Bug 1458748

Hi Jarod,
Have you reviewed this patch? Are you planning to apply it?
Let me know if there's something you want me to do.

Thanks,
Ram

(In reply to Ram Amrani from comment #2)
> Hi Jarod,
> Have you reviewed this patch? Are you planning to apply it?
> Let me know if there's something you want me to do.

Didn't see the patch. Generally, you want to attach or inline the patch, or at least it's full description, rather than just tack a single-line short hash description to the very end of a comment that is easily missed when triaging your way through a mountain of bugzilla emails.

That said, the patch itself looks fine, but I don't know if it's possible to get it into the release at this point in time. More likely needs to be 7.5 material, with a possible z-stream backport for 7.4 if this is considered critical enough. However, from the patch description, it seems this only relates to a loopback QP, so I suspect it really isn't that critical.

> Didn't see the patch. Generally, you want to attach or inline the patch, or
> at least it's full description, rather than just tack a single-line short
> hash description to the very end of a comment that is easily missed when
> triaging your way through a mountain of bugzilla emails.


I thought that since the rdma-core is a small project in the comparison to the kernel that it'd be enough to share the short form.

> That said, the patch itself looks fine, but I don't know if it's possible to
> get it into the release at this point in time. More likely needs to be 7.5
> material, with a possible z-stream backport for 7.4 if this is considered
> critical enough. However, from the patch description, it seems this only
> relates to a loopback QP, so I suspect it really isn't that critical.

I agree this probably isn't that critical as Open MPI, for example, recognizes loopback and doesn't start RDMA to begin with.
If we stumble into a generic RDMA application that'll be problematic.

Jarod,
Regarding bugs like this - are you just going to take the latest rdma-core for 7.4z and/or 7.5 or are you going to backport specific patches?
If the latter, do you require my assistance in posting them to rkml or do you it yourself?

Thanks,
Ram

(In reply to Ram Amrani from comment #5)
> Jarod,
> Regarding bugs like this - are you just going to take the latest rdma-core
> for 7.4z and/or 7.5 or are you going to backport specific patches?
> If the latter, do you require my assistance in posting them to rkml or do
> you it yourself?
> 
> Thanks,
> Ram

rdma-core is a userspace pkg; only kernel patches go to rkml.

Jarod has uber-privs as pkg owner to modify it at will.  Testing keeps him honest! ;-)

Jarod can chime in differently, or save his breath:
rdma-core is in synch with the kernel now -- it's own little OFED-like world.  So, Jarod backports the rdma-core that is in synch with the backport base of kernel RDMA in the RHEL release.  Bug fixes btwn major releases of rdma-core are picked when deemed necessary for release requirements.

We'll definitely do an updated rdma-core for 7.5, to match up with what we do kernel-wise, like Don said. Anything going back to z-stream is going to be a selective backport, we won't do a major version bump there. This particular bug should be addressed by a 7.5 rebase, but I don't currently see a strong argument to worry about backporting it to 7.4-z as well. I do appreciate particular desired patches/bugfixes being pointed out though, even better if they're tested and known to fix reported bugs.

(In reply to Jarod Wilson from comment #7)
Is this bz resolved in 7.5?

Yes, we actually picked up this fix via the rebase to rdma-core v15 in 7.5.

*** This bug has been marked as a duplicate of bug 1494607 ***