[PATCH] audit: fixes in audit_log_drain()

[PATCH] audit: fixes in audit_log_drain()

[Tommy Christensen]

[-- Attachment #1: Type: text/plain, Size: 166 bytes --]

Would this be for you, James?

o Fix skb leak
o Don't send shared skb's to netlink_unicast

Signed-off-by: Tommy S. Christensen <tommy.christensen@tpack.net>

-Tommy

[-- Attachment #2: audit.c.patch --]
[-- Type: text/plain, Size: 834 bytes --]

--- linux-2.6.11-rc1-bk3/kernel/audit.c	2005-01-12 14:54:15.000000000 +0100
+++ linux-2.6.11-work/kernel/audit.c	2005-01-15 23:51:25.399453674 +0100
@@ -483,6 +483,7 @@
 
 	while ((skb = skb_dequeue(&ab->sklist))) {
 		int retval = 0;
+		struct sk_buff *nskb;
 
 		if (audit_pid) {
 			if (ab->nlh) {
@@ -492,13 +493,17 @@
 				ab->nlh->nlmsg_seq   = 0;
 				ab->nlh->nlmsg_pid   = ab->pid;
 			}
-			skb_get(skb); /* because netlink_* frees */
-			retval = netlink_unicast(audit_sock, skb, audit_pid,
-						 MSG_DONTWAIT);
+			retval = -ENOMEM;
+			nskb = skb_clone(skb);
+			if (nskb)
+				retval = netlink_unicast(audit_sock, nskb,
+							 audit_pid,
+							 MSG_DONTWAIT);
 		}
 		if (retval == -EAGAIN && ab->count < 5) {
 			++ab->count;
 			audit_log_end_irq(ab);
+			kfree_skb(skb);
 			return 1;
 		}
 		if (retval < 0) {

Re: [PATCH] audit: fixes in audit_log_drain()

[Tommy Christensen]

Tommy Christensen wrote:
> o Don't send shared skb's to netlink_unicast

Never mind. Herbert Xu made a better fix for this in netlink.

-Tommy