2009-06-24 02:30:16

by Yong Wang

[permalink] [raw]
Subject: [PATCH -tip] perf_counter/x86: Set global control MSR correctly

Previous code made an assumption that the power on value of global
control MSR has enabled all fixed and general purpose counters properly.
However, this is not the case for certain Intel processors, such as
Atom. Each enable bit in IA32_PERF_GLOBAL_CTRL is AND'ed with the
enable bits for all privilege levels in the respective IA32_PERFEVTSELx
or IA32_PERF_FIXED_CTR_CTRL MSRs to start/stop the counting of
respective counters. Counting is enabled if the AND'ed results is true;
counting is disabled when the result is false. The end result is that
all fixed counters are always disabled on Atom processors because the
assumption is just invalid.

Reported-by: stephane eranian <[email protected]>
Signed-off-by: Yong Wang <[email protected]>

---
arch/x86/kernel/cpu/perf_counter.c | 10 +---------
1 file changed, 1 insertion(+), 9 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index 22eb3a1..a310d19 100644
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -969,13 +969,6 @@ fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc)
if (!x86_pmu.num_counters_fixed)
return -1;

- /*
- * Quirk, IA32_FIXED_CTRs do not work on current Atom processors:
- */
- if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
- boot_cpu_data.x86_model == 28)
- return -1;
-
event = hwc->config & ARCH_PERFMON_EVENT_MASK;

if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS)))
@@ -1428,8 +1421,6 @@ static int intel_pmu_init(void)
*/
x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3);

- rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
-
/*
* Install the hw-cache-events table:
*/
@@ -1514,6 +1505,7 @@ void __init init_hw_perf_counters(void)

perf_counter_mask |=
((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED;
+ x86_pmu.intel_ctrl = perf_counter_mask;

perf_counters_lapic_init();
register_die_notifier(&perf_counter_nmi_notifier);


2009-06-24 09:01:35

by Yong Wang

[permalink] [raw]
Subject: [tip:perfcounters/urgent] perf_counter, x86: Set global control MSR correctly

Commit-ID: c14dab5c0782ef632742963a66276a195418a63c
Gitweb: http://git.kernel.org/tip/c14dab5c0782ef632742963a66276a195418a63c
Author: Yong Wang <[email protected]>
AuthorDate: Wed, 24 Jun 2009 10:13:24 +0800
Committer: Ingo Molnar <[email protected]>
CommitDate: Wed, 24 Jun 2009 10:51:24 +0200

perf_counter, x86: Set global control MSR correctly

Previous code made an assumption that the power on value of global
control MSR has enabled all fixed and general purpose counters properly.

However, this is not the case for certain Intel processors, such as
Atom - and it might also be firmware dependent.

Each enable bit in IA32_PERF_GLOBAL_CTRL is AND'ed with the
enable bits for all privilege levels in the respective IA32_PERFEVTSELx
or IA32_PERF_FIXED_CTR_CTRL MSRs to start/stop the counting of
respective counters. Counting is enabled if the AND'ed results is true;
counting is disabled when the result is false.

The end result is that all fixed counters are always disabled on Atom
processors because the assumption is just invalid.

Fix this by not initializing the ctrl-mask out of the global MSR,
but setting it to perf_counter_mask.

Reported-by: Stephane Eranian <[email protected]>
Signed-off-by: Yong Wang <[email protected]>
Cc: Arjan van de Ven <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Paul Mackerras <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>


---
arch/x86/kernel/cpu/perf_counter.c | 10 +---------
1 files changed, 1 insertions(+), 9 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index 22eb3a1..a310d19 100644
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -969,13 +969,6 @@ fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc)
if (!x86_pmu.num_counters_fixed)
return -1;

- /*
- * Quirk, IA32_FIXED_CTRs do not work on current Atom processors:
- */
- if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
- boot_cpu_data.x86_model == 28)
- return -1;
-
event = hwc->config & ARCH_PERFMON_EVENT_MASK;

if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS)))
@@ -1428,8 +1421,6 @@ static int intel_pmu_init(void)
*/
x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3);

- rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
-
/*
* Install the hw-cache-events table:
*/
@@ -1514,6 +1505,7 @@ void __init init_hw_perf_counters(void)

perf_counter_mask |=
((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED;
+ x86_pmu.intel_ctrl = perf_counter_mask;

perf_counters_lapic_init();
register_die_notifier(&perf_counter_nmi_notifier);