, Oct 6, 2021 at 1:15 PM Peter Xu <peterx@xxxxxxxxxx> wrote:
> > >
> > > On Thu, Sep 30, 2021 at 02:53:08PM -0700, Yang Shi wrote:
> > > > @@ -1148,8 +1148,12 @@ static int __get_hwpoison_page(struct page *page)
> > > >               return -EBUSY;
> > > >
> > > >       if (get_page_unless_zero(head)) {
> > > > -             if (head == compound_head(page))
> > > > +             if (head == compound_head(page)) {
> > > > +                     if (PageTransHuge(head))
> > > > +                             SetPageHasHWPoisoned(head);
> > > > +
> > > >                       return 1;
> > > > +             }
> > > >
> > > >               pr_info("Memory failure: %#lx cannot catch tail\n",
> > > >                       page_to_pfn(page));
> > >
> > > Sorry for the late comments.
> > >
> > > I'm wondering whether it's ideal to set this bit here, as get_hwpoison_page()
> > > sounds like a pure helper to get a refcount out of a sane hwpoisoned page.  I'm
> > > afraid there can be side effect that we set this without being noticed, so I'm
> > > also wondering we should keep it in memory_failure().
> > >
> > > Quotting comments for get_hwpoison_page():
> > >
> > >  * get_hwpoison_page() takes a page refcount of an error page to handle memory
> > >  * error on it, after checking that the error page is in a well-defined state
> > >  * (defined as a page-type we can successfully handle the memor error on it,
> > >  * such as LRU page and hugetlb page).
> > >
> > > For example, I see that both unpoison_memory() and soft_offline_page() will
> > > call it too, does it mean that we'll also set the bits e.g. even when we want
> > > to inject an unpoison event too?
> >
> > unpoison_memory() should be not a problem since it will just bail out
> > once THP is met as the comment says:
> >
> > /*
> > * unpoison_memory() can encounter thp only when the thp is being
> > * worked by memory_failure() and the page lock is not held yet.
> > * In such case, we yield to memory_failure() and make unpoison fail.
> > */
> >
> >
> > And I think we should set the flag for soft offline too, right? The
> > soft offline does set the hwpoison flag for the corrupted sub page and
> > doesn't split file THP, so it should be captured by page fault as
> > well. And yes for poison injection.
> 
> Err... I must be blind. The soft offline does *NOT* set hwpoison flag
> for any page. So your comment does stand. The flag should be set
> outside get_hwpoison_page().

I saw that page_handle_poison() sets it, so perhaps we do need to take care of
soft offline?  Though I still think the extra bit should be set outside of the
get_hwpoison_page() function.

Another thing is I noticed soft_offline_in_use_page() will still ignore file
backed split.  I'm not sure whether it means we'd better also handle that case
as well, so shmem thp can be split there too?

-- 
Peter Xu