On Mon, Feb 05, 2018 at 02:28:02PM -0600, Steve Wise wrote: > > On Mon, Feb 05, 2018 at 09:33:07AM -0600, Steve Wise wrote: > > > > > Perhaps we add a device-unique (or globally unique) identifier as part > of > > > the restrack struct and use that for GET/SET? > > > > What use is an ID if the user can't associate that with something > > meaningful? > > It uniquely identifies the widget you want to GET or SET. Ideally, having > it be something meaningful is nice, but if we want to support GET/SET on > various single objects, we need identifiers. And assigning them as part of > restrack would do the trick. > > However, I don't see GET/SET being useful for QP, CQ, PD, CM_ID, nor MR > resources. Just DUMP. I do see perhaps devices and links having GET/SET. > For instance setting up stuff for RXE and SIW soft devices. Well exactly.. I don't expect GET/SET for restrack objects.. > > For userspace, a tuple of the filehandle and existing the per-file > > handle # would make some sense as a unique ID, but userspace is going > > to have a pointer not necessarily the handle number :( > > I'm not following you here. Who does filehandle/handle# pertain to rdmatool > fetching objects via netlnk and possibly setting them? It is the other side of the question - when rdmatool presents me an object how do I figure out what it is in 'my world' as userspace. Like a PD for instance.. If I have two in a process how do I know which is which compared to a dump? Jason -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
