[Xen-devel] Re: [PATCH] Add a new style of passing GSO packets to frontends.

[Ian Campbell <Ian.Campbell@xxxxxxxxxxxxx>]

On Sat, 2010-07-03 at 08:22 +0100, Jeremy Fitzhardinge wrote:
> On 07/02/2010 10:28 AM, Paul Durrant wrote:
> > feature-gso-tcpv4-prefix uses precedes the packet data passed to
> > the frontend with a ring entry that contains the necessary
> > metadata. This style of GSO passing is required for Citrix
> > Windows PV Drivers.
> >
> > Signed-off-by: Paul Durrant <paul.durrant@xxxxxxxxxx>
> > Cc: Ian Campbell <ian.campbell@xxxxxxxxxx>
> > ---
> >  drivers/xen/netback/common.h     |    3 +-
> >  drivers/xen/netback/netback.c    |   43 
> > ++++++++++++++++++++++++++++++++++---
> >  drivers/xen/netback/xenbus.c     |   17 +++++++++++---
> >  include/xen/interface/io/netif.h |    4 +++
> >  4 files changed, 58 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/xen/netback/common.h b/drivers/xen/netback/common.h
> > index 857778c..1cbc4ff 100644
> > --- a/drivers/xen/netback/common.h
> > +++ b/drivers/xen/netback/common.h
> > @@ -82,7 +82,8 @@ struct xen_netif {
> >     int smart_poll;
> >  
> >     /* Internal feature information. */
> > -   u8 can_queue:1; /* can queue packets for receiver? */
> > +   u8 can_queue:1;     /* can queue packets for receiver? */
> > +   u8 gso_prefix:1;    /* use a prefix segment for GSO information */
> >  
> >     /* Allow netif_be_start_xmit() to peek ahead in the rx request
> >      * ring.  This is a prediction of what rx_req_cons will be once
> > diff --git a/drivers/xen/netback/netback.c b/drivers/xen/netback/netback.c
> > index c8f5c1b..93f0686 100644
> > --- a/drivers/xen/netback/netback.c
> > +++ b/drivers/xen/netback/netback.c
> > @@ -313,8 +313,12 @@ int netif_be_start_xmit(struct sk_buff *skb, struct 
> > net_device *dev)
> >  
> >     netbk = &xen_netbk[netif->group];
> >  
> > +   /* Drop the packet if the netif is not up or there is no carrier. */
> > +   if (unlikely(!netif_schedulable(netif)))
> > +           goto drop;
> > +
> >     /* Drop the packet if the target domain has no receive buffers. */
> > -   if (unlikely(!netif_schedulable(netif) || netbk_queue_full(netif)))
> > +   if (unlikely(netbk_queue_full(netif)))
> >             goto drop;
> >   
> 
> Are these related to the gso negotiation or a separate fix?  If they're
> separate, could I have it as a separate patch with its own description
> of the change (and if not, perhaps some comment about how this relates
> to the rest of the patch)?

I think it is just splitting the existing || clause into two separate if
statements with their own descriptive comment? IOW it's an unrelated
cleanup I guess?

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel