llc: SOCK_DGRAM interface fixes

The datagram interface of LLC is broken in a couple of ways. These were discovered when trying to use it to build an out-of-kernel version of STP.

First it didn't pass the source address of the received packet in recvfrom(). It needs to copy the source address of received LLC packets into the socket control block. At the same time fix a security issue because there was uninitialized data leakage. Every recvfrom call was just copying out old data.

Second, LLC should not merge multiple packets in one receive call on datagram sockets. LLC should preserve packet boundaries on SOCK_DGRAM.

This fix goes against the old historical comments about UNIX98 semantics but without this fix SOCK_DGRAM is broken and useless. So either ANK's interpretation was incorect or UNIX98 standard was wrong.

---
 net/llc/af_llc.c  |   20 ++++++++------------
 net/llc/llc_sap.c |    4 ++--
 2 files changed, 10 insertions(+), 14 deletions(-)

index d6cfe84..2652ead 100644
@@ -784,24 +784,20 @@ static int llc_ui_recvmsg(struct kiocb *
 		copied += used;
 		len -= used;
 
-		if (used + offset < skb->len)
-			continue;
-
 		if (!(flags & MSG_PEEK)) {
 			sk_eat_skb(sk, skb, 0);
 			*seq = 0;
 		}
+
+		/* For non stream protcols we get one packet per recvmsg call */
+		if (sk->sk_type != SOCK_STREAM)
+			goto copy_uaddr;
+
+		/* Partial read */
+		if (used + offset < skb->len)
+			continue;
 	} while (len > 0);
 
-	/* 
-	 * According to UNIX98, msg_name/msg_namelen are ignored
-	 * on connected socket. -ANK
-	 * But... af_llc still doesn't have separate sets of methods for
-	 * SOCK_DGRAM and SOCK_STREAM :-( So we have to do this test, will
-	 * eventually fix this tho :-) -acme
-	 */
-	if (sk->sk_type == SOCK_DGRAM)
-		goto copy_uaddr;
 out:
 	release_sock(sk);
 	return copied;

index 20c4eb5..42eb0c3 100644
@@ -51,10 +51,10 @@ void llc_save_primitive(struct sock *sk,
 {
 	struct sockaddr_llc *addr;
 
-	if (skb->sk->sk_type == SOCK_STREAM) /* See UNIX98 */
-		return;
        /* save primitive for use by the user. */
 	addr		  = llc_ui_skb_cb(skb);
+
+	memset(addr, 0, sizeof(*addr));
 	addr->sllc_family = sk->sk_family;
 	addr->sllc_arphrd = skb->dev->type;
 	addr->sllc_test   = prim == LLC_TEST_PRIM;

1.4.0

I don't doubt it, most of the testing was done on the SOCK_STREAM code, Jay Schullist worked a bit on the dgram side of things but that was long ago.

The plan is to use sk->sk_prot-> in the same way as the INET transport protocols, to clearly separate STREAM from DGRAM code, etc. Its yet another thing on my longish backlog, sigh.

Dave, please apply.

- Arnaldo

From: Arnaldo Carvalho de Melo <[email protected]>
Date: Thu, 03 Aug 2006 13:38:51 -0300

Applied, thanks everyone.

This doesn't apply at all to the current -stable tree, sorry.

greg k-h

Hello!

Just found this reference to me. :-)

The comment migrated from tcp.c. It is only about connected SOCK_STREAM sockets, I do not see how it can make SOCK_DGRAM broken or useless.

That UNIX98 statement allowed to avoid expensive callback to protocol specific setup of address in tcp_recvmsg().

Alexey

The code was in the place where the source information was being copied from the data portion to the cb portion of the skb. The original code did the save only for SOCK_STREAM.
The cb portion is needed later to produce the address portion of the receive handling which is critical with SOCK_DGRAM in LLC.

So the comment made sense for TCP but not LLC.

Em Tue, Aug 08, 2006 at 08:36:18PM +0400, Alexey Kuznetsov escreveu:

Its just that the work to get the AF_LLC code closer to the AF_INET structure was suspended before the SOCK_STREAM code was separated from the SOCK_DGRAM one, moving PF_LLC to sk->sk_prot-> land, so the comment is indeed about SOCK_STREAM part, not DGRAM.

- Arnaldo