[PATCH 2/3] perf_counter: allow arch to supply event misc flags and instruction pointer

[PATCH 2/3] perf_counter: allow arch to supply event misc flags and instruction pointer

[Paul Mackerras]

At present the values we put in overflow events for the misc flags
indicating processor mode and the instruction pointer are obtained
using the standard user_mode() and instruction_pointer() functions.
Those functions tell you where the performance monitor interrupt
was taken, which might not be exactly where the counter overflow
occurred, for example because interrupts were disabled at the point
where the overflow occurred, or because the processor had many
instructions in flight and chose to complete some more instructions
beyond the one that caused the counter overflow.

Some architectures (e.g. powerpc) can supply more precise information
about where the counter overflow occurred and the processor mode at
that point.  This introduces new functions, perf_misc_flags() and
perf_instruction_pointer(), which arch code can override to provide
more precise information if available.  They have default implementations
which are identical to the existing code.

This also adds a new misc flag value, PERF_EVENT_MISC_HYPERVISOR, for
the case where a counter overflow occurred in the hypervisor.

Signed-off-by: Paul Mackerras <paulus@samba.org>
---
 include/linux/perf_counter.h |    7 +++++++
 kernel/perf_counter.c        |    5 ++---
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index 614f921..2b12d98 100644
--- a/include/linux/perf_counter.h
+++ b/include/linux/perf_counter.h
@@ -213,6 +213,7 @@ struct perf_counter_mmap_page {
 #define PERF_EVENT_MISC_KERNEL		(1 << 0)
 #define PERF_EVENT_MISC_USER		(1 << 1)
 #define PERF_EVENT_MISC_OVERFLOW	(1 << 2)
+#define PERF_EVENT_MISC_HYPERVISOR	(1 << 3)
 
 struct perf_event_header {
 	__u32	type;
@@ -588,6 +589,12 @@ extern int sysctl_perf_counter_mlock;
 
 extern void perf_counter_init(void);
 
+#ifndef perf_misc_flags
+#define perf_misc_flags(regs)	(user_mode(regs) ? PERF_EVENT_MISC_USER : \
+				 PERF_EVENT_MISC_KERNEL)
+#define perf_instruction_pointer(regs)	instruction_pointer(regs)
+#endif
+
 #else
 static inline void
 perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index ff166c1..89568de 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -1978,11 +1978,10 @@ static void perf_counter_output(struct perf_counter *counter,
 	header.size = sizeof(header);
 
 	header.misc = PERF_EVENT_MISC_OVERFLOW;
-	header.misc |= user_mode(regs) ?
-		PERF_EVENT_MISC_USER : PERF_EVENT_MISC_KERNEL;
+	header.misc |= perf_misc_flags(regs);
 
 	if (record_type & PERF_RECORD_IP) {
-		ip = instruction_pointer(regs);
+		ip = perf_instruction_pointer(regs);
 		header.type |= PERF_RECORD_IP;
 		header.size += sizeof(ip);
 	}
-- 
1.5.5.rc3.7.gba13

Re: [PATCH 2/3] perf_counter: allow arch to supply event misc flags and instruction pointer

[Peter Zijlstra]

On Thu, 2009-05-14 at 13:29 +1000, Paul Mackerras wrote:
> At present the values we put in overflow events for the misc flags
> indicating processor mode and the instruction pointer are obtained
> using the standard user_mode() and instruction_pointer() functions.
> Those functions tell you where the performance monitor interrupt
> was taken, which might not be exactly where the counter overflow
> occurred, for example because interrupts were disabled at the point
> where the overflow occurred, or because the processor had many
> instructions in flight and chose to complete some more instructions
> beyond the one that caused the counter overflow.
> 
> Some architectures (e.g. powerpc) can supply more precise information
> about where the counter overflow occurred and the processor mode at
> that point.  This introduces new functions, perf_misc_flags() and
> perf_instruction_pointer(), which arch code can override to provide
> more precise information if available.  They have default implementations
> which are identical to the existing code.
> 
> This also adds a new misc flag value, PERF_EVENT_MISC_HYPERVISOR, for
> the case where a counter overflow occurred in the hypervisor.
> 
> Signed-off-by: Paul Mackerras <paulus@samba.org>
> ---
>  include/linux/perf_counter.h |    7 +++++++
>  kernel/perf_counter.c        |    5 ++---
>  2 files changed, 9 insertions(+), 3 deletions(-)
> 
> diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
> index 614f921..2b12d98 100644
> --- a/include/linux/perf_counter.h
> +++ b/include/linux/perf_counter.h
> @@ -213,6 +213,7 @@ struct perf_counter_mmap_page {
>  #define PERF_EVENT_MISC_KERNEL		(1 << 0)
>  #define PERF_EVENT_MISC_USER		(1 << 1)
>  #define PERF_EVENT_MISC_OVERFLOW	(1 << 2)
> +#define PERF_EVENT_MISC_HYPERVISOR	(1 << 3)

Ah, I think I've taught userspace that either !USER && !KERNEL or USER
&& KERNEL is HV information, since neither really makes sense :-)

>  struct perf_event_header {
>  	__u32	type;
> @@ -588,6 +589,12 @@ extern int sysctl_perf_counter_mlock;
>  
>  extern void perf_counter_init(void);
>  
> +#ifndef perf_misc_flags
> +#define perf_misc_flags(regs)	(user_mode(regs) ? PERF_EVENT_MISC_USER : \
> +				 PERF_EVENT_MISC_KERNEL)
> +#define perf_instruction_pointer(regs)	instruction_pointer(regs)
> +#endif
> +
>  #else
>  static inline void
>  perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
> diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
> index ff166c1..89568de 100644
> --- a/kernel/perf_counter.c
> +++ b/kernel/perf_counter.c
> @@ -1978,11 +1978,10 @@ static void perf_counter_output(struct perf_counter *counter,
>  	header.size = sizeof(header);
>  
>  	header.misc = PERF_EVENT_MISC_OVERFLOW;
> -	header.misc |= user_mode(regs) ?
> -		PERF_EVENT_MISC_USER : PERF_EVENT_MISC_KERNEL;
> +	header.misc |= perf_misc_flags(regs);
>  
>  	if (record_type & PERF_RECORD_IP) {
> -		ip = instruction_pointer(regs);
> +		ip = perf_instruction_pointer(regs);
>  		header.type |= PERF_RECORD_IP;
>  		header.size += sizeof(ip);
>  	}

Re: [PATCH 2/3] perf_counter: allow arch to supply event misc flags and instruction pointer

[Paul Mackerras]

Peter Zijlstra writes:

> On Thu, 2009-05-14 at 13:29 +1000, Paul Mackerras wrote:

> >  #define PERF_EVENT_MISC_KERNEL		(1 << 0)
> >  #define PERF_EVENT_MISC_USER		(1 << 1)
> >  #define PERF_EVENT_MISC_OVERFLOW	(1 << 2)
> > +#define PERF_EVENT_MISC_HYPERVISOR	(1 << 3)
> 
> Ah, I think I've taught userspace that either !USER && !KERNEL or USER
> && KERNEL is HV information, since neither really makes sense :-)

OK, should we do something like this?

#define PERF_EVENT_MISC_CPUMODE_MASK	(3 << 0)
#define PERF_EVENT_MISC_CPUMODE_UNKNOWN	(0 << 0)
#define PERF_EVENT_MISC_USER		(1 << 0)
#define PERF_EVENT_MISC_KERNEL		(2 << 0)
#define PERF_EVENT_MISC_HYPERVISOR	(3 << 0)

Paul.

Re: [PATCH 2/3] perf_counter: allow arch to supply event misc flags and instruction pointer

[Peter Zijlstra]

On Thu, 2009-05-14 at 19:47 +1000, Paul Mackerras wrote:
> Peter Zijlstra writes:
> 
> > On Thu, 2009-05-14 at 13:29 +1000, Paul Mackerras wrote:
> 
> > >  #define PERF_EVENT_MISC_KERNEL		(1 << 0)
> > >  #define PERF_EVENT_MISC_USER		(1 << 1)
> > >  #define PERF_EVENT_MISC_OVERFLOW	(1 << 2)
> > > +#define PERF_EVENT_MISC_HYPERVISOR	(1 << 3)
> > 
> > Ah, I think I've taught userspace that either !USER && !KERNEL or USER
> > && KERNEL is HV information, since neither really makes sense :-)
> 
> OK, should we do something like this?
> 
> #define PERF_EVENT_MISC_CPUMODE_MASK	(3 << 0)
> #define PERF_EVENT_MISC_CPUMODE_UNKNOWN	(0 << 0)
> #define PERF_EVENT_MISC_USER		(1 << 0)
> #define PERF_EVENT_MISC_KERNEL		(2 << 0)
> #define PERF_EVENT_MISC_HYPERVISOR	(3 << 0)

Yep, that reads nicer :-)