[PATCH] namespaces: fix exit race by splitting exit

Fix exit race by splitting the nsproxy putting into two pieces. First
piece reduces the nsproxy refcount. If we dropped the last reference, then
it puts the mnt_ns, and returns the nsproxy as a hint to the caller. Else
it returns NULL. The second piece of exiting task namespaces sets
tsk->nsproxy to NULL, and drops the references to other namespaces and
frees the nsproxy only if an nsproxy was passed in.

A little awkward and should probably be reworked, but hopefully it fixes
the NFS oops.

Signed-off-by: Serge E. Hallyn <[e-mail]>
Cc: Herbert Poetzl <[e-mail]>
Cc: Oleg Nesterov <[e-mail]>
Cc: "Eric W. Biederman" <[e-mail]>
Cc: Cedric Le Goater <[e-mail]>
Cc: Daniel Hokka Zakrisson <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

include/linux/nsproxy.h

···
 int copy_namespaces(int flags, struct task_struct *tsk);
 void get_task_namespaces(struct task_struct *tsk);
 void free_nsproxy(struct nsproxy *ns);
+struct nsproxy *put_nsproxy(struct nsproxy *ns);
 
-static inline void put_nsproxy(struct nsproxy *ns)
+static inline void finalize_put_nsproxy(struct nsproxy *ns)
 {
-	if (atomic_dec_and_test(&ns->count)) {
+	if (ns)
 		free_nsproxy(ns);
-	}
 }
 
-static inline void exit_task_namespaces(struct task_struct *p)
+static inline void put_and_finalize_nsproxy(struct nsproxy *ns)
 {
-	struct nsproxy *ns = p->nsproxy;
-	if (ns) {
-		task_lock(p);
-		p->nsproxy = NULL;
-		task_unlock(p);
-		put_nsproxy(ns);
-	}
+	finalize_put_nsproxy(put_nsproxy(ns));
+}
+
+static inline struct nsproxy *preexit_task_namespaces(struct task_struct *p)
+{
+	return put_nsproxy(p->nsproxy);
+}
+
+static inline void exit_task_namespaces(struct task_struct *p,
+						struct nsproxy *ns)
+{
+	task_lock(p);
+	p->nsproxy = NULL;
+	task_unlock(p);
+	finalize_put_nsproxy(ns);
 }
 #endif

kernel/exit.c

···
 	current->fs = fs;
 	atomic_inc(&fs->count);
 
-	exit_task_namespaces(current);
+	put_and_finalize_nsproxy(current->nsproxy);
 	current->nsproxy = init_task.nsproxy;
 	get_task_namespaces(current);
 
···
 fastcall NORET_TYPE void do_exit(long code)
 {
 	struct task_struct *tsk = current;
+	struct nsproxy *ns;
 	int group_dead;
 
 	profile_task_exit(tsk);
···
 
 	tsk->exit_code = code;
 	proc_exit_connector(tsk);
+	ns = preexit_task_namespaces(tsk);
 	exit_notify(tsk);
-	exit_task_namespaces(tsk);
+	exit_task_namespaces(tsk, ns);
 #ifdef CONFIG_NUMA
 	mpol_free(tsk->mempolicy);
 	tsk->mempolicy = NULL;

kernel/fork.c

···
 	return p;
 
 bad_fork_cleanup_namespaces:
-	exit_task_namespaces(p);
+	put_and_finalize_nsproxy(p->nsproxy);
 bad_fork_cleanup_keys:
 	exit_keys(p);
 bad_fork_cleanup_mm:
···
 	}
 
 	if (new_nsproxy)
-		put_nsproxy(new_nsproxy);
+		put_and_finalize_nsproxy(new_nsproxy);
 
 bad_unshare_cleanup_ipc:
 	if (new_ipc)

kernel/nsproxy.c

···
 		goto out_pid;
 
 out:
-	put_nsproxy(old_ns);
+	put_and_finalize_nsproxy(old_ns);
 	return err;
 
 out_pid:
···
 	tsk->nsproxy = old_ns;
 	kfree(new_ns);
 	goto out;
+}
+
+struct nsproxy *put_nsproxy(struct nsproxy *ns)
+{
+	if (ns) {
+		if (atomic_dec_and_test(&ns->count)) {
+			if (ns->mnt_ns) {
+				put_mnt_ns(ns->mnt_ns);
+				ns->mnt_ns = NULL;
+			}
+			return ns;
+		}
+	}
+	return NULL;
 }
 
 void free_nsproxy(struct nsproxy *ns)