Skip to contentSkip to conversation

perf report: Fix a memory leak for perf_env on AMD

PATCH Applied Namhyung Kim 2 participants 4 messages Mar 11, 2025 – Mar 25, 2025 perf events

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

PATCH 1/1 perf report: Fix a memory leak for perf_env on AMD

The env.pmu_mapping can be leaked when it reads data from a pipe on AMD. For a pipe data, it reads the header data including pmu_mapping from PERF_RECORD_HEADER_FEATURE runtime. But it's already set in:

perf_session__new()

    __perf_session__new()
      evlist__init_trace_event_sample_raw()
        evlist__has_amd_ibs()
          perf_env__nr_pmu_mappings()

Then it'll overwrite that when it processes the HEADER_FEATURE record. Here's a report from address sanitizer.

Direct leak of 2689 byte(s) in 1 object(s) allocated from:

    #0 0x7fed8f814596 in realloc ../../../../src/libsanitizer/lsan/lsan_interceptors.cpp:98
    #1 0x5595a7d416b1 in strbuf_grow util/strbuf.c:64
    #2 0x5595a7d414ef in strbuf_init util/strbuf.c:25
    #3 0x5595a7d0f4b7 in perf_env__read_pmu_mappings util/env.c:362
    #4 0x5595a7d12ab7 in perf_env__nr_pmu_mappings util/env.c:517
    #5 0x5595a7d89d2f in evlist__has_amd_ibs util/amd-sample-raw.c:315
    #6 0x5595a7d87fb2 in evlist__init_trace_event_sample_raw util/sample-raw.c:23
    #7 0x5595a7d7f893 in __perf_session__new util/session.c:179
    #8 0x5595a7b79572 in perf_session__new util/session.h:115
    #9 0x5595a7b7e9dc in cmd_report builtin-report.c:1603
    #10 0x5595a7c019eb in run_builtin perf.c:351
    #11 0x5595a7c01c92 in handle_internal_command perf.c:404
    #12 0x5595a7c01deb in run_argv perf.c:448
    #13 0x5595a7c02134 in main perf.c:556
    #14 0x7fed85833d67 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58

Let's free the existing pmu_mapping data if any.

---
 tools/perf/util/header.c | 2 ++
 1 file changed, 2 insertions(+)
tools/perf/util/header.c +2 -0
index 1900965f87527948..e3cdc3b7b4ab2409 100644
@@ -2770,6 +2770,8 @@ static int process_pmu_mappings(struct feat_fd *ff, void *data __maybe_unused)
 		free(name);
 		pmu_num--;
 	}
+	/* AMD may set it by evlist__has_amd_ibs() from perf_session__new() */
+	free(ff->ph->env.pmu_mappings);
 	ff->ph->env.pmu_mappings = strbuf_detach(&sb, NULL);
 	return 0;
 
Signature · 1 line
2.49.0.rc0.332.g42c0ae87b1-goog
reviewer
#1 lore raw
Quoted 31 lines · Namhyung Kim

The env.pmu_mapping can be leaked when it reads data from a pipe on AMD. For a pipe data, it reads the header data including pmu_mapping from PERF_RECORD_HEADER_FEATURE runtime. But it's already set in:

perf_session__new()

    __perf_session__new()
      evlist__init_trace_event_sample_raw()
        evlist__has_amd_ibs()
          perf_env__nr_pmu_mappings()

Then it'll overwrite that when it processes the HEADER_FEATURE record. Here's a report from address sanitizer.

Direct leak of 2689 byte(s) in 1 object(s) allocated from:

    #0 0x7fed8f814596 in realloc ../../../../src/libsanitizer/lsan/lsan_interceptors.cpp:98
    #1 0x5595a7d416b1 in strbuf_grow util/strbuf.c:64
    #2 0x5595a7d414ef in strbuf_init util/strbuf.c:25
    #3 0x5595a7d0f4b7 in perf_env__read_pmu_mappings util/env.c:362
    #4 0x5595a7d12ab7 in perf_env__nr_pmu_mappings util/env.c:517
    #5 0x5595a7d89d2f in evlist__has_amd_ibs util/amd-sample-raw.c:315
    #6 0x5595a7d87fb2 in evlist__init_trace_event_sample_raw util/sample-raw.c:23
    #7 0x5595a7d7f893 in __perf_session__new util/session.c:179
    #8 0x5595a7b79572 in perf_session__new util/session.h:115
    #9 0x5595a7b7e9dc in cmd_report builtin-report.c:1603
    #10 0x5595a7c019eb in run_builtin perf.c:351
    #11 0x5595a7c01c92 in handle_internal_command perf.c:404
    #12 0x5595a7c01deb in run_argv perf.c:448
    #13 0x5595a7c02134 in main perf.c:556
    #14 0x7fed85833d67 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
Quoted 1 line

Let's free the existing pmu_mapping data if any.

Can we simplify evlist__has_amd_ibs as there seems no reason to be creating/parsing strings. For example:

```
bool evlist__has_amd_ibs(struct evlist *evlist)
{
   return perf_pmus__find("ibs_op") != NULL ||
perf_pmus__find("ibs_fetch") != NULL;
}
```

Similarly the variables ibs_fetch_type and ibs_op_type can be found with perf_pmus__find("ibs_op")->type and perf_pmus__find("ibs_op")->type respectively. It probably makes more sense to just cache access to the PMUs rather than do the finds all the time.

I wonder there is a problem with the pmus alternative that currently the pmus are global and not part of say the session, not loaded from the perf.data file, etc. So use on a different machine than the recording could be broken by what I suggest. I guess then, just throwing in the free is simplest.

Thanks,
Ian

Quoted 22 lines
---
 tools/perf/util/header.c | 2 ++
 1 file changed, 2 insertions(+)
tools/perf/util/header.c +2 -1
index 1900965f87527948..e3cdc3b7b4ab2409 100644
@@ -2770,6 +2770,8 @@ static int process_pmu_mappings(struct feat_fd *ff, void *data __maybe_unused)
                free(name);
                pmu_num--;
        }
+       /* AMD may set it by evlist__has_amd_ibs() from perf_session__new() */
+       free(ff->ph->env.pmu_mappings);
        ff->ph->env.pmu_mappings = strbuf_detach(&sb, NULL);
        return 0;
 
--

2.49.0.rc0.332.g42c0ae87b1-goog

author maintainer
#2 lore raw
Quoted 36 lines · Ian Rogers
Quoted 31 lines · Namhyung Kim

The env.pmu_mapping can be leaked when it reads data from a pipe on AMD. For a pipe data, it reads the header data including pmu_mapping from PERF_RECORD_HEADER_FEATURE runtime. But it's already set in:

perf_session__new()

    __perf_session__new()
      evlist__init_trace_event_sample_raw()
        evlist__has_amd_ibs()
          perf_env__nr_pmu_mappings()

Then it'll overwrite that when it processes the HEADER_FEATURE record. Here's a report from address sanitizer.

Direct leak of 2689 byte(s) in 1 object(s) allocated from:

    #0 0x7fed8f814596 in realloc ../../../../src/libsanitizer/lsan/lsan_interceptors.cpp:98
    #1 0x5595a7d416b1 in strbuf_grow util/strbuf.c:64
    #2 0x5595a7d414ef in strbuf_init util/strbuf.c:25
    #3 0x5595a7d0f4b7 in perf_env__read_pmu_mappings util/env.c:362
    #4 0x5595a7d12ab7 in perf_env__nr_pmu_mappings util/env.c:517
    #5 0x5595a7d89d2f in evlist__has_amd_ibs util/amd-sample-raw.c:315
    #6 0x5595a7d87fb2 in evlist__init_trace_event_sample_raw util/sample-raw.c:23
    #7 0x5595a7d7f893 in __perf_session__new util/session.c:179
    #8 0x5595a7b79572 in perf_session__new util/session.h:115
    #9 0x5595a7b7e9dc in cmd_report builtin-report.c:1603
    #10 0x5595a7c019eb in run_builtin perf.c:351
    #11 0x5595a7c01c92 in handle_internal_command perf.c:404
    #12 0x5595a7c01deb in run_argv perf.c:448
    #13 0x5595a7c02134 in main perf.c:556
    #14 0x7fed85833d67 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
Quoted 1 line

Let's free the existing pmu_mapping data if any.

Can we simplify evlist__has_amd_ibs as there seems no reason to be creating/parsing strings. For example:

Quoted 7 lines
```
bool evlist__has_amd_ibs(struct evlist *evlist)
{
   return perf_pmus__find("ibs_op") != NULL ||
perf_pmus__find("ibs_fetch") != NULL;
}
```

I think the idea is to find the pmu from the recorded data file not at runtime.

Quoted 11 lines

Similarly the variables ibs_fetch_type and ibs_op_type can be found with perf_pmus__find("ibs_op")->type and perf_pmus__find("ibs_op")->type respectively. It probably makes more sense to just cache access to the PMUs rather than do the finds all the time.

I wonder there is a problem with the pmus alternative that currently the pmus are global and not part of say the session, not loaded from the perf.data file, etc. So use on a different machine than the recording could be broken by what I suggest. I guess then, just throwing in the free is simplest.

Agreed. It could be saved in a session or env.

For now, let's free the string to fix the leak quickly.

Thanks,
Namhyung

Quoted 24 lines
Quoted 24 lines
---
 tools/perf/util/header.c | 2 ++
 1 file changed, 2 insertions(+)
tools/perf/util/header.c +2 -1
index 1900965f87527948..e3cdc3b7b4ab2409 100644
@@ -2770,6 +2770,8 @@ static int process_pmu_mappings(struct feat_fd *ff, void *data __maybe_unused)
                free(name);
                pmu_num--;
        }
+       /* AMD may set it by evlist__has_amd_ibs() from perf_session__new() */
+       free(ff->ph->env.pmu_mappings);
        ff->ph->env.pmu_mappings = strbuf_detach(&sb, NULL);
        return 0;
 
--

2.49.0.rc0.332.g42c0ae87b1-goog

author maintainer Applied
#3 lore raw
In reply to Namhyung Kim
Quoted 11 lines · Namhyung Kim

The env.pmu_mapping can be leaked when it reads data from a pipe on AMD. For a pipe data, it reads the header data including pmu_mapping from PERF_RECORD_HEADER_FEATURE runtime. But it's already set in:

perf_session__new()

    __perf_session__new()
      evlist__init_trace_event_sample_raw()
        evlist__has_amd_ibs()
          perf_env__nr_pmu_mappings()

[...]

Applied to perf-tools-next, thanks!

Best regards,
Namhyung