| Message ID | YIIo1HuyNmhDeiNx@slm.duckdns.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [for-5.13/block] blk-iocost: don't ignore vrate_min on QD contention | expand |
On 4/22/21 7:54 PM, Tejun Heo wrote: > ioc_adjust_base_vrate() ignored vrate_min when rq_wait_pct indicates that > there is QD contention. The reasoning was that QD depletion always reliably > indicates device saturation and thus it's safe to override user specified > vrate_min. However, this sometimes leads to unnecessary throttling, > especially on really fast devices, because vrate adjustments have delays and > inertia. It also confuses users because the behavior violates the explicitly > specified configuration. > > This patch drops the special case handling so that vrate_min is always > applied. Applied, thanks.
diff --git a/block/blk-iocost.c b/block/blk-iocost.c index 98d656bdb42b7..e0c4baa018578 100644 --- a/block/blk-iocost.c +++ b/block/blk-iocost.c @@ -987,10 +987,6 @@ static void ioc_adjust_base_vrate(struct ioc *ioc, u32 rq_wait_pct, return; } - /* rq_wait signal is always reliable, ignore user vrate_min */ - if (rq_wait_pct > RQ_WAIT_BUSY_PCT) - vrate_min = VRATE_MIN; - /* * If vrate is out of bounds, apply clamp gradually as the * bounds can change abruptly. Otherwise, apply busy_level
ioc_adjust_base_vrate() ignored vrate_min when rq_wait_pct indicates that there is QD contention. The reasoning was that QD depletion always reliably indicates device saturation and thus it's safe to override user specified vrate_min. However, this sometimes leads to unnecessary throttling, especially on really fast devices, because vrate adjustments have delays and inertia. It also confuses users because the behavior violates the explicitly specified configuration. This patch drops the special case handling so that vrate_min is always applied. Signed-off-by: Tejun Heo <tj@kernel.org> --- block/blk-iocost.c | 4 ---- 1 file changed, 4 deletions(-)