[ofa-general] Re: [PATCH] infiniband: add support for invalidate stag

[Steve Wise]

On Tue, 2007-05-01 at 06:57 +0300, Michael S. Tsirkin wrote:
> > Quoting mhagen at iol.unh.edu <mhagen at iol.unh.edu>:
> > Subject: [PATCH] infiniband: add support for invalidate stag
> > 
> > Patch to add support for the iWARP verbs SEND with INV and SEND with SE
> > and INV.
> > 
> > --- linux-2.6.21.1/include/rdma/ib_verbs.h    2007-04-28
> > 15:35:02.677618096 -0400
> > +++ linux-2.6.21.1/include/rdma/ib_verbs.h    2007-04-28
> > 15:29:16.200290656 -0400
> > @@ -611,7 +611,8 @@ enum ib_send_flags {
> >      IB_SEND_FENCE        = 1,
> >      IB_SEND_SIGNALED    = (1<<1),
> >      IB_SEND_SOLICITED    = (1<<2),
> > -    IB_SEND_INLINE        = (1<<3)
> > +    IB_SEND_INLINE        = (1<<3),
> > +    IB_SEND_INVALIDATE    = (1<<4)
> >  };
> > 
> >  struct ib_sge {
> > @@ -646,6 +647,9 @@ struct ib_send_wr {
> >              u16    pkey_index; /* valid for GSI only */
> >              u8    port_num;   /* valid for DR SMPs on switch only */
> >          } ud;
> > +        struct {
> > +            u32    rkey;
> > +        } invalidate;
> >      } wr;
> >  };
> 
> Shouldn't this rather be part of rc wr?

What's an rc wr?

He added the invalidate struct to the union part of the ib_send_wr.  Its
analogous to the rdma struct in that union in that its additional values
passed in the send wr for an iwarp send w/invalidate and send-se
w/invalidate.  The seems reasonable to me...






>