From: Tejun Heo <tj@kernel.org>
To: Shaohua Li <shli@fb.com>
Cc: linux-block@vger.kernel.org, linux-kernel@vger.kernel.org,
kernel-team@fb.com, axboe@fb.com, vgoyal@redhat.com
Subject: Re: [PATCH V5 10/17] blk-throttle: make bandwidth change smooth
Date: Mon, 9 Jan 2017 15:28:44 -0500 [thread overview]
Message-ID: <20170109202844.GN12827@mtj.duckdns.org> (raw)
In-Reply-To: <390c68366acef5f3ce6ac6c5ce868826f07fd993.1481833017.git.shli@fb.com>
Hello,
On Thu, Dec 15, 2016 at 12:33:01PM -0800, Shaohua Li wrote:
> static uint64_t tg_bps_limit(struct throtl_grp *tg, int rw)
> {
> struct blkcg_gq *blkg = tg_to_blkg(tg);
> + struct throtl_data *td;
> uint64_t ret;
>
> if (cgroup_subsys_on_dfl(io_cgrp_subsys) && !blkg->parent)
> return U64_MAX;
> - return tg->bps[rw][tg->td->limit_index];
> +
> + td = tg->td;
> + ret = tg->bps[rw][td->limit_index];
> + if (td->limit_index == LIMIT_MAX && tg->bps[rw][LIMIT_LOW] !=
> + tg->bps[rw][LIMIT_MAX]) {
> + uint64_t increase;
> +
> + if (td->scale < 4096 && time_after_eq(jiffies,
Hmm... why do we need to limit scale to 4096? As 4096 is a big
number, this is only theoretical but this means that if max is more
then 2048 times low, that will never be reached, right?
> + td->low_upgrade_time + td->scale * td->throtl_slice)) {
> + unsigned int time = jiffies - td->low_upgrade_time;
> +
> + td->scale = time / td->throtl_slice;
> + }
> + increase = (tg->bps[rw][LIMIT_LOW] >> 1) * td->scale;
> + ret = min(tg->bps[rw][LIMIT_MAX],
> + tg->bps[rw][LIMIT_LOW] + increase);
> + }
> + return ret;
> }
I think the code can use some comments.
> static unsigned int tg_iops_limit(struct throtl_grp *tg, int rw)
> {
> struct blkcg_gq *blkg = tg_to_blkg(tg);
> + struct throtl_data *td;
> unsigned int ret;
>
> if (cgroup_subsys_on_dfl(io_cgrp_subsys) && !blkg->parent)
> return UINT_MAX;
> - return tg->iops[rw][tg->td->limit_index];
> +
> + td = tg->td;
> + ret = tg->iops[rw][td->limit_index];
> + if (td->limit_index == LIMIT_MAX && tg->iops[rw][LIMIT_LOW] !=
> + tg->iops[rw][LIMIT_MAX]) {
> + uint64_t increase;
> +
> + if (td->scale < 4096 && time_after_eq(jiffies,
> + td->low_upgrade_time + td->scale * td->throtl_slice)) {
> + unsigned int time = jiffies - td->low_upgrade_time;
> +
> + td->scale = time / td->throtl_slice;
> + }
> +
> + increase = (tg->iops[rw][LIMIT_LOW] >> 1) * td->scale;
> + ret = min(tg->iops[rw][LIMIT_MAX],
> + tg->iops[rw][LIMIT_LOW] + (unsigned int)increase);
Would it be worthwhile to factor the common part into a helper?
> @@ -1662,6 +1702,13 @@ static void throtl_upgrade_state(struct throtl_data *td)
>
> static void throtl_downgrade_state(struct throtl_data *td, int new)
> {
> + td->scale /= 2;
> +
> + if (td->scale) {
> + td->low_upgrade_time = jiffies - td->scale * td->throtl_slice;
> + return;
> + }
Cool, so linear increase and exponential backdown. Yeah, that makes
sense to me but let's please document it.
Thanks.
--
tejun
next prev parent reply other threads:[~2017-01-09 20:29 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-12-15 20:32 [PATCH V5 00/17] blk-throttle: add .low limit Shaohua Li
2016-12-15 20:32 ` [PATCH V5 01/17] blk-throttle: use U64_MAX/UINT_MAX to replace -1 Shaohua Li
2016-12-15 20:32 ` [PATCH V5 02/17] blk-throttle: prepare support multiple limits Shaohua Li
2016-12-15 20:32 ` [PATCH V5 03/17] blk-throttle: add .low interface Shaohua Li
2017-01-09 16:35 ` Tejun Heo
2016-12-15 20:32 ` [PATCH V5 04/17] blk-throttle: configure bps/iops limit for cgroup in low limit Shaohua Li
2017-01-09 17:35 ` Tejun Heo
2016-12-15 20:32 ` [PATCH V5 05/17] blk-throttle: add upgrade logic for LIMIT_LOW state Shaohua Li
2017-01-09 18:40 ` Tejun Heo
2017-01-09 19:46 ` Tejun Heo
2016-12-15 20:32 ` [PATCH V5 06/17] blk-throttle: add downgrade logic Shaohua Li
2016-12-15 20:32 ` [PATCH V5 07/17] blk-throttle: make sure expire time isn't too big Shaohua Li
2017-01-09 19:54 ` Tejun Heo
2016-12-15 20:32 ` [PATCH V5 08/17] blk-throttle: make throtl_slice tunable Shaohua Li
2017-01-09 20:08 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 09/17] blk-throttle: detect completed idle cgroup Shaohua Li
2017-01-09 20:13 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 10/17] blk-throttle: make bandwidth change smooth Shaohua Li
2017-01-09 20:28 ` Tejun Heo [this message]
2016-12-15 20:33 ` [PATCH V5 11/17] blk-throttle: add a simple idle detection Shaohua Li
2017-01-09 20:56 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 12/17] blk-throttle: add interface to configure idle time threshold Shaohua Li
2017-01-09 20:58 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 13/17] blk-throttle: ignore idle cgroup limit Shaohua Li
2017-01-09 21:01 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 14/17] blk-throttle: add interface for per-cgroup target latency Shaohua Li
2017-01-09 21:14 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 15/17] block: track request size in blk_issue_stat Shaohua Li
2016-12-16 2:01 ` kbuild test robot
2017-01-09 21:17 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 16/17] blk-throttle: add a mechanism to estimate IO latency Shaohua Li
2017-01-09 21:39 ` Tejun Heo
2016-12-15 20:33 ` [PATCH V5 17/17] blk-throttle: add latency target support Shaohua Li
2017-01-09 21:46 ` [PATCH V5 00/17] blk-throttle: add .low limit Tejun Heo
2017-01-09 22:27 ` Shaohua Li
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170109202844.GN12827@mtj.duckdns.org \
--to=tj@kernel.org \
--cc=axboe@fb.com \
--cc=kernel-team@fb.com \
--cc=linux-block@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=shli@fb.com \
--cc=vgoyal@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).