Re: [PATCH v1] perf intel-pt: Fix memory sanitizer use-of-uninitialized-value
From: Adrian Hunter
Date: Mon Mar 25 2024 - 18:26:15 EST
On 25/03/24 17:44, Ian Rogers wrote:
> On Wed, Mar 20, 2024 at 9:26 AM Ian Rogers <irogers@xxxxxxxxxx> wrote:
>>
>> Running the test "Miscellaneous Intel PT testing" with a build with
>> -fsanitize=memory and -fsanitize-memory-track-origins I saw:
>>
>> ```
>> ==1257749==WARNING: MemorySanitizer: use-of-uninitialized-value
>> #0 0x5581c00a76b3 in intel_pt_sample_flags tools/perf/util/intel-pt.c:1527:17
>> #1 0x5581c00c5cf6 in intel_pt_run_decoder tools/perf/util/intel-pt.c:2961:3
>> #2 0x5581c00968f8 in intel_pt_process_timeless_queues tools/perf/util/intel-pt.c:3074:4
>> #3 0x5581c007cf49 in intel_pt_process_event tools/perf/util/intel-pt.c:3482:10
>> #4 0x5581bffa269a in auxtrace__process_event tools/perf/util/auxtrace.c:2830:9
>> #5 0x5581bfb826c0 in perf_session__deliver_event tools/perf/util/session.c:1649:8
>> #6 0x5581bfba1d7f in perf_session__process_event tools/perf/util/session.c:1891:9
>> #7 0x5581bfba82e4 in process_simple tools/perf/util/session.c:2452:9
>> #8 0x5581bfbabcc3 in reader__read_event tools/perf/util/session.c:2381:14
>> #9 0x5581bfbad942 in reader__process_events tools/perf/util/session.c:2430:8
>> #10 0x5581bfb78256 in __perf_session__process_events tools/perf/util/session.c:2477:8
>> #11 0x5581bfb702c4 in perf_session__process_events tools/perf/util/session.c:2643:9
>> #12 0x5581bf2da266 in __cmd_script tools/perf/builtin-script.c:2855:8
>> #13 0x5581bf2bfcdd in cmd_script tools/perf/builtin-script.c:4402:8
>> #14 0x5581bf67004b in run_builtin tools/perf/perf.c:350:11
>> #15 0x5581bf66b8ea in handle_internal_command tools/perf/perf.c:403:8
>> #16 0x5581bf66f527 in run_argv tools/perf/perf.c:447:2
>> #17 0x5581bf669d2d in main tools/perf/perf.c:561:3
>>
>> Uninitialized value was stored to memory at
>> #0 0x5581c005ddf8 in intel_pt_walk_insn tools/perf/util/intel-pt-decoder/intel-pt-decoder.c:1256:25
>> #1 0x5581c001a932 in intel_pt_walk_fup tools/perf/util/intel-pt-decoder/intel-pt-decoder.c:1428:9
>> #2 0x5581c000f76c in intel_pt_walk_trace tools/perf/util/intel-pt-decoder/intel-pt-decoder.c:3299:10
>> #3 0x5581c000899b in intel_pt_decode tools/perf/util/intel-pt-decoder/intel-pt-decoder.c:3988:10
>> #4 0x5581c00c5180 in intel_pt_run_decoder tools/perf/util/intel-pt.c:2941:11
>> #5 0x5581c00968f8 in intel_pt_process_timeless_queues tools/perf/util/intel-pt.c:3074:4
>> #6 0x5581c007cf49 in intel_pt_process_event tools/perf/util/intel-pt.c:3482:10
>> #7 0x5581bffa269a in auxtrace__process_event tools/perf/util/auxtrace.c:2830:9
>> #8 0x5581bfb826c0 in perf_session__deliver_event tools/perf/util/session.c:1649:8
>> #9 0x5581bfba1d7f in perf_session__process_event tools/perf/util/session.c:1891:9
>> #10 0x5581bfba82e4 in process_simple tools/perf/util/session.c:2452:9
>> #11 0x5581bfbabcc3 in reader__read_event tools/perf/util/session.c:2381:14
>> #12 0x5581bfbad942 in reader__process_events tools/perf/util/session.c:2430:8
>> #13 0x5581bfb78256 in __perf_session__process_events tools/perf/util/session.c:2477:8
>> #14 0x5581bfb702c4 in perf_session__process_events tools/perf/util/session.c:2643:9
>> #15 0x5581bf2da266 in __cmd_script tools/perf/builtin-script.c:2855:8
>> #16 0x5581bf2bfcdd in cmd_script tools/perf/builtin-script.c:4402:8
>> #17 0x5581bf67004b in run_builtin tools/perf/perf.c:350:11
>> #18 0x5581bf66b8ea in handle_internal_command tools/perf/perf.c:403:8
>> #19 0x5581bf66f527 in run_argv tools/perf/perf.c:447:2
>> ```
>>
>> Adding a curly brace initializer for the intel_pt_insn in
>> intel_pt_walk_fup rectifies this, however, there may be other issues
>> lurking behind this so initialize all similar instances.
>>
>> Fixes: f4aa081949e7 ("perf tools: Add Intel PT decoder")
>> Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
>
> Adrian, could you take a look at this for the sake of tests passing with msan.
I did have a look, but I found msan so slow and full of errors
that it never gave any results.
However, it is easy enough to instrument some debugging which is
what I am still looking at.
Just initializing intel_pt_insn is not quite right, so I need
to check things a bit more.
>
> Thanks,
> Ian
>
>> ---
>> tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 8 ++++----
>> 1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
>> index b450178e3420..b4a95af2e4cc 100644
>> --- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
>> +++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
>> @@ -1115,7 +1115,7 @@ static void intel_pt_sample_insn(struct intel_pt_decoder *decoder)
>> */
>> static void intel_pt_sample_fup_insn(struct intel_pt_decoder *decoder)
>> {
>> - struct intel_pt_insn intel_pt_insn;
>> + struct intel_pt_insn intel_pt_insn = {};
>> uint64_t max_insn_cnt, insn_cnt = 0;
>> int err;
>>
>> @@ -1418,7 +1418,7 @@ static inline bool intel_pt_fup_with_nlip(struct intel_pt_decoder *decoder,
>>
>> static int intel_pt_walk_fup(struct intel_pt_decoder *decoder)
>> {
>> - struct intel_pt_insn intel_pt_insn;
>> + struct intel_pt_insn intel_pt_insn = {};
>> uint64_t ip;
>> int err;
>>
>> @@ -1461,7 +1461,7 @@ static int intel_pt_walk_fup(struct intel_pt_decoder *decoder)
>>
>> static int intel_pt_walk_tip(struct intel_pt_decoder *decoder)
>> {
>> - struct intel_pt_insn intel_pt_insn;
>> + struct intel_pt_insn intel_pt_insn = {};
>> int err;
>>
>> err = intel_pt_walk_insn(decoder, &intel_pt_insn, 0);
>> @@ -1626,7 +1626,7 @@ static int intel_pt_emulated_ptwrite(struct intel_pt_decoder *decoder)
>>
>> static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)
>> {
>> - struct intel_pt_insn intel_pt_insn;
>> + struct intel_pt_insn intel_pt_insn = {};
>> int err;
>>
>> while (1) {
>> --
>> 2.44.0.291.gc1ea87d7ee-goog
>>