Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 5 Sep 2007 23:16:35 +0200
From:      Max Laier <max@love2party.net>
To:        freebsd-pf@freebsd.org
Subject:   Re: pfsync errors
Message-ID:  <200709052316.41257.max@love2party.net>
In-Reply-To: <e667a90b0709051331x35cafdfw50ee0be40969aa30@mail.gmail.com>
References:  <e667a90b0709051331x35cafdfw50ee0be40969aa30@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
--nextPart3147102.ViYZi2FlqM
Content-Type: multipart/mixed;
  boundary="Boundary-01=_0yx3GPUlzgcylMK"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

--Boundary-01=_0yx3GPUlzgcylMK
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

On Wednesday 05 September 2007, Rian Shelley wrote:
> As far as I can tell, am having the same problem described by bill
> marquette. I have two firewalls using pfsync, where the secondary
> firewall just increases its state count steadily.
>
> I created a simple libpcap program to watch the pfsync headers flowing
> by, and i see types 8, 4, 2, which are PFSYNC_ACT_UREQ,
> PFSYNC_ACT_UPD_C, PFSYNC_ACT_UPD. I dont see any of type 3 or 5, which
> are the ones that delete state. As far as i can tell, states are
> pumped across the link, but never removed and are left to time out on
> their own.

Very good observation.  I don't quite believe that you don't see *any*=20
three or fives, but I do see that those would get lost most easily.  The=20
problem stems from the way states are purged in 3.7/RELENG_6.  Newer pf=20
4.1/(soon to be)RELENG_7 splits the state removal.

I'm attaching a *very* experimental *HACK* that might help the situation. =
=20
I believe however, that you would be better off with moving to=20
4.1/RELENG_6 (patches at [1]) or 4.1/RELENG_7 as soon as it's done.  The=20
state purge is one of the biggest weaknesses of 3.7/RELENG_6 which isn't=20
easily solveable.

Another way to go is setting the queuelength for the internal processing=20
queue to something insanely high (1000+).  This will most likely work=20
around the problem at the cost of burning (mbuf) memory.

[1] http://people.freebsd.org/~mlaier/PF41/

=2D-=20
/"\  Best regards,                      | mlaier@freebsd.org
\ /  Max Laier                          | ICQ #67774661
 X   http://pf4freebsd.love2party.net/  | mlaier@EFnet
/ \  ASCII Ribbon Campaign              | Against HTML Mail and News

--Boundary-01=_0yx3GPUlzgcylMK
Content-Type: text/x-diff; charset="iso-8859-1"; name="pf_purge.hack.diff"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment;
	filename="pf_purge.hack.diff"

Index: pf.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
RCS file: /usr/store/mlaier/fcvs/src/sys/contrib/pf/net/pf.c,v
retrieving revision 1.34.2.5
diff -u -r1.34.2.5 pf.c
=2D-- pf.c	28 Jul 2007 22:32:57 -0000	1.34.2.5
+++ pf.c	9 Aug 2007 20:51:42 -0000
@@ -1145,17 +1145,34 @@
 	pf_status.states--;
 }
=20
+static struct pf_state *pf_purge_pickup;
+
 void
 pf_purge_expired_states(void)
 {
 	struct pf_state		*cur, *next;
+	int			 max_purge =3D 5000;
+
+	if (pf_purge_pickup !=3D NULL)
+		cur =3D pf_purge_pickup;
+	else
+		cur =3D RB_MIN(pf_state_tree_id, &tree_id);
=20
=2D	for (cur =3D RB_MIN(pf_state_tree_id, &tree_id);
=2D	    cur; cur =3D next) {
+	pf_purge_pickup =3D NULL;
+	for (;cur && max_purge; max_purge--, cur =3D next) {
 		next =3D RB_NEXT(pf_state_tree_id, &tree_id, cur);
 		if (pf_state_expires(cur) <=3D time_second)
 			pf_purge_expired_state(cur);
 	}
+	if (max_purge) {
+		cur =3D RB_MIN(pf_state_tree_id, &tree_id);
+		for (;cur && max_purge; max_purge--, cur =3D next) {
+			next =3D RB_NEXT(pf_state_tree_id, &tree_id, cur);
+			if (pf_state_expires(cur) <=3D time_second)
+				pf_purge_expired_state(cur);
+		}
+	}
+	pf_purge_pickup =3D cur;
 }
=20
 int

--Boundary-01=_0yx3GPUlzgcylMK--

--nextPart3147102.ViYZi2FlqM
Content-Type: application/pgp-signature; name=signature.asc 
Content-Description: This is a digitally signed message part.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)

iD8DBQBG3xy5XyyEoT62BG0RAqk5AJ0TvzCPnRQN3bs2wlPBSQCxPVzIggCeJ0Gi
/qKGaIoHHDv41N9aJS+zLVQ=
=L2wy
-----END PGP SIGNATURE-----

--nextPart3147102.ViYZi2FlqM--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200709052316.41257.max>