On Wed, Nov 18, 2026 at 10:52 PM Numan Siddique <[email protected]> wrote:
>
> On Wed, Nov 18, 2026 at 5:45 PM Brian Haley <[email protected]> wrote:
> >
> > Thanks for the change. Just an FYI someone reported the same issue
> > against Neutron just a few days ago at [0], I had barely looked into it
> > when I saw this.
> >
> > [0] https://bugs.launchpad.net/neutron/+bug/2174452
> >
> > -Brian
> >
> > On 11/17/26 9:51 AM, Numan via dev wrote:
> > > From: Loke Berne <[email protected]>
> > >
> > > A logical switch with DNS records gets a datapath-wide flow in the
> > > ls_in_dns_lookup stage matching "udp.dst == 53" and applying the
> > > dns_lookup() action.  Any UDP packet to port 53 entering the logical
> > > switch, including unsolicited ones coming from the outside through a
> > > router port (e.g. from the internet to a NAT'ed address), is thus
> > > handed to ovn-controller, where it is parsed, before any ACL is
> > > evaluated.  An ACL can't stop this traffic, and a flood of such packets
> > > can saturate ovn-controller and ovs-vswitchd on the chassis hosting the
> > > targeted logical port.
> > >
> > > Add a priority 200 logical flow for each router port of a logical switch
> > > with DNS records, matching on "inport" and advancing the packet to the
> > > next stage without any lookup.  The priority 100 lookup flow is
> > > unchanged, so the packets entering through any other port are handled
> > > as before, and the number of added logical flows only depends on the
> > > number of router ports of the logical switch, not on its number of
> > > ports.
> > >
> > > Signed-off-by: Loke Berne <[email protected]>
>
> Thanks Loke for the patch and Brian for pointing to the neutron bug.
>
> Just FYI I'm not the author of this patch.
>
> The intention of the native DNS feature when it was added was to provide the
> native DNS resolution for the DNS requests originating from the VIF
> logical ports of
> a switch.  I think when a DNS request packet is destined to a VIF
> logical port and
> enters from outside of the switch (via a router port), it should not
> be handled by
> the OVN native dns resolver.
>
> I agree with the approach used in this patch.
>
> I've a few small nit comments in the test code of this patch.
> It would be good to address this:
>
> Even otherwise it's fine with me.
>
> Acked-by: Numan Siddique <[email protected]>
>
> I'll wait for any concerns from other maintainers before applying.
>

Thanks Loke.

I went ahead and applied the patch to main and backported to
branch-26.09 and branch-26.03.

Before applying I did the below changes.

---------------------
diff --git a/tests/ovn-northd.at b/tests/ovn-northd.at
index b0c2c22df5..813d9e365e 100644
--- a/tests/ovn-northd.at
+++ b/tests/ovn-northd.at
@@ -13198,7 +13198,10 @@ for i in 2 3 4; do
     check ovn-nbctl lsp-set-addresses sw0-vm$i "00:00:00:00:00:1$i 10.0.0.1$i"
 done
 check ovn-nbctl --wait=sb sync
-AT_CHECK([dns_lookup_flows | wc -l], [0], [3
+AT_CHECK([dns_lookup_flows], [0], [dnl
+  table=??(ls_in_dns_lookup   ), priority=0    , match=(1), action=(next;)
+  table=??(ls_in_dns_lookup   ), priority=100  , match=(udp.dst ==
53), action=(reg0[[4]] = dns_lookup(); next;)
+  table=??(ls_in_dns_lookup   ), priority=200  , match=(inport ==
"sw0-lr0"), action=(next;)
 ])

 # A second router port gets its own skip flow, and loses it when removed.
@@ -13207,11 +13210,19 @@ check ovn-nbctl lsp-set-type sw0-lr1 router
 check ovn-nbctl lsp-set-addresses sw0-lr1 router
 check ovn-nbctl lsp-set-options sw0-lr1 router-port=lrp1
 check ovn-nbctl --wait=sb sync
-AT_CHECK([dns_lookup_flows | grep -c 'priority=200'], [0], [2
+AT_CHECK([dns_lookup_flows], [0], [dnl
+  table=??(ls_in_dns_lookup   ), priority=0    , match=(1), action=(next;)
+  table=??(ls_in_dns_lookup   ), priority=100  , match=(udp.dst ==
53), action=(reg0[[4]] = dns_lookup(); next;)
+  table=??(ls_in_dns_lookup   ), priority=200  , match=(inport ==
"sw0-lr0"), action=(next;)
+  table=??(ls_in_dns_lookup   ), priority=200  , match=(inport ==
"sw0-lr1"), action=(next;)
 ])
+
 check ovn-nbctl lsp-del sw0-lr1
 check ovn-nbctl --wait=sb sync
-AT_CHECK([dns_lookup_flows | grep -c 'priority=200'], [0], [1
+AT_CHECK([dns_lookup_flows], [0], [dnl
+  table=??(ls_in_dns_lookup   ), priority=0    , match=(1), action=(next;)
+  table=??(ls_in_dns_lookup   ), priority=100  , match=(udp.dst ==
53), action=(reg0[[4]] = dns_lookup(); next;)
+  table=??(ls_in_dns_lookup   ), priority=200  , match=(inport ==
"sw0-lr0"), action=(next;)
 ])

 # The DNS response flows are unchanged.
diff --git a/northd/northd.c b/northd/northd.c
index f4cb9999b0..fa85ae6175 100644
--- a/northd/northd.c
+++ b/northd/northd.c
@@ -11571,11 +11571,7 @@ build_lswitch_dhcp_and_dns_defaults(struct
ovn_datapath *od,

 /* Logical switch ingress table 22 and 23: DNS lookup and response
 * priority 100 flows.
-*
-* Packets entering the logical switch through a router port come from
-* outside of the switch (e.g. from the internet through a NAT'ed address) and
-* must not be handed to ovn-controller: see
-* build_lswitch_dns_lookup_skip_router_port(). */
+*/
 static void
 build_lswitch_dns_lookup_and_response(struct ovn_datapath *od,
                                       struct lflow_table *lflows,
-----------------------

Thanks
Numan

> Thanks
> Numan
>
> > > ---
> > >   Documentation/ref/ovn-logical-flows.7.rst |  7 +++
> > >   northd/northd.c                           | 26 +++++++-
> > >   tests/ovn-northd.at                       | 77 +++++++++++++++++++++++
> > >   3 files changed, 109 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/Documentation/ref/ovn-logical-flows.7.rst 
> > > b/Documentation/ref/ovn-logical-flows.7.rst
> > > index 913ebbc83..4692c17cc 100644
> > > --- a/Documentation/ref/ovn-logical-flows.7.rst
> > > +++ b/Documentation/ref/ovn-logical-flows.7.rst
> > > @@ -1390,6 +1390,13 @@ Ingress Table 30 DNS Lookup
> > >   This table looks up and resolves the DNS names to the corresponding 
> > > configured
> > >   IP address(es).
> > >
> > > +- For each logical switch port ``P`` of type ``router`` of a logical 
> > > switch
> > > +  that is configured with DNS records, a priority-200 logical flow which
> > > +  matches packets with ``inport`` == ``P`` and advances them to the next 
> > > table
> > > +  without any DNS lookup. Packets coming from the routed side of the 
> > > logical
> > > +  switch, e.g. from the internet, are thus never handed to 
> > > ``ovn-controller``
> > > +  and are subject to the ACLs like any other UDP traffic.
> > > +
> > >   - A priority-100 logical flow for each logical switch datapath if it is
> > >     configured with DNS records, which matches the IPv4 and IPv6 packets 
> > > with
> > >     ``udp.dst`` = 53 and applies the action ``dns_lookup`` and advances 
> > > the packet
> > > diff --git a/northd/northd.c b/northd/northd.c
> > > index 7585767c9..a858a313c 100644
> > > --- a/northd/northd.c
> > > +++ b/northd/northd.c
> > > @@ -11568,7 +11568,11 @@ build_lswitch_dhcp_and_dns_defaults(struct 
> > > ovn_datapath *od,
> > >
> > >   /* Logical switch ingress table 22 and 23: DNS lookup and response
> > >   * priority 100 flows.
> > > -*/
> > > +*
> > > +* Packets entering the logical switch through a router port come from
> > > +* outside of the switch (e.g. from the internet through a NAT'ed 
> > > address) and
> > > +* must not be handed to ovn-controller: see
> > > +* build_lswitch_dns_lookup_skip_router_port(). */
> > >   static void
> > >   build_lswitch_dns_lookup_and_response(struct ovn_datapath *od,
> > >                                         struct lflow_table *lflows,
> > > @@ -11598,6 +11602,25 @@ build_lswitch_dns_lookup_and_response(struct 
> > > ovn_datapath *od,
> > >                     dns_match, dns_action, lflow_ref);
> > >   }
> > >
> > > +/* Table 22: DNS lookup, priority 200 flow that skips the lookup for the
> > > + * packets entering the logical switch through a router port, so that
> > > + * unsolicited DNS traffic coming from the routed side never reaches
> > > + * ovn-controller, and is subject to the ACLs like any other UDP 
> > > traffic. */
> > > +static void
> > > +build_lswitch_dns_lookup_skip_router_port(struct ovn_port *op,
> > > +                                          struct lflow_table *lflows,
> > > +                                          struct ds *match)
> > > +{
> > > +    ovs_assert(op->nbsp);
> > > +    if (!lsp_is_router(op->nbsp) || !ls_has_dns_records(op->od->nbs)) {
> > > +        return;
> > > +    }
> > > +    ds_clear(match);
> > > +    ds_put_format(match, "inport == %s", op->json_key);
> > > +    ovn_lflow_add(lflows, op->od, S_SWITCH_IN_DNS_LOOKUP, 200,
> > > +                  ds_cstr(match), "next;", op->lflow_ref);
> > > +}
> > > +
> > >   /* Table 24: External port. Drop ARP request for router ips from
> > >    * external ports  on chassis not binding those ports.
> > >    * This makes the router pipeline to be run only on the chassis
> > > @@ -20772,6 +20795,7 @@ build_lswitch_and_lrouter_iterate_by_lsp(struct 
> > > ovn_port *op,
> > >       build_lswitch_arp_nd_responder_known_ips(op, lflows, ls_ports,
> > >                                                meter_groups, actions, 
> > > match);
> > >       build_lswitch_dhcp_options_and_response(op, lflows, meter_groups);
> > > +    build_lswitch_dns_lookup_skip_router_port(op, lflows, match);
> > >       build_lswitch_external_port(op, lflows);
> > >       build_lswitch_icmp_packet_toobig_admin_flows(op, lflows, match, 
> > > actions);
> > >       build_lswitch_ip_unicast_lookup(op, lflows, actions,
> > > diff --git a/tests/ovn-northd.at b/tests/ovn-northd.at
> > > index e2b8424d2..de5241841 100644
> > > --- a/tests/ovn-northd.at
> > > +++ b/tests/ovn-northd.at
> > > @@ -13101,6 +13101,83 @@ check ovn-nbctl clear Logical_Switch sw0 
> > > dns_records
> > >   check ovn-nbctl --wait=sb sync
> > >   wait_row_count sb:DNS 0
> > >
> > > +OVN_CLEANUP_NORTHD
> > > +AT_CLEANUP
> > > +])
> > > +
> > > +OVN_FOR_EACH_NORTHD_NO_HV([
> > > +AT_SETUP([DNS lookup skipped for packets entering through router ports])
> > > +ovn_start
> > > +
> > > +check ovn-nbctl lr-add lr0
> > > +check ovn-nbctl lrp-add lr0 lrp0 00:00:00:00:00:01 10.0.0.1/24
> > > +check ovn-nbctl lrp-add lr0 lrp1 00:00:00:00:00:02 10.0.0.2/24
> > > +
> > > +check ovn-nbctl ls-add sw0
> > > +check ovn-nbctl lsp-add sw0 sw0-vm1
> > > +check ovn-nbctl lsp-set-addresses sw0-vm1 "00:00:00:00:00:11 10.0.0.11"
> > > +check ovn-nbctl lsp-add sw0 sw0-lr0
> > > +check ovn-nbctl lsp-set-type sw0-lr0 router
> > > +check ovn-nbctl lsp-set-addresses sw0-lr0 router
> > > +check ovn-nbctl lsp-set-options sw0-lr0 router-port=lrp0
> > > +
> > > +dns_uuid=$(ovn-nbctl create DNS records={})
> > > +check ovn-nbctl set DNS $dns_uuid records:vm1.ovn.org="10.0.0.11"
> > > +
> > > +dns_lookup_flows() {
> > > +    ovn-sbctl dump-flows sw0 | grep -w "ls_in_dns_lookup" | 
> > > ovn_strip_lflows | sort
> > > +}
> > > +
> > > +# Without DNS records, there is no lookup flow.
> > > +check ovn-nbctl --wait=sb sync
> > > +AT_CHECK([dns_lookup_flows], [0], [dnl
> > > +  table=??(ls_in_dns_lookup   ), priority=0    , match=(1), 
> > > action=(next;)
> > > +])
> > > +
> > > +# With DNS records, packets entering through the router port skip the 
> > > lookup,
> > > +# all the others are looked up, whatever the number of ports.
> > > +check ovn-nbctl set Logical_Switch sw0 dns_records="$dns_uuid"
> > > +check ovn-nbctl --wait=sb sync
> > > +AT_CHECK([dns_lookup_flows], [0], [dnl
> > > +  table=??(ls_in_dns_lookup   ), priority=0    , match=(1), 
> > > action=(next;)
> > > +  table=??(ls_in_dns_lookup   ), priority=100  , match=(udp.dst == 53), 
> > > action=(reg0[[4]] = dns_lookup(); next;)
> > > +  table=??(ls_in_dns_lookup   ), priority=200  , match=(inport == 
> > > "sw0-lr0"), action=(next;)
> > > +])
> > > +
> > > +# Adding VIF ports does not add lookup flows.
> > > +for i in 2 3 4; do
> > > +    check ovn-nbctl lsp-add sw0 sw0-vm$i
> > > +    check ovn-nbctl lsp-set-addresses sw0-vm$i "00:00:00:00:00:1$i 
> > > 10.0.0.1$i"
> > > +done
> > > +check ovn-nbctl --wait=sb sync
> > > +AT_CHECK([dns_lookup_flows | wc -l], [0], [3
> > > +])
> > > +
> > > +# A second router port gets its own skip flow, and loses it when removed.
> > > +check ovn-nbctl lsp-add sw0 sw0-lr1
> > > +check ovn-nbctl lsp-set-type sw0-lr1 router
> > > +check ovn-nbctl lsp-set-addresses sw0-lr1 router
> > > +check ovn-nbctl lsp-set-options sw0-lr1 router-port=lrp1
> > > +check ovn-nbctl --wait=sb sync
> > > +AT_CHECK([dns_lookup_flows | grep -c 'priority=200'], [0], [2
> > > +])
>
> Instead of grepping here with "-c" option I'd prefer you just do
> something like below
>
> AT_CHECK([dns_lookup_flows], [0], [dnl
>   table=??(ls_in_dns_lookup   ), priority=0    , match=(1), action=(next;)
>   table=??(ls_in_dns_lookup   ), priority=100  , match=(udp.dst ==
> 53), action=(reg0[[4]] = dns_lookup(); next;)
>   table=??(ls_in_dns_lookup   ), priority=200  , match=(inport ==
> "sw0-lr0"), action=(next;)
>   table=??(ls_in_dns_lookup   ), priority=200  , match=(inport ==
> "sw0-lr1"), action=(next;)
> ])
>
> Same for below.
>
> Thanks
> Numan
>
>
> > > +check ovn-nbctl lsp-del sw0-lr1
> > > +check ovn-nbctl --wait=sb sync
> > > +AT_CHECK([dns_lookup_flows | grep -c 'priority=200'], [0], [1
> > > +])
> > > +
> > > +# The DNS response flows are unchanged.
> > > +AT_CHECK([ovn-sbctl dump-flows sw0 | grep -w "ls_in_dns_response" | grep 
> > > -c "udp.dst == 53"], [0], [2
> > > +])
> > > +
> > > +# Clearing the DNS records removes every lookup flow again.
> > > +check ovn-nbctl clear Logical_Switch sw0 dns_records
> > > +check ovn-nbctl --wait=sb sync
> > > +AT_CHECK([dns_lookup_flows], [0], [dnl
> > > +  table=??(ls_in_dns_lookup   ), priority=0    , match=(1), 
> > > action=(next;)
> > > +])
> > > +
> > > +OVN_CLEANUP_NORTHD
> > >   AT_CLEANUP
> > >   ])
> > >
> >
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to