Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1390325 > unrolled thread

[PATCH v0 0/3] perf, pt: Intel PT updates

Back to article view | Back to linux.kernel

Contents

  [PATCH v0 0/3] perf, pt: Intel PT updates Alexander Shishkin <alexander.shishkin@linux.intel.com> - 2016-04-28 17:40 +0200
    [PATCH v0 2/3] perf/x86/intel/pt: Export cpu frequency ratios needed by PT decoders Alexander Shishkin <alexander.shishkin@linux.intel.com> - 2016-04-28 17:40 +0200
      Re: [PATCH v0 2/3] perf/x86/intel/pt: Export cpu frequency ratios needed by PT decoders Alexander Shishkin <alexander.shishkin@linux.intel.com> - 2016-04-28 18:00 +0200
        [tip:perf/core] perf/x86/intel/pt: Export CPU frequency ratios  needed by PT decoders tip-bot for Alexander Shishkin <tipbot@zytor.com> - 2016-05-05 12:00 +0200
      Re: [PATCH v0 2/3] perf/x86/intel/pt: Export cpu frequency ratios  needed by PT decoders Borislav Petkov <bp@alien8.de> - 2016-05-03 16:00 +0200

#1390325 — [PATCH v0 0/3] perf, pt: Intel PT updates

Hi Peter,

Here are some patches for Intel PT driver, should all be
self-evident. 2/3 has been posted before in a slightly different form,
but it's been quite a while and everybody should have forgotten it by
now.

Alexander Shishkin (3):
  perf/x86/intel/pt: Convert ACCESS_ONCE()s
  perf/x86/intel/pt: Export cpu frequency ratios needed by PT decoders
  perf, x86: Bypass PT vs LBR exclusivity if the core supports it

 arch/x86/events/core.c       |  6 +++++
 arch/x86/events/intel/core.c |  1 +
 arch/x86/events/intel/pt.c   | 60 +++++++++++++++++++++++++++++++++++++++++---
 arch/x86/events/intel/pt.h   |  6 +++++
 arch/x86/events/perf_event.h |  1 +
 5 files changed, 71 insertions(+), 3 deletions(-)

-- 
2.8.0.rc3

[toc] | [next] | [standalone]

#1390328 — [PATCH v0 2/3] perf/x86/intel/pt: Export cpu frequency ratios needed by PT decoders

In order for Intel PT decoders to infer correct crystal clock and bus
frequencies that are required to correctly decode timing information
from a PT stream (MTC and CBR packets), export them as sysfs attributes:

  * max_nonturbo_ratio: ratio between the invariant TSC and base clock;
  * tsc_art_ratio: TSC to core crystal clock ratio.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
---
 arch/x86/events/intel/pt.c | 54 ++++++++++++++++++++++++++++++++++++++++++++++
 arch/x86/events/intel/pt.h |  6 ++++++
 2 files changed, 60 insertions(+)

diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c
index e88915ff98..bf0b0fe33d 100644
--- a/arch/x86/events/intel/pt.c
+++ b/arch/x86/events/intel/pt.c
@@ -125,9 +125,46 @@ static struct attribute_group pt_format_group = {
 	.attrs	= pt_formats_attr,
 };
 
+static ssize_t
+pt_timing_attr_show(struct device *dev, struct device_attribute *attr,
+		    char *page)
+{
+	struct perf_pmu_events_attr *pmu_attr =
+		container_of(attr, struct perf_pmu_events_attr, attr);
+
+	switch (pmu_attr->id) {
+	case 0:
+		return sprintf(page, "%lu\n", pt_pmu.max_nonturbo_ratio);
+	case 1:
+		return sprintf(page, "%u:%u\n",
+			       pt_pmu.tsc_art_num,
+			       pt_pmu.tsc_art_den);
+	default:
+		break;
+	}
+
+	return -EINVAL;
+}
+
+PMU_EVENT_ATTR(max_nonturbo_ratio, timing_attr_max_nonturbo_ratio, 0,
+	       pt_timing_attr_show);
+PMU_EVENT_ATTR(tsc_art_ratio, timing_attr_tsc_art_ratio, 1,
+	       pt_timing_attr_show);
+
+static struct attribute *pt_timing_attr[] = {
+	&timing_attr_max_nonturbo_ratio.attr.attr,
+	&timing_attr_tsc_art_ratio.attr.attr,
+	NULL,
+};
+
+static struct attribute_group pt_timing_group = {
+	.attrs	= pt_timing_attr,
+};
+
 static const struct attribute_group *pt_attr_groups[] = {
 	&pt_cap_group,
 	&pt_format_group,
+	&pt_timing_group,
 	NULL,
 };
 
@@ -140,6 +177,23 @@ static int __init pt_pmu_hw_init(void)
 	int ret;
 	long i;
 
+	rdmsrl(MSR_PLATFORM_INFO, reg);
+	pt_pmu.max_nonturbo_ratio = (reg & 0xff00) >> 8;
+
+	/*
+	 * if available, read in TSC to core crystal clock ratio,
+	 * otherwise, zero for numerator stands for "not enumerated"
+	 * as per SDM
+	 */
+	if (boot_cpu_data.cpuid_level >= CPUID_TSC_LEAF) {
+		u32 eax, ebx, ecx, edx;
+
+		cpuid(CPUID_TSC_LEAF, &eax, &ebx, &ecx, &edx);
+
+		pt_pmu.tsc_art_num = ebx;
+		pt_pmu.tsc_art_den = eax;
+	}
+
 	if (boot_cpu_has(X86_FEATURE_VMX)) {
 		/*
 		 * Intel SDM, 36.5 "Tracing post-VMXON" says that
diff --git a/arch/x86/events/intel/pt.h b/arch/x86/events/intel/pt.h
index 3abb5f5ccc..e5385b1fa0 100644
--- a/arch/x86/events/intel/pt.h
+++ b/arch/x86/events/intel/pt.h
@@ -48,6 +48,9 @@ struct topa_entry {
 #define PT_CPUID_LEAVES		2
 #define PT_CPUID_REGS_NUM	4 /* number of regsters (eax, ebx, ecx, edx) */
 
+/* TSC to Core Crystal Clock Ratio */
+#define CPUID_TSC_LEAF		0x15
+
 enum pt_capabilities {
 	PT_CAP_max_subleaf = 0,
 	PT_CAP_cr3_filtering,
@@ -66,6 +69,9 @@ struct pt_pmu {
 	struct pmu		pmu;
 	u32			caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES];
 	bool			vmx;
+	unsigned long		max_nonturbo_ratio;
+	unsigned int		tsc_art_num;
+	unsigned int		tsc_art_den;
 };
 
 /**
-- 
2.8.0.rc3

[toc] | [prev] | [next] | [standalone]

#1390345 — Re: [PATCH v0 2/3] perf/x86/intel/pt: Export cpu frequency ratios needed by PT decoders

Alexander Shishkin <alexander.shishkin@linux.intel.com> writes:

> In order for Intel PT decoders to infer correct crystal clock and bus
> frequencies that are required to correctly decode timing information
> from a PT stream (MTC and CBR packets), export them as sysfs attributes:
>
>   * max_nonturbo_ratio: ratio between the invariant TSC and base clock;
>   * tsc_art_ratio: TSC to core crystal clock ratio.
>
> Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>

With a slightly better changelog:

From d0fbc4e45fb633e464cbbd61d6aacd5a476fd50d Mon Sep 17 00:00:00 2001
From: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Date: Wed, 19 Aug 2015 17:02:10 +0300
Subject: [PATCH] perf/x86/intel/pt: Export cpu frequency ratios needed by PT
 decoders

Intel PT decoders need access to various bits of timing related
information to be able to correctly decode timing packets from a PT
stream (MTC and CBR packets). This patch exports all the necessary
bits as sysfs attributes for the sake of consistency:

  * max_nonturbo_ratio: ratio between the invariant TSC and base clock;
  * tsc_art_ratio: TSC to core crystal clock ratio (also available as
    CPUID.15H).

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
---
 arch/x86/events/intel/pt.c | 54 ++++++++++++++++++++++++++++++++++++++++++++++
 arch/x86/events/intel/pt.h |  6 ++++++
 2 files changed, 60 insertions(+)

diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c
index e88915ff98..bf0b0fe33d 100644
--- a/arch/x86/events/intel/pt.c
+++ b/arch/x86/events/intel/pt.c
@@ -125,9 +125,46 @@ static struct attribute_group pt_format_group = {
 	.attrs	= pt_formats_attr,
 };
 
+static ssize_t
+pt_timing_attr_show(struct device *dev, struct device_attribute *attr,
+		    char *page)
+{
+	struct perf_pmu_events_attr *pmu_attr =
+		container_of(attr, struct perf_pmu_events_attr, attr);
+
+	switch (pmu_attr->id) {
+	case 0:
+		return sprintf(page, "%lu\n", pt_pmu.max_nonturbo_ratio);
+	case 1:
+		return sprintf(page, "%u:%u\n",
+			       pt_pmu.tsc_art_num,
+			       pt_pmu.tsc_art_den);
+	default:
+		break;
+	}
+
+	return -EINVAL;
+}
+
+PMU_EVENT_ATTR(max_nonturbo_ratio, timing_attr_max_nonturbo_ratio, 0,
+	       pt_timing_attr_show);
+PMU_EVENT_ATTR(tsc_art_ratio, timing_attr_tsc_art_ratio, 1,
+	       pt_timing_attr_show);
+
+static struct attribute *pt_timing_attr[] = {
+	&timing_attr_max_nonturbo_ratio.attr.attr,
+	&timing_attr_tsc_art_ratio.attr.attr,
+	NULL,
+};
+
+static struct attribute_group pt_timing_group = {
+	.attrs	= pt_timing_attr,
+};
+
 static const struct attribute_group *pt_attr_groups[] = {
 	&pt_cap_group,
 	&pt_format_group,
+	&pt_timing_group,
 	NULL,
 };
 
@@ -140,6 +177,23 @@ static int __init pt_pmu_hw_init(void)
 	int ret;
 	long i;
 
+	rdmsrl(MSR_PLATFORM_INFO, reg);
+	pt_pmu.max_nonturbo_ratio = (reg & 0xff00) >> 8;
+
+	/*
+	 * if available, read in TSC to core crystal clock ratio,
+	 * otherwise, zero for numerator stands for "not enumerated"
+	 * as per SDM
+	 */
+	if (boot_cpu_data.cpuid_level >= CPUID_TSC_LEAF) {
+		u32 eax, ebx, ecx, edx;
+
+		cpuid(CPUID_TSC_LEAF, &eax, &ebx, &ecx, &edx);
+
+		pt_pmu.tsc_art_num = ebx;
+		pt_pmu.tsc_art_den = eax;
+	}
+
 	if (boot_cpu_has(X86_FEATURE_VMX)) {
 		/*
 		 * Intel SDM, 36.5 "Tracing post-VMXON" says that
diff --git a/arch/x86/events/intel/pt.h b/arch/x86/events/intel/pt.h
index 3abb5f5ccc..e5385b1fa0 100644
--- a/arch/x86/events/intel/pt.h
+++ b/arch/x86/events/intel/pt.h
@@ -48,6 +48,9 @@ struct topa_entry {
 #define PT_CPUID_LEAVES		2
 #define PT_CPUID_REGS_NUM	4 /* number of regsters (eax, ebx, ecx, edx) */
 
+/* TSC to Core Crystal Clock Ratio */
+#define CPUID_TSC_LEAF		0x15
+
 enum pt_capabilities {
 	PT_CAP_max_subleaf = 0,
 	PT_CAP_cr3_filtering,
@@ -66,6 +69,9 @@ struct pt_pmu {
 	struct pmu		pmu;
 	u32			caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES];
 	bool			vmx;
+	unsigned long		max_nonturbo_ratio;
+	unsigned int		tsc_art_num;
+	unsigned int		tsc_art_den;
 };
 
 /**
-- 
2.8.0.rc3

[toc] | [prev] | [next] | [standalone]

#1395011 — [tip:perf/core] perf/x86/intel/pt: Export CPU frequency ratios needed by PT decoders

Commit-ID:  65c7e6f1c4810e9bce935520f44f6d2613cd1b40
Gitweb:     http://git.kernel.org/tip/65c7e6f1c4810e9bce935520f44f6d2613cd1b40
Author:     Alexander Shishkin <alexander.shishkin@linux.intel.com>
AuthorDate: Wed, 19 Aug 2015 17:02:10 +0300
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 5 May 2016 10:16:28 +0200

perf/x86/intel/pt: Export CPU frequency ratios needed by PT decoders

Intel PT decoders need access to various bits of timing related
information to be able to correctly decode timing packets from a PT
stream (MTC and CBR packets). This patch exports all the necessary
bits as sysfs attributes for the sake of consistency:

  * max_nonturbo_ratio: ratio between the invariant TSC and base clock;

  * tsc_art_ratio: TSC to core crystal clock ratio (also available as CPUID.15H).

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/87zisdvibe.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/events/intel/pt.c | 54 ++++++++++++++++++++++++++++++++++++++++++++++
 arch/x86/events/intel/pt.h |  6 ++++++
 2 files changed, 60 insertions(+)

diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c
index 2d1ce2c..c3a359c 100644
--- a/arch/x86/events/intel/pt.c
+++ b/arch/x86/events/intel/pt.c
@@ -127,9 +127,46 @@ static struct attribute_group pt_format_group = {
 	.attrs	= pt_formats_attr,
 };
 
+static ssize_t
+pt_timing_attr_show(struct device *dev, struct device_attribute *attr,
+		    char *page)
+{
+	struct perf_pmu_events_attr *pmu_attr =
+		container_of(attr, struct perf_pmu_events_attr, attr);
+
+	switch (pmu_attr->id) {
+	case 0:
+		return sprintf(page, "%lu\n", pt_pmu.max_nonturbo_ratio);
+	case 1:
+		return sprintf(page, "%u:%u\n",
+			       pt_pmu.tsc_art_num,
+			       pt_pmu.tsc_art_den);
+	default:
+		break;
+	}
+
+	return -EINVAL;
+}
+
+PMU_EVENT_ATTR(max_nonturbo_ratio, timing_attr_max_nonturbo_ratio, 0,
+	       pt_timing_attr_show);
+PMU_EVENT_ATTR(tsc_art_ratio, timing_attr_tsc_art_ratio, 1,
+	       pt_timing_attr_show);
+
+static struct attribute *pt_timing_attr[] = {
+	&timing_attr_max_nonturbo_ratio.attr.attr,
+	&timing_attr_tsc_art_ratio.attr.attr,
+	NULL,
+};
+
+static struct attribute_group pt_timing_group = {
+	.attrs	= pt_timing_attr,
+};
+
 static const struct attribute_group *pt_attr_groups[] = {
 	&pt_cap_group,
 	&pt_format_group,
+	&pt_timing_group,
 	NULL,
 };
 
@@ -142,6 +179,23 @@ static int __init pt_pmu_hw_init(void)
 	int ret;
 	long i;
 
+	rdmsrl(MSR_PLATFORM_INFO, reg);
+	pt_pmu.max_nonturbo_ratio = (reg & 0xff00) >> 8;
+
+	/*
+	 * if available, read in TSC to core crystal clock ratio,
+	 * otherwise, zero for numerator stands for "not enumerated"
+	 * as per SDM
+	 */
+	if (boot_cpu_data.cpuid_level >= CPUID_TSC_LEAF) {
+		u32 eax, ebx, ecx, edx;
+
+		cpuid(CPUID_TSC_LEAF, &eax, &ebx, &ecx, &edx);
+
+		pt_pmu.tsc_art_num = ebx;
+		pt_pmu.tsc_art_den = eax;
+	}
+
 	if (boot_cpu_has(X86_FEATURE_VMX)) {
 		/*
 		 * Intel SDM, 36.5 "Tracing post-VMXON" says that
diff --git a/arch/x86/events/intel/pt.h b/arch/x86/events/intel/pt.h
index ca64599..efffa4a 100644
--- a/arch/x86/events/intel/pt.h
+++ b/arch/x86/events/intel/pt.h
@@ -82,6 +82,9 @@ struct topa_entry {
 #define PT_CPUID_LEAVES		2
 #define PT_CPUID_REGS_NUM	4 /* number of regsters (eax, ebx, ecx, edx) */
 
+/* TSC to Core Crystal Clock Ratio */
+#define CPUID_TSC_LEAF		0x15
+
 enum pt_capabilities {
 	PT_CAP_max_subleaf = 0,
 	PT_CAP_cr3_filtering,
@@ -102,6 +105,9 @@ struct pt_pmu {
 	struct pmu		pmu;
 	u32			caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES];
 	bool			vmx;
+	unsigned long		max_nonturbo_ratio;
+	unsigned int		tsc_art_num;
+	unsigned int		tsc_art_den;
 };
 
 /**

[toc] | [prev] | [next] | [standalone]

#1393432 — Re: [PATCH v0 2/3] perf/x86/intel/pt: Export cpu frequency ratios needed by PT decoders

On Thu, Apr 28, 2016 at 06:35:45PM +0300, Alexander Shishkin wrote:
> In order for Intel PT decoders to infer correct crystal clock and bus
> frequencies that are required to correctly decode timing information
> from a PT stream (MTC and CBR packets), export them as sysfs attributes:
> 
>   * max_nonturbo_ratio: ratio between the invariant TSC and base clock;
>   * tsc_art_ratio: TSC to core crystal clock ratio.

I guess new sysfs nodes need to be documented:

Documentation/ABI/...

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web