netlink: fix append of whole original message on negative ack.

PATCH In review M Lammertink 4 participants 4 messages Oct 23, 2008

Subsystems (2)

Networking Network drivers

lore

Share & tools

lore

M Lammertink

author

2008-10-23 18:18 UTC #0 lore raw

PATCH 1/1 netlink: fix append of whole original message on negative ack.

From: Michel Lammertink <mlammertink <at> ti-wmc.nl>
Signed-off-by: Michel Lammertink <mlammertink <at> ti-wmc.nl>

Netlink can return an negative acknowledgement, which AFAICT should include an error code and the whole original message. This patch fixes the size argument given to the __nlmsg_put call. Without this patch only the original header was attached.
---

Can anyone explain to me why struct nlmsgerr contains an int error and the nlmsghdr struct. The comment in netlink_ack states that the whole message should be appended. Why not only include an int error in struct nlmsgerr and just append the original message?

Comments appreciated.

net/netlink/af_netlink.c +1 -1

diff -uprN linux-2.6.27.3.orig/net/netlink/af_netlink.c linux-2.6.27.3/net/netlink/af_netlink.c
@@ -1656,7 +1656,7 @@ void netlink_ack(struct sk_buff *in_skb,
 	}
 
 	rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
-			  NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
+			  NLMSG_ERROR, payload, 0);
 	errmsg = nlmsg_data(rep);
 	errmsg->error = err;
 	memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));

David Miller

maintainer

2008-10-23 20:45 UTC #1 lore raw

From: M Lammertink <[email protected]>
Date: Thu, 23 Oct 2008 20:17:51 +0200

You'll get a much better review of this if you send it to the netdev mailing list, which is where all the networking developers are subscribed. They don't read linux-kernel

Added to CC:

Quoted 27 lines

From: Michel Lammertink <mlammertink <at> ti-wmc.nl>
Signed-off-by: Michel Lammertink <mlammertink <at> ti-wmc.nl>
Netlink can return an negative acknowledgement, which AFAICT should include an error code and the whole original message. This patch fixes the size argument given to the __nlmsg_put call. Without this patch only the original header was attached.
---
Can anyone explain to me why struct nlmsgerr contains an int error and the nlmsghdr struct. The comment in netlink_ack states that the whole message should be appended. Why not only include an int error in struct nlmsgerr and just append the original message?
Comments appreciated.
diff -uprN linux-2.6.27.3.orig/net/netlink/af_netlink.c linux-2.6.27.3/net/netlink/af_netlink.c
net/netlink/af_netlink.c +1 -2
@@ -1656,7 +1656,7 @@ void netlink_ack(struct sk_buff *in_skb,
 	}
 
 	rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
-			  NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
+			  NLMSG_ERROR, payload, 0);
 	errmsg = nlmsg_data(rep);
 	errmsg->error = err;
 	memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Evgeniy Polyakov

2008-10-23 20:55 UTC #2 lore raw

Hi.

Quoted 4 lines · David Miller

Quoted 4 lines
Netlink can return an negative acknowledgement, which AFAICT should include an error code and the whole original message. This patch fixes the size argument given to the __nlmsg_put call. Without this patch only the original header was attached.
---
Can anyone explain to me why struct nlmsgerr contains an int error and the nlmsghdr struct. The comment in netlink_ack states that the whole message should be appended. Why not only include an int error in struct nlmsgerr and just append the original message?

Having header (even seq/pid only) is enough to find a message for that userspace which cares. Looks like your patch can break existing applications which are used to work with header only, even if that was a mistake? Although application is supposed to check the length, it still can expect only a header. I have to note, that I do not know any such application though.

Signature · 1 line

	Evgeniy Polyakov

Patrick McHardy

2008-10-23 21:05 UTC #3 lore raw

In reply to David Miller

Quoted 13 lines · David Miller

From: M Lammertink <[email protected]>
Date: Thu, 23 Oct 2008 20:17:51 +0200
Quoted 10 lines
Can anyone explain to me why struct nlmsgerr contains an int error and the nlmsghdr struct. The comment in netlink_ack states that the whole message should be appended. Why not only include an int error in struct nlmsgerr and just append the original message?
Comments appreciated.
diff -uprN linux-2.6.27.3.orig/net/netlink/af_netlink.c linux-2.6.27.3/net/netlink/af_netlink.c
net/netlink/af_netlink.c +0 -0
@@ -1656,7 +1656,7 @@ void netlink_ack(struct sk_buff *in_skb,
 	}
 

Quoted 3 lines

net/netlink/af_netlink.c void netlink_ack(struct sk_buff *in_skb,
Quoted 3 lines
net/netlink/af_netlink.c +1 -1
 	rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
-			  NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
+			  NLMSG_ERROR, payload, 0)

You can't include the full message since the resulting message might exceed the 64k netlink limit.