Re: [PATCH v3 06/48] perf cpumap: Switch cpu_map__build_map to cpu function
From: Arnaldo Carvalho de Melo
Date: Tue Jan 11 2022 - 10:12:09 EST
Em Mon, Jan 10, 2022 at 09:41:12PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Mon, Jan 10, 2022 at 02:29:58PM -0800, Ian Rogers escreveu:
> > On Mon, Jan 10, 2022 at 1:34 PM Arnaldo Carvalho de Melo
> > <acme@xxxxxxxxxx> wrote:
> > >
> > > Em Mon, Jan 10, 2022 at 06:23:09PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > > Em Mon, Jan 10, 2022 at 06:03:04PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > > > Em Mon, Jan 10, 2022 at 05:46:52PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > > > > At this point we need to remove the now unused perf_env__get_cpu(), and
> > > > > > then deal with the fallout in the following patches that needlessly
> > > > > > modify this unused function.
> > > > > >
> > > > > > Doing that now.
> > > > >
> > > > > Done, force pushed to tmp.perf/perf_cpu, now it is going again thru the
> > > > > build containers, a better start now:
> > >
> > > > So far so good:
> > >
> > > perf test ok, just that one that is failing, but before this series as
> > > well:
> > >
> > > 89: perf all metricgroups test : FAILED!
> >
> > Could you add the verbose output? I suspect it is fixed by this:
> > https://lore.kernel.org/lkml/20211223183948.3423989-1-irogers@xxxxxxxxxx/
>
> I'll try it.
>
> BTW, here are the final tests, problems with arm 32 and 64-bit, I'll go thru
> them tomorrow, to fix in the cset the problem appears, so that we keep it
> bisectable, else you can get what I have in tmp.perf/perf_cpu, fix it and
> tomorrow I'll retest.
>
> 27 85.67 debian:11 : Ok gcc (Debian 10.2.1-6) 10.2.1 20210110 , Debian clang version 11.0.1-2
> 28 98.65 debian:experimental : Ok gcc (Debian 11.2.0-13) 11.2.0 , Debian clang version 13.0.0-9+b2
> 29 8.69 debian:experimental-x-arm64 : FAIL gcc version 11.2.0 (Debian 11.2.0-9)
> arch/arm64/util/pmu.c: In function 'pmu_events_map__find':
> arch/arm64/util/pmu.c:18:35: error: invalid operands to binary != (have 'int' and 'struct perf_cpu')
> 18 | if (pmu->cpus->nr != cpu__max_cpu())
> | ~~~~~~~~~~~~~ ^~ ~~~~~~~~~~~~~~
> | | |
> | int struct perf_cpu
> make[5]: *** [/git/perf-5.16.0/tools/build/Makefile.build:139: util] Error 2
> make[4]: *** [/git/perf-5.16.0/tools/build/Makefile.build:139: arm64] Error 2
> make[3]: *** [/git/perf-5.16.0/tools/build/Makefile.build:139: arch] Error 2
> 30 19.15 debian:experimental-x-mips : Ok mips-linux-gnu-gcc (Debian 10.2.1-6) 10.2.1 20210110
Fixed with the patch below squashed into 48/48 where the problem
was introduced.
tmp.perf/perf_cpu updated.
[perfbuilder@five ~]$ dm debian:experimental-x-arm64 ubuntu:16.04-x-arm ubuntu:16.04-x-arm64 ubuntu:18.04-x-arm ubuntu:18.04-x-arm64
1 26.08 debian:experimental-x-arm64 : Ok aarch64-linux-gnu-gcc (Debian 11.2.0-9) 11.2.0
2 21.68 ubuntu:16.04-x-arm : Ok arm-linux-gnueabihf-gcc (Ubuntu/Linaro 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609
3 21.53 ubuntu:16.04-x-arm64 : Ok aarch64-linux-gnu-gcc (Ubuntu/Linaro 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609
4 61.22 ubuntu:18.04-x-arm : Ok arm-linux-gnueabihf-gcc (Ubuntu/Linaro 7.5.0-3ubuntu1~18.04) 7.5.0
5 108.41 ubuntu:18.04-x-arm64 : Ok aarch64-linux-gnu-gcc (Ubuntu/Linaro 7.5.0-3ubuntu1~18.04) 7.5.0
BUILD_TARBALL_HEAD=5610e6f0b9e64a02dfc906c9db9d18fd198200d9
6 238.918
[perfbuilder@five ~]$
diff --git a/tools/perf/arch/arm/util/cs-etm.c b/tools/perf/arch/arm/util/cs-etm.c
index 129c0272d65b8596..2e8b2c4365a0003f 100644
--- a/tools/perf/arch/arm/util/cs-etm.c
+++ b/tools/perf/arch/arm/util/cs-etm.c
@@ -203,9 +203,11 @@ static int cs_etm_set_option(struct auxtrace_record *itr,
struct perf_cpu_map *online_cpus = perf_cpu_map__new(NULL);
/* Set option of each CPU we have */
- for (i = 0; i < cpu__max_cpu(); i++) {
- if (!perf_cpu_map__has(event_cpus, i) ||
- !perf_cpu_map__has(online_cpus, i))
+ for (i = 0; i < cpu__max_cpu().cpu; i++) {
+ struct perf_cpu cpu = { .cpu = i, };
+
+ if (!perf_cpu_map__has(event_cpus, cpu) ||
+ !perf_cpu_map__has(online_cpus, cpu))
continue;
if (option & BIT(ETM_OPT_CTXTID)) {
@@ -522,9 +524,11 @@ cs_etm_info_priv_size(struct auxtrace_record *itr __maybe_unused,
/* cpu map is not empty, we have specific CPUs to work with */
if (!perf_cpu_map__empty(event_cpus)) {
- for (i = 0; i < cpu__max_cpu(); i++) {
- if (!perf_cpu_map__has(event_cpus, i) ||
- !perf_cpu_map__has(online_cpus, i))
+ for (i = 0; i < cpu__max_cpu().cpu; i++) {
+ struct perf_cpu cpu = { .cpu = i, };
+
+ if (!perf_cpu_map__has(event_cpus, cpu) ||
+ !perf_cpu_map__has(online_cpus, cpu))
continue;
if (cs_etm_is_ete(itr, i))
@@ -536,8 +540,10 @@ cs_etm_info_priv_size(struct auxtrace_record *itr __maybe_unused,
}
} else {
/* get configuration for all CPUs in the system */
- for (i = 0; i < cpu__max_cpu(); i++) {
- if (!perf_cpu_map__has(online_cpus, i))
+ for (i = 0; i < cpu__max_cpu().cpu; i++) {
+ struct perf_cpu cpu = { .cpu = i, };
+
+ if (!perf_cpu_map__has(online_cpus, cpu))
continue;
if (cs_etm_is_ete(itr, i))
@@ -722,8 +728,10 @@ static int cs_etm_info_fill(struct auxtrace_record *itr,
} else {
/* Make sure all specified CPUs are online */
for (i = 0; i < perf_cpu_map__nr(event_cpus); i++) {
- if (perf_cpu_map__has(event_cpus, i) &&
- !perf_cpu_map__has(online_cpus, i))
+ struct perf_cpu cpu = { .cpu = i, };
+
+ if (perf_cpu_map__has(event_cpus, cpu) &&
+ !perf_cpu_map__has(online_cpus, cpu))
return -EINVAL;
}
@@ -743,9 +751,12 @@ static int cs_etm_info_fill(struct auxtrace_record *itr,
offset = CS_ETM_SNAPSHOT + 1;
- for (i = 0; i < cpu__max_cpu() && offset < priv_size; i++)
- if (perf_cpu_map__has(cpu_map, i))
+ for (i = 0; i < cpu__max_cpu().cpu && offset < priv_size; i++) {
+ struct perf_cpu cpu = { .cpu = i, };
+
+ if (perf_cpu_map__has(cpu_map, cpu))
cs_etm_get_metadata(i, &offset, itr, info);
+ }
perf_cpu_map__put(online_cpus);
diff --git a/tools/perf/arch/arm64/util/pmu.c b/tools/perf/arch/arm64/util/pmu.c
index d3a18f9c85f6a731..79124bba713e5f0d 100644
--- a/tools/perf/arch/arm64/util/pmu.c
+++ b/tools/perf/arch/arm64/util/pmu.c
@@ -15,7 +15,7 @@ const struct pmu_events_map *pmu_events_map__find(void)
* The cpumap should cover all CPUs. Otherwise, some CPUs may
* not support some events or have different event IDs.
*/
- if (pmu->cpus->nr != cpu__max_cpu())
+ if (pmu->cpus->nr != cpu__max_cpu().cpu)
return NULL;
return perf_pmu__find_map(pmu);