[ofa-general] [PATCH RFC] RDMA: New Memory Extensions.

[Steve Wise]

Roland Dreier wrote:
> A few quick comments (more later):
> 
>  > - New device capability flag added: IB_DEVICE_MEMORY_EXTENSIONS indicates
>  > device support for this feature.
> 
> We still have time before 2.6.26 comes out.  Rather than moving
> IB_DEVICE_SEND_W_INV to a new bit number, I think it might be better to
> just remove IB_DEVICE_SEND_W_INV and make IB_DEVICE_MEMORY_EXTENSIONS
> (maybe "MM_EXTENSIONS" or "MEMORY_MANAGEMENT_EXTENSIONS" is better?)
> imply real send-with-invalidate support... so 2.6.26 won't have
> send-with-invalidate and 2.6.27 will have all of the IB base MM exts
> (and iWARP equivs) in one capability bit.
> 
> Any thoughts either way?  I'll post a trial balloon patch tomorrow.
> 

Sounds fine to me.  As you've seen, I don't like to type, so 
MM_EXTENSIONS seems better. :)

> 
> Second question -- IB BMME and iWARP talk about a key portion (least
> significant byte) of STag/L_Key/R_Key as being under consumer control.
> Do we want to expose that as part of this API?  Basically it means we
> need to add a way for the consumer to pass in a new L_Key/STag as part
> of a lot of calls.

I left it out from this first pass because we don't expose any of that 
in the existing RDMA API.  Currently the iwarp providers make up their 
own keys.  EG:  ib_reg_phys_mr() should also allow passing in the key, 
at least according to iWARP verbs.  But I don't really see the need...

Steve.