| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20141216211427.GA23877@thinkpad>
Date: Tue, 16 Dec 2014 22:14:27 +0100
From: Stephane Eranian <eranian@...gle.com>
To: linux-kernel@...r.kernel.org
Cc: peterz@...radead.org, mingo@...e.hu, ak@...ux.intel.com,
jolsa@...hat.com, acme@...hat.com, dsahern@...il.com,
dzickus@...hat.com, rfowles@...hat.com, jmario@...hat.com,
namhyung@...nel.org
Subject: [PATCH v2] perf mem: enable simultaneous sampling of loads/stores
This patch modifies perf mem to default to smapling loads
and stores simultaneously. It could only do one or the other
before yet there was not hardware restriction preventing
simultaneous collection. With this patch, one run is sufficient
to collect both.
It is still possible to smaple only loads or stores by using the
-t option:
$ perf mem -t load rec
$ perf mem -t load rep
Or
$ perf mem -t store rec
$ perf mem -t store rep
The perf report TUI will show one event at a time. The store
output will contain a Weight column which will be empty.
In V2, we update the man page and simplify the initialization
of the argv vector to cmd_*().
Signed-off-by: Stephane Eranian <eranian@...gle.com>
---
tools/perf/Documentation/perf-mem.txt | 9 +--
tools/perf/builtin-mem.c | 123 +++++++++++++++++++++++++++-------
2 files changed, 105 insertions(+), 27 deletions(-)
diff --git a/tools/perf/Documentation/perf-mem.txt b/tools/perf/Documentation/perf-mem.txt
index 1d78a40..43310d8 100644
--- a/tools/perf/Documentation/perf-mem.txt
+++ b/tools/perf/Documentation/perf-mem.txt
@@ -12,11 +12,12 @@ SYNOPSIS
DESCRIPTION
-----------
-"perf mem -t <TYPE> record" runs a command and gathers memory operation data
+"perf mem record" runs a command and gathers memory operation data
from it, into perf.data. Perf record options are accepted and are passed through.
-"perf mem -t <TYPE> report" displays the result. It invokes perf report with the
-right set of options to display a memory access profile.
+"perf mem report" displays the result. It invokes perf report with the
+right set of options to display a memory access profile. By default, loads
+and stores are sampled. Use the -t option to limit to loads or stores.
Note that on Intel systems the memory latency reported is the use-latency,
not the pure load (or store latency). Use latency includes any pipeline
@@ -29,7 +30,7 @@ OPTIONS
-t::
--type=::
- Select the memory operation type: load or store (default: load)
+ Select the memory operation type: load or store (default: load,store)
-D::
--dump-raw-samples=::
diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c
index 24db6ff..1eded0a 100644
--- a/tools/perf/builtin-mem.c
+++ b/tools/perf/builtin-mem.c
@@ -7,10 +7,13 @@
#include "util/session.h"
#include "util/data.h"
-#define MEM_OPERATION_LOAD "load"
-#define MEM_OPERATION_STORE "store"
+#define MEM_OPERATION_LOAD 0x1
+#define MEM_OPERATION_STORE 0x2
-static const char *mem_operation = MEM_OPERATION_LOAD;
+/*
+ * default to both load an store sampling
+ */
+static int mem_operation = MEM_OPERATION_LOAD | MEM_OPERATION_STORE;
struct perf_mem {
struct perf_tool tool;
@@ -25,26 +28,30 @@ static int __cmd_record(int argc, const char **argv)
{
int rec_argc, i = 0, j;
const char **rec_argv;
- char event[64];
int ret;
- rec_argc = argc + 4;
+ rec_argc = argc + 7; /* max number of arguments */
rec_argv = calloc(rec_argc + 1, sizeof(char *));
if (!rec_argv)
return -1;
- rec_argv[i++] = strdup("record");
- if (!strcmp(mem_operation, MEM_OPERATION_LOAD))
- rec_argv[i++] = strdup("-W");
- rec_argv[i++] = strdup("-d");
- rec_argv[i++] = strdup("-e");
+ rec_argv[i++] = "record";
- if (strcmp(mem_operation, MEM_OPERATION_LOAD))
- sprintf(event, "cpu/mem-stores/pp");
- else
- sprintf(event, "cpu/mem-loads/pp");
+ if (mem_operation & MEM_OPERATION_LOAD)
+ rec_argv[i++] = "-W";
+
+ rec_argv[i++] = "-d";
+
+ if (mem_operation & MEM_OPERATION_LOAD) {
+ rec_argv[i++] = "-e";
+ rec_argv[i++] = "cpu/mem-loads/pp";
+ }
+
+ if (mem_operation & MEM_OPERATION_STORE) {
+ rec_argv[i++] = "-e";
+ rec_argv[i++] = "cpu/mem-stores/pp";
+ }
- rec_argv[i++] = strdup(event);
for (j = 1; j < argc; j++, i++)
rec_argv[i] = argv[j];
@@ -162,17 +169,17 @@ static int report_events(int argc, const char **argv, struct perf_mem *mem)
if (!rep_argv)
return -1;
- rep_argv[i++] = strdup("report");
- rep_argv[i++] = strdup("--mem-mode");
- rep_argv[i++] = strdup("-n"); /* display number of samples */
+ rep_argv[i++] = "report";
+ rep_argv[i++] = "--mem-mode";
+ rep_argv[i++] = "-n"; /* display number of samples */
/*
* there is no weight (cost) associated with stores, so don't print
* the column
*/
- if (strcmp(mem_operation, MEM_OPERATION_LOAD))
- rep_argv[i++] = strdup("--sort=mem,sym,dso,symbol_daddr,"
- "dso_daddr,tlb,locked");
+ if (!(mem_operation & MEM_OPERATION_LOAD))
+ rep_argv[i++] = "--sort=mem,sym,dso,symbol_daddr,"
+ "dso_daddr,tlb,locked";
for (j = 1; j < argc; j++, i++)
rep_argv[i] = argv[j];
@@ -182,6 +189,75 @@ static int report_events(int argc, const char **argv, struct perf_mem *mem)
return ret;
}
+struct mem_mode {
+ const char *name;
+ int mode;
+};
+
+#define MEM_OPT(n, m) \
+ { .name = n, .mode = (m) }
+
+#define MEM_END { .name = NULL }
+
+static const struct mem_mode mem_modes[]={
+ MEM_OPT("load", MEM_OPERATION_LOAD),
+ MEM_OPT("store", MEM_OPERATION_STORE),
+ MEM_END
+};
+
+static int
+parse_mem_ops(const struct option *opt, const char *str, int unset)
+{
+ int *mode = (int *)opt->value;
+ const struct mem_mode *m;
+ char *s, *os = NULL, *p;
+ int ret = -1;
+
+ if (unset)
+ return 0;
+
+ /* str may be NULL in case no arg is passed to -t */
+ if (str) {
+ /* because str is read-only */
+ s = os = strdup(str);
+ if (!s)
+ return -1;
+
+ /* reset mode */
+ *mode = 0;
+
+ for (;;) {
+ p = strchr(s, ',');
+ if (p)
+ *p = '\0';
+
+ for (m = mem_modes; m->name; m++) {
+ if (!strcasecmp(s, m->name))
+ break;
+ }
+ if (!m->name) {
+ fprintf(stderr, "unknown sampling op %s,"
+ " check man page\n", s);
+ goto error;
+ }
+
+ *mode |= m->mode;
+
+ if (!p)
+ break;
+
+ s = p + 1;
+ }
+ }
+ ret = 0;
+
+ if (*mode == 0)
+ *mode = MEM_OPERATION_LOAD;
+error:
+ free(os);
+ return ret;
+}
+
int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
{
struct stat st;
@@ -199,8 +275,9 @@ int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
.input_name = "perf.data",
};
const struct option mem_options[] = {
- OPT_STRING('t', "type", &mem_operation,
- "type", "memory operations(load/store)"),
+ OPT_CALLBACK('t', "type", &mem_operation,
+ "type", "memory operations(load,store) Default load,store",
+ parse_mem_ops),
OPT_BOOLEAN('D', "dump-raw-samples", &mem.dump_raw,
"dump raw samples in ASCII"),
OPT_BOOLEAN('U', "hide-unresolved", &mem.hide_unresolved,
--
1.9.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists