Skip to contentSkip to conversation

perf, x86: Revamp PEBS event selection

PATCH In review Andi Kleen 4 participants 15 messages Jul 12, 2014 – Jul 21, 2014

Subsystems (2)
lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

PATCH 1/2 perf, x86: Revamp PEBS event selection

From: Andi Kleen <[email protected]>

The basic idea is that it does not make sense to list all PEBS events individually. The list is very long, sometimes outdated and the hardware doesn't need it. If an event does not support PEBS it will just not count, there is no security issue.

This vastly simplifies the PEBS event selection. It also speeds up the scheduling because the scheduler doesn't have to walk as many constraints.

Bugs fixed:
- We do not allow setting forbidden flags with PEBS anymore (SDM 18.9.4), except for the special cycle event.
This is done using a new constraint macro that also
matches on the event flags.
- We now allow DataLA on all Haswell events, not just
a small subset. In general all PEBS events that tag memory accesses support DataLA on Haswell. Otherwise the reported address is just zero. This allows address profiling
on vastly more events.
- We did not allow all PEBS events on Haswell:
We were missing some valid subevents in d1-d2 (MEM_LOAD_UOPS_RETIRED.*, MEM_LOAD_UOPS_RETIRED_L3_HIT_RETIRED.*)

This includes the changes proposed by Stephane earlier and obsoletes his patchkit (except for some changes on pre Sandy Bridge/Silvermont CPUs)

I only did Sandy Bridge and Silvermont and later so far, mostly because these are the parts I could directly confirm the hardware behavior with hardware architects. Also I do not believe the older CPUs have any
missing events in their PEBS list, so there's no pressing
need to change them.

I did not implement the flag proposed by Peter to allow setting forbidden flags. If really needed this could be implemented on to of this patch.

Cc: [email protected]
v2: Fix broken store events on SNB/IVB (Stephane Eranian)
v3: More fixes. Rename some arguments (Stephane Eranian) Update description.

---
 arch/x86/include/asm/perf_event.h         |  8 +++
 arch/x86/kernel/cpu/perf_event.h          | 18 +++++--
 arch/x86/kernel/cpu/perf_event_intel_ds.c | 88 +++++++------------------------
 3 files changed, 39 insertions(+), 75 deletions(-)
arch/x86/include/asm/perf_event.h +8 -0
index 8249df4..8dfc9fd 100644
@@ -51,6 +51,14 @@
 	 ARCH_PERFMON_EVENTSEL_EDGE  |	\
 	 ARCH_PERFMON_EVENTSEL_INV   |	\
 	 ARCH_PERFMON_EVENTSEL_CMASK)
+#define X86_ALL_EVENT_FLAGS  			\
+	(ARCH_PERFMON_EVENTSEL_EDGE |  		\
+	 ARCH_PERFMON_EVENTSEL_INV | 		\
+	 ARCH_PERFMON_EVENTSEL_CMASK | 		\
+	 ARCH_PERFMON_EVENTSEL_ANY | 		\
+	 ARCH_PERFMON_EVENTSEL_PIN_CONTROL | 	\
+	 HSW_IN_TX | 				\
+	 HSW_IN_TX_CHECKPOINTED)
 #define AMD64_RAW_EVENT_MASK		\
 	(X86_RAW_EVENT_MASK          |  \
 	 AMD64_EVENTSEL_EVENT)
arch/x86/kernel/cpu/perf_event.h +13 -5
index a22a34e9..8f32af0 100644
@@ -262,16 +262,24 @@ struct cpu_hw_events {
 	EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK)
 
 #define INTEL_PLD_CONSTRAINT(c, n)	\
-	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
+	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
 			   HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_LDLAT)
 
 #define INTEL_PST_CONSTRAINT(c, n)	\
-	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
+	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
 			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST)
 
-/* DataLA version of store sampling without extra enable bit. */
-#define INTEL_PST_HSW_CONSTRAINT(c, n)	\
-	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
+/* Event constraint, but match on all event flags too. */
+#define INTEL_FLAGS_EVENT_CONSTRAINT(c, n) \
+	EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS)
+
+/* Check only flags, but allow all event/umask */
+#define INTEL_ALL_EVENT_CONSTRAINT(code, n)	\
+	EVENT_CONSTRAINT(code, n, X86_ALL_EVENT_FLAGS)
+
+/* Same as above, but enable DataLA */
+#define INTEL_ALL_EVENT_CONSTRAINT_DATALA(code, n) \
+	__EVENT_CONSTRAINT(code, n, X86_ALL_EVENT_FLAGS, \
 			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST_HSW)
 
 /*
arch/x86/kernel/cpu/perf_event_intel_ds.c +18 -70

114 lines not on this page · raw message

Signature · 1 line
1.9.3
author
#1 lore raw

PATCH 2/2 perf, x86: Don't mark DataLA addresses as store

From: Andi Kleen <[email protected]>

Haswell supports reporting the data address for a range of PEBS events, including:

UOPS_RETIRED.ALL
MEM_UOPS_RETIRED.STLB_MISS_LOADS
MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.LOCK_LOADS
MEM_UOPS_RETIRED.SPLIT_LOADS
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_LOADS
MEM_UOPS_RETIRED.ALL_STORES
MEM_LOAD_UOPS_RETIRED.L1_HIT
MEM_LOAD_UOPS_RETIRED.L2_HIT
MEM_LOAD_UOPS_RETIRED.L3_HIT
MEM_LOAD_UOPS_RETIRED.L1_MISS
MEM_LOAD_UOPS_RETIRED.L2_MISS
MEM_LOAD_UOPS_RETIRED.L3_MISS
MEM_LOAD_UOPS_RETIRED.HIT_LFB
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_NONE
MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM

This facility was already enabled earlier with the original Haswell perf changes.

However these addresses were always reports as stores by perf, which is wrong, as they could be loads too. The hardware does not distinguish loads and stores for these instructions, so there's no (cheap) way for the profiler to find out.

Change the type to PERF_MEM_OP_NA instead.

---
 arch/x86/kernel/cpu/perf_event_intel_ds.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/kernel/cpu/perf_event_intel_ds.c +1 -1
index 64b4be9..13baa7c 100644
@@ -114,7 +114,7 @@ static u64 precise_store_data_hsw(struct perf_event *event, u64 status)
 	u64 cfg = event->hw.config & INTEL_ARCH_EVENT_MASK;
 
 	dse.val = 0;
-	dse.mem_op = PERF_MEM_OP_STORE;
+	dse.mem_op = PERF_MEM_OP_NA;
 	dse.mem_lvl = PERF_MEM_LVL_NA;
 
 	/*
Signature · 1 line
1.9.3
#2 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 29 lines · Andi Kleen

From: Andi Kleen <[email protected]>

Haswell supports reporting the data address for a range of PEBS events, including:

UOPS_RETIRED.ALL
MEM_UOPS_RETIRED.STLB_MISS_LOADS
MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.LOCK_LOADS
MEM_UOPS_RETIRED.SPLIT_LOADS
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_LOADS
MEM_UOPS_RETIRED.ALL_STORES
MEM_LOAD_UOPS_RETIRED.L1_HIT
MEM_LOAD_UOPS_RETIRED.L2_HIT
MEM_LOAD_UOPS_RETIRED.L3_HIT
MEM_LOAD_UOPS_RETIRED.L1_MISS
MEM_LOAD_UOPS_RETIRED.L2_MISS
MEM_LOAD_UOPS_RETIRED.L3_MISS
MEM_LOAD_UOPS_RETIRED.HIT_LFB
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_NONE
MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM

This facility was already enabled earlier with the original Haswell perf changes.

Quoted 7 lines

However these addresses were always reports as stores by perf, which is wrong, as they could be loads too. The hardware does not distinguish loads and stores for these instructions, so there's no (cheap) way for the profiler to find out.

Change the type to PERF_MEM_OP_NA instead.

You could do better if you tagged the event during setup as load vs. store. And then you could simply propagate the flag to the data source struct.

Quoted 21 lines
---
 arch/x86/kernel/cpu/perf_event_intel_ds.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/kernel/cpu/perf_event_intel_ds.c +1 -2
index 64b4be9..13baa7c 100644
@@ -114,7 +114,7 @@ static u64 precise_store_data_hsw(struct perf_event *event, u64 status)
        u64 cfg = event->hw.config & INTEL_ARCH_EVENT_MASK;
 
        dse.val = 0;
-       dse.mem_op = PERF_MEM_OP_STORE;
+       dse.mem_op = PERF_MEM_OP_NA;
        dse.mem_lvl = PERF_MEM_LVL_NA;
 
        /*
--

1.9.3

#3 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 2 lines

You could do better if you tagged the event during setup as load vs. store. And then you could simply propagate the flag to the data source struct.

This would require listing all PEBS events in the table again. The whole point of the other patch was to get rid of that.

Besides it wouldn't work for a range of events (like UOPS_RETIRED.ALL)

-Andi

#4 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Andi,

Quoted 8 lines · Andi Kleen
Quoted 2 lines

You could do better if you tagged the event during setup as load vs. store. And then you could simply propagate the flag to the data source struct.

This would require listing all PEBS events in the table again. The whole point of the other patch was to get rid of that.

Besides it wouldn't work for a range of events (like UOPS_RETIRED.ALL)

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Was working before for the mem-loads, mem-stores events.

#5 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 9 lines

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Was working before for the mem-loads, mem-stores events.

Ok. Would it be enough if it only worked for "mem-stores" and not all PEBS events?

Otherwise have to go back to an even larger PEBS constraint table for HSW than before.

-Andi

Signature · 1 line
[email protected] -- Speaking for myself only
#6 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 5 lines · Andi Kleen
Quoted 5 lines

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

Quoted 8 lines
Quoted 4 lines

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Was working before for the mem-loads, mem-stores events.

Ok. Would it be enough if it only worked for "mem-stores" and not all PEBS events?

Ok, do that at a minimum.

Quoted 7 lines

Otherwise have to go back to an even larger PEBS constraint table for HSW than before.

-Andi

--
[email protected] -- Speaking for myself only

#7 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 9 lines · Stephane Eranian
Quoted 8 lines · Andi Kleen
Quoted 8 lines

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Was working before for the mem-loads, mem-stores events.

Ok. Would it be enough if it only worked for "mem-stores" and not all PEBS events?

Ok, do that at a minimum.

But if I recall the PEBS stores events were not that many to begin with.

Quoted 7 lines
Quoted 7 lines

Otherwise have to go back to an even larger PEBS constraint table for HSW than before.

-Andi

--
[email protected] -- Speaking for myself only

#8 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 10 lines · Stephane Eranian
Quoted 9 lines · Stephane Eranian
Quoted 8 lines · Andi Kleen
Quoted 8 lines

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Quoted 8 lines
Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Was working before for the mem-loads, mem-stores events.

Ok. Would it be enough if it only worked for "mem-stores" and not all PEBS events?

Ok, do that at a minimum.

But if I recall the PEBS stores events were not that many to begin with.

Yes, there are only three store events:

MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_STORES

These can be added.

But most others are loads, so if you wanted loads too (besides mem-loads) it would be nearly a full list.

-Andi

maintainer
#9 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 10 lines · Andi Kleen

Yes, there are only three store events:

MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_STORES

These can be added.

But most others are loads, so if you wanted loads too (besides mem-loads) it would be nearly a full list.

Of that list you had earlier:

MEM_UOPS_RETIRED.STLB_MISS_LOADS
MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.LOCK_LOADS
MEM_UOPS_RETIRED.SPLIT_LOADS
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_LOADS
MEM_UOPS_RETIRED.ALL_STORES

There's only 4 loads and (as you already said) 3 stores.

That's 7 events total, that's not nearly a full list.

The other events:

UOPS_RETIRED.ALL
MEM_LOAD_UOPS_RETIRED.L1_HIT
MEM_LOAD_UOPS_RETIRED.L2_HIT
MEM_LOAD_UOPS_RETIRED.L3_HIT
MEM_LOAD_UOPS_RETIRED.L1_MISS
MEM_LOAD_UOPS_RETIRED.L2_MISS
MEM_LOAD_UOPS_RETIRED.L3_MISS
MEM_LOAD_UOPS_RETIRED.HIT_LFB
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_NONE
MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM

are unclear on their type and should indeed be NA.

#10 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 29 lines · Peter Zijlstra
Quoted 10 lines · Andi Kleen

Yes, there are only three store events:

MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_STORES

These can be added.

But most others are loads, so if you wanted loads too (besides mem-loads) it would be nearly a full list.

Of that list you had earlier:

MEM_UOPS_RETIRED.STLB_MISS_LOADS
MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.LOCK_LOADS
MEM_UOPS_RETIRED.SPLIT_LOADS
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_LOADS
MEM_UOPS_RETIRED.ALL_STORES

There's only 4 loads and (as you already said) 3 stores.

That's 7 events total, that's not nearly a full list.

The other events:

Quoted 14 lines

UOPS_RETIRED.ALL
MEM_LOAD_UOPS_RETIRED.L1_HIT
MEM_LOAD_UOPS_RETIRED.L2_HIT
MEM_LOAD_UOPS_RETIRED.L3_HIT
MEM_LOAD_UOPS_RETIRED.L1_MISS
MEM_LOAD_UOPS_RETIRED.L2_MISS
MEM_LOAD_UOPS_RETIRED.L3_MISS
MEM_LOAD_UOPS_RETIRED.HIT_LFB
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_NONE
MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM

Those are loads uops.

Quoted 1 line

are unclear on their type and should indeed be NA.

#11 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 44 lines · Stephane Eranian
Quoted 29 lines · Peter Zijlstra
Quoted 10 lines · Andi Kleen

Yes, there are only three store events:

MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_STORES

These can be added.

But most others are loads, so if you wanted loads too (besides mem-loads) it would be nearly a full list.

Of that list you had earlier:

MEM_UOPS_RETIRED.STLB_MISS_LOADS
MEM_UOPS_RETIRED.STLB_MISS_STORES
MEM_UOPS_RETIRED.LOCK_LOADS
MEM_UOPS_RETIRED.SPLIT_LOADS
MEM_UOPS_RETIRED.SPLIT_STORES
MEM_UOPS_RETIRED.ALL_LOADS
MEM_UOPS_RETIRED.ALL_STORES

There's only 4 loads and (as you already said) 3 stores.

That's 7 events total, that's not nearly a full list.

The other events:

Quoted 14 lines

UOPS_RETIRED.ALL
MEM_LOAD_UOPS_RETIRED.L1_HIT
MEM_LOAD_UOPS_RETIRED.L2_HIT
MEM_LOAD_UOPS_RETIRED.L3_HIT
MEM_LOAD_UOPS_RETIRED.L1_MISS
MEM_LOAD_UOPS_RETIRED.L2_MISS
MEM_LOAD_UOPS_RETIRED.L3_MISS
MEM_LOAD_UOPS_RETIRED.HIT_LFB
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM
MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_NONE
MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM

Quoted 2 lines

Those are loads uops.

I suggest we add those back as loads. We cannot really loose precision in the info returned.

Quoted 1 line
Quoted 1 line

are unclear on their type and should indeed be NA.

author
#12 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

In reply to Stephane Eranian
Quoted 9 lines · Stephane Eranian
Quoted 8 lines · Andi Kleen
Quoted 8 lines

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Quoted 6 lines
Quoted 5 lines
Quoted 1 line

Was working before for the mem-loads, mem-stores events.

Ok. Would it be enough if it only worked for "mem-stores" and not all PEBS events?

Ok, do that at a minimum.

I fixed it now. However it turned out that "perf mem report" actually not report mem_op, only mem_lvl.

You may want to fix that separately.

-Andi

#13 lore raw

Re: [PATCH 2/2] perf, x86: Don't mark DataLA addresses as store

Quoted 15 lines · Andi Kleen
Quoted 14 lines · Stephane Eranian
Quoted 13 lines · Andi Kleen
Quoted 9 lines

I have a problem with this patch.

It makes: perf mem -t store rec record OP_NA for the store. It was recording OP_STORE before.

I think we need to keep LD/ST info. This is useful for analysis especially if we collect loads/stores simultaneously.

Was working before for the mem-loads, mem-stores events.

Ok. Would it be enough if it only worked for "mem-stores" and not all PEBS events?

Quoted 7 lines
Quoted 1 line

Ok, do that at a minimum.

I fixed it now. However it turned out that "perf mem report" actually not report mem_op, only mem_lvl.

You may want to fix that separately.

It is because it currently record loads or store but not both simultaneously. Once I allow load/store in a single run, it will print the mem_op.

Quoted 1 line

-Andi

Reviewed-by on v1 1/2
#14 lore raw
In reply to Andi Kleen
Quoted 39 lines · Andi Kleen

From: Andi Kleen <[email protected]>

The basic idea is that it does not make sense to list all PEBS events individually. The list is very long, sometimes outdated and the hardware doesn't need it. If an event does not support PEBS it will just not count, there is no security issue.

This vastly simplifies the PEBS event selection. It also speeds up the scheduling because the scheduler doesn't have to walk as many constraints.

Bugs fixed:
- We do not allow setting forbidden flags with PEBS anymore (SDM 18.9.4), except for the special cycle event.
This is done using a new constraint macro that also
matches on the event flags.
- We now allow DataLA on all Haswell events, not just
a small subset. In general all PEBS events that tag memory accesses support DataLA on Haswell. Otherwise the reported address is just zero. This allows address profiling
on vastly more events.
- We did not allow all PEBS events on Haswell:
We were missing some valid subevents in d1-d2 (MEM_LOAD_UOPS_RETIRED.*, MEM_LOAD_UOPS_RETIRED_L3_HIT_RETIRED.*)

This includes the changes proposed by Stephane earlier and obsoletes his patchkit (except for some changes on pre Sandy Bridge/Silvermont CPUs)

I only did Sandy Bridge and Silvermont and later so far, mostly because these are the parts I could directly confirm the hardware behavior with hardware architects. Also I do not believe the older CPUs have any
missing events in their PEBS list, so there's no pressing
need to change them.

I did not implement the flag proposed by Peter to allow setting forbidden flags. If really needed this could be implemented on to of this patch.

Quoted 5 lines

Cc: [email protected]
v2: Fix broken store events on SNB/IVB (Stephane Eranian)
v3: More fixes. Rename some arguments (Stephane Eranian) Update description.

Works now for me on SNB/HSW.

Quoted 182 lines
---
 arch/x86/include/asm/perf_event.h         |  8 +++
 arch/x86/kernel/cpu/perf_event.h          | 18 +++++--
 arch/x86/kernel/cpu/perf_event_intel_ds.c | 88 +++++++------------------------
 3 files changed, 39 insertions(+), 75 deletions(-)
arch/x86/include/asm/perf_event.h +8 -0
index 8249df4..8dfc9fd 100644
@@ -51,6 +51,14 @@
         ARCH_PERFMON_EVENTSEL_EDGE  |  \
         ARCH_PERFMON_EVENTSEL_INV   |  \
         ARCH_PERFMON_EVENTSEL_CMASK)
+#define X86_ALL_EVENT_FLAGS                    \
+       (ARCH_PERFMON_EVENTSEL_EDGE |           \
+        ARCH_PERFMON_EVENTSEL_INV |            \
+        ARCH_PERFMON_EVENTSEL_CMASK |          \
+        ARCH_PERFMON_EVENTSEL_ANY |            \
+        ARCH_PERFMON_EVENTSEL_PIN_CONTROL |    \
+        HSW_IN_TX |                            \
+        HSW_IN_TX_CHECKPOINTED)
 #define AMD64_RAW_EVENT_MASK           \
        (X86_RAW_EVENT_MASK          |  \
         AMD64_EVENTSEL_EVENT)
arch/x86/kernel/cpu/perf_event.h +13 -5
index a22a34e9..8f32af0 100644
@@ -262,16 +262,24 @@ struct cpu_hw_events {
        EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK)
 
 #define INTEL_PLD_CONSTRAINT(c, n)     \
-       __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
+       __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
                           HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_LDLAT)
 
 #define INTEL_PST_CONSTRAINT(c, n)     \
-       __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
+       __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
                          HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST)
 
-/* DataLA version of store sampling without extra enable bit. */
-#define INTEL_PST_HSW_CONSTRAINT(c, n) \
-       __EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
+/* Event constraint, but match on all event flags too. */
+#define INTEL_FLAGS_EVENT_CONSTRAINT(c, n) \
+       EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS)
+
+/* Check only flags, but allow all event/umask */
+#define INTEL_ALL_EVENT_CONSTRAINT(code, n)    \
+       EVENT_CONSTRAINT(code, n, X86_ALL_EVENT_FLAGS)
+
+/* Same as above, but enable DataLA */
+#define INTEL_ALL_EVENT_CONSTRAINT_DATALA(code, n) \
+       __EVENT_CONSTRAINT(code, n, X86_ALL_EVENT_FLAGS, \
                          HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST_HSW)
 
 /*
arch/x86/kernel/cpu/perf_event_intel_ds.c +18 -71
index 980970c..64b4be9 100644
@@ -567,28 +567,10 @@ struct event_constraint intel_atom_pebs_event_constraints[] = {
 };
 
 struct event_constraint intel_slm_pebs_event_constraints[] = {
-       INTEL_UEVENT_CONSTRAINT(0x0103, 0x1), /* REHABQ.LD_BLOCK_ST_FORWARD_PS */
-       INTEL_UEVENT_CONSTRAINT(0x0803, 0x1), /* REHABQ.LD_SPLITS_PS */
-       INTEL_UEVENT_CONSTRAINT(0x0204, 0x1), /* MEM_UOPS_RETIRED.L2_HIT_LOADS_PS */
-       INTEL_UEVENT_CONSTRAINT(0x0404, 0x1), /* MEM_UOPS_RETIRED.L2_MISS_LOADS_PS */
-       INTEL_UEVENT_CONSTRAINT(0x0804, 0x1), /* MEM_UOPS_RETIRED.DTLB_MISS_LOADS_PS */
-       INTEL_UEVENT_CONSTRAINT(0x2004, 0x1), /* MEM_UOPS_RETIRED.HITM_PS */
-       INTEL_UEVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY_PS */
-       INTEL_UEVENT_CONSTRAINT(0x00c4, 0x1), /* BR_INST_RETIRED.ALL_BRANCHES_PS */
-       INTEL_UEVENT_CONSTRAINT(0x7ec4, 0x1), /* BR_INST_RETIRED.JCC_PS */
-       INTEL_UEVENT_CONSTRAINT(0xbfc4, 0x1), /* BR_INST_RETIRED.FAR_BRANCH_PS */
-       INTEL_UEVENT_CONSTRAINT(0xebc4, 0x1), /* BR_INST_RETIRED.NON_RETURN_IND_PS */
-       INTEL_UEVENT_CONSTRAINT(0xf7c4, 0x1), /* BR_INST_RETIRED.RETURN_PS */
-       INTEL_UEVENT_CONSTRAINT(0xf9c4, 0x1), /* BR_INST_RETIRED.CALL_PS */
-       INTEL_UEVENT_CONSTRAINT(0xfbc4, 0x1), /* BR_INST_RETIRED.IND_CALL_PS */
-       INTEL_UEVENT_CONSTRAINT(0xfdc4, 0x1), /* BR_INST_RETIRED.REL_CALL_PS */
-       INTEL_UEVENT_CONSTRAINT(0xfec4, 0x1), /* BR_INST_RETIRED.TAKEN_JCC_PS */
-       INTEL_UEVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_MISP_RETIRED.ALL_BRANCHES_PS */
-       INTEL_UEVENT_CONSTRAINT(0x7ec5, 0x1), /* BR_INST_MISP_RETIRED.JCC_PS */
-       INTEL_UEVENT_CONSTRAINT(0xebc5, 0x1), /* BR_INST_MISP_RETIRED.NON_RETURN_IND_PS */
-       INTEL_UEVENT_CONSTRAINT(0xf7c5, 0x1), /* BR_INST_MISP_RETIRED.RETURN_PS */
-       INTEL_UEVENT_CONSTRAINT(0xfbc5, 0x1), /* BR_INST_MISP_RETIRED.IND_CALL_PS */
-       INTEL_UEVENT_CONSTRAINT(0xfec5, 0x1), /* BR_INST_MISP_RETIRED.TAKEN_JCC_PS */
+       /* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
+       INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
+       /* Allow all events as PEBS with no flags */
+       INTEL_ALL_EVENT_CONSTRAINT(0, 0x1),
        EVENT_CONSTRAINT_END
 };
 
@@ -624,68 +606,34 @@ struct event_constraint intel_westmere_pebs_event_constraints[] = {
 
 struct event_constraint intel_snb_pebs_event_constraints[] = {
        INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
-       INTEL_UEVENT_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
-       INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */
-       INTEL_EVENT_CONSTRAINT(0xc4, 0xf),    /* BR_INST_RETIRED.* */
-       INTEL_EVENT_CONSTRAINT(0xc5, 0xf),    /* BR_MISP_RETIRED.* */
        INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */
        INTEL_PST_CONSTRAINT(0x02cd, 0x8),    /* MEM_TRANS_RETIRED.PRECISE_STORES */
-       INTEL_EVENT_CONSTRAINT(0xd0, 0xf),    /* MEM_UOP_RETIRED.* */
-       INTEL_EVENT_CONSTRAINT(0xd1, 0xf),    /* MEM_LOAD_UOPS_RETIRED.* */
-       INTEL_EVENT_CONSTRAINT(0xd2, 0xf),    /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */
-       INTEL_EVENT_CONSTRAINT(0xd3, 0xf),    /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */
-       INTEL_UEVENT_CONSTRAINT(0x02d4, 0xf), /* MEM_LOAD_UOPS_MISC_RETIRED.LLC_MISS */
+       /* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
+       INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
+       /* Allow all events as PEBS with no flags */
+       INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
        EVENT_CONSTRAINT_END
 };
 
 struct event_constraint intel_ivb_pebs_event_constraints[] = {
         INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
-        INTEL_UEVENT_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
-        INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */
-        INTEL_EVENT_CONSTRAINT(0xc4, 0xf),    /* BR_INST_RETIRED.* */
-        INTEL_EVENT_CONSTRAINT(0xc5, 0xf),    /* BR_MISP_RETIRED.* */
         INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.LAT_ABOVE_THR */
        INTEL_PST_CONSTRAINT(0x02cd, 0x8),    /* MEM_TRANS_RETIRED.PRECISE_STORES */
-        INTEL_EVENT_CONSTRAINT(0xd0, 0xf),    /* MEM_UOP_RETIRED.* */
-        INTEL_EVENT_CONSTRAINT(0xd1, 0xf),    /* MEM_LOAD_UOPS_RETIRED.* */
-        INTEL_EVENT_CONSTRAINT(0xd2, 0xf),    /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */
-        INTEL_EVENT_CONSTRAINT(0xd3, 0xf),    /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.* */
+       /* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
+       INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
+       /* Allow all events as PEBS with no flags */
+       INTEL_ALL_EVENT_CONSTRAINT(0, 0xf),
         EVENT_CONSTRAINT_END
 };
 
 struct event_constraint intel_hsw_pebs_event_constraints[] = {
        INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
-       INTEL_PST_HSW_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
-       INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */
-       INTEL_EVENT_CONSTRAINT(0xc4, 0xf),    /* BR_INST_RETIRED.* */
-       INTEL_UEVENT_CONSTRAINT(0x01c5, 0xf), /* BR_MISP_RETIRED.CONDITIONAL */
-       INTEL_UEVENT_CONSTRAINT(0x04c5, 0xf), /* BR_MISP_RETIRED.ALL_BRANCHES */
-       INTEL_UEVENT_CONSTRAINT(0x20c5, 0xf), /* BR_MISP_RETIRED.NEAR_TAKEN */
-       INTEL_PLD_CONSTRAINT(0x01cd, 0x8),    /* MEM_TRANS_RETIRED.* */
-       /* MEM_UOPS_RETIRED.STLB_MISS_LOADS */
-       INTEL_UEVENT_CONSTRAINT(0x11d0, 0xf),
-       /* MEM_UOPS_RETIRED.STLB_MISS_STORES */
-       INTEL_UEVENT_CONSTRAINT(0x12d0, 0xf),
-       INTEL_UEVENT_CONSTRAINT(0x21d0, 0xf), /* MEM_UOPS_RETIRED.LOCK_LOADS */
-       INTEL_UEVENT_CONSTRAINT(0x41d0, 0xf), /* MEM_UOPS_RETIRED.SPLIT_LOADS */
-       /* MEM_UOPS_RETIRED.SPLIT_STORES */
-       INTEL_UEVENT_CONSTRAINT(0x42d0, 0xf),
-       INTEL_UEVENT_CONSTRAINT(0x81d0, 0xf), /* MEM_UOPS_RETIRED.ALL_LOADS */
-       INTEL_PST_HSW_CONSTRAINT(0x82d0, 0xf), /* MEM_UOPS_RETIRED.ALL_STORES */
-       INTEL_UEVENT_CONSTRAINT(0x01d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L1_HIT */
-       INTEL_UEVENT_CONSTRAINT(0x02d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L2_HIT */
-       INTEL_UEVENT_CONSTRAINT(0x04d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L3_HIT */
-       /* MEM_LOAD_UOPS_RETIRED.HIT_LFB */
-       INTEL_UEVENT_CONSTRAINT(0x40d1, 0xf),
-       /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS */
-       INTEL_UEVENT_CONSTRAINT(0x01d2, 0xf),
-       /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT */
-       INTEL_UEVENT_CONSTRAINT(0x02d2, 0xf),
-       /* MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM */
-       INTEL_UEVENT_CONSTRAINT(0x01d3, 0xf),
-       INTEL_UEVENT_CONSTRAINT(0x04c8, 0xf), /* HLE_RETIRED.Abort */
-       INTEL_UEVENT_CONSTRAINT(0x04c9, 0xf), /* RTM_RETIRED.Abort */
-
+       INTEL_PLD_CONSTRAINT(0x01cd, 0xf),    /* MEM_TRANS_RETIRED.* */
+       /* UOPS_RETIRED.ALL, inv=1, cmask=16 (cycles:p). */
+       INTEL_FLAGS_EVENT_CONSTRAINT(0x108001c2, 0xf),
+       /* Allow all events as PEBS with no flags */
+       /* We allow DATALA for all PEBS events, will be 0 if not supported */
+       INTEL_ALL_EVENT_CONSTRAINT_DATALA(0, 0xf),
        EVENT_CONSTRAINT_END
 };
 
--

1.9.3