Nouveau - Sep 2020 - [PATCH v3] mm/thp: fix __split_huge_pmd_locked() for migration PMD

A migrating transparent huge page has to already be unmapped. Otherwise,
the page could be modified while it is being copied to a new page and
data could be lost. The function __split_huge_pmd() checks for a PMD
migration entry before calling __split_huge_pmd_locked() leading one to
think that __split_huge_pmd_locked() can handle splitting a migrating PMD.
However, the code always increments the page->_mapcount and adjusts the
memory control group accounting assuming the page is mapped.
Also, if the PMD entry is a migration PMD entry, the call to
is_huge_zero_pmd(*pmd) is incorrect because it calls pmd_pfn(pmd) instead
of migration_entry_to_pfn(pmd_to_swp_entry(pmd)).
Fix these problems by checking for a PMD migration entry.

Fixes: 84c3fc4e9c56 ("mm: thp: check pmd migration entry in common
path")
cc: stable at vger.kernel.org # 4.14+
Signed-off-by: Ralph Campbell <rcampbell at nvidia.com>
Reviewed-by: Yang Shi <shy828301 at gmail.com>
Reviewed-by: Zi Yan <ziy at nvidia.com>
---

No changes in v3 to this patch, just added reviewed-by and fixes to the
change log and sending this as a separate patch from the rest of the
series ("mm/hmm/nouveau: add THP migration to migrate_vma_*").
I'll hold off resending the series without this patch unless there are
changes needed.

 mm/huge_memory.c | 42 +++++++++++++++++++++++-------------------
 1 file changed, 23 insertions(+), 19 deletions(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 2a468a4acb0a..606d712d9505 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2023,7 +2023,7 @@ static void __split_huge_pmd_locked(struct vm_area_struct
*vma, pmd_t *pmd,
 		put_page(page);
 		add_mm_counter(mm, mm_counter_file(page), -HPAGE_PMD_NR);
 		return;
-	} else if (is_huge_zero_pmd(*pmd)) {
+	} else if (pmd_trans_huge(*pmd) && is_huge_zero_pmd(*pmd)) {
 		/*
 		 * FIXME: Do we want to invalidate secondary mmu by calling
 		 * mmu_notifier_invalidate_range() see comments below inside
@@ -2117,30 +2117,34 @@ static void __split_huge_pmd_locked(struct
vm_area_struct *vma, pmd_t *pmd,
 		pte = pte_offset_map(&_pmd, addr);
 		BUG_ON(!pte_none(*pte));
 		set_pte_at(mm, addr, pte, entry);
-		atomic_inc(&page[i]._mapcount);
-		pte_unmap(pte);
-	}
-
-	/*
-	 * Set PG_double_map before dropping compound_mapcount to avoid
-	 * false-negative page_mapped().
-	 */
-	if (compound_mapcount(page) > 1 && !TestSetPageDoubleMap(page)) {
-		for (i = 0; i < HPAGE_PMD_NR; i++)
+		if (!pmd_migration)
 			atomic_inc(&page[i]._mapcount);
+		pte_unmap(pte);
 	}
 
-	lock_page_memcg(page);
-	if (atomic_add_negative(-1, compound_mapcount_ptr(page))) {
-		/* Last compound_mapcount is gone. */
-		__dec_lruvec_page_state(page, NR_ANON_THPS);
-		if (TestClearPageDoubleMap(page)) {
-			/* No need in mapcount reference anymore */
+	if (!pmd_migration) {
+		/*
+		 * Set PG_double_map before dropping compound_mapcount to avoid
+		 * false-negative page_mapped().
+		 */
+		if (compound_mapcount(page) > 1 &&
+		    !TestSetPageDoubleMap(page)) {
 			for (i = 0; i < HPAGE_PMD_NR; i++)
-				atomic_dec(&page[i]._mapcount);
+				atomic_inc(&page[i]._mapcount);
+		}
+
+		lock_page_memcg(page);
+		if (atomic_add_negative(-1, compound_mapcount_ptr(page))) {
+			/* Last compound_mapcount is gone. */
+			__dec_lruvec_page_state(page, NR_ANON_THPS);
+			if (TestClearPageDoubleMap(page)) {
+				/* No need in mapcount reference anymore */
+				for (i = 0; i < HPAGE_PMD_NR; i++)
+					atomic_dec(&page[i]._mapcount);
+			}
 		}
+		unlock_page_memcg(page);
 	}
-	unlock_page_memcg(page);
 
 	smp_wmb(); /* make pte visible before pmd */
 	pmd_populate(mm, pmd, pgtable);
-- 
2.20.1

Hi

[This is an automated email]

This commit has been processed because it contains a "Fixes:" tag
fixing commit: 84c3fc4e9c56 ("mm: thp: check pmd migration entry in common
path").

The bot has tested the following trees: v5.8.6, v5.4.62, v4.19.143, v4.14.196.

v5.8.6: Build OK!
v5.4.62: Failed to apply! Possible dependencies:
    0d1c20722ab3 ("mm: memcontrol: switch to native NR_FILE_PAGES and
NR_SHMEM counters")
    3fba69a56e16 ("mm: memcontrol: drop @compound parameter from memcg
charging API")
    468c398233da ("mm: memcontrol: switch to native NR_ANON_THPS
counter")
    72981e0e7b60 ("userfaultfd: wp: add UFFDIO_COPY_MODE_WP")
    83d116c53058 ("mm: fix double page fault on arm64 if PTE_AF is
cleared")
    85b9f46e8ea4 ("mm, thp: track fallbacks due to failed memcg charges
separately")
    92855270ff08 ("mm/memcontrol.c: cleanup some useless code")
    be5d0a74c62d ("mm: memcontrol: switch to native NR_ANON_MAPPED
counter")
    c23a0c99793f ("mm/migrate: clean up some minor coding style")
    dcdf11ee1441 ("mm, shmem: add vmstat for hugepage fallback")
    f4129ea3591a ("mm: fix NUMA node file count error in
replace_page_cache()")
    ffe945e633b5 ("khugepaged: do not stop collapse if less than half PTEs
are referenced")

v4.19.143: Failed to apply! Possible dependencies:
    0ac261042084 ("x86/irq/64: Init hardirq_stack_ptr during CPU
hotplug")
    0d1c20722ab3 ("mm: memcontrol: switch to native NR_FILE_PAGES and
NR_SHMEM counters")
    117ed4548541 ("x86/irq/64: Remove stack overflow debug code")
    231c4846b106 ("x86/irq/32: Make irq stack a character array")
    30842211506e ("x86/exceptions: Remove unused stack defines on
32bit")
    39656e83dab9 ("mm: lift the x86_32 PAE version of gup_get_pte to common
code")
    468c398233da ("mm: memcontrol: switch to native NR_ANON_THPS
counter")
    4f44b8f0b33b ("x86/irq/64: Remove a hardcoded irq_stack_union
access")
    66c7ceb47f62 ("x86/irq/32: Handle irq stack allocation failure
proper")
    758a2e312228 ("x86/irq/64: Rename irq_stack_ptr to
hardirq_stack_ptr")
    7ac870747988 ("x86/vdso: Switch to generic vDSO implementation")
    99cb0dbd47a1 ("mm,thp: add read-only THP support for (non-shmem)
FS")
    a754fe2b76d1 ("x86/irq/32: Rename hard/softirq_stack to
hard/softirq_stack_ptr")
    aa641c287b2f ("x86/irq/32: Define IRQ_STACK_SIZE")
    be5d0a74c62d ("mm: memcontrol: switch to native NR_ANON_MAPPED
counter")
    df835e7083be ("x86/irq/64: Sanitize the top/bottom confusion")
    e6401c130931 ("x86/irq/64: Split the IRQ stack into its own
pages")

v4.14.196: Failed to apply! Possible dependencies:
    050e9baa9dc9 ("Kbuild: rename CC_STACKPROTECTOR[_STRONG] config
variables")
    0d1c20722ab3 ("mm: memcontrol: switch to native NR_FILE_PAGES and
NR_SHMEM counters")
    117ed4548541 ("x86/irq/64: Remove stack overflow debug code")
    152e93af3cfe ("mm, thp: Do not make pmd/pud dirty without a
reason")
    2a61f4747eea ("stack-protector: test compiler capability in Kconfig and
drop AUTO mode")
    2b8383927525 ("Makefile: move stack-protector compiler breakage test
earlier")
    2bc2f688fdf8 ("Makefile: move stack-protector availability out of
Kconfig")
    39656e83dab9 ("mm: lift the x86_32 PAE version of gup_get_pte to common
code")
    44c6dc940b19 ("Makefile: introduce CONFIG_CC_STACKPROTECTOR_AUTO")
    4645b9fe84bf ("mm/mmu_notifier: avoid call to invalidate_range() in
range_end()")
    468c398233da ("mm: memcontrol: switch to native NR_ANON_THPS
counter")
    59c66c5f8c4f ("mm: factor out page cache page freeing into a separate
function")
    5ecc4d852c03 ("mm: factor out checks and accounting from
__delete_from_page_cache()")
    76253fbc8fbf ("mm: move accounting updates before
page_cache_tree_delete()")
    7ac870747988 ("x86/vdso: Switch to generic vDSO implementation")
    8373b7d9d174 ("Documentation: kconfig: add recommended way to describe
compiler support")
    99cb0dbd47a1 ("mm,thp: add read-only THP support for (non-shmem)
FS")
    be5d0a74c62d ("mm: memcontrol: switch to native NR_ANON_MAPPED
counter")
    e6401c130931 ("x86/irq/64: Split the IRQ stack into its own
pages")
    fadae2953072 ("thp: use mm_file_counter to determine update which rss
counter")


NOTE: The patch will not be queued to stable trees until it is upstream.

How should we proceed with this patch?

-- 
Thanks
Sasha