Re: [PATCH net-next] eth: mlx5: link NAPI instances to queues and IRQs

From: Joe Damato
Date: Wed Feb 07 2024 - 09:40:26 EST


On Wed, Feb 07, 2024 at 03:23:47PM +0200, Tariq Toukan wrote:
>
>
> On 06/02/2024 21:23, Joe Damato wrote:
> >On Tue, Feb 06, 2024 at 09:10:27PM +0200, Tariq Toukan wrote:
> >>
> >>
> >>On 06/02/2024 19:12, Joe Damato wrote:
> >>>On Tue, Feb 06, 2024 at 10:11:28AM +0200, Tariq Toukan wrote:
> >>>>
> >>>>
> >>>>On 06/02/2024 3:03, Joe Damato wrote:
> >>>>>Make mlx5 compatible with the newly added netlink queue GET APIs.
> >>>>>
> >>>>>Signed-off-by: Joe Damato <jdamato@xxxxxxxxxx>
>
> ...
>
> >
> >OK, well I tweaked the v3 I had queued based on your feedback. I am
> >definitiely not an mlx5 expert, so I have no idea if it's correct.
> >
> >The changes can be summed up as:
> > - mlx5e_activate_channel and mlx5e_deactivate_channel to use
> > netif_queue_set_napi for each mlx5e_txqsq as it is
> > activated/deactivated. I assumed sq->txq_ix is the correct index, but I
> > have no idea.
> > - mlx5e_activate_qos_sq and mlx5e_deactivate_qos_sq to handle the QOS/HTB
> > case, similar to the above.
> > - IRQ storage removed
> >
> >If you think that sounds vaguely correct, I can send the v3 tomorrow when
> >it has been >24hrs as per Rahul's request.
> >
>
> Sounds correct.
> Please go on and send when it's time so we can review.

OK, I'll send it a bit later today. After looking at it again just now, I
am wondering if the PTP txqsq case needs to be handled, as well.