Re: [net-next v39] mctp pcc: Implement MCTP over PCC Transport

[Adam Young]

All the other will be taken in as changes

On 5/6/26 23:02, Jeremy Kerr wrote:
> > +static void mctp_pcc_tx_done(struct mbox_client *c, void *mssg, int r)
> > +{
> > +       struct mctp_pcc_ndev *mctp_pcc_ndev;
> > +       struct sk_buff *skb = mssg;
> > +
> > +       /*
> > +        * If there is a packet in flight during driver cleanup
> > +        * It may have been freed already.
> > +        */
> > +       if (!mssg)
> > +               return;
> > +       /*
> > +        * If the return code is non-zero, we should not report the packet
> > +        * as transmitted.  However, we are in IRQ context right now, and we
> > +        * cannot safely write transmission statistics.
> > +        */
> This reads as if you're not updating stats at all, but you do so in
> mctp_pcc_tx_prepare(). I don't think this comment is necessary - if
> you really want to mention this, add a comment on the
> dev_dstats_tx_add() to indicate why you're calling it early.

This comment is in prep for a fairly large change in the PCC layer to 
address it.

This statistic should be reported in tx_done, but cannot be done safely 
yet.  The fix is to get tx_done out of a hard-irq handler. I will submit 
that as a follow on changes to mailbox/pcc.c and mctp-pc.c