Re: [PATCH net 3/3] bnxt_en: Update MRU and RSS table of RSS contexts on queue reset

[Michael Chan <michael.chan@broadcom.com>]

[-- Attachment #1: Type: text/plain, Size: 1349 bytes --]

On Tue, May 20, 2025 at 6:51 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Tue, 20 May 2025 18:38:45 -0700 Michael Chan wrote:
> > On Tue, May 20, 2025 at 6:28 PM Jakub Kicinski <kuba@kernel.org> wrote:
> > > On Mon, 19 May 2025 13:41:30 -0700 Michael Chan wrote:
> > > > @@ -15987,6 +16005,7 @@ static int bnxt_queue_stop(struct net_device *dev, void *qmem, int idx)
> > > >
> > > >               bnxt_set_vnic_mru_p5(bp, vnic, 0);
> > > >       }
> > > > +     bnxt_set_rss_ctx_vnic_mru(bp, 0);
> > > >       /* Make sure NAPI sees that the VNIC is disabled */
> > > >       synchronize_net();
> > > >       rxr = &bp->rx_ring[idx];
> > >
> > > What does setting MRU to zero do? All traffic will be dropped?
> > > Traffic will no longer be filtered based on MRU?  Or.. ?
> >
> > That VNIC with MRU set to zero will not receive any more traffic.
> > This step was recommended by the FW team when we first started working
> > with David to implement the queue_mgmt_ops.
>
> Shutting down traffic to ZC queues is one thing, but now you
> seem to be walking all RSS contexts and shutting them all down.
> The whole point of the queue API is to avoid shutting down
> the entire device.

The existing code has been setting the MRU to 0 for the default RSS
context's VNIC.  They found that this sequence was reliable.

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4196 bytes --]