khugepaged might_sleep() warn due to CONFIG_HIGHPTE=y

PATCH In review Ville Syrjälä 4 participants 6 messages Oct 29, 2019 – Oct 30, 2019

Subsystems (2)

Memory management THP

lore

Share & tools

lore

Ville Syrjälä

author

2019-10-29 20:15 UTC #0 lore raw

Hi,

I got some khugepaged spew on a 32bit x86:

[  217.490026] BUG: sleeping function called from invalid context at include/linux/mmu_notifier.h:346
[  217.492826] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: khugepaged
[  217.495589] INFO: lockdep is turned off.
[  217.498371] CPU: 1 PID: 25 Comm: khugepaged Not tainted 5.4.0-rc5-elk+ #206
[  217.501233] Hardware name: System manufacturer P5Q-EM/P5Q-EM, BIOS 2203    07/08/2009
[  217.501697] Call Trace:
[  217.501697]  dump_stack+0x66/0x8e
[  217.501697]  ___might_sleep.cold.96+0x95/0xa6
[  217.501697]  __might_sleep+0x2e/0x80
[  217.501697]  collapse_huge_page.isra.51+0x5ac/0x1360
[  217.501697]  ? __alloc_pages_nodemask+0xec/0xf80
[  217.501697]  ? __alloc_pages_nodemask+0x191/0xf80
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  khugepaged+0x9a9/0x20f0
[  217.501697]  ? _raw_spin_unlock+0x21/0x30
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  ? wait_woken+0xa0/0xa0
[  217.501697]  kthread+0xf5/0x110
[  217.501697]  ? collapse_pte_mapped_thp+0x3b0/0x3b0
[  217.501697]  ? kthread_create_worker_on_cpu+0x20/0x20
[  217.501697]  ret_from_fork+0x2e/0x38

Looks like it's due to CONFIG_HIGHPTE=y pte_offset_map()->kmap_atomic() vs. mmu_notifier_invalidate_range_start().

My naive idea would be to just reorder those things, but not sure if there's some magic ordering constraint here. At least the machine still boots when I do it :)

mm/khugepaged.c +4 -3

index 0a1b4b484ac5..f05d27b7183d 100644
@@ -1028,12 +1028,13 @@ static void collapse_huge_page(struct mm_struct *mm,
 
 	anon_vma_lock_write(vma->anon_vma);
 
-	pte = pte_offset_map(pmd, address);
-	pte_ptl = pte_lockptr(mm, pmd);
-
 	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
 				address, address + HPAGE_PMD_SIZE);
 	mmu_notifier_invalidate_range_start(&range);
+
+	pte = pte_offset_map(pmd, address);
+	pte_ptl = pte_lockptr(mm, pmd);
+
 	pmd_ptl = pmd_lock(mm, pmd); /* probably unnecessary */
 	/*
 	 * After this gup_fast can't run anymore. This also removes

Signature · 2 lines

Ville Syrjälä
Intel

Andrew Morton

maintainer

2019-10-29 21:02 UTC #1 lore raw

Quoted 45 lines · On Tue, 29 Oct 2019 22:15:13 +0200 Ville Syrjälä…

Hi,

I got some khugepaged spew on a 32bit x86:

[  217.490026] BUG: sleeping function called from invalid context at include/linux/mmu_notifier.h:346
[  217.492826] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: khugepaged
[  217.495589] INFO: lockdep is turned off.
[  217.498371] CPU: 1 PID: 25 Comm: khugepaged Not tainted 5.4.0-rc5-elk+ #206
[  217.501233] Hardware name: System manufacturer P5Q-EM/P5Q-EM, BIOS 2203    07/08/2009
[  217.501697] Call Trace:
[  217.501697]  dump_stack+0x66/0x8e
[  217.501697]  ___might_sleep.cold.96+0x95/0xa6
[  217.501697]  __might_sleep+0x2e/0x80
[  217.501697]  collapse_huge_page.isra.51+0x5ac/0x1360
[  217.501697]  ? __alloc_pages_nodemask+0xec/0xf80
[  217.501697]  ? __alloc_pages_nodemask+0x191/0xf80
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  khugepaged+0x9a9/0x20f0
[  217.501697]  ? _raw_spin_unlock+0x21/0x30
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  ? wait_woken+0xa0/0xa0
[  217.501697]  kthread+0xf5/0x110
[  217.501697]  ? collapse_pte_mapped_thp+0x3b0/0x3b0
[  217.501697]  ? kthread_create_worker_on_cpu+0x20/0x20
[  217.501697]  ret_from_fork+0x2e/0x38

Looks like it's due to CONFIG_HIGHPTE=y pte_offset_map()->kmap_atomic() vs. mmu_notifier_invalidate_range_start().

My naive idea would be to just reorder those things, but not sure if there's some magic ordering constraint here. At least the machine still boots when I do it :)

mm/khugepaged.c +0 -3

index 0a1b4b484ac5..f05d27b7183d 100644
@@ -1028,12 +1028,13 @@ static void collapse_huge_page(struct mm_struct *mm,
 
 	anon_vma_lock_write(vma->anon_vma);
 
-	pte = pte_offset_map(pmd, address);
-	pte_ptl = pte_lockptr(mm, pmd);
-

mm/khugepaged.c static void collapse_huge_page(struct mm_struct *mm, · line 1040

Quoted 10 lines

mm/khugepaged.c +4 -0

 	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
 				address, address + HPAGE_PMD_SIZE);
 	mmu_notifier_invalidate_range_start(&range);
+
+	pte = pte_offset_map(pmd, address);
+	pte_ptl = pte_lockptr(mm, pmd);
+
 	pmd_ptl = pmd_lock(mm, pmd); /* probably unnecessary */
 	/*
 	 * After this gup_fast can't run anymore. This also removes

Looks good to me. Can you resend it with a signoff please?

Ville Syrjala

author

2019-10-29 21:25 UTC #2 lore raw

PATCH 1/1 mm/khugepaged: Fix might_sleep() warn with CONFIG_HIGHPTE=y

From: Ville Syrjälä <[email protected]>

I got some khugepaged spew on a 32bit x86:

[  217.490026] BUG: sleeping function called from invalid context at include/linux/mmu_notifier.h:346
[  217.492826] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: khugepaged
[  217.495589] INFO: lockdep is turned off.
[  217.498371] CPU: 1 PID: 25 Comm: khugepaged Not tainted 5.4.0-rc5-elk+ #206
[  217.501233] Hardware name: System manufacturer P5Q-EM/P5Q-EM, BIOS 2203    07/08/2009
[  217.501697] Call Trace:
[  217.501697]  dump_stack+0x66/0x8e
[  217.501697]  ___might_sleep.cold.96+0x95/0xa6
[  217.501697]  __might_sleep+0x2e/0x80
[  217.501697]  collapse_huge_page.isra.51+0x5ac/0x1360
[  217.501697]  ? __alloc_pages_nodemask+0xec/0xf80
[  217.501697]  ? __alloc_pages_nodemask+0x191/0xf80
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  khugepaged+0x9a9/0x20f0
[  217.501697]  ? _raw_spin_unlock+0x21/0x30
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  ? wait_woken+0xa0/0xa0
[  217.501697]  kthread+0xf5/0x110
[  217.501697]  ? collapse_pte_mapped_thp+0x3b0/0x3b0
[  217.501697]  ? kthread_create_worker_on_cpu+0x20/0x20
[  217.501697]  ret_from_fork+0x2e/0x38

Looks like it's due to CONFIG_HIGHPTE=y pte_offset_map()->kmap_atomic() vs. mmu_notifier_invalidate_range_start(). Let's do the naive approach and just reorder the two operations.

Cc [email protected]

Cc Andrew Morton <[email protected]>

Cc "Kirill A. Shutemov" <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Ingo Molnar <[email protected]>

Cc Borislav Petkov <[email protected]>

Cc "H. Peter Anvin" <[email protected]>

Cc [email protected]

Signed-off-by Ville Syrjälä <[email protected]>

---
 mm/khugepaged.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

mm/khugepaged.c +4 -3

index 0a1b4b484ac5..f05d27b7183d 100644
@@ -1028,12 +1028,13 @@ static void collapse_huge_page(struct mm_struct *mm,
 
 	anon_vma_lock_write(vma->anon_vma);
 
-	pte = pte_offset_map(pmd, address);
-	pte_ptl = pte_lockptr(mm, pmd);
-
 	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
 				address, address + HPAGE_PMD_SIZE);
 	mmu_notifier_invalidate_range_start(&range);
+
+	pte = pte_offset_map(pmd, address);
+	pte_ptl = pte_lockptr(mm, pmd);
+
 	pmd_ptl = pmd_lock(mm, pmd); /* probably unnecessary */
 	/*
 	 * After this gup_fast can't run anymore. This also removes

Signature · 1 line

2.23.0

Andrew Morton

maintainer

2019-10-29 21:43 UTC #3 lore raw

Re: [PATCH] mm/khugepaged: Fix might_sleep() warn with CONFIG_HIGHPTE=y

Quoted 26 lines · On Tue, 29 Oct 2019 23:25:13 +0200 Ville Syrjala…

From: Ville Syrjälä <[email protected]>

I got some khugepaged spew on a 32bit x86:

[  217.490026] BUG: sleeping function called from invalid context at include/linux/mmu_notifier.h:346
[  217.492826] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: khugepaged
[  217.495589] INFO: lockdep is turned off.
[  217.498371] CPU: 1 PID: 25 Comm: khugepaged Not tainted 5.4.0-rc5-elk+ #206
[  217.501233] Hardware name: System manufacturer P5Q-EM/P5Q-EM, BIOS 2203    07/08/2009
[  217.501697] Call Trace:
[  217.501697]  dump_stack+0x66/0x8e
[  217.501697]  ___might_sleep.cold.96+0x95/0xa6
[  217.501697]  __might_sleep+0x2e/0x80
[  217.501697]  collapse_huge_page.isra.51+0x5ac/0x1360
[  217.501697]  ? __alloc_pages_nodemask+0xec/0xf80
[  217.501697]  ? __alloc_pages_nodemask+0x191/0xf80
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  khugepaged+0x9a9/0x20f0
[  217.501697]  ? _raw_spin_unlock+0x21/0x30
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  ? wait_woken+0xa0/0xa0
[  217.501697]  kthread+0xf5/0x110
[  217.501697]  ? collapse_pte_mapped_thp+0x3b0/0x3b0
[  217.501697]  ? kthread_create_worker_on_cpu+0x20/0x20
[  217.501697]  ret_from_fork+0x2e/0x38

Quoted 3 lines

Looks like it's due to CONFIG_HIGHPTE=y pte_offset_map()->kmap_atomic() vs. mmu_notifier_invalidate_range_start(). Let's do the naive approach and just reorder the two operations.

Thanks.

I'll add

Fixes 810e24e009cf71 ("mm/mmu_notifiers: annotate with might_sleep()")

Cc <[email protected]>

although 810e24e009cf71 merely exposed the bug.

Jason Gunthorpe

2019-10-29 23:15 UTC #5 lore raw

In reply to Ville Syrjälä

Quoted 40 lines · Ville Syrjälä

Hi,

I got some khugepaged spew on a 32bit x86:

[  217.490026] BUG: sleeping function called from invalid context at include/linux/mmu_notifier.h:346
[  217.492826] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: khugepaged
[  217.495589] INFO: lockdep is turned off.
[  217.498371] CPU: 1 PID: 25 Comm: khugepaged Not tainted 5.4.0-rc5-elk+ #206
[  217.501233] Hardware name: System manufacturer P5Q-EM/P5Q-EM, BIOS 2203    07/08/2009
[  217.501697] Call Trace:
[  217.501697]  dump_stack+0x66/0x8e
[  217.501697]  ___might_sleep.cold.96+0x95/0xa6
[  217.501697]  __might_sleep+0x2e/0x80
[  217.501697]  collapse_huge_page.isra.51+0x5ac/0x1360
[  217.501697]  ? __alloc_pages_nodemask+0xec/0xf80
[  217.501697]  ? __alloc_pages_nodemask+0x191/0xf80
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  khugepaged+0x9a9/0x20f0
[  217.501697]  ? _raw_spin_unlock+0x21/0x30
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  ? wait_woken+0xa0/0xa0
[  217.501697]  kthread+0xf5/0x110
[  217.501697]  ? collapse_pte_mapped_thp+0x3b0/0x3b0
[  217.501697]  ? kthread_create_worker_on_cpu+0x20/0x20
[  217.501697]  ret_from_fork+0x2e/0x38

Looks like it's due to CONFIG_HIGHPTE=y pte_offset_map()->kmap_atomic() vs. mmu_notifier_invalidate_range_start().

My naive idea would be to just reorder those things, but not sure if there's some magic ordering constraint here. At least the machine still boots when I do it :)

mm/khugepaged.c +1 -0

index 0a1b4b484ac5..f05d27b7183d 100644
+++ b/mm/khugepaged.c

@@ -1028,12 +1028,13 @@ static void collapse_huge_page(struct mm_struct *mm,
 
 	anon_vma_lock_write(vma->anon_vma);

mm/khugepaged.c static void collapse_huge_page(struct mm_struct *mm, · line 1037

Quoted 10 lines

mm/khugepaged.c +4 -3

-	pte = pte_offset_map(pmd, address);
-	pte_ptl = pte_lockptr(mm, pmd);
-
 	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
 				address, address + HPAGE_PMD_SIZE);
 	mmu_notifier_invalidate_range_start(&range);
+
+	pte = pte_offset_map(pmd, address);
+	pte_ptl = pte_lockptr(mm, pmd);
+

Since pte and pte_ptl don't leak into invalidate_range_start this seems reasonable to me..

Good catch with the new debugging!

Jason

Kirill A. Shutemov

Acked-by on v1 1/1

2019-10-30 07:08 UTC #4 lore raw

Re: [PATCH] mm/khugepaged: Fix might_sleep() warn with CONFIG_HIGHPTE=y

In reply to Ville Syrjala

Quoted 30 lines · Ville Syrjala

From: Ville Syrjälä <[email protected]>

I got some khugepaged spew on a 32bit x86:

[  217.490026] BUG: sleeping function called from invalid context at include/linux/mmu_notifier.h:346
[  217.492826] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: khugepaged
[  217.495589] INFO: lockdep is turned off.
[  217.498371] CPU: 1 PID: 25 Comm: khugepaged Not tainted 5.4.0-rc5-elk+ #206
[  217.501233] Hardware name: System manufacturer P5Q-EM/P5Q-EM, BIOS 2203    07/08/2009
[  217.501697] Call Trace:
[  217.501697]  dump_stack+0x66/0x8e
[  217.501697]  ___might_sleep.cold.96+0x95/0xa6
[  217.501697]  __might_sleep+0x2e/0x80
[  217.501697]  collapse_huge_page.isra.51+0x5ac/0x1360
[  217.501697]  ? __alloc_pages_nodemask+0xec/0xf80
[  217.501697]  ? __alloc_pages_nodemask+0x191/0xf80
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  khugepaged+0x9a9/0x20f0
[  217.501697]  ? _raw_spin_unlock+0x21/0x30
[  217.501697]  ? trace_hardirqs_on+0x4a/0xf0
[  217.501697]  ? wait_woken+0xa0/0xa0
[  217.501697]  kthread+0xf5/0x110
[  217.501697]  ? collapse_pte_mapped_thp+0x3b0/0x3b0
[  217.501697]  ? kthread_create_worker_on_cpu+0x20/0x20
[  217.501697]  ret_from_fork+0x2e/0x38

Looks like it's due to CONFIG_HIGHPTE=y pte_offset_map()->kmap_atomic() vs. mmu_notifier_invalidate_range_start(). Let's do the naive approach and just reorder the two operations.

Quoted 9 lines

Cc [email protected]
Cc Andrew Morton <[email protected]>
Cc "Kirill A. Shutemov" <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
Cc [email protected]
Signed-off-by Ville Syrjälä <[email protected]>

Acked-by Kirill A. Shutemov <[email protected]>

Signature · 1 line

 Kirill A. Shutemov