Skip to contentSkip to conversation

powerpc: Fix possible deadlock on page fault

PATCH In review Aneesh Kumar K.V 3 participants 5 messages Sep 5, 2013 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc: Fix possible deadlock on page fault

From: "Aneesh Kumar K.V" <[email protected]>

stack_grow_into/14082 is trying to acquire lock:
(&mm->mmap_sem){++++++}, at: [<c000000000206d28>] .might_fault+0x78/0xe0

but task is already holding lock:
(&mm->mmap_sem){++++++}, at: [<c0000000007ffd8c>] .do_page_fault+0x24c/0x910

other info that might help us debug this:

  Possible unsafe locking scenario:
        CPU0
        ----

lock(&mm->mmap_sem);
lock(&mm->mmap_sem);

*** DEADLOCK ***

May be due to missing lock nesting notation

1 lock held by stack_grow_into/14082:
#0: (&mm->mmap_sem){++++++}, at: [<c0000000007ffd8c>] .do_page_fault+0x24c/0x910

 stack backtrace:
 CPU: 21 PID: 14082 Comm: stack_grow_into Not tainted 3.10.0-10.el7.ppc64.debug #1
 Call Trace:
 [c0000003d396b850] [c000000000016e7c] .show_stack+0x7c/0x1f0 (unreliable)
 [c0000003d396b920] [c000000000813fc8] .dump_stack+0x28/0x3c
 [c0000003d396b990] [c000000000124b90] .__lock_acquire+0x1640/0x1800
 [c0000003d396bab0] [c00000000012570c] .lock_acquire+0xac/0x250
 [c0000003d396bb80] [c000000000206d54] .might_fault+0xa4/0xe0
 [c0000003d396bbf0] [c0000000007ffe2c] .do_page_fault+0x2ec/0x910
 [c0000003d396be30] [c0000000000092e8] handle_page_fault+0x10/0x30
---
 arch/powerpc/mm/fault.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
arch/powerpc/mm/fault.c +9 -2
index 8726779..dc2902a 100644
@@ -206,7 +206,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 	int trap = TRAP(regs);
  	int is_exec = trap == 0x400;
 	int fault;
-	int rc = 0;
+	int rc = 0, store_update;
 
 #if !(defined(CONFIG_4xx) || defined(CONFIG_BOOKE))
 	/*
@@ -280,6 +280,13 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	store_update = store_updates_sp(regs);
+
 	/* When running in the kernel we expect faults to occur only to
 	 * addresses in user space.  All other faults represent errors in the
 	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
@@ -346,7 +353,7 @@ retry:
 		 * expand the stack rather than segfaulting.
 		 */
 		if (address + 2048 < uregs->gpr[1]
-		    && (!user_mode(regs) || !store_updates_sp(regs)))
+		    && (!user_mode(regs) || !store_update))
 			goto bad_area;
 	}
 	if (expand_stack(vma, address))
Signature · 1 line
1.8.1.2
#1 lore raw

arch/powerpc/mm/fault.c int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, · line 288

Quoted 10 lines · Aneesh Kumar K.V
arch/powerpc/mm/fault.c +6 -0
@@ -280,6 +280,13 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	store_update = store_updates_sp(regs);

We should only call store_updates_sp() if user_mode(regs); that was the previous behaviour.

Paul.

#2 lore raw
Quoted 6 lines · Paul Mackerras
Quoted 4 lines · Aneesh Kumar K.V
(quoted diff) +0 -0
@@ -280,6 +280,13 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
Quoted 9 lines

Review int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, · line 288

Quoted 6 lines
(quoted diff) +6 -0
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	store_update = store_updates_sp(regs);

We should only call store_updates_sp() if user_mode(regs); that was the previous behaviour.

Updated to

arch/powerpc/mm/fault.c +10 -3
index 8726779..fad7af6 100644
@@ -206,7 +206,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 	int trap = TRAP(regs);
  	int is_exec = trap == 0x400;
 	int fault;
-	int rc = 0;
+	int rc = 0, store_update = 0;
 
 #if !(defined(CONFIG_4xx) || defined(CONFIG_BOOKE))
 	/*
@@ -280,6 +280,14 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	if (user_mode(regs))
+		store_update = store_updates_sp(regs);
+
 	/* When running in the kernel we expect faults to occur only to
 	 * addresses in user space.  All other faults represent errors in the
 	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
@@ -345,8 +353,7 @@ retry:
 		 * between the last mapped region and the stack will
 		 * expand the stack rather than segfaulting.
 		 */
-		if (address + 2048 < uregs->gpr[1]
-		    && (!user_mode(regs) || !store_updates_sp(regs)))
+		if (address + 2048 < uregs->gpr[1] && !store_update)
 			goto bad_area;
 	}
 	if (expand_stack(vma, address))
#3 lore raw
Quoted 25 lines · Aneesh Kumar K.V
Quoted 6 lines · Paul Mackerras
Quoted 4 lines · Aneesh Kumar K.V
arch/powerpc/mm/fault.c +0 -0
@@ -280,6 +280,13 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
Quoted 9 lines

arch/powerpc/mm/fault.c int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, · line 289

Quoted 6 lines
arch/powerpc/mm/fault.c +6 -0
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	store_update = store_updates_sp(regs);

We should only call store_updates_sp() if user_mode(regs); that was the previous behaviour.

Updated to

arch/powerpc/mm/fault.c +0 -0
index 8726779..fad7af6 100644

arch/powerpc/mm/fault.c int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, · line 209

Quoted 5 lines
arch/powerpc/mm/fault.c +1 -1
@@ -206,7 +206,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 	int trap = TRAP(regs);
  	int is_exec = trap == 0x400;
 	int fault;
-	int rc = 0;
+	int rc = 0, store_update = 0;

Keep the "sp", in the name, it's confusing otherwise. It's not just about "store update", it's about specifically recognizing instructions used to update the stack frame in order to let them and only them significantly lower the stack pointer.

Cheers,
Ben.

Quoted 27 lines
arch/powerpc/mm/fault.c +9 -2
 #if !(defined(CONFIG_4xx) || defined(CONFIG_BOOKE))
 	/*
@@ -280,6 +280,14 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	if (user_mode(regs))
+		store_update = store_updates_sp(regs);
+
 	/* When running in the kernel we expect faults to occur only to
 	 * addresses in user space.  All other faults represent errors in the
 	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
@@ -345,8 +353,7 @@ retry:
 		 * between the last mapped region and the stack will
 		 * expand the stack rather than segfaulting.
 		 */
-		if (address + 2048 < uregs->gpr[1]
-		    && (!user_mode(regs) || !store_updates_sp(regs)))
+		if (address + 2048 < uregs->gpr[1] && !store_update)
 			goto bad_area;
 	}
 	if (expand_stack(vma, address))
#4 lore raw
Quoted 21 lines · Benjamin Herrenschmidt
Quoted 20 lines · Aneesh Kumar K.V
Quoted 6 lines · Paul Mackerras
Quoted 4 lines · Aneesh Kumar K.V
arch/powerpc/mm/fault.c +0 -0
@@ -280,6 +280,13 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 
Quoted 9 lines

arch/powerpc/mm/fault.c int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, · line 289

Quoted 6 lines
arch/powerpc/mm/fault.c +6 -0
+	/*
+	 * We want to do this outside mmap_sem, because reading code around nip
+	 * can result in fault, which will cause a deadlock when called with
+	 * mmap_sem held
+	 */
+	store_update = store_updates_sp(regs);

We should only call store_updates_sp() if user_mode(regs); that was the previous behaviour.

Updated to

Quoted 16 lines

arch/powerpc/mm/fault.c int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address, · line 209

Quoted 10 lines
arch/powerpc/mm/fault.c +1 -1
index 8726779..fad7af6 100644
@@ -206,7 +206,7 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 	int trap = TRAP(regs);
  	int is_exec = trap == 0x400;
 	int fault;
-	int rc = 0;
+	int rc = 0, store_update = 0;

Keep the "sp", in the name, it's confusing otherwise. It's not just about "store update", it's about specifically recognizing instructions used to update the stack frame in order to let them and only them significantly lower the stack pointer.

Ok will do that. I posted a v2. So will wait for other feedback before i post a new version.

-aneesh