Format specifier issue when building kernel

[Richard Guy Briggs]

On Thu, Oct 31, 2013 at 12:25:55PM -0700, William Roberts wrote:
> > +                       if (msg_type != AUDIT_USER_TTY) {
> > +                               char fmt[64];
> > +                               strcat(fmt, " msg='%.");
> > +                               strcat(fmt, "AUDIT_MESSAGE_TEXT_MAX");
> > +                               strcat(fmt, "s'");
> > +                               audit_log_format(ab, fmt, (char *)data);
> > +                       } else {
> 
> I am ok with this. In fact I was going to do this the first time, but I
> thought their would be some explicit reason to avoid the additional
> run time overhead as the concat could be made at compile time.

Ok, this was in danger of starting with fmt in an unknown state.  Latest
patch: 

diff --git a/kernel/audit.c b/kernel/audit.c
@@ -148,6 +148,8 @@ DEFINE_MUTEX(audit_cmd_mutex);
  * should be at least that large. */
 #define AUDIT_BUFSIZ 1024
 
+char usermsg_format[64] = "";
+
 /* AUDIT_MAXFREE is the number of empty audit_buffers we keep on the
  * audit_freelist.  Doing so eliminates many kmalloc/kfree calls. */
 #define AUDIT_MAXFREE  (2*NR_CPUS)
@@ -714,11 +716,15 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 					break;
 			}
 			audit_log_common_recv_msg(&ab, msg_type);
-			if (msg_type != AUDIT_USER_TTY)
-				audit_log_format(ab,
-						 " msg='%.AUDIT_MESSAGE_TEXT_MAXs'",
+			if (msg_type != AUDIT_USER_TTY) {
+				if (unlikely(usermsg_format[0] == 0))
+					snprintf(usermsg_format,
+						sizeof(usermsg_format),
+						" msg=\'%%.%ds\'", 
+						AUDIT_MESSAGE_TEXT_MAX);
+				audit_log_format(ab, usermsg_format,
 						 (char *)data);
-			else {
+			} else {
 				int size;
 
 				audit_log_format(ab, " data=");

- RGB

--
Richard Guy Briggs <rbriggs at redhat.com>
Senior Software Engineer
Kernel Security
AMER ENG Base Operating Systems
Remote, Ottawa, Canada
Voice: +1.647.777.2635
Internal: (81) 32635
Alt: +1.613.693.0684x3545