Hi all,
in the same way of this patch, why dst_entry are stored for
RAW socket ? In case of specific IPSec rules for ICMPv6,
xfrm state can be different for the same destination.
Attached, a proposed patch.
Regards,
Nicolas
[IPV6] Don't store dst_entry for RAW socket
Signed-off-by: Nicolas DICHTEL <[EMAIL PROTECTED]>
Mitsuru KANDA a écrit :
Hello,
I recreated a patch for unconnected udpv6 socket checking.
(Fixed forgetting dst refcnt decrement from the previous patch.)
Regards,
-mk
At Thu, 22 Sep 2005 23:44:05 -0700 (PDT),
"David S. Miller" <[EMAIL PROTECTED]> wrote:
...
I think you're patch adds a route leak. If we elide the
ip6_dst_store() where does the 'dst' reference go?
In fact, I think the UDPv6 change might have done the same thing.
Mitsuru-san?
ipv4 code like this does explicit dst_clone() when storeing
the 'dst' to the socket, but I can't see where ipv6 is doing
this. Perhaps ipv6 dst's work a little differently. :-)
--- linux-2.6.15.2/net/ipv6/raw.c 2006-01-31 07:25:07.000000000 +0100
+++ linux-2.6.15.2-new/net/ipv6/raw.c 2006-02-06 11:46:13.000000000 +0100
@@ -814,10 +814,7 @@
err = rawv6_push_pending_frames(sk, &fl, rp);
}
done:
- ip6_dst_store(sk, dst,
- ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
- &np->daddr : NULL);
-
+ dst_release(dst);
release_sock(sk);
out:
fl6_sock_release(flowlabel);
