perf, x86: Improve the PEBS ABI

Rename perf_event_attr::precise to perf_event_attr::precise_ip and
widen it to 2 bits. This new field describes the required precision of
the PERF_SAMPLE_IP field:

  0 - SAMPLE_IP can have arbitrary skid
  1 - SAMPLE_IP must have constant skid
  2 - SAMPLE_IP requested to have 0 skid
  3 - SAMPLE_IP must have 0 skid

And modify the Intel PEBS code accordingly. The PEBS implementation
now supports up to precise_ip == 2, where we perform the IP fixup.

Also s/PERF_RECORD_MISC_EXACT/&_IP/ to clarify its meaning, this bit
should be set for each PERF_SAMPLE_IP field known to match the actual
instruction triggering the event.

This new scheme allows for a PEBS mode that uses the buffer for more
than a single event.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

6 files changed, 60 insertions, 23 deletions

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 4a3f1f2b9b9..27fa9eeed02 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -488,6 +488,21 @@ static int x86_setup_perfctr(struct perf_event *event)
 
 static int x86_pmu_hw_config(struct perf_event *event)
 {
+	if (event->attr.precise_ip) {
+		int precise = 0;
+
+		/* Support for constant skid */
+		if (x86_pmu.pebs)
+			precise++;
+
+		/* Support for IP fixup */
+		if (x86_pmu.lbr_nr)
+			precise++;
+
+		if (event->attr.precise_ip > precise)
+			return -EOPNOTSUPP;
+	}
+
 	/*
 	 * Generate PMC IRQs:
 	 * (keep 'enabled' bit clear for now)
@@ -1780,7 +1795,7 @@ unsigned long perf_misc_flags(struct pt_regs *regs)
 	}
 
 	if (regs->flags & PERF_EFLAGS_EXACT)
-		misc |= PERF_RECORD_MISC_EXACT;
+		misc |= PERF_RECORD_MISC_EXACT_IP;
 
 	return misc;
 }
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index a4b56ac425c..fdbc652d3fe 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -563,7 +563,7 @@ static void intel_pmu_disable_event(struct perf_event *event)
 
 	x86_pmu_disable_event(event);
 
-	if (unlikely(event->attr.precise))
+	if (unlikely(event->attr.precise_ip))
 		intel_pmu_pebs_disable(event);
 }
 
@@ -615,7 +615,7 @@ static void intel_pmu_enable_event(struct perf_event *event)
 		return;
 	}
 
-	if (unlikely(event->attr.precise))
+	if (unlikely(event->attr.precise_ip))
 		intel_pmu_pebs_enable(event);
 
 	__x86_pmu_enable_event(hwc, ARCH_PERFMON_EVENTSEL_ENABLE);
diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c
index 35056f715e9..18018d1311c 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_ds.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c
@@ -307,7 +307,7 @@ intel_pebs_constraints(struct perf_event *event)
 {
 	struct event_constraint *c;
 
-	if (!event->attr.precise)
+	if (!event->attr.precise_ip)
 		return NULL;
 
 	if (x86_pmu.pebs_constraints) {
@@ -330,7 +330,7 @@ static void intel_pmu_pebs_enable(struct perf_event *event)
 	cpuc->pebs_enabled |= 1ULL << hwc->idx;
 	WARN_ON_ONCE(cpuc->enabled);
 
-	if (x86_pmu.intel_cap.pebs_trap)
+	if (x86_pmu.intel_cap.pebs_trap && event->attr.precise_ip > 1)
 		intel_pmu_lbr_enable(event);
 }
 
@@ -345,7 +345,7 @@ static void intel_pmu_pebs_disable(struct perf_event *event)
 
 	hwc->config |= ARCH_PERFMON_EVENTSEL_INT;
 
-	if (x86_pmu.intel_cap.pebs_trap)
+	if (x86_pmu.intel_cap.pebs_trap && event->attr.precise_ip > 1)
 		intel_pmu_lbr_disable(event);
 }
 
@@ -485,7 +485,7 @@ static void __intel_pmu_pebs_event(struct perf_event *event,
 	regs.bp = pebs->bp;
 	regs.sp = pebs->sp;
 
-	if (intel_pmu_pebs_fixup_ip(regs))
+	if (event->attr.precise_ip > 1 && intel_pmu_pebs_fixup_ip(&regs))
 		regs.flags |= PERF_EFLAGS_EXACT;
 	else
 		regs.flags &= ~PERF_EFLAGS_EXACT;
@@ -518,7 +518,7 @@ static void intel_pmu_drain_pebs_core(struct pt_regs *iregs)
 
 	WARN_ON_ONCE(!event);
 
-	if (!event->attr.precise)
+	if (!event->attr.precise_ip)
 		return;
 
 	n = top - at;
@@ -570,7 +570,7 @@ static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs)
 
 			WARN_ON_ONCE(!event);
 
-			if (!event->attr.precise)
+			if (!event->attr.precise_ip)
 				continue;
 
 			if (__test_and_set_bit(bit, (unsigned long *)&status))
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 6be4a0f9137..23cd0057a68 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -203,9 +203,19 @@ struct perf_event_attr {
 				enable_on_exec :  1, /* next exec enables     */
 				task           :  1, /* trace fork/exit       */
 				watermark      :  1, /* wakeup_watermark      */
-				precise        :  1, /* OoO invariant counter */
-
-				__reserved_1   : 48;
+				/*
+				 * precise_ip:
+				 *
+				 *  0 - SAMPLE_IP can have arbitrary skid
+				 *  1 - SAMPLE_IP must have constant skid
+				 *  2 - SAMPLE_IP requested to have 0 skid
+				 *  3 - SAMPLE_IP must have 0 skid
+				 *
+				 *  See also PERF_RECORD_MISC_EXACT_IP
+				 */
+				precise_ip     :  2, /* skid constraint       */
+
+				__reserved_1   : 47;
 
 	union {
 		__u32		wakeup_events;	  /* wakeup every n events */
@@ -296,7 +306,12 @@ struct perf_event_mmap_page {
 #define PERF_RECORD_MISC_GUEST_KERNEL		(4 << 0)
 #define PERF_RECORD_MISC_GUEST_USER		(5 << 0)
 
-#define PERF_RECORD_MISC_EXACT			(1 << 14)
+/*
+ * Indicates that the content of PERF_SAMPLE_IP points to
+ * the actual instruction that triggered the event. See also
+ * perf_event_attr::precise_ip.
+ */
+#define PERF_RECORD_MISC_EXACT_IP		(1 << 14)
 /*
  * Reserve the last bit to indicate some extended misc field
  */
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 3de397764cb..ed9b5b6905f 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -1021,7 +1021,7 @@ static void event__process_sample(const event_t *self,
 		return;
 	}
 
-	if (self->header.misc & PERF_RECORD_MISC_EXACT)
+	if (self->header.misc & PERF_RECORD_MISC_EXACT_IP)
 		exact_samples++;
 
 	if (event__preprocess_sample(self, session, &al, symbol_filter) < 0 ||
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index bc8b7e61420..ae7f5917935 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -654,10 +654,6 @@ parse_raw_event(const char **strp, struct perf_event_attr *attr)
 		return EVT_FAILED;
 	n = hex2u64(str + 1, &config);
 	if (n > 0) {
-		if (str[n+1] == 'p') {
-			attr->precise = 1;
-			n++;
-		}
 		*strp = str + n + 1;
 		attr->type = PERF_TYPE_RAW;
 		attr->config = config;
@@ -692,19 +688,29 @@ static enum event_result
 parse_event_modifier(const char **strp, struct perf_event_attr *attr)
 {
 	const char *str = *strp;
-	int eu = 1, ek = 1, eh = 1;
+	int exclude = 0;
+	int eu = 0, ek = 0, eh = 0, precise = 0;
 
 	if (*str++ != ':')
 		return 0;
 	while (*str) {
-		if (*str == 'u')
+		if (*str == 'u') {
+			if (!exclude)
+				exclude = eu = ek = eh = 1;
 			eu = 0;
-		else if (*str == 'k')
+		} else if (*str == 'k') {
+			if (!exclude)
+				exclude = eu = ek = eh = 1;
 			ek = 0;
-		else if (*str == 'h')
+		} else if (*str == 'h') {
+			if (!exclude)
+				exclude = eu = ek = eh = 1;
 			eh = 0;
-		else
+		} else if (*str == 'p') {
+			precise++;
+		} else
 			break;
+
 		++str;
 	}
 	if (str >= *strp + 2) {
@@ -712,6 +718,7 @@ parse_event_modifier(const char **strp, struct perf_event_attr *attr)
 		attr->exclude_user   = eu;
 		attr->exclude_kernel = ek;
 		attr->exclude_hv     = eh;
+		attr->precise_ip     = precise;
 		return 1;
 	}
 	return 0;