OSDN Git Service

iser-target: Use WQ_UNBOUND for completion workqueue
authorSagi Grimberg <sagig@mellanox.com>
Sun, 25 Jan 2015 17:09:50 +0000 (19:09 +0200)
committerNicholas Bellinger <nab@linux-iscsi.org>
Wed, 4 Feb 2015 18:55:03 +0000 (10:55 -0800)
commit631af550621071d56abe2edbb63d9afd4f4dafcf
tree713dd06e37ab7bebdbfdb62affb1089b3b46bbae
parentf64d2792dde1423af05e661abcf28e7fab9c6e28
iser-target: Use WQ_UNBOUND for completion workqueue

Bound workqueues might be too restrictive since they allow
only a single core per session for processing completions.
WQ_UNBOUND will allow bouncing to another CPU if the running
CPU is currently busy. Luckily, our workqueues are NUMA aware
and will first try to bounce within the same NUMA socket.
My measurements with NULL backend devices show that there is
no (noticeable) additional latency as a result of the change.
I'd expect even to gain performance when working with fast
devices that also allocate MSIX interrupt vectors.

While we're at it, make it WQ_HIGHPRI since processing
completions is really a high priority for performance.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Reported-by: Moussa Ba <moussaba@micron.com>
Signed-off-by: Moussa Ba <moussaba@micron.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
drivers/infiniband/ulp/isert/ib_isert.c