[ofw] crash in mlx4 driver

[Leonid Keller]

I've added this field (p_hca_obj) for WinVerbs, because it needs it now
for performing QueryCa.
Before that WinVerbs was using p_hca_dev field, which was wrong, because
this field contains HCA's PDO.
In fact, WinVerbs can perform QueryCa using the existing interface like
IBAL does.
If Sean will change WinVerbs to do that, he can remove this field as
unnecessary.  
 

> -----Original Message-----
> From: ofw-bounces at lists.openfabrics.org 
> [mailto:ofw-bounces at lists.openfabrics.org] On Behalf Of Fab Tillier
> Sent: Friday, March 13, 2009 1:52 AM
> To: Sean Hefty; ofw at lists.openfabrics.org
> Subject: RE: [ofw] crash in mlx4 driver
> 
> > This may be related - I noticed that a change was committed 
> into SVN 
> > which modified struct _RDMA_INTERFACE_VERBS.  However, the 
> neither the 
> > interface GUID or version was updated.  This breaks the 
> existing drivers.
> >
> > typedef struct _RDMA_INTERFACE_VERBS
> > {
> >         INTERFACE               InterfaceHeader;
> >         ci_interface_t  Verbs;
> >         void *                  p_hca_obj;              
> <--- new field
> >
> > }       RDMA_INTERFACE_VERBS;
> >
> > Maybe I missed this patch on the list...  In any case, struct 
> > _ci_interface already contains this exact member.  Is there 
> a reason 
> > to include this in both places?  Can we please revert this 
> change (and 
> > the corresponding changes to winverbs/winmad) to avoid breaking the 
> > interface?
> 
> Note that while ci_interface already contains this member, it 
> is used as the internal CA handle of the HCA driver when 
> retrieving the RDMA_INTERFACE_VERBS, but as a pointer to the 
> PDO when registering the HCA with IBAL.  Note also that 
> there's no comment about the dual use of this field, though 
> there is a comment about being used as a PDO pointer.
> 
> Looking at the comment for the checkin that changed this 
> file, it appears to have been done in relation to discovery 
> of HCAs by IBAL, to remove the convoluted PnP 
> notification/registration mechanisms previously needed for 
> the HCA to discover the IB Bus driver (that was a different 
> devnode).  Looking at the bus driver code the member doesn't 
> seem to be referenced in ibbus, so perhaps this is a stale 
> change that shouldn't have been committed?
> 
> -Fab
> _______________________________________________
> ofw mailing list
> ofw at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
>