Bug 832725 (CVE-2022-0492)

Summary:	sys-kernel/gentoo-sources kernel/cgroup/cgroup-v1.c - privesc in cgroup_release_agent_write (CVE-2022-0492)
Product:	Gentoo Linux	Reporter:	Hank Leininger <hlein>
Component:	Current packages	Assignee:	Gentoo Security <security>
Status:	RESOLVED FIXED
Severity:	normal	CC:	jstein, kernel
Priority:	Normal	Keywords:	SECURITY
Version:	unspecified
Hardware:	All
OS:	Linux
URL:	https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=24f6008564183aa120d07c03d9289519c2fe02af
Whiteboard:
Package list:		Runtime testing required:	---

Description Hank Leininger 2022-02-04 22:23:46 UTC

From https://marc.info/?l=oss-security&m=164399879422272&w=4:

"It has been discovered that under certain circumstances, the Linux kernel's
cgroups v1 release_agent feature can be used to escalate privilege and
bypass namespace isolation unexpectedly.

CVE-2022-0492 has been assigned to this issue, which is corrected by
requiring CAP_SYS_ADMIN in the initial user namespace when setting
release_agent. This has been included upstream in commit
24f6008564183aa120d07c03d9289519c2fe02af."

From https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=24f6008564183aa120d07c03d9289519c2fe02af:

"The cgroup release_agent is called with call_usermodehelper.  The function
call_usermodehelper starts the release_agent with a full set fo capabilities.
Therefore require capabilities when setting the release_agaent."

AIUI, that means anyone who can call cgroup_release_agent_write can end up with all capabilities, prior to the fix?

Some minor spelunking showed that cgroup_release_agent_write appeared in more-or-less its current form in 2008. So any Linux kernel from v2.6.26 onward, with CONFIG_CGROUPS, may be vulnerable?

Comment 1 John Helmert III archtester

2022-08-16 22:59:34 UTC

$ fix_in_what_release 24f6008564183aa120d07c03d9289519c2fe02af
4.9.301 4.14.266 4.19.229 5.4.177 5.10.97 5.15.20 5.16.6

I can't find a security-supported kernel that has any vulnerable version, and since we don't issue GLSAs for kernels, we should be all done here.