Re: [PATCH] perf data convert: Fix segfault when converting to json on arm64

[Ilkka Koskinen <ilkka@os.amperecomputing.com>]

[-- Attachment #1: Type: text/plain, Size: 2085 bytes --]


On Tue, 16 Jan 2024, Namhyung Kim wrote:
> Hello,
>
> On Fri, Jan 12, 2024 at 3:35 AM James Clark <james.clark@arm.com> wrote:
>>
>>
>>
>> On 11/01/2024 23:29, Ilkka Koskinen wrote:
>>> Arm64 doesn't have Model in /proc/cpuinfo and, thus, cpu_desc doesn't get
>>> assigned.
>>>
>>> Running
>>>       $ perf data convert --to-json perf.data.json
>>>
>>> ends up calling output_json_string() with NULL pointer, which causes a
>>> segmentation fault.
>>>
>>> Signed-off-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
>>> ---
>>>  tools/perf/util/data-convert-json.c | 5 +++++
>>>  1 file changed, 5 insertions(+)
>>>
>>> diff --git a/tools/perf/util/data-convert-json.c b/tools/perf/util/data-convert-json.c
>>> index 5bb3c2ba95ca..5d6de1cef546 100644
>>> --- a/tools/perf/util/data-convert-json.c
>>> +++ b/tools/perf/util/data-convert-json.c
>>> @@ -97,6 +97,11 @@ static void output_json_format(FILE *out, bool comma, int depth, const char *for
>>>  static void output_json_key_string(FILE *out, bool comma, int depth,
>>>               const char *key, const char *value)
>>>  {
>>> +     if (!value) {
>>> +             pr_info("No value set for key %s\n", key);
>>> +             return;
>>> +     }
>>> +
>>>       output_json_delimiters(out, comma, depth);
>>>       output_json_string(out, key);
>>>       fputs(": ", out);
>>
>>
>> It looks like this would hide new errors on any of the other fields that
>> output_json_key_string() is called on. Maybe it would be better to only
>> wrap the call to output cpu_desc with the if? If that's the only one
>> that we think is optional, and even better only do it for arm64.
>>
>> I mention this because the test for 'perf data convert' only checks for
>> valid json syntax, but not any fields. So we might want to avoid others
>> going missing.
>
> Makes sense.  Ilkka, can you send v2 with this?

I initially considered the choice James suggested but I kind of thought
that pr_info() might be enough. However, I don't have strong prefence on
either. I'll send an updated version soon.

Cheers, Ilkka

>
> Thanks,
> Namhyung
>