Documentation/admin-guide/cgroup-v2.rst | 9 ++++++ include/linux/misc_cgroup.h | 2 ++ kernel/cgroup/misc.c | 41 +++++++++++++++++++++++++ 3 files changed, 52 insertions(+)
Introduce misc.peak to record the historical maximum usage of the
resource, as in some scenarios the value of misc.max could be
adjusted based on the peak usage of the resource.
Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com>
---
v5: change the type of watermark to atomic64_t
v4: fix the issue of unconditionally updating the watermark
v3: fix while (0)
v2: use cmpxchg to update the watermark
---
Documentation/admin-guide/cgroup-v2.rst | 9 ++++++
include/linux/misc_cgroup.h | 2 ++
kernel/cgroup/misc.c | 41 +++++++++++++++++++++++++
3 files changed, 52 insertions(+)
diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
index ae0fdb6fc618..468a95379009 100644
--- a/Documentation/admin-guide/cgroup-v2.rst
+++ b/Documentation/admin-guide/cgroup-v2.rst
@@ -2646,6 +2646,15 @@ Miscellaneous controller provides 3 interface files. If two misc resources (res_
res_a 3
res_b 0
+ misc.peak
+ A read-only flat-keyed file shown in all cgroups. It shows the
+ historical maximum usage of the resources in the cgroup and its
+ children.::
+
+ $ cat misc.peak
+ res_a 10
+ res_b 8
+
misc.max
A read-write flat-keyed file shown in the non root cgroups. Allowed
maximum usage of the resources in the cgroup and its children.::
diff --git a/include/linux/misc_cgroup.h b/include/linux/misc_cgroup.h
index e799b1f8d05b..ba02e04b7165 100644
--- a/include/linux/misc_cgroup.h
+++ b/include/linux/misc_cgroup.h
@@ -30,11 +30,13 @@ struct misc_cg;
/**
* struct misc_res: Per cgroup per misc type resource
* @max: Maximum limit on the resource.
+ * @watermark: Historical maximum usage of the resource.
* @usage: Current usage of the resource.
* @events: Number of times, the resource limit exceeded.
*/
struct misc_res {
u64 max;
+ atomic64_t watermark;
atomic64_t usage;
atomic64_t events;
};
diff --git a/kernel/cgroup/misc.c b/kernel/cgroup/misc.c
index 79a3717a5803..b92daf5d234d 100644
--- a/kernel/cgroup/misc.c
+++ b/kernel/cgroup/misc.c
@@ -121,6 +121,19 @@ static void misc_cg_cancel_charge(enum misc_res_type type, struct misc_cg *cg,
misc_res_name[type]);
}
+static void misc_cg_update_watermark(struct misc_res *res, u64 new_usage)
+{
+ u64 old;
+
+ while (true) {
+ old = atomic64_read(&res->watermark);
+ if (new_usage <= old)
+ break;
+ if (atomic64_cmpxchg(&res->watermark, old, new_usage) == old)
+ break;
+ }
+}
+
/**
* misc_cg_try_charge() - Try charging the misc cgroup.
* @type: Misc res type to charge.
@@ -159,6 +172,7 @@ int misc_cg_try_charge(enum misc_res_type type, struct misc_cg *cg, u64 amount)
ret = -EBUSY;
goto err_charge;
}
+ misc_cg_update_watermark(res, new_usage);
}
return 0;
@@ -307,6 +321,29 @@ static int misc_cg_current_show(struct seq_file *sf, void *v)
return 0;
}
+/**
+ * misc_cg_peak_show() - Show the peak usage of the misc cgroup.
+ * @sf: Interface file
+ * @v: Arguments passed
+ *
+ * Context: Any context.
+ * Return: 0 to denote successful print.
+ */
+static int misc_cg_peak_show(struct seq_file *sf, void *v)
+{
+ int i;
+ u64 watermark;
+ struct misc_cg *cg = css_misc(seq_css(sf));
+
+ for (i = 0; i < MISC_CG_RES_TYPES; i++) {
+ watermark = atomic64_read(&cg->res[i].watermark);
+ if (READ_ONCE(misc_res_capacity[i]) || watermark)
+ seq_printf(sf, "%s %llu\n", misc_res_name[i], watermark);
+ }
+
+ return 0;
+}
+
/**
* misc_cg_capacity_show() - Show the total capacity of misc res on the host.
* @sf: Interface file
@@ -357,6 +394,10 @@ static struct cftype misc_cg_files[] = {
.name = "current",
.seq_show = misc_cg_current_show,
},
+ {
+ .name = "peak",
+ .seq_show = misc_cg_peak_show,
+ },
{
.name = "capacity",
.seq_show = misc_cg_capacity_show,
--
2.34.1On Tue, 02 Jul 2024 19:36:46 -0500, Xiu Jianfeng <xiujianfeng@huawei.com> wrote: > Introduce misc.peak to record the historical maximum usage of the > resource, as in some scenarios the value of misc.max could be > adjusted based on the peak usage of the resource. > Acked-by: haitao.huang@linux.intel.com BTW, from SGX point of view, I think it also makes sense to show peak usage in the root group. Thanks Haitao
On Wed, Jul 03, 2024 at 12:36:46AM +0000, Xiu Jianfeng wrote: > Introduce misc.peak to record the historical maximum usage of the > resource, as in some scenarios the value of misc.max could be > adjusted based on the peak usage of the resource. > > Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com> Applied to cgroup/for-6.11. Thanks. -- tejun
On 7/3/24 6:06 AM, Xiu Jianfeng wrote:
> Introduce misc.peak to record the historical maximum usage of the
> resource, as in some scenarios the value of misc.max could be
> adjusted based on the peak usage of the resource.
>
[...]
> /**
> * misc_cg_capacity_show() - Show the total capacity of misc res on the host.
> * @sf: Interface file
> @@ -357,6 +394,10 @@ static struct cftype misc_cg_files[] = {
> .name = "current",
> .seq_show = misc_cg_current_show,
> },
> + {
> + .name = "peak",
> + .seq_show = misc_cg_peak_show,
> + },
> {
> .name = "capacity",
> .seq_show = misc_cg_capacity_show,
The patch looks good to me after the atomic conversion. Sorry for bringing up
this question so late into the discussion. Given that misc.max is available
only for non-root cgroups, does it make sense for misc.peak too, available
for non-root cgroups only?
--
Thanks,
Kamalesh
On Wed, Jul 03, 2024 at 03:13:29PM +0530, Kamalesh Babulal wrote: ... > The patch looks good to me after the atomic conversion. Sorry for bringing up > this question so late into the discussion. Given that misc.max is available > only for non-root cgroups, does it make sense for misc.peak too, available > for non-root cgroups only? It's more tied to the usage - misc.current. For memcg, memory.current is only on non-root cgroups, so is peak. For misc, as misc.current exists for the root cgroup, it makes sense for .peak to be there too. Thanks. -- tejun
On 7/3/24 11:33 PM, Tejun Heo wrote: > On Wed, Jul 03, 2024 at 03:13:29PM +0530, Kamalesh Babulal wrote: > ... >> The patch looks good to me after the atomic conversion. Sorry for bringing up >> this question so late into the discussion. Given that misc.max is available >> only for non-root cgroups, does it make sense for misc.peak too, available >> for non-root cgroups only? > > It's more tied to the usage - misc.current. For memcg, memory.current is > only on non-root cgroups, so is peak. For misc, as misc.current exists for > the root cgroup, it makes sense for .peak to be there too. > Thank you so much for explaining in the context of misc controller. -- Thanks, Kamalesh
© 2016 - 2025 Red Hat, Inc.