Roland Dreier wrote:
OK, I checked in changes to libibverbs and the kernel uverbs to handle
cleaning up stale events when destroying a CQ/QP/SRQ. All the changes
are in svn r3279.
The changes require a kernel ABI bump. The new libibverbs works with
both the old kernel and new kernel, but the old libibverbs will only
work with the old kernel. So in other words, if you upgrade your
kernel, then make sure you upgrade libibverbs as well. If you upgrade
libibverbs, then you don't have to upgrade your kernel but you can if
you want. (Confused yet? Or should I write still more?)
I did some light testing but I don't have any tests that generate lots
of async events. Sean and Arlin, if you could retest uDAPL or
whatever was choking on QP connected events, that would be great.
The regress.sh dapltest set to 40x40 crashes the system. All I see on
the console is a call trace of ib_verbs: ib_uverbs_event_poll+58. I
will try and get more of the oops info.
-arlin
Thanks,
Roland
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
