Can you please open a bugzila bug for this issue? Thanks Tzachi
> -----Original Message----- > From: Usha Srinivasan [mailto:[email protected]] > Sent: Monday, February 08, 2010 11:02 PM > To: Sean Hefty; Tzachi Dar > Cc: 'ofw_list' > Subject: RE: ctrl-c of appl > > I interpreted driver to mean hca driver, hence my confusion. > I agree that this bug is not the fault of the user > application, but lies somewhere between the application and > the hca driver. > > If someone has a fix for this problem in ual (or kal), I will > be happy to test it. > > -----Original Message----- > From: Sean Hefty [mailto:[email protected]] > Sent: Monday, February 08, 2010 3:41 PM > To: Usha Srinivasan; Tzachi Dar > Cc: 'ofw_list' > Subject: RE: ctrl-c of appl > > >Ok, I'm not sure if I agree that the problem I reported below with > >deregistermr not being called is the fault of the driver. > > I believe that Tzachi was simply stating that the problem is > with some kernel driver. I.e. it's not a library or user > space application issue. > > >When I get this bug check, it is because ibal has NOT called > >DeregisterMR; it only calls DestroyQp, DestroyCq and > UmCloseCa. How is > >the driver supposed to determine which MRs to automatically > deregister? > >An application could have created an MR for use with multiple QPs. > >And, I don't think there is a way to connect an MR with a CA > ctx. Feel free to correct me if I'm wrong. > > In this case, resource tracking is done with IBAL, so IBAL > should be responsible for tracking the memory registrations > and deregistering the memory on application failure. > Winverbs has a similar responsibility. > > - Sean > > _______________________________________________ ofw mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
