[PATCH v3 1/4] pidfd: restrict task access ioctls to the caller's pid namespace

[Chen Linxuan via B4 Relay]

From: Chen Linxuan <me at black-desk.cn>

pidfds can cross pid namespace boundaries: a task may obtain one for a
process outside of its own pid namespace hierarchy through SO_PASSPIDFD
or SO_PEERPIDFD, via SCM_RIGHTS, or through file descriptor inheritance.

The namespace ioctls gate access with a ptrace check against the
target's credentials only. That check does not account for the pid
namespace relationship between the caller and the target, so a task
inside a pid namespace can use a pidfd obtained across that boundary to
acquire namespace file descriptors of a task outside of its hierarchy.

procfs exposes the equivalent information only for tasks visible in the
reader's pid namespace, and PIDFD_GET_INFO already answers with -EREMOTE
for targets outside the caller's hierarchy. Extend that policy to the
namespace ioctls: fail with -EREMOTE unless the target is visible in the
caller's active pid namespace. Callers in ancestor pid namespaces are
unaffected.

Unlike PIDFD_GET_INFO, the visibility check is performed after the task
lookup here, so reaped targets outside of the caller's pid namespace
hierarchy answer -ESRCH instead of -EREMOTE.

Link: https://lore.kernel.org/all/CAG48ez1T5-tUHVK_iHsgUtWa-inVdrZxA3E-ZCW0iENJvasKfg@mail.gmail.com
Link: https://lore.kernel.org/all/20260925-gewischt-auftrag-tierzucht-c153cb3f9f74@brauner
Suggested-by: Jann Horn <jannh at google.com>
Suggested-by: Christian Brauner <brauner at kernel.org>
Assisted-by: LLM
Signed-off-by: Chen Linxuan <me at black-desk.cn>
---
 fs/pidfs.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/fs/pidfs.c b/fs/pidfs.c
index a6a643f15d08..791f7efceabe 100644
--- a/fs/pidfs.c
+++ b/fs/pidfs.c
@@ -556,6 +556,14 @@ static long pidfd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 	if (arg)
 		return -EINVAL;
 
+	/*
+	 * Require the target to be visible in the caller's pid namespace
+	 * for operations that grant access to its resources, mirroring
+	 * procfs and PIDFD_GET_INFO.
+	 */
+	if (!pid_in_current_pidns(pidfd_pid(file)))
+		return -EREMOTE;
+
 	/*
 	 * We're trying to open a file descriptor to the namespace so perform a
 	 * filesystem cred ptrace check. Hold @task's exec_update_lock for the

-- 
2.55.0