two optimizations in the busy loop wait in class_cleanup():
(1) remove unnecessary spinlocks protecting obd_conn_inprogress
access
(2) replace cond_resched by more suitable yield to unconditionally
yield the CPU if the condition check fails.
Xyratex-bug-id: MRP-1991
Signed-off-by: Alexander Zarochentsev <alexander_zarochentsev@xyratex.com>
Change-Id: I9079b2d04afda7d6db902392c6dcb216706a7ae4
Reviewed-on: http://review.whamcloud.com/11198
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Tested-by: Jenkins
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
}
/* Leave this on forever */
obd->obd_stopping = 1;
+ spin_unlock(&obd->obd_dev_lock);
/* wait for already-arrived-connections to finish. */
- while (obd->obd_conn_inprogress > 0) {
- spin_unlock(&obd->obd_dev_lock);
-
- cond_resched();
-
- spin_lock(&obd->obd_dev_lock);
- }
- spin_unlock(&obd->obd_dev_lock);
+ while (obd->obd_conn_inprogress > 0)
+ yield();
+ smp_rmb();
if (lcfg->lcfg_bufcount >= 2 && LUSTRE_CFG_BUFLEN(lcfg, 1) > 0) {
for (flag = lustre_cfg_string(lcfg, 1); *flag != 0; flag++)