[PATCH 1/2] drm/amdgpu: individualize amdgpu entity allocation per HW IP
Luben Tuikov
luben.tuikov at amd.com
Wed Jan 22 16:24:30 UTC 2020
On 2020-01-22 4:32 a.m., Nirmoy wrote:
>
> On 1/22/20 12:34 AM, Luben Tuikov wrote:
>>
>>> + for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
>>> + struct amdgpu_ctx_entity *entity = &ctx->entities[i][j];
>>>
>>> - kfree(entity->fences);
>>> - entity->fences = NULL;
>>> + kfree(entity->fences);
>>> + entity->fences = NULL;
>>> + }
>>> +
>>> +error_cleanup_entity_memory:
>> "cleanup" refers to something spilled, or something to be collected.
> It is collecting memory.
But it really is doing a "kfree()". Freeing previously allocated
memory. The traditional working is most accessible, in my opinion.
>> (Or winning in a wagered game of chance.) Also at "module_exit", maybe.
>>
>> The kernel has traditionally used "unroll" and "free" for this. Now, since
>> you're unrolling the loop (notice that it sits after the "return 0;"), then
>> you can backtrack and name it like this:
>>
>> Error_unroll_free1:
>> for ( ; i >= 0; i--)
>> free(my_array[i]);
>>
> I prefer the existing style makes it more readable.
>
>
> Regards,
>
> Nirmoy
>
More information about the amd-gfx
mailing list