2022-08-31 23:08:40

by Josh Don

Subject: [PATCH] cgroup: add pids.peak interface for pids controller

pids.peak tracks the high watermark of usage for number of pids. This
helps give a better baseline on which to set pids.max. Polling
pids.current isn't really feasible, since it would potentially miss
short-lived spikes.

This interface is analogous to memory.peak.

Signed-off-by: Josh Don <[email protected]>
---
kernel/cgroup/pids.c | 37 +++++++++++++++++++++++++++++++++++--
1 file changed, 35 insertions(+), 2 deletions(-)

diff --git a/kernel/cgroup/pids.c b/kernel/cgroup/pids.c
index 511af87f685e..7695e60bcb40 100644
--- a/kernel/cgroup/pids.c
+++ b/kernel/cgroup/pids.c
@@ -47,6 +47,7 @@ struct pids_cgroup {
*/
atomic64_t counter;
atomic64_t limit;
+ int64_t watermark;

/* Handle for "pids.events" */
struct cgroup_file events_file;
@@ -85,6 +86,16 @@ static void pids_css_free(struct cgroup_subsys_state *css)
kfree(css_pids(css));
}

+static void pids_update_watermark(struct pids_cgroup *p, int64_t nr_pids)
+{
+ /*
+ * This is racy, but we don't need perfectly accurate tallying of
+ * the watermark, and this lets us avoid extra atomic overhead.
+ */
+ if (nr_pids > READ_ONCE(p->watermark))
+ WRITE_ONCE(p->watermark, nr_pids);
+}
+
/**
* pids_cancel - uncharge the local pid count
* @pids: the pid cgroup state
@@ -128,8 +139,11 @@ static void pids_charge(struct pids_cgroup *pids, int num)
{
struct pids_cgroup *p;

- for (p = pids; parent_pids(p); p = parent_pids(p))
- atomic64_add(num, &p->counter);
+ for (p = pids; parent_pids(p); p = parent_pids(p)) {
+ int64_t new = atomic64_add_return(num, &p->counter);
+
+ pids_update_watermark(p, new);
+ }
}

/**
@@ -156,6 +170,12 @@ static int pids_try_charge(struct pids_cgroup *pids, int num)
*/
if (new > limit)
goto revert;
+
+ /*
+ * Not technically accurate if we go over limit somewhere up
+ * the hierarchy, but that's tolerable for the watermark.
+ */
+ pids_update_watermark(p, new);
}

return 0;
@@ -311,6 +331,14 @@ static s64 pids_current_read(struct cgroup_subsys_state *css,
return atomic64_read(&pids->counter);
}

+static s64 pids_peak_read(struct cgroup_subsys_state *css,
+ struct cftype *cft)
+{
+ struct pids_cgroup *pids = css_pids(css);
+
+ return READ_ONCE(pids->watermark);
+}
+
static int pids_events_show(struct seq_file *sf, void *v)
{
struct pids_cgroup *pids = css_pids(seq_css(sf));
@@ -331,6 +359,11 @@ static struct cftype pids_files[] = {
.read_s64 = pids_current_read,
.flags = CFTYPE_NOT_ON_ROOT,
},
+ {
+ .name = "peak",
+ .flags = CFTYPE_NOT_ON_ROOT,
+ .read_s64 = pids_peak_read,
+ },
{
.name = "events",
.seq_show = pids_events_show,
--
2.37.2.672.g94769d06f0-goog

2022-09-04 19:30:59

by Tejun Heo

Subject: Re: [PATCH] cgroup: add pids.peak interface for pids controller

On Wed, Aug 31, 2022 at 03:49:03PM -0700, Josh Don wrote:
> pids.peak tracks the high watermark of usage for number of pids. This
> helps give a better baseline on which to set pids.max. Polling
> pids.current isn't really feasible, since it would potentially miss
> short-lived spikes.
>
> This interface is analogous to memory.peak.
>
> Signed-off-by: Josh Don <[email protected]>

Applied to cgroup/for-6.1.

Thanks.

--
tejun