| Message ID | 20200724071111.35593-3-liwei391@huawei.com (mailing list archive) |
|---|---|
| State | Mainlined |
| Commit | 3e43d79da1dc9731aee24d8b3e8059a4d2297bfc |
| Headers | show |
| Series | perf tools: Fix record failure when mixed with ARM SPE event | expand |
On Fri, Jul 24, 2020 at 03:11:11PM +0800, Wei Li wrote: > - Firstly, the function auxtrace_record__init() will be invoked only > once, the variable "arm_spe_pmus" will not be used afterwards, thus > we don't need to check "arm_spe_pmus" is NULL or not; > - Another reason is, even though SPE is micro-architecture dependent, > but so far it only supports "statistical-profiling-extension-v1" and > we have no chance to use multiple SPE's PMU events in Perf command. I find the above changelog somewhat out of touch with the patch itself. The only thing that is happening here is the removal of a useless check and a fix for a memory leak. Once again whether Arnaldo wants to make the changes by hand or not you may have to resubmit. Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org> > > So remove the useless check code to make it clear. > > Signed-off-by: Wei Li <liwei391@huawei.com> > --- > tools/perf/arch/arm/util/auxtrace.c | 9 ++++----- > 1 file changed, 4 insertions(+), 5 deletions(-) > > diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c > index 28a5d0c18b1d..b187bddbd01a 100644 > --- a/tools/perf/arch/arm/util/auxtrace.c > +++ b/tools/perf/arch/arm/util/auxtrace.c > @@ -57,17 +57,15 @@ struct auxtrace_record > struct evsel *evsel; > bool found_etm = false; > struct perf_pmu *found_spe = NULL; > - static struct perf_pmu **arm_spe_pmus = NULL; > - static int nr_spes = 0; > + struct perf_pmu **arm_spe_pmus = NULL; > + int nr_spes = 0; > int i = 0; > > if (!evlist) > return NULL; > > cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME); > - > - if (!arm_spe_pmus) > - arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > + arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > evlist__for_each_entry(evlist, evsel) { > if (cs_etm_pmu && > @@ -84,6 +82,7 @@ struct auxtrace_record > } > } > } > + free(arm_spe_pmus); > > if (found_etm && found_spe) { > pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n"); > -- > 2.17.1 >
Em Mon, Jul 27, 2020 at 02:34:36PM -0600, Mathieu Poirier escreveu: > On Fri, Jul 24, 2020 at 03:11:11PM +0800, Wei Li wrote: > > - Firstly, the function auxtrace_record__init() will be invoked only > > once, the variable "arm_spe_pmus" will not be used afterwards, thus > > we don't need to check "arm_spe_pmus" is NULL or not; > > - Another reason is, even though SPE is micro-architecture dependent, > > but so far it only supports "statistical-profiling-extension-v1" and > > we have no chance to use multiple SPE's PMU events in Perf command. > > I find the above changelog somewhat out of touch with the patch itself. The > only thing that is happening here is the removal of a useless check and a fix > for a memory leak. > > Once again whether Arnaldo wants to make the changes by hand or not you may have > to resubmit. I'll fix it now, thanks for reviewing. - Arnaldo > Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org> > > > > > So remove the useless check code to make it clear. > > > > Signed-off-by: Wei Li <liwei391@huawei.com> > > --- > > tools/perf/arch/arm/util/auxtrace.c | 9 ++++----- > > 1 file changed, 4 insertions(+), 5 deletions(-) > > > > diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c > > index 28a5d0c18b1d..b187bddbd01a 100644 > > --- a/tools/perf/arch/arm/util/auxtrace.c > > +++ b/tools/perf/arch/arm/util/auxtrace.c > > @@ -57,17 +57,15 @@ struct auxtrace_record > > struct evsel *evsel; > > bool found_etm = false; > > struct perf_pmu *found_spe = NULL; > > - static struct perf_pmu **arm_spe_pmus = NULL; > > - static int nr_spes = 0; > > + struct perf_pmu **arm_spe_pmus = NULL; > > + int nr_spes = 0; > > int i = 0; > > > > if (!evlist) > > return NULL; > > > > cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME); > > - > > - if (!arm_spe_pmus) > > - arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > + arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > > > evlist__for_each_entry(evlist, evsel) { > > if (cs_etm_pmu && > > @@ -84,6 +82,7 @@ struct auxtrace_record > > } > > } > > } > > + free(arm_spe_pmus); > > > > if (found_etm && found_spe) { > > pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n"); > > -- > > 2.17.1 > >
Em Mon, Jul 27, 2020 at 02:34:36PM -0600, Mathieu Poirier escreveu: > On Fri, Jul 24, 2020 at 03:11:11PM +0800, Wei Li wrote: > > - Firstly, the function auxtrace_record__init() will be invoked only > > once, the variable "arm_spe_pmus" will not be used afterwards, thus > > we don't need to check "arm_spe_pmus" is NULL or not; > > - Another reason is, even though SPE is micro-architecture dependent, > > but so far it only supports "statistical-profiling-extension-v1" and > > we have no chance to use multiple SPE's PMU events in Perf command. > > I find the above changelog somewhat out of touch with the patch itself. The > only thing that is happening here is the removal of a useless check and a fix > for a memory leak. Humm, I think the original intent of that code was to cache the results of find_all_arm_spe_pmus(), as the variable it is assigned to is static. So not a leak, as there was that static reference to it to reuse it later, but that is strange in a function named "__init()" which usually is called only once, anyway, so I think that the paragraph with "Firstly" is kinda ok, but confusing, I think it should read: - auxtrace_record__init() is called only once, so there is no point in using a static variable to cache the results of find_all_arm_spe_pmus(), make it local and free the results after use. The second paragraph is SPE specific, so I'm not qualified to judge on it. I'm replacing the first paragraph with the version I wrote and keep it in my local branch, please holler if you think I misunderstood. - Arnaldo > Once again whether Arnaldo wants to make the changes by hand or not you may have > to resubmit. > > Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org> > > > > > So remove the useless check code to make it clear. > > > > Signed-off-by: Wei Li <liwei391@huawei.com> > > --- > > tools/perf/arch/arm/util/auxtrace.c | 9 ++++----- > > 1 file changed, 4 insertions(+), 5 deletions(-) > > > > diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c > > index 28a5d0c18b1d..b187bddbd01a 100644 > > --- a/tools/perf/arch/arm/util/auxtrace.c > > +++ b/tools/perf/arch/arm/util/auxtrace.c > > @@ -57,17 +57,15 @@ struct auxtrace_record > > struct evsel *evsel; > > bool found_etm = false; > > struct perf_pmu *found_spe = NULL; > > - static struct perf_pmu **arm_spe_pmus = NULL; > > - static int nr_spes = 0; > > + struct perf_pmu **arm_spe_pmus = NULL; > > + int nr_spes = 0; > > int i = 0; > > > > if (!evlist) > > return NULL; > > > > cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME); > > - > > - if (!arm_spe_pmus) > > - arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > + arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > > > evlist__for_each_entry(evlist, evsel) { > > if (cs_etm_pmu && > > @@ -84,6 +82,7 @@ struct auxtrace_record > > } > > } > > } > > + free(arm_spe_pmus); > > > > if (found_etm && found_spe) { > > pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n"); > > -- > > 2.17.1 > >
Hi Arnaldo, On Tue, Jul 28, 2020 at 09:02:20AM -0300, Arnaldo Carvalho de Melo wrote: > Em Mon, Jul 27, 2020 at 02:34:36PM -0600, Mathieu Poirier escreveu: > > On Fri, Jul 24, 2020 at 03:11:11PM +0800, Wei Li wrote: > > > - Firstly, the function auxtrace_record__init() will be invoked only > > > once, the variable "arm_spe_pmus" will not be used afterwards, thus > > > we don't need to check "arm_spe_pmus" is NULL or not; > > > - Another reason is, even though SPE is micro-architecture dependent, > > > but so far it only supports "statistical-profiling-extension-v1" and > > > we have no chance to use multiple SPE's PMU events in Perf command. > > > > I find the above changelog somewhat out of touch with the patch itself. The > > only thing that is happening here is the removal of a useless check and a fix > > for a memory leak. > > Humm, I think the original intent of that code was to cache the results > of find_all_arm_spe_pmus(), as the variable it is assigned to is static. > > So not a leak, as there was that static reference to it to reuse it > later, but that is strange in a function named "__init()" which usually > is called only once, anyway, so I think that the paragraph with > "Firstly" is kinda ok, but confusing, I think it should read: > > - auxtrace_record__init() is called only once, so there is no point in > using a static variable to cache the results of > find_all_arm_spe_pmus(), make it local and free the results after use. > > The second paragraph is SPE specific, so I'm not qualified to judge on > it. > > I'm replacing the first paragraph with the version I wrote and keep it > in my local branch, please holler if you think I misunderstood. Thanks a lot for this. These two paragraphs were coming from reviewing and comments, but I think your rephrasing is very sufficient to describe what this patch is doing :) Thanks, Leo > > Once again whether Arnaldo wants to make the changes by hand or not you may have > > to resubmit. > > > > Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org> > > > > > > > > So remove the useless check code to make it clear. > > > > > > Signed-off-by: Wei Li <liwei391@huawei.com> > > > --- > > > tools/perf/arch/arm/util/auxtrace.c | 9 ++++----- > > > 1 file changed, 4 insertions(+), 5 deletions(-) > > > > > > diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c > > > index 28a5d0c18b1d..b187bddbd01a 100644 > > > --- a/tools/perf/arch/arm/util/auxtrace.c > > > +++ b/tools/perf/arch/arm/util/auxtrace.c > > > @@ -57,17 +57,15 @@ struct auxtrace_record > > > struct evsel *evsel; > > > bool found_etm = false; > > > struct perf_pmu *found_spe = NULL; > > > - static struct perf_pmu **arm_spe_pmus = NULL; > > > - static int nr_spes = 0; > > > + struct perf_pmu **arm_spe_pmus = NULL; > > > + int nr_spes = 0; > > > int i = 0; > > > > > > if (!evlist) > > > return NULL; > > > > > > cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME); > > > - > > > - if (!arm_spe_pmus) > > > - arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > > + arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > > > > > evlist__for_each_entry(evlist, evsel) { > > > if (cs_etm_pmu && > > > @@ -84,6 +82,7 @@ struct auxtrace_record > > > } > > > } > > > } > > > + free(arm_spe_pmus); > > > > > > if (found_etm && found_spe) { > > > pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n"); > > > -- > > > 2.17.1 > > > > > -- > > - Arnaldo
On Tue, 28 Jul 2020 at 06:02, Arnaldo Carvalho de Melo <acme@kernel.org> wrote: > > Em Mon, Jul 27, 2020 at 02:34:36PM -0600, Mathieu Poirier escreveu: > > On Fri, Jul 24, 2020 at 03:11:11PM +0800, Wei Li wrote: > > > - Firstly, the function auxtrace_record__init() will be invoked only > > > once, the variable "arm_spe_pmus" will not be used afterwards, thus > > > we don't need to check "arm_spe_pmus" is NULL or not; > > > - Another reason is, even though SPE is micro-architecture dependent, > > > but so far it only supports "statistical-profiling-extension-v1" and > > > we have no chance to use multiple SPE's PMU events in Perf command. > > > > I find the above changelog somewhat out of touch with the patch itself. The > > only thing that is happening here is the removal of a useless check and a fix > > for a memory leak. > > Humm, I think the original intent of that code was to cache the results > of find_all_arm_spe_pmus(), as the variable it is assigned to is static. Correct, but as you pointed out below the function is called only once. And there is still a leak as that memory is never freed. > > So not a leak, as there was that static reference to it to reuse it > later, but that is strange in a function named "__init()" which usually > is called only once, anyway, so I think that the paragraph with > "Firstly" is kinda ok, but confusing, I think it should read: > > - auxtrace_record__init() is called only once, so there is no point in > using a static variable to cache the results of > find_all_arm_spe_pmus(), make it local and free the results after use. This is exactly what this patch does and what the changelog should read. > > The second paragraph is SPE specific, so I'm not qualified to judge on > it. > > I'm replacing the first paragraph with the version I wrote and keep it > in my local branch, please holler if you think I misunderstood. > There is no point for the next paragraph, it has no relevance to what the code is doing. Thanks for the editing. > - Arnaldo > > > Once again whether Arnaldo wants to make the changes by hand or not you may have > > to resubmit. > > > > Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org> > > > > > > > > So remove the useless check code to make it clear. > > > > > > Signed-off-by: Wei Li <liwei391@huawei.com> > > > --- > > > tools/perf/arch/arm/util/auxtrace.c | 9 ++++----- > > > 1 file changed, 4 insertions(+), 5 deletions(-) > > > > > > diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c > > > index 28a5d0c18b1d..b187bddbd01a 100644 > > > --- a/tools/perf/arch/arm/util/auxtrace.c > > > +++ b/tools/perf/arch/arm/util/auxtrace.c > > > @@ -57,17 +57,15 @@ struct auxtrace_record > > > struct evsel *evsel; > > > bool found_etm = false; > > > struct perf_pmu *found_spe = NULL; > > > - static struct perf_pmu **arm_spe_pmus = NULL; > > > - static int nr_spes = 0; > > > + struct perf_pmu **arm_spe_pmus = NULL; > > > + int nr_spes = 0; > > > int i = 0; > > > > > > if (!evlist) > > > return NULL; > > > > > > cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME); > > > - > > > - if (!arm_spe_pmus) > > > - arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > > + arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); > > > > > > evlist__for_each_entry(evlist, evsel) { > > > if (cs_etm_pmu && > > > @@ -84,6 +82,7 @@ struct auxtrace_record > > > } > > > } > > > } > > > + free(arm_spe_pmus); > > > > > > if (found_etm && found_spe) { > > > pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n"); > > > -- > > > 2.17.1 > > > > > -- > > - Arnaldo
diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c index 28a5d0c18b1d..b187bddbd01a 100644 --- a/tools/perf/arch/arm/util/auxtrace.c +++ b/tools/perf/arch/arm/util/auxtrace.c @@ -57,17 +57,15 @@ struct auxtrace_record struct evsel *evsel; bool found_etm = false; struct perf_pmu *found_spe = NULL; - static struct perf_pmu **arm_spe_pmus = NULL; - static int nr_spes = 0; + struct perf_pmu **arm_spe_pmus = NULL; + int nr_spes = 0; int i = 0; if (!evlist) return NULL; cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME); - - if (!arm_spe_pmus) - arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); + arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err); evlist__for_each_entry(evlist, evsel) { if (cs_etm_pmu && @@ -84,6 +82,7 @@ struct auxtrace_record } } } + free(arm_spe_pmus); if (found_etm && found_spe) { pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n");
- Firstly, the function auxtrace_record__init() will be invoked only once, the variable "arm_spe_pmus" will not be used afterwards, thus we don't need to check "arm_spe_pmus" is NULL or not; - Another reason is, even though SPE is micro-architecture dependent, but so far it only supports "statistical-profiling-extension-v1" and we have no chance to use multiple SPE's PMU events in Perf command. So remove the useless check code to make it clear. Signed-off-by: Wei Li <liwei391@huawei.com> --- tools/perf/arch/arm/util/auxtrace.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-)