Re: [PATCH for-next] RDMA/efa: Properly handle unexpected AQ completions

["Margolin, Michael" <mrgolin@amazon.com>]

Thanks Gal, we indeed shouldn't see such errors on regular basis and the 
print is sufficient to help tracking and debugging in case it appears.


Michael

On 5/16/2024 1:54 PM, Gal Pressman wrote:
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.
>
>
>
> On 13/05/2024 9:46, Michael Margolin wrote:
>> Do not try to handle admin command completion if it has an unexpected
>> command id and print a relevant error message.
>>
>> Reviewed-by: Firas Jahjah <firasj@amazon.com>
>> Reviewed-by: Yehuda Yitschak <yehuday@amazon.com>
>> Signed-off-by: Michael Margolin <mrgolin@amazon.com>
>> ---
>>   static void efa_com_handle_admin_completion(struct efa_com_admin_queue *aq)
>>   {
>>        struct efa_admin_acq_entry *cqe;
>>        u16 queue_size_mask;
>> -     u16 comp_num = 0;
>> +     u16 comp_cmds = 0;
>>        u8 phase;
>> +     int err;
>>        u16 ci;
>>
>>        queue_size_mask = aq->depth - 1;
>> @@ -453,10 +456,12 @@ static void efa_com_handle_admin_completion(struct efa_com_admin_queue *aq)
>>                 * phase bit was validated
>>                 */
>>                dma_rmb();
>> -             efa_com_handle_single_admin_completion(aq, cqe);
>> +             err = efa_com_handle_single_admin_completion(aq, cqe);
>> +             if (!err)
>> +                     comp_cmds++;
> I would count the unexpected completions as well.
> Regardless, I would definitely add a counter to track these (hopefully)
> rare cases.
>
> Whatever you decide:
> Reviewed-by: Gal Pressman <gal.pressman@linux.dev>