Rename nsproxy.pid_ns and fix a related security bug

This is a security bug.

The follow-up will fix nsproxy to discourage this type of issue from happening again.

Cc [email protected]

Signed-off-by Andy Lutomirski <[email protected]>

---
 net/core/scm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

net/core/scm.c +1 -1

index 03795d0..b4da80b 100644
@@ -54,7 +54,7 @@ static __inline__ int scm_check_creds(struct ucred *creds)
 		return -EINVAL;
 
 	if ((creds->pid == task_tgid_vnr(current) ||
-	     ns_capable(current->nsproxy->pid_ns->user_ns, CAP_SYS_ADMIN)) &&
+	     ns_capable(task_active_pid_ns(current)->user_ns, CAP_SYS_ADMIN)) &&
 	    ((uid_eq(uid, cred->uid)   || uid_eq(uid, cred->euid) ||
 	      uid_eq(uid, cred->suid)) || nsown_capable(CAP_SETUID)) &&
 	    ((gid_eq(gid, cred->gid)   || gid_eq(gid, cred->egid) ||

nsproxy.pid_ns is *not* the task's pid namespace. The name should clarify that.

This makes it more obvious that setns on a pid namespace is weird -- it won't change the pid namespace shown in procfs.

...

+ * The pid namespace is an exception -- it's accessed using
+ * task_active_pid_ns. The pid namespace here is the
+ * namespace that children will use.
+ *
* 'count' is the number of tasks holding a reference.
* The count for each namespace, then, will be the number
* of nsproxies pointing to it, not the number of tasks.

nsproxy.pid_ns is *not* the task's pid namespace. The name should clarify that.

This makes it more obvious that setns on a pid namespace is weird -- it won't change the pid namespace shown in procfs.

Signed-off-by Andy Lutomirski <[email protected]>

---
 include/linux/nsproxy.h |  6 +++++-
 kernel/fork.c           |  5 +++--
 kernel/nsproxy.c        | 27 ++++++++++++++-------------
 kernel/pid_namespace.c  |  4 ++--
 4 files changed, 24 insertions(+), 18 deletions(-)

include/linux/nsproxy.h +5 -1

index 10e5947..b4ec59d 100644
@@ -14,6 +14,10 @@ struct fs_struct;
  * A structure to contain pointers to all per-process
  * namespaces - fs (mount), uts, network, sysvipc, etc.
  *
+ * The pid namespace is an exception -- it's accessed using
+ * task_active_pid_ns.  The pid namespace here is the
+ * namespace that children will use.
+ *
  * 'count' is the number of tasks holding a reference.
  * The count for each namespace, then, will be the number
  * of nsproxies pointing to it, not the number of tasks.

@@ -27,7 +31,7 @@ struct nsproxy {
 	struct uts_namespace *uts_ns;
 	struct ipc_namespace *ipc_ns;
 	struct mnt_namespace *mnt_ns;
-	struct pid_namespace *pid_ns;
+	struct pid_namespace *pid_ns_for_children;
 	struct net 	     *net_ns;
 };
 extern struct nsproxy init_nsproxy;

kernel/fork.c +3 -2

index e23bb19..bf46287 100644
@@ -1177,7 +1177,8 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 	 * don't allow the creation of threads.
 	 */
 	if ((clone_flags & (CLONE_VM|CLONE_NEWPID)) &&
-	    (task_active_pid_ns(current) != current->nsproxy->pid_ns))
+	    (task_active_pid_ns(current) !=
+	     current->nsproxy->pid_ns_for_children))
 		return ERR_PTR(-EINVAL);
 
 	retval = security_task_create(clone_flags);

@@ -1351,7 +1352,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 
 	if (pid != &init_struct_pid) {
 		retval = -ENOMEM;
-		pid = alloc_pid(p->nsproxy->pid_ns);
+		pid = alloc_pid(p->nsproxy->pid_ns_for_children);
 		if (!pid)
 			goto bad_fork_cleanup_io;
 	}

kernel/nsproxy.c +14 -13

index 364ceab..997cbb9 100644
@@ -29,15 +29,15 @@
 static struct kmem_cache *nsproxy_cachep;
 
 struct nsproxy init_nsproxy = {
-	.count	= ATOMIC_INIT(1),
-	.uts_ns	= &init_uts_ns,
+	.count			= ATOMIC_INIT(1),
+	.uts_ns			= &init_uts_ns,
 #if defined(CONFIG_POSIX_MQUEUE) || defined(CONFIG_SYSVIPC)
-	.ipc_ns	= &init_ipc_ns,
+	.ipc_ns			= &init_ipc_ns,
 #endif
-	.mnt_ns	= NULL,
-	.pid_ns	= &init_pid_ns,
+	.mnt_ns			= NULL,
+	.pid_ns_for_children	= &init_pid_ns,
 #ifdef CONFIG_NET
-	.net_ns	= &init_net,
+	.net_ns			= &init_net,
 #endif
 };
 

@@ -85,9 +85,10 @@ static struct nsproxy *create_new_namespaces(unsigned long flags,
 		goto out_ipc;
 	}
 
-	new_nsp->pid_ns = copy_pid_ns(flags, user_ns, tsk->nsproxy->pid_ns);
-	if (IS_ERR(new_nsp->pid_ns)) {
-		err = PTR_ERR(new_nsp->pid_ns);
+	new_nsp->pid_ns_for_children =
+		copy_pid_ns(flags, user_ns, tsk->nsproxy->pid_ns_for_children);
+	if (IS_ERR(new_nsp->pid_ns_for_children)) {
+		err = PTR_ERR(new_nsp->pid_ns_for_children);
 		goto out_pid;
 	}
 

@@ -100,8 +101,8 @@ static struct nsproxy *create_new_namespaces(unsigned long flags,
 	return new_nsp;
 
 out_net:
-	if (new_nsp->pid_ns)
-		put_pid_ns(new_nsp->pid_ns);
+	if (new_nsp->pid_ns_for_children)
+		put_pid_ns(new_nsp->pid_ns_for_children);
 out_pid:
 	if (new_nsp->ipc_ns)
 		put_ipc_ns(new_nsp->ipc_ns);

@@ -174,8 +175,8 @@ void free_nsproxy(struct nsproxy *ns)
 		put_uts_ns(ns->uts_ns);
 	if (ns->ipc_ns)
 		put_ipc_ns(ns->ipc_ns);
-	if (ns->pid_ns)
-		put_pid_ns(ns->pid_ns);
+	if (ns->pid_ns_for_children)
+		put_pid_ns(ns->pid_ns_for_children);
 	put_net(ns->net_ns);
 	kmem_cache_free(nsproxy_cachep, ns);
 }

kernel/pid_namespace.c +2 -2

index 6917e8e..601bb36 100644
@@ -349,8 +349,8 @@ static int pidns_install(struct nsproxy *nsproxy, void *ns)
 	if (ancestor != active)
 		return -EINVAL;
 
-	put_pid_ns(nsproxy->pid_ns);
-	nsproxy->pid_ns = get_pid_ns(new);
+	put_pid_ns(nsproxy->pid_ns_for_children);
+	nsproxy->pid_ns_for_children = get_pid_ns(new);
 	return 0;
 }

commit 92f28d973cce45ef5823209aab3138eb45d8b349
Author: Eric W. Biederman <[email protected]>
Date: Fri Mar 15 01:03:33 2013 -0700

scm: Require CAP_SYS_ADMIN over the current pidns to spoof pids.

Eric fell for my bogus claim that nsproxy->pid_ns was the current' process's pid ns. This isn't true.

Let's fix the bug and rename pid_ns so that no one gets this wrong again.

Signed-off-by Andy Lutomirski <[email protected]>

From: Andy Lutomirski <[email protected]>
Date: Thu, 22 Aug 2013 11:39:14 -0700

Quoted 6 lines

commit 92f28d973cce45ef5823209aab3138eb45d8b349
Author: Eric W. Biederman <[email protected]>
Date: Fri Mar 15 01:03:33 2013 -0700

scm: Require CAP_SYS_ADMIN over the current pidns to spoof pids.

Quoted 6 lines

Eric fell for my bogus claim that nsproxy->pid_ns was the current' process's pid ns. This isn't true.

Let's fix the bug and rename pid_ns so that no one gets this wrong again.

Signed-off-by Andy Lutomirski <[email protected]>

Eric please take the time to review these changes, thanks.

Eric fell for my bogus claim that nsproxy->pid_ns was the current' process's pid ns. This isn't true.

Let's fix the bug and rename pid_ns so that no one gets this wrong again.

Signed-off-by Andy Lutomirski <[email protected]>