Re: [Intel-gfx] mm/huge_memory: do not clobber swp_entry_t during THP split

[Tvrtko Ursulin]

On 24/10/2022 15:23, Mel Gorman wrote:
On Mon, Oct 24, 2022 at 02:04:50PM +0100, Tvrtko Ursulin wrote:

Hi Mel, mm experts,

With 6.1-rc2 we started hitting the WARN_ON added in 71e2d666ef85 ("mm/huge_memory: 
do not clobber swp_entry_t during THP split") in i915 automated CI:


Thanks for the report. As shmem pages pages are allocated via vma_alloc_folio
and are compound pages, can you try the following patch please?  If it
still triggers, please post the new oops as it'll include the tail page
information.

--8<--
From: Hugh Dickins <hu...@google.com>
Subject: [PATCH] mm: prep_compound_tail() clear page->private

Although page allocation always clears page->private in the first page
or head page of an allocation, it has never made a point of clearing
page->private in the tails (though 0 is often what is already there).

But now commit 71e2d666ef85 ("mm/huge_memory: do not clobber swp_entry_t
during THP split") issues a warning when page_tail->private is found to
be non-0 (unless it's swapcache).

Change that warning to dump page_tail (which also dumps head), instead
of just the head: so far we have seen dead000000000122, dead000000000003,
dead000000000001 or 0000000000000002 in the raw output for tail private.

We could just delete the warning, but today's consensus appears to want
page->private to be 0, unless there's a good reason for it to be set:
so now clear it in prep_compound_tail() (more general than just for THP;
but not for high order allocation, which makes no pass down the tails).

Fixes: 71e2d666ef85 ("mm/huge_memory: do not clobber swp_entry_t during THP 
split")
Signed-off-by: Hugh Dickins <hu...@google.com>
Cc: Mel Gorman <mgor...@techsingularity.net>
Cc: Matthew Wilcox (Oracle) <wi...@infradead.org>
Cc: <sta...@vger.kernel.org>
---
  mm/huge_memory.c | 2 +-
  mm/page_alloc.c  | 1 +
  2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 03fc7e5edf07..561a42567477 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2462,7 +2462,7 @@ static void __split_huge_page_tail(struct page *head, int 
tail,
         * Fix up and warn once if private is unexpectedly set.
         */
        if (!folio_test_swapcache(page_folio(head))) {
-               VM_WARN_ON_ONCE_PAGE(page_tail->private != 0, head);
+               VM_WARN_ON_ONCE_PAGE(page_tail->private != 0, page_tail);
                page_tail->private = 0;
        }
  
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index b5a6c815ae28..218b28ee49ed 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -807,6 +807,7 @@ static void prep_compound_tail(struct page *head, int 
tail_idx)
  
  	p->mapping = TAIL_MAPPING;
        set_compound_head(p, head);
+       set_page_private(p, 0);
  }
  
  void prep_compound_page(struct page *page, unsigned int order)

The patch seems to fix our CI runs. Is it considered final version? If 
so I can temporarily put it in until it arrives via the next rc - 
assuming that would be the flow from upstream pov?

Regards,

Tvrtko