Hi,
To speed up socket lookups it is possible to use the existing pf
state. For UDP this is currently disabled as there were some
stability issues. I have strong indication that the crashes were
fixed when reference counting for pf states was implemented. So I
would like to reenable this code. Apart from the speed improvement
it also makes corner cases for pf divert-to more reliable.
ok?
bluhm
Index: netinet/udp_usrreq.c
===================================================================
RCS file: /data/mirror/openbsd/cvs/src/sys/netinet/udp_usrreq.c,v
retrieving revision 1.245
diff -u -p -r1.245 udp_usrreq.c
--- netinet/udp_usrreq.c 1 Dec 2017 10:33:33 -0000 1.245
+++ netinet/udp_usrreq.c 5 Apr 2018 11:56:49 -0000
@@ -513,7 +513,7 @@ udp_input(struct mbuf **mp, int *offp, i
/*
* Locate pcb for datagram.
*/
-#if NPF > 0 && 0 /* currently disabled */
+#if NPF > 0
inp = pf_inp_lookup(m);
#endif
if (inp == NULL) {
