[openib-general] [PATCH][UVERBS][RFC] Exporting device node_type to user mode

[Tom Tucker]

Roland:

Thinking about this a little more and having read some less than
flattering commentary on various mailing list about sysfs, ABI,
differences between distros, etc... Is using sysfs for device attributes
the right approach here, or should be bite the bullet and update the
kernel-abi?


On Thu, 2006-04-20 at 19:57 -0500, Tom Tucker wrote:
> On Thu, 2006-04-20 at 14:13 -0700, Roland Dreier wrote:
> >     Tom> In order to support transport independent behavior for
> >     Tom> user-mode RDMA CMA clients we need to export the node_type to
> >     Tom> the user mode device attributes structure. The reason for
> >     Tom> this is that the user-mode CMA needs to behave differently
> >     Tom> for iWARP vs. IB transports when migrating QP state at
> >     Tom> connection setup and tear down.
> > 
> > Adding node_type to the libibverbs API is OK (for the 1.1 release
> > series...), but I think it would be better to read the existing
> > /sys/class/infiniband/<xxx>/node_type field in sysfs rather than
> > adding it in to the query stuff.
> > 
> 
> Ok -- no problem. Are there rules/guidelines that govern the device
> attributes that belong in sys/class/infiniband vs. attributes that
> belong in device_attr?
> 
> >  - R.
> 
> _______________________________________________
> openib-general mailing list
> openib-general at openib.org
> http://openib.org/mailman/listinfo/openib-general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general