Skip to contentSkip to conversation

Ftrace: irqsoff tracer may cause stack overflow

PATCH In review Li Yi 4 participants 8 messages Jan 8, 2010

Subsystems (4)
lore
Share & tools
lore

Li Yi

author
#0 lore raw

PATCH 1/1 Ftrace: irqsoff tracer may cause stack overflow

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of, e.g. trace_hardirqs_off().

trace_hardirqs_off() -> start_critical_timing() -> atomic_inc() -> atomic_add_return() -> local_irq_save() -> trace_hardirqs_off()

---
 include/asm-generic/atomic.h |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)
include/asm-generic/atomic.h +0 -0
index c99c64d..c3e5f4b 100644
@@ -60,11 +60,11 @@ static inline int atomic_add_return(int i, atomic_t

*v)
unsigned long flags;
int temp;

- local_irq_save(flags);
+ raw_local_irq_save(flags);
temp = v->counter;
temp += i;
v->counter = temp;
- local_irq_restore(flags);
+ raw_local_irq_restore(flags);

return temp;
}
@@ -82,11 +82,11 @@ static inline int atomic_sub_return(int i, atomic_t *v)
unsigned long flags;
int temp;

- local_irq_save(flags);
+ raw_local_irq_save(flags);
temp = v->counter;
temp -= i;
v->counter = temp;
- local_irq_restore(flags);
+ raw_local_irq_restore(flags);

return temp;
}
@@ -139,9 +139,9 @@ static inline void atomic_clear_mask(unsigned long mask, unsigned long *addr)
unsigned long flags;

mask = ~mask;
- local_irq_save(flags);
+ raw_local_irq_save(flags);
*addr &= mask;
- local_irq_restore(flags);
+ raw_local_irq_restore(flags);
}

#define atomic_xchg(ptr, v) (xchg(&(ptr)->counter, (v)))

Signature · 1 line
1.6.0.4
#1 lore raw
Quoted 10 lines · Li Yi

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of, e.g. trace_hardirqs_off().

trace_hardirqs_off() -> start_critical_timing() -> atomic_inc() -> atomic_add_return() -> local_irq_save() -> trace_hardirqs_off()

---
 include/asm-generic/atomic.h |   12 ++++++------

Arnd is watching over asm-generic/ atm

Review static inline void atomic_clear_mask(unsigned long · line 139

Quoted 6 lines
(quoted diff) +0 -0
@@ -60,11 +60,11 @@ static inline int atomic_add_return(int i, atomic_t
 *v)
@@ -82,11 +82,11 @@ static inline int atomic_sub_return(int i, atomic_t
 *v)
@@ -139,9 +139,9 @@ static inline void atomic_clear_mask(unsigned long
 mask, unsigned long *addr)

looks like your mailer has line wrapping problems ... might want to fix that. maybe just find a smtp server in ADI shanghai's office to send to and use sendemail.smtpersver in your ~/.gitconfig ? -mike

author
#2 lore raw
Quoted 12 lines · Mike Frysinger

Review static inline void atomic_clear_mask(unsigned long · line 139

Quoted 6 lines
(quoted diff) +0 -0
@@ -60,11 +60,11 @@ static inline int atomic_add_return(int i, atomic_t
 *v)
@@ -82,11 +82,11 @@ static inline int atomic_sub_return(int i, atomic_t
 *v)
@@ -139,9 +139,9 @@ static inline void atomic_clear_mask(unsigned long
 mask, unsigned long *addr)

looks like your mailer has line wrapping problems ... might want to fix that. maybe just find a smtp server in ADI shanghai's office to send to and use sendemail.smtpersver in your ~/.gitconfig ? -mike

Thanks. Disabled auto-wrap and resend:

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of trace_hardirqs_off().

trace_hardirqs_off() -> start_critical_timing() -> atomic_inc() -> atomic_add_return() -> local_irq_save() -> trace_hardirqs_off()

---
 include/asm-generic/atomic.h |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)
include/asm-generic/atomic.h +6 -6
index c99c64d..c3e5f4b 100644
@@ -60,11 +60,11 @@ static inline int atomic_add_return(int i, atomic_t *v)
 	unsigned long flags;
 	int temp;
 
-	local_irq_save(flags);
+	raw_local_irq_save(flags);
 	temp = v->counter;
 	temp += i;
 	v->counter = temp;
-	local_irq_restore(flags);
+	raw_local_irq_restore(flags);
 
 	return temp;
 }
@@ -82,11 +82,11 @@ static inline int atomic_sub_return(int i, atomic_t *v)
 	unsigned long flags;
 	int temp;
 
-	local_irq_save(flags);
+	raw_local_irq_save(flags);
 	temp = v->counter;
 	temp -= i;
 	v->counter = temp;
-	local_irq_restore(flags);
+	raw_local_irq_restore(flags);
 
 	return temp;
 }
@@ -139,9 +139,9 @@ static inline void atomic_clear_mask(unsigned long mask, unsigned long *addr)
 	unsigned long flags;
 
 	mask = ~mask;
-	local_irq_save(flags);
+	raw_local_irq_save(flags);
 	*addr &= mask;
-	local_irq_restore(flags);
+	raw_local_irq_restore(flags);
 }
 
 #define atomic_xchg(ptr, v)		(xchg(&(ptr)->counter, (v)))
Signature · 1 line
1.6.0.4
#3 lore raw
Quoted 12 lines · Li Yi
Quoted 10 lines · Mike Frysinger

Review static inline void atomic_clear_mask(unsigned long · line 139

Quoted 6 lines
(quoted diff) +0 -0
@@ -60,11 +60,11 @@ static inline int atomic_add_return(int i, atomic_t
 *v)
@@ -82,11 +82,11 @@ static inline int atomic_sub_return(int i, atomic_t
 *v)
@@ -139,9 +139,9 @@ static inline void atomic_clear_mask(unsigned long
 mask, unsigned long *addr)

looks like your mailer has line wrapping problems ... might want to fix that.  maybe just find a smtp server in ADI shanghai's office to send to and use sendemail.smtpersver in your ~/.gitconfig ?

Quoted 3 lines

Thanks. Disabled auto-wrap and resend:

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of trace_hardirqs_off().

well now you have the opposite problem ... your changelog needs to be line wrapped
-mike

#4 lore raw
In reply to Li Yi
Quoted 8 lines · Li Yi

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of, e.g. trace_hardirqs_off().

trace_hardirqs_off() -> start_critical_timing() -> atomic_inc() -> atomic_add_return() -> local_irq_save() -> trace_hardirqs_off()

Good catch!

However, may be we should keep the local_irq_save there and have __raw_atomic_* versions only for tracing.

It's better to keep track of most irq disabled sites.

Why not something like the following (untested):

include/asm-generic/atomic.h +14 -6
index c99c64d..ffc6772 100644
@@ -17,6 +17,14 @@
 #error not SMP safe
 #endif
 
+#ifdef __ATOMIC_NEED_RAW_IRQ_SAVE
+#define __atomic_op_irq_save(f)		raw_local_irq_save(f)
+#define __atomic_op_irq_restore(f)	raw_local_irq_restore(f)
+#else
+#define __atomic_op_irq_save(f)		local_irq_save(f)
+#define __atomic_op_irq_restore(f)	local_irq_restore(f)
+#endif
+
 /*
  * Atomic operations that C can't guarantee us.  Useful for
  * resource counting etc..
@@ -60,11 +68,11 @@ static inline int atomic_add_return(int i, atomic_t *v)
 	unsigned long flags;
 	int temp;
 
-	local_irq_save(flags);
+	__atomic_op_irq_save(flags);
 	temp = v->counter;
 	temp += i;
 	v->counter = temp;
-	local_irq_restore(flags);
+	__atomic_op_irq_restore(flags);
 
 	return temp;
 }
@@ -82,11 +90,11 @@ static inline int atomic_sub_return(int i, atomic_t *v)
 	unsigned long flags;
 	int temp;
 
-	local_irq_save(flags);
+	__atomic_op_irq_save(flags);
 	temp = v->counter;
 	temp -= i;
 	v->counter = temp;
-	local_irq_restore(flags);
+	__atomic_op_irq_restore(flags);
 
 	return temp;
 }
@@ -139,9 +147,9 @@ static inline void atomic_clear_mask(unsigned long mask, unsigned long *addr)
 	unsigned long flags;
 
 	mask = ~mask;
-	local_irq_save(flags);
+	__atomic_op_irq_save(flags);
 	*addr &= mask;
-	local_irq_restore(flags);
+	__atomic_op_irq_restore(flags);
 }
 
 #define atomic_xchg(ptr, v)		(xchg(&(ptr)->counter, (v)))
kernel/trace/trace_irqsoff.c +3 -0
index 2974bc7..6bcb1d1 100644
@@ -9,6 +9,9 @@
  *  Copyright (C) 2004-2006 Ingo Molnar
  *  Copyright (C) 2004 William Lee Irwin III
  */
+
+#define __ATOMIC_NEED_RAW_IRQ_SAVE
+
 #include <linux/kallsyms.h>
 #include <linux/debugfs.h>
 #include <linux/uaccess.h>
author
#5 lore raw
Quoted 12 lines · Frederic Weisbecker
Quoted 8 lines · Li Yi

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of, e.g. trace_hardirqs_off().

trace_hardirqs_off() -> start_critical_timing() -> atomic_inc() -> atomic_add_return() -> local_irq_save() -> trace_hardirqs_off()

Quoted 8 lines

Good catch!

However, may be we should keep the local_irq_save there and have __raw_atomic_* versions only for tracing.

It's better to keep track of most irq disabled sites.

Why not something like the following (untested):

Yes. I agree this is better solution.

-Yi

maintainer
#6 lore raw
In reply to Frederic Weisbecker
Quoted 5 lines · Frederic Weisbecker
Quoted 4 lines · Li Yi

"irqsoff" tracer may cause stack overflow on architectures using asm-generic/atomic.h, due to recursive invoking of, e.g. trace_hardirqs_off().

Quoted 8 lines
Quoted 4 lines

trace_hardirqs_off() -> start_critical_timing() -> atomic_inc() -> atomic_add_return() -> local_irq_save() -> trace_hardirqs_off()

Good catch!

Yes, nice catch indeed. Hmm, I wonder if lockdep has any issues here as well? /me looks, no it uses current->lockdep_recursion++, where trace hardirqsoff uses atomic_inc_return :-/

Quoted 14 lines

However, may be we should keep the local_irq_save there and have __raw_atomic_* versions only for tracing.

It's better to keep track of most irq disabled sites.

Why not something like the following (untested):

include/asm-generic/atomic.h +0 -0
index c99c64d..ffc6772 100644

include/asm-generic/atomic.h line 18

Quoted 2 lines
include/asm-generic/atomic.h +0 -0
@@ -17,6 +17,14 @@
 #error not SMP safe
 #endif

Comment needed here:

/*
* The irqsoff tracer uses atomic_inc_return to prevent recursion.
* Unfortunately, in this file, atomic_inc_return disables interrupts
* which causes the recursion the irqsoff trace was trying to prevent. *
* The irqsoff tracer will define __ATOMIC_NEED_RAW_IRQ_SAVE before
* including this file, which will make the atomic_inc_return use
* the raw versions of interrupts disabling. This will allow other
* users of the atomic_inc_return to still have the interrupt
* disabling be traced, but will prevent the recursion by the
* irqsoff tracer itself.
*/

Quoted 58 lines
include/asm-generic/atomic.h +14 -6
+#ifdef __ATOMIC_NEED_RAW_IRQ_SAVE
+#define __atomic_op_irq_save(f)		raw_local_irq_save(f)
+#define __atomic_op_irq_restore(f)	raw_local_irq_restore(f)
+#else
+#define __atomic_op_irq_save(f)		local_irq_save(f)
+#define __atomic_op_irq_restore(f)	local_irq_restore(f)
+#endif
+
 /*
  * Atomic operations that C can't guarantee us.  Useful for
  * resource counting etc..
@@ -60,11 +68,11 @@ static inline int atomic_add_return(int i, atomic_t *v)
 	unsigned long flags;
 	int temp;
 
-	local_irq_save(flags);
+	__atomic_op_irq_save(flags);
 	temp = v->counter;
 	temp += i;
 	v->counter = temp;
-	local_irq_restore(flags);
+	__atomic_op_irq_restore(flags);
 
 	return temp;
 }
@@ -82,11 +90,11 @@ static inline int atomic_sub_return(int i, atomic_t *v)
 	unsigned long flags;
 	int temp;
 
-	local_irq_save(flags);
+	__atomic_op_irq_save(flags);
 	temp = v->counter;
 	temp -= i;
 	v->counter = temp;
-	local_irq_restore(flags);
+	__atomic_op_irq_restore(flags);
 
 	return temp;
 }
@@ -139,9 +147,9 @@ static inline void atomic_clear_mask(unsigned long mask, unsigned long *addr)
 	unsigned long flags;
 
 	mask = ~mask;
-	local_irq_save(flags);
+	__atomic_op_irq_save(flags);
 	*addr &= mask;
-	local_irq_restore(flags);
+	__atomic_op_irq_restore(flags);
 }
 
 #define atomic_xchg(ptr, v)		(xchg(&(ptr)->counter, (v)))
kernel/trace/trace_irqsoff.c +0 -0
index 2974bc7..6bcb1d1 100644

kernel/trace/trace_irqsoff.c line 17

Quoted 9 lines
kernel/trace/trace_irqsoff.c +3 -0
@@ -9,6 +9,9 @@
  *  Copyright (C) 2004-2006 Ingo Molnar
  *  Copyright (C) 2004 William Lee Irwin III
  */
+
+#define __ATOMIC_NEED_RAW_IRQ_SAVE
+
 #include <linux/kallsyms.h>
 #include <linux/debugfs.h>
 #include <linux/uaccess.h>

I wonder if we could just use a per_cpu variable and increment that instead. Since the irqsoff tracer only gets called with interrupts disabled (and the preemptoff with preemption disabled), a per_cpu variable should be protected well.

-- Steve

#7 lore raw
Quoted 16 lines · Steven Rostedt

On Fri, 2010-01-08 at 06:18 +0100, Frederic Weisbecker wrote:
Comment needed here:

/*
* The irqsoff tracer uses atomic_inc_return to prevent recursion.
* Unfortunately, in this file, atomic_inc_return disables interrupts
* which causes the recursion the irqsoff trace was trying to prevent. *
* The irqsoff tracer will define __ATOMIC_NEED_RAW_IRQ_SAVE before
* including this file, which will make the atomic_inc_return use
* the raw versions of interrupts disabling. This will allow other
* users of the atomic_inc_return to still have the interrupt
* disabling be traced, but will prevent the recursion by the
* irqsoff tracer itself.
*/

Yep, that was a first catch, just to ping opinions, it was even not tested :)

Quoted 4 lines

I wonder if we could just use a per_cpu variable and increment that instead. Since the irqsoff tracer only gets called with interrupts disabled (and the preemptoff with preemption disabled), a per_cpu variable should be protected well.

Doh! I thought about that but feared about preempt_disable recursion. I didn't realize this code was under such context already.

True, that's indeed a much better idea!