On 05/03/2016 07:31 AM, Christoph Hellwig wrote:
But running xfstests triggered the following (mlx4
hardware; SRP initiator and LIO target running on the same server and
communicating over loopback):
I can reproduce this, thanks. The issue was that my implementation
of keeping the MRs around when getting an -EAGAIN for setting up new
RDMA R/W contexts wasn't correct. To make it properly work we'd need
a pointer to the send ioctx from the recv ioctx. I don't feel safe
to make this change at this point, and given that force_mr is only
a debug option for the SRP target until RDMA/CM support goes in I
think we should be fine without it. I'll resend the series once I
get feedback from the buildbot, and I'd be happy if you could review
it quickly.
Thanks for the analysis. I'm fine with this patch series going upstream.
Bart.
--
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
