On 21-May-2021, at 12:33 PM, kernel test robot <oliver.sang@intel.com> wrote:



Greeting,

FYI, we noticed the following commit (built with gcc-9):

commit: 06e5ca746c07380dfe0e4c3e10c34a6daa69eae6 ("perf tools: Support pipeline stage cycles for powerpc")
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master


in testcase: perf-test
version: perf-test-x86_64-git-1_20210508
with following parameters:

type: lkp
ucode: 0xde

test-description: The internal Perf Test suite.


on test machine: 4 threads 1 sockets Intel(R) Core(TM) i7-7567U CPU @ 3.50GHz with 32G memory

caused below changes (please refer to attached dmesg/kmsg for entire log/backtrace):




If you fix the issue, kindly add following tag
Reported-by: kernel test robot <oliver.sang@intel.com>

Direct leak of 736 byte(s) in 1 object(s) allocated from:
  #0 0x7f9cb9a16720 in __interceptor_realloc (/usr/lib/x86_64-linux-gnu/libasan.so.5+0xe9720)
  #1 0x55e00bbd08ca in strbuf_grow util/strbuf.c:64
  #2 0x55e00bbd0d42 in strbuf_add util/strbuf.c:86
  #3 0x55e00bd458be in strbuf_addstr util/strbuf.h:87
  #4 0x55e00bd66ddb in add_key util/sort.c:3329
  #5 0x55e00bd67045 in add_sort_string util/sort.c:3342
  #6 0x55e00bd67369 in sort_help util/sort.c:3368
  #7 0x55e00b8a1f00 in cmd_report /usr/src/perf_selftests-x86_64-rhel-8.3-06e5ca746c07380dfe0e4c3e10c34a6daa69eae6/tools/perf/builtin-report.c:1177
  #8 0x55e00bae47e2 in run_builtin /usr/src/perf_selftests-x86_64-rhel-8.3-06e5ca746c07380dfe0e4c3e10c34a6daa69eae6/tools/perf/perf.c:313
  #9 0x55e00bae4f8d in handle_internal_command /usr/src/perf_selftests-x86_64-rhel-8.3-06e5ca746c07380dfe0e4c3e10c34a6daa69eae6/tools/perf/perf.c:365
  #10 0x55e00bae54af in run_argv /usr/src/perf_selftests-x86_64-rhel-8.3-06e5ca746c07380dfe0e4c3e10c34a6daa69eae6/tools/perf/perf.c:409
  #11 0x55e00bae5e8f in main /usr/src/perf_selftests-x86_64-rhel-8.3-06e5ca746c07380dfe0e4c3e10c34a6daa69eae6/tools/perf/perf.c:539
  #12 0x7f9cb8c0309a in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2409a)

Hi,

With upstream perf tool compiled with "-fsanitize=undefined -fsanitize=address"
and running "perf mem report”, I could reproduce the above trace. However I found we have
already existing similar traces apart from new one reported. Example:

Direct leak of 736 byte(s) in 1 object(s) allocated from:
   #0 0x7fce92e16ffe in __interceptor_realloc (/lib/x86_64-linux-gnu/libasan.so.5+0x10dffe)
   #1 0x563b86f16cbc in strbuf_grow util/strbuf.c:64
   #2 0x563b86f16cbc in strbuf_addv util/strbuf.c:112
   #3 0x563b86f1953d in strbuf_addf util/strbuf.c:134
   #4 0x563b871124c3 in add_key util/sort.c:3351
   #5 0x563b871124c3 in add_sort_string util/sort.c:3361
   #6 0x563b871124c3 in sort_help util/sort.c:3387
   #7 0x563b86bc9b7c in cmd_report /home/athira/linux/tools/perf/builtin-report.c:1205
   #8 0x563b86c9a36a in report_events /home/athira/linux/tools/perf/builtin-mem.c:350
   #9 0x563b86c9a36a in cmd_mem /home/athira/linux/tools/perf/builtin-mem.c:492
   #10 0x563b86dfdc37 in run_builtin /home/athira/linux/tools/perf/perf.c:313
   #11 0x563b86dfe769 in handle_internal_command /home/athira/linux/tools/perf/perf.c:365
   #12 0x563b86b49ca1 in run_argv /home/athira/linux/tools/perf/perf.c:409
   #13 0x563b86b49ca1 in main /home/athira/linux/tools/perf/perf.c:539
   #14 0x7fce922a90b2 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x270b2)

Here it is from "strbuf_addf" . The new trace from “strbuf_addstr" started appearing since the patch:
commit: 06e5ca746c07380dfe0e4c3e10c34a6daa69eae6 ("perf tools: Support pipeline stage cycles for powerpc")
added a new sort key and code is now touching “strbuf_addstr" also. But the reason for detected leak in
both cases are same.

Checking further, these doesn't seems a real memory leak issue. This is actually used for sort order help string
and IIUC there is no need to explicitly free it.

Thanks
Athira

To reproduce:

      git clone https://github.com/intel/lkp-tests.git
      cd lkp-tests
      bin/lkp install                job.yaml  # job file is attached in this email
      bin/lkp split-job --compatible job.yaml  # generate the yaml file for lkp run
      bin/lkp run                    generated-yaml-file



---
0DAY/LKP+ Test Infrastructure                   Open Source Technology Center
https://lists.01.org/hyperkitty/list/lkp@lists.01.org       Intel Corporation

Thanks,
Oliver Sang

<config-5.12.0-rc2-00111-g06e5ca746c07><job-script.txt><kmsg.xz><perf-test.txt><job.yaml><reproduce.txt>