block: strip out locking optimization in put_io_context()
authorTejun Heo <tj@kernel.org>
Tue, 7 Feb 2012 06:51:30 +0000 (07:51 +0100)
committerJens Axboe <axboe@kernel.dk>
Tue, 7 Feb 2012 06:51:30 +0000 (07:51 +0100)
commit11a3122f6cf2d988a77eb8883d0fc49cd013a6d5
treeded8ea8a2982754ff0c58448a7ed2e59487104cb
parent822bfa51ce44f2c63c300fdb76dc99c4d5a5ca9f
block: strip out locking optimization in put_io_context()

put_io_context() performed a complex trylock dancing to avoid
deferring ioc release to workqueue.  It was also broken on UP because
trylock was always assumed to succeed which resulted in unbalanced
preemption count.

While there are ways to fix the UP breakage, even the most
pathological microbench (forced ioc allocation and tight fork/exit
loop) fails to show any appreciable performance benefit of the
optimization.  Strip it out.  If there turns out to be workloads which
are affected by this change, simpler optimization from the discussion
thread can be applied later.

Signed-off-by: Tejun Heo <tj@kernel.org>
LKML-Reference: <1328514611.21268.66.camel@sli10-conroe>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/blk-cgroup.c
block/blk-core.c
block/blk-ioc.c
block/cfq-iosched.c
fs/ioprio.c
include/linux/blkdev.h
include/linux/iocontext.h
kernel/fork.c