Skip to contentSkip to conversation

Possible bug in netlink_recvmsg()

DISCUSSION David Howells 2 participants 3 messages Apr 13, 2007 – Apr 19, 2007

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

As I understand it, according to the recvmsg() manual page, if the packet being returned is larger than the buffer provided, and the protocol does not support piecemeal reception of data, then:

(1) the buffer should be filled,

(2) MSG_TRUNC should be set in msg_flags, and

(3) the length of the full packet, including the discarded bit should be returned.

AF_NETLINK sockets, however, do not do (3). See this bit in netlink_recvmsg():

	copied = skb->len;
	if (len < copied) {
		msg->msg_flags |= MSG_TRUNC;
		copied = len;
	}

Or is this only true if the caller of recvmsg() passes MSG_TRUNC in?

David

maintainer
#1 lore raw

From: David Howells <[email protected]>
Date: Fri, 13 Apr 2007 19:08:31 +0100

Quoted 9 lines

AF_NETLINK sockets, however, do not do (3). See this bit in netlink_recvmsg():

	copied = skb->len;
	if (len < copied) {
		msg->msg_flags |= MSG_TRUNC;
		copied = len;
	}

Or is this only true if the caller of recvmsg() passes MSG_TRUNC in?

See this fix in my net-2.6.22 tree:

commit ad495d7b6cfcd1bc2eaf06c42699be0bb5d84234

Author: David S. Miller <[email protected]>
Date:   Tue Mar 6 17:02:35 2007 -0800

[NETLINK]: Mirror UDP MSG_TRUNC semantics.

If the user passes MSG_TRUNC in via msg_flags, return the full packet size not the truncated size.

Idea from Herbert Xu and Thomas Graf.

net/netlink/af_netlink.c +3 -0
index e73d8f5..b31a732 100644
@@ -1242,6 +1242,9 @@ static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
 
 	scm_recv(sock, msg, siocb->scm, flags);
 
+	if (flags & MSG_TRUNC)
+		copied = skb->len;
+
 out:
 	netlink_rcv_wake(sk);
 	return err ? : copied;
#2 lore raw
Quoted 5 lines · David Miller

See this fix in my net-2.6.22 tree:

commit ad495d7b6cfcd1bc2eaf06c42699be0bb5d84234

Author: David S. Miller <[email protected]>
Date:   Tue Mar 6 17:02:35 2007 -0800

Ummm... That seems to conflict with something in your net-2.6 tree. Which one should I use?

David