Skip to contentSkip to conversation

powerpc/64s/hash: Fix hash_preload running with interrupts enabled

PATCH Applied Nicholas Piggin 4 participants 7 messages Jul 27, 2020 – Aug 2, 2020 PowerPC

lore
Share & tools
lore

Nicholas Piggin

author reviewer
#0 lore raw

PATCH 1/1 powerpc/64s/hash: Fix hash_preload running with interrupts enabled

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

cpu 0x49e: Vector: 100 (System Reset) at [c00000001a4f7d70]

      pc: c000000000072dc8: hash_page_mm+0x8/0x800
      lr: c00000000000c5a4: do_hash_page+0x24/0x38
      sp: c0002ac1cc69ac70
     msr: 8000000000081033
    current = 0xc0002ac1cc602e00
    paca    = 0xc00000001de1f280   irqmask: 0x03   irq_happened: 0x01
      pid   = 20118, comm = pread2_processe

Linux version 5.8.0-rc6-00345-g1fad14f18bc6
49e:mon> t
[c0002ac1cc69ac70] c00000000000c5a4 do_hash_page+0x24/0x38 (unreliable)
--- Exception: 300 (Data Access) at c00000000008fa60 __copy_tofrom_user_power7+0x20c/0x7ac
[link register ] c000000000335d10 copy_from_user_nofault+0xf0/0x150
[c0002ac1cc69af70] c00032bf9fa3c880 (unreliable)
[c0002ac1cc69afa0] c000000000109df0 read_user_stack_64+0x70/0xf0 [c0002ac1cc69afd0] c000000000109fcc perf_callchain_user_64+0x15c/0x410 [c0002ac1cc69b060] c000000000109c00 perf_callchain_user+0x20/0x40 [c0002ac1cc69b080] c00000000031c6cc get_perf_callchain+0x25c/0x360 [c0002ac1cc69b120] c000000000316b50 perf_callchain+0x70/0xa0
[c0002ac1cc69b140] c000000000316ddc perf_prepare_sample+0x25c/0x790 [c0002ac1cc69b1a0] c000000000317350 perf_event_output_forward+0x40/0xb0 [c0002ac1cc69b220] c000000000306138 __perf_event_overflow+0x88/0x1a0 [c0002ac1cc69b270] c00000000010cf70 record_and_restart+0x230/0x750 [c0002ac1cc69b620] c00000000010d69c perf_event_interrupt+0x20c/0x510 [c0002ac1cc69b730] c000000000027d9c performance_monitor_exception+0x4c/0x60 [c0002ac1cc69b750] c00000000000b2f8 performance_monitor_common_virt+0x1b8/0x1c0 --- Exception: f00 (Performance Monitor) at c0000000000cb5b0 pSeries_lpar_hpte_insert+0x0/0x160
[link register ] c0000000000846f0 __hash_page_64K+0x210/0x540
[c0002ac1cc69ba50] 0000000000000000 (unreliable)
[c0002ac1cc69bb00] c000000000073ae0 update_mmu_cache+0x390/0x3a0 [c0002ac1cc69bb70] c00000000037f024 wp_page_copy+0x364/0xce0
[c0002ac1cc69bc20] c00000000038272c do_wp_page+0xdc/0xa60
[c0002ac1cc69bc70] c0000000003857bc handle_mm_fault+0xb9c/0x1b60 [c0002ac1cc69bd50] c00000000006c434 __do_page_fault+0x314/0xc90 [c0002ac1cc69be20] c00000000000c5c8 handle_page_fault+0x10/0x2c
--- Exception: 300 (Data Access) at 00007fff8c861fe8
SP (7ffff6b19660) is in userspace

Fixes 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the

caller")

---
 arch/powerpc/kernel/exceptions-64s.S  | 14 +++++++++++---
 arch/powerpc/mm/book3s64/hash_utils.c | 25 +++++++++++++++++++++++++
 arch/powerpc/perf/core-book3s.c       |  6 ++++++
 3 files changed, 42 insertions(+), 3 deletions(-)
arch/powerpc/kernel/exceptions-64s.S +11 -3
index 0fc8bad878b2..446e54c3f71e 100644
@@ -3072,10 +3072,18 @@ do_hash_page:
 	ori	r0,r0,DSISR_BAD_FAULT_64S@l
 	and.	r0,r5,r0		/* weird error? */
 	bne-	handle_page_fault	/* if not, try to insert a HPTE */
+
+	/*
+	 * If we are in an "NMI" (e.g., an interrupt when soft-disabled), then
+	 * don't call hash_page, just fail the fault. This is required to
+	 * prevent re-entrancy problems in the hash code, namely perf
+	 * interrupts hitting while something holds H_PAGE_BUSY, and taking a
+	 * hash fault. See the comment in hash_preload().
+	 */
 	ld	r11, PACA_THREAD_INFO(r13)
-	lwz	r0,TI_PREEMPT(r11)	/* If we're in an "NMI" */
-	andis.	r0,r0,NMI_MASK@h	/* (i.e. an irq when soft-disabled) */
-	bne	77f			/* then don't call hash_page now */
+	lwz	r0,TI_PREEMPT(r11)
+	andis.	r0,r0,NMI_MASK@h
+	bne	77f
 
 	/*
 	 * r3 contains the trap number
arch/powerpc/mm/book3s64/hash_utils.c +25 -0
index 468169e33c86..9b9f92ad0e7a 100644
@@ -1559,6 +1559,7 @@ static void hash_preload(struct mm_struct *mm, pte_t *ptep, unsigned long ea,
 	pgd_t *pgdir;
 	int rc, ssize, update_flags = 0;
 	unsigned long access = _PAGE_PRESENT | _PAGE_READ | (is_exec ? _PAGE_EXEC : 0);
+	unsigned long flags;
 
 	BUG_ON(get_region_id(ea) != USER_REGION_ID);
 
@@ -1592,6 +1593,28 @@ static void hash_preload(struct mm_struct *mm, pte_t *ptep, unsigned long ea,
 		return;
 #endif /* CONFIG_PPC_64K_PAGES */
 
+	/*
+	 * __hash_page_* must run with interrupts off, as it sets the
+	 * H_PAGE_BUSY bit. It's possible for perf interrupts to hit at any
+	 * time and may take a hash fault reading the user stack, see
+	 * read_user_stack_slow() in the powerpc/perf code.
+	 *
+	 * If that takes a hash fault on the same page as we lock here, it
+	 * will bail out when seeing H_PAGE_BUSY set, and retry the access
+	 * leading to an infinite loop.
+	 *
+	 * Disabling interrupts here does not prevent perf interrupts, but it
+	 * will prevent them taking hash faults (see the NMI test in
+	 * do_hash_page), then read_user_stack's copy_from_user_nofault will
+	 * fail and perf will fall back to read_user_stack_slow(), which
+	 * walks the Linux page tables.
+	 *
+	 * Interrupts must also be off for the duration of the
+	 * mm_is_thread_local test and update, to prevent preempt running the
+	 * mm on another CPU (XXX: this may be racy vs kthread_use_mm).
+	 */
+	local_irq_save(flags);
+
 	/* Is that local to this CPU ? */
 	if (mm_is_thread_local(mm))
 		update_flags |= HPTE_LOCAL_UPDATE;
@@ -1614,6 +1637,8 @@ static void hash_preload(struct mm_struct *mm, pte_t *ptep, unsigned long ea,
 				   mm_ctx_user_psize(&mm->context),
 				   mm_ctx_user_psize(&mm->context),
 				   pte_val(*ptep));
+
+	local_irq_restore(flags);
 }
 
 /*
arch/powerpc/perf/core-book3s.c +6 -0
index cd6a742ac6ef..01d70280d287 100644
@@ -2179,6 +2179,12 @@ static void __perf_event_interrupt(struct pt_regs *regs)
 
 	perf_read_regs(regs);
 
+	/*
+	 * If perf interrupts hit in a local_irq_disable (soft-masked) region,
+	 * we consider them as NMIs. This is required to prevent hash faults on
+	 * user addresses when reading callchains. See the NMI test in
+	 * do_hash_page.
+	 */
 	nmi = perf_intr_is_nmi(regs);
 	if (nmi)
 		nmi_enter();
Signature · 1 line
2.23.0
Tested-by on v1 1/1
#1 lore raw
Quoted 53 lines · Nicholas Piggin

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

cpu 0x49e: Vector: 100 (System Reset) at [c00000001a4f7d70]
pc: c000000000072dc8: hash_page_mm+0x8/0x800
lr: c00000000000c5a4: do_hash_page+0x24/0x38
sp: c0002ac1cc69ac70
msr: 8000000000081033
current = 0xc0002ac1cc602e00
paca = 0xc00000001de1f280 irqmask: 0x03 irq_happened: 0x01
pid = 20118, comm = pread2_processe
Linux version 5.8.0-rc6-00345-g1fad14f18bc6
49e:mon> t
[c0002ac1cc69ac70] c00000000000c5a4 do_hash_page+0x24/0x38 (unreliable)
--- Exception: 300 (Data Access) at c00000000008fa60 __copy_tofrom_user_power7+0x20c/0x7ac
[link register ] c000000000335d10 copy_from_user_nofault+0xf0/0x150
[c0002ac1cc69af70] c00032bf9fa3c880 (unreliable)
[c0002ac1cc69afa0] c000000000109df0 read_user_stack_64+0x70/0xf0 [c0002ac1cc69afd0] c000000000109fcc perf_callchain_user_64+0x15c/0x410 [c0002ac1cc69b060] c000000000109c00 perf_callchain_user+0x20/0x40 [c0002ac1cc69b080] c00000000031c6cc get_perf_callchain+0x25c/0x360 [c0002ac1cc69b120] c000000000316b50 perf_callchain+0x70/0xa0
[c0002ac1cc69b140] c000000000316ddc perf_prepare_sample+0x25c/0x790 [c0002ac1cc69b1a0] c000000000317350 perf_event_output_forward+0x40/0xb0 [c0002ac1cc69b220] c000000000306138 __perf_event_overflow+0x88/0x1a0 [c0002ac1cc69b270] c00000000010cf70 record_and_restart+0x230/0x750 [c0002ac1cc69b620] c00000000010d69c perf_event_interrupt+0x20c/0x510 [c0002ac1cc69b730] c000000000027d9c performance_monitor_exception+0x4c/0x60 [c0002ac1cc69b750] c00000000000b2f8 performance_monitor_common_virt+0x1b8/0x1c0 --- Exception: f00 (Performance Monitor) at c0000000000cb5b0 pSeries_lpar_hpte_insert+0x0/0x160
[link register ] c0000000000846f0 __hash_page_64K+0x210/0x540
[c0002ac1cc69ba50] 0000000000000000 (unreliable)
[c0002ac1cc69bb00] c000000000073ae0 update_mmu_cache+0x390/0x3a0 [c0002ac1cc69bb70] c00000000037f024 wp_page_copy+0x364/0xce0
[c0002ac1cc69bc20] c00000000038272c do_wp_page+0xdc/0xa60
[c0002ac1cc69bc70] c0000000003857bc handle_mm_fault+0xb9c/0x1b60 [c0002ac1cc69bd50] c00000000006c434 __do_page_fault+0x314/0xc90 [c0002ac1cc69be20] c00000000000c5c8 handle_page_fault+0x10/0x2c
--- Exception: 300 (Data Access) at 00007fff8c861fe8
SP (7ffff6b19660) is in userspace

Quoted 6 lines
Fixes 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the

caller")

Hi,

Tested with the patch and it fixes the lockups I was seeing with my test run. Thanks for the fix.

Quoted 102 lines

---

arch/powerpc/kernel/exceptions-64s.S  | 14 +++++++++++---
arch/powerpc/mm/book3s64/hash_utils.c | 25 +++++++++++++++++++++++++
arch/powerpc/perf/core-book3s.c       |  6 ++++++
3 files changed, 42 insertions(+), 3 deletions(-)
arch/powerpc/kernel/exceptions-64s.S +11 -3
index 0fc8bad878b2..446e54c3f71e 100644
@@ -3072,10 +3072,18 @@ do_hash_page:
 	ori	r0,r0,DSISR_BAD_FAULT_64S@l
 	and.	r0,r5,r0		/* weird error? */
 	bne-	handle_page_fault	/* if not, try to insert a HPTE */
+
+	/*
+	 * If we are in an "NMI" (e.g., an interrupt when soft-disabled), then
+	 * don't call hash_page, just fail the fault. This is required to
+	 * prevent re-entrancy problems in the hash code, namely perf
+	 * interrupts hitting while something holds H_PAGE_BUSY, and taking a
+	 * hash fault. See the comment in hash_preload().
+	 */
 	ld	r11, PACA_THREAD_INFO(r13)
-	lwz	r0,TI_PREEMPT(r11)	/* If we're in an "NMI" */
-	andis.	r0,r0,NMI_MASK@h	/* (i.e. an irq when soft-disabled) */
-	bne	77f			/* then don't call hash_page now */
+	lwz	r0,TI_PREEMPT(r11)
+	andis.	r0,r0,NMI_MASK@h
+	bne	77f
 
 	/*
 	 * r3 contains the trap number
arch/powerpc/mm/book3s64/hash_utils.c +25 -0
index 468169e33c86..9b9f92ad0e7a 100644
@@ -1559,6 +1559,7 @@ static void hash_preload(struct mm_struct *mm, pte_t *ptep, unsigned long ea,
 	pgd_t *pgdir;
 	int rc, ssize, update_flags = 0;
 	unsigned long access = _PAGE_PRESENT | _PAGE_READ | (is_exec ? _PAGE_EXEC : 0);
+	unsigned long flags;
 
 	BUG_ON(get_region_id(ea) != USER_REGION_ID);
 
@@ -1592,6 +1593,28 @@ static void hash_preload(struct mm_struct *mm, pte_t *ptep, unsigned long ea,
 		return;
 #endif /* CONFIG_PPC_64K_PAGES */
 
+	/*
+	 * __hash_page_* must run with interrupts off, as it sets the
+	 * H_PAGE_BUSY bit. It's possible for perf interrupts to hit at any
+	 * time and may take a hash fault reading the user stack, see
+	 * read_user_stack_slow() in the powerpc/perf code.
+	 *
+	 * If that takes a hash fault on the same page as we lock here, it
+	 * will bail out when seeing H_PAGE_BUSY set, and retry the access
+	 * leading to an infinite loop.
+	 *
+	 * Disabling interrupts here does not prevent perf interrupts, but it
+	 * will prevent them taking hash faults (see the NMI test in
+	 * do_hash_page), then read_user_stack's copy_from_user_nofault will
+	 * fail and perf will fall back to read_user_stack_slow(), which
+	 * walks the Linux page tables.
+	 *
+	 * Interrupts must also be off for the duration of the
+	 * mm_is_thread_local test and update, to prevent preempt running the
+	 * mm on another CPU (XXX: this may be racy vs kthread_use_mm).
+	 */
+	local_irq_save(flags);
+
 	/* Is that local to this CPU ? */
 	if (mm_is_thread_local(mm))
 		update_flags |= HPTE_LOCAL_UPDATE;
@@ -1614,6 +1637,8 @@ static void hash_preload(struct mm_struct *mm, pte_t *ptep, unsigned long ea,
 				   mm_ctx_user_psize(&mm->context),
 				   mm_ctx_user_psize(&mm->context),
 				   pte_val(*ptep));
+
+	local_irq_restore(flags);
 }
 
 /*
arch/powerpc/perf/core-book3s.c +6 -0
index cd6a742ac6ef..01d70280d287 100644
@@ -2179,6 +2179,12 @@ static void __perf_event_interrupt(struct pt_regs *regs)
 
 	perf_read_regs(regs);
 
+	/*
+	 * If perf interrupts hit in a local_irq_disable (soft-masked) region,
+	 * we consider them as NMIs. This is required to prevent hash faults on
+	 * user addresses when reading callchains. See the NMI test in
+	 * do_hash_page.
+	 */
 	nmi = perf_intr_is_nmi(regs);
 	if (nmi)
 		nmi_enter();
Signature · 1 line
2.23.0
#2 lore raw
Quoted 59 lines · Athira Rajeev
Quoted 51 lines · Nicholas Piggin

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

cpu 0x49e: Vector: 100 (System Reset) at [c00000001a4f7d70]
pc: c000000000072dc8: hash_page_mm+0x8/0x800
lr: c00000000000c5a4: do_hash_page+0x24/0x38
sp: c0002ac1cc69ac70
msr: 8000000000081033
current = 0xc0002ac1cc602e00
paca = 0xc00000001de1f280 irqmask: 0x03 irq_happened: 0x01
pid = 20118, comm = pread2_processe
Linux version 5.8.0-rc6-00345-g1fad14f18bc6
49e:mon> t
[c0002ac1cc69ac70] c00000000000c5a4 do_hash_page+0x24/0x38 (unreliable)
--- Exception: 300 (Data Access) at c00000000008fa60 __copy_tofrom_user_power7+0x20c/0x7ac
[link register ] c000000000335d10 copy_from_user_nofault+0xf0/0x150
[c0002ac1cc69af70] c00032bf9fa3c880 (unreliable)
[c0002ac1cc69afa0] c000000000109df0 read_user_stack_64+0x70/0xf0 [c0002ac1cc69afd0] c000000000109fcc perf_callchain_user_64+0x15c/0x410 [c0002ac1cc69b060] c000000000109c00 perf_callchain_user+0x20/0x40 [c0002ac1cc69b080] c00000000031c6cc get_perf_callchain+0x25c/0x360 [c0002ac1cc69b120] c000000000316b50 perf_callchain+0x70/0xa0
[c0002ac1cc69b140] c000000000316ddc perf_prepare_sample+0x25c/0x790 [c0002ac1cc69b1a0] c000000000317350 perf_event_output_forward+0x40/0xb0 [c0002ac1cc69b220] c000000000306138 __perf_event_overflow+0x88/0x1a0 [c0002ac1cc69b270] c00000000010cf70 record_and_restart+0x230/0x750 [c0002ac1cc69b620] c00000000010d69c perf_event_interrupt+0x20c/0x510 [c0002ac1cc69b730] c000000000027d9c performance_monitor_exception+0x4c/0x60 [c0002ac1cc69b750] c00000000000b2f8 performance_monitor_common_virt+0x1b8/0x1c0 --- Exception: f00 (Performance Monitor) at c0000000000cb5b0 pSeries_lpar_hpte_insert+0x0/0x160
[link register ] c0000000000846f0 __hash_page_64K+0x210/0x540
[c0002ac1cc69ba50] 0000000000000000 (unreliable)
[c0002ac1cc69bb00] c000000000073ae0 update_mmu_cache+0x390/0x3a0 [c0002ac1cc69bb70] c00000000037f024 wp_page_copy+0x364/0xce0
[c0002ac1cc69bc20] c00000000038272c do_wp_page+0xdc/0xa60
[c0002ac1cc69bc70] c0000000003857bc handle_mm_fault+0xb9c/0x1b60 [c0002ac1cc69bd50] c00000000006c434 __do_page_fault+0x314/0xc90 [c0002ac1cc69be20] c00000000000c5c8 handle_page_fault+0x10/0x2c
--- Exception: 300 (Data Access) at 00007fff8c861fe8
SP (7ffff6b19660) is in userspace

Quoted 6 lines
Fixes 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the

caller")

Quoted 6 lines

Hi,

Tested with the patch and it fixes the lockups I was seeing with my test run. Thanks for the fix.

Thanks for testing.

What test are you running?

cheers

#3 lore raw
Quoted 66 lines · Michael Ellerman
Quoted 61 lines · Athira Rajeev
Quoted 51 lines · Nicholas Piggin

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

cpu 0x49e: Vector: 100 (System Reset) at [c00000001a4f7d70]
pc: c000000000072dc8: hash_page_mm+0x8/0x800
lr: c00000000000c5a4: do_hash_page+0x24/0x38
sp: c0002ac1cc69ac70
msr: 8000000000081033
current = 0xc0002ac1cc602e00
paca = 0xc00000001de1f280 irqmask: 0x03 irq_happened: 0x01
pid = 20118, comm = pread2_processe
Linux version 5.8.0-rc6-00345-g1fad14f18bc6
49e:mon> t
[c0002ac1cc69ac70] c00000000000c5a4 do_hash_page+0x24/0x38 (unreliable)
--- Exception: 300 (Data Access) at c00000000008fa60 __copy_tofrom_user_power7+0x20c/0x7ac
[link register ] c000000000335d10 copy_from_user_nofault+0xf0/0x150
[c0002ac1cc69af70] c00032bf9fa3c880 (unreliable)
[c0002ac1cc69afa0] c000000000109df0 read_user_stack_64+0x70/0xf0 [c0002ac1cc69afd0] c000000000109fcc perf_callchain_user_64+0x15c/0x410 [c0002ac1cc69b060] c000000000109c00 perf_callchain_user+0x20/0x40 [c0002ac1cc69b080] c00000000031c6cc get_perf_callchain+0x25c/0x360 [c0002ac1cc69b120] c000000000316b50 perf_callchain+0x70/0xa0
[c0002ac1cc69b140] c000000000316ddc perf_prepare_sample+0x25c/0x790 [c0002ac1cc69b1a0] c000000000317350 perf_event_output_forward+0x40/0xb0 [c0002ac1cc69b220] c000000000306138 __perf_event_overflow+0x88/0x1a0 [c0002ac1cc69b270] c00000000010cf70 record_and_restart+0x230/0x750 [c0002ac1cc69b620] c00000000010d69c perf_event_interrupt+0x20c/0x510 [c0002ac1cc69b730] c000000000027d9c performance_monitor_exception+0x4c/0x60 [c0002ac1cc69b750] c00000000000b2f8 performance_monitor_common_virt+0x1b8/0x1c0 --- Exception: f00 (Performance Monitor) at c0000000000cb5b0 pSeries_lpar_hpte_insert+0x0/0x160
[link register ] c0000000000846f0 __hash_page_64K+0x210/0x540
[c0002ac1cc69ba50] 0000000000000000 (unreliable)
[c0002ac1cc69bb00] c000000000073ae0 update_mmu_cache+0x390/0x3a0 [c0002ac1cc69bb70] c00000000037f024 wp_page_copy+0x364/0xce0
[c0002ac1cc69bc20] c00000000038272c do_wp_page+0xdc/0xa60
[c0002ac1cc69bc70] c0000000003857bc handle_mm_fault+0xb9c/0x1b60 [c0002ac1cc69bd50] c00000000006c434 __do_page_fault+0x314/0xc90 [c0002ac1cc69be20] c00000000000c5c8 handle_page_fault+0x10/0x2c
--- Exception: 300 (Data Access) at 00007fff8c861fe8
SP (7ffff6b19660) is in userspace

Quoted 6 lines
Fixes 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the

caller")

Hi,

Quoted 8 lines
Quoted 4 lines

Tested with the patch and it fixes the lockups I was seeing with my test run. Thanks for the fix.

Thanks for testing.

What test are you running?

Hi Michael

I was running “perf record” and Unixbench tests ( https://github.com/kdlucas/byte-unixbench ) in parallel where we were getting soft lockups

1. Perf command run:

# perf record -a -g -c 10000000 -o <data_file> sleep 60

2. Unixbench tests
# Run -q -c <nr_threads> spawn

Wtth the fix, perf completes successfully.

Thanks
Athira

Quoted 2 lines

cheers

#4 lore raw
Quoted 75 lines · Athira Rajeev
Quoted 64 lines · Michael Ellerman
Quoted 61 lines · Athira Rajeev
Quoted 51 lines · Nicholas Piggin

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

cpu 0x49e: Vector: 100 (System Reset) at [c00000001a4f7d70]
pc: c000000000072dc8: hash_page_mm+0x8/0x800
lr: c00000000000c5a4: do_hash_page+0x24/0x38
sp: c0002ac1cc69ac70
msr: 8000000000081033
current = 0xc0002ac1cc602e00
paca = 0xc00000001de1f280 irqmask: 0x03 irq_happened: 0x01
pid = 20118, comm = pread2_processe
Linux version 5.8.0-rc6-00345-g1fad14f18bc6
49e:mon> t
[c0002ac1cc69ac70] c00000000000c5a4 do_hash_page+0x24/0x38 (unreliable)
--- Exception: 300 (Data Access) at c00000000008fa60 __copy_tofrom_user_power7+0x20c/0x7ac
[link register ] c000000000335d10 copy_from_user_nofault+0xf0/0x150
[c0002ac1cc69af70] c00032bf9fa3c880 (unreliable)
[c0002ac1cc69afa0] c000000000109df0 read_user_stack_64+0x70/0xf0 [c0002ac1cc69afd0] c000000000109fcc perf_callchain_user_64+0x15c/0x410 [c0002ac1cc69b060] c000000000109c00 perf_callchain_user+0x20/0x40 [c0002ac1cc69b080] c00000000031c6cc get_perf_callchain+0x25c/0x360 [c0002ac1cc69b120] c000000000316b50 perf_callchain+0x70/0xa0
[c0002ac1cc69b140] c000000000316ddc perf_prepare_sample+0x25c/0x790 [c0002ac1cc69b1a0] c000000000317350 perf_event_output_forward+0x40/0xb0 [c0002ac1cc69b220] c000000000306138 __perf_event_overflow+0x88/0x1a0 [c0002ac1cc69b270] c00000000010cf70 record_and_restart+0x230/0x750 [c0002ac1cc69b620] c00000000010d69c perf_event_interrupt+0x20c/0x510 [c0002ac1cc69b730] c000000000027d9c performance_monitor_exception+0x4c/0x60 [c0002ac1cc69b750] c00000000000b2f8 performance_monitor_common_virt+0x1b8/0x1c0 --- Exception: f00 (Performance Monitor) at c0000000000cb5b0 pSeries_lpar_hpte_insert+0x0/0x160
[link register ] c0000000000846f0 __hash_page_64K+0x210/0x540
[c0002ac1cc69ba50] 0000000000000000 (unreliable)
[c0002ac1cc69bb00] c000000000073ae0 update_mmu_cache+0x390/0x3a0 [c0002ac1cc69bb70] c00000000037f024 wp_page_copy+0x364/0xce0
[c0002ac1cc69bc20] c00000000038272c do_wp_page+0xdc/0xa60
[c0002ac1cc69bc70] c0000000003857bc handle_mm_fault+0xb9c/0x1b60 [c0002ac1cc69bd50] c00000000006c434 __do_page_fault+0x314/0xc90 [c0002ac1cc69be20] c00000000000c5c8 handle_page_fault+0x10/0x2c
--- Exception: 300 (Data Access) at 00007fff8c861fe8
SP (7ffff6b19660) is in userspace

Quoted 6 lines
Fixes 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the

caller")

Hi,

Quoted 8 lines
Quoted 4 lines

Tested with the patch and it fixes the lockups I was seeing with my test run. Thanks for the fix.

Thanks for testing.

What test are you running?

Hi Michael

Quoted 7 lines

I was running “perf record” and Unixbench tests ( https://github.com/kdlucas/byte-unixbench ) in parallel where we were getting soft lockups

1. Perf command run:

# perf record -a -g -c 10000000 -o <data_file> sleep 60

2. Unixbench tests
# Run -q -c <nr_threads> spawn

Thanks, I can reproduce it with that.

cheers

#5 lore raw
Quoted 82 lines · Michael Ellerman
Quoted 77 lines · Athira Rajeev
Quoted 64 lines · Michael Ellerman
Quoted 61 lines · Athira Rajeev
Quoted 51 lines · Nicholas Piggin

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

cpu 0x49e: Vector: 100 (System Reset) at [c00000001a4f7d70]
pc: c000000000072dc8: hash_page_mm+0x8/0x800
lr: c00000000000c5a4: do_hash_page+0x24/0x38
sp: c0002ac1cc69ac70
msr: 8000000000081033
current = 0xc0002ac1cc602e00
paca = 0xc00000001de1f280 irqmask: 0x03 irq_happened: 0x01
pid = 20118, comm = pread2_processe
Linux version 5.8.0-rc6-00345-g1fad14f18bc6
49e:mon> t
[c0002ac1cc69ac70] c00000000000c5a4 do_hash_page+0x24/0x38 (unreliable)
--- Exception: 300 (Data Access) at c00000000008fa60 __copy_tofrom_user_power7+0x20c/0x7ac
[link register ] c000000000335d10 copy_from_user_nofault+0xf0/0x150
[c0002ac1cc69af70] c00032bf9fa3c880 (unreliable)
[c0002ac1cc69afa0] c000000000109df0 read_user_stack_64+0x70/0xf0 [c0002ac1cc69afd0] c000000000109fcc perf_callchain_user_64+0x15c/0x410 [c0002ac1cc69b060] c000000000109c00 perf_callchain_user+0x20/0x40 [c0002ac1cc69b080] c00000000031c6cc get_perf_callchain+0x25c/0x360 [c0002ac1cc69b120] c000000000316b50 perf_callchain+0x70/0xa0
[c0002ac1cc69b140] c000000000316ddc perf_prepare_sample+0x25c/0x790 [c0002ac1cc69b1a0] c000000000317350 perf_event_output_forward+0x40/0xb0 [c0002ac1cc69b220] c000000000306138 __perf_event_overflow+0x88/0x1a0 [c0002ac1cc69b270] c00000000010cf70 record_and_restart+0x230/0x750 [c0002ac1cc69b620] c00000000010d69c perf_event_interrupt+0x20c/0x510 [c0002ac1cc69b730] c000000000027d9c performance_monitor_exception+0x4c/0x60 [c0002ac1cc69b750] c00000000000b2f8 performance_monitor_common_virt+0x1b8/0x1c0 --- Exception: f00 (Performance Monitor) at c0000000000cb5b0 pSeries_lpar_hpte_insert+0x0/0x160
[link register ] c0000000000846f0 __hash_page_64K+0x210/0x540
[c0002ac1cc69ba50] 0000000000000000 (unreliable)
[c0002ac1cc69bb00] c000000000073ae0 update_mmu_cache+0x390/0x3a0 [c0002ac1cc69bb70] c00000000037f024 wp_page_copy+0x364/0xce0
[c0002ac1cc69bc20] c00000000038272c do_wp_page+0xdc/0xa60
[c0002ac1cc69bc70] c0000000003857bc handle_mm_fault+0xb9c/0x1b60 [c0002ac1cc69bd50] c00000000006c434 __do_page_fault+0x314/0xc90 [c0002ac1cc69be20] c00000000000c5c8 handle_page_fault+0x10/0x2c
--- Exception: 300 (Data Access) at 00007fff8c861fe8
SP (7ffff6b19660) is in userspace

Quoted 6 lines
Fixes 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the

caller")

Hi,

Quoted 8 lines
Quoted 4 lines

Tested with the patch and it fixes the lockups I was seeing with my test run. Thanks for the fix.

Thanks for testing.

What test are you running?

Hi Michael

I was running “perf record” and Unixbench tests ( https://github.com/kdlucas/byte-unixbench ) in parallel where we were getting soft lockups

Quoted 7 lines
Quoted 5 lines

1. Perf command run:

# perf record -a -g -c 10000000 -o <data_file> sleep 60

2. Unixbench tests
# Run -q -c <nr_threads> spawn

Thanks, I can reproduce it with that.

Sure Michael

Quoted 2 lines

cheers

#6 lore raw
In reply to Nicholas Piggin
Quoted 6 lines · Nicholas Piggin

Commit 2f92447f9f96 ("powerpc/book3s64/hash: Use the pte_t address from the caller") removed the local_irq_disable from hash_preload, but it was required for more than just the page table walk: the hash pte busy bit is effectively a lock which may be taken in interrupt context, and the local update flag test must not be preempted before it's used.

Quoted 6 lines

This solves apparent lockups with perf interrupting __hash_page_64K. If get_perf_callchain then also takes a hash fault on the same page while it is already locked, it will loop forever taking hash faults, which looks like this:

[...]

Applied to powerpc/fixes.

[1/1] powerpc/64s/hash: Fix hash_preload running with interrupts enabled

      https://git.kernel.org/powerpc/c/909adfc66b9a1db21b5e8733e9ebfa6cd5135d74

cheers