As explained in the commit 76f969e8948d8 ("cgroup: cgroup v2 freezer"),
the original freezer is imperfect, some users may unwittingly rely on it
when there exists the alternative of v2. Print a message when it happens
and explain that in the docs.
Signed-off-by: Michal Koutný <mkoutny@xxxxxxxx>
---
Documentation/admin-guide/cgroup-v1/freezer-subsystem.rst | 4 ++++
kernel/cgroup/legacy_freezer.c | 6 ++++--
2 files changed, 8 insertions(+), 2 deletions(-)
diff --git a/Documentation/admin-guide/cgroup-v1/freezer-subsystem.rst b/Documentation/admin-guide/cgroup-v1/freezer-subsystem.rst
index 582d3427de3f8..a964aff373b19 100644
--- a/Documentation/admin-guide/cgroup-v1/freezer-subsystem.rst
+++ b/Documentation/admin-guide/cgroup-v1/freezer-subsystem.rst
@@ -125,3 +125,7 @@ to unfreeze all tasks in the container::
This is the basic mechanism which should do the right thing for user space task
in a simple scenario.
+
+This freezer implementation is affected by shortcomings (see commit
+76f969e8948d8 ("cgroup: cgroup v2 freezer")) and cgroup v2 freezer is
+recommended.
diff --git a/kernel/cgroup/legacy_freezer.c b/kernel/cgroup/legacy_freezer.c
index 074653f964c1d..039d1eb2f215b 100644
--- a/kernel/cgroup/legacy_freezer.c
+++ b/kernel/cgroup/legacy_freezer.c
@@ -430,9 +430,11 @@ static ssize_t freezer_write(struct kernfs_open_file *of,
if (strcmp(buf, freezer_state_strs(0)) == 0)
freeze = false;
- else if (strcmp(buf, freezer_state_strs(CGROUP_FROZEN)) == 0)
+ else if (strcmp(buf, freezer_state_strs(CGROUP_FROZEN)) == 0) {
+ pr_info_once("Freezing with imperfect legacy cgroup freezer. "
+ "See cgroup.freeze of cgroup v2\n");
freeze = true;
- else
+ } else
return -EINVAL;
freezer_change_state(css_freezer(of_css(of)), freeze);
--
2.48.1
