JakeHillion/scx-upstream
1
0
Fork 0
mirror of https://github.com/sched-ext/scx.git synced 2024-11-24 20:00:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

scx_bpfland: always honor average nvcsw in lowlatency mode

Browse Source 
Keep evaluating the average number of voluntary context switches for
each task when lowlatency mode is enabled, even when interactive tasks
classification is disabled (via `-c 0`).

The average nvcsw is also used in lowlatency mode to evaluate the
proportional bonus to the tasks' deadline and it shouldn't be ignored
when interactive tasks classification is disabled. Moreover, make sure
that such bonus never exceeds the starvation threshold.

Keep in mind that it is still possible to disable the periodic average
nvcsw evaluation with `-c 0`, without specifying `--lowlatency`.

Fixes: 6a22853 ("scx_bpfland: introduce --lowlatency option")
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
This commit is contained in:
Andrea Righi 2024-08-23 11:14:47 +02:00
parent e238e01e8e
commit 5a08855a86
1 changed files with 10 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
scheds/rust/scx_bpfland/src/bpf/main.bpf.c
View File

@ -384,8 +384,7 @@ static bool task_avg_nvcsw(struct task_struct *p)
*/
static inline u64 task_deadline(struct task_struct *p)
{
u64 dl_boost = lowlatency ?
MIN(task_avg_nvcsw(p), nvcsw_max_thresh) * slice_ns : 0;
u64 dl_boost = lowlatency ? task_avg_nvcsw(p) * slice_ns : 0;
/*
* Limit the vruntime to (vtime_now - slice_ns_lag) to avoid
@ -406,8 +405,11 @@ static inline u64 task_deadline(struct task_struct *p)
* Return the task's deadline as its vruntime, with a bonus that is
* proportional to the task's average number of voluntary context
* switches.
*
* Also make sure the bonus is limited to the starvation threshold (to
* prevent starvation).
*/
return p->scx.dsq_vtime - dl_boost;
return p->scx.dsq_vtime - MIN(dl_boost, starvation_thresh_ns);
}
/*
@ -1062,18 +1064,19 @@ void BPF_STRUCT_OPS(bpfland_stopping, struct task_struct *p, bool runnable)
* Evaluate the average number of voluntary context switches per second
* using an exponentially weighted moving average, see calc_avg().
*/
if (!lowlatency && !is_nvcsw_enabled())
return;
delta_t = (s64)(now - tctx->nvcsw_ts);
if (is_nvcsw_enabled() && delta_t > NSEC_PER_SEC) {
if (delta_t > NSEC_PER_SEC) {
u64 delta_nvcsw = p->nvcsw - tctx->nvcsw;
u64 avg_nvcsw = delta_nvcsw * NSEC_PER_SEC / delta_t;
/*
* Evaluate the average nvcsw for the task, limited to the
* range [0 .. nvcsw_max_thresh * 8] to prevent excessive
* spikes.
* range [0 .. 1000] to prevent excessive spikes.
*/
tctx->avg_nvcsw = calc_avg_clamp(tctx->avg_nvcsw, avg_nvcsw,
0, nvcsw_max_thresh << 3);
0, MAX(nvcsw_max_thresh, 1000));
tctx->nvcsw = p->nvcsw;
tctx->nvcsw_ts = now;
@ -1098,10 +1101,6 @@ void BPF_STRUCT_OPS(bpfland_stopping, struct task_struct *p, bool runnable)
* Reresh task status: interactive or regular.
*/
update_task_interactive(tctx);
dbg_msg("%d (%s) avg_nvcsw = %llu [%s]",
p->pid, p->comm, tctx->avg_nvcsw,
tctx->avg_nvcsw < nvcsw_avg_thresh ? "regular" : "interactive");
}
}