userns: detect initial namespace by inode

The current uid_map heuristic treats a private user namespace with an identity uid_map as the initial user namespace. systemd 260 can create that shape with PrivateUsers=full, causing callers to believe they can perform initial-namespace-only operations such as cgroup device BPF setup.

Use /proc/self/ns/user's inode instead and compare it with the kernel's PROC_USER_INIT_INO value. This detects the initial user namespace directly instead of inferring it from the uid_map layout.

Related: containers/crun#2150
Signed-off-by: Alvaro Leiva Geisse <aleivag@gmail.com>

Author: aleivag

userns/userns_linux.go

@@ -5,8 +5,12 @@ import (
 	"fmt"
 	"os"
 	"sync"
+	"syscall"
 )
 
+// See PROC_USER_INIT_INO in https://github.com/torvalds/linux/blob/v7.1/include/uapi/linux/nsfs.h#L50.
+const procUserInitIno = 0xEFFFFFFD
+
 var inUserNS = sync.OnceValue(runningInUserNS)
 
 // runningInUserNS detects whether we are currently running in a user namespace.
@@ -17,6 +21,11 @@ var inUserNS = sync.OnceValue(runningInUserNS)
 // [libcontainer/runc]: https://github.com/opencontainers/runc/blob/3778ae603c706494fd1e2c2faf83b406e38d687d/libcontainer/userns/userns_linux.go#L12-L49
 // [lcx/incus]: https://github.com/lxc/incus/blob/e45085dd42f826b3c8c3228e9733c0b6f998eafe/shared/util.go#L678-L700
 func runningInUserNS() bool {
+	var st syscall.Stat_t
+	if err := syscall.Stat("/proc/self/ns/user", &st); err == nil {
+		return st.Ino != procUserInitIno
+	}
+
 	file, err := os.Open("/proc/self/uid_map")
 	if err != nil {
 		// This kernel-provided file only exists if user namespaces are supported.
@@ -30,12 +39,9 @@ func runningInUserNS() bool {
 		return false
 	}
 
-	return uidMapInUserNS(string(l))
-}
-
-func uidMapInUserNS(uidMap string) bool {
+	uidMap := string(l)
 	if uidMap == "" {
-		// File exist but empty (the initial state when userns is created,
+		// File exists but is empty (the initial state when userns is created,
 		// see user_namespaces(7)).
 		return true
 	}