Commit 7ca5b2e6 authored by Tejun Heo's avatar Tejun Heo Committed by Jens Axboe

blk-iocost: streamline vtime margin and timer slack handling

The margin handling was pretty inconsistent.

* ioc->margin_us and ioc->inuse_margin_vtime were used as vtime margin
  thresholds. However, the two are in different units with the former
  requiring conversion to vtime on use.

* iocg_kick_waitq() was using a quarter of WAITQ_TIMER_MARGIN_PCT of
  period_us as the timer slack - ~1.2%. While iocg_kick_delay() was using a
  quarter of ioc->margin_us - ~12.5%. There aren't strong reasons to use
  different values for the two.

This patch cleans up margin and timer slack handling:

* vtime margins are now recorded in ioc->margins.{min, max} on period
  duration changes and used consistently.

* Timer slack is now 1% of period_us and recorded in ioc->timer_slack_ns and
  used consistently for iocg_kick_waitq() and iocg_kick_delay().

The only functional change is shortening of timer slack. No meaningful
visible change is expected.
Signed-off-by: default avatarTejun Heo <tj@kernel.org>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
parent ce95570a
......@@ -221,11 +221,11 @@ enum {
* serves as its IO credit buffer. Surplus weight adjustment is
* immediately canceled if the vtime margin runs below 10%.
*/
MARGIN_PCT = 50,
INUSE_MARGIN_PCT = 10,
MARGIN_MIN_PCT = 10,
MARGIN_MAX_PCT = 50,
/* Have some play in waitq timer operations */
WAITQ_TIMER_MARGIN_PCT = 5,
/* Have some play in timer operations */
TIMER_SLACK_PCT = 1,
/*
* vtime can wrap well within a reasonable uptime when vrate is
......@@ -374,6 +374,11 @@ struct ioc_params {
u32 too_slow_vrate_pct;
};
struct ioc_margins {
s64 min;
s64 max;
};
struct ioc_missed {
local_t nr_met;
local_t nr_missed;
......@@ -395,8 +400,9 @@ struct ioc {
bool enabled;
struct ioc_params params;
struct ioc_margins margins;
u32 period_us;
u32 margin_us;
u32 timer_slack_ns;
u64 vrate_min;
u64 vrate_max;
......@@ -415,7 +421,6 @@ struct ioc {
atomic64_t cur_period; /* inc'd each period */
int busy_level; /* saturation history */
u64 inuse_margin_vtime;
bool weights_updated;
atomic_t hweight_gen; /* for lazy hweights */
......@@ -678,6 +683,16 @@ static void iocg_commit_bio(struct ioc_gq *iocg, struct bio *bio, u64 cost)
#define CREATE_TRACE_POINTS
#include <trace/events/iocost.h>
static void ioc_refresh_margins(struct ioc *ioc)
{
struct ioc_margins *margins = &ioc->margins;
u32 period_us = ioc->period_us;
u64 vrate = atomic64_read(&ioc->vtime_rate);
margins->min = (period_us * MARGIN_MIN_PCT / 100) * vrate;
margins->max = (period_us * MARGIN_MAX_PCT / 100) * vrate;
}
/* latency Qos params changed, update period_us and all the dependent params */
static void ioc_refresh_period_us(struct ioc *ioc)
{
......@@ -711,9 +726,10 @@ static void ioc_refresh_period_us(struct ioc *ioc)
/* calculate dependent params */
ioc->period_us = period_us;
ioc->margin_us = period_us * MARGIN_PCT / 100;
ioc->inuse_margin_vtime = DIV64_U64_ROUND_UP(
period_us * VTIME_PER_USEC * INUSE_MARGIN_PCT, 100);
ioc->timer_slack_ns = div64_u64(
(u64)period_us * NSEC_PER_USEC * TIMER_SLACK_PCT,
100);
ioc_refresh_margins(ioc);
}
static int ioc_autop_idx(struct ioc *ioc)
......@@ -1031,7 +1047,7 @@ static bool iocg_activate(struct ioc_gq *iocg, struct ioc_now *now)
{
struct ioc *ioc = iocg->ioc;
u64 last_period, cur_period, max_period_delta;
u64 vtime, vmargin, vmin;
u64 vtime, vmin;
int i;
/*
......@@ -1077,8 +1093,7 @@ static bool iocg_activate(struct ioc_gq *iocg, struct ioc_now *now)
*/
max_period_delta = DIV64_U64_ROUND_UP(VTIME_VALID_DUR, ioc->period_us);
vtime = atomic64_read(&iocg->vtime);
vmargin = ioc->margin_us * now->vrate;
vmin = now->vnow - vmargin;
vmin = now->vnow - ioc->margins.max;
if (last_period + max_period_delta < cur_period ||
time_before64(vtime, vmin)) {
......@@ -1121,8 +1136,6 @@ static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
struct ioc *ioc = iocg->ioc;
struct blkcg_gq *blkg = iocg_to_blkg(iocg);
u64 vtime = atomic64_read(&iocg->vtime);
u64 vmargin = ioc->margin_us * now->vrate;
u64 margin_ns = ioc->margin_us * NSEC_PER_USEC;
u64 delta_ns, expires, oexpires;
u32 hw_inuse;
......@@ -1142,7 +1155,7 @@ static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
return false;
}
if (!atomic_read(&blkg->use_delay) &&
time_before_eq64(vtime, now->vnow + vmargin))
time_before_eq64(vtime, now->vnow + ioc->margins.max))
return false;
/* use delay */
......@@ -1154,11 +1167,11 @@ static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
/* if already active and close enough, don't bother */
oexpires = ktime_to_ns(hrtimer_get_softexpires(&iocg->delay_timer));
if (hrtimer_is_queued(&iocg->delay_timer) &&
abs(oexpires - expires) <= margin_ns / 4)
abs(oexpires - expires) <= ioc->timer_slack_ns)
return true;
hrtimer_start_range_ns(&iocg->delay_timer, ns_to_ktime(expires),
margin_ns / 4, HRTIMER_MODE_ABS);
ioc->timer_slack_ns, HRTIMER_MODE_ABS);
return true;
}
......@@ -1206,8 +1219,6 @@ static void iocg_kick_waitq(struct ioc_gq *iocg, struct ioc_now *now)
{
struct ioc *ioc = iocg->ioc;
struct iocg_wake_ctx ctx = { .iocg = iocg };
u64 margin_ns = (u64)(ioc->period_us *
WAITQ_TIMER_MARGIN_PCT / 100) * NSEC_PER_USEC;
u64 vdebt, vshortage, expires, oexpires;
s64 vbudget;
u32 hw_inuse;
......@@ -1243,20 +1254,20 @@ static void iocg_kick_waitq(struct ioc_gq *iocg, struct ioc_now *now)
if (WARN_ON_ONCE(ctx.vbudget >= 0))
return;
/* determine next wakeup, add a quarter margin to guarantee chunking */
/* determine next wakeup, add a timer margin to guarantee chunking */
vshortage = -ctx.vbudget;
expires = now->now_ns +
DIV64_U64_ROUND_UP(vshortage, now->vrate) * NSEC_PER_USEC;
expires += margin_ns / 4;
expires += ioc->timer_slack_ns;
/* if already active and close enough, don't bother */
oexpires = ktime_to_ns(hrtimer_get_softexpires(&iocg->waitq_timer));
if (hrtimer_is_queued(&iocg->waitq_timer) &&
abs(oexpires - expires) <= margin_ns / 4)
abs(oexpires - expires) <= ioc->timer_slack_ns)
return;
hrtimer_start_range_ns(&iocg->waitq_timer, ns_to_ktime(expires),
margin_ns / 4, HRTIMER_MODE_ABS);
ioc->timer_slack_ns, HRTIMER_MODE_ABS);
}
static enum hrtimer_restart iocg_waitq_timer_fn(struct hrtimer *timer)
......@@ -1399,7 +1410,7 @@ static void ioc_timer_fn(struct timer_list *timer)
/* calc usages and see whether some weights need to be moved around */
list_for_each_entry(iocg, &ioc->active_iocgs, active_list) {
u64 vdone, vtime, vusage, vmargin, vmin;
u64 vdone, vtime, vusage, vmin;
u32 hw_active, hw_inuse, usage;
/*
......@@ -1450,8 +1461,7 @@ static void ioc_timer_fn(struct timer_list *timer)
}
/* see whether there's surplus vtime */
vmargin = ioc->margin_us * now.vrate;
vmin = now.vnow - vmargin;
vmin = now.vnow - ioc->margins.max;
iocg->has_surplus = false;
......@@ -1623,8 +1633,7 @@ static void ioc_timer_fn(struct timer_list *timer)
nr_surpluses);
atomic64_set(&ioc->vtime_rate, vrate);
ioc->inuse_margin_vtime = DIV64_U64_ROUND_UP(
ioc->period_us * vrate * INUSE_MARGIN_PCT, 100);
ioc_refresh_margins(ioc);
} else if (ioc->busy_level != prev_busy_level || nr_lagging) {
trace_iocost_ioc_vrate_adj(ioc, atomic64_read(&ioc->vtime_rate),
missed_ppm, rq_wait_pct, nr_lagging,
......@@ -1754,7 +1763,7 @@ static void ioc_rqos_throttle(struct rq_qos *rqos, struct bio *bio)
current_hweight(iocg, &hw_active, &hw_inuse);
if (hw_inuse < hw_active &&
time_after_eq64(vtime + ioc->inuse_margin_vtime, now.vnow)) {
time_after_eq64(vtime + ioc->margins.min, now.vnow)) {
TRACE_IOCG_PATH(inuse_reset, iocg, &now,
iocg->inuse, iocg->weight, hw_inuse, hw_active);
spin_lock_irq(&ioc->lock);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment