* [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics
@ 2021-03-24 14:30 Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 1/4] perf stat: Basic support for iostat in perf Alexander Antonov
` (4 more replies)
0 siblings, 5 replies; 6+ messages in thread
From: Alexander Antonov @ 2021-03-24 14:30 UTC (permalink / raw)
To: acme
Cc: linux-kernel, jolsa, ak, alexander.shishkin, mark.rutland,
namhyung, irogers, mingo, peterz, alexey.v.bayduraev,
alexander.antonov
The previous version can be found at:
v4: https://lkml.kernel.org/r/20210203135830.38568-1-alexander.antonov@linux.intel.com/
Changes in this revision are:
v4 -> v5:
- Addressed comments from Namhyung Kim:
1. Removed AGGR_PCIE_PORT aggregation mode
2. Added iostat_prepare() function
3. Moved implementation specific fprintf() calls to separate x86-related function
4. Fixed code-related issues
- Moved __weak iostat's functions to separate util/iostat.c file
The previous version can be found at:
v3: https://lkml.kernel.org/r/20210126080619.30275-1-alexander.antonov@linux.intel.com/
Changes in this revision are:
v3 -> v4:
- Addressed comment from Namhyung Kim:
1. Removed NULL-termination of root ports list
The previous version can be found at:
v2: https://lkml.kernel.org/r/20201223130320.3930-1-alexander.antonov@linux.intel.com
Changes in this revision are:
v2 -> v3:
- Addressed comments from Namhyung Kim:
1. Removed perf_device pointer from evsel structure. Use priv field instead
2. Renamed 'iiostat' to 'iostat'
3. Renamed 'show' mode to 'list' mode
4. Renamed iiostat_delete_root_ports() to iiostat_release() and
iostat_show_root_ports() to iostat_list()
The previous version can be found at:
v1: https://lkml.kernel.org/r/20201210090340.14358-1-alexander.antonov@linux.intel.com
Changes in this revision are:
v1 -> v2:
- Addressed comment from Arnaldo Carvalho de Melo:
1. Using 'perf iiostat' subcommand instead of 'perf stat --iiostat':
- Added perf-iiostat.sh script to use short command
- Updated manual pages to get help for 'perf iiostat'
- Added 'perf-iiostat' to perf's gitignore file
Mode is intended to provide four I/O performance metrics in MB per each
root port:
- Inbound Read: I/O devices below root port read from the host memory
- Inbound Write: I/O devices below root port write to the host memory
- Outbound Read: CPU reads from I/O devices below root port
- Outbound Write: CPU writes to I/O devices below root port
Each metric requiries only one uncore event which increments at every 4B
transfer in corresponding direction. The formulas to compute metrics
are generic:
#EventCount * 4B / (1024 * 1024)
Note: iostat introduces new perf data aggregation mode - per PCIe root port
hence -e and -M options are not supported.
Usage examples:
1. List all PCIe root ports (example for 2-S platform):
$ perf iostat list
S0-uncore_iio_0<0000:00>
S1-uncore_iio_0<0000:80>
S0-uncore_iio_1<0000:17>
S1-uncore_iio_1<0000:85>
S0-uncore_iio_2<0000:3a>
S1-uncore_iio_2<0000:ae>
S0-uncore_iio_3<0000:5d>
S1-uncore_iio_3<0000:d7>
2. Collect metrics for all PCIe root ports:
$ perf iostat -- dd if=/dev/zero of=/dev/nvme0n1 bs=1M oflag=direct
357708+0 records in
357707+0 records out
375083606016 bytes (375 GB, 349 GiB) copied, 215.974 s, 1.7 GB/s
Performance counter stats for 'system wide':
port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound Write(MB)
0000:00 1 0 2 3
0000:80 0 0 0 0
0000:17 352552 43 0 21
0000:85 0 0 0 0
0000:3a 3 0 0 0
0000:ae 0 0 0 0
0000:5d 0 0 0 0
0000:d7 0 0 0 0
3. Collect metrics for comma separated list of PCIe root ports:
$ perf iostat 0000:17,0:3a -- dd if=/dev/zero of=/dev/nvme0n1 bs=1M oflag=direct
357708+0 records in
357707+0 records out
375083606016 bytes (375 GB, 349 GiB) copied, 197.08 s, 1.9 GB/s
Performance counter stats for 'system wide':
port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound Write(MB)
0000:17 358559 44 0 22
0000:3a 3 2 0 0
197.081983474 seconds time elapsed
Alexander Antonov (4):
perf stat: Basic support for iostat in perf
perf stat: Helper functions for PCIe root ports list in iostat mode
perf stat: Enable iostat mode for x86 platforms
perf: Update .gitignore file
tools/perf/.gitignore | 1 +
tools/perf/Documentation/perf-iostat.txt | 88 +++++
tools/perf/Makefile.perf | 5 +-
tools/perf/arch/x86/util/Build | 1 +
tools/perf/arch/x86/util/iostat.c | 470 +++++++++++++++++++++++
tools/perf/builtin-stat.c | 21 +-
tools/perf/command-list.txt | 1 +
tools/perf/perf-iostat.sh | 12 +
tools/perf/util/Build | 1 +
tools/perf/util/iostat.c | 53 +++
tools/perf/util/iostat.h | 47 +++
tools/perf/util/stat-display.c | 40 +-
tools/perf/util/stat-shadow.c | 5 +-
tools/perf/util/stat.h | 1 +
14 files changed, 733 insertions(+), 13 deletions(-)
create mode 100644 tools/perf/Documentation/perf-iostat.txt
create mode 100644 tools/perf/arch/x86/util/iostat.c
create mode 100644 tools/perf/perf-iostat.sh
create mode 100644 tools/perf/util/iostat.c
create mode 100644 tools/perf/util/iostat.h
base-commit: 6859bc0e78c6a699599cbb21404fdb6c8125da74
--
2.19.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v5 1/4] perf stat: Basic support for iostat in perf
2021-03-24 14:30 [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Alexander Antonov
@ 2021-03-24 14:30 ` Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 2/4] perf stat: Helper functions for PCIe root ports list in iostat mode Alexander Antonov
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Alexander Antonov @ 2021-03-24 14:30 UTC (permalink / raw)
To: acme
Cc: linux-kernel, jolsa, ak, alexander.shishkin, mark.rutland,
namhyung, irogers, mingo, peterz, alexey.v.bayduraev,
alexander.antonov
Add basic flow for a new iostat mode in perf. Mode is intended to
provide four I/O performance metrics per each PCIe root port: Inbound Read,
Inbound Write, Outbound Read, Outbound Write.
The actual code to compute the metrics and attribute it to
root port is in follow-on patches.
Signed-off-by: Alexander Antonov <alexander.antonov@linux.intel.com>
---
tools/perf/builtin-stat.c | 21 +++++++++++++-
tools/perf/util/Build | 1 +
tools/perf/util/iostat.c | 53 ++++++++++++++++++++++++++++++++++
tools/perf/util/iostat.h | 47 ++++++++++++++++++++++++++++++
tools/perf/util/stat-display.c | 40 ++++++++++++++++++-------
tools/perf/util/stat-shadow.c | 5 +++-
tools/perf/util/stat.h | 1 +
7 files changed, 156 insertions(+), 12 deletions(-)
create mode 100644 tools/perf/util/iostat.c
create mode 100644 tools/perf/util/iostat.h
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 2e2e4a8345ea..4cef64ce9261 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -68,6 +68,7 @@
#include "util/affinity.h"
#include "util/pfm.h"
#include "util/bpf_counter.h"
+#include "util/iostat.h"
#include "asm/bug.h"
#include <linux/time64.h>
@@ -212,7 +213,8 @@ static struct perf_stat_config stat_config = {
.walltime_nsecs_stats = &walltime_nsecs_stats,
.big_num = true,
.ctl_fd = -1,
- .ctl_fd_ack = -1
+ .ctl_fd_ack = -1,
+ .iostat_run = false,
};
static bool cpus_map_matched(struct evsel *a, struct evsel *b)
@@ -1247,6 +1249,9 @@ static struct option stat_options[] = {
"\t\t\t Optionally send control command completion ('ack\\n') to ack-fd descriptor.\n"
"\t\t\t Alternatively, ctl-fifo / ack-fifo will be opened and used as ctl-fd / ack-fd.",
parse_control_option),
+ OPT_CALLBACK_OPTARG(0, "iostat", &evsel_list, &stat_config, "default",
+ "measure I/O performance metrics provided by arch/platform",
+ iostat_parse),
OPT_END()
};
@@ -2320,6 +2325,17 @@ int cmd_stat(int argc, const char **argv)
goto out;
}
+ if (stat_config.iostat_run) {
+ status = iostat_prepare(evsel_list, &stat_config);
+ if (status)
+ goto out;
+ if (iostat_mode == IOSTAT_LIST) {
+ iostat_list(evsel_list, &stat_config);
+ goto out;
+ } else if (verbose)
+ iostat_list(evsel_list, &stat_config);
+ }
+
if (add_default_attributes())
goto out;
@@ -2495,6 +2511,9 @@ int cmd_stat(int argc, const char **argv)
perf_stat__exit_aggr_mode();
evlist__free_stats(evsel_list);
out:
+ if (stat_config.iostat_run)
+ iostat_release(evsel_list);
+
zfree(&stat_config.walltime_run);
if (smi_cost && smi_reset)
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index e3e12f9d4733..7dd815712d60 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -102,6 +102,7 @@ perf-y += rwsem.o
perf-y += thread-stack.o
perf-y += spark.o
perf-y += topdown.o
+perf-y += iostat.o
perf-y += stream.o
perf-$(CONFIG_AUXTRACE) += auxtrace.o
perf-$(CONFIG_AUXTRACE) += intel-pt-decoder/
diff --git a/tools/perf/util/iostat.c b/tools/perf/util/iostat.c
new file mode 100644
index 000000000000..57dd49da28fe
--- /dev/null
+++ b/tools/perf/util/iostat.c
@@ -0,0 +1,53 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "util/iostat.h"
+#include "util/debug.h"
+
+enum iostat_mode_t iostat_mode = IOSTAT_NONE;
+
+__weak int iostat_prepare(struct evlist *evlist __maybe_unused,
+ struct perf_stat_config *config __maybe_unused)
+{
+ return -1;
+}
+
+__weak int iostat_parse(const struct option *opt __maybe_unused,
+ const char *str __maybe_unused,
+ int unset __maybe_unused)
+{
+ pr_err("iostat mode is not supported on current platform\n");
+ return -1;
+}
+
+__weak void iostat_list(struct evlist *evlist __maybe_unused,
+ struct perf_stat_config *config __maybe_unused)
+{
+}
+
+__weak void iostat_release(struct evlist *evlist __maybe_unused)
+{
+}
+
+__weak void iostat_print_header_prefix(struct perf_stat_config *config __maybe_unused)
+{
+}
+
+__weak void iostat_print_metric(struct perf_stat_config *config __maybe_unused,
+ struct evsel *evsel __maybe_unused,
+ struct perf_stat_output_ctx *out __maybe_unused)
+{
+}
+
+__weak void iostat_prefix(struct evlist *evlist __maybe_unused,
+ struct perf_stat_config *config __maybe_unused,
+ char *prefix __maybe_unused,
+ struct timespec *ts __maybe_unused)
+{
+}
+
+__weak void iostat_print_counters(struct evlist *evlist __maybe_unused,
+ struct perf_stat_config *config __maybe_unused,
+ struct timespec *ts __maybe_unused,
+ char *prefix __maybe_unused,
+ iostat_print_counter_t print_cnt_cb __maybe_unused)
+{
+}
diff --git a/tools/perf/util/iostat.h b/tools/perf/util/iostat.h
new file mode 100644
index 000000000000..23c1c46a331a
--- /dev/null
+++ b/tools/perf/util/iostat.h
@@ -0,0 +1,47 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * perf iostat
+ *
+ * Copyright (C) 2020, Intel Corporation
+ *
+ * Authors: Alexander Antonov <alexander.antonov@linux.intel.com>
+ */
+
+#ifndef _IOSTAT_H
+#define _IOSTAT_H
+
+#include <subcmd/parse-options.h>
+#include "util/stat.h"
+#include "util/parse-events.h"
+#include "util/evlist.h"
+
+struct option;
+struct perf_stat_config;
+struct evlist;
+struct timespec;
+
+enum iostat_mode_t {
+ IOSTAT_NONE = -1,
+ IOSTAT_RUN = 0,
+ IOSTAT_LIST = 1
+};
+
+extern enum iostat_mode_t iostat_mode;
+
+typedef void (*iostat_print_counter_t)(struct perf_stat_config *, struct evsel *, char *);
+
+int iostat_prepare(struct evlist *evlist, struct perf_stat_config *config);
+int iostat_parse(const struct option *opt, const char *str,
+ int unset __maybe_unused);
+void iostat_list(struct evlist *evlist, struct perf_stat_config *config);
+void iostat_release(struct evlist *evlist);
+void iostat_prefix(struct evlist *evlist, struct perf_stat_config *config,
+ char *prefix, struct timespec *ts);
+void iostat_print_header_prefix(struct perf_stat_config *config);
+void iostat_print_metric(struct perf_stat_config *config, struct evsel *evsel,
+ struct perf_stat_output_ctx *out);
+void iostat_print_counters(struct evlist *evlist,
+ struct perf_stat_config *config, struct timespec *ts,
+ char *prefix, iostat_print_counter_t print_cnt_cb);
+
+#endif /* _IOSTAT_H */
diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index 7f09cdaf5b60..aabe78dc6836 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -17,6 +17,7 @@
#include "cgroup.h"
#include <api/fs/fs.h>
#include "util.h"
+#include "iostat.h"
#define CNTR_NOT_SUPPORTED "<not supported>"
#define CNTR_NOT_COUNTED "<not counted>"
@@ -310,6 +311,11 @@ static void print_metric_header(struct perf_stat_config *config,
struct outstate *os = ctx;
char tbuf[1024];
+ /* In case of iostat, print metric header for first root port only */
+ if (config->iostat_run &&
+ os->evsel->priv != os->evsel->evlist->selected->priv)
+ return;
+
if (!valid_only_metric(unit))
return;
unit = fixunit(tbuf, os->evsel, unit);
@@ -952,8 +958,11 @@ static void print_metric_headers(struct perf_stat_config *config,
if (config->csv_output) {
if (config->interval)
fputs("time,", config->output);
- fputs(aggr_header_csv[config->aggr_mode], config->output);
+ if (!config->iostat_run)
+ fputs(aggr_header_csv[config->aggr_mode], config->output);
}
+ if (config->iostat_run)
+ iostat_print_header_prefix(config);
/* Print metrics headers only */
evlist__for_each_entry(evlist, counter) {
@@ -983,7 +992,8 @@ static void print_interval(struct perf_stat_config *config,
if (config->interval_clear)
puts(CONSOLE_CLEAR);
- sprintf(prefix, "%6lu.%09lu%s", (unsigned long) ts->tv_sec, ts->tv_nsec, config->csv_sep);
+ if (!config->iostat_run)
+ sprintf(prefix, "%6lu.%09lu%s", (unsigned long) ts->tv_sec, ts->tv_nsec, config->csv_sep);
if ((num_print_interval == 0 && !config->csv_output) || config->interval_clear) {
switch (config->aggr_mode) {
@@ -1019,9 +1029,11 @@ static void print_interval(struct perf_stat_config *config,
break;
case AGGR_GLOBAL:
default:
- fprintf(output, "# time");
- if (!metric_only)
- fprintf(output, " counts %*s events\n", unit_width, "unit");
+ if (!config->iostat_run) {
+ fprintf(output, "# time");
+ if (!metric_only)
+ fprintf(output, " counts %*s events\n", unit_width, "unit");
+ }
case AGGR_UNSET:
break;
}
@@ -1214,6 +1226,9 @@ void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *conf
struct evsel *counter;
char buf[64], *prefix = NULL;
+ if (config->iostat_run)
+ evlist->selected = evlist__first(evlist);
+
if (interval)
print_interval(config, evlist, prefix = buf, ts);
else
@@ -1226,7 +1241,7 @@ void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *conf
print_metric_headers(config, evlist, prefix, false);
if (num_print_iv++ == 25)
num_print_iv = 0;
- if (config->aggr_mode == AGGR_GLOBAL && prefix)
+ if (config->aggr_mode == AGGR_GLOBAL && prefix && !config->iostat_run)
fprintf(config->output, "%s", prefix);
}
@@ -1243,11 +1258,16 @@ void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *conf
}
break;
case AGGR_GLOBAL:
- evlist__for_each_entry(evlist, counter) {
- print_counter_aggr(config, counter, prefix);
+ if (config->iostat_run)
+ iostat_print_counters(evlist, config, ts, prefix = buf,
+ print_counter_aggr);
+ else {
+ evlist__for_each_entry(evlist, counter) {
+ print_counter_aggr(config, counter, prefix);
+ }
+ if (metric_only)
+ fputc('\n', config->output);
}
- if (metric_only)
- fputc('\n', config->output);
break;
case AGGR_NONE:
if (metric_only)
diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
index 3f800e71126f..39967a45f55b 100644
--- a/tools/perf/util/stat-shadow.c
+++ b/tools/perf/util/stat-shadow.c
@@ -11,6 +11,7 @@
#include "cgroup.h"
#include "units.h"
#include <linux/zalloc.h>
+#include "iostat.h"
/*
* AGGR_GLOBAL: Use CPU 0
@@ -962,7 +963,9 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
struct metric_event *me;
int num = 1;
- if (evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
+ if (config->iostat_run) {
+ iostat_print_metric(config, evsel, out);
+ } else if (evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
total = runtime_stat_avg(st, STAT_CYCLES, cpu, &rsd);
if (total) {
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 41107b8deac5..96ed00eb609e 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -132,6 +132,7 @@ struct perf_stat_config {
bool metric_no_merge;
bool stop_read_counter;
bool quiet;
+ bool iostat_run;
FILE *output;
unsigned int interval;
unsigned int timeout;
--
2.19.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v5 2/4] perf stat: Helper functions for PCIe root ports list in iostat mode
2021-03-24 14:30 [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 1/4] perf stat: Basic support for iostat in perf Alexander Antonov
@ 2021-03-24 14:30 ` Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 3/4] perf stat: Enable iostat mode for x86 platforms Alexander Antonov
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Alexander Antonov @ 2021-03-24 14:30 UTC (permalink / raw)
To: acme
Cc: linux-kernel, jolsa, ak, alexander.shishkin, mark.rutland,
namhyung, irogers, mingo, peterz, alexey.v.bayduraev,
alexander.antonov
Introduce helper functions to control PCIe root ports list.
These helpers will be used in the follow-up patch.
Signed-off-by: Alexander Antonov <alexander.antonov@linux.intel.com>
---
tools/perf/arch/x86/util/iostat.c | 110 ++++++++++++++++++++++++++++++
1 file changed, 110 insertions(+)
create mode 100644 tools/perf/arch/x86/util/iostat.c
diff --git a/tools/perf/arch/x86/util/iostat.c b/tools/perf/arch/x86/util/iostat.c
new file mode 100644
index 000000000000..c4471f8efa5e
--- /dev/null
+++ b/tools/perf/arch/x86/util/iostat.c
@@ -0,0 +1,110 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * perf iostat
+ *
+ * Copyright (C) 2020, Intel Corporation
+ *
+ * Authors: Alexander Antonov <alexander.antonov@linux.intel.com>
+ */
+
+#include <api/fs/fs.h>
+#include <linux/kernel.h>
+#include <linux/err.h>
+#include <limits.h>
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <dirent.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <regex.h>
+#include "util/cpumap.h"
+#include "util/debug.h"
+#include "util/iostat.h"
+#include "util/counts.h"
+#include "path.h"
+
+struct iio_root_port {
+ u32 domain;
+ u8 bus;
+ u8 die;
+ u8 pmu_idx;
+ int idx;
+};
+
+struct iio_root_ports_list {
+ struct iio_root_port **rps;
+ int nr_entries;
+};
+
+static void iio_root_port_show(FILE *output,
+ const struct iio_root_port * const rp)
+{
+ if (output && rp)
+ fprintf(output, "S%d-uncore_iio_%d<%04x:%02x>\n",
+ rp->die, rp->pmu_idx, rp->domain, rp->bus);
+}
+
+static struct iio_root_port *iio_root_port_new(u32 domain, u8 bus,
+ u8 die, u8 pmu_idx)
+{
+ struct iio_root_port *p = calloc(1, sizeof(*p));
+
+ if (p) {
+ p->domain = domain;
+ p->bus = bus;
+ p->die = die;
+ p->pmu_idx = pmu_idx;
+ }
+ return p;
+}
+
+static void iio_root_ports_list_free(struct iio_root_ports_list *list)
+{
+ int idx;
+
+ if (list) {
+ for (idx = 0; idx < list->nr_entries; idx++)
+ free(list->rps[idx]);
+ free(list->rps);
+ free(list);
+ }
+}
+
+static struct iio_root_port *iio_root_port_find_by_notation(
+ const struct iio_root_ports_list * const list, u32 domain, u8 bus)
+{
+ int idx;
+ struct iio_root_port *rp;
+
+ if (list) {
+ for (idx = 0; idx < list->nr_entries; idx++) {
+ rp = list->rps[idx];
+ if (rp && rp->domain == domain && rp->bus == bus)
+ return rp;
+ }
+ }
+ return NULL;
+}
+
+static int iio_root_ports_list_insert(struct iio_root_ports_list *list,
+ struct iio_root_port * const rp)
+{
+ struct iio_root_port **tmp_buf;
+
+ if (list && rp) {
+ rp->idx = list->nr_entries++;
+ tmp_buf = realloc(list->rps,
+ list->nr_entries * sizeof(*list->rps));
+ if (!tmp_buf) {
+ pr_err("Failed to realloc memory\n");
+ return -ENOMEM;
+ }
+ tmp_buf[rp->idx] = rp;
+ list->rps = tmp_buf;
+ }
+ return 0;
+}
--
2.19.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v5 3/4] perf stat: Enable iostat mode for x86 platforms
2021-03-24 14:30 [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 1/4] perf stat: Basic support for iostat in perf Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 2/4] perf stat: Helper functions for PCIe root ports list in iostat mode Alexander Antonov
@ 2021-03-24 14:30 ` Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 4/4] perf: Update .gitignore file Alexander Antonov
2021-03-30 7:48 ` [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Namhyung Kim
4 siblings, 0 replies; 6+ messages in thread
From: Alexander Antonov @ 2021-03-24 14:30 UTC (permalink / raw)
To: acme
Cc: linux-kernel, jolsa, ak, alexander.shishkin, mark.rutland,
namhyung, irogers, mingo, peterz, alexey.v.bayduraev,
alexander.antonov
This functionality is based on recently introduced sysfs attributes
for Intel® Xeon® Scalable processor family (code name Skylake-SP):
Commit bb42b3d39781 ("perf/x86/intel/uncore: Expose an Uncore unit to
IIO PMON mapping")
Mode is intended to provide four I/O performance metrics in MB per each
PCIe root port:
- Inbound Read: I/O devices below root port read from the host memory
- Inbound Write: I/O devices below root port write to the host memory
- Outbound Read: CPU reads from I/O devices below root port
- Outbound Write: CPU writes to I/O devices below root port
Each metric requiries only one uncore event which increments at every 4B
transfer in corresponding direction. The formulas to compute metrics
are generic:
#EventCount * 4B / (1024 * 1024)
Signed-off-by: Alexander Antonov <alexander.antonov@linux.intel.com>
---
tools/perf/Documentation/perf-iostat.txt | 88 ++++++
tools/perf/Makefile.perf | 5 +-
tools/perf/arch/x86/util/Build | 1 +
tools/perf/arch/x86/util/iostat.c | 360 +++++++++++++++++++++++
tools/perf/command-list.txt | 1 +
tools/perf/perf-iostat.sh | 12 +
6 files changed, 466 insertions(+), 1 deletion(-)
create mode 100644 tools/perf/Documentation/perf-iostat.txt
create mode 100644 tools/perf/perf-iostat.sh
diff --git a/tools/perf/Documentation/perf-iostat.txt b/tools/perf/Documentation/perf-iostat.txt
new file mode 100644
index 000000000000..165176944031
--- /dev/null
+++ b/tools/perf/Documentation/perf-iostat.txt
@@ -0,0 +1,88 @@
+perf-iostat(1)
+===============
+
+NAME
+----
+perf-iostat - Show I/O performance metrics
+
+SYNOPSIS
+--------
+[verse]
+'perf iostat' list
+'perf iostat' <ports> -- <command> [<options>]
+
+DESCRIPTION
+-----------
+Mode is intended to provide four I/O performance metrics per each PCIe root port:
+
+- Inbound Read - I/O devices below root port read from the host memory, in MB
+
+- Inbound Write - I/O devices below root port write to the host memory, in MB
+
+- Outbound Read - CPU reads from I/O devices below root port, in MB
+
+- Outbound Write - CPU writes to I/O devices below root port, in MB
+
+OPTIONS
+-------
+<command>...::
+ Any command you can specify in a shell.
+
+list::
+ List all PCIe root ports.
+
+<ports>::
+ Select the root ports for monitoring. Comma-separated list is supported.
+
+EXAMPLES
+--------
+
+1. List all PCIe root ports (example for 2-S platform):
+
+ $ perf iostat list
+ S0-uncore_iio_0<0000:00>
+ S1-uncore_iio_0<0000:80>
+ S0-uncore_iio_1<0000:17>
+ S1-uncore_iio_1<0000:85>
+ S0-uncore_iio_2<0000:3a>
+ S1-uncore_iio_2<0000:ae>
+ S0-uncore_iio_3<0000:5d>
+ S1-uncore_iio_3<0000:d7>
+
+2. Collect metrics for all PCIe root ports:
+
+ $ perf iostat -- dd if=/dev/zero of=/dev/nvme0n1 bs=1M oflag=direct
+ 357708+0 records in
+ 357707+0 records out
+ 375083606016 bytes (375 GB, 349 GiB) copied, 215.974 s, 1.7 GB/s
+
+ Performance counter stats for 'system wide':
+
+ port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound Write(MB)
+ 0000:00 1 0 2 3
+ 0000:80 0 0 0 0
+ 0000:17 352552 43 0 21
+ 0000:85 0 0 0 0
+ 0000:3a 3 0 0 0
+ 0000:ae 0 0 0 0
+ 0000:5d 0 0 0 0
+ 0000:d7 0 0 0 0
+
+3. Collect metrics for comma-separated list of PCIe root ports:
+
+ $ perf iostat 0000:17,0:3a -- dd if=/dev/zero of=/dev/nvme0n1 bs=1M oflag=direct
+ 357708+0 records in
+ 357707+0 records out
+ 375083606016 bytes (375 GB, 349 GiB) copied, 197.08 s, 1.9 GB/s
+
+ Performance counter stats for 'system wide':
+
+ port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound Write(MB)
+ 0000:17 358559 44 0 22
+ 0000:3a 3 2 0 0
+
+ 197.081983474 seconds time elapsed
+
+SEE ALSO
+--------
+linkperf:perf-stat[1]
\ No newline at end of file
diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
index a7d768fdc8a1..3b3a452f4862 100644
--- a/tools/perf/Makefile.perf
+++ b/tools/perf/Makefile.perf
@@ -283,6 +283,7 @@ SCRIPT_SH =
SCRIPT_SH += perf-archive.sh
SCRIPT_SH += perf-with-kcore.sh
+SCRIPT_SH += perf-iostat.sh
grep-libs = $(filter -l%,$(1))
strip-libs = $(filter-out -l%,$(1))
@@ -948,6 +949,8 @@ endif
$(INSTALL) $(OUTPUT)perf-archive -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
$(call QUIET_INSTALL, perf-with-kcore) \
$(INSTALL) $(OUTPUT)perf-with-kcore -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
+ $(call QUIET_INSTALL, perf-iostat) \
+ $(INSTALL) $(OUTPUT)perf-iostat -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)'
ifndef NO_LIBAUDIT
$(call QUIET_INSTALL, strace/groups) \
$(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(STRACE_GROUPS_INSTDIR_SQ)'; \
@@ -1041,7 +1044,7 @@ bpf-skel-clean:
$(call QUIET_CLEAN, bpf-skel) $(RM) -r $(SKEL_TMP_OUT) $(SKELETONS)
clean:: $(LIBTRACEEVENT)-clean $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBPERF)-clean fixdep-clean python-clean bpf-skel-clean
- $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(LANG_BINDINGS)
+ $(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf-iostat $(LANG_BINDINGS)
$(Q)find $(if $(OUTPUT),$(OUTPUT),.) -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete
$(Q)$(RM) $(OUTPUT).config-detected
$(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf perf-read-vdso32 perf-read-vdsox32 $(OUTPUT)pmu-events/jevents $(OUTPUT)$(LIBJVMTI).so
diff --git a/tools/perf/arch/x86/util/Build b/tools/perf/arch/x86/util/Build
index 0c72d418932e..dbeb04cb336e 100644
--- a/tools/perf/arch/x86/util/Build
+++ b/tools/perf/arch/x86/util/Build
@@ -9,6 +9,7 @@ perf-y += event.o
perf-y += evlist.o
perf-y += mem-events.o
perf-y += evsel.o
+perf-y += iostat.o
perf-$(CONFIG_DWARF) += dwarf-regs.o
perf-$(CONFIG_BPF_PROLOGUE) += dwarf-regs.o
diff --git a/tools/perf/arch/x86/util/iostat.c b/tools/perf/arch/x86/util/iostat.c
index c4471f8efa5e..d63acb782b63 100644
--- a/tools/perf/arch/x86/util/iostat.c
+++ b/tools/perf/arch/x86/util/iostat.c
@@ -27,6 +27,36 @@
#include "util/counts.h"
#include "path.h"
+#ifndef MAX_PATH
+#define MAX_PATH 1024
+#endif
+
+#define UNCORE_IIO_PMU_PATH "devices/uncore_iio_%d"
+#define SYSFS_UNCORE_PMU_PATH "%s/"UNCORE_IIO_PMU_PATH
+#define PLATFORM_MAPPING_PATH UNCORE_IIO_PMU_PATH"/die%d"
+
+/*
+ * Each metric requiries one IIO event which increments at every 4B transfer
+ * in corresponding direction. The formulas to compute metrics are generic:
+ * #EventCount * 4B / (1024 * 1024)
+ */
+static const char * const iostat_metrics[] = {
+ "Inbound Read(MB)",
+ "Inbound Write(MB)",
+ "Outbound Read(MB)",
+ "Outbound Write(MB)",
+};
+
+static inline int iostat_metrics_count(void)
+{
+ return sizeof(iostat_metrics) / sizeof(char *);
+}
+
+static const char *iostat_metric_by_idx(int idx)
+{
+ return *(iostat_metrics + idx % iostat_metrics_count());
+}
+
struct iio_root_port {
u32 domain;
u8 bus;
@@ -40,6 +70,8 @@ struct iio_root_ports_list {
int nr_entries;
};
+static struct iio_root_ports_list *root_ports;
+
static void iio_root_port_show(FILE *output,
const struct iio_root_port * const rp)
{
@@ -108,3 +140,331 @@ static int iio_root_ports_list_insert(struct iio_root_ports_list *list,
}
return 0;
}
+
+static int iio_mapping(u8 pmu_idx, struct iio_root_ports_list * const list)
+{
+ char *buf;
+ char path[MAX_PATH];
+ u32 domain;
+ u8 bus;
+ struct iio_root_port *rp;
+ size_t size;
+ int ret;
+
+ for (int die = 0; die < cpu__max_node(); die++) {
+ scnprintf(path, MAX_PATH, PLATFORM_MAPPING_PATH, pmu_idx, die);
+ if (sysfs__read_str(path, &buf, &size) < 0) {
+ if (pmu_idx)
+ goto out;
+ pr_err("Mode iostat is not supported\n");
+ return -1;
+ }
+ ret = sscanf(buf, "%04x:%02hhx", &domain, &bus);
+ free(buf);
+ if (ret != 2) {
+ pr_err("Invalid mapping data: iio_%d; die%d\n",
+ pmu_idx, die);
+ return -1;
+ }
+ rp = iio_root_port_new(domain, bus, die, pmu_idx);
+ if (!rp || iio_root_ports_list_insert(list, rp)) {
+ free(rp);
+ return -ENOMEM;
+ }
+ }
+out:
+ return 0;
+}
+
+static u8 iio_pmu_count(void)
+{
+ u8 pmu_idx = 0;
+ char path[MAX_PATH];
+ const char *sysfs = sysfs__mountpoint();
+
+ if (sysfs) {
+ for (;; pmu_idx++) {
+ snprintf(path, sizeof(path), SYSFS_UNCORE_PMU_PATH,
+ sysfs, pmu_idx);
+ if (access(path, F_OK) != 0)
+ break;
+ }
+ }
+ return pmu_idx;
+}
+
+static int iio_root_ports_scan(struct iio_root_ports_list **list)
+{
+ int ret = -ENOMEM;
+ struct iio_root_ports_list *tmp_list;
+ u8 pmu_count = iio_pmu_count();
+
+ if (!pmu_count) {
+ pr_err("Unsupported uncore pmu configuration\n");
+ return -1;
+ }
+
+ tmp_list = calloc(1, sizeof(*tmp_list));
+ if (!tmp_list)
+ goto err;
+
+ for (u8 pmu_idx = 0; pmu_idx < pmu_count; pmu_idx++) {
+ ret = iio_mapping(pmu_idx, tmp_list);
+ if (ret)
+ break;
+ }
+err:
+ if (!ret)
+ *list = tmp_list;
+ else
+ iio_root_ports_list_free(tmp_list);
+
+ return ret;
+}
+
+static int iio_root_port_parse_str(u32 *domain, u8 *bus, char *str)
+{
+ int ret;
+ regex_t regex;
+ /*
+ * Expected format domain:bus:
+ * Valid domain range [0:ffff]
+ * Valid bus range [0:ff]
+ * Example: 0000:af, 0:3d, 01:7
+ */
+ regcomp(®ex, "^([a-f0-9A-F]{1,}):([a-f0-9A-F]{1,2})", REG_EXTENDED);
+ ret = regexec(®ex, str, 0, NULL, 0);
+ if (ret || sscanf(str, "%08x:%02hhx", domain, bus) != 2)
+ pr_warning("Unrecognized root port format: %s\n"
+ "Please use the following format:\n"
+ "\t [domain]:[bus]\n"
+ "\t for example: 0000:3d\n", str);
+
+ regfree(®ex);
+ return ret;
+}
+
+static int iio_root_ports_list_filter(struct iio_root_ports_list **list,
+ const char *filter)
+{
+ char *tok, *tmp, *filter_copy = NULL;
+ struct iio_root_port *rp;
+ u32 domain;
+ u8 bus;
+ int ret = -ENOMEM;
+ struct iio_root_ports_list *tmp_list = calloc(1, sizeof(*tmp_list));
+
+ if (!tmp_list)
+ goto err;
+
+ filter_copy = strdup(filter);
+ if (!filter_copy)
+ goto err;
+
+ for (tok = strtok_r(filter_copy, ",", &tmp); tok;
+ tok = strtok_r(NULL, ",", &tmp)) {
+ if (!iio_root_port_parse_str(&domain, &bus, tok)) {
+ rp = iio_root_port_find_by_notation(*list, domain, bus);
+ if (rp) {
+ (*list)->rps[rp->idx] = NULL;
+ ret = iio_root_ports_list_insert(tmp_list, rp);
+ if (ret) {
+ free(rp);
+ goto err;
+ }
+ } else if (!iio_root_port_find_by_notation(tmp_list,
+ domain, bus))
+ pr_warning("Root port %04x:%02x were not found\n",
+ domain, bus);
+ }
+ }
+
+ if (tmp_list->nr_entries == 0) {
+ pr_err("Requested root ports were not found\n");
+ ret = -EINVAL;
+ }
+err:
+ iio_root_ports_list_free(*list);
+ if (ret)
+ iio_root_ports_list_free(tmp_list);
+ else
+ *list = tmp_list;
+
+ free(filter_copy);
+ return ret;
+}
+
+static int iostat_event_group(struct evlist *evl,
+ struct iio_root_ports_list *list)
+{
+ int ret;
+ int idx;
+ const char *iostat_cmd_template =
+ "{uncore_iio_%x/event=0x83,umask=0x04,ch_mask=0xF,fc_mask=0x07/,\
+ uncore_iio_%x/event=0x83,umask=0x01,ch_mask=0xF,fc_mask=0x07/,\
+ uncore_iio_%x/event=0xc0,umask=0x04,ch_mask=0xF,fc_mask=0x07/,\
+ uncore_iio_%x/event=0xc0,umask=0x01,ch_mask=0xF,fc_mask=0x07/}";
+ const int len_template = strlen(iostat_cmd_template) + 1;
+ struct evsel *evsel = NULL;
+ int metrics_count = iostat_metrics_count();
+ char *iostat_cmd = calloc(len_template, 1);
+
+ if (!iostat_cmd)
+ return -ENOMEM;
+
+ for (idx = 0; idx < list->nr_entries; idx++) {
+ sprintf(iostat_cmd, iostat_cmd_template,
+ list->rps[idx]->pmu_idx, list->rps[idx]->pmu_idx,
+ list->rps[idx]->pmu_idx, list->rps[idx]->pmu_idx);
+ ret = parse_events(evl, iostat_cmd, NULL);
+ if (ret)
+ goto err;
+ }
+
+ evlist__for_each_entry(evl, evsel) {
+ evsel->priv = list->rps[evsel->idx / metrics_count];
+ }
+ list->nr_entries = 0;
+err:
+ iio_root_ports_list_free(list);
+ free(iostat_cmd);
+ return ret;
+}
+
+int iostat_prepare(struct evlist *evlist, struct perf_stat_config *config)
+{
+ if (evlist->core.nr_entries > 0) {
+ pr_warning("The -e and -M options are not supported."
+ "All chosen events/metrics will be dropped\n");
+ evlist__delete(evlist);
+ evlist = evlist__new();
+ if (!evlist)
+ return -ENOMEM;
+ }
+
+ config->metric_only = true;
+ config->aggr_mode = AGGR_GLOBAL;
+
+ return iostat_event_group(evlist, root_ports);
+}
+
+int iostat_parse(const struct option *opt, const char *str,
+ int unset __maybe_unused)
+{
+ int ret;
+ struct perf_stat_config *config = (struct perf_stat_config *)opt->data;
+
+ ret = iio_root_ports_scan(&root_ports);
+ if (!ret) {
+ config->iostat_run = true;
+ if (!str)
+ iostat_mode = IOSTAT_RUN;
+ else if (!strcmp(str, "list"))
+ iostat_mode = IOSTAT_LIST;
+ else {
+ iostat_mode = IOSTAT_RUN;
+ ret = iio_root_ports_list_filter(&root_ports, str);
+ }
+ }
+ return ret;
+}
+
+void iostat_list(struct evlist *evlist, struct perf_stat_config *config)
+{
+ struct evsel *evsel;
+ struct iio_root_port *rp = NULL;
+
+ evlist__for_each_entry(evlist, evsel) {
+ if (rp != evsel->priv) {
+ rp = evsel->priv;
+ iio_root_port_show(config->output, rp);
+ }
+ }
+}
+
+void iostat_release(struct evlist *evlist)
+{
+ struct evsel *evsel;
+ struct iio_root_port *rp = NULL;
+
+ evlist__for_each_entry(evlist, evsel) {
+ if (rp != evsel->priv) {
+ rp = evsel->priv;
+ free(evsel->priv);
+ }
+ }
+}
+
+void iostat_prefix(struct evlist *evlist,
+ struct perf_stat_config *config,
+ char *prefix, struct timespec *ts)
+{
+ struct iio_root_port *rp = evlist->selected->priv;
+
+ if (rp) {
+ if (ts)
+ sprintf(prefix, "%6lu.%09lu%s%04x:%02x%s",
+ ts->tv_sec, ts->tv_nsec,
+ config->csv_sep, rp->domain, rp->bus,
+ config->csv_sep);
+ else
+ sprintf(prefix, "%04x:%02x%s", rp->domain, rp->bus,
+ config->csv_sep);
+ }
+}
+
+void iostat_print_header_prefix(struct perf_stat_config *config)
+{
+ if (config->csv_output)
+ fputs("port,", config->output);
+ else if (config->interval)
+ fprintf(config->output, "# time port ");
+ else
+ fprintf(config->output, " port ");
+}
+
+void iostat_print_metric(struct perf_stat_config *config, struct evsel *evsel,
+ struct perf_stat_output_ctx *out)
+{
+ double iostat_value = 0;
+ u64 prev_count_val = 0;
+ const char *iostat_metric = iostat_metric_by_idx(evsel->idx);
+ u8 die = ((struct iio_root_port *)evsel->priv)->die;
+ struct perf_counts_values *count = perf_counts(evsel->counts, die, 0);
+
+ if (count->run && count->ena) {
+ if (evsel->prev_raw_counts && !out->force_header) {
+ struct perf_counts_values *prev_count =
+ perf_counts(evsel->prev_raw_counts, die, 0);
+
+ prev_count_val = prev_count->val;
+ prev_count->val = count->val;
+ }
+ iostat_value = (count->val - prev_count_val) /
+ ((double) count->run / count->ena);
+ }
+ out->print_metric(config, out->ctx, NULL, "%8.0f", iostat_metric,
+ iostat_value / (256 * 1024));
+}
+
+void iostat_print_counters(struct evlist *evlist,
+ struct perf_stat_config *config, struct timespec *ts,
+ char *prefix, iostat_print_counter_t print_cnt_cb)
+{
+ void *perf_device = NULL;
+ struct evsel *counter = evlist__first(evlist);
+
+ evlist__set_selected(evlist, counter);
+ iostat_prefix(evlist, config, prefix, ts);
+ fprintf(config->output, "%s", prefix);
+ evlist__for_each_entry(evlist, counter) {
+ perf_device = evlist->selected->priv;
+ if (perf_device && perf_device != counter->priv) {
+ evlist__set_selected(evlist, counter);
+ iostat_prefix(evlist, config, prefix, ts);
+ fprintf(config->output, "\n%s", prefix);
+ }
+ print_cnt_cb(config, counter, prefix);
+ }
+ fputc('\n', config->output);
+}
diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt
index 825a12e8d694..4aa034aefa33 100644
--- a/tools/perf/command-list.txt
+++ b/tools/perf/command-list.txt
@@ -14,6 +14,7 @@ perf-config mainporcelain common
perf-evlist mainporcelain common
perf-ftrace mainporcelain common
perf-inject mainporcelain common
+perf-iostat mainporcelain common
perf-kallsyms mainporcelain common
perf-kmem mainporcelain common
perf-kvm mainporcelain common
diff --git a/tools/perf/perf-iostat.sh b/tools/perf/perf-iostat.sh
new file mode 100644
index 000000000000..e562f252d56f
--- /dev/null
+++ b/tools/perf/perf-iostat.sh
@@ -0,0 +1,12 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# perf iostat
+# Alexander Antonov <alexander.antonov@linux.intel.com>
+
+if [[ "$1" == "list" ]] || [[ "$1" =~ ([a-f0-9A-F]{1,}):([a-f0-9A-F]{1,2})(,)? ]]; then
+ DELIMITER="="
+else
+ DELIMITER=" "
+fi
+
+perf stat --iostat$DELIMITER$*
--
2.19.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v5 4/4] perf: Update .gitignore file
2021-03-24 14:30 [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Alexander Antonov
` (2 preceding siblings ...)
2021-03-24 14:30 ` [PATCH v5 3/4] perf stat: Enable iostat mode for x86 platforms Alexander Antonov
@ 2021-03-24 14:30 ` Alexander Antonov
2021-03-30 7:48 ` [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Namhyung Kim
4 siblings, 0 replies; 6+ messages in thread
From: Alexander Antonov @ 2021-03-24 14:30 UTC (permalink / raw)
To: acme
Cc: linux-kernel, jolsa, ak, alexander.shishkin, mark.rutland,
namhyung, irogers, mingo, peterz, alexey.v.bayduraev,
alexander.antonov
After a "make -C tools/perf", git reports the following untracked file:
perf-iostat
Add this generated file to perf's .gitignore file.
Signed-off-by: Alexander Antonov <alexander.antonov@linux.intel.com>
---
tools/perf/.gitignore | 1 +
1 file changed, 1 insertion(+)
diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore
index f3f84781fd74..e555e9729758 100644
--- a/tools/perf/.gitignore
+++ b/tools/perf/.gitignore
@@ -20,6 +20,7 @@ perf.data.old
output.svg
perf-archive
perf-with-kcore
+perf-iostat
tags
TAGS
cscope*
--
2.19.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics
2021-03-24 14:30 [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Alexander Antonov
` (3 preceding siblings ...)
2021-03-24 14:30 ` [PATCH v5 4/4] perf: Update .gitignore file Alexander Antonov
@ 2021-03-30 7:48 ` Namhyung Kim
4 siblings, 0 replies; 6+ messages in thread
From: Namhyung Kim @ 2021-03-30 7:48 UTC (permalink / raw)
To: Alexander Antonov
Cc: Arnaldo Carvalho de Melo, linux-kernel, Jiri Olsa, Andi Kleen,
Alexander Shishkin, Mark Rutland, Ian Rogers, Ingo Molnar,
Peter Zijlstra, Alexey Bayduraev
Hello,
On Wed, Mar 24, 2021 at 11:30 PM Alexander Antonov
<alexander.antonov@linux.intel.com> wrote:
>
> The previous version can be found at:
> v4: https://lkml.kernel.org/r/20210203135830.38568-1-alexander.antonov@linux.intel.com/
> Changes in this revision are:
> v4 -> v5:
> - Addressed comments from Namhyung Kim:
> 1. Removed AGGR_PCIE_PORT aggregation mode
> 2. Added iostat_prepare() function
> 3. Moved implementation specific fprintf() calls to separate x86-related function
> 4. Fixed code-related issues
> - Moved __weak iostat's functions to separate util/iostat.c file
>
> The previous version can be found at:
> v3: https://lkml.kernel.org/r/20210126080619.30275-1-alexander.antonov@linux.intel.com/
> Changes in this revision are:
> v3 -> v4:
> - Addressed comment from Namhyung Kim:
> 1. Removed NULL-termination of root ports list
>
> The previous version can be found at:
> v2: https://lkml.kernel.org/r/20201223130320.3930-1-alexander.antonov@linux.intel.com
>
> Changes in this revision are:
> v2 -> v3:
> - Addressed comments from Namhyung Kim:
> 1. Removed perf_device pointer from evsel structure. Use priv field instead
> 2. Renamed 'iiostat' to 'iostat'
> 3. Renamed 'show' mode to 'list' mode
> 4. Renamed iiostat_delete_root_ports() to iiostat_release() and
> iostat_show_root_ports() to iostat_list()
>
> The previous version can be found at:
> v1: https://lkml.kernel.org/r/20201210090340.14358-1-alexander.antonov@linux.intel.com
>
> Changes in this revision are:
> v1 -> v2:
> - Addressed comment from Arnaldo Carvalho de Melo:
> 1. Using 'perf iiostat' subcommand instead of 'perf stat --iiostat':
> - Added perf-iiostat.sh script to use short command
> - Updated manual pages to get help for 'perf iiostat'
> - Added 'perf-iiostat' to perf's gitignore file
>
> Mode is intended to provide four I/O performance metrics in MB per each
> root port:
> - Inbound Read: I/O devices below root port read from the host memory
> - Inbound Write: I/O devices below root port write to the host memory
> - Outbound Read: CPU reads from I/O devices below root port
> - Outbound Write: CPU writes to I/O devices below root port
>
> Each metric requiries only one uncore event which increments at every 4B
> transfer in corresponding direction. The formulas to compute metrics
> are generic:
> #EventCount * 4B / (1024 * 1024)
>
> Note: iostat introduces new perf data aggregation mode - per PCIe root port
> hence -e and -M options are not supported.
>
> Usage examples:
>
> 1. List all PCIe root ports (example for 2-S platform):
> $ perf iostat list
> S0-uncore_iio_0<0000:00>
> S1-uncore_iio_0<0000:80>
> S0-uncore_iio_1<0000:17>
> S1-uncore_iio_1<0000:85>
> S0-uncore_iio_2<0000:3a>
> S1-uncore_iio_2<0000:ae>
> S0-uncore_iio_3<0000:5d>
> S1-uncore_iio_3<0000:d7>
>
> 2. Collect metrics for all PCIe root ports:
> $ perf iostat -- dd if=/dev/zero of=/dev/nvme0n1 bs=1M oflag=direct
> 357708+0 records in
> 357707+0 records out
> 375083606016 bytes (375 GB, 349 GiB) copied, 215.974 s, 1.7 GB/s
>
> Performance counter stats for 'system wide':
>
> port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound Write(MB)
> 0000:00 1 0 2 3
> 0000:80 0 0 0 0
> 0000:17 352552 43 0 21
> 0000:85 0 0 0 0
> 0000:3a 3 0 0 0
> 0000:ae 0 0 0 0
> 0000:5d 0 0 0 0
> 0000:d7 0 0 0 0
>
> 3. Collect metrics for comma separated list of PCIe root ports:
> $ perf iostat 0000:17,0:3a -- dd if=/dev/zero of=/dev/nvme0n1 bs=1M oflag=direct
> 357708+0 records in
> 357707+0 records out
> 375083606016 bytes (375 GB, 349 GiB) copied, 197.08 s, 1.9 GB/s
>
> Performance counter stats for 'system wide':
>
> port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound Write(MB)
> 0000:17 358559 44 0 22
> 0000:3a 3 2 0 0
>
> 197.081983474 seconds time elapsed
>
> Alexander Antonov (4):
> perf stat: Basic support for iostat in perf
> perf stat: Helper functions for PCIe root ports list in iostat mode
> perf stat: Enable iostat mode for x86 platforms
> perf: Update .gitignore file
Thanks for your work!
Acked-by: Namhyung Kim <namhyung@kernel.org>
Thanks,
Namhyung
>
> tools/perf/.gitignore | 1 +
> tools/perf/Documentation/perf-iostat.txt | 88 +++++
> tools/perf/Makefile.perf | 5 +-
> tools/perf/arch/x86/util/Build | 1 +
> tools/perf/arch/x86/util/iostat.c | 470 +++++++++++++++++++++++
> tools/perf/builtin-stat.c | 21 +-
> tools/perf/command-list.txt | 1 +
> tools/perf/perf-iostat.sh | 12 +
> tools/perf/util/Build | 1 +
> tools/perf/util/iostat.c | 53 +++
> tools/perf/util/iostat.h | 47 +++
> tools/perf/util/stat-display.c | 40 +-
> tools/perf/util/stat-shadow.c | 5 +-
> tools/perf/util/stat.h | 1 +
> 14 files changed, 733 insertions(+), 13 deletions(-)
> create mode 100644 tools/perf/Documentation/perf-iostat.txt
> create mode 100644 tools/perf/arch/x86/util/iostat.c
> create mode 100644 tools/perf/perf-iostat.sh
> create mode 100644 tools/perf/util/iostat.c
> create mode 100644 tools/perf/util/iostat.h
>
>
> base-commit: 6859bc0e78c6a699599cbb21404fdb6c8125da74
> --
> 2.19.1
>
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-03-30 7:49 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-24 14:30 [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 1/4] perf stat: Basic support for iostat in perf Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 2/4] perf stat: Helper functions for PCIe root ports list in iostat mode Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 3/4] perf stat: Enable iostat mode for x86 platforms Alexander Antonov
2021-03-24 14:30 ` [PATCH v5 4/4] perf: Update .gitignore file Alexander Antonov
2021-03-30 7:48 ` [PATCH v5 0/4] perf stat: Introduce iostat mode to provide I/O performance metrics Namhyung Kim
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).