[PATCH v4 0/3] Fix CONT-PTE/PMD size hugetlb issue when unmapping or migrating

[PATCH v4 0/3] Fix CONT-PTE/PMD size hugetlb issue when unmapping or migrating

[Baolin Wang]

Hi,

Now migrating a hugetlb page or unmapping a poisoned hugetlb page, we'll
use ptep_clear_flush() and set_pte_at() to nuke the page table entry
and remap it, and this is incorrect for CONT-PTE or CONT-PMD size hugetlb
page, which will cause potential data consistent issue. This patch set
will change to use hugetlb related APIs to fix this issue, please find
details in each patch. Thanks.

Note: Mike pointed out the huge_ptep_get() will only return the one specific
value, and it would not take into account the dirty or young bits of CONT-PTE/PMDs
like the huge_ptep_get_and_clear() [1]. This inconsistent issue is not introduced
by this patch set, and will address this issue in another thread [2]. Meanwhile
the uffd for hugetlb case [3] pointed by Gerald also need another patch to address.

[1] https://lore.kernel.org/linux-mm/85bd80b4-b4fd-0d3f-a2e5-149559f2f387@oracle.com/
[2] https://lore.kernel.org/all/cover.1651998586.git.baolin.wang@linux.alibaba.com/
[3] https://lore.kernel.org/linux-mm/20220503120343.6264e126@thinkpad/

Changes from v3:
 - Fix building errors for !CONFIG_MMU.

Changes from v2:
 - Collect reviewed tags from Muchun and Mike.
 - Drop the unnecessary casting in hugetlb.c.
 - Fix building errors with adding dummy functions for !CONFIG_HUGETLB_PAGE.

Changes from v1:
 - Add acked tag from Mike.
 - Update some commit message.
 - Add VM_BUG_ON in try_to_unmap() for hugetlb case.
 - Add an explict void casting for huge_ptep_clear_flush() in hugetlb.c.

Baolin Wang (3):
  mm: change huge_ptep_clear_flush() to return the original pte
  mm: rmap: Fix CONT-PTE/PMD size hugetlb issue when migration
  mm: rmap: Fix CONT-PTE/PMD size hugetlb issue when unmapping

 arch/arm64/include/asm/hugetlb.h   |  4 +--
 arch/arm64/mm/hugetlbpage.c        | 12 +++-----
 arch/ia64/include/asm/hugetlb.h    |  5 +--
 arch/mips/include/asm/hugetlb.h    |  9 ++++--
 arch/parisc/include/asm/hugetlb.h  |  5 +--
 arch/powerpc/include/asm/hugetlb.h |  9 ++++--
 arch/s390/include/asm/hugetlb.h    |  6 ++--
 arch/sh/include/asm/hugetlb.h      |  5 +--
 arch/sparc/include/asm/hugetlb.h   |  5 +--
 include/asm-generic/hugetlb.h      |  4 +--
 include/linux/hugetlb.h            | 11 +++++++
 mm/rmap.c                          | 63 ++++++++++++++++++++++++--------------
 12 files changed, 87 insertions(+), 51 deletions(-)

-- 
1.8.3.1

[PATCH v4 1/3] mm: change huge_ptep_clear_flush() to return the original pte

[Baolin Wang]

It is incorrect to use ptep_clear_flush() to nuke a hugetlb page
table when unmapping or migrating a hugetlb page, and will change
to use huge_ptep_clear_flush() instead in the following patches.

So this is a preparation patch, which changes the huge_ptep_clear_flush()
to return the original pte to help to nuke a hugetlb page table.

Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
---
 arch/arm64/include/asm/hugetlb.h   |  4 ++--
 arch/arm64/mm/hugetlbpage.c        | 12 +++++-------
 arch/ia64/include/asm/hugetlb.h    |  5 +++--
 arch/mips/include/asm/hugetlb.h    |  9 ++++++---
 arch/parisc/include/asm/hugetlb.h  |  5 +++--
 arch/powerpc/include/asm/hugetlb.h |  9 ++++++---
 arch/s390/include/asm/hugetlb.h    |  6 +++---
 arch/sh/include/asm/hugetlb.h      |  5 +++--
 arch/sparc/include/asm/hugetlb.h   |  5 +++--
 include/asm-generic/hugetlb.h      |  4 ++--
 10 files changed, 36 insertions(+), 28 deletions(-)

diff --git a/arch/arm64/include/asm/hugetlb.h b/arch/arm64/include/asm/hugetlb.h
index 1242f71..616b2ca 100644
--- a/arch/arm64/include/asm/hugetlb.h
+++ b/arch/arm64/include/asm/hugetlb.h
@@ -39,8 +39,8 @@ extern pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
 extern void huge_ptep_set_wrprotect(struct mm_struct *mm,
 				    unsigned long addr, pte_t *ptep);
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-extern void huge_ptep_clear_flush(struct vm_area_struct *vma,
-				  unsigned long addr, pte_t *ptep);
+extern pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+				   unsigned long addr, pte_t *ptep);
 #define __HAVE_ARCH_HUGE_PTE_CLEAR
 extern void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
 			   pte_t *ptep, unsigned long sz);
diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
index cbace1c..ca8e65c 100644
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -486,19 +486,17 @@ void huge_ptep_set_wrprotect(struct mm_struct *mm,
 		set_pte_at(mm, addr, ptep, pfn_pte(pfn, hugeprot));
 }
 
-void huge_ptep_clear_flush(struct vm_area_struct *vma,
-			   unsigned long addr, pte_t *ptep)
+pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+			    unsigned long addr, pte_t *ptep)
 {
 	size_t pgsize;
 	int ncontig;
 
-	if (!pte_cont(READ_ONCE(*ptep))) {
-		ptep_clear_flush(vma, addr, ptep);
-		return;
-	}
+	if (!pte_cont(READ_ONCE(*ptep)))
+		return ptep_clear_flush(vma, addr, ptep);
 
 	ncontig = find_num_contig(vma->vm_mm, addr, ptep, &pgsize);
-	clear_flush(vma->vm_mm, addr, ptep, pgsize, ncontig);
+	return get_clear_flush(vma->vm_mm, addr, ptep, pgsize, ncontig);
 }
 
 static int __init hugetlbpage_init(void)
diff --git a/arch/ia64/include/asm/hugetlb.h b/arch/ia64/include/asm/hugetlb.h
index 7e46ebd..026ead4 100644
--- a/arch/ia64/include/asm/hugetlb.h
+++ b/arch/ia64/include/asm/hugetlb.h
@@ -23,9 +23,10 @@ static inline int is_hugepage_only_range(struct mm_struct *mm,
 #define is_hugepage_only_range is_hugepage_only_range
 
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long addr, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
 {
+	return *ptep;
 }
 
 #include <asm-generic/hugetlb.h>
diff --git a/arch/mips/include/asm/hugetlb.h b/arch/mips/include/asm/hugetlb.h
index c214440..fd69c88 100644
--- a/arch/mips/include/asm/hugetlb.h
+++ b/arch/mips/include/asm/hugetlb.h
@@ -43,16 +43,19 @@ static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long addr, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
 {
+	pte_t pte;
+
 	/*
 	 * clear the huge pte entry firstly, so that the other smp threads will
 	 * not get old pte entry after finishing flush_tlb_page and before
 	 * setting new huge pte entry
 	 */
-	huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
+	pte = huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
 	flush_tlb_page(vma, addr);
+	return pte;
 }
 
 #define __HAVE_ARCH_HUGE_PTE_NONE
diff --git a/arch/parisc/include/asm/hugetlb.h b/arch/parisc/include/asm/hugetlb.h
index a69cf9e..f7f078c 100644
--- a/arch/parisc/include/asm/hugetlb.h
+++ b/arch/parisc/include/asm/hugetlb.h
@@ -28,9 +28,10 @@ static inline int prepare_hugepage_range(struct file *file,
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long addr, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
 {
+	return *ptep;
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_SET_WRPROTECT
diff --git a/arch/powerpc/include/asm/hugetlb.h b/arch/powerpc/include/asm/hugetlb.h
index 6a1a1ac..8a5674f 100644
--- a/arch/powerpc/include/asm/hugetlb.h
+++ b/arch/powerpc/include/asm/hugetlb.h
@@ -43,11 +43,14 @@ static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long addr, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
 {
-	huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
+	pte_t pte;
+
+	pte = huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
 	flush_hugetlb_page(vma, addr);
+	return pte;
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_SET_ACCESS_FLAGS
diff --git a/arch/s390/include/asm/hugetlb.h b/arch/s390/include/asm/hugetlb.h
index 32c3fd6..f22beda 100644
--- a/arch/s390/include/asm/hugetlb.h
+++ b/arch/s390/include/asm/hugetlb.h
@@ -50,10 +50,10 @@ static inline void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
 		set_pte(ptep, __pte(_SEGMENT_ENTRY_EMPTY));
 }
 
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long address, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long address, pte_t *ptep)
 {
-	huge_ptep_get_and_clear(vma->vm_mm, address, ptep);
+	return huge_ptep_get_and_clear(vma->vm_mm, address, ptep);
 }
 
 static inline int huge_ptep_set_access_flags(struct vm_area_struct *vma,
diff --git a/arch/sh/include/asm/hugetlb.h b/arch/sh/include/asm/hugetlb.h
index ae4de7b..4d3ba39 100644
--- a/arch/sh/include/asm/hugetlb.h
+++ b/arch/sh/include/asm/hugetlb.h
@@ -21,9 +21,10 @@ static inline int prepare_hugepage_range(struct file *file,
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long addr, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
 {
+	return *ptep;
 }
 
 static inline void arch_clear_hugepage_flags(struct page *page)
diff --git a/arch/sparc/include/asm/hugetlb.h b/arch/sparc/include/asm/hugetlb.h
index 53838a1..0a26cca 100644
--- a/arch/sparc/include/asm/hugetlb.h
+++ b/arch/sparc/include/asm/hugetlb.h
@@ -21,9 +21,10 @@ pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 			      pte_t *ptep);
 
 #define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
-					 unsigned long addr, pte_t *ptep)
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
 {
+	return *ptep;
 }
 
 #define __HAVE_ARCH_HUGE_PTEP_SET_WRPROTECT
diff --git a/include/asm-generic/hugetlb.h b/include/asm-generic/hugetlb.h
index 896f341..a57d667 100644
--- a/include/asm-generic/hugetlb.h
+++ b/include/asm-generic/hugetlb.h
@@ -84,10 +84,10 @@ static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
 #endif
 
 #ifndef __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
-static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
 		unsigned long addr, pte_t *ptep)
 {
-	ptep_clear_flush(vma, addr, ptep);
+	return ptep_clear_flush(vma, addr, ptep);
 }
 #endif
 
-- 
1.8.3.1

[PATCH v4 2/3] mm: rmap: Fix CONT-PTE/PMD size hugetlb issue when migration

[Baolin Wang]

On some architectures (like ARM64), it can support CONT-PTE/PMD size
hugetlb, which means it can support not only PMD/PUD size hugetlb:
2M and 1G, but also CONT-PTE/PMD size: 64K and 32M if a 4K page
size specified.

When migrating a hugetlb page, we will get the relevant page table
entry by huge_pte_offset() only once to nuke it and remap it with
a migration pte entry. This is correct for PMD or PUD size hugetlb,
since they always contain only one pmd entry or pud entry in the
page table.

However this is incorrect for CONT-PTE and CONT-PMD size hugetlb,
since they can contain several continuous pte or pmd entry with
same page table attributes. So we will nuke or remap only one pte
or pmd entry for this CONT-PTE/PMD size hugetlb page, which is
not expected for hugetlb migration. The problem is we can still
continue to modify the subpages' data of a hugetlb page during
migrating a hugetlb page, which can cause a serious data consistent
issue, since we did not nuke the page table entry and set a
migration pte for the subpages of a hugetlb page.

To fix this issue, we should change to use huge_ptep_clear_flush()
to nuke a hugetlb page table, and remap it with set_huge_pte_at()
and set_huge_swap_pte_at() when migrating a hugetlb page, which
already considered the CONT-PTE or CONT-PMD size hugetlb.

Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
---
 include/linux/hugetlb.h | 11 +++++++++++
 mm/rmap.c               | 24 ++++++++++++++++++------
 2 files changed, 29 insertions(+), 6 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 306d6ef..abde66e 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -1093,6 +1093,17 @@ static inline void set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr
 					pte_t *ptep, pte_t pte, unsigned long sz)
 {
 }
+
+static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
+					  unsigned long addr, pte_t *ptep)
+{
+	return *ptep;
+}
+
+static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+				   pte_t *ptep, pte_t pte)
+{
+}
 #endif	/* CONFIG_HUGETLB_PAGE */
 
 static inline spinlock_t *huge_pte_lock(struct hstate *h,
diff --git a/mm/rmap.c b/mm/rmap.c
index 94d6b24..4e96daf 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1926,13 +1926,15 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
 					break;
 				}
 			}
+
+			/* Nuke the hugetlb page table entry */
+			pteval = huge_ptep_clear_flush(vma, address, pvmw.pte);
 		} else {
 			flush_cache_page(vma, address, pte_pfn(*pvmw.pte));
+			/* Nuke the page table entry. */
+			pteval = ptep_clear_flush(vma, address, pvmw.pte);
 		}
 
-		/* Nuke the page table entry. */
-		pteval = ptep_clear_flush(vma, address, pvmw.pte);
-
 		/* Set the dirty flag on the folio now the pte is gone. */
 		if (pte_dirty(pteval))
 			folio_mark_dirty(folio);
@@ -2017,7 +2019,10 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
 			pte_t swp_pte;
 
 			if (arch_unmap_one(mm, vma, address, pteval) < 0) {
-				set_pte_at(mm, address, pvmw.pte, pteval);
+				if (folio_test_hugetlb(folio))
+					set_huge_pte_at(mm, address, pvmw.pte, pteval);
+				else
+					set_pte_at(mm, address, pvmw.pte, pteval);
 				ret = false;
 				page_vma_mapped_walk_done(&pvmw);
 				break;
@@ -2026,7 +2031,10 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
 				       !anon_exclusive, subpage);
 			if (anon_exclusive &&
 			    page_try_share_anon_rmap(subpage)) {
-				set_pte_at(mm, address, pvmw.pte, pteval);
+				if (folio_test_hugetlb(folio))
+					set_huge_pte_at(mm, address, pvmw.pte, pteval);
+				else
+					set_pte_at(mm, address, pvmw.pte, pteval);
 				ret = false;
 				page_vma_mapped_walk_done(&pvmw);
 				break;
@@ -2052,7 +2060,11 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
 				swp_pte = pte_swp_mksoft_dirty(swp_pte);
 			if (pte_uffd_wp(pteval))
 				swp_pte = pte_swp_mkuffd_wp(swp_pte);
-			set_pte_at(mm, address, pvmw.pte, swp_pte);
+			if (folio_test_hugetlb(folio))
+				set_huge_swap_pte_at(mm, address, pvmw.pte,
+						     swp_pte, vma_mmu_pagesize(vma));
+			else
+				set_pte_at(mm, address, pvmw.pte, swp_pte);
 			trace_set_migration_pte(address, pte_val(swp_pte),
 						compound_order(&folio->page));
 			/*
-- 
1.8.3.1

[PATCH v4 3/3] mm: rmap: Fix CONT-PTE/PMD size hugetlb issue when unmapping

[Baolin Wang]

On some architectures (like ARM64), it can support CONT-PTE/PMD size
hugetlb, which means it can support not only PMD/PUD size hugetlb:
2M and 1G, but also CONT-PTE/PMD size: 64K and 32M if a 4K page
size specified.

When unmapping a hugetlb page, we will get the relevant page table
entry by huge_pte_offset() only once to nuke it. This is correct
for PMD or PUD size hugetlb, since they always contain only one
pmd entry or pud entry in the page table.

However this is incorrect for CONT-PTE and CONT-PMD size hugetlb,
since they can contain several continuous pte or pmd entry with
same page table attributes, so we will nuke only one pte or pmd
entry for this CONT-PTE/PMD size hugetlb page.

And now try_to_unmap() is only passed a hugetlb page in the case
where the hugetlb page is poisoned. Which means now we will unmap
only one pte entry for a CONT-PTE or CONT-PMD size poisoned hugetlb
page, and we can still access other subpages of a CONT-PTE or CONT-PMD
size poisoned hugetlb page, which will cause serious issues possibly.

So we should change to use huge_ptep_clear_flush() to nuke the
hugetlb page table to fix this issue, which already considered
CONT-PTE and CONT-PMD size hugetlb.

We've already used set_huge_swap_pte_at() to set a poisoned
swap entry for a poisoned hugetlb page. Meanwhile adding a VM_BUG_ON()
to make sure the passed hugetlb page is poisoned in try_to_unmap().

Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
---
 mm/rmap.c | 39 ++++++++++++++++++++++-----------------
 1 file changed, 22 insertions(+), 17 deletions(-)

diff --git a/mm/rmap.c b/mm/rmap.c
index 4e96daf..219e287 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1528,6 +1528,11 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
 
 		if (folio_test_hugetlb(folio)) {
 			/*
+			 * The try_to_unmap() is only passed a hugetlb page
+			 * in the case where the hugetlb page is poisoned.
+			 */
+			VM_BUG_ON_PAGE(!PageHWPoison(subpage), subpage);
+			/*
 			 * huge_pmd_unshare may unmap an entire PMD page.
 			 * There is no way of knowing exactly which PMDs may
 			 * be cached for this mm, so we must flush them all.
@@ -1562,28 +1567,28 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
 					break;
 				}
 			}
+			pteval = huge_ptep_clear_flush(vma, address, pvmw.pte);
 		} else {
 			flush_cache_page(vma, address, pte_pfn(*pvmw.pte));
-		}
-
-		/*
-		 * Nuke the page table entry. When having to clear
-		 * PageAnonExclusive(), we always have to flush.
-		 */
-		if (should_defer_flush(mm, flags) && !anon_exclusive) {
 			/*
-			 * We clear the PTE but do not flush so potentially
-			 * a remote CPU could still be writing to the folio.
-			 * If the entry was previously clean then the
-			 * architecture must guarantee that a clear->dirty
-			 * transition on a cached TLB entry is written through
-			 * and traps if the PTE is unmapped.
+			 * Nuke the page table entry. When having to clear
+			 * PageAnonExclusive(), we always have to flush.
 			 */
-			pteval = ptep_get_and_clear(mm, address, pvmw.pte);
+			if (should_defer_flush(mm, flags) && !anon_exclusive) {
+				/*
+				 * We clear the PTE but do not flush so potentially
+				 * a remote CPU could still be writing to the folio.
+				 * If the entry was previously clean then the
+				 * architecture must guarantee that a clear->dirty
+				 * transition on a cached TLB entry is written through
+				 * and traps if the PTE is unmapped.
+				 */
+				pteval = ptep_get_and_clear(mm, address, pvmw.pte);
 
-			set_tlb_ubc_flush_pending(mm, pte_dirty(pteval));
-		} else {
-			pteval = ptep_clear_flush(vma, address, pvmw.pte);
+				set_tlb_ubc_flush_pending(mm, pte_dirty(pteval));
+			} else {
+				pteval = ptep_clear_flush(vma, address, pvmw.pte);
+			}
 		}
 
 		/*
-- 
1.8.3.1

Re: [PATCH v4 2/3] mm: rmap: Fix CONT-PTE/PMD size hugetlb issue when migration

[David Hildenbrand]

On 11.05.22 14:04, Baolin Wang wrote:
> On some architectures (like ARM64), it can support CONT-PTE/PMD size
> hugetlb, which means it can support not only PMD/PUD size hugetlb:
> 2M and 1G, but also CONT-PTE/PMD size: 64K and 32M if a 4K page
> size specified.
> 
> When migrating a hugetlb page, we will get the relevant page table
> entry by huge_pte_offset() only once to nuke it and remap it with
> a migration pte entry. This is correct for PMD or PUD size hugetlb,
> since they always contain only one pmd entry or pud entry in the
> page table.
> 
> However this is incorrect for CONT-PTE and CONT-PMD size hugetlb,
> since they can contain several continuous pte or pmd entry with
> same page table attributes. So we will nuke or remap only one pte
> or pmd entry for this CONT-PTE/PMD size hugetlb page, which is
> not expected for hugetlb migration. The problem is we can still
> continue to modify the subpages' data of a hugetlb page during
> migrating a hugetlb page, which can cause a serious data consistent
> issue, since we did not nuke the page table entry and set a
> migration pte for the subpages of a hugetlb page.
> 
> To fix this issue, we should change to use huge_ptep_clear_flush()
> to nuke a hugetlb page table, and remap it with set_huge_pte_at()
> and set_huge_swap_pte_at() when migrating a hugetlb page, which
> already considered the CONT-PTE or CONT-PMD size hugetlb.
> 
> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> Reviewed-by: Muchun Song <songmuchun@bytedance.com>
> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>

Makes sense to me.

Acked-by: David Hildenbrand <david@redhat.com>


-- 
Thanks,

David / dhildenb

Re: [PATCH v4 3/3] mm: rmap: Fix CONT-PTE/PMD size hugetlb issue when unmapping

[David Hildenbrand]

On 11.05.22 14:04, Baolin Wang wrote:
> On some architectures (like ARM64), it can support CONT-PTE/PMD size
> hugetlb, which means it can support not only PMD/PUD size hugetlb:
> 2M and 1G, but also CONT-PTE/PMD size: 64K and 32M if a 4K page
> size specified.
> 
> When unmapping a hugetlb page, we will get the relevant page table
> entry by huge_pte_offset() only once to nuke it. This is correct
> for PMD or PUD size hugetlb, since they always contain only one
> pmd entry or pud entry in the page table.
> 
> However this is incorrect for CONT-PTE and CONT-PMD size hugetlb,
> since they can contain several continuous pte or pmd entry with
> same page table attributes, so we will nuke only one pte or pmd
> entry for this CONT-PTE/PMD size hugetlb page.
> 
> And now try_to_unmap() is only passed a hugetlb page in the case
> where the hugetlb page is poisoned. Which means now we will unmap
> only one pte entry for a CONT-PTE or CONT-PMD size poisoned hugetlb
> page, and we can still access other subpages of a CONT-PTE or CONT-PMD
> size poisoned hugetlb page, which will cause serious issues possibly.
> 
> So we should change to use huge_ptep_clear_flush() to nuke the
> hugetlb page table to fix this issue, which already considered
> CONT-PTE and CONT-PMD size hugetlb.
> 
> We've already used set_huge_swap_pte_at() to set a poisoned
> swap entry for a poisoned hugetlb page. Meanwhile adding a VM_BUG_ON()
> to make sure the passed hugetlb page is poisoned in try_to_unmap().
> 
> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> Reviewed-by: Muchun Song <songmuchun@bytedance.com>
> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
> ---
>  mm/rmap.c | 39 ++++++++++++++++++++++-----------------
>  1 file changed, 22 insertions(+), 17 deletions(-)
> 
> diff --git a/mm/rmap.c b/mm/rmap.c
> index 4e96daf..219e287 100644
> --- a/mm/rmap.c
> +++ b/mm/rmap.c
> @@ -1528,6 +1528,11 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
>  
>  		if (folio_test_hugetlb(folio)) {
>  			/*
> +			 * The try_to_unmap() is only passed a hugetlb page
> +			 * in the case where the hugetlb page is poisoned.
> +			 */
> +			VM_BUG_ON_PAGE(!PageHWPoison(subpage), subpage);
> +			/*
>  			 * huge_pmd_unshare may unmap an entire PMD page.
>  			 * There is no way of knowing exactly which PMDs may
>  			 * be cached for this mm, so we must flush them all.
> @@ -1562,28 +1567,28 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
>  					break;
>  				}
>  			}
> +			pteval = huge_ptep_clear_flush(vma, address, pvmw.pte);
>  		} else {
>  			flush_cache_page(vma, address, pte_pfn(*pvmw.pte));
> -		}
> -
> -		/*
> -		 * Nuke the page table entry. When having to clear
> -		 * PageAnonExclusive(), we always have to flush.
> -		 */
> -		if (should_defer_flush(mm, flags) && !anon_exclusive) {
>  			/*
> -			 * We clear the PTE but do not flush so potentially
> -			 * a remote CPU could still be writing to the folio.
> -			 * If the entry was previously clean then the
> -			 * architecture must guarantee that a clear->dirty
> -			 * transition on a cached TLB entry is written through
> -			 * and traps if the PTE is unmapped.
> +			 * Nuke the page table entry. When having to clear
> +			 * PageAnonExclusive(), we always have to flush.
>  			 */
> -			pteval = ptep_get_and_clear(mm, address, pvmw.pte);
> +			if (should_defer_flush(mm, flags) && !anon_exclusive) {
> +				/*
> +				 * We clear the PTE but do not flush so potentially
> +				 * a remote CPU could still be writing to the folio.
> +				 * If the entry was previously clean then the
> +				 * architecture must guarantee that a clear->dirty
> +				 * transition on a cached TLB entry is written through
> +				 * and traps if the PTE is unmapped.
> +				 */
> +				pteval = ptep_get_and_clear(mm, address, pvmw.pte);
>  
> -			set_tlb_ubc_flush_pending(mm, pte_dirty(pteval));
> -		} else {
> -			pteval = ptep_clear_flush(vma, address, pvmw.pte);
> +				set_tlb_ubc_flush_pending(mm, pte_dirty(pteval));
> +			} else {
> +				pteval = ptep_clear_flush(vma, address, pvmw.pte);
> +			}
>  		}
>  
>  		/*

LGTM

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Thanks,

David / dhildenb