[Intel-gfx] [PATCH i-g-t] intel_gpu_top: Fix cleanup on old kernels / unsupported GPU

Das, Nirmoy nirmoy.das at linux.intel.com
Mon Jan 23 11:09:26 UTC 2023


Hi Tvrtko,

On 1/23/2023 11:24 AM, Tvrtko Ursulin wrote:
>
> On 20/01/2023 14:29, Das, Nirmoy wrote:
>> Hi Tvrtko,
>>
>> On 1/19/2023 4:54 PM, Tvrtko Ursulin wrote:
>>> From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>>>
>>> Avoid trying to dereference null engines on exit when there are either
>>> none which are supported, or kernel does not have i915 PMU support.
>>>
>>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>>> ---
>>>   tools/intel_gpu_top.c | 10 +++++++---
>>>   1 file changed, 7 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/tools/intel_gpu_top.c b/tools/intel_gpu_top.c
>>> index 7aa233570463..517dc2995d26 100644
>>> --- a/tools/intel_gpu_top.c
>>> +++ b/tools/intel_gpu_top.c
>>> @@ -448,6 +448,9 @@ static void free_engines(struct engines *engines)
>>>       };
>>>       unsigned int i;
>>> +    if (!engines)
>>> +        return;
>>> +
>>
>>
>> This check should happen before the  struct pmu_counter free_list 
>> generation.
>
> I could, but it doesn't have to, not sure what reasons for should you 
> have in mind?

In my tree, I see:

"

static void free_engines(struct engines *engines)

{
         struct pmu_counter **pmu, *free_list[] = {
                 &engines->r_gpu,
                 &engines->r_pkg,
                 &engines->imc_reads,
                 &engines->imc_writes,
                 NULL
         };

         unsigned int i;

"

Maybe I am missing something here, wouldn't those dereferences of 
"engines" segfault if the engines  is NULL.


Regards,

Nirmoy

>
> Regards,
>
> Tvrtko
>
>>
>> Regards,
>>
>> Nirmoy
>>
>>>       for (pmu = &free_list[0]; *pmu; pmu++) {
>>>           if ((*pmu)->present)
>>>               free((char *)(*pmu)->units);
>>> @@ -2568,7 +2571,7 @@ int main(int argc, char **argv)
>>>               "Failed to detect engines! (%s)\n(Kernel 4.16 or newer 
>>> is required for i915 PMU support.)\n",
>>>               strerror(errno));
>>>           ret = EXIT_FAILURE;
>>> -        goto err;
>>> +        goto err_engines;
>>>       }
>>>       ret = pmu_init(engines);
>>> @@ -2585,7 +2588,7 @@ int main(int argc, char **argv)
>>>   "More information can be found at 'Perf events and tool security' 
>>> document:\n"
>>> "https://www.kernel.org/doc/html/latest/admin-guide/perf-security.html\n");
>>>           ret = EXIT_FAILURE;
>>> -        goto err;
>>> +        goto err_pmu;
>>>       }
>>>       ret = EXIT_SUCCESS;
>>> @@ -2699,8 +2702,9 @@ int main(int argc, char **argv)
>>>           free_clients(clients);
>>>       free(codename);
>>> -err:
>>> +err_pmu:
>>>       free_engines(engines);
>>> +err_engines:
>>>       free(pmu_device);
>>>   exit:
>>>       igt_devices_free();


More information about the Intel-gfx mailing list