audit: don't ever sleep on a command record/message

Sleeping on a command record/message in audit_log_start() could slow
something, e.g. auditd, from doing something important, e.g. clean
shutdown, which could present problems on a heavily loaded system.
This patch allows tasks to bypass any queue restrictions if they are
logging a command record/message.

Signed-off-by: Paul Moore <paul@paul-moore.com>
diff --git a/kernel/audit.c b/kernel/audit.c
index b447a6b..f20eee0 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -1488,11 +1488,19 @@
 	if (unlikely(!audit_filter(type, AUDIT_FILTER_TYPE)))
 		return NULL;
 
-	/* don't ever fail/sleep on auditd since we need auditd to drain the
-	 * queue; also, when we are checking for auditd, compare PIDs using
-	 * task_tgid_vnr() since auditd_pid is set in audit_receive_msg() using
-	 * a PID anchored in the caller's namespace */
-	if (!(audit_pid && audit_pid == task_tgid_vnr(current))) {
+	/* don't ever fail/sleep on these two conditions:
+	 * 1. auditd generated record - since we need auditd to drain the
+	 *    queue; also, when we are checking for auditd, compare PIDs using
+	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
+	 *    using a PID anchored in the caller's namespace
+	 * 2. audit command message - record types 1000 through 1099 inclusive
+	 *    are command messages/records used to manage the kernel subsystem
+	 *    and the audit userspace, blocking on these messages could cause
+	 *    problems under load so don't do it (note: not all of these
+	 *    command types are valid as record types, but it is quicker to
+	 *    just check two ints than a series of ints in a if/switch stmt) */
+	if (!((audit_pid && audit_pid == task_tgid_vnr(current)) ||
+	      (type >= 1000 && type <= 1099))) {
 		long sleep_time = audit_backlog_wait_time;
 
 		while (audit_backlog_limit &&