blk: optimization for classic polling
authorNitesh Shetty <nj.shetty@samsung.com>
Tue, 13 Feb 2018 15:48:12 +0000 (21:18 +0530)
committerJens Axboe <axboe@kernel.dk>
Tue, 13 Feb 2018 16:12:04 +0000 (09:12 -0700)
commit67b4110f8c8d16e588d7730db8e8b01b32c1bd8b
treeb8564dcbfef53bae1b43777ff84e90abb9d61b72
parent178e834c47b0d01352c48730235aae69898fbc02
blk: optimization for classic polling

This removes the dependency on interrupts to wake up task. Set task
state as TASK_RUNNING, if need_resched() returns true,
while polling for IO completion.
Earlier, polling task used to sleep, relying on interrupt to wake it up.
This made some IO take very long when interrupt-coalescing is enabled in
NVMe.

Reference:
http://lists.infradead.org/pipermail/linux-nvme/2018-February/015435.html

Changes since v2->v3:
-using __set_current_state() instead of set_current_state()

Changes since v1->v2:
-setting task state once in blk_poll, instead of multiple
callers.

Signed-off-by: Nitesh Shetty <nj.shetty@samsung.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/blk-mq.c