Prev: [PATCH v2] Documentation/sysctl/vm.txt typo
Next: perf tools: allow cross compiling with DWARF support
From: Will Deacon on 25 Jun 2010 07:20
Hi Peter,

On Thu, 2010-06-24 at 15:28 +0100, Peter Zijlstra wrote:
> These patches prepare the perf code for multiple pmus (no user
> interface yet, Lin Ming is working on that). These patches remove all
> weak functions and rework the struct pmu interface.
>
> The patches are boot tested on x86_64 and compile tested on: powerpc
> (!fsl, what config is that?), sparc and arm (sorry no SH compiler)
>

I tried to test these for ARM, but I get compilation errors. Maybe you
forgot to enable perf in the ARM .config? Anyway, the compiler log is:

CC arch/arm/kernel/perf_event.o
arch/arm/kernel/perf_event.c: In function 'validate_event':
arch/arm/kernel/perf_event.c:342: error: 'pmu' undeclared (first use in this function)
arch/arm/kernel/perf_event.c:342: error: (Each undeclared identifier is reported only once
arch/arm/kernel/perf_event.c:342: error: for each function it appears in.)
arch/arm/kernel/perf_event.c: In function 'armpmu_event_init':
arch/arm/kernel/perf_event.c:536: warning: return makes integer from pointer without a cast
arch/arm/kernel/perf_event.c: In function 'init_hw_perf_events':
arch/arm/kernel/perf_event.c:3037: error: expected ';' before 'return'
arch/arm/kernel/perf_event.c:3038: warning: no return statement in function returning non-void
make[1]: *** [arch/arm/kernel/perf_event.o] Error 1
make: *** [arch/arm/kernel] Error 2

I think there's a missing semicolon, an ERR_PTR that should be a PTR_ERR
and an ordering issue with the pmu struct.

Cheers,

Will

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 25 Jun 2010 07:20
On Fri, 2010-06-25 at 12:11 +0100, Will Deacon wrote:
> Hi Peter,
>
> On Thu, 2010-06-24 at 15:28 +0100, Peter Zijlstra wrote:
> > These patches prepare the perf code for multiple pmus (no user
> > interface yet, Lin Ming is working on that). These patches remove all
> > weak functions and rework the struct pmu interface.
> >
> > The patches are boot tested on x86_64 and compile tested on: powerpc
> > (!fsl, what config is that?), sparc and arm (sorry no SH compiler)
> >
>
> I tried to test these for ARM, but I get compilation errors. Maybe you
> forgot to enable perf in the ARM .config? Anyway, the compiler log is:
>
> CC arch/arm/kernel/perf_event.o
> arch/arm/kernel/perf_event.c: In function 'validate_event':
> arch/arm/kernel/perf_event.c:342: error: 'pmu' undeclared (first use in this function)
> arch/arm/kernel/perf_event.c:342: error: (Each undeclared identifier is reported only once
> arch/arm/kernel/perf_event.c:342: error: for each function it appears in.)
> arch/arm/kernel/perf_event.c: In function 'armpmu_event_init':
> arch/arm/kernel/perf_event.c:536: warning: return makes integer from pointer without a cast
> arch/arm/kernel/perf_event.c: In function 'init_hw_perf_events':
> arch/arm/kernel/perf_event.c:3037: error: expected ';' before 'return'
> arch/arm/kernel/perf_event.c:3038: warning: no return statement in function returning non-void
> make[1]: *** [arch/arm/kernel/perf_event.o] Error 1
> make: *** [arch/arm/kernel] Error 2
>
> I think there's a missing semicolon, an ERR_PTR that should be a PTR_ERR
> and an ordering issue with the pmu struct.

I seem to have lost a refresh before sending the emails, please check:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/linux-2.6-perf.git perf-pmu

I pushed out updated patches there.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Will Deacon on 25 Jun 2010 10:40
Hello,

On Fri, 2010-06-25 at 12:16 +0100, Peter Zijlstra wrote:
> I seem to have lost a refresh before sending the emails, please check:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/peterz/linux-2.6-perf.git perf-pmu
>
> I pushed out updated patches there.

Ok, I rebuilt my Kernel and perf tools from that tree and tested it on a
quad-core ARMv7 board. Per-task counters appear to work (software and
hardware) but pinned hardware counters always return 0:


root(a)will-lucid-nfs:~# perf stat -a -e cs -e cycles -e instructions -- ls
linux-2.6 tmp

Performance counter stats for 'ls':

33 context-switches
0 cycles
0 instructions # 0.000 IPC

0.028572009 seconds time elapsed


It's odd if only ARM is affected in this way. Do pinned events still work
for other people?

Will

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 25 Jun 2010 11:00
On Fri, 2010-06-25 at 15:36 +0100, Will Deacon wrote:
> Hello,
>
> On Fri, 2010-06-25 at 12:16 +0100, Peter Zijlstra wrote:
> > I seem to have lost a refresh before sending the emails, please check:
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/peterz/linux-2.6-perf.git perf-pmu
> >
> > I pushed out updated patches there.
>
> Ok, I rebuilt my Kernel and perf tools from that tree and tested it on a
> quad-core ARMv7 board.

Ooh, neat toy ;-)

> Per-task counters appear to work (software and
> hardware) but pinned hardware counters always return 0:
>
>
> root(a)will-lucid-nfs:~# perf stat -a -e cs -e cycles -e instructions -- ls
> linux-2.6 tmp
>
> Performance counter stats for 'ls':
>
> 33 context-switches
> 0 cycles
> 0 instructions # 0.000 IPC
>
> 0.028572009 seconds time elapsed
>
>
> It's odd if only ARM is affected in this way. Do pinned events still work
> for other people?

/me goes build that exact tree on his x86_64.. and gets:

# perf stat -a -e cs -e cycles -e instructions -- ls > /dev/null

Performance counter stats for 'ls':

51 context-switches
24963513 cycles
9225808 instructions # 0.370 IPC

0.002389051 seconds time elapsed


Not exactly sure how I could have messed up the ARM architecture code to
make this happen though... will have a peek.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Matt Fleming on 26 Jun 2010 07:30
On Thu, 24 Jun 2010 16:28:04 +0200, Peter Zijlstra <a.p.zijlstra(a)chello.nl> wrote:
> These patches prepare the perf code for multiple pmus (no user
> interface yet, Lin Ming is working on that). These patches remove all
> weak functions and rework the struct pmu interface.
>
> The patches are boot tested on x86_64 and compile tested on: powerpc
> (!fsl, what config is that?), sparc and arm (sorry no SH compiler)

Hi Peter,

I tried your SH changes and I needed to apply this patch to get it to
compile. I haven't run the code yet, but I'll do that later today.

---

diff --git a/arch/sh/kernel/perf_event.c b/arch/sh/kernel/perf_event.c
index 50ff852..6bfbaec 100644
--- a/arch/sh/kernel/perf_event.c
+++ b/arch/sh/kernel/perf_event.c
@@ -251,7 +251,7 @@ static void sh_pmu_del(struct perf_event *event, int flags)
perf_event_update_userpage(event);
}

-static int sh_pmu_enable(struct perf_event *event, int flags)
+static int sh_pmu_add(struct perf_event *event, int flags)
{
struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
struct hw_perf_event *hwc = &event->hw;
@@ -283,7 +283,7 @@ static void sh_pmu_read(struct perf_event *event)
sh_perf_event_update(event, &event->hw, event->hw.idx);
}

-static in sh_pmu_event_init(struct perf_event *event)
+static int sh_pmu_event_init(struct perf_event *event)
{
int err = __hw_perf_event_init(event);
if (unlikely(err)) {
@@ -345,15 +345,15 @@ sh_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
return NOTIFY_OK;
}

-int __cpuinit register_sh_pmu(struct sh_pmu *pmu)
+int __cpuinit register_sh_pmu(struct sh_pmu *_pmu)
{
if (sh_pmu)
return -EBUSY;
- sh_pmu = pmu;
+ sh_pmu = _pmu;

- pr_info("Performance Events: %s support registered\n", pmu->name);
+ pr_info("Performance Events: %s support registered\n", _pmu->name);

- WARN_ON(pmu->num_events > MAX_HWEVENTS);
+ WARN_ON(_pmu->num_events > MAX_HWEVENTS);

perf_pmu_register(&pmu);
perf_cpu_notifier(sh_pmu_notifier);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 |  Next  |  Last
Pages: 1 2 3 4 5
Prev: [PATCH v2] Documentation/sysctl/vm.txt typo
Next: perf tools: allow cross compiling with DWARF support