On Fri, Apr 03, 2009 at 06:28:45PM -0700, Joe Eykholt wrote: > Chris Leech wrote: > > On Fri, Apr 03, 2009 at 03:48:37PM -0700, Joe Eykholt wrote: > >>> -----Original Message----- > >>> From: [email protected]<mailto:[email protected]> > >>> [mailto:[email protected]] On Behalf Of Chris Leech > >>> Sent: Friday, April 03, 2009 3:21 PM > >>> To: [email protected]<mailto:[email protected]> > >>> Subject: [Open-FCoE] [PATCH] fcoe: use ETH_P_FIP for > >>> skb->protocol of FIPframes > >>> > >>> FIP frames should leave the fcoe layer with skb->protocol set > >>> to ETH_P_FIP, not ETH_P_802_3. > >> I think the reason 802_3 was set was to accommodate the tc command, to get > >> the right 803.1p tags. Maybe that's been solved some other way by now. > >> > >> How is skb->protocol used on output? > > > > For the FCoE offload support we wanted traffic set to ETH_P_FCOE so that > > a network driver could tell what type of frame it was without peeking > > into the header. Other than that I think it's just a matter of correctness > > and matching what other protocols do. With FCoE data traffic changed to > > using ETH_P_FCOE, it doesn't make sense to me to not have FIP using > > ETH_P_FIP. > > > > I was never really clear what the issue with the tc filters for DCB was, > > when I looked at it there didn't seem to be a problem with using the > > real ethertype. That may have changed with the use of the skb_edit > > action to assign a transmit queue, if fcoeplumb was previously using a > > action to set the priority bits? > > > > - Chris > > I don't remember either. Maybe it was a much earlier version of tc that > had some limitation. > > Thanks for fixing up fcoeplumb. I was wondering if we should add a filter > to put FIP in the same queues. FIP is less sensitive to drops, but maybe > we should do that anyway. What do you think?
I agree, FIP should be kept in the same traffic class as FCoE. _______________________________________________ devel mailing list [email protected] http://www.open-fcoe.org/mailman/listinfo/devel
