| Message ID | 1520345084-42646-2-git-send-email-agustinv@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Em Tue, Mar 06, 2018 at 09:04:42AM -0500, Agustin Vega-Frias escreveu: > Starting on v4.12 event parsing code for dynamic pmu events already > supports prefix-based matching of multiple pmus when creating dynamic > events. E.g., in a system with the following dynamic pmus: > > mypmu_0 > mypmu_1 > mypmu_2 > mypmu_4 > > passing mypmu/<config>/ as an event spec will result in the creation > of the event in all of the pmus. This change expands this matching > through the use of fnmatch so glob-like expressions can be used to > create events in multiple pmus. E.g., in the system described above > if a user only wants to create the event in mypmu_0 and mypmu_1, > mypmu_[01]/<config>/ can be passed. > > Change-Id: Icb25653fc5d5239c20f3bffdfdf4ab4c9c9bb20b > Signed-off-by: Agustin Vega-Frias <agustinv@codeaurora.org> > --- > tools/perf/Documentation/perf-list.txt | 8 +++++++- > tools/perf/Documentation/perf-stat.txt | 13 +++++++++++++ > tools/perf/util/parse-events.l | 2 +- > tools/perf/util/parse-events.y | 19 ++++++++++++------- > 4 files changed, 33 insertions(+), 9 deletions(-) > > diff --git a/tools/perf/Documentation/perf-list.txt b/tools/perf/Documentation/perf-list.txt > index e2a897a..2549c34 100644 > --- a/tools/perf/Documentation/perf-list.txt > +++ b/tools/perf/Documentation/perf-list.txt > @@ -141,7 +141,13 @@ on the first memory controller on socket 0 of a Intel Xeon system > > Each memory controller has its own PMU. Measuring the complete system > bandwidth would require specifying all imc PMUs (see perf list output), > -and adding the values together. > +and adding the values together. To simplify creation of multiple events, > +prefix and glob matching is supported in the PMU name, and the prefix > +'uncore_' is also ignored when performing the match. So the command above > +can be expanded to all memory controllers by using the syntaxes: > + > + perf stat -C 0 -a imc/cas_count_read/,imc/cas_count_write/ -I 1000 ... > + perf stat -C 0 -a *imc*/cas_count_read/,*imc*/cas_count_write/ -I 1000 ... > > This example measures the combined core power every second > > diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt > index 823fce7..5ee954f 100644 > --- a/tools/perf/Documentation/perf-stat.txt > +++ b/tools/perf/Documentation/perf-stat.txt > @@ -49,6 +49,13 @@ report:: > parameters are defined by corresponding entries in > /sys/bus/event_source/devices/<pmu>/format/* > > + Note that the last two syntaxes support prefix and glob matching in > + the PMU name to simplify creation of events accross multiple instances > + of the same type of PMU in large systems (e.g. memory controller PMUs). > + Multiple PMU instances are typical for uncore PMUs, so the prefix > + 'uncore_' is also ignored when performing this match. > + > + > -i:: > --no-inherit:: > child tasks do not inherit counters > @@ -246,6 +253,12 @@ taskset. > --no-merge:: > Do not merge results from same PMUs. > > +When multiple events are created from a single event alias, stat will, > +by default, aggregate the event counts and show the result in a single > +row. This option disables that behavior and shows the individual events > +and counts. Aliases are listed immediately after the Kernel PMU events > +by perf list. > + > --smi-cost:: > Measure SMI cost if msr/aperf/ and msr/smi/ events are supported. > > diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l > index 655ecff..a1a01b1 100644 > --- a/tools/perf/util/parse-events.l > +++ b/tools/perf/util/parse-events.l > @@ -175,7 +175,7 @@ bpf_source [^,{}]+\.c[a-zA-Z0-9._]* > num_dec [0-9]+ > num_hex 0x[a-fA-F0-9]+ > num_raw_hex [a-fA-F0-9]+ > -name [a-zA-Z_*?][a-zA-Z0-9_*?.]* > +name [a-zA-Z_*?\[\]][a-zA-Z0-9_*?.\[\]]* > name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?.:]* > drv_cfg_term [a-zA-Z0-9_\.]+(=[a-zA-Z0-9_*?\.:]+)? > /* If you add a modifier you need to update check_modifier() */ > diff --git a/tools/perf/util/parse-events.y b/tools/perf/util/parse-events.y > index e81a20e..316ac07 100644 > --- a/tools/perf/util/parse-events.y > +++ b/tools/perf/util/parse-events.y > @@ -8,6 +8,7 @@ > > #define YYDEBUG 1 > > +#include <fnmatch.h> > #include <linux/compiler.h> > #include <linux/list.h> > #include <linux/types.h> > @@ -234,21 +235,25 @@ PE_NAME opt_event_config > if (parse_events_add_pmu(_parse_state, list, $1, $2)) { > struct perf_pmu *pmu = NULL; > int ok = 0; > + char *pattern; > > - while ((pmu = perf_pmu__scan(pmu)) != NULL) { > - char *name = pmu->name; > + if (asprintf(&pattern, "?(uncore_)%s*", $1) < 0) > + YYABORT; > > - if (!strncmp(name, "uncore_", 7) && > - strncmp($1, "uncore_", 7)) > - name += 7; > - if (!strncmp($1, name, strlen($1))) { > - if (parse_events_copy_term_list(orig_terms, &terms)) > + while ((pmu = perf_pmu__scan(pmu)) != NULL) { > + if (!fnmatch(pattern, pmu->name, FNM_EXTMATCH)) { > + if (parse_events_copy_term_list(orig_terms, &terms)) { > + free(pattern); This breaks the build in at least Alpine Linux 3.4, 3.5 and 3.6 and android-ndk:r15c-arm Alpine 3.6 build failure: CC /tmp/build/perf/util/parse-events-bison.o util/parse-events.y: In function 'parse_events_parse': util/parse-events.y:244:37: error: 'FNM_EXTMATCH' undeclared (first use in this function) if (!fnmatch(pattern, pmu->name, FNM_EXTMATCH)) { ^~~~~~~~~~~~ util/parse-events.y:244:37: note: each undeclared identifier is reported only once for each function it appears in CC /tmp/build/perf/bench/mem-functions.o mv: can't rename '/tmp/build/perf/util/.parse-events-bison.o.tmp': No such file or directory make[4]: *** [/git/linux/tools/build/Makefile.build:96: /tmp/build/perf/util/parse-events-bison.o] Error 1 make[3]: *** [/git/linux/tools/build/Makefile.build:139: util] Error 2 make[2]: *** [Makefile.perf:623: /tmp/build/perf/libperf-in.o] Error 2 make[2]: *** Waiting for unfinished jobs.... Android NDK r15c ARM, cross built from, IIRC, Fedora 24: CC /tmp/build/perf/util/parse-events-bison.o util/parse-events.y: In function 'parse_events_parse': util/parse-events.y:244:37: error: 'FNM_EXTMATCH' undeclared (first use in this function) if (!fnmatch(pattern, pmu->name, FNM_EXTMATCH)) { ^ util/parse-events.y:244:37: note: each undeclared identifier is reported only once for each function it appears in mv: cannot stat '/tmp/build/perf/util/.parse-events-bison.o.tmp': No such file or directory make[4]: *** [/git/linux/tools/build/Makefile.build:96: /tmp/build/perf/util/parse-events-bison.o] Error 1 make[4]: *** Waiting for unfinished jobs.... CC /tmp/build/perf/builtin-kmem.o make[3]: *** [/git/linux/tools/build/Makefile.build:139: util] Error 2 make[2]: *** [Makefile.perf:623: /tmp/build/perf/libperf-in.o] Error 2 Probably others, I'll investigate later, feel free to do it if you have the time now :-) Extra info: 1 46.29 alpine:3.4 : FAIL gcc (Alpine 5.3.0) 5.3.0 2 53.37 alpine:3.5 : FAIL gcc (Alpine 6.2.1) 6.2.1 20160822 3 48.42 alpine:3.6 : FAIL gcc (Alpine 6.3.0) 6.3.0 4 35.88 alpine:edge : FAIL gcc (Alpine 6.4.0) 6.4.0 5 96.79 amazonlinux:1 : Ok gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-11) 6 116.40 amazonlinux:2 : Ok gcc (GCC) 7.2.1 20170915 (Red Hat 7.2.1-2) 7 34.30 android-ndk:r12b-arm : FAIL arm-linux-androideabi-gcc (GCC) 4.9.x 20150123 (prerelease) 8 33.69 android-ndk:r15c-arm : FAIL arm-linux-androideabi-gcc (GCC) 4.9.x 20150123 (prerelease) 9 60.50 centos:5 : Ok gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-55) 10 88.54 centos:6 : Ok gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-18) 11 108.89 centos:7 : Ok gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-16) The builds are still ongoing. - Arnaldo
On 2018-03-07 11:02, Arnaldo Carvalho de Melo wrote: > Em Tue, Mar 06, 2018 at 09:04:42AM -0500, Agustin Vega-Frias escreveu: [snip] > This breaks the build in at least Alpine Linux 3.4, 3.5 and 3.6 and > android-ndk:r15c-arm > > Alpine 3.6 build failure: > > CC /tmp/build/perf/util/parse-events-bison.o > util/parse-events.y: In function 'parse_events_parse': > util/parse-events.y:244:37: error: 'FNM_EXTMATCH' undeclared (first > use in this function) > if (!fnmatch(pattern, pmu->name, FNM_EXTMATCH)) { > ^~~~~~~~~~~~ > util/parse-events.y:244:37: note: each undeclared identifier is > reported only once for each function it appears in > CC /tmp/build/perf/bench/mem-functions.o > mv: can't rename '/tmp/build/perf/util/.parse-events-bison.o.tmp': No > such file or directory > make[4]: *** [/git/linux/tools/build/Makefile.build:96: > /tmp/build/perf/util/parse-events-bison.o] Error 1 > make[3]: *** [/git/linux/tools/build/Makefile.build:139: util] Error 2 > make[2]: *** [Makefile.perf:623: /tmp/build/perf/libperf-in.o] Error 2 > make[2]: *** Waiting for unfinished jobs.... > > Android NDK r15c ARM, cross built from, IIRC, Fedora 24: > > CC /tmp/build/perf/util/parse-events-bison.o > util/parse-events.y: In function 'parse_events_parse': > util/parse-events.y:244:37: error: 'FNM_EXTMATCH' undeclared (first > use in this function) > if (!fnmatch(pattern, pmu->name, FNM_EXTMATCH)) { > ^ > util/parse-events.y:244:37: note: each undeclared identifier is > reported only once for each function it appears in > mv: cannot stat '/tmp/build/perf/util/.parse-events-bison.o.tmp': No > such file or directory > make[4]: *** [/git/linux/tools/build/Makefile.build:96: > /tmp/build/perf/util/parse-events-bison.o] Error 1 > make[4]: *** Waiting for unfinished jobs.... > CC /tmp/build/perf/builtin-kmem.o > make[3]: *** [/git/linux/tools/build/Makefile.build:139: util] Error 2 > make[2]: *** [Makefile.perf:623: /tmp/build/perf/libperf-in.o] Error 2 > > > Probably others, I'll investigate later, feel free to do it if you have > the time now :-) > Sorry about that. That's probably because FNM_EXTMATCH is a GNU extension, not POSIX, and the Alpine and Android runtimes likely don't implement that... I'll send a fix reverting back to the strncmp to ignore the uncore_ prefix, and dropping that extension. AgustÃn > Extra info: > > > 1 46.29 alpine:3.4 : FAIL gcc (Alpine 5.3.0) > 5.3.0 > 2 53.37 alpine:3.5 : FAIL gcc (Alpine 6.2.1) > 6.2.1 20160822 > 3 48.42 alpine:3.6 : FAIL gcc (Alpine 6.3.0) > 6.3.0 > 4 35.88 alpine:edge : FAIL gcc (Alpine 6.4.0) > 6.4.0 > 5 96.79 amazonlinux:1 : Ok gcc (GCC) 4.8.5 > 20150623 (Red Hat 4.8.5-11) > 6 116.40 amazonlinux:2 : Ok gcc (GCC) 7.2.1 > 20170915 (Red Hat 7.2.1-2) > 7 34.30 android-ndk:r12b-arm : FAIL > arm-linux-androideabi-gcc (GCC) 4.9.x 20150123 (prerelease) > 8 33.69 android-ndk:r15c-arm : FAIL > arm-linux-androideabi-gcc (GCC) 4.9.x 20150123 (prerelease) > 9 60.50 centos:5 : Ok gcc (GCC) 4.1.2 > 20080704 (Red Hat 4.1.2-55) > 10 88.54 centos:6 : Ok gcc (GCC) 4.4.7 > 20120313 (Red Hat 4.4.7-18) > 11 108.89 centos:7 : Ok gcc (GCC) 4.8.5 > 20150623 (Red Hat 4.8.5-16) > > > The builds are still ongoing. > > - Arnaldo
> Sorry about that. That's probably because FNM_EXTMATCH is a GNU extension, > not POSIX, and the Alpine and Android runtimes likely don't implement > that... > I'll send a fix reverting back to the strncmp to ignore the uncore_ prefix, > and dropping that extension. Just don't set it? Even the basic glob patterns are useful. -Andi
Em Wed, Mar 07, 2018 at 10:54:15AM -0800, Andi Kleen escreveu: > > Sorry about that. That's probably because FNM_EXTMATCH is a GNU extension, > > not POSIX, and the Alpine and Android runtimes likely don't implement > > that... > > I'll send a fix reverting back to the strncmp to ignore the uncore_ prefix, > > and dropping that extension. > > Just don't set it? Even the basic glob patterns are useful. Or use: #ifndef FNM_EXTMATCH #define FNM_EXTMATCH 0 #endif So on systems without it, its not used, while on GNU systems, we have that functionality (pretty fancy, someone may need that... ;-)) - Arnaldo
Em Wed, Mar 07, 2018 at 04:05:09PM -0300, Arnaldo Carvalho de Melo escreveu: > Em Wed, Mar 07, 2018 at 10:54:15AM -0800, Andi Kleen escreveu: > > > Sorry about that. That's probably because FNM_EXTMATCH is a GNU extension, > > > not POSIX, and the Alpine and Android runtimes likely don't implement > > > that... > > > I'll send a fix reverting back to the strncmp to ignore the uncore_ prefix, > > > and dropping that extension. > > > > Just don't set it? Even the basic glob patterns are useful. > > Or use: > > #ifndef FNM_EXTMATCH > #define FNM_EXTMATCH 0 > #endif > > So on systems without it, its not used, while on GNU systems, we have > that functionality (pretty fancy, someone may need that... ;-)) I'll try this route, btw, no need to send more patches for now. - Arnaldo
diff --git a/tools/perf/Documentation/perf-list.txt b/tools/perf/Documentation/perf-list.txt index e2a897a..2549c34 100644 --- a/tools/perf/Documentation/perf-list.txt +++ b/tools/perf/Documentation/perf-list.txt @@ -141,7 +141,13 @@ on the first memory controller on socket 0 of a Intel Xeon system Each memory controller has its own PMU. Measuring the complete system bandwidth would require specifying all imc PMUs (see perf list output), -and adding the values together. +and adding the values together. To simplify creation of multiple events, +prefix and glob matching is supported in the PMU name, and the prefix +'uncore_' is also ignored when performing the match. So the command above +can be expanded to all memory controllers by using the syntaxes: + + perf stat -C 0 -a imc/cas_count_read/,imc/cas_count_write/ -I 1000 ... + perf stat -C 0 -a *imc*/cas_count_read/,*imc*/cas_count_write/ -I 1000 ... This example measures the combined core power every second diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt index 823fce7..5ee954f 100644 --- a/tools/perf/Documentation/perf-stat.txt +++ b/tools/perf/Documentation/perf-stat.txt @@ -49,6 +49,13 @@ report:: parameters are defined by corresponding entries in /sys/bus/event_source/devices/<pmu>/format/* + Note that the last two syntaxes support prefix and glob matching in + the PMU name to simplify creation of events accross multiple instances + of the same type of PMU in large systems (e.g. memory controller PMUs). + Multiple PMU instances are typical for uncore PMUs, so the prefix + 'uncore_' is also ignored when performing this match. + + -i:: --no-inherit:: child tasks do not inherit counters @@ -246,6 +253,12 @@ taskset. --no-merge:: Do not merge results from same PMUs. +When multiple events are created from a single event alias, stat will, +by default, aggregate the event counts and show the result in a single +row. This option disables that behavior and shows the individual events +and counts. Aliases are listed immediately after the Kernel PMU events +by perf list. + --smi-cost:: Measure SMI cost if msr/aperf/ and msr/smi/ events are supported. diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l index 655ecff..a1a01b1 100644 --- a/tools/perf/util/parse-events.l +++ b/tools/perf/util/parse-events.l @@ -175,7 +175,7 @@ bpf_source [^,{}]+\.c[a-zA-Z0-9._]* num_dec [0-9]+ num_hex 0x[a-fA-F0-9]+ num_raw_hex [a-fA-F0-9]+ -name [a-zA-Z_*?][a-zA-Z0-9_*?.]* +name [a-zA-Z_*?\[\]][a-zA-Z0-9_*?.\[\]]* name_minus [a-zA-Z_*?][a-zA-Z0-9\-_*?.:]* drv_cfg_term [a-zA-Z0-9_\.]+(=[a-zA-Z0-9_*?\.:]+)? /* If you add a modifier you need to update check_modifier() */ diff --git a/tools/perf/util/parse-events.y b/tools/perf/util/parse-events.y index e81a20e..316ac07 100644 --- a/tools/perf/util/parse-events.y +++ b/tools/perf/util/parse-events.y @@ -8,6 +8,7 @@ #define YYDEBUG 1 +#include <fnmatch.h> #include <linux/compiler.h> #include <linux/list.h> #include <linux/types.h> @@ -234,21 +235,25 @@ PE_NAME opt_event_config if (parse_events_add_pmu(_parse_state, list, $1, $2)) { struct perf_pmu *pmu = NULL; int ok = 0; + char *pattern; - while ((pmu = perf_pmu__scan(pmu)) != NULL) { - char *name = pmu->name; + if (asprintf(&pattern, "?(uncore_)%s*", $1) < 0) + YYABORT; - if (!strncmp(name, "uncore_", 7) && - strncmp($1, "uncore_", 7)) - name += 7; - if (!strncmp($1, name, strlen($1))) { - if (parse_events_copy_term_list(orig_terms, &terms)) + while ((pmu = perf_pmu__scan(pmu)) != NULL) { + if (!fnmatch(pattern, pmu->name, FNM_EXTMATCH)) { + if (parse_events_copy_term_list(orig_terms, &terms)) { + free(pattern); YYABORT; + } if (!parse_events_add_pmu(_parse_state, list, pmu->name, terms)) ok++; parse_events_terms__delete(terms); } } + + free(pattern); + if (!ok) YYABORT; }
Starting on v4.12 event parsing code for dynamic pmu events already supports prefix-based matching of multiple pmus when creating dynamic events. E.g., in a system with the following dynamic pmus: mypmu_0 mypmu_1 mypmu_2 mypmu_4 passing mypmu/<config>/ as an event spec will result in the creation of the event in all of the pmus. This change expands this matching through the use of fnmatch so glob-like expressions can be used to create events in multiple pmus. E.g., in the system described above if a user only wants to create the event in mypmu_0 and mypmu_1, mypmu_[01]/<config>/ can be passed. Change-Id: Icb25653fc5d5239c20f3bffdfdf4ab4c9c9bb20b Signed-off-by: Agustin Vega-Frias <agustinv@codeaurora.org> --- tools/perf/Documentation/perf-list.txt | 8 +++++++- tools/perf/Documentation/perf-stat.txt | 13 +++++++++++++ tools/perf/util/parse-events.l | 2 +- tools/perf/util/parse-events.y | 19 ++++++++++++------- 4 files changed, 33 insertions(+), 9 deletions(-)