[PATCH 1/5] cred: record how far up CAP_SETFCAP reaches
sashiko-bot at kernel.org
sashiko-bot at kernel.org
Tue Oct 6 15:58:26 UTC 2026
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] cred: use-after-put of user_ns in userns_install()
--
commit 4646fadaf7eaab13c361fe3e2861493a87f5e140
Author: Josef Bacik <josef at toxicpanda.com>
cred: record how far up CAP_SETFCAP reaches
This patch introduces cred->setfcap_level to track how far up the user
namespace hierarchy a CAP_SETFCAP capability extends. It sets this value
in set_cred_user_ns() when entering a new user namespace.
> diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c
> index 1b23d819d3983..6b45df3a8d826 100644
> --- a/kernel/user_namespace.c
> +++ b/kernel/user_namespace.c
> @@ -44,6 +44,9 @@ static void dec_user_namespaces(struct ucounts *ucounts)
>
> static void set_cred_user_ns(struct cred *cred, struct user_namespace *user_ns)
> {
> + /* The last chance to see what we can do outside of the new namespace. */
> + cred->setfcap_level = cap_setfcap_level(cred, user_ns->parent);
> +
[Severity: Low]
Could this result in a logical use-after-put of cred->user_ns when called
from userns_install()?
Looking at userns_install() in kernel/user_namespace.c, the reference to
cred->user_ns is relinquished right before calling set_cred_user_ns():
userns_install() {
...
put_user_ns(cred->user_ns);
set_cred_user_ns(cred, get_user_ns(user_ns));
...
}
Inside the newly added call to cap_setfcap_level(), the provided cred
object is dereferenced. Specifically, cap_setfcap_level() checks
cap_capable(), which inspects cred->user_ns, and later uses
cred->user_ns->level as a fallback.
While the object is independently kept alive by current->real_cred->user_ns
during the setns system call, does this pattern violate object lifecycle
semantics by relying on a logically dropped pointer?
Should the reference to cred->user_ns be saved to a temporary variable and
put only after set_cred_user_ns returns?
> /* Start with the same capabilities as init but useless for doing
> * anything as the capabilities are bound to the new user namespace.
> */
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261006-b4-setfcap-userns-v1-0-f47e7ed66072@toxicpanda.com?part=1
More information about the Linux-security-module-archive
mailing list