Skip to contentSkip to conversation

perf/x86: various RAPL improvements and fixes

PATCH In review Stephane Eranian 3 participants 8 messages Jan 8, 2014 – Jan 13, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover perf/x86: various RAPL improvements and fixes

This short patch series provides the following improvements to the RAPL support for perf_events:
- support for RAPL PP1 energy counter
- fix struct perf_event active_list field initialization

The PP1 counter maps to the builtin graphic card for client processors. that is why it is called energy-gpu.

The new energy event is:
- name: power/energy-gpu/
- code: 0x4
- unit: Joules
- scaling factor: 2^-32 Joules.

In v2, we renamed energy-pp1 to energy-gpu. We dropped the Haswell Celeron patch because it is committed already. But we add a fix for the broken initialization of the active_entry list.

Stephane Eranian (2):
perf/x86: fix active_entry initialization
perf/x86: add RAPL PP1 energy counter support

 arch/x86/kernel/cpu/perf_event_intel_rapl.c |   31 +++++++++++++++++++++------
 include/linux/perf_event.h                  |    6 ++----
 kernel/events/core.c                        |    2 ++
 3 files changed, 28 insertions(+), 11 deletions(-)
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH v2 1/2 perf/x86: fix active_entry initialization

This patch fixes a problem with the initialization of the struct perf_event active_entry field. It is defined inside an anonymous union and was initialized in perf_event_alloc() using INIT_LIST_HEAD(). However at that time, we do not know whether the event is going to use active_entry or hlist_entry (SW). Or at last, we don't want to make that determination there. The problem is that hlist and list_head are not initialized the same way. One is okay with NULL (from kzmalloc), the other needs to pointers to point to self.

This patch resolves this problem by dropping the union. This will avoid problems later on, if someone starts using active_entry or hlist_entry without verifying that they actually overlap. This also solves the initialization problem.

---
 include/linux/perf_event.h |    6 ++----
 kernel/events/core.c       |    2 ++
 2 files changed, 4 insertions(+), 4 deletions(-)
include/linux/perf_event.h +2 -4
index 8f4a70f..e56b07f 100644
@@ -319,10 +319,8 @@ struct perf_event {
 	 */
 	struct list_head		migrate_entry;
 
-	union {
-		struct hlist_node	hlist_entry;
-		struct list_head	active_entry;
-	};
+	struct hlist_node		hlist_entry;
+	struct list_head		active_entry;
 	int				nr_siblings;
 	int				group_flags;
 	struct perf_event		*group_leader;
kernel/events/core.c +2 -0
index 4b743b2..74fdef5 100644
@@ -6685,6 +6685,8 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	INIT_LIST_HEAD(&event->sibling_list);
 	INIT_LIST_HEAD(&event->rb_entry);
 	INIT_LIST_HEAD(&event->active_entry);
+	INIT_HLIST_NODE(&event->hlist_entry);
+
 
 	init_waitqueue_head(&event->waitq);
 	init_irq_work(&event->pending, perf_pending_event);
Signature · 1 line
1.7.9.5
#2 lore raw

[tip:perf/core] perf/x86: Fix active_entry initialization

perf/x86: Fix active_entry initialization
Commit-ID:  f3ae75de98c4bac145a87d830c156c96f9414022
Gitweb:     http://git.kernel.org/tip/f3ae75de98c4bac145a87d830c156c96f9414022
Author:     Stephane Eranian <[email protected]>
AuthorDate: Wed, 8 Jan 2014 11:15:52 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sun, 12 Jan 2014 10:16:07 +0100

perf/x86: Fix active_entry initialization

This patch fixes a problem with the initialization of the struct perf_event active_entry field. It is defined inside an anonymous union and was initialized in perf_event_alloc() using INIT_LIST_HEAD(). However at that time, we do not know whether the event is going to use active_entry or hlist_entry (SW). Or at last, we don't want to make that determination there. The problem is that hlist and list_head are not initialized the same way. One is okay with NULL (from kzmalloc), the other needs to pointers to point to self.

This patch resolves this problem by dropping the union. This will avoid problems later on, if someone starts using active_entry or hlist_entry without verifying that they actually overlap. This also solves the initialization problem.

---
 include/linux/perf_event.h | 6 ++----
 kernel/events/core.c       | 2 ++
 2 files changed, 4 insertions(+), 4 deletions(-)
include/linux/perf_event.h +2 -4
index 8f4a70f..e56b07f 100644
@@ -319,10 +319,8 @@ struct perf_event {
 	 */
 	struct list_head		migrate_entry;
 
-	union {
-		struct hlist_node	hlist_entry;
-		struct list_head	active_entry;
-	};
+	struct hlist_node		hlist_entry;
+	struct list_head		active_entry;
 	int				nr_siblings;
 	int				group_flags;
 	struct perf_event		*group_leader;
kernel/events/core.c +2 -0
index 89d34f9..c3b6c27 100644
@@ -6670,6 +6670,8 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	INIT_LIST_HEAD(&event->sibling_list);
 	INIT_LIST_HEAD(&event->rb_entry);
 	INIT_LIST_HEAD(&event->active_entry);
+	INIT_HLIST_NODE(&event->hlist_entry);
+
 
 	init_waitqueue_head(&event->waitq);
 	init_irq_work(&event->pending, perf_pending_event);
maintainer
#3 lore raw

Re: [tip:perf/core] perf/x86: Fix active_entry initialization

include/linux/perf_event.h struct perf_event { · line 323

Quoted 6 lines · tip-bot for Stephane Eranian
include/linux/perf_event.h +2 -4
-	union {
-		struct hlist_node	hlist_entry;
-		struct list_head	active_entry;
-	};
+	struct hlist_node		hlist_entry;
+	struct list_head		active_entry;

Fwiw, we could probably stick both of those in hw_perf_event.

the software event's hrtimer isn't the biggest entry anymore, so we could add the hlist_entry in there, and we can start a new entry for RAPL events.

#4 lore raw

Re: [tip:perf/core] perf/x86: Fix active_entry initialization

Quoted 9 lines · Peter Zijlstra

include/linux/perf_event.h struct perf_event { · line 323

Quoted 6 lines · tip-bot for Stephane Eranian
include/linux/perf_event.h +2 -4
-     union {
-             struct hlist_node       hlist_entry;
-             struct list_head        active_entry;
-     };
+     struct hlist_node               hlist_entry;
+     struct list_head                active_entry;
Quoted 5 lines

Fwiw, we could probably stick both of those in hw_perf_event.

the software event's hrtimer isn't the biggest entry anymore, so we could add the hlist_entry in there, and we can start a new entry for RAPL events.

Yeah, that would be possible.
The other thing I noticed too in another uncore patch I am preparing is that we could also use that active_entry to avoid tracking association of uncore events to counters to update them in the hrtimer routine. All we need there is the list of active event on the uncore PMU and we iterate over them.

#5 lore raw

PATCH v2 2/2 perf/x86: add RAPL PP1 energy counter support

In reply to Stephane Eranian

This patch adds support for the RAPL energy counter PP1 (Power Plane 1).

On client processors, it usually corresponds to the energy consumption of the builtin graphic card. That is why the sysfs event is called energy-gpu.

New event:
- name: power/energy-gpu/
- code: event=0x4
- unit: 2^-32 Joules

On processors without graphics, this should count 0. The patch only enables this event on client processors.

---
 arch/x86/kernel/cpu/perf_event_intel_rapl.c |   31 +++++++++++++++++++++------
 1 file changed, 24 insertions(+), 7 deletions(-)
arch/x86/kernel/cpu/perf_event_intel_rapl.c +24 -7
index 0e3754e..5ad35ad 100644
@@ -27,6 +27,10 @@
  *	  event: rapl_energy_dram
  *    perf code: 0x3
  *
+ * dram counter: consumption of the builtin-gpu domain (client only)
+ *	  event: rapl_energy_gpu
+ *    perf code: 0x4
+ *
  * We manage those counters as free running (read-only). They may be
  * use simultaneously by other tools, such as turbostat.
  *
@@ -55,10 +59,13 @@
 #define INTEL_RAPL_PKG		0x2	/* pseudo-encoding */
 #define RAPL_IDX_RAM_NRG_STAT	2	/* DRAM */
 #define INTEL_RAPL_RAM		0x3	/* pseudo-encoding */
+#define RAPL_IDX_PP1_NRG_STAT	3	/* DRAM */
+#define INTEL_RAPL_PP1		0x4	/* pseudo-encoding */
 
 /* Clients have PP0, PKG */
 #define RAPL_IDX_CLN	(1<<RAPL_IDX_PP0_NRG_STAT|\
-			 1<<RAPL_IDX_PKG_NRG_STAT)
+			 1<<RAPL_IDX_PKG_NRG_STAT|\
+			 1<<RAPL_IDX_PP1_NRG_STAT)
 
 /* Servers have PP0, PKG, RAM */
 #define RAPL_IDX_SRV	(1<<RAPL_IDX_PP0_NRG_STAT|\
@@ -315,6 +322,10 @@ static int rapl_pmu_event_init(struct perf_event *event)
 		bit = RAPL_IDX_RAM_NRG_STAT;
 		msr = MSR_DRAM_ENERGY_STATUS;
 		break;
+	case INTEL_RAPL_PP1:
+		bit = RAPL_IDX_PP1_NRG_STAT;
+		msr = MSR_PP1_ENERGY_STATUS;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -367,19 +378,22 @@ static struct attribute_group rapl_pmu_attr_group = {
 };
 
 EVENT_ATTR_STR(energy-cores, rapl_cores, "event=0x01");
-EVENT_ATTR_STR(energy-pkg  , rapl_pkg, "event=0x02");
-EVENT_ATTR_STR(energy-ram  , rapl_ram, "event=0x03");
+EVENT_ATTR_STR(energy-pkg  ,   rapl_pkg, "event=0x02");
+EVENT_ATTR_STR(energy-ram  ,   rapl_ram, "event=0x03");
+EVENT_ATTR_STR(energy-gpu  ,   rapl_gpu, "event=0x04");
 
 EVENT_ATTR_STR(energy-cores.unit, rapl_cores_unit, "Joules");
-EVENT_ATTR_STR(energy-pkg.unit  , rapl_pkg_unit, "Joules");
-EVENT_ATTR_STR(energy-ram.unit  , rapl_ram_unit, "Joules");
+EVENT_ATTR_STR(energy-pkg.unit  ,   rapl_pkg_unit, "Joules");
+EVENT_ATTR_STR(energy-ram.unit  ,   rapl_ram_unit, "Joules");
+EVENT_ATTR_STR(energy-gpu.unit  ,   rapl_gpu_unit, "Joules");
 
 /*
  * we compute in 0.23 nJ increments regardless of MSR
  */
 EVENT_ATTR_STR(energy-cores.scale, rapl_cores_scale, "2.3283064365386962890625e-10");
-EVENT_ATTR_STR(energy-pkg.scale, rapl_pkg_scale, "2.3283064365386962890625e-10");
-EVENT_ATTR_STR(energy-ram.scale, rapl_ram_scale, "2.3283064365386962890625e-10");
+EVENT_ATTR_STR(energy-pkg.scale,     rapl_pkg_scale, "2.3283064365386962890625e-10");
+EVENT_ATTR_STR(energy-ram.scale,     rapl_ram_scale, "2.3283064365386962890625e-10");
+EVENT_ATTR_STR(energy-gpu.scale,     rapl_gpu_scale, "2.3283064365386962890625e-10");
 
 static struct attribute *rapl_events_srv_attr[] = {
 	EVENT_PTR(rapl_cores),
@@ -399,12 +413,15 @@ static struct attribute *rapl_events_srv_attr[] = {
 static struct attribute *rapl_events_cln_attr[] = {
 	EVENT_PTR(rapl_cores),
 	EVENT_PTR(rapl_pkg),
+	EVENT_PTR(rapl_gpu),
 
 	EVENT_PTR(rapl_cores_unit),
 	EVENT_PTR(rapl_pkg_unit),
+	EVENT_PTR(rapl_gpu_unit),
 
 	EVENT_PTR(rapl_cores_scale),
 	EVENT_PTR(rapl_pkg_scale),
+	EVENT_PTR(rapl_gpu_scale),
 	NULL,
 };
 
Signature · 1 line
1.7.9.5
#6 lore raw

[tip:perf/core] perf/x86/intel: Add Intel RAPL PP1 energy counter support

perf/x86/intel: Add Intel RAPL PP1 energy counter support
Commit-ID:  f228c5b882602697a1adb50d61ff688b0df1eced
Gitweb:     http://git.kernel.org/tip/f228c5b882602697a1adb50d61ff688b0df1eced
Author:     Stephane Eranian <[email protected]>
AuthorDate: Wed, 8 Jan 2014 11:15:53 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sun, 12 Jan 2014 10:16:08 +0100

perf/x86/intel: Add Intel RAPL PP1 energy counter support

This patch adds support for the Intel RAPL energy counter PP1 (Power Plane 1).

On client processors, it usually corresponds to the energy consumption of the builtin graphic card. That is why the sysfs event is called energy-gpu.

New event:
- name: power/energy-gpu/
- code: event=0x4
- unit: 2^-32 Joules

On processors without graphics, this should count 0. The patch only enables this event on client processors.

---
 arch/x86/kernel/cpu/perf_event_intel_rapl.c | 31 ++++++++++++++++++++++-------
 1 file changed, 24 insertions(+), 7 deletions(-)
arch/x86/kernel/cpu/perf_event_intel_rapl.c +24 -7
index 0e3754e..5ad35ad 100644
@@ -27,6 +27,10 @@
  *	  event: rapl_energy_dram
  *    perf code: 0x3
  *
+ * dram counter: consumption of the builtin-gpu domain (client only)
+ *	  event: rapl_energy_gpu
+ *    perf code: 0x4
+ *
  * We manage those counters as free running (read-only). They may be
  * use simultaneously by other tools, such as turbostat.
  *
@@ -55,10 +59,13 @@
 #define INTEL_RAPL_PKG		0x2	/* pseudo-encoding */
 #define RAPL_IDX_RAM_NRG_STAT	2	/* DRAM */
 #define INTEL_RAPL_RAM		0x3	/* pseudo-encoding */
+#define RAPL_IDX_PP1_NRG_STAT	3	/* DRAM */
+#define INTEL_RAPL_PP1		0x4	/* pseudo-encoding */
 
 /* Clients have PP0, PKG */
 #define RAPL_IDX_CLN	(1<<RAPL_IDX_PP0_NRG_STAT|\
-			 1<<RAPL_IDX_PKG_NRG_STAT)
+			 1<<RAPL_IDX_PKG_NRG_STAT|\
+			 1<<RAPL_IDX_PP1_NRG_STAT)
 
 /* Servers have PP0, PKG, RAM */
 #define RAPL_IDX_SRV	(1<<RAPL_IDX_PP0_NRG_STAT|\
@@ -315,6 +322,10 @@ static int rapl_pmu_event_init(struct perf_event *event)
 		bit = RAPL_IDX_RAM_NRG_STAT;
 		msr = MSR_DRAM_ENERGY_STATUS;
 		break;
+	case INTEL_RAPL_PP1:
+		bit = RAPL_IDX_PP1_NRG_STAT;
+		msr = MSR_PP1_ENERGY_STATUS;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -367,19 +378,22 @@ static struct attribute_group rapl_pmu_attr_group = {
 };
 
 EVENT_ATTR_STR(energy-cores, rapl_cores, "event=0x01");
-EVENT_ATTR_STR(energy-pkg  , rapl_pkg, "event=0x02");
-EVENT_ATTR_STR(energy-ram  , rapl_ram, "event=0x03");
+EVENT_ATTR_STR(energy-pkg  ,   rapl_pkg, "event=0x02");
+EVENT_ATTR_STR(energy-ram  ,   rapl_ram, "event=0x03");
+EVENT_ATTR_STR(energy-gpu  ,   rapl_gpu, "event=0x04");
 
 EVENT_ATTR_STR(energy-cores.unit, rapl_cores_unit, "Joules");
-EVENT_ATTR_STR(energy-pkg.unit  , rapl_pkg_unit, "Joules");
-EVENT_ATTR_STR(energy-ram.unit  , rapl_ram_unit, "Joules");
+EVENT_ATTR_STR(energy-pkg.unit  ,   rapl_pkg_unit, "Joules");
+EVENT_ATTR_STR(energy-ram.unit  ,   rapl_ram_unit, "Joules");
+EVENT_ATTR_STR(energy-gpu.unit  ,   rapl_gpu_unit, "Joules");
 
 /*
  * we compute in 0.23 nJ increments regardless of MSR
  */
 EVENT_ATTR_STR(energy-cores.scale, rapl_cores_scale, "2.3283064365386962890625e-10");
-EVENT_ATTR_STR(energy-pkg.scale, rapl_pkg_scale, "2.3283064365386962890625e-10");
-EVENT_ATTR_STR(energy-ram.scale, rapl_ram_scale, "2.3283064365386962890625e-10");
+EVENT_ATTR_STR(energy-pkg.scale,     rapl_pkg_scale, "2.3283064365386962890625e-10");
+EVENT_ATTR_STR(energy-ram.scale,     rapl_ram_scale, "2.3283064365386962890625e-10");
+EVENT_ATTR_STR(energy-gpu.scale,     rapl_gpu_scale, "2.3283064365386962890625e-10");
 
 static struct attribute *rapl_events_srv_attr[] = {
 	EVENT_PTR(rapl_cores),
@@ -399,12 +413,15 @@ static struct attribute *rapl_events_srv_attr[] = {
 static struct attribute *rapl_events_cln_attr[] = {
 	EVENT_PTR(rapl_cores),
 	EVENT_PTR(rapl_pkg),
+	EVENT_PTR(rapl_gpu),
 
 	EVENT_PTR(rapl_cores_unit),
 	EVENT_PTR(rapl_pkg_unit),
+	EVENT_PTR(rapl_gpu_unit),
 
 	EVENT_PTR(rapl_cores_scale),
 	EVENT_PTR(rapl_pkg_scale),
+	EVENT_PTR(rapl_gpu_scale),
 	NULL,
 };
 
maintainer
#7 lore raw
In reply to Stephane Eranian
Quoted 14 lines · Stephane Eranian

This short patch series provides the following improvements to the RAPL support for perf_events:
- support for RAPL PP1 energy counter
- fix struct perf_event active_list field initialization

The PP1 counter maps to the builtin graphic card for client processors. that is why it is called energy-gpu.

The new energy event is:
- name: power/energy-gpu/
- code: 0x4
- unit: Joules
- scaling factor: 2^-32 Joules.

Quoted 4 lines

In v2, we renamed energy-pp1 to energy-gpu. We dropped the Haswell Celeron patch because it is committed already. But we add a fix for the broken initialization of the active_entry list.

Thanks!