Skip to contentSkip to conversation

perf/x86/intel/uncore: Provide alias for IIO free-running boxes on SKX

PATCH In review Jin Yao 2 participants 4 messages Sep 4, 2018 – Sep 10, 2018

Subsystems (2)
lore
Share & tools
lore

Jin Yao

author
#0 lore raw

PATCH v2 1/1 perf/x86/intel/uncore: Provide alias for IIO free-running boxes on SKX

For Skylake Server, Linux has supported a number of free running counters that collect counts of IO clocks/Bandwidth/Utilization.

For example, to collect the inbound bandwidth,

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_free_running_0
uncore_iio_free_running_1
uncore_iio_free_running_2
uncore_iio_free_running_3
uncore_iio_free_running_4
uncore_iio_free_running_5

root@skx /sys/devices# perf stat -a -e uncore_iio_free_running_2/bw_in_port0/ ^C
Performance counter stats for 'system wide':

            153.19 MiB  uncore_iio_free_running_2/bw_in_port0/

       8.037701069 seconds time elapsed

While it's hard for user to understanding what the box the uncore_iio_free_running_N means.

This patch provides aliases for the boxes. With this patch, for example,

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_cbdma
uncore_iio_mcp0
uncore_iio_mcp1
uncore_iio_pcie0
uncore_iio_pcie1
uncore_iio_pcie2

root@skx ~# perf stat -a -e uncore_iio_pcie1/bw_in_port0/ ^C
Performance counter stats for 'system wide':

            153.12 MiB  uncore_iio_pcie1/bw_in_port0/

       8.469790720 seconds time elapsed

v2:
---
Previously, it used braces around initialized string. For example, { "iio_cbdma" }.

The braces around scalar initializer would cause the build warning. In v2, it removes the braces.

---
 arch/x86/events/intel/uncore.c       |  9 +++++++--
 arch/x86/events/intel/uncore.h       |  1 +
 arch/x86/events/intel/uncore_snbep.c | 10 ++++++++++
 3 files changed, 18 insertions(+), 2 deletions(-)
arch/x86/events/intel/uncore.c +7 -2
index 27a4614..6c6615f 100644
@@ -812,8 +812,13 @@ static int uncore_pmu_register(struct intel_uncore_pmu *pmu)
 		else
 			sprintf(pmu->name, "uncore");
 	} else {
-		sprintf(pmu->name, "uncore_%s_%d", pmu->type->name,
-			pmu->pmu_idx);
+		if (pmu->type->alias && pmu->pmu_idx < pmu->type->num_boxes) {
+			sprintf(pmu->name, "uncore_%s",
+				pmu->type->alias[pmu->pmu_idx]);
+		} else {
+			sprintf(pmu->name, "uncore_%s_%d", pmu->type->name,
+				pmu->pmu_idx);
+		}
 	}
 
 	ret = perf_pmu_register(&pmu->pmu, pmu->name, -1);
arch/x86/events/intel/uncore.h +1 -0
index e17ab88..db9fd9a 100644
@@ -44,6 +44,7 @@ struct freerunning_counters;
 
 struct intel_uncore_type {
 	const char *name;
+	const char **alias;
 	int num_counters;
 	int num_boxes;
 	int perf_ctr_bits;
arch/x86/events/intel/uncore_snbep.c +10 -0
index 51d7c11..75f14a3 100644
@@ -3596,8 +3596,18 @@ static const struct attribute_group skx_uncore_iio_freerunning_format_group = {
 	.attrs = skx_uncore_iio_freerunning_formats_attr,
 };
 
+static const char *skx_uncore_iio_free_running_aliases[] = {
+	"iio_cbdma",
+	"iio_pcie0",
+	"iio_pcie1",
+	"iio_pcie2",
+	"iio_mcp0",
+	"iio_mcp1",
+};
+
 static struct intel_uncore_type skx_uncore_iio_free_running = {
 	.name			= "iio_free_running",
+	.alias			= skx_uncore_iio_free_running_aliases,
 	.num_counters		= 17,
 	.num_boxes		= 6,
 	.num_freerunning_types	= SKX_IIO_FREERUNNING_TYPE_MAX,
Signature · 1 line
2.7.4
maintainer
#1 lore raw
Quoted 14 lines · Jin Yao

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_free_running_0
uncore_iio_free_running_1
uncore_iio_free_running_2
uncore_iio_free_running_3
uncore_iio_free_running_4
uncore_iio_free_running_5

Quoted 13 lines

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_cbdma
uncore_iio_mcp0
uncore_iio_mcp1
uncore_iio_pcie0
uncore_iio_pcie1
uncore_iio_pcie2

I think I'm ok with that, except of course for people that have "free_running_#" in their scripts now and will to wtf when they upgrade their kernel.

Do we care about them?

author
#2 lore raw
Quoted 29 lines · Peter Zijlstra
Quoted 14 lines · Jin Yao

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_free_running_0
uncore_iio_free_running_1
uncore_iio_free_running_2
uncore_iio_free_running_3
uncore_iio_free_running_4
uncore_iio_free_running_5

Quoted 13 lines

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_cbdma
uncore_iio_mcp0
uncore_iio_mcp1
uncore_iio_pcie0
uncore_iio_pcie1
uncore_iio_pcie2

Quoted 6 lines

I think I'm ok with that, except of course for people that have "free_running_#" in their scripts now and will to wtf when they upgrade their kernel.

Do we care about them?

Yes, that may be a potential issue but maybe it's not since we really don't know if some people have used uncore_iio_free_running_# in their scripts or not.

I write this patch is because I always forget the meaning of uncore_iio_free_running_# so I have to go back to check the document "Intel Xeon Processor Scalable Memory Family Uncore Performance Monitoring" again and again to find the box definition. I guess other people may have similar trouble.

Maybe we wait some time to see more feedback from community?

Thanks
Jin Yao

author
#3 lore raw

One week nearly passed, I guess there is no any objections for this patch from community. :)

Thanks
Jin Yao

Quoted 47 lines · Yao
Quoted 29 lines · Peter Zijlstra
Quoted 14 lines · Jin Yao

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_free_running_0
uncore_iio_free_running_1
uncore_iio_free_running_2
uncore_iio_free_running_3
uncore_iio_free_running_4
uncore_iio_free_running_5

Quoted 13 lines

root@skx /sys/devices# ls | grep uncore_iio
uncore_iio_0
uncore_iio_1
uncore_iio_2
uncore_iio_3
uncore_iio_4
uncore_iio_5
uncore_iio_cbdma
uncore_iio_mcp0
uncore_iio_mcp1
uncore_iio_pcie0
uncore_iio_pcie1
uncore_iio_pcie2

Quoted 6 lines

I think I'm ok with that, except of course for people that have "free_running_#" in their scripts now and will to wtf when they upgrade their kernel.

Do we care about them?

Yes, that may be a potential issue but maybe it's not since we really don't know if some people have used uncore_iio_free_running_# in their scripts or not.

I write this patch is because I always forget the meaning of uncore_iio_free_running_# so I have to go back to check the document "Intel Xeon Processor Scalable Memory Family Uncore Performance Monitoring" again and again to find the box definition. I guess other people may have similar trouble.

Maybe we wait some time to see more feedback from community?

Thanks
Jin Yao