Re: [PATCH 2/5] docs/cgroup: add entry for cgroup.kill

On Thu, Apr 29, 2021 at 08:22:03PM -0700, Roman Gushchin wrote:

On Thu, Apr 29, 2021 at 02:01:10PM +0200, Christian Brauner wrote:

quoted

From: Christian Brauner <christian.brauner-GeWIH/nMZzLQT0dZR+AlfA@public.gmane.org>

Give a brief overview of the cgroup.kill functionality.

Cc: Roman Gushchin <redacted>
Cc: Tejun Heo <redacted>
Cc: cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Signed-off-by: Christian Brauner <christian.brauner-GeWIH/nMZzLQT0dZR+AlfA@public.gmane.org>
---
 Documentation/admin-guide/cgroup-v2.rst | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
index 64c62b979f2f..c9f656a84590 100644
--- a/Documentation/admin-guide/cgroup-v2.rst
+++ b/Documentation/admin-guide/cgroup-v2.rst

@@ -949,6 +949,23 @@ All cgroup core files are prefixed with "cgroup."
 	it's possible to delete a frozen (and empty) cgroup, as well as
 	create new sub-cgroups.
 
+  cgroup.kill
+	A write-only single value file which exists in non-root cgroups.
+	The only allowed value is "1".
+
+	Writing "1" to the file causes the cgroup and all descendant cgroups to
+	be killed. This means that all processes located in the affected cgroup
+	tree will be killed via SIGKILL.
+
+	Killing a cgroup tree will deal with concurrent forks appropriately and
+	is protected against migrations. If callers require strict guarantees
+	they can issue the cgroup.kill request after a freezing the cgroup via
+	cgroup.freeze.

Hm, is it necessarily? What additional guarantees adds using the freezer?

Every new process that get's added is frozen. So even if the a process
ends up escaping the cgroup.kill request somehow it will be frozen in
the cgroup and can't itself fork again right away. So you could do:

echo 1 > cgroup.freeze
wait for frozen notification

echo 1 > cgroup.kill
wait for unpopulated notification