diff options
| author | Harshad Shirwadkar <harshadshirwadkar@gmail.com> | 2019-10-05 21:59:27 +0300 |
|---|---|---|
| committer | Greg Kroah-Hartman <gregkh@linuxfoundation.org> | 2019-10-17 23:45:16 +0300 |
| commit | 345c03a0defb07505c3c099bf68f30187f00dae2 (patch) | |
| tree | b24e895a096ca94115ee4f4f7a09f3ab3cfbbd12 /block/blk-rq-qos.c | |
| parent | d855a5f2dec7d2f48a7270abf92dfb3a46b5a581 (diff) | |
| download | linux-345c03a0defb07505c3c099bf68f30187f00dae2.tar.xz | |
blk-wbt: fix performance regression in wbt scale_up/scale_down
commit b84477d3ebb96294f87dc3161e53fa8fe22d9bfd upstream.
scale_up wakes up waiters after scaling up. But after scaling max, it
should not wake up more waiters as waiters will not have anything to
do. This patch fixes this by making scale_up (and also scale_down)
return when threshold is reached.
This bug causes increased fdatasync latency when fdatasync and dd
conv=sync are performed in parallel on 4.19 compared to 4.14. This
bug was introduced during refactoring of blk-wbt code.
Fixes: a79050434b45 ("blk-rq-qos: refactor out common elements of blk-wbt")
Cc: stable@vger.kernel.org
Cc: Josef Bacik <jbacik@fb.com>
Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Diffstat (limited to 'block/blk-rq-qos.c')
| -rw-r--r-- | block/blk-rq-qos.c | 14 |
1 files changed, 9 insertions, 5 deletions
diff --git a/block/blk-rq-qos.c b/block/blk-rq-qos.c index 0005dfd568dd..43bcd4e7a7f9 100644 --- a/block/blk-rq-qos.c +++ b/block/blk-rq-qos.c @@ -148,24 +148,27 @@ bool rq_depth_calc_max_depth(struct rq_depth *rqd) return ret; } -void rq_depth_scale_up(struct rq_depth *rqd) +/* Returns true on success and false if scaling up wasn't possible */ +bool rq_depth_scale_up(struct rq_depth *rqd) { /* * Hit max in previous round, stop here */ if (rqd->scaled_max) - return; + return false; rqd->scale_step--; rqd->scaled_max = rq_depth_calc_max_depth(rqd); + return true; } /* * Scale rwb down. If 'hard_throttle' is set, do it quicker, since we - * had a latency violation. + * had a latency violation. Returns true on success and returns false if + * scaling down wasn't possible. */ -void rq_depth_scale_down(struct rq_depth *rqd, bool hard_throttle) +bool rq_depth_scale_down(struct rq_depth *rqd, bool hard_throttle) { /* * Stop scaling down when we've hit the limit. This also prevents @@ -173,7 +176,7 @@ void rq_depth_scale_down(struct rq_depth *rqd, bool hard_throttle) * keep up. */ if (rqd->max_depth == 1) - return; + return false; if (rqd->scale_step < 0 && hard_throttle) rqd->scale_step = 0; @@ -182,6 +185,7 @@ void rq_depth_scale_down(struct rq_depth *rqd, bool hard_throttle) rqd->scaled_max = false; rq_depth_calc_max_depth(rqd); + return true; } void rq_qos_exit(struct request_queue *q) |