[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAP-5=fV_-Z85b_DMCCPFo1FU2dyUY9ELNEJz6fx+JgZg4iPj7g@mail.gmail.com>
Date: Wed, 23 Oct 2024 19:42:56 -0700
From: Ian Rogers <irogers@...gle.com>
To: Namhyung Kim <namhyung@...nel.org>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
Arnaldo Carvalho de Melo <acme@...nel.org>, Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>, Jiri Olsa <jolsa@...nel.org>,
Adrian Hunter <adrian.hunter@...el.com>, Kan Liang <kan.liang@...ux.intel.com>,
Thomas Richter <tmricht@...ux.ibm.com>, James Clark <james.clark@...aro.org>,
linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] perf check: Add sanitizer feature and use to avoid
test failure
On Wed, Oct 23, 2024 at 4:24 PM Namhyung Kim <namhyung@...nel.org> wrote:
>
> On Wed, Oct 23, 2024 at 03:19:38PM -0700, Namhyung Kim wrote:
> > On Tue, Oct 22, 2024 at 11:18:18PM -0700, Namhyung Kim wrote:
> > > On Tue, Oct 22, 2024 at 10:39:36AM -0700, Ian Rogers wrote:
> > > > Sure, the reproduction is trivial, just add -fsanitize=address, so I'm
> > > > surprised you're not already seeing it:
> > > > ```
> > > > $ perf test annotate -v
> > > > --- start ---
> > > > test child forked, pid 444258
> > > > 68e8a0-68e96b l noploop
> > > > perf does have symbol 'noploop'
> > > > Basic perf annotate test
> > > > : 0 0x68e8a0 <noploop>:
> > > > 0.00 : 68e8a0: pushq %rbp
> > > > 0.00 : 68e8a1: movq %rsp, %rbp
> > > > 0.00 : 68e8a4: subq $0x30, %rsp
> > [...]
> > > > 0.00 : 92d6: shrl %cl, %edx
> > > > 0.00 : 92d8: movl %edx, %ecx
> > > > 0.00 : 92da: movq %rax, %rdx
> > > > Basic annotate [Failed: missing disasm output when specifying the target symbol]
> > >
> > > Hmm.. this is strange. The error message says it failed when it
> > > specified the target symbol (noploop) for perf annotate.
> > >
> > > As it's the dominant symbol, it should have the same output for the
> > > first function (noploop) whether it has target symbol or not and it
> > > should match the disasm_regex. I'm curious how it can fail here.
> >
> > Hmm.. ok. For some reason, it wasn't failed when I add DEBUG=1.
>
> Oh, now I'm seeing why. We skip perf_session__delete() on !DEBUG build.
> :(
>
> >
> > Without DEBUG, I can see it now.
> >
> > =================================================================
> > ==1053492==ERROR: LeakSanitizer: detected memory leaks
> >
> > Direct leak of 33 byte(s) in 1 object(s) allocated from:
> > #0 0x7f1ad78edd20 in strdup ../../../../src/libsanitizer/asan/asan_interceptors.cpp:566
> > #1 0x55eda19cb76f in perf_data__open (linux/tools/perf/perf+0x65276f) (BuildId: 6fc1b7cdc123c7bd586ce55ea8b727875f42cda2)
> > #2 0x55eda18ffafa in __perf_session__new (linux/tools/perf/perf+0x586afa) (BuildId: 6fc1b7cdc123c7bd586ce55ea8b727875f42cda2)
> > #3 0x55eda15485d3 in cmd_annotate (linux/tools/perf/perf+0x1cf5d3) (BuildId: 6fc1b7cdc123c7bd586ce55ea8b727875f42cda2)
> > #4 0x55eda1695467 in run_builtin (linux/tools/perf/perf+0x31c467) (BuildId: 6fc1b7cdc123c7bd586ce55ea8b727875f42cda2)
> > #5 0x55eda1695c0e in handle_internal_command (linux/tools/perf/perf+0x31cc0e) (BuildId: 6fc1b7cdc123c7bd586ce55ea8b727875f42cda2)
> > #6 0x55eda153ba72 in main (linux/tools/perf/perf+0x1c2a72) (BuildId: 6fc1b7cdc123c7bd586ce55ea8b727875f42cda2)
> > #7 0x7f1acda43b89 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
> >
> > SUMMARY: AddressSanitizer: 33 byte(s) leaked in 1 allocation(s).
> > Unexpected signal in test_basic
> >
> > No idea how it can leak the data->file.path (that's what I can find
> > where strdup is used in the function).
>
> Maybe we need to revisit how much speed up it can give.
I think this is a different issue and not sanitizer related except for
the sanitizer catching the deliberate leak. My full build command line
is:
```
$ make -C tools/perf O=/tmp/perf DEBUG=1 EXTRA_CFLAGS="-O0 -g
-fno-omit-frame-pointer -fsanitize=address" CC=clang CXX=clang++
HOSTCC=clang
```
Of which the -O0 may be the thing that is making the difference and
breaking the test with sanitizers for me.
Thanks,
Ian
Powered by blists - more mailing lists