Fix rule eviction order for AUDIT_DIR

If syscall removes the root of subtree being watched, we
definitely do not want the rules refering that subtree
to be destroyed without the syscall in question having
a chance to match them.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

4 files changed, 72 insertions, 23 deletions

diff --git a/kernel/audit.c b/kernel/audit.c
index 6194c50..defc2e6 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -133,7 +133,7 @@ static DECLARE_WAIT_QUEUE_HEAD(kauditd_wait);
 static DECLARE_WAIT_QUEUE_HEAD(audit_backlog_wait);
 
 /* Serialize requests from userspace. */
-static DEFINE_MUTEX(audit_cmd_mutex);
+DEFINE_MUTEX(audit_cmd_mutex);
 
 /* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
  * audit records.  Since printk uses a 1024 byte buffer, this buffer
@@ -505,21 +505,6 @@ int audit_send_list(void *_dest)
 	return 0;
 }
 
-#ifdef CONFIG_AUDIT_TREE
-static int prune_tree_thread(void *unused)
-{
-	mutex_lock(&audit_cmd_mutex);
-	audit_prune_trees();
-	mutex_unlock(&audit_cmd_mutex);
-	return 0;
-}
-
-void audit_schedule_prune(void)
-{
-	kthread_run(prune_tree_thread, NULL, "audit_prune_tree");
-}
-#endif
-
 struct sk_buff *audit_make_reply(int pid, int seq, int type, int done,
 				 int multi, void *payload, int size)
 {
diff --git a/kernel/audit.h b/kernel/audit.h
index bb1c0d6..208687b 100644
--- a/kernel/audit.h
+++ b/kernel/audit.h
@@ -128,10 +128,9 @@ extern int audit_add_tree_rule(struct audit_krule *);
 extern int audit_remove_tree_rule(struct audit_krule *);
 extern void audit_trim_trees(void);
 extern int audit_tag_tree(char *old, char *new);
-extern void audit_schedule_prune(void);
-extern void audit_prune_trees(void);
 extern const char *audit_tree_path(struct audit_tree *);
 extern void audit_put_tree(struct audit_tree *);
+extern void audit_kill_trees(struct list_head *);
 #else
 #define audit_remove_tree_rule(rule) BUG()
 #define audit_add_tree_rule(rule) -EINVAL
@@ -140,6 +139,7 @@ extern void audit_put_tree(struct audit_tree *);
 #define audit_put_tree(tree) (void)0
 #define audit_tag_tree(old, new) -EINVAL
 #define audit_tree_path(rule) ""	/* never called */
+#define audit_kill_trees(list) BUG()
 #endif
 
 extern char *audit_unpack_string(void **, size_t *, size_t);
@@ -158,7 +158,10 @@ static inline int audit_signal_info(int sig, struct task_struct *t)
 	return 0;
 }
 extern void audit_filter_inodes(struct task_struct *, struct audit_context *);
+extern struct list_head *audit_killed_trees(void);
 #else
 #define audit_signal_info(s,t) AUDIT_DISABLED
 #define audit_filter_inodes(t,c) AUDIT_DISABLED
 #endif
+
+extern struct mutex audit_cmd_mutex;
diff --git a/kernel/audit_tree.c b/kernel/audit_tree.c
index 3ff0731..2451dc6 100644
--- a/kernel/audit_tree.c
+++ b/kernel/audit_tree.c
@@ -2,6 +2,7 @@
 #include <linux/inotify.h>
 #include <linux/namei.h>
 #include <linux/mount.h>
+#include <linux/kthread.h>
 
 struct audit_tree;
 struct audit_chunk;
@@ -517,6 +518,8 @@ static void trim_marked(struct audit_tree *tree)
 	}
 }
 
+static void audit_schedule_prune(void);
+
 /* called with audit_filter_mutex */
 int audit_remove_tree_rule(struct audit_krule *rule)
 {
@@ -822,10 +825,11 @@ int audit_tag_tree(char *old, char *new)
 
 /*
  * That gets run when evict_chunk() ends up needing to kill audit_tree.
- * Runs from a separate thread, with audit_cmd_mutex held.
+ * Runs from a separate thread.
  */
-void audit_prune_trees(void)
+static int prune_tree_thread(void *unused)
 {
+	mutex_lock(&audit_cmd_mutex);
 	mutex_lock(&audit_filter_mutex);
 
 	while (!list_empty(&prune_list)) {
@@ -842,6 +846,40 @@ void audit_prune_trees(void)
 	}
 
 	mutex_unlock(&audit_filter_mutex);
+	mutex_unlock(&audit_cmd_mutex);
+	return 0;
+}
+
+static void audit_schedule_prune(void)
+{
+	kthread_run(prune_tree_thread, NULL, "audit_prune_tree");
+}
+
+/*
+ * ... and that one is done if evict_chunk() decides to delay until the end
+ * of syscall.  Runs synchronously.
+ */
+void audit_kill_trees(struct list_head *list)
+{
+	mutex_lock(&audit_cmd_mutex);
+	mutex_lock(&audit_filter_mutex);
+
+	while (!list_empty(list)) {
+		struct audit_tree *victim;
+
+		victim = list_entry(list->next, struct audit_tree, list);
+		kill_rules(victim);
+		list_del_init(&victim->list);
+
+		mutex_unlock(&audit_filter_mutex);
+
+		prune_one(victim);
+
+		mutex_lock(&audit_filter_mutex);
+	}
+
+	mutex_unlock(&audit_filter_mutex);
+	mutex_unlock(&audit_cmd_mutex);
 }
 
 /*
@@ -852,6 +890,8 @@ void audit_prune_trees(void)
 static void evict_chunk(struct audit_chunk *chunk)
 {
 	struct audit_tree *owner;
+	struct list_head *postponed = audit_killed_trees();
+	int need_prune = 0;
 	int n;
 
 	if (chunk->dead)
@@ -867,15 +907,21 @@ static void evict_chunk(struct audit_chunk *chunk)
 		owner->root = NULL;
 		list_del_init(&owner->same_root);
 		spin_unlock(&hash_lock);
-		kill_rules(owner);
-		list_move(&owner->list, &prune_list);
-		audit_schedule_prune();
+		if (!postponed) {
+			kill_rules(owner);
+			list_move(&owner->list, &prune_list);
+			need_prune = 1;
+		} else {
+			list_move(&owner->list, postponed);
+		}
 		spin_lock(&hash_lock);
 	}
 	list_del_rcu(&chunk->hash);
 	for (n = 0; n < chunk->count; n++)
 		list_del_init(&chunk->owners[n].list);
 	spin_unlock(&hash_lock);
+	if (need_prune)
+		audit_schedule_prune();
 	mutex_unlock(&audit_filter_mutex);
 }
 
diff --git a/kernel/auditsc.c b/kernel/auditsc.c
index 2de95d1..68d3c6a 100644
--- a/kernel/auditsc.c
+++ b/kernel/auditsc.c
@@ -199,6 +199,7 @@ struct audit_context {
 
 	struct audit_tree_refs *trees, *first_trees;
 	int tree_count;
+	struct list_head killed_trees;
 
 	int type;
 	union {
@@ -853,6 +854,7 @@ static inline struct audit_context *audit_alloc_context(enum audit_state state)
 	if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
 		return NULL;
 	audit_zero_context(context, state);
+	INIT_LIST_HEAD(&context->killed_trees);
 	return context;
 }
 
@@ -1545,6 +1547,8 @@ void audit_free(struct task_struct *tsk)
 	/* that can happen only if we are called from do_exit() */
 	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
 		audit_log_exit(context, tsk);
+	if (!list_empty(&context->killed_trees))
+		audit_kill_trees(&context->killed_trees);
 
 	audit_free_context(context);
 }
@@ -1688,6 +1692,9 @@ void audit_syscall_exit(int valid, long return_code)
 	context->in_syscall = 0;
 	context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
 
+	if (!list_empty(&context->killed_trees))
+		audit_kill_trees(&context->killed_trees);
+
 	if (context->previous) {
 		struct audit_context *new_context = context->previous;
 		context->previous  = NULL;
@@ -2521,3 +2528,11 @@ void audit_core_dumps(long signr)
 	audit_log_format(ab, " sig=%ld", signr);
 	audit_log_end(ab);
 }
+
+struct list_head *audit_killed_trees(void)
+{
+	struct audit_context *ctx = current->audit_context;
+	if (likely(!ctx || !ctx->in_syscall))
+		return NULL;
+	return &ctx->killed_trees;
+}