IB/rdmavt: Use per-CPU reference count for MRs
authorSebastian Sanchez <sebastian.sanchez@intel.com>
Wed, 8 Feb 2017 13:26:31 +0000 (05:26 -0800)
committerDoug Ledford <dledford@redhat.com>
Sun, 19 Feb 2017 14:18:35 +0000 (09:18 -0500)
commit338adfdddf6abe89726e1146ad3102ce9663a634
treee509ec29d80f6cd81097ffea338a8a7222ec7389
parentf3e862cb6894389a35d0beb10f73d62eb3317beb
IB/rdmavt: Use per-CPU reference count for MRs

Having per-CPU reference count for each MR prevents
cache-line bouncing across the system. Thus, it
prevents bottlenecks. Use per-CPU reference counts
per MR.

The per-CPU reference count for FMRs is used in
atomic mode to allow accurate testing of the busy
state. Other MR types run in per-CPU mode MR until
they're freed.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
drivers/infiniband/sw/rdmavt/mr.c
include/rdma/rdmavt_mr.h