Michael S. Tsirkin wrote:
Make mad module use a single workqueue rather than a per-port
workqueue. This way, we'll have less clutter on systems with
a lot of ports.
Signed-off-by: Michael S. Tsirkin <[EMAIL PROTECTED]>
Thinking about it, why would we *want* a per-port thread?
What do you guys think about the following?
As a bonus, this makes it easier to renice the mad thread
for people that want to do this.
Indeed, today the mad module creates thread per device port, and the cm
module creates thread per cpu, so if the system has 16 cores and two
hcas each with two ports, the IB stack would create 20 threads just for
the sake of the mad and cm modules... I also think it would be good if
the mad module would create one thread and the cm module as well.
Sean - does it make sense to you to change the CM for that matter?
I brought below listing of the kernel threads on an rh4 u3 / ofed 1.2
system.
Or.
root 1 0 0 Jul15 ? 00:00:00 init [5]
root 2 1 0 Jul15 ? 00:00:00 [migration/0]
root 3 1 0 Jul15 ? 00:00:00 [ksoftirqd/0]
root 4 1 0 Jul15 ? 00:00:00 [migration/1]
root 5 1 0 Jul15 ? 00:00:00 [ksoftirqd/1]
root 6 1 0 Jul15 ? 00:00:00 [migration/2]
root 7 1 0 Jul15 ? 00:00:00 [ksoftirqd/2]
root 8 1 0 Jul15 ? 00:00:00 [migration/3]
root 9 1 0 Jul15 ? 00:00:00 [ksoftirqd/3]
root 10 1 0 Jul15 ? 00:00:00 [events/0]
root 11 1 0 Jul15 ? 00:00:00 [events/1]
root 12 1 0 Jul15 ? 00:00:00 [events/2]
root 13 1 0 Jul15 ? 00:00:00 [events/3]
root 14 10 0 Jul15 ? 00:00:00 [khelper]
root 15 10 0 Jul15 ? 00:00:00 [kacpid]
root 60 10 0 Jul15 ? 00:00:00 [kblockd/0]
root 61 10 0 Jul15 ? 00:00:00 [kblockd/1]
root 62 10 0 Jul15 ? 00:00:00 [kblockd/2]
root 63 10 0 Jul15 ? 00:00:00 [kblockd/3]
root 64 1 0 Jul15 ? 00:00:00 [khubd]
root 73 10 0 Jul15 ? 00:00:00 [pdflush]
root 74 10 0 Jul15 ? 00:00:00 [pdflush]
root 76 10 0 Jul15 ? 00:00:00 [aio/0]
root 77 10 0 Jul15 ? 00:00:00 [aio/1]
root 78 10 0 Jul15 ? 00:00:00 [aio/2]
root 79 10 0 Jul15 ? 00:00:00 [aio/3]
root 75 1 0 Jul15 ? 00:00:00 [kswapd0]
root 152 1 0 Jul15 ? 00:00:00 [kseriod]
root 230 12 0 Jul15 ? 00:00:00 [ata/0]
root 231 12 0 Jul15 ? 00:00:00 [ata/1]
root 232 12 0 Jul15 ? 00:00:00 [ata/2]
root 233 12 0 Jul15 ? 00:00:00 [ata/3]
root 239 1 0 Jul15 ? 00:00:00 [scsi_eh_0]
root 267 1 0 Jul15 ? 00:00:01 [kjournald]
root 1372 10 0 Jul15 ? 00:00:00 [ib_mcast]
root 1377 10 0 Jul15 ? 00:00:00 [ib_cm/0]
root 1378 10 0 Jul15 ? 00:00:00 [ib_cm/1]
root 1379 10 0 Jul15 ? 00:00:00 [ib_cm/2]
root 1380 10 0 Jul15 ? 00:00:00 [ib_cm/3]
root 1401 13 0 Jul15 ? 00:00:00 [ipoib]
root 1418 12 0 Jul15 ? 00:00:00 [mthcacatas]
root 1421 13 0 Jul15 ? 00:00:00 [ib_mad1]
root 1422 10 0 Jul15 ? 00:00:00 [ib_mad2]
root 2132 10 0 Jul15 ? 00:00:00 [kauditd]
root 2209 10 0 Jul15 ? 00:00:00 [kmpathd/0]
root 2210 10 0 Jul15 ? 00:00:00 [kmpathd/1]
root 2211 10 0 Jul15 ? 00:00:00 [kmpathd/2]
root 2212 10 0 Jul15 ? 00:00:00 [kmpathd/3]
root 2220 10 0 Jul15 ? 00:00:00 [kmirrord]
root 2221 10 0 Jul15 ? 00:00:00 [kmir_mon]
root 3330 11 0 Jul15 ? 00:00:00 [local_sa]
root 3335 13 0 Jul15 ? 00:00:00 [ib_addr_wq]
root 3340 11 0 Jul15 ? 00:00:00 [iw_cm_wq]
root 3345 12 0 Jul15 ? 00:00:00 [rdma_cm_wq]
root 3350 13 0 Jul15 ? 00:00:00 [sdp]
root 3358 12 0 Jul15 ? 00:00:00 [krdsd]
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
