Skip to contentSkip to conversation

mm/hugetlb: support FOLL_FORCE|FOLL_WRITE

PATCH Superseded Guillaume Morin 4 participants 6 messages Dec 5, 2024 – Dec 6, 2024

Subsystems (5)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 mm/hugetlb: support FOLL_FORCE|FOLL_WRITE

Eric reported that PTRACE_POKETEXT fails when applications use hugetlb for mapping text using huge pages. Before commit 1d8d14641fd9 ("mm/hugetlb: support write-faults in shared mappings"), PTRACE_POKETEXT worked by accident, but it was buggy and silently ended up mapping pages writable into the page tables even though VM_WRITE was not set.

In general, FOLL_FORCE|FOLL_WRITE does currently not work with hugetlb. Let's implement FOLL_FORCE|FOLL_WRITE properly for hugetlb, such that what used to work in the past by accident now properly works, allowing applications using hugetlb for text etc. to get properly debugged.

This change might also be required to implement uprobes support for hugetlb [1].

[1] https://lore.kernel.org/lkml/[email protected]/

Cc Muchun Song <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Eric Hagberg <[email protected]>

---
Changes in v2:
- Improved commit message
Changes in v3:
- Fix potential unitialized mem access in follow_huge_pud
- define pud_soft_dirty when soft dirty is not enabled

 include/linux/pgtable.h |  5 +++
 mm/gup.c                | 99 +++++++++++++++++++++--------------------
 mm/hugetlb.c            | 20 +++++----
 3 files changed, 66 insertions(+), 58 deletions(-)
include/linux/pgtable.h +5 -0
index adef9d6e9b1b..9335d7c82d20 100644
@@ -1422,6 +1422,11 @@ static inline int pmd_soft_dirty(pmd_t pmd)
 	return 0;
 }
 
+static inline int pud_soft_dirty(pud_t pud)
+{
+	return 0;
+}
+
 static inline pte_t pte_mksoft_dirty(pte_t pte)
 {
 	return pte;
mm/gup.c +50 -49

146 lines not on this page · raw message

mm/hugetlb.c +11 -9
index ea2ed8e301ef..52517b7ce308 100644
@@ -5169,6 +5169,13 @@ static void set_huge_ptep_writable(struct vm_area_struct *vma,
 		update_mmu_cache(vma, address, ptep);
 }
 
+static void set_huge_ptep_maybe_writable(struct vm_area_struct *vma,
+					 unsigned long address, pte_t *ptep)
+{
+	if (vma->vm_flags & VM_WRITE)
+		set_huge_ptep_writable(vma, address, ptep);
+}
+
 bool is_hugetlb_entry_migration(pte_t pte)
 {
 	swp_entry_t swp;
@@ -5802,13 +5809,6 @@ static vm_fault_t hugetlb_wp(struct folio *pagecache_folio,
 	if (!unshare && huge_pte_uffd_wp(pte))
 		return 0;
 
-	/*
-	 * hugetlb does not support FOLL_FORCE-style write faults that keep the
-	 * PTE mapped R/O such as maybe_mkwrite() would do.
-	 */
-	if (WARN_ON_ONCE(!unshare && !(vma->vm_flags & VM_WRITE)))
-		return VM_FAULT_SIGSEGV;
-
 	/* Let's take out MAP_SHARED mappings first. */
 	if (vma->vm_flags & VM_MAYSHARE) {
 		set_huge_ptep_writable(vma, vmf->address, vmf->pte);
@@ -5837,7 +5837,8 @@ static vm_fault_t hugetlb_wp(struct folio *pagecache_folio,
 			SetPageAnonExclusive(&old_folio->page);
 		}
 		if (likely(!unshare))
-			set_huge_ptep_writable(vma, vmf->address, vmf->pte);
+			set_huge_ptep_maybe_writable(vma, vmf->address,
+						     vmf->pte);
 
 		delayacct_wpcopy_end();
 		return 0;
@@ -5943,7 +5944,8 @@ static vm_fault_t hugetlb_wp(struct folio *pagecache_folio,
 	spin_lock(vmf->ptl);
 	vmf->pte = hugetlb_walk(vma, vmf->address, huge_page_size(h));
 	if (likely(vmf->pte && pte_same(huge_ptep_get(mm, vmf->address, vmf->pte), pte))) {
-		pte_t newpte = make_huge_pte(vma, &new_folio->page, !unshare);
+		const bool writable = !unshare && (vma->vm_flags & VM_WRITE);
+		pte_t newpte = make_huge_pte(vma, &new_folio->page, writable);
 
 		/* Break COW or unshare */
 		huge_ptep_clear_flush(vma, vmf->address, vmf->pte);
Signature · 4 lines
2.39.1

-- 
Guillaume Morin <[email protected]>
#1 lore raw

Hi Guillaume and s390 folks,

Quoted 97 lines · Guillaume Morin

Eric reported that PTRACE_POKETEXT fails when applications use hugetlb for mapping text using huge pages. Before commit 1d8d14641fd9 ("mm/hugetlb: support write-faults in shared mappings"), PTRACE_POKETEXT worked by accident, but it was buggy and silently ended up mapping pages writable into the page tables even though VM_WRITE was not set.

In general, FOLL_FORCE|FOLL_WRITE does currently not work with hugetlb. Let's implement FOLL_FORCE|FOLL_WRITE properly for hugetlb, such that what used to work in the past by accident now properly works, allowing applications using hugetlb for text etc. to get properly debugged.

This change might also be required to implement uprobes support for hugetlb [1].

[1] https://lore.kernel.org/lkml/[email protected]/

Cc Muchun Song <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Eric Hagberg <[email protected]>

---
Changes in v2:
- Improved commit message
Changes in v3:
- Fix potential unitialized mem access in follow_huge_pud
- define pud_soft_dirty when soft dirty is not enabled

 include/linux/pgtable.h |  5 +++
 mm/gup.c                | 99 +++++++++++++++++++++--------------------
 mm/hugetlb.c            | 20 +++++----
 3 files changed, 66 insertions(+), 58 deletions(-)
include/linux/pgtable.h +5 -0
index adef9d6e9b1b..9335d7c82d20 100644
@@ -1422,6 +1422,11 @@ static inline int pmd_soft_dirty(pmd_t pmd)
 	return 0;
 }
 
+static inline int pud_soft_dirty(pud_t pud)
+{
+	return 0;
+}
+
 static inline pte_t pte_mksoft_dirty(pte_t pte)
 {
 	return pte;
mm/gup.c +31 -0
index 746070a1d8bf..cc3eae458013 100644
@@ -587,6 +587,33 @@ static struct folio *try_grab_folio_fast(struct page *page, int refs,
 }
 #endif	/* CONFIG_HAVE_GUP_FAST */
 
+/* Common code for can_follow_write_* */
+static inline bool can_follow_write_common(struct page *page,
+		struct vm_area_struct *vma, unsigned int flags)
+{
+	/* Maybe FOLL_FORCE is set to override it? */
+	if (!(flags & FOLL_FORCE))
+		return false;
+
+	/* But FOLL_FORCE has no effect on shared mappings */
+	if (vma->vm_flags & (VM_MAYSHARE | VM_SHARED))
+		return false;
+
+	/* ... or read-only private ones */
+	if (!(vma->vm_flags & VM_MAYWRITE))
+		return false;
+
+	/* ... or already writable ones that just need to take a write fault */
+	if (vma->vm_flags & VM_WRITE)
+		return false;
+
+	/*
+	 * See can_change_pte_writable(): we broke COW and could map the page
+	 * writable if we have an exclusive anonymous page ...
+	 */
+	return page && PageAnon(page) && PageAnonExclusive(page);
+}
+
 static struct page *no_page_table(struct vm_area_struct *vma,
 				  unsigned int flags, unsigned long address)
 {
@@ -613,6 +640,22 @@ static struct page *no_page_table(struct vm_area_struct *vma,
 }
 
 #ifdef CONFIG_PGTABLE_HAS_HUGE_LEAVES
+/* FOLL_FORCE can write to even unwritable PUDs in COW mappings. */
+static inline bool can_follow_write_pud(pud_t pud, struct page *page,
+					struct vm_area_struct *vma,
+					unsigned int flags)

mm/gup.c static struct page *no_page_table(struct vm_area_struct *vma, · line 656

Quoted 10 lines
mm/gup.c +10 -0
+{
+	/* If the pud is writable, we can write to the page. */
+	if (pud_write(pud))
+		return true;
+
+	if (!can_follow_write_common(page, vma, flags))
+		return false;
+
+	/* ... and a write-fault isn't required for other reasons. */
+	return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);

This looks to be one of the first uses of pud_soft_dirty() in a generic part of the tree from what I can tell, which shows that s390 is lacking it despite setting CONFIG_HAVE_ARCH_SOFT_DIRTY:

  $ make -skj"$(nproc)" ARCH=s390 CROSS_COMPILE=s390-linux- mrproper defconfig mm/gup.o
  mm/gup.c: In function 'can_follow_write_pud':
  mm/gup.c:665:48: error: implicit declaration of function 'pud_soft_dirty'; did you mean 'pmd_soft_dirty'? [-Wimplicit-function-declaration]
    665 |         return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);
        |                                                ^~~~~~~~~~~~~~
        |                                                pmd_soft_dirty

Is this expected?

Cheers,
Nathan

Quoted 158 lines
mm/gup.c +9 -49
+}
+
 static struct page *follow_huge_pud(struct vm_area_struct *vma,
 				    unsigned long addr, pud_t *pudp,
 				    int flags, struct follow_page_context *ctx)
@@ -625,13 +668,16 @@ static struct page *follow_huge_pud(struct vm_area_struct *vma,
 
 	assert_spin_locked(pud_lockptr(mm, pudp));
 
-	if ((flags & FOLL_WRITE) && !pud_write(pud))
+	pfn += (addr & ~PUD_MASK) >> PAGE_SHIFT;
+	page = pfn_to_page(pfn);
+
+	if ((flags & FOLL_WRITE) &&
+	    !can_follow_write_pud(pud, page, vma, flags))
 		return NULL;
 
 	if (!pud_present(pud))
 		return NULL;
 
-	pfn += (addr & ~PUD_MASK) >> PAGE_SHIFT;
 
 	if (IS_ENABLED(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD) &&
 	    pud_devmap(pud)) {
@@ -653,8 +699,6 @@ static struct page *follow_huge_pud(struct vm_area_struct *vma,
 			return ERR_PTR(-EFAULT);
 	}
 
-	page = pfn_to_page(pfn);
-
 	if (!pud_devmap(pud) && !pud_write(pud) &&
 	    gup_must_unshare(vma, flags, page))
 		return ERR_PTR(-EMLINK);
@@ -677,27 +721,7 @@ static inline bool can_follow_write_pmd(pmd_t pmd, struct page *page,
 	if (pmd_write(pmd))
 		return true;
 
-	/* Maybe FOLL_FORCE is set to override it? */
-	if (!(flags & FOLL_FORCE))
-		return false;
-
-	/* But FOLL_FORCE has no effect on shared mappings */
-	if (vma->vm_flags & (VM_MAYSHARE | VM_SHARED))
-		return false;
-
-	/* ... or read-only private ones */
-	if (!(vma->vm_flags & VM_MAYWRITE))
-		return false;
-
-	/* ... or already writable ones that just need to take a write fault */
-	if (vma->vm_flags & VM_WRITE)
-		return false;
-
-	/*
-	 * See can_change_pte_writable(): we broke COW and could map the page
-	 * writable if we have an exclusive anonymous page ...
-	 */
-	if (!page || !PageAnon(page) || !PageAnonExclusive(page))
+	if (!can_follow_write_common(page, vma, flags))
 		return false;
 
 	/* ... and a write-fault isn't required for other reasons. */
@@ -798,27 +822,7 @@ static inline bool can_follow_write_pte(pte_t pte, struct page *page,
 	if (pte_write(pte))
 		return true;
 
-	/* Maybe FOLL_FORCE is set to override it? */
-	if (!(flags & FOLL_FORCE))
-		return false;
-
-	/* But FOLL_FORCE has no effect on shared mappings */
-	if (vma->vm_flags & (VM_MAYSHARE | VM_SHARED))
-		return false;
-
-	/* ... or read-only private ones */
-	if (!(vma->vm_flags & VM_MAYWRITE))
-		return false;
-
-	/* ... or already writable ones that just need to take a write fault */
-	if (vma->vm_flags & VM_WRITE)
-		return false;
-
-	/*
-	 * See can_change_pte_writable(): we broke COW and could map the page
-	 * writable if we have an exclusive anonymous page ...
-	 */
-	if (!page || !PageAnon(page) || !PageAnonExclusive(page))
+	if (!can_follow_write_common(page, vma, flags))
 		return false;
 
 	/* ... and a write-fault isn't required for other reasons. */
@@ -1285,9 +1289,6 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
 		if (!(vm_flags & VM_WRITE) || (vm_flags & VM_SHADOW_STACK)) {
 			if (!(gup_flags & FOLL_FORCE))
 				return -EFAULT;
-			/* hugetlb does not support FOLL_FORCE|FOLL_WRITE. */
-			if (is_vm_hugetlb_page(vma))
-				return -EFAULT;
 			/*
 			 * We used to let the write,force case do COW in a
 			 * VM_MAYWRITE VM_SHARED !VM_WRITE vma, so ptrace could
mm/hugetlb.c +11 -9
index ea2ed8e301ef..52517b7ce308 100644
@@ -5169,6 +5169,13 @@ static void set_huge_ptep_writable(struct vm_area_struct *vma,
 		update_mmu_cache(vma, address, ptep);
 }
 
+static void set_huge_ptep_maybe_writable(struct vm_area_struct *vma,
+					 unsigned long address, pte_t *ptep)
+{
+	if (vma->vm_flags & VM_WRITE)
+		set_huge_ptep_writable(vma, address, ptep);
+}
+
 bool is_hugetlb_entry_migration(pte_t pte)
 {
 	swp_entry_t swp;
@@ -5802,13 +5809,6 @@ static vm_fault_t hugetlb_wp(struct folio *pagecache_folio,
 	if (!unshare && huge_pte_uffd_wp(pte))
 		return 0;
 
-	/*
-	 * hugetlb does not support FOLL_FORCE-style write faults that keep the
-	 * PTE mapped R/O such as maybe_mkwrite() would do.
-	 */
-	if (WARN_ON_ONCE(!unshare && !(vma->vm_flags & VM_WRITE)))
-		return VM_FAULT_SIGSEGV;
-
 	/* Let's take out MAP_SHARED mappings first. */
 	if (vma->vm_flags & VM_MAYSHARE) {
 		set_huge_ptep_writable(vma, vmf->address, vmf->pte);
@@ -5837,7 +5837,8 @@ static vm_fault_t hugetlb_wp(struct folio *pagecache_folio,
 			SetPageAnonExclusive(&old_folio->page);
 		}
 		if (likely(!unshare))
-			set_huge_ptep_writable(vma, vmf->address, vmf->pte);
+			set_huge_ptep_maybe_writable(vma, vmf->address,
+						     vmf->pte);
 
 		delayacct_wpcopy_end();
 		return 0;
@@ -5943,7 +5944,8 @@ static vm_fault_t hugetlb_wp(struct folio *pagecache_folio,
 	spin_lock(vmf->ptl);
 	vmf->pte = hugetlb_walk(vma, vmf->address, huge_page_size(h));
 	if (likely(vmf->pte && pte_same(huge_ptep_get(mm, vmf->address, vmf->pte), pte))) {
-		pte_t newpte = make_huge_pte(vma, &new_folio->page, !unshare);
+		const bool writable = !unshare && (vma->vm_flags & VM_WRITE);
+		pte_t newpte = make_huge_pte(vma, &new_folio->page, writable);
 
 		/* Break COW or unshare */
 		huge_ptep_clear_flush(vma, vmf->address, vmf->pte);
Signature · 4 lines
2.39.1

-- 
Guillaume Morin <[email protected]>
#2 lore raw
Quoted 116 lines · Nathan Chancellor

Hi Guillaume and s390 folks,

Quoted 97 lines · Guillaume Morin

Eric reported that PTRACE_POKETEXT fails when applications use hugetlb for mapping text using huge pages. Before commit 1d8d14641fd9 ("mm/hugetlb: support write-faults in shared mappings"), PTRACE_POKETEXT worked by accident, but it was buggy and silently ended up mapping pages writable into the page tables even though VM_WRITE was not set.

In general, FOLL_FORCE|FOLL_WRITE does currently not work with hugetlb. Let's implement FOLL_FORCE|FOLL_WRITE properly for hugetlb, such that what used to work in the past by accident now properly works, allowing applications using hugetlb for text etc. to get properly debugged.

This change might also be required to implement uprobes support for hugetlb [1].

[1] https://lore.kernel.org/lkml/[email protected]/

Cc Muchun Song <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Eric Hagberg <[email protected]>

---
Changes in v2:
- Improved commit message
Changes in v3:
- Fix potential unitialized mem access in follow_huge_pud
- define pud_soft_dirty when soft dirty is not enabled

 include/linux/pgtable.h |  5 +++
 mm/gup.c                | 99 +++++++++++++++++++++--------------------
 mm/hugetlb.c            | 20 +++++----
 3 files changed, 66 insertions(+), 58 deletions(-)
include/linux/pgtable.h +5 -0
index adef9d6e9b1b..9335d7c82d20 100644
@@ -1422,6 +1422,11 @@ static inline int pmd_soft_dirty(pmd_t pmd)
 	return 0;
 }
 
+static inline int pud_soft_dirty(pud_t pud)
+{
+	return 0;
+}
+
 static inline pte_t pte_mksoft_dirty(pte_t pte)
 {
 	return pte;
mm/gup.c +31 -0
index 746070a1d8bf..cc3eae458013 100644
@@ -587,6 +587,33 @@ static struct folio *try_grab_folio_fast(struct page *page, int refs,
 }
 #endif	/* CONFIG_HAVE_GUP_FAST */
 
+/* Common code for can_follow_write_* */
+static inline bool can_follow_write_common(struct page *page,
+		struct vm_area_struct *vma, unsigned int flags)
+{
+	/* Maybe FOLL_FORCE is set to override it? */
+	if (!(flags & FOLL_FORCE))
+		return false;
+
+	/* But FOLL_FORCE has no effect on shared mappings */
+	if (vma->vm_flags & (VM_MAYSHARE | VM_SHARED))
+		return false;
+
+	/* ... or read-only private ones */
+	if (!(vma->vm_flags & VM_MAYWRITE))
+		return false;
+
+	/* ... or already writable ones that just need to take a write fault */
+	if (vma->vm_flags & VM_WRITE)
+		return false;
+
+	/*
+	 * See can_change_pte_writable(): we broke COW and could map the page
+	 * writable if we have an exclusive anonymous page ...
+	 */
+	return page && PageAnon(page) && PageAnonExclusive(page);
+}
+
 static struct page *no_page_table(struct vm_area_struct *vma,
 				  unsigned int flags, unsigned long address)
 {
@@ -613,6 +640,22 @@ static struct page *no_page_table(struct vm_area_struct *vma,
 }
 
 #ifdef CONFIG_PGTABLE_HAS_HUGE_LEAVES
+/* FOLL_FORCE can write to even unwritable PUDs in COW mappings. */
+static inline bool can_follow_write_pud(pud_t pud, struct page *page,
+					struct vm_area_struct *vma,
+					unsigned int flags)

mm/gup.c static struct page *no_page_table(struct vm_area_struct *vma, · line 656

Quoted 10 lines
mm/gup.c +10 -0
+{
+	/* If the pud is writable, we can write to the page. */
+	if (pud_write(pud))
+		return true;
+
+	if (!can_follow_write_common(page, vma, flags))
+		return false;
+
+	/* ... and a write-fault isn't required for other reasons. */
+	return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);

This looks to be one of the first uses of pud_soft_dirty() in a generic part of the tree from what I can tell, which shows that s390 is lacking it despite setting CONFIG_HAVE_ARCH_SOFT_DIRTY:

Quoted 8 lines
  $ make -skj"$(nproc)" ARCH=s390 CROSS_COMPILE=s390-linux- mrproper defconfig mm/gup.o
  mm/gup.c: In function 'can_follow_write_pud':
  mm/gup.c:665:48: error: implicit declaration of function 'pud_soft_dirty'; did you mean 'pmd_soft_dirty'? [-Wimplicit-function-declaration]
    665 |         return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);
        |                                                ^~~~~~~~~~~~~~
        |                                                pmd_soft_dirty

Is this expected?

Yikes! It does look like an oversight in the s390 code since as you said it has CONFIG_HAVE_ARCH_SOFT_DIRTY and pud_mkdirty seems to be setting _REGION3_ENTRY_SOFT_DIRTY. But I'll let the s390 folks opine.

I don't mind dropping the pud part of the change (even if that's a bit of a shame) if it's causing too many issues.

Signature · 1 line
Guillaume Morin <[email protected]>
maintainer
#3 lore raw
Quoted 28 lines · Guillaume Morin
Quoted 27 lines · Nathan Chancellor
Quoted 5 lines

#ifdef CONFIG_PGTABLE_HAS_HUGE_LEAVES

mm/gup.c +4 -0
+/* FOLL_FORCE can write to even unwritable PUDs in COW mappings. */
+static inline bool can_follow_write_pud(pud_t pud, struct page *page,
+					struct vm_area_struct *vma,
+					unsigned int flags)

mm/gup.c static struct page *no_page_table(struct vm_area_struct *vma, · line 656

Quoted 10 lines
mm/gup.c +10 -0
+{
+	/* If the pud is writable, we can write to the page. */
+	if (pud_write(pud))
+		return true;
+
+	if (!can_follow_write_common(page, vma, flags))
+		return false;
+
+	/* ... and a write-fault isn't required for other reasons. */
+	return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);

This looks to be one of the first uses of pud_soft_dirty() in a generic part of the tree from what I can tell, which shows that s390 is lacking it despite setting CONFIG_HAVE_ARCH_SOFT_DIRTY:

  $ make -skj"$(nproc)" ARCH=s390 CROSS_COMPILE=s390-linux- mrproper defconfig mm/gup.o
  mm/gup.c: In function 'can_follow_write_pud':
  mm/gup.c:665:48: error: implicit declaration of function 'pud_soft_dirty'; did you mean 'pmd_soft_dirty'? [-Wimplicit-function-declaration]
    665 |         return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);
        |                                                ^~~~~~~~~~~~~~
        |                                                pmd_soft_dirty
Quoted 8 lines
Quoted 1 line

Is this expected?

Yikes! It does look like an oversight in the s390 code since as you said it has CONFIG_HAVE_ARCH_SOFT_DIRTY and pud_mkdirty seems to be setting _REGION3_ENTRY_SOFT_DIRTY. But I'll let the s390 folks opine.

I don't mind dropping the pud part of the change (even if that's a bit of a shame) if it's causing too many issues.

It would be quite easy to add pud_soft_dirty() etc. helper functions for s390, but I think that would be the wrong answer to this problem.

s390 implements pud_mkdirty(), but it is only used in the context of HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD, which s390 doesn't support. So this function should probably be removed from s390's pgtable.h.

Similar the pud_soft_dirty() and friends helper functions should only be implemented if common code support for soft dirty would exist, which is currently not the case. Otherwise similar fallbacks like for pmd_soft_dirty() (-> include/linux/pgtable.h) would also need to be implemented.

So IMHO the right fix (at this time) seems to be to remove the above pud part of your patch, and in addition we should probably also drop the partially implemented pud level soft dirty bits in s390 code, since that is dead code and might cause even more confusion in future.

Does that make sense?

#4 lore raw
Quoted 51 lines · Heiko Carstens
Quoted 28 lines · Guillaume Morin
Quoted 27 lines · Nathan Chancellor
Quoted 5 lines
  #ifdef CONFIG_PGTABLE_HAS_HUGE_LEAVES
mm/gup.c +4 -0
+/* FOLL_FORCE can write to even unwritable PUDs in COW mappings. */
+static inline bool can_follow_write_pud(pud_t pud, struct page *page,
+					struct vm_area_struct *vma,
+					unsigned int flags)

mm/gup.c static struct page *no_page_table(struct vm_area_struct *vma, · line 656

Quoted 10 lines
mm/gup.c +10 -0
+{
+	/* If the pud is writable, we can write to the page. */
+	if (pud_write(pud))
+		return true;
+
+	if (!can_follow_write_common(page, vma, flags))
+		return false;
+
+	/* ... and a write-fault isn't required for other reasons. */
+	return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);

This looks to be one of the first uses of pud_soft_dirty() in a generic part of the tree from what I can tell, which shows that s390 is lacking it despite setting CONFIG_HAVE_ARCH_SOFT_DIRTY:

   $ make -skj"$(nproc)" ARCH=s390 CROSS_COMPILE=s390-linux- mrproper defconfig mm/gup.o
   mm/gup.c: In function 'can_follow_write_pud':
   mm/gup.c:665:48: error: implicit declaration of function 'pud_soft_dirty'; did you mean 'pmd_soft_dirty'? [-Wimplicit-function-declaration]
     665 |         return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);
         |                                                ^~~~~~~~~~~~~~
         |                                                pmd_soft_dirty
Quoted 8 lines
Quoted 1 line

Is this expected?

Yikes! It does look like an oversight in the s390 code since as you said it has CONFIG_HAVE_ARCH_SOFT_DIRTY and pud_mkdirty seems to be setting _REGION3_ENTRY_SOFT_DIRTY. But I'll let the s390 folks opine.

I don't mind dropping the pud part of the change (even if that's a bit of a shame) if it's causing too many issues.

It would be quite easy to add pud_soft_dirty() etc. helper functions for s390, but I think that would be the wrong answer to this problem.

s390 implements pud_mkdirty(), but it is only used in the context of HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD, which s390 doesn't support. So this function should probably be removed from s390's pgtable.h.

Similar the pud_soft_dirty() and friends helper functions should only be implemented if common code support for soft dirty would exist, which is currently not the case. Otherwise similar fallbacks like for pmd_soft_dirty() (-> include/linux/pgtable.h) would also need to be implemented.

Quoted 6 lines

So IMHO the right fix (at this time) seems to be to remove the above pud part of your patch, and in addition we should probably also drop the partially implemented pud level soft dirty bits in s390 code, since that is dead code and might cause even more confusion in future.

Does that make sense?

As hugetlb does not support softdirty, and PUDs are currently only possible (weird DAX thing put aside) with hugetlb, it makes sense to drop the pud softdirty thingy.

Signature · 3 lines
Cheers,

David / dhildenb
#5 lore raw
Quoted 42 lines · David Hildenbrand
Quoted 40 lines · Heiko Carstens
Quoted 12 lines · Guillaume Morin
Quoted 11 lines · Nathan Chancellor

This looks to be one of the first uses of pud_soft_dirty() in a generic part of the tree from what I can tell, which shows that s390 is lacking it despite setting CONFIG_HAVE_ARCH_SOFT_DIRTY:

   $ make -skj"$(nproc)" ARCH=s390 CROSS_COMPILE=s390-linux- mrproper defconfig mm/gup.o
   mm/gup.c: In function 'can_follow_write_pud':
   mm/gup.c:665:48: error: implicit declaration of function 'pud_soft_dirty'; did you mean 'pmd_soft_dirty'? [-Wimplicit-function-declaration]
     665 |         return !vma_soft_dirty_enabled(vma) || pud_soft_dirty(pud);
         |                                                ^~~~~~~~~~~~~~
         |                                                pmd_soft_dirty
Quoted 8 lines
Quoted 1 line

Is this expected?

Yikes! It does look like an oversight in the s390 code since as you said it has CONFIG_HAVE_ARCH_SOFT_DIRTY and pud_mkdirty seems to be setting _REGION3_ENTRY_SOFT_DIRTY. But I'll let the s390 folks opine.

I don't mind dropping the pud part of the change (even if that's a bit of a shame) if it's causing too many issues.

It would be quite easy to add pud_soft_dirty() etc. helper functions for s390, but I think that would be the wrong answer to this problem.

s390 implements pud_mkdirty(), but it is only used in the context of HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD, which s390 doesn't support. So this function should probably be removed from s390's pgtable.h.

Similar the pud_soft_dirty() and friends helper functions should only be implemented if common code support for soft dirty would exist, which is currently not the case. Otherwise similar fallbacks like for pmd_soft_dirty() (-> include/linux/pgtable.h) would also need to be implemented.

So IMHO the right fix (at this time) seems to be to remove the above pud part of your patch, and in addition we should probably also drop the partially implemented pud level soft dirty bits in s390 code, since that is dead code and might cause even more confusion in future.

Quoted 5 lines
Quoted 1 line

Does that make sense?

As hugetlb does not support softdirty, and PUDs are currently only possible (weird DAX thing put aside) with hugetlb, it makes sense to drop the pud softdirty thingy.

Thanks all. I dropped the check and the dummy definition I had to add for i386 in v4 [1]

[1] https://lore.kernel.org/linux-mm/[email protected]/T/#u

Signature · 1 line
Guillaume Morin <[email protected]>