Re: [PATCH v2 0/4] perf timechart: Small optimization for backtrace

[Namhyung Kim]

Hi Ian,

On Thu, Jun 04, 2026 at 03:38:05PM -0700, Ian Rogers wrote:
> On Thu, Jun 4, 2026 at 2:49 PM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
> >
> > Hello,
> >
> > I found an issue int timechart backtrace handling during the last review.
> > The goal is to reduce unnecessary work in generating backtrace string.
> >
> > v2 changes)
> >  * fix memory leaks when backtrace is not used
> >  * copy backtrace when it's used twice
> 
> Hi Namhyung,
> 
> thanks for cleaning this up and the fixes! Running a build of:
> $ make -C tools/perf O=/tmp/perf DEBUG=1
> EXTRA_CFLAGS="-fno-omit-frame-pointer -O0 -g -fsanitize=address" clean
> all
> 
> I get an error but the test harness makes it look like the test is
> passing (I added set -x to the test and removed the "> /dev/null
> 2>&1"):
> ```
> $ sudo /tmp/perf/perf test -vv timechart
> 136: perf timechart tests
> 136: perf timechart tests:
> ---- start ----
> test child forked, pid 937043
> + err=0
> ++ mktemp /tmp/__perf_timechart_test.perf.data.XXXXX
> + perfdata=/tmp/__perf_timechart_test.perf.data.cXMvK
> ++ mktemp /tmp/__perf_timechart_test.output.XXXXX.svg
> + output=/tmp/__perf_timechart_test.output.xZXfy.svg
> + trap trap_cleanup EXIT TERM INT
> + perf check feature -q libtraceevent
> + test_timechart
> + echo 'Basic perf timechart test'
> Basic perf timechart test
> + perf timechart record -o /tmp/__perf_timechart_test.perf.data.cXMvK true
> [ perf record: Woken up 1 times to write data ]
> [ perf record: Captured and wrote 4.135 MB
> /tmp/__perf_timechart_test.perf.data.cXMvK (748 samples) ]
> 
> =================================================================
> ==937049==ERROR: LeakSanitizer: detected memory leaks
> 
> Direct leak of 136 byte(s) in 1 object(s) allocated from:
>    #0 0x7fd211119a23 in calloc
> ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:74
>    #1 0x555c44044416 in timechart__record ~/linux/tools/perf/builtin-timech
> art.c:1895
>    #2 0x555c44046c19 in cmd_timechart ~/linux/tools/perf/builtin-timechart.
> c:2073
>    #3 0x555c440828a4 in run_builtin ~/linux/tools/perf/perf.c:348
>    #4 0x555c44082d42 in handle_internal_command ~/linux/tools/perf/perf.c:3
> 98
>    #5 0x555c44083110 in run_argv ~/linux/tools/perf/perf.c:442
>    #6 0x555c44083667 in main ~/linux/tools/perf/perf.c:549
>    #7 0x7fd206e29f76 in __libc_start_call_main
> ../sysdeps/nptl/libc_start_call_main.h:58
>    #8 0x7fffcf8a352a  ([stack]+0x2052a)
> 
> Direct leak of 10 byte(s) in 2 object(s) allocated from:
>    #0 0x7fd211113e20 in strdup
> ../../../../src/libsanitizer/asan/asan_interceptors.cpp:593
>    #1 0x555c440444a1 in timechart__record ~/linux/tools/perf/builtin-timech
> art.c:1902
>    #2 0x555c44046c19 in cmd_timechart ~/linux/tools/perf/builtin-timechart.
> c:2073
>    #3 0x555c440828a4 in run_builtin ~/linux/tools/perf/perf.c:348
>    #4 0x555c44082d42 in handle_internal_command ~/linux/tools/perf/perf.c:3
> 98
>    #5 0x555c44083110 in run_argv ~/linux/tools/perf/perf.c:442
>    #6 0x555c44083667 in main ~/linux/tools/perf/perf.c:549
>    #7 0x7fd206e29f76 in __libc_start_call_main
> ../sysdeps/nptl/libc_start_call_main.h:58
>    #8 0x7fffcf8a352a  ([stack]+0x2052a)
> 
> Indirect leak of 44 byte(s) in 4 object(s) allocated from:
>    #0 0x7fd211113e20 in strdup
> ../../../../src/libsanitizer/asan/asan_interceptors.cpp:593
>    #1 0x555c440445ef in timechart__record ~/linux/tools/perf/builtin-timech
> art.c:1908
>    #2 0x555c44046c19 in cmd_timechart ~/linux/tools/perf/builtin-timechart.
> c:2073
>    #3 0x555c440828a4 in run_builtin ~/linux/tools/perf/perf.c:348
>    #4 0x555c44082d42 in handle_internal_command ~/linux/tools/perf/perf.c:3
> 98
>    #5 0x555c44083110 in run_argv ~/linux/tools/perf/perf.c:442
>    #6 0x555c44083667 in main ~/linux/tools/perf/perf.c:549
>    #7 0x7fd206e29f76 in __libc_start_call_main
> ../sysdeps/nptl/libc_start_call_main.h:58
>    #8 0x7fffcf8a352a  ([stack]+0x2052a)
> 
> Indirect leak of 41 byte(s) in 4 object(s) allocated from:
>    #0 0x7fd211113e20 in strdup
> ../../../../src/libsanitizer/asan/asan_interceptors.cpp:593
>    #1 0x555c44044696 in timechart__record ~/linux/tools/perf/builtin-timech
> art.c:1911
>    #2 0x555c44046c19 in cmd_timechart ~/linux/tools/perf/builtin-timechart.
> c:2073
>    #3 0x555c440828a4 in run_builtin ~/linux/tools/perf/perf.c:348
>    #4 0x555c44082d42 in handle_internal_command ~/linux/tools/perf/perf.c:3
> 98
>    #5 0x555c44083110 in run_argv ~/linux/tools/perf/perf.c:442
>    #6 0x555c44083667 in main ~/linux/tools/perf/perf.c:549
>    #7 0x7fd206e29f76 in __libc_start_call_main
> ../sysdeps/nptl/libc_start_call_main.h:58
>    #8 0x7fffcf8a352a  ([stack]+0x2052a)
> 
> Indirect leak of 11 byte(s) in 4 object(s) allocated from:
>    #0 0x7fd211113e20 in strdup
> ../../../../src/libsanitizer/asan/asan_interceptors.cpp:593
>    #1 0x555c440444a1 in timechart__record ~/linux/tools/perf/builtin-timech
> art.c:1902
>    #2 0x555c44046c19 in cmd_timechart ~/linux/tools/perf/builtin-timechart.
> c:2073
>    #3 0x555c440828a4 in run_builtin ~/linux/tools/perf/perf.c:348
>    #4 0x555c44082d42 in handle_internal_command ~/linux/tools/perf/perf.c:3
> 98
>    #5 0x555c44083110 in run_argv ~/linux/tools/perf/perf.c:442
>    #6 0x555c44083667 in main ~/linux/tools/perf/perf.c:549
>    #7 0x7fd206e29f76 in __libc_start_call_main
> ../sysdeps/nptl/libc_start_call_main.h:58
>    #8 0x7fffcf8a352a  ([stack]+0x2052a)
> 
> SUMMARY: AddressSanitizer: 242 byte(s) leaked in 15 allocation(s).
> + echo 'Basic perf timechart test [Skipped: perf timechart record
> failed (permissions/events?)]'
> Basic perf timechart test [Skipped: perf timechart record failed
> (permissions/events?)]
> + return
> + cleanup
> + rm -f /tmp/__perf_timechart_test.perf.data.cXMvK
> + rm -f /tmp/__perf_timechart_test.output.xZXfy.svg
> + trap - EXIT TERM INT
> + exit 0
> ---- end(0) ----
> 136: perf timechart tests
> : Ok
> ```
> 
> It was this test that I was trying to make leak sanitizer clean when
> doing my fixes.

I think these leaks are pre-existing.  I don't see where it frees cpu
samples and other events.  I'll take a look and send a fix separately.

Thanks,
Namhyung