Re: [PATCH] sfc: adjust efx->xdp_tx_queue_count with the real number of initialized queues

[Edward Cree <ecree.xilinx@gmail.com>]

On 29/04/2021 15:49, Ignat Korchagin wrote:
> On Thu, Apr 29, 2021 at 3:22 PM Edward Cree <ecree.xilinx@gmail.com> wrote:
>>
>> On 27/04/2021 22:09, Ignat Korchagin wrote:
>>> +     if (xdp_queue_number)
>> Wait, why is this guard condition needed?
>> What happens if we had nonzero efx->xdp_tx_queue_count initially, but we end up
>>  with no TXQs available for XDP at all (so xdp_queue_number == 0)?
>>
>> -ed
> 
> My thoughts were: efx->xdp_tx_queue_count is originally used to
> allocate efx->xdp_tx_queues.
> So, if xdp_queue_number ends up being 0, we should keep
> efx->xdp_tx_queue_count positive not
> to forget to release efx->xdp_tx_queues (because most checks are
> efx->xdp_tx_queue_count && efx->xdp_tx_queues).
Well, we allocated it in this function, so could we not just free it
 (and NULL it) if we get here with xdp_queue_number == 0?
Assuming it even makes sense for those checks to be that conjunction,
 and not just efx->xdp_tx_queues.

> I'm not familiar enough with SFC internals to definitely say if it is
> even possible to have
> xdp_queue_number == 0 while having efx->xdp_tx_queue_count > 0
If it's possible for us to get xdp_queue_number != efx->xdp_tx_queue_count
 at all (which I can't remember exactly how it happens, but I think it's a
 case of not getting as many VIs back from firmware as we wanted, which
 happens after the initial determination of numbers of queues & channels),
 then it's possible that our number of available TXQs is reduced far
 enough that we don't have any left for XDP.
At least, I think so; this part of the driver confuses me too :S

-ed