linux-kernel - Re: [PATCH 4/4] cgroup: Do not report unavailable v1 controllers in /proc/cgroups

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <7sbzasggfk3elhvxsd5mtuzd4yo3c64wuzkaulr7yqybpfxwuh@g6dcatriw7hx>
Date: Fri, 18 Jul 2025 11:18:54 +0200
From: Michal Koutný <mkoutny@...e.com>
To: Tejun Heo <tj@...nel.org>, Ben Hutchings <ben@...adent.org.uk>
Cc: cgroups@...r.kernel.org, linux-kernel@...r.kernel.org, 
	linux-mm@...ck.org, Zefan Li <lizefan.x@...edance.com>, 
	Johannes Weiner <hannes@...xchg.org>, Michal Hocko <mhocko@...nel.org>, 
	Roman Gushchin <roman.gushchin@...ux.dev>, Shakeel Butt <shakeel.butt@...ux.dev>, 
	Muchun Song <muchun.song@...ux.dev>, Andrew Morton <akpm@...ux-foundation.org>, 
	Chen Ridong <chenridong@...wei.com>, 1108294@...s.debian.org
Subject: Re: [PATCH 4/4] cgroup: Do not report unavailable v1 controllers in
 /proc/cgroups

On Fri, Jul 11, 2025 at 12:15:07PM -1000, Tejun Heo <tj@...nel.org> wrote:
> I think we still want to deprecate /proc/cgroups but given that there are
> impacted users maybe we can bring it back under a boottime param w/ warning?

Something like below? (I don't change the log level.)

Ben, the affected Java users could modify it at boot time. I saw your
revert is in v6.12, so you may also want backport of a0ab1453226d8 to
give the users a message. (I realize current->comm in the message would
be even more instructive.)

-- >8 --

From ace88e9e3a77ff3fe86aee4b7a5866b3bfd2df58 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Michal=20Koutn=C3=BD?= <mkoutny@...e.com>
Date: Thu, 17 Jul 2025 17:38:47 +0200
Subject: [PATCH] cgroup: Add compatibility option for content of /proc/cgroups
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

/proc/cgroups lists only v1 controllers by default, however, this is
only enforced since the commit af000ce85293b ("cgroup: Do not report
unavailable v1 controllers in /proc/cgroups") and there is software in
the wild that uses content of /proc/cgroups to decide on availability of
v2 (sic) controllers.

Add a boottime param that can bring back the previous behavior for
setups where the check in the software cannot be changed and it causes
e.g. unintended OOMs.

Also, this patch takes out cgrp_v1_visible from cgroup1_subsys_absent()
guard since it's only important to check which hierarchy (v1 vs v2) the
subsys is attached to. This has no effect on the printed message but
the code is cleaner since cgrp_v1_visible is really about mounted
hierarchies, not the content of /proc/cgroups.

Link: https://lore.kernel.org/r/b26b60b7d0d2a5ecfd2f3c45f95f32922ed24686.camel@decadent.org.uk
Fixes: af000ce85293b ("cgroup: Do not report unavailable v1 controllers in /proc/cgroups")
Fixes: a0ab1453226d8 ("cgroup: Print message when /proc/cgroups is read on v2-only system")
Signed-off-by: Michal Koutný <mkoutny@...e.com>
---
 Documentation/admin-guide/kernel-parameters.txt |  8 ++++++++
 kernel/cgroup/cgroup-v1.c                       | 14 ++++++++++++--
 2 files changed, 20 insertions(+), 2 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 07e22ba5bfe34..f6d317e1674d6 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -633,6 +633,14 @@
 			named mounts. Specifying both "all" and "named" disables
 			all v1 hierarchies.
 
+	cgroup_v1_proc=	[KNL] Show also missing controllers in /proc/cgroups
+			Format: { "true" | "false" }
+			/proc/cgroups lists only v1 controllers by default.
+			This compatibility option enables listing also v2
+			controllers (whose v1 code is not compiled!), so that
+			semi-legacy software can check this file to decide
+			about usage of v2 (sic) controllers.
+
 	cgroup_favordynmods= [KNL] Enable or Disable favordynmods.
 			Format: { "true" | "false" }
 			Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS.
diff --git a/kernel/cgroup/cgroup-v1.c b/kernel/cgroup/cgroup-v1.c
index fa24c032ed6fe..2a4a387f867ab 100644
--- a/kernel/cgroup/cgroup-v1.c
+++ b/kernel/cgroup/cgroup-v1.c
@@ -32,6 +32,9 @@ static u16 cgroup_no_v1_mask;
 /* disable named v1 mounts */
 static bool cgroup_no_v1_named;
 
+/* Show unavailable controllers in /proc/cgroups */
+static bool proc_show_all;
+
 /*
  * pidlist destructions need to be flushed on cgroup destruction.  Use a
  * separate workqueue as flush domain.
@@ -683,10 +686,11 @@ int proc_cgroupstats_show(struct seq_file *m, void *v)
 	 */
 
 	for_each_subsys(ss, i) {
-		if (cgroup1_subsys_absent(ss))
-			continue;
 		cgrp_v1_visible |= ss->root != &cgrp_dfl_root;
 
+		if (!proc_show_all && cgroup1_subsys_absent(ss))
+			continue;
+
 		seq_printf(m, "%s\t%d\t%d\t%d\n",
 			   ss->legacy_name, ss->root->hierarchy_id,
 			   atomic_read(&ss->root->nr_cgrps),
@@ -1359,3 +1363,9 @@ static int __init cgroup_no_v1(char *str)
 	return 1;
 }
 __setup("cgroup_no_v1=", cgroup_no_v1);
+
+static int __init cgroup_v1_proc(char *str)
+{
+	return (kstrtobool(str, &proc_show_all) == 0);
+}
+__setup("cgroup_v1_proc=", cgroup_v1_proc);
-- 
2.50.0


Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)