* [PATCH] perf/x86/intel: Apply early ACK for small core
@ 2021-07-08 19:54 kan.liang
2021-07-09 9:49 ` Peter Zijlstra
0 siblings, 1 reply; 3+ messages in thread
From: kan.liang @ 2021-07-08 19:54 UTC (permalink / raw)
To: peterz, mingo, linux-kernel; +Cc: ak, Kan Liang, stable
From: Kan Liang <kan.liang@linux.intel.com>
A warning as below may be occasionally triggered in an ADL machine when
these conditions occur,
- Two perf record commands run one by one. Both record a PEBS event.
- Both runs on small cores.
- They have different adaptive PEBS configuration (PEBS_DATA_CFG).
[ 673.663291] WARNING: CPU: 4 PID: 9874 at
arch/x86/events/intel/ds.c:1743
setup_pebs_adaptive_sample_data+0x55e/0x5b0
[ 673.663348] RIP: 0010:setup_pebs_adaptive_sample_data+0x55e/0x5b0
[ 673.663357] Call Trace:
[ 673.663357] <NMI>
[ 673.663357] intel_pmu_drain_pebs_icl+0x48b/0x810
[ 673.663360] perf_event_nmi_handler+0x41/0x80
[ 673.663368] </NMI>
[ 673.663370] __perf_event_task_sched_in+0x2c2/0x3a0
Different from the big core, the small core requires the ACK before
re-enabling counters in the NMI handler, otherwise a stale PEBS record
may be dumped into the later NMI handler, which trigger the warning.
Add late_ack in the struct x86_hybrid_pmu to track the late_ack flag for
different types of PMUs. Apply late ACK only for the big cores on an
Alder Lake machine.
The existing hybrid() macro has a compile error when taking address of a
bit-field variable. Add a new macro hybrid_bit() to get the bit-field
value of a given PMU.
Fixes: f83d2f91d259 ("perf/x86/intel: Add Alder Lake Hybrid support")
Reported-by: Ammy Yi <ammy.yi@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Cc: stable@vger.kernel.org
---
arch/x86/events/intel/core.c | 11 +++++------
arch/x86/events/perf_event.h | 12 ++++++++++++
2 files changed, 17 insertions(+), 6 deletions(-)
diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c
index a0dfa82..430a24d 100644
--- a/arch/x86/events/intel/core.c
+++ b/arch/x86/events/intel/core.c
@@ -2904,14 +2904,13 @@ static int handle_pmi_common(struct pt_regs *regs, u64 status)
*/
static int intel_pmu_handle_irq(struct pt_regs *regs)
{
- struct cpu_hw_events *cpuc;
+ struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+ bool late_ack = hybrid_bit(cpuc->pmu, late_ack);
int loops;
u64 status;
int handled;
int pmu_enabled;
- cpuc = this_cpu_ptr(&cpu_hw_events);
-
/*
* Save the PMU state.
* It needs to be restored when leaving the handler.
@@ -2921,7 +2920,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
* No known reason to not always do late ACK,
* but just in case do it opt-in.
*/
- if (!x86_pmu.late_ack)
+ if (!late_ack)
apic_write(APIC_LVTPC, APIC_DM_NMI);
intel_bts_disable_local();
cpuc->enabled = 0;
@@ -2969,7 +2968,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
* have been reset. This avoids spurious NMIs on
* Haswell CPUs.
*/
- if (x86_pmu.late_ack)
+ if (late_ack)
apic_write(APIC_LVTPC, APIC_DM_NMI);
return handled;
}
@@ -6116,7 +6115,6 @@ __init int intel_pmu_init(void)
static_branch_enable(&perf_is_hybrid);
x86_pmu.num_hybrid_pmus = X86_HYBRID_NUM_PMUS;
- x86_pmu.late_ack = true;
x86_pmu.pebs_aliases = NULL;
x86_pmu.pebs_prec_dist = true;
x86_pmu.pebs_block = true;
@@ -6154,6 +6152,7 @@ __init int intel_pmu_init(void)
pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX];
pmu->name = "cpu_core";
pmu->cpu_type = hybrid_big;
+ pmu->late_ack = true;
if (cpu_feature_enabled(X86_FEATURE_HYBRID_CPU)) {
pmu->num_counters = x86_pmu.num_counters + 2;
pmu->num_counters_fixed = x86_pmu.num_counters_fixed + 1;
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index bc8836b..40fa3b1 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -656,6 +656,8 @@ struct x86_hybrid_pmu {
struct event_constraint *event_constraints;
struct event_constraint *pebs_constraints;
struct extra_reg *extra_regs;
+
+ unsigned int late_ack:1;
};
static __always_inline struct x86_hybrid_pmu *hybrid_pmu(struct pmu *pmu)
@@ -686,6 +688,16 @@ extern struct static_key_false perf_is_hybrid;
__Fp; \
}))
+#define hybrid_bit(_pmu, _field) \
+({ \
+ bool __Fp = x86_pmu._field; \
+ \
+ if (is_hybrid() && (_pmu)) \
+ __Fp = hybrid_pmu(_pmu)->_field; \
+ \
+ __Fp; \
+})
+
enum hybrid_pmu_type {
hybrid_big = 0x40,
hybrid_small = 0x20,
--
2.7.4
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] perf/x86/intel: Apply early ACK for small core
2021-07-08 19:54 [PATCH] perf/x86/intel: Apply early ACK for small core kan.liang
@ 2021-07-09 9:49 ` Peter Zijlstra
2021-07-09 12:31 ` Liang, Kan
0 siblings, 1 reply; 3+ messages in thread
From: Peter Zijlstra @ 2021-07-09 9:49 UTC (permalink / raw)
To: kan.liang; +Cc: mingo, linux-kernel, ak, stable
On Thu, Jul 08, 2021 at 12:54:33PM -0700, kan.liang@linux.intel.com wrote:
> @@ -2921,7 +2920,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
> * No known reason to not always do late ACK,
> * but just in case do it opt-in.
> */
^^^ comment is now seriously out of date. Can you please update it?
> - if (!x86_pmu.late_ack)
> + if (!late_ack)
> apic_write(APIC_LVTPC, APIC_DM_NMI);
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] perf/x86/intel: Apply early ACK for small core
2021-07-09 9:49 ` Peter Zijlstra
@ 2021-07-09 12:31 ` Liang, Kan
0 siblings, 0 replies; 3+ messages in thread
From: Liang, Kan @ 2021-07-09 12:31 UTC (permalink / raw)
To: Peter Zijlstra; +Cc: mingo, linux-kernel, ak, stable
On 7/9/2021 5:49 AM, Peter Zijlstra wrote:
> On Thu, Jul 08, 2021 at 12:54:33PM -0700, kan.liang@linux.intel.com wrote:
>> @@ -2921,7 +2920,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
>> * No known reason to not always do late ACK,
>> * but just in case do it opt-in.
>> */
>
> ^^^ comment is now seriously out of date. Can you please update it?
>
Sure, I will update it and send a V2 patch.
Thanks,
Kan
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-07-09 12:31 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-08 19:54 [PATCH] perf/x86/intel: Apply early ACK for small core kan.liang
2021-07-09 9:49 ` Peter Zijlstra
2021-07-09 12:31 ` Liang, Kan
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).