Skip to contentSkip to conversation

net: bridge: forwarding of unknown IPv4/IPv6/MAC BUM traffic

RFC PATCH In review Joachim Wiberg 3 participants 38 messages Apr 11, 2022 – Apr 13, 2022

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

cover net: bridge: forwarding of unknown IPv4/IPv6/MAC BUM traffic

This is a proposal to improve forwarding control of BUM traffic to the bridge itself. Another, related, issue regarding loss of function when an IP multicast router is detected, is also brought up here which can be a separate series in v2.

First, we add BROPT_BCAST_FLOOD, BROPT_UNICAST_FLOOD, BROPT_MCAST_FLOOD flags for unknown traffic to the bridge itself, with netlink support and a selftest. We ensure backwards compatible forwarding behavior is preserved by enabling these flags by default. Please note however, these flags do not affect the behavior of IFF_PROMISC on the bridge interface.

Second, and with the above in place, we set out to verify flooding of unknown *and* known multicast to regular bridge ports, including the bridge itself. We use tcpdump -p in the tests to ensure we do not trigger IFF_PROIMISC. Unknown multicast should be forwarded according to the MCAST_FLOOD flag, and known multicast according the MDB and to multicast router ports.

We find that forwarding of unknown IP multicast stops[1] as soon as a multicast router is known. Affecting all ports, including the bridge itself (this series). The root cause for this is the classification mrouters_only in br_multicast_rcv().

Dropping this classification of unknown IP multicast and moving the multicast flow handling from br_multicast_flood() to br_flood() fixes this problem. The bridge now properly floods all unknown multicast, and this can now be controlled using the MCAST_FLOOD flag on all ports including the bridge itself.

The use of br_flood() opens up the need for the multicast 'rport' API, but that is behind CONFIG_BRIDGE_IGMP_SNOOPING, which is the primary reason for this being an RFC series. I'd love some feedback on how to go about all this, opening up that API and even the take on the issue as a whole.

Honestly, despite taking great care to not change the bridge's default behavior the patch series in itself propose quite radical changes that alone mandate RFC status at this point. There has been some discussion already on this in 20220410220324.4c3l3idubwi3w6if@skbuf and I expect more disucssion here.

Note: this series builds upon my previous patch for host l2 mdb entries,

      [email protected], extending the test
      bridge_mdb.sh

Best regards
/Joachim

[1]: MAC multicast is not affected.

net/bridge/br_device.c                        |   4 +
 net/bridge/br_forward.c                       |  11 +
 net/bridge/br_input.c                         |  11 +-
 net/bridge/br_multicast.c                     |   6 +-
 net/bridge/br_netlink.c                       | 170 +++++++---
 net/bridge/br_private.h                       |   7 +-
 net/bridge/br_switchdev.c                     |   8 +-
 net/bridge/br_sysfs_if.c                      |   2 +-
 .../drivers/net/ocelot/tc_flower_chains.sh    |  24 +-
 .../testing/selftests/net/forwarding/Makefile |   3 +-
 .../selftests/net/forwarding/bridge_flood.sh  | 170 ++++++++++
 .../selftests/net/forwarding/bridge_mdb.sh    | 321 +++++++++++++++++-
 tools/testing/selftests/net/forwarding/lib.sh |  33 +-
 13 files changed, 683 insertions(+), 87 deletions(-)
 create mode 100755 tools/testing/selftests/net/forwarding/bridge_flood.sh
 mode change 100644 => 100755 tools/testing/selftests/net/forwarding/lib.sh
#1 lore raw

RFC PATCH 1/13 net: bridge: add control of bum flooding to bridge itself

The bridge itself is also a port, but unfortunately it does not (yet) have a 'struct net_bridge_port'. However, in many cases we want to treat it as a proper port so concessions have been made, e.g., NULL port or host_joined attributes.

This patch is an attempt to more of the same by adding support for controlling flooding of unknown broadcast/unicast/multicast to the bridge. Something we often also want to control in an offloaded switching fabric.

---
 net/bridge/br_device.c  |  4 ++++
 net/bridge/br_input.c   | 11 ++++++++---
 net/bridge/br_private.h |  3 +++
 3 files changed, 15 insertions(+), 3 deletions(-)
net/bridge/br_device.c +4 -0
index 8d6bab244c4a..0aa7d21ac82c 100644
@@ -526,6 +526,10 @@ void br_dev_setup(struct net_device *dev)
 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 	dev->max_mtu = ETH_MAX_MTU;
 
+	br_opt_toggle(br, BROPT_UNICAST_FLOOD, 1);
+	br_opt_toggle(br, BROPT_MCAST_FLOOD, 1);
+	br_opt_toggle(br, BROPT_BCAST_FLOOD, 1);
+
 	br_netfilter_rtable_init(br);
 	br_stp_timer_init(br);
 	br_multicast_init(br);
net/bridge/br_input.c +8 -3
index 196417859c4a..d439b876bdf5 100644
@@ -118,7 +118,8 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 		/* by definition the broadcast is also a multicast address */
 		if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
 			pkt_type = BR_PKT_BROADCAST;
-			local_rcv = true;
+			if (br_opt_get(br, BROPT_BCAST_FLOOD))
+				local_rcv = true;
 		} else {
 			pkt_type = BR_PKT_MULTICAST;
 			if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
@@ -161,12 +162,16 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 			}
 			mcast_hit = true;
 		} else {
-			local_rcv = true;
-			br->dev->stats.multicast++;
+			if (br_opt_get(br, BROPT_MCAST_FLOOD)) {
+				local_rcv = true;
+				br->dev->stats.multicast++;
+			}
 		}
 		break;
 	case BR_PKT_UNICAST:
 		dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
+		if (!dst && br_opt_get(br, BROPT_UNICAST_FLOOD))
+			local_rcv = true;
 		break;
 	default:
 		break;
net/bridge/br_private.h +3 -0
index 18ccc3d5d296..683bd0ee4c64 100644
@@ -449,6 +449,9 @@ enum net_bridge_opts {
 	BROPT_VLAN_BRIDGE_BINDING,
 	BROPT_MCAST_VLAN_SNOOPING_ENABLED,
 	BROPT_MST_ENABLED,
+	BROPT_UNICAST_FLOOD,
+	BROPT_MCAST_FLOOD,
+	BROPT_BCAST_FLOOD,
 };
 
 struct net_bridge {
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH RFC net-next 01/13] net: bridge: add control of bum flooding to bridge itself

Quoted 22 lines · Joachim Wiberg

The bridge itself is also a port, but unfortunately it does not (yet) have a 'struct net_bridge_port'. However, in many cases we want to treat it as a proper port so concessions have been made, e.g., NULL port or host_joined attributes.

This patch is an attempt to more of the same by adding support for controlling flooding of unknown broadcast/unicast/multicast to the bridge. Something we often also want to control in an offloaded switching fabric.

---
 net/bridge/br_device.c  |  4 ++++
 net/bridge/br_input.c   | 11 ++++++++---
 net/bridge/br_private.h |  3 +++
 3 files changed, 15 insertions(+), 3 deletions(-)
net/bridge/br_device.c +0 -0
index 8d6bab244c4a..0aa7d21ac82c 100644

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 529

Quoted 4 lines
net/bridge/br_device.c +1 -0
@@ -526,6 +526,10 @@ void br_dev_setup(struct net_device *dev)
 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 	dev->max_mtu = ETH_MAX_MTU;
 
+	br_opt_toggle(br, BROPT_UNICAST_FLOOD, 1);

This one must be false by default. It changes current default behaviour. Unknown unicast is not currently passed up to the bridge if the port is not in promisc mode, this will change it. You'll have to make it consistent with promisc (e.g. one way would be for promisc always to enable unicast flood and it won't be possible to be disabled while promisc).

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 531

Quoted 2 lines
net/bridge/br_device.c +2 -0
+	br_opt_toggle(br, BROPT_MCAST_FLOOD, 1);
+	br_opt_toggle(br, BROPT_BCAST_FLOOD, 1);

s/1/true/ for consistency

Quoted 25 lines
net/bridge/br_device.c +1 -0
+
 	br_netfilter_rtable_init(br);
 	br_stp_timer_init(br);
 	br_multicast_init(br);
net/bridge/br_input.c +3 -3
index 196417859c4a..d439b876bdf5 100644
@@ -118,7 +118,8 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 		/* by definition the broadcast is also a multicast address */
 		if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
 			pkt_type = BR_PKT_BROADCAST;
-			local_rcv = true;
+			if (br_opt_get(br, BROPT_BCAST_FLOOD))
+				local_rcv = true;
 		} else {
 			pkt_type = BR_PKT_MULTICAST;
 			if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
@@ -161,12 +162,16 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 			}
 			mcast_hit = true;
 		} else {
-			local_rcv = true;
-			br->dev->stats.multicast++;
+			if (br_opt_get(br, BROPT_MCAST_FLOOD)) {

net/bridge/br_input.c int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb · line 175

Quoted 10 lines
net/bridge/br_input.c +5 -0
+				local_rcv = true;
+				br->dev->stats.multicast++;
+			}
 		}
 		break;
 	case BR_PKT_UNICAST:
 		dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
+		if (!dst && br_opt_get(br, BROPT_UNICAST_FLOOD))
+			local_rcv = true;
 		break;

This adds new tests for all fast paths for host traffic, especially the port - port communication which is the most critical one. Please at least move the unicast test to the "else" block of "if (dst)" later.

The other tests can be moved to host only code too, but would require bigger changes. Please try to keep the impact on the fast-path at minimum.

Quoted 16 lines
net/bridge/br_input.c +0 -0
 	default:
 		break;
net/bridge/br_private.h +3 -0
index 18ccc3d5d296..683bd0ee4c64 100644
@@ -449,6 +449,9 @@ enum net_bridge_opts {
 	BROPT_VLAN_BRIDGE_BINDING,
 	BROPT_MCAST_VLAN_SNOOPING_ENABLED,
 	BROPT_MST_ENABLED,
+	BROPT_UNICAST_FLOOD,
+	BROPT_MCAST_FLOOD,
+	BROPT_BCAST_FLOOD,
 };
 
 struct net_bridge {
#3 lore raw

Re: [PATCH RFC net-next 01/13] net: bridge: add control of bum flooding to bridge itself

Quoted 26 lines · Nikolay Aleksandrov
Quoted 25 lines · Joachim Wiberg

The bridge itself is also a port, but unfortunately it does not (yet) have a 'struct net_bridge_port'. However, in many cases we want to treat it as a proper port so concessions have been made, e.g., NULL port or host_joined attributes.

This patch is an attempt to more of the same by adding support for controlling flooding of unknown broadcast/unicast/multicast to the bridge. Something we often also want to control in an offloaded switching fabric.

---
 net/bridge/br_device.c  |  4 ++++
 net/bridge/br_input.c   | 11 ++++++++---
 net/bridge/br_private.h |  3 +++
 3 files changed, 15 insertions(+), 3 deletions(-)
net/bridge/br_device.c +0 -0
index 8d6bab244c4a..0aa7d21ac82c 100644
@@ -526,6 +526,10 @@ void br_dev_setup(struct net_device *dev)
 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 	dev->max_mtu = ETH_MAX_MTU;
 
Quoted 4 lines

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 529

Quoted 1 line
net/bridge/br_device.c +1 -0
+	br_opt_toggle(br, BROPT_UNICAST_FLOOD, 1);

This one must be false by default. It changes current default behaviour. Unknown unicast is not currently passed up to the bridge if the port is

s/port/bridge/ in promisc mode

Quoted 69 lines

not in promisc mode, this will change it. You'll have to make it consistent with promisc (e.g. one way would be for promisc always to enable unicast flood and it won't be possible to be disabled while promisc).

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 531

Quoted 2 lines
net/bridge/br_device.c +2 -0
+	br_opt_toggle(br, BROPT_MCAST_FLOOD, 1);
+	br_opt_toggle(br, BROPT_BCAST_FLOOD, 1);

s/1/true/ for consistency

Quoted 25 lines
net/bridge/br_device.c +1 -0
+
 	br_netfilter_rtable_init(br);
 	br_stp_timer_init(br);
 	br_multicast_init(br);
net/bridge/br_input.c +3 -3
index 196417859c4a..d439b876bdf5 100644
@@ -118,7 +118,8 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 		/* by definition the broadcast is also a multicast address */
 		if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
 			pkt_type = BR_PKT_BROADCAST;
-			local_rcv = true;
+			if (br_opt_get(br, BROPT_BCAST_FLOOD))
+				local_rcv = true;
 		} else {
 			pkt_type = BR_PKT_MULTICAST;
 			if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
@@ -161,12 +162,16 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 			}
 			mcast_hit = true;
 		} else {
-			local_rcv = true;
-			br->dev->stats.multicast++;
+			if (br_opt_get(br, BROPT_MCAST_FLOOD)) {

net/bridge/br_input.c int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb · line 175

Quoted 10 lines
net/bridge/br_input.c +5 -0
+				local_rcv = true;
+				br->dev->stats.multicast++;
+			}
 		}
 		break;
 	case BR_PKT_UNICAST:
 		dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
+		if (!dst && br_opt_get(br, BROPT_UNICAST_FLOOD))
+			local_rcv = true;
 		break;

This adds new tests for all fast paths for host traffic, especially the port - port communication which is the most critical one. Please at least move the unicast test to the "else" block of "if (dst)" later.

The other tests can be moved to host only code too, but would require bigger changes. Please try to keep the impact on the fast-path at minimum.

Quoted 16 lines
net/bridge/br_input.c +0 -0
 	default:
 		break;
net/bridge/br_private.h +3 -0
index 18ccc3d5d296..683bd0ee4c64 100644
@@ -449,6 +449,9 @@ enum net_bridge_opts {
 	BROPT_VLAN_BRIDGE_BINDING,
 	BROPT_MCAST_VLAN_SNOOPING_ENABLED,
 	BROPT_MST_ENABLED,
+	BROPT_UNICAST_FLOOD,
+	BROPT_MCAST_FLOOD,
+	BROPT_BCAST_FLOOD,
 };
 
 struct net_bridge {
#4 lore raw

Re: [PATCH RFC net-next 01/13] net: bridge: add control of bum flooding to bridge itself

In reply to Nikolay Aleksandrov
Quoted 10 lines · Nikolay Aleksandrov

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 528

Quoted 4 lines · Joachim Wiberg
net/bridge/br_device.c +1 -0
@@ -526,6 +526,10 @@ void br_dev_setup(struct net_device *dev)
 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 	dev->max_mtu = ETH_MAX_MTU;
+	br_opt_toggle(br, BROPT_UNICAST_FLOOD, 1);

This one must be false by default. It changes current default behaviour. Unknown unicast is not currently passed up to the bridge if the port is not in promisc mode, this will change it. You'll have to make it consistent with promisc (e.g. one way would be for promisc always to enable unicast flood and it won't be possible to be disabled while promisc).

Ouch, my bad! Will look into how to let this have as little impact as possible. I like your semantics there, promisc should always win.

Quoted 4 lines

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 531

Quoted 2 lines
net/bridge/br_device.c +2 -0
+	br_opt_toggle(br, BROPT_MCAST_FLOOD, 1);
+	br_opt_toggle(br, BROPT_BCAST_FLOOD, 1);

s/1/true/ for consistency

Of course, thanks!

Quoted 24 lines
Quoted 24 lines
net/bridge/br_input.c +6 -3
@@ -118,7 +118,8 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 		/* by definition the broadcast is also a multicast address */
 		if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
 			pkt_type = BR_PKT_BROADCAST;
-			local_rcv = true;
+			if (br_opt_get(br, BROPT_BCAST_FLOOD))
+				local_rcv = true;
 		} else {
 			pkt_type = BR_PKT_MULTICAST;
 			if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
@@ -161,12 +162,16 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 			}
 			mcast_hit = true;
 		} else {
-			local_rcv = true;
-			br->dev->stats.multicast++;
+			if (br_opt_get(br, BROPT_MCAST_FLOOD)) {
+				local_rcv = true;
+				br->dev->stats.multicast++;
+			}
 		}
 		break;
 	case BR_PKT_UNICAST:
 		dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
Quoted 8 lines

net/bridge/br_input.c int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb · line 175

Quoted 3 lines
net/bridge/br_input.c +2 -0
+		if (!dst && br_opt_get(br, BROPT_UNICAST_FLOOD))
+			local_rcv = true;
 		break;

This adds new tests for all fast paths for host traffic, especially the port - port communication which is the most critical one. Please at least move the unicast test to the "else" block of "if (dst)" later.

OK, will fix!

Quoted 3 lines

The other tests can be moved to host only code too, but would require bigger changes. Please try to keep the impact on the fast-path at minimum.

Interesting, you mean by speculatively setting local_rcv = true and postpone the decsion to br_pass_frame_up(), right? Yeah that would indeed be a bit more work.

#5 lore raw

Re: [PATCH RFC net-next 01/13] net: bridge: add control of bum flooding to bridge itself

Quoted 57 lines · Joachim Wiberg
Quoted 10 lines · Nikolay Aleksandrov

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 528

Quoted 4 lines · Joachim Wiberg
net/bridge/br_device.c +1 -0
@@ -526,6 +526,10 @@ void br_dev_setup(struct net_device *dev)
 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 	dev->max_mtu = ETH_MAX_MTU;
+	br_opt_toggle(br, BROPT_UNICAST_FLOOD, 1);

This one must be false by default. It changes current default behaviour. Unknown unicast is not currently passed up to the bridge if the port is not in promisc mode, this will change it. You'll have to make it consistent with promisc (e.g. one way would be for promisc always to enable unicast flood and it won't be possible to be disabled while promisc).

Ouch, my bad! Will look into how to let this have as little impact as possible. I like your semantics there, promisc should always win.

Quoted 4 lines

net/bridge/br_device.c void br_dev_setup(struct net_device *dev) · line 531

Quoted 2 lines
net/bridge/br_device.c +2 -0
+	br_opt_toggle(br, BROPT_MCAST_FLOOD, 1);
+	br_opt_toggle(br, BROPT_BCAST_FLOOD, 1);

s/1/true/ for consistency

Of course, thanks!

Quoted 24 lines
Quoted 24 lines
net/bridge/br_input.c +6 -3
@@ -118,7 +118,8 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 		/* by definition the broadcast is also a multicast address */
 		if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
 			pkt_type = BR_PKT_BROADCAST;
-			local_rcv = true;
+			if (br_opt_get(br, BROPT_BCAST_FLOOD))
+				local_rcv = true;
 		} else {
 			pkt_type = BR_PKT_MULTICAST;
 			if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
@@ -161,12 +162,16 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
 			}
 			mcast_hit = true;
 		} else {
-			local_rcv = true;
-			br->dev->stats.multicast++;
+			if (br_opt_get(br, BROPT_MCAST_FLOOD)) {
+				local_rcv = true;
+				br->dev->stats.multicast++;
+			}
 		}
 		break;
 	case BR_PKT_UNICAST:
 		dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
Quoted 8 lines

net/bridge/br_input.c int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb · line 175

Quoted 3 lines
net/bridge/br_input.c +2 -0
+		if (!dst && br_opt_get(br, BROPT_UNICAST_FLOOD))
+			local_rcv = true;
 		break;

This adds new tests for all fast paths for host traffic, especially the port - port communication which is the most critical one. Please at least move the unicast test to the "else" block of "if (dst)" later.

OK, will fix!

Quoted 7 lines
Quoted 3 lines

The other tests can be moved to host only code too, but would require bigger changes. Please try to keep the impact on the fast-path at minimum.

Interesting, you mean by speculatively setting local_rcv = true and postpone the decsion to br_pass_frame_up(), right? Yeah that would indeed be a bit more work.

Yes, I was thinking maybe local_rcv can become an enum with an exact reason for the local_rcv, so if it's > 0 do the local_rcv and br_pass_frame_up() then can make a proper decision without passing all of the vars. I haven't tried it, so not sure if it's feasible. :)

#6 lore raw

Re: [PATCH RFC net-next 01/13] net: bridge: add control of bum flooding to bridge itself

Quoted 8 lines · Nikolay Aleksandrov
Quoted 3 lines · Joachim Wiberg

Interesting, you mean by speculatively setting local_rcv = true and postpone the decsion to br_pass_frame_up(), right? Yeah that would indeed be a bit more work.

Yes, I was thinking maybe local_rcv can become an enum with an exact reason for the local_rcv, so if it's > 0 do the local_rcv and br_pass_frame_up() then can make a proper decision without passing all of the vars. I haven't tried it, so not sure if it's feasible. :)

Ah, yeah that could definitely work. Thanks, I'll keep that in mind! :)

#7 lore raw

RFC PATCH 2/13 net: bridge: rename br_switchdev_set_port_flag() to .._dev_flag()

In reply to Joachim Wiberg

The br_switchdev_set_port_flag() function only uses the p->dev member of the port, and we also want to reuse it for the bridge itself. Change the first argument and the name of the function to match.

---
 net/bridge/br_netlink.c   | 2 +-
 net/bridge/br_private.h   | 4 ++--
 net/bridge/br_switchdev.c | 8 ++++----
 net/bridge/br_sysfs_if.c  | 2 +-
 4 files changed, 8 insertions(+), 8 deletions(-)
net/bridge/br_netlink.c +1 -1
index 200ad05b296f..7fca8ff13ec7 100644
@@ -942,7 +942,7 @@ static int br_setport(struct net_bridge_port *p, struct nlattr *tb[],
 
 	changed_mask = old_flags ^ p->flags;
 
-	err = br_switchdev_set_port_flag(p, p->flags, changed_mask, extack);
+	err = br_switchdev_set_dev_flag(p->dev, p->flags, changed_mask, extack);
 	if (err) {
 		p->flags = old_flags;
 		return err;
net/bridge/br_private.h +2 -2
index 683bd0ee4c64..6c230be6fbf5 100644
@@ -2038,7 +2038,7 @@ void nbp_switchdev_frame_mark(const struct net_bridge_port *p,
 			      struct sk_buff *skb);
 bool nbp_switchdev_allowed_egress(const struct net_bridge_port *p,
 				  const struct sk_buff *skb);
-int br_switchdev_set_port_flag(struct net_bridge_port *p,
+int br_switchdev_set_dev_flag(struct net_device *dev,
 			       unsigned long flags,
 			       unsigned long mask,
 			       struct netlink_ext_ack *extack);
@@ -2108,7 +2108,7 @@ static inline bool nbp_switchdev_allowed_egress(const struct net_bridge_port *p,
 	return true;
 }
 
-static inline int br_switchdev_set_port_flag(struct net_bridge_port *p,
+static inline int br_switchdev_set_dev_flag(struct net_device *dev,
 					     unsigned long flags,
 					     unsigned long mask,
 					     struct netlink_ext_ack *extack)
net/bridge/br_switchdev.c +4 -4
index 8cc44c367231..3b4df63e96a6 100644
@@ -74,13 +74,13 @@ bool nbp_switchdev_allowed_egress(const struct net_bridge_port *p,
 #define BR_PORT_FLAGS_HW_OFFLOAD (BR_LEARNING | BR_FLOOD | \
 				  BR_MCAST_FLOOD | BR_BCAST_FLOOD | BR_PORT_LOCKED)
 
-int br_switchdev_set_port_flag(struct net_bridge_port *p,
+int br_switchdev_set_dev_flag(struct net_device *dev,
 			       unsigned long flags,
 			       unsigned long mask,
 			       struct netlink_ext_ack *extack)
 {
 	struct switchdev_attr attr = {
-		.orig_dev = p->dev,
+		.orig_dev = dev,
 	};
 	struct switchdev_notifier_port_attr_info info = {
 		.attr = &attr,
@@ -96,7 +96,7 @@ int br_switchdev_set_port_flag(struct net_bridge_port *p,
 	attr.u.brport_flags.mask = mask;
 
 	/* We run from atomic context here */
-	err = call_switchdev_notifiers(SWITCHDEV_PORT_ATTR_SET, p->dev,
+	err = call_switchdev_notifiers(SWITCHDEV_PORT_ATTR_SET, dev,
 				       &info.info, extack);
 	err = notifier_to_errno(err);
 	if (err == -EOPNOTSUPP)
@@ -112,7 +112,7 @@ int br_switchdev_set_port_flag(struct net_bridge_port *p,
 	attr.id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS;
 	attr.flags = SWITCHDEV_F_DEFER;
 
-	err = switchdev_port_attr_set(p->dev, &attr, extack);
+	err = switchdev_port_attr_set(dev, &attr, extack);
 	if (err) {
 		if (extack && !extack->_msg)
 			NL_SET_ERR_MSG_MOD(extack,
net/bridge/br_sysfs_if.c +1 -1
index 07fa76080512..c222c68105f1 100644
@@ -69,7 +69,7 @@ static int store_flag(struct net_bridge_port *p, unsigned long v,
 		flags &= ~mask;
 
 	if (flags != p->flags) {
-		err = br_switchdev_set_port_flag(p, flags, mask, &extack);
+		err = br_switchdev_set_dev_flag(p->dev, flags, mask, &extack);
 		if (err) {
 			netdev_err(p->dev, "%s\n", extack._msg);
 			return err;
Signature · 1 line
2.25.1
#8 lore raw

RFC PATCH 3/13 net: bridge: minor refactor of br_setlink() for readability

In reply to Joachim Wiberg

The br_setlink() function extracts the struct net_bridge pointer a bit sloppy. It's easy to interpret the code wrong. This patch attempts to clear things up a bit.

---
 net/bridge/br_netlink.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
net/bridge/br_netlink.c +6 -4
index 7fca8ff13ec7..8f4297287b32 100644
@@ -1040,6 +1040,8 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 		return 0;
 
 	p = br_port_get_rtnl(dev);
+	if (p)
+		br = p->br;
 	/* We want to accept dev as bridge itself if the AF_SPEC
 	 * is set to see if someone is setting vlan info on the bridge
 	 */
@@ -1055,17 +1057,17 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 			if (err)
 				return err;
 
-			spin_lock_bh(&p->br->lock);
+			spin_lock_bh(&br->lock);
 			err = br_setport(p, tb, extack);
-			spin_unlock_bh(&p->br->lock);
+			spin_unlock_bh(&br->lock);
 		} else {
 			/* Binary compatibility with old RSTP */
 			if (nla_len(protinfo) < sizeof(u8))
 				return -EINVAL;
 
-			spin_lock_bh(&p->br->lock);
+			spin_lock_bh(&br->lock);
 			err = br_set_port_state(p, nla_get_u8(protinfo));
-			spin_unlock_bh(&p->br->lock);
+			spin_unlock_bh(&br->lock);
 		}
 		if (err)
 			goto out;
Signature · 1 line
2.25.1
#9 lore raw

Re: [PATCH RFC net-next 03/13] net: bridge: minor refactor of br_setlink() for readability

Quoted 9 lines · Joachim Wiberg

The br_setlink() function extracts the struct net_bridge pointer a bit sloppy. It's easy to interpret the code wrong. This patch attempts to clear things up a bit.

---
 net/bridge/br_netlink.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

I think you can make it more straight-forward, remove the first br = netdev_priv and do something like (completely untested):
...
struct net_bridge_port *p = NULL;
...
if (netif_is_bridge_master(dev)) {
br = netdev_priv(dev);
} else {
p = br_port_get_rtnl(dev);
if (WARN_ON(!p))

	return -EINVAL;

br = p->br;
}

So br is always and only set in this block.

Quoted 35 lines
net/bridge/br_netlink.c +6 -4
index 7fca8ff13ec7..8f4297287b32 100644
@@ -1040,6 +1040,8 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 		return 0;
 
 	p = br_port_get_rtnl(dev);
+	if (p)
+		br = p->br;
 	/* We want to accept dev as bridge itself if the AF_SPEC
 	 * is set to see if someone is setting vlan info on the bridge
 	 */
@@ -1055,17 +1057,17 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 			if (err)
 				return err;
 
-			spin_lock_bh(&p->br->lock);
+			spin_lock_bh(&br->lock);
 			err = br_setport(p, tb, extack);
-			spin_unlock_bh(&p->br->lock);
+			spin_unlock_bh(&br->lock);
 		} else {
 			/* Binary compatibility with old RSTP */
 			if (nla_len(protinfo) < sizeof(u8))
 				return -EINVAL;
 
-			spin_lock_bh(&p->br->lock);
+			spin_lock_bh(&br->lock);
 			err = br_set_port_state(p, nla_get_u8(protinfo));
-			spin_unlock_bh(&p->br->lock);
+			spin_unlock_bh(&br->lock);
 		}
 		if (err)
 			goto out;
#10 lore raw

Re: [PATCH RFC net-next 03/13] net: bridge: minor refactor of br_setlink() for readability

Quoted 14 lines · Nikolay Aleksandrov
Quoted 3 lines · Joachim Wiberg

The br_setlink() function extracts the struct net_bridge pointer a bit sloppy. It's easy to interpret the code wrong. This patch attempts to clear things up a bit.

I think you can make it more straight-forward, remove the first br = netdev_priv and do something like (completely untested):
...
struct net_bridge_port *p = NULL;
...
if (netif_is_bridge_master(dev)) {
br = netdev_priv(dev);
} else {
p = br_port_get_rtnl(dev);
if (WARN_ON(!p))

Quoted 5 lines
	return -EINVAL;

br = p->br;
}

So br is always and only set in this block.

Yes, this is much better, thank you! I took the misguided approach of minmizing my change. I'll update and include in the non-RFC patch series I send next.

#11 lore raw

RFC PATCH 4/13 net: bridge: netlink support for controlling BUM flooding to bridge

In reply to Joachim Wiberg

This patch adds netlink support for controlling the new broadcast, unicast, and multicast flooding flags to the bridge itself.

The messy part is in br_setport(), which re-indents a large block of code for the port settings. To reduce code duplication a few new variables have been added; new_flags and dev. The latter is used for the recently renamed br_switchdev_set_dev_flag(), which can now be used by underlying switching fabric drivers as another source of information when controlling flooding of unknown BUM traffic to the CPU port.

---
 net/bridge/br_netlink.c | 160 ++++++++++++++++++++++++++++++----------
 1 file changed, 123 insertions(+), 37 deletions(-)
net/bridge/br_netlink.c +123 -37
index 8f4297287b32..68bbf703b31a 100644
@@ -225,13 +225,29 @@ static inline size_t br_nlmsg_size(struct net_device *dev, u32 filter_mask)
 		+ nla_total_size(4); /* IFLA_BRPORT_BACKUP_PORT */
 }
 
-static int br_port_fill_attrs(struct sk_buff *skb,
+static int br_port_fill_attrs(struct sk_buff *skb, const struct net_bridge *br,
 			      const struct net_bridge_port *p)
 {
-	u8 mode = !!(p->flags & BR_HAIRPIN_MODE);
 	struct net_bridge_port *backup_p;
 	u64 timerval;
+	u8 mode;
 
+	if (!p) {
+		if (!br)
+			return -EINVAL;
+
+		if (nla_put_u8(skb, IFLA_BRPORT_UNICAST_FLOOD,
+			       br_opt_get(br, BROPT_UNICAST_FLOOD)) ||
+		    nla_put_u8(skb, IFLA_BRPORT_MCAST_FLOOD,
+			       br_opt_get(br, BROPT_MCAST_FLOOD)) ||
+		    nla_put_u8(skb, IFLA_BRPORT_BCAST_FLOOD,
+			       br_opt_get(br, BROPT_BCAST_FLOOD)))
+			return -EMSGSIZE;
+
+		return 0;
+	}
+
+	mode = !!(p->flags & BR_HAIRPIN_MODE);
 	if (nla_put_u8(skb, IFLA_BRPORT_STATE, p->state) ||
 	    nla_put_u16(skb, IFLA_BRPORT_PRIORITY, p->priority) ||
 	    nla_put_u32(skb, IFLA_BRPORT_COST, p->path_cost) ||
@@ -475,11 +491,11 @@ static int br_fill_ifinfo(struct sk_buff *skb,
 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
 		goto nla_put_failure;
 
-	if (event == RTM_NEWLINK && port) {
+	if (event == RTM_NEWLINK) {
 		struct nlattr *nest;
 
 		nest = nla_nest_start(skb, IFLA_PROTINFO);
-		if (nest == NULL || br_port_fill_attrs(skb, port) < 0)
+		if (!nest || br_port_fill_attrs(skb, br, port) < 0)
 			goto nla_put_failure;
 		nla_nest_end(skb, nest);
 	}
@@ -911,43 +927,113 @@ static void br_set_port_flag(struct net_bridge_port *p, struct nlattr *tb[],
 		p->flags &= ~mask;
 }
 
+/* Map bridge options to brport flags */
+static unsigned long br_boolopt_map_flags(struct br_boolopt_multi *bm)
+{
+	unsigned long bitmap = bm->optmask;
+	unsigned long bitmask = 0;
+	int opt_id;
+
+	for_each_set_bit(opt_id, &bitmap, BR_BOOLOPT_MAX) {
+		if (!(bm->optval & BIT(opt_id)))
+			continue;
+
+		switch (opt_id) {
+		case BROPT_UNICAST_FLOOD:
+			bitmask |= BR_FLOOD;
+			break;
+		case BROPT_MCAST_FLOOD:
+			bitmask |= BR_MCAST_FLOOD;
+			break;
+		case BROPT_BCAST_FLOOD:
+			bitmask |= BR_BCAST_FLOOD;
+			break;
+		}
+	}
+
+	return bitmask;
+}
+
+static void br_set_bropt(struct net_bridge *br, struct nlattr *tb[],
+			 int attrtype, enum net_bridge_opts opt)
+{
+	if (!tb[attrtype])
+		return;
+
+	br_opt_toggle(br, opt, !!nla_get_u8(tb[attrtype]));
+}
+
+#define BROPT_MASK (BROPT_UNICAST_FLOOD | BROPT_MCAST_FLOOD | BROPT_MCAST_FLOOD)
+
 /* Process bridge protocol info on port */
-static int br_setport(struct net_bridge_port *p, struct nlattr *tb[],
-		      struct netlink_ext_ack *extack)
+static int br_setport(struct net_bridge *br, struct net_bridge_port *p,
+		      struct nlattr *tb[], struct netlink_ext_ack *extack)
 {
-	unsigned long old_flags, changed_mask;
+	unsigned long old_flags, new_flags, changed_mask;
+	struct br_boolopt_multi old_opts = {
+		.optmask = BROPT_MASK
+	};
 	bool br_vlan_tunnel_old;
+	struct net_device *dev;
 	int err;
 
-	old_flags = p->flags;
-	br_vlan_tunnel_old = (old_flags & BR_VLAN_TUNNEL) ? true : false;
-
-	br_set_port_flag(p, tb, IFLA_BRPORT_MODE, BR_HAIRPIN_MODE);
-	br_set_port_flag(p, tb, IFLA_BRPORT_GUARD, BR_BPDU_GUARD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_FAST_LEAVE,
-			 BR_MULTICAST_FAST_LEAVE);
-	br_set_port_flag(p, tb, IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK);
-	br_set_port_flag(p, tb, IFLA_BRPORT_LEARNING, BR_LEARNING);
-	br_set_port_flag(p, tb, IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_TO_UCAST,
-			 BR_MULTICAST_TO_UNICAST);
-	br_set_port_flag(p, tb, IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP, BR_PROXYARP);
-	br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP_WIFI, BR_PROXYARP_WIFI);
-	br_set_port_flag(p, tb, IFLA_BRPORT_VLAN_TUNNEL, BR_VLAN_TUNNEL);
-	br_set_port_flag(p, tb, IFLA_BRPORT_NEIGH_SUPPRESS, BR_NEIGH_SUPPRESS);
-	br_set_port_flag(p, tb, IFLA_BRPORT_ISOLATED, BR_ISOLATED);
-	br_set_port_flag(p, tb, IFLA_BRPORT_LOCKED, BR_PORT_LOCKED);
-
-	changed_mask = old_flags ^ p->flags;
-
-	err = br_switchdev_set_dev_flag(p->dev, p->flags, changed_mask, extack);
+	if (p) {
+		old_flags = p->flags;
+		br_vlan_tunnel_old = (old_flags & BR_VLAN_TUNNEL) ? true : false;
+
+		br_set_port_flag(p, tb, IFLA_BRPORT_MODE, BR_HAIRPIN_MODE);
+		br_set_port_flag(p, tb, IFLA_BRPORT_GUARD, BR_BPDU_GUARD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_FAST_LEAVE,
+				 BR_MULTICAST_FAST_LEAVE);
+		br_set_port_flag(p, tb, IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK);
+		br_set_port_flag(p, tb, IFLA_BRPORT_LEARNING, BR_LEARNING);
+		br_set_port_flag(p, tb, IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_TO_UCAST,
+				 BR_MULTICAST_TO_UNICAST);
+		br_set_port_flag(p, tb, IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP, BR_PROXYARP);
+		br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP_WIFI, BR_PROXYARP_WIFI);
+		br_set_port_flag(p, tb, IFLA_BRPORT_VLAN_TUNNEL, BR_VLAN_TUNNEL);
+		br_set_port_flag(p, tb, IFLA_BRPORT_NEIGH_SUPPRESS, BR_NEIGH_SUPPRESS);
+		br_set_port_flag(p, tb, IFLA_BRPORT_ISOLATED, BR_ISOLATED);
+		br_set_port_flag(p, tb, IFLA_BRPORT_LOCKED, BR_PORT_LOCKED);
+
+		new_flags = p->flags;
+		dev = p->dev;
+	} else {
+		struct br_boolopt_multi opts = {
+			.optmask = BROPT_MASK
+		};
+
+		br_boolopt_multi_get(br, &old_opts);
+		old_flags = br_boolopt_map_flags(&old_opts);
+
+		br_set_bropt(br, tb, IFLA_BRPORT_UNICAST_FLOOD, BROPT_UNICAST_FLOOD);
+		br_set_bropt(br, tb, IFLA_BRPORT_MCAST_FLOOD, BROPT_MCAST_FLOOD);
+		br_set_bropt(br, tb, IFLA_BRPORT_BCAST_FLOOD, BROPT_BCAST_FLOOD);
+
+		br_boolopt_multi_get(br, &opts);
+		new_flags = br_boolopt_map_flags(&opts);
+		dev = br->dev;
+	}
+
+	changed_mask = old_flags ^ new_flags;
+
+	err = br_switchdev_set_dev_flag(dev, new_flags, changed_mask, extack);
 	if (err) {
-		p->flags = old_flags;
+		if (!p)
+			br_boolopt_multi_toggle(br, &old_opts, extack);
+		else
+			p->flags = old_flags;
+
 		return err;
 	}
 
+	/* Skip the rest for the bridge itself, for now */
+	if (!p)
+		return 0;
+
 	if (br_vlan_tunnel_old && !(p->flags & BR_VLAN_TUNNEL))
 		nbp_vlan_tunnel_info_flush(p);
 
@@ -1048,7 +1134,7 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 	if (!p && !afspec)
 		return -EINVAL;
 
-	if (p && protinfo) {
+	if (protinfo) {
 		if (protinfo->nla_type & NLA_F_NESTED) {
 			err = nla_parse_nested_deprecated(tb, IFLA_BRPORT_MAX,
 							  protinfo,
@@ -1058,9 +1144,9 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 				return err;
 
 			spin_lock_bh(&br->lock);
-			err = br_setport(p, tb, extack);
+			err = br_setport(br, p, tb, extack);
 			spin_unlock_bh(&br->lock);
-		} else {
+		} else if (p) {
 			/* Binary compatibility with old RSTP */
 			if (nla_len(protinfo) < sizeof(u8))
 				return -EINVAL;
@@ -1153,7 +1239,7 @@ static int br_port_slave_changelink(struct net_device *brdev,
 		return 0;
 
 	spin_lock_bh(&br->lock);
-	ret = br_setport(br_port_get_rtnl(dev), data, extack);
+	ret = br_setport(br, br_port_get_rtnl(dev), data, extack);
 	spin_unlock_bh(&br->lock);
 
 	return ret;
@@ -1163,7 +1249,7 @@ static int br_port_fill_slave_info(struct sk_buff *skb,
 				   const struct net_device *brdev,
 				   const struct net_device *dev)
 {
-	return br_port_fill_attrs(skb, br_port_get_rtnl(dev));
+	return br_port_fill_attrs(skb, NULL, br_port_get_rtnl(dev));
 }
 
 static size_t br_port_get_slave_size(const struct net_device *brdev,
Signature · 1 line
2.25.1
#12 lore raw

Re: [PATCH RFC net-next 04/13] net: bridge: netlink support for controlling BUM flooding to bridge

Quoted 10 lines · Joachim Wiberg

This patch adds netlink support for controlling the new broadcast, unicast, and multicast flooding flags to the bridge itself.

The messy part is in br_setport(), which re-indents a large block of code for the port settings. To reduce code duplication a few new variables have been added; new_flags and dev. The latter is used for the recently renamed br_switchdev_set_dev_flag(), which can now be used by underlying switching fabric drivers as another source of information when controlling flooding of unknown BUM traffic to the CPU port.

Quoted 2 lines

---

Absolutely not. This is just wrong on a few levels and way too hacky.

Please separate the bridge handling altogether and make it clean. No need to integrate it with the port handling, also I think you've missed a few things about bool options, more below...

For boolopts examples you can check BR_BOOLOPT_NO_LL_LEARN, BR_BOOLOPT_MCAST_VLAN_SNOOPING and BR_BOOLOPT_MST_ENABLE.

Quoted 21 lines
 net/bridge/br_netlink.c | 160 ++++++++++++++++++++++++++++++----------
 1 file changed, 123 insertions(+), 37 deletions(-)
net/bridge/br_netlink.c +3 -2
index 8f4297287b32..68bbf703b31a 100644
@@ -225,13 +225,29 @@ static inline size_t br_nlmsg_size(struct net_device *dev, u32 filter_mask)
 		+ nla_total_size(4); /* IFLA_BRPORT_BACKUP_PORT */
 }
 
-static int br_port_fill_attrs(struct sk_buff *skb,
+static int br_port_fill_attrs(struct sk_buff *skb, const struct net_bridge *br,
 			      const struct net_bridge_port *p)
 {
-	u8 mode = !!(p->flags & BR_HAIRPIN_MODE);
 	struct net_bridge_port *backup_p;
 	u64 timerval;
+	u8 mode;
 
+	if (!p) {

net/bridge/br_netlink.c static inline size_t br_nlmsg_size(struct net_device *dev, u32 filter_mask) · line 245

Quoted 10 lines
net/bridge/br_netlink.c +10 -0
+		if (!br)
+			return -EINVAL;
+
+		if (nla_put_u8(skb, IFLA_BRPORT_UNICAST_FLOOD,
+			       br_opt_get(br, BROPT_UNICAST_FLOOD)) ||
+		    nla_put_u8(skb, IFLA_BRPORT_MCAST_FLOOD,
+			       br_opt_get(br, BROPT_MCAST_FLOOD)) ||
+		    nla_put_u8(skb, IFLA_BRPORT_BCAST_FLOOD,
+			       br_opt_get(br, BROPT_BCAST_FLOOD)))
+			return -EMSGSIZE;

No. Bool opts are already exposed through IFLA_BR_MULTI_BOOLOPT.

Quoted 51 lines
net/bridge/br_netlink.c +32 -2
+
+		return 0;
+	}
+
+	mode = !!(p->flags & BR_HAIRPIN_MODE);
 	if (nla_put_u8(skb, IFLA_BRPORT_STATE, p->state) ||
 	    nla_put_u16(skb, IFLA_BRPORT_PRIORITY, p->priority) ||
 	    nla_put_u32(skb, IFLA_BRPORT_COST, p->path_cost) ||
@@ -475,11 +491,11 @@ static int br_fill_ifinfo(struct sk_buff *skb,
 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
 		goto nla_put_failure;
 
-	if (event == RTM_NEWLINK && port) {
+	if (event == RTM_NEWLINK) {
 		struct nlattr *nest;
 
 		nest = nla_nest_start(skb, IFLA_PROTINFO);
-		if (nest == NULL || br_port_fill_attrs(skb, port) < 0)
+		if (!nest || br_port_fill_attrs(skb, br, port) < 0)
 			goto nla_put_failure;
 		nla_nest_end(skb, nest);
 	}
@@ -911,43 +927,113 @@ static void br_set_port_flag(struct net_bridge_port *p, struct nlattr *tb[],
 		p->flags &= ~mask;
 }
 
+/* Map bridge options to brport flags */
+static unsigned long br_boolopt_map_flags(struct br_boolopt_multi *bm)
+{
+	unsigned long bitmap = bm->optmask;
+	unsigned long bitmask = 0;
+	int opt_id;
+
+	for_each_set_bit(opt_id, &bitmap, BR_BOOLOPT_MAX) {
+		if (!(bm->optval & BIT(opt_id)))
+			continue;
+
+		switch (opt_id) {
+		case BROPT_UNICAST_FLOOD:
+			bitmask |= BR_FLOOD;
+			break;
+		case BROPT_MCAST_FLOOD:
+			bitmask |= BR_MCAST_FLOOD;
+			break;
+		case BROPT_BCAST_FLOOD:
+			bitmask |= BR_BCAST_FLOOD;
+			break;
+		}
+	}
+
+	return bitmask;

net/bridge/br_netlink.c static void br_set_port_flag(struct net_bridge_port *p, struct nlattr *tb[], · line 964

Quoted 10 lines
net/bridge/br_netlink.c +10 -0
+}
+
+static void br_set_bropt(struct net_bridge *br, struct nlattr *tb[],
+			 int attrtype, enum net_bridge_opts opt)
+{
+	if (!tb[attrtype])
+		return;
+
+	br_opt_toggle(br, opt, !!nla_get_u8(tb[attrtype]));
+}

These must be controlled via the boolopt api and attributes, not through additional nl attributes.

Quoted 113 lines
net/bridge/br_netlink.c +64 -29
+
+#define BROPT_MASK (BROPT_UNICAST_FLOOD | BROPT_MCAST_FLOOD | BROPT_MCAST_FLOOD)
+
 /* Process bridge protocol info on port */
-static int br_setport(struct net_bridge_port *p, struct nlattr *tb[],
-		      struct netlink_ext_ack *extack)
+static int br_setport(struct net_bridge *br, struct net_bridge_port *p,
+		      struct nlattr *tb[], struct netlink_ext_ack *extack)
 {
-	unsigned long old_flags, changed_mask;
+	unsigned long old_flags, new_flags, changed_mask;
+	struct br_boolopt_multi old_opts = {
+		.optmask = BROPT_MASK
+	};
 	bool br_vlan_tunnel_old;
+	struct net_device *dev;
 	int err;
 
-	old_flags = p->flags;
-	br_vlan_tunnel_old = (old_flags & BR_VLAN_TUNNEL) ? true : false;
-
-	br_set_port_flag(p, tb, IFLA_BRPORT_MODE, BR_HAIRPIN_MODE);
-	br_set_port_flag(p, tb, IFLA_BRPORT_GUARD, BR_BPDU_GUARD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_FAST_LEAVE,
-			 BR_MULTICAST_FAST_LEAVE);
-	br_set_port_flag(p, tb, IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK);
-	br_set_port_flag(p, tb, IFLA_BRPORT_LEARNING, BR_LEARNING);
-	br_set_port_flag(p, tb, IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_TO_UCAST,
-			 BR_MULTICAST_TO_UNICAST);
-	br_set_port_flag(p, tb, IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD);
-	br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP, BR_PROXYARP);
-	br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP_WIFI, BR_PROXYARP_WIFI);
-	br_set_port_flag(p, tb, IFLA_BRPORT_VLAN_TUNNEL, BR_VLAN_TUNNEL);
-	br_set_port_flag(p, tb, IFLA_BRPORT_NEIGH_SUPPRESS, BR_NEIGH_SUPPRESS);
-	br_set_port_flag(p, tb, IFLA_BRPORT_ISOLATED, BR_ISOLATED);
-	br_set_port_flag(p, tb, IFLA_BRPORT_LOCKED, BR_PORT_LOCKED);
-
-	changed_mask = old_flags ^ p->flags;
-
-	err = br_switchdev_set_dev_flag(p->dev, p->flags, changed_mask, extack);
+	if (p) {
+		old_flags = p->flags;
+		br_vlan_tunnel_old = (old_flags & BR_VLAN_TUNNEL) ? true : false;
+
+		br_set_port_flag(p, tb, IFLA_BRPORT_MODE, BR_HAIRPIN_MODE);
+		br_set_port_flag(p, tb, IFLA_BRPORT_GUARD, BR_BPDU_GUARD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_FAST_LEAVE,
+				 BR_MULTICAST_FAST_LEAVE);
+		br_set_port_flag(p, tb, IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK);
+		br_set_port_flag(p, tb, IFLA_BRPORT_LEARNING, BR_LEARNING);
+		br_set_port_flag(p, tb, IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_MCAST_TO_UCAST,
+				 BR_MULTICAST_TO_UNICAST);
+		br_set_port_flag(p, tb, IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD);
+		br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP, BR_PROXYARP);
+		br_set_port_flag(p, tb, IFLA_BRPORT_PROXYARP_WIFI, BR_PROXYARP_WIFI);
+		br_set_port_flag(p, tb, IFLA_BRPORT_VLAN_TUNNEL, BR_VLAN_TUNNEL);
+		br_set_port_flag(p, tb, IFLA_BRPORT_NEIGH_SUPPRESS, BR_NEIGH_SUPPRESS);
+		br_set_port_flag(p, tb, IFLA_BRPORT_ISOLATED, BR_ISOLATED);
+		br_set_port_flag(p, tb, IFLA_BRPORT_LOCKED, BR_PORT_LOCKED);
+
+		new_flags = p->flags;
+		dev = p->dev;
+	} else {
+		struct br_boolopt_multi opts = {
+			.optmask = BROPT_MASK
+		};
+
+		br_boolopt_multi_get(br, &old_opts);
+		old_flags = br_boolopt_map_flags(&old_opts);
+
+		br_set_bropt(br, tb, IFLA_BRPORT_UNICAST_FLOOD, BROPT_UNICAST_FLOOD);
+		br_set_bropt(br, tb, IFLA_BRPORT_MCAST_FLOOD, BROPT_MCAST_FLOOD);
+		br_set_bropt(br, tb, IFLA_BRPORT_BCAST_FLOOD, BROPT_BCAST_FLOOD);
+
+		br_boolopt_multi_get(br, &opts);
+		new_flags = br_boolopt_map_flags(&opts);
+		dev = br->dev;
+	}
+
+	changed_mask = old_flags ^ new_flags;
+
+	err = br_switchdev_set_dev_flag(dev, new_flags, changed_mask, extack);
 	if (err) {
-		p->flags = old_flags;
+		if (!p)
+			br_boolopt_multi_toggle(br, &old_opts, extack);
+		else
+			p->flags = old_flags;
+
 		return err;
 	}
 
+	/* Skip the rest for the bridge itself, for now */
+	if (!p)
+		return 0;
+
 	if (br_vlan_tunnel_old && !(p->flags & BR_VLAN_TUNNEL))
 		nbp_vlan_tunnel_info_flush(p);
 
@@ -1048,7 +1134,7 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 	if (!p && !afspec)
 		return -EINVAL;
 
-	if (p && protinfo) {
+	if (protinfo) {
 		if (protinfo->nla_type & NLA_F_NESTED) {
 			err = nla_parse_nested_deprecated(tb, IFLA_BRPORT_MAX,
 							  protinfo,

net/bridge/br_netlink.c int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags, · line 1147

Quoted 5 lines
net/bridge/br_netlink.c +1 -1
@@ -1058,9 +1144,9 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 				return err;
 
 			spin_lock_bh(&br->lock);
-			err = br_setport(p, tb, extack);
+			err = br_setport(br, p, tb, extack);

setport is for *port* only...

Quoted 24 lines
net/bridge/br_netlink.c +3 -3
 			spin_unlock_bh(&br->lock);
-		} else {
+		} else if (p) {
 			/* Binary compatibility with old RSTP */
 			if (nla_len(protinfo) < sizeof(u8))
 				return -EINVAL;
@@ -1153,7 +1239,7 @@ static int br_port_slave_changelink(struct net_device *brdev,
 		return 0;
 
 	spin_lock_bh(&br->lock);
-	ret = br_setport(br_port_get_rtnl(dev), data, extack);
+	ret = br_setport(br, br_port_get_rtnl(dev), data, extack);
 	spin_unlock_bh(&br->lock);
 
 	return ret;
@@ -1163,7 +1249,7 @@ static int br_port_fill_slave_info(struct sk_buff *skb,
 				   const struct net_device *brdev,
 				   const struct net_device *dev)
 {
-	return br_port_fill_attrs(skb, br_port_get_rtnl(dev));
+	return br_port_fill_attrs(skb, NULL, br_port_get_rtnl(dev));
 }
 
 static size_t br_port_get_slave_size(const struct net_device *brdev,
#13 lore raw

Re: [PATCH RFC net-next 04/13] net: bridge: netlink support for controlling BUM flooding to bridge

Quoted 10 lines · Nikolay Aleksandrov
Quoted 6 lines · Joachim Wiberg

The messy part is in br_setport(), which re-indents a large block of code for the port settings. To reduce code duplication a few new variables have been added; new_flags and dev. The latter is used for the recently renamed br_switchdev_set_dev_flag(), which can now be used by underlying switching fabric drivers as another source of information when controlling flooding of unknown BUM traffic to the CPU port.

Absolutely not. This is just wrong on a few levels and way too hacky. Please separate the bridge handling altogether and make it clean. No need to integrate it with the port handling,

OK, I'll have a go at it.

Quoted 5 lines

also I think you've missed a few things about bool options, more below...

For boolopts examples you can check BR_BOOLOPT_NO_LL_LEARN, BR_BOOLOPT_MCAST_VLAN_SNOOPING and BR_BOOLOPT_MST_ENABLE.

Ah yes, will read up on those, thanks!

Quoted 8 lines

net/bridge/br_netlink.c static inline size_t br_nlmsg_size(struct net_device *dev, u32 filter_mask) · line 245

Quoted 7 lines
net/bridge/br_netlink.c +7 -0
+		if (nla_put_u8(skb, IFLA_BRPORT_UNICAST_FLOOD,
+			       br_opt_get(br, BROPT_UNICAST_FLOOD)) ||
+		    nla_put_u8(skb, IFLA_BRPORT_MCAST_FLOOD,
+			       br_opt_get(br, BROPT_MCAST_FLOOD)) ||
+		    nla_put_u8(skb, IFLA_BRPORT_BCAST_FLOOD,
+			       br_opt_get(br, BROPT_BCAST_FLOOD)))
+			return -EMSGSIZE;

No. Bool opts are already exposed through IFLA_BR_MULTI_BOOLOPT.

Aha, there it is, awesome!

Quoted 10 lines

net/bridge/br_netlink.c static void br_set_port_flag(struct net_bridge_port *p, struct nlattr *tb[], · line 964

Quoted 8 lines
net/bridge/br_netlink.c +8 -0
+static void br_set_bropt(struct net_bridge *br, struct nlattr *tb[],
+			 int attrtype, enum net_bridge_opts opt)
+{
+	if (!tb[attrtype])
+		return;
+
+	br_opt_toggle(br, opt, !!nla_get_u8(tb[attrtype]));
+}

These must be controlled via the boolopt api and attributes, not through additional nl attributes.

Understood.

Quoted 7 lines

net/bridge/br_netlink.c int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags, · line 1147

Quoted 6 lines
net/bridge/br_netlink.c +1 -1
@@ -1058,9 +1144,9 @@ int br_setlink(struct net_device *dev, struct nlmsghdr *nlh, u16 flags,
 				return err;
 
 			spin_lock_bh(&br->lock);
-			err = br_setport(p, tb, extack);
+			err = br_setport(br, p, tb, extack);

setport is for *port* only...

A-firm.

Thank you for the honest review. Netlink is still much of a mystery in many ways to me.

Best regards
/Joachim

#14 lore raw

RFC PATCH 5/13 selftests: forwarding: add TCPDUMP_EXTRA_FLAGS to lib.sh

In reply to Joachim Wiberg

For some use-cases we may want to change the tcpdump flags used in tcpdump_start(). For instance, observing interfaces without the PROMISC flag, e.g. to see what's really being forwarded to the bridge interface.

---
 tools/testing/selftests/net/forwarding/lib.sh | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
 mode change 100644 => 100755 tools/testing/selftests/net/forwarding/lib.sh
tools/testing/selftests/net/forwarding/lib.sh +7 -1
old mode 100644
new mode 100755
index 664b9ecaf228..00cdcab7accf
@@ -1369,7 +1369,13 @@ tcpdump_start()
 		capuser="-Z $SUDO_USER"
 	fi
 
-	$ns_cmd tcpdump -e -n -Q in -i $if_name \
+	if [ -z $TCPDUMP_EXTRA_FLAGS ]; then
+		extra_flags=""
+	else
+		extra_flags="$TCPDUMP_EXTRA_FLAGS"
+	fi
+
+	$ns_cmd tcpdump $extra_flags -e -n -Q in -i $if_name \
 		-s 65535 -B 32768 $capuser -w $capfile > "$capout" 2>&1 &
 	cappid=$!
 
Signature · 1 line
2.25.1
maintainer
#15 lore raw

Re: [PATCH RFC net-next 05/13] selftests: forwarding: add TCPDUMP_EXTRA_FLAGS to lib.sh

Quoted 18 lines · Joachim Wiberg

For some use-cases we may want to change the tcpdump flags used in tcpdump_start(). For instance, observing interfaces without the PROMISC flag, e.g. to see what's really being forwarded to the bridge interface.

---
 tools/testing/selftests/net/forwarding/lib.sh | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
 mode change 100644 => 100755 tools/testing/selftests/net/forwarding/lib.sh
tools/testing/selftests/net/forwarding/lib.sh +0 -0
old mode 100644
new mode 100755
index 664b9ecaf228..00cdcab7accf
@@ -1369,7 +1369,13 @@ tcpdump_start()
 		capuser="-Z $SUDO_USER"

tools/testing/selftests/net/forwarding/lib.sh tcpdump_start() · line 1378

Quoted 10 lines
tools/testing/selftests/net/forwarding/lib.sh +7 -1
 	fi
 
-	$ns_cmd tcpdump -e -n -Q in -i $if_name \
+	if [ -z $TCPDUMP_EXTRA_FLAGS ]; then
+		extra_flags=""
+	else
+		extra_flags="$TCPDUMP_EXTRA_FLAGS"
+	fi
+
+	$ns_cmd tcpdump $extra_flags -e -n -Q in -i $if_name \

Could you call directly "$ns_cmd tcpdump $TCPDUMP_EXTRA_FLAGS ..." here, without an intermediary "extra_flags" global variable which holds the same content?

You could initialize it just like the way other variables are initialized, at the beginning of lib.sh:

TCPDUMP_EXTRA_FLAGS=${TCPDUMP_EXTRA_FLAGS:=}

Quoted 6 lines
tools/testing/selftests/net/forwarding/lib.sh +0 -0
 		-s 65535 -B 32768 $capuser -w $capfile > "$capout" 2>&1 &
 	cappid=$!
 
Signature · 1 line
2.25.1
#16 lore raw

Re: [PATCH RFC net-next 05/13] selftests: forwarding: add TCPDUMP_EXTRA_FLAGS to lib.sh

Quoted 14 lines · Vladimir Oltean

tools/testing/selftests/net/forwarding/lib.sh tcpdump_start() · line 1378

Quoted 8 lines · Joachim Wiberg
tools/testing/selftests/net/forwarding/lib.sh +7 -1
-	$ns_cmd tcpdump -e -n -Q in -i $if_name \
+	if [ -z $TCPDUMP_EXTRA_FLAGS ]; then
+		extra_flags=""
+	else
+		extra_flags="$TCPDUMP_EXTRA_FLAGS"
+	fi
+
+	$ns_cmd tcpdump $extra_flags -e -n -Q in -i $if_name \

Could you call directly "$ns_cmd tcpdump $TCPDUMP_EXTRA_FLAGS ..." here, without an intermediary "extra_flags" global variable which holds the same content?

Quoted 4 lines

You could initialize it just like the way other variables are initialized, at the beginning of lib.sh:

TCPDUMP_EXTRA_FLAGS=${TCPDUMP_EXTRA_FLAGS:=}

Ah, yes of course. Will fix in the next drop!

/J

#17 lore raw

RFC PATCH 6/13 selftests: forwarding: multiple instances in tcpdump helper

In reply to Joachim Wiberg

Extend tcpdump_start() & C:o to handle multiple instances. Useful when observing bridge operation, e.g., unicast learning/flooding, and any case of multicast distribution (to these ports but not that one ...).

This means the interface argument is now a mandatory argument to all tcpdump_*() functions, hence the changes to the ocelot flower test.

---
 .../drivers/net/ocelot/tc_flower_chains.sh    | 24 +++++++++---------
 tools/testing/selftests/net/forwarding/lib.sh | 25 +++++++++++++------
 2 files changed, 30 insertions(+), 19 deletions(-)
tools/testing/selftests/drivers/net/ocelot/tc_flower_chains.sh +12 -12
index eaf8a04a7ca5..7e684e27a682 100755
@@ -215,15 +215,15 @@ test_vlan_pop()
 
 	sleep 1
 
-	tcpdump_stop
+	tcpdump_stop $eth2
 
-	if tcpdump_show | grep -q "$eth3_mac > $eth2_mac, ethertype IPv4"; then
+	if tcpdump_show $eth2 | grep -q "$eth3_mac > $eth2_mac, ethertype IPv4"; then
 		echo "OK"
 	else
 		echo "FAIL"
 	fi
 
-	tcpdump_cleanup
+	tcpdump_cleanup $eth2
 }
 
 test_vlan_push()
@@ -236,15 +236,15 @@ test_vlan_push()
 
 	sleep 1
 
-	tcpdump_stop
+	tcpdump_stop $eth3.100
 
-	if tcpdump_show | grep -q "$eth2_mac > $eth3_mac"; then
+	if tcpdump_show $eth3.100 | grep -q "$eth2_mac > $eth3_mac"; then
 		echo "OK"
 	else
 		echo "FAIL"
 	fi
 
-	tcpdump_cleanup
+	tcpdump_cleanup $eth3.100
 }
 
 test_vlan_ingress_modify()
@@ -267,15 +267,15 @@ test_vlan_ingress_modify()
 
 	sleep 1
 
-	tcpdump_stop
+	tcpdump_stop $eth2
 
-	if tcpdump_show | grep -q "$eth3_mac > $eth2_mac, .* vlan 300"; then
+	if tcpdump_show $eth2 | grep -q "$eth3_mac > $eth2_mac, .* vlan 300"; then
 		echo "OK"
 	else
 		echo "FAIL"
 	fi
 
-	tcpdump_cleanup
+	tcpdump_cleanup $eth2
 
 	tc filter del dev $eth0 ingress chain $(IS1 2) pref 3
 
@@ -305,15 +305,15 @@ test_vlan_egress_modify()
 
 	sleep 1
 
-	tcpdump_stop
+	tcpdump_stop $eth2
 
-	if tcpdump_show | grep -q "$eth3_mac > $eth2_mac, .* vlan 300"; then
+	if tcpdump_show $eth2 | grep -q "$eth3_mac > $eth2_mac, .* vlan 300"; then
 		echo "OK"
 	else
 		echo "FAIL"
 	fi
 
-	tcpdump_cleanup
+	tcpdump_cleanup $eth2
 
 	tc filter del dev $eth1 egress chain $(ES0) pref 3
 	tc qdisc del dev $eth1 clsact
tools/testing/selftests/net/forwarding/lib.sh +18 -7
index 00cdcab7accf..20a6d6b2f389 100755
@@ -1349,13 +1349,17 @@ stop_traffic()
 	{ kill %% && wait %%; } 2>/dev/null
 }
 
+declare -A cappid
+declare -A capfile
+declare -A capout
+
 tcpdump_start()
 {
 	local if_name=$1; shift
 	local ns=$1; shift
 
-	capfile=$(mktemp)
-	capout=$(mktemp)
+	capfile[$if_name]=$(mktemp)
+	capout[$if_name]=$(mktemp)
 
 	if [ -z $ns ]; then
 		ns_cmd=""
@@ -1376,26 +1380,33 @@ tcpdump_start()
 	fi
 
 	$ns_cmd tcpdump $extra_flags -e -n -Q in -i $if_name \
-		-s 65535 -B 32768 $capuser -w $capfile > "$capout" 2>&1 &
-	cappid=$!
+		-s 65535 -B 32768 $capuser -w ${capfile[$if_name]} > "${capout[$if_name]}" 2>&1 &
+	cappid[$if_name]=$!
 
 	sleep 1
 }
 
 tcpdump_stop()
 {
-	$ns_cmd kill $cappid
+	local if_name=$1
+	local pid=${cappid[$if_name]}
+
+	$ns_cmd kill "$pid" && wait "$pid"
 	sleep 1
 }
 
 tcpdump_cleanup()
 {
-	rm $capfile $capout
+	local if_name=$1
+
+	rm ${capfile[$if_name]} ${capout[$if_name]}
 }
 
 tcpdump_show()
 {
-	tcpdump -e -n -r $capfile 2>&1
+	local if_name=$1
+
+	tcpdump -e -n -r ${capfile[$if_name]} 2>&1
 }
 
 # return 0 if the packet wasn't seen on host2_if or 1 if it was
Signature · 1 line
2.25.1
maintainer Reviewed-by on v1 6/13
#18 lore raw

Re: [PATCH RFC net-next 06/13] selftests: forwarding: multiple instances in tcpdump helper

Quoted 9 lines · Joachim Wiberg

Extend tcpdump_start() & C:o to handle multiple instances. Useful when observing bridge operation, e.g., unicast learning/flooding, and any case of multicast distribution (to these ports but not that one ...).

This means the interface argument is now a mandatory argument to all tcpdump_*() functions, hence the changes to the ocelot flower test.

---

#19 lore raw

RFC PATCH 7/13 selftests: forwarding: new test, verify bridge flood flags

In reply to Joachim Wiberg

Test per-port flood control flags of unknown BUM traffic by injecting bc/uc/mc on one bridge port and verifying it being forwarded to both the bridge itself and another regular bridge port.

---
 .../testing/selftests/net/forwarding/Makefile |   3 +-
 .../selftests/net/forwarding/bridge_flood.sh  | 170 ++++++++++++++++++
 2 files changed, 172 insertions(+), 1 deletion(-)
 create mode 100755 tools/testing/selftests/net/forwarding/bridge_flood.sh
tools/testing/selftests/net/forwarding/Makefile +2 -1
index ae80c2aef577..873fa61d1ee1 100644
@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0+ OR MIT
 
-TEST_PROGS = bridge_igmp.sh \
+TEST_PROGS = bridge_flood.sh \
+	bridge_igmp.sh \
 	bridge_locked_port.sh \
 	bridge_mdb.sh \
 	bridge_port_isolation.sh \
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +170 -0

170 lines not on this page · raw message

Signature · 1 line
2.25.1
maintainer
#20 lore raw

Re: [PATCH RFC net-next 07/13] selftests: forwarding: new test, verify bridge flood flags

Quoted 30 lines · Joachim Wiberg

Test per-port flood control flags of unknown BUM traffic by injecting bc/uc/mc on one bridge port and verifying it being forwarded to both the bridge itself and another regular bridge port.

---
 .../testing/selftests/net/forwarding/Makefile |   3 +-
 .../selftests/net/forwarding/bridge_flood.sh  | 170 ++++++++++++++++++
 2 files changed, 172 insertions(+), 1 deletion(-)
 create mode 100755 tools/testing/selftests/net/forwarding/bridge_flood.sh
tools/testing/selftests/net/forwarding/Makefile +2 -1
index ae80c2aef577..873fa61d1ee1 100644
@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0+ OR MIT
 
-TEST_PROGS = bridge_igmp.sh \
+TEST_PROGS = bridge_flood.sh \
+	bridge_igmp.sh \
 	bridge_locked_port.sh \
 	bridge_mdb.sh \
 	bridge_port_isolation.sh \
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +0 -0
new file mode 100755
index 000000000000..1966c960d705

tools/testing/selftests/net/forwarding/bridge_flood.sh line 8

Quoted 8 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +8 -0
@@ -0,0 +1,170 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+#
+# Verify per-port flood control flags of unknown BUM traffic.
+#
+#                     br0
+#                    /   \
+#                  h1     h2

I think the picture is slightly inaccurate. From it I understand that h1 and h2 are bridge ports, but they are stations attached to the real bridge ports, swp1 and swp2. Maybe it would be good to draw all interfaces.

tools/testing/selftests/net/forwarding/bridge_flood.sh line 14

Quoted 6 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +6 -0
+#
+# We inject bc/uc/mc on h1, toggle the three flood flags for
+# both br0 and h2, then verify that traffic is flooded as per
+# the flags, and nowhere else.
+#
+#set -x

stray debug line

tools/testing/selftests/net/forwarding/bridge_flood.sh line 24

Quoted 10 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +10 -0
+
+ALL_TESTS="br_flood_unknown_bc_test br_flood_unknown_uc_test br_flood_unknown_mc_test"
+NUM_NETIFS=4
+
+SRC_MAC="00:de:ad:be:ef:00"
+GRP_IP4="225.1.2.3"
+GRP_MAC="01:00:01:c0:ff:ee"
+GRP_IP6="ff02::42"
+
+BC_PKT="ff:ff:ff:ff:ff:ff $SRC_MAC 00:04 48:45:4c:4f"

HELO to you too

tools/testing/selftests/net/forwarding/bridge_flood.sh line 29

Quoted 5 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +5 -0
+UC_PKT="02:00:01:c0:ff:ee $SRC_MAC 00:04 48:45:4c:4f"
+MC_PKT="01:00:5e:01:02:03 $SRC_MAC 08:00 45:00 00:20 c2:10 00:00 ff 11 12:b2 01:02:03:04 e1:01:02:03 04:d2 10:e1 00:0c 6e:84 48:45:4c:4f"
+
+# Disable promisc to ensure we only receive flooded frames
+export TCPDUMP_EXTRA_FLAGS="-pl"

Exporting should be required only for sub-shells, doesn't apply when you source a script.

tools/testing/selftests/net/forwarding/bridge_flood.sh line 41

Quoted 12 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +12 -0
+
+source lib.sh
+
+h1=${NETIFS[p1]}
+h2=${NETIFS[p3]}
+swp1=${NETIFS[p2]}
+swp2=${NETIFS[p4]}
+
+#
+# Port mappings and flood flag pattern to set/detect
+#
+declare -A ports=([br0]=br0 [$swp1]=$h1 [$swp2]=$h2)

Maybe you could populate the "ports" and the "flagN" arrays in the same order, i.e. bridge first for all?

Also, to be honest, a generic name like "ports" is hard to digest, especially since you have another generic variable name "iface". Maybe "brports" and "station" is a little bit more specific?

tools/testing/selftests/net/forwarding/bridge_flood.sh line 45

Quoted 4 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +4 -0
+declare -A flag1=([$swp1]=off [$swp2]=off [br0]=off)
+declare -A flag2=([$swp1]=off [$swp2]=on  [br0]=off)
+declare -A flag3=([$swp1]=off [$swp2]=on  [br0]=on )
+declare -A flag4=([$swp1]=off [$swp2]=off [br0]=on )

If it's not too much, maybe these could be called "flags_pass1", etc. Again, it was a bit hard to digest on first read.

Quoted 45 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +45 -0
+
+switch_create()
+{
+	ip link add dev br0 type bridge
+
+	for port in ${!ports[@]}; do
+		[ "$port" != "br0" ] && ip link set dev $port master br0
+		ip link set dev $port up
+	done
+}
+
+switch_destroy()
+{
+	for port in ${!ports[@]}; do
+		ip link set dev $port down
+	done
+	ip link del dev br0
+}
+
+setup_prepare()
+{
+	vrf_prepare
+
+	let i=1
+	for iface in ${ports[@]}; do
+		[ "$iface" = "br0" ] && continue
+		simple_if_init $iface 192.0.2.$i/24 2001:db8:1::$i/64
+		let i=$((i + 1))
+	done
+
+	switch_create
+}
+
+cleanup()
+{
+	pre_cleanup
+	switch_destroy
+
+	let i=1
+	for iface in ${ports[@]}; do
+		[ "$iface" = "br0" ] && continue
+		simple_if_fini $iface 192.0.2.$i/24 2001:db8:1::$i/64
+		let i=$((i + 1))
+	done
+

tools/testing/selftests/net/forwarding/bridge_flood.sh line 100

Quoted 10 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +10 -0
+	vrf_cleanup
+}
+
+do_flood_unknown()
+{
+	local type=$1
+	local pass=$2
+	local flag=$3
+	local pkt=$4
+	local -n flags=$5

I find it slightly less confusing if "flag" and "flags" are next to each other in the parameter list, since they're related.

Quoted 18 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +18 -0
+
+	RET=0
+	for port in ${!ports[@]}; do
+		if [ "$port" = "br0" ]; then
+			self="self"
+		else
+			self=""
+		fi
+		bridge link set dev $port $flag ${flags[$port]} $self
+		check_err $? "Failed setting $port $flag ${flags[$port]}"
+	done
+
+	for iface in ${ports[@]}; do
+		tcpdump_start $iface
+	done
+
+	$MZ -q $h1 "$pkt"
+	sleep 1

tools/testing/selftests/net/forwarding/bridge_flood.sh line 128

Quoted 10 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +10 -0
+
+	for iface in ${ports[@]}; do
+		tcpdump_stop $iface
+	done
+
+	for port in ${!ports[@]}; do
+		iface=${ports[$port]}
+
+#		echo "Dumping PCAP from $iface, expecting ${flags[$port]}:"
+#		tcpdump_show $iface

Do something about the commented lines.

tools/testing/selftests/net/forwarding/bridge_flood.sh line 129

Quoted 1 line
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +1 -0
+		tcpdump_show $iface |grep -q "$SRC_MAC"

Space between pipe and grep.

tools/testing/selftests/net/forwarding/bridge_flood.sh line 132

Quoted 3 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +3 -0
+		rc=$?
+
+		check_err_fail "${flags[$port]} = on"  $? "failed flooding from $h1 to port $port"

I think the "failed" word here is superfluous, since check_err_fail already says "$what succeeded, but should have failed".

Quoted 41 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +38 -0
+		check_err_fail "${flags[$port]} = off" $? "flooding from $h1 to port $port"
+	done
+
+	log_test "flood unknown $type pass $pass/4"
+}
+
+br_flood_unknown_bc_test()
+{
+	do_flood_unknown BC 1 bcast_flood "$BC_PKT" flag1
+	do_flood_unknown BC 2 bcast_flood "$BC_PKT" flag2
+	do_flood_unknown BC 3 bcast_flood "$BC_PKT" flag3
+	do_flood_unknown BC 4 bcast_flood "$BC_PKT" flag4
+}
+
+br_flood_unknown_uc_test()
+{
+	do_flood_unknown UC 1 flood "$UC_PKT" flag1
+	do_flood_unknown UC 2 flood "$UC_PKT" flag2
+	do_flood_unknown UC 3 flood "$UC_PKT" flag3
+	do_flood_unknown UC 4 flood "$UC_PKT" flag4
+}
+
+br_flood_unknown_mc_test()
+{
+	do_flood_unknown MC 1 mcast_flood "$MC_PKT" flag1
+	do_flood_unknown MC 2 mcast_flood "$MC_PKT" flag2
+	do_flood_unknown MC 3 mcast_flood "$MC_PKT" flag3
+	do_flood_unknown MC 4 mcast_flood "$MC_PKT" flag4
+}
+
+trap cleanup EXIT
+
+setup_prepare
+setup_wait
+
+tests_run
+
+exit $EXIT_STATUS
Signature · 1 line
2.25.1
#21 lore raw

Re: [PATCH RFC net-next 07/13] selftests: forwarding: new test, verify bridge flood flags

Quoted 10 lines · Vladimir Oltean

tools/testing/selftests/net/forwarding/bridge_flood.sh line 8

Quoted 5 lines · Joachim Wiberg
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +5 -0
+# Verify per-port flood control flags of unknown BUM traffic.
+#
+#                     br0
+#                    /   \
+#                  h1     h2

I think the picture is slightly inaccurate. From it I understand that h1 and h2 are bridge ports, but they are stations attached to the real bridge ports, swp1 and swp2. Maybe it would be good to draw all interfaces.

Hmm, yeah either that or drop it entirely. I sort of assumed everyone knew about the h<-[veth]->swp (or actual cable) setup, but you're right this is a bit unclear. Me and Tobias have internally used h<-->p (for host<-->bridge-port) and other similar nomenclature. Finding a good name that fits easily, and is still readable, in ASCII drawings is hard. I'll give it a go in the next drop, thanks!

Quoted 2 lines
Quoted 1 line

+#set -x

stray debug line

thx

Quoted 4 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 29

Quoted 2 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +2 -0
+# Disable promisc to ensure we only receive flooded frames
+export TCPDUMP_EXTRA_FLAGS="-pl"

Exporting should be required only for sub-shells, doesn't apply when you source a script.

Ah thanks, will fix!

Quoted 4 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 41

Quoted 2 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +2 -0
+# Port mappings and flood flag pattern to set/detect
+declare -A ports=([br0]=br0 [$swp1]=$h1 [$swp2]=$h2)

Maybe you could populate the "ports" and the "flagN" arrays in the same order, i.e. bridge first for all?

Good point, thanks!

Quoted 3 lines

Also, to be honest, a generic name like "ports" is hard to digest, especially since you have another generic variable name "iface". Maybe "brports" and "station" is a little bit more specific?

Is there a common naming standard between bridge tests, or is it more important to be consistent the test overview (test heading w/ picture)?

Anyway, I'll have a look at the naming for the next drop.

Quoted 6 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 45

Quoted 4 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +4 -0
+declare -A flag1=([$swp1]=off [$swp2]=off [br0]=off)
+declare -A flag2=([$swp1]=off [$swp2]=on  [br0]=off)
+declare -A flag3=([$swp1]=off [$swp2]=on  [br0]=on )
+declare -A flag4=([$swp1]=off [$swp2]=off [br0]=on )

If it's not too much, maybe these could be called "flags_pass1", etc. Again, it was a bit hard to digest on first read.

More like flags_pass_fail, but since its the flooding flags, maybe flood_patternN would be better?

Quoted 9 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 100

Quoted 7 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +7 -0
+do_flood_unknown()
+{
+	local type=$1
+	local pass=$2
+	local flag=$3
+	local pkt=$4
+	local -n flags=$5

I find it slightly less confusing if "flag" and "flags" are next to each other in the parameter list, since they're related.

Hmm, OK.

Quoted 3 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 128

Quoted 2 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +2 -0
+#		echo "Dumping PCAP from $iface, expecting ${flags[$port]}:"
+#		tcpdump_show $iface

Do something about the commented lines.

Oups, thanks!

Quoted 2 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 129

Quoted 1 line
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +1 -0
+		tcpdump_show $iface |grep -q "$SRC_MAC"

Space between pipe and grep.

Will fix!

Quoted 3 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 132

Quoted 1 line
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +1 -0
+		check_err_fail "${flags[$port]} = on"  $? "failed flooding from $h1 to port $port"

I think the "failed" word here is superfluous, since check_err_fail already says "$what succeeded, but should have failed".

Ah, good point!

Thank you for the review! <3

/J

maintainer
#22 lore raw

Re: [PATCH RFC net-next 07/13] selftests: forwarding: new test, verify bridge flood flags

Quoted 12 lines · Joachim Wiberg
Quoted 10 lines · Vladimir Oltean

tools/testing/selftests/net/forwarding/bridge_flood.sh line 8

Quoted 5 lines · Joachim Wiberg
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +5 -0
+# Verify per-port flood control flags of unknown BUM traffic.
+#
+#                     br0
+#                    /   \
+#                  h1     h2

I think the picture is slightly inaccurate. From it I understand that h1 and h2 are bridge ports, but they are stations attached to the real bridge ports, swp1 and swp2. Maybe it would be good to draw all interfaces.

Quoted 6 lines

Hmm, yeah either that or drop it entirely. I sort of assumed everyone knew about the h<-[veth]->swp (or actual cable) setup, but you're right this is a bit unclear. Me and Tobias have internally used h<-->p (for host<-->bridge-port) and other similar nomenclature. Finding a good name that fits easily, and is still readable, in ASCII drawings is hard. I'll give it a go in the next drop, thanks!

I wasn't thinking of anything too fancy, this would do I guess.

             br0
            /   \

h1 --- swp1 swp2 --- h2

Quoted 8 lines
Quoted 3 lines

Also, to be honest, a generic name like "ports" is hard to digest, especially since you have another generic variable name "iface". Maybe "brports" and "station" is a little bit more specific?

Is there a common naming standard between bridge tests, or is it more important to be consistent the test overview (test heading w/ picture)?

Anyway, I'll have a look at the naming for the next drop.

Even if there is a common naming standard in the selftests I wouldn't know it. I just found the naming here to be vague enough that it is confusing. If there are other examples of "port" + "iface" please feel free to disregard.

Quoted 9 lines
Quoted 6 lines

tools/testing/selftests/net/forwarding/bridge_flood.sh line 45

Quoted 4 lines
tools/testing/selftests/net/forwarding/bridge_flood.sh new file +4 -0
+declare -A flag1=([$swp1]=off [$swp2]=off [br0]=off)
+declare -A flag2=([$swp1]=off [$swp2]=on  [br0]=off)
+declare -A flag3=([$swp1]=off [$swp2]=on  [br0]=on )
+declare -A flag4=([$swp1]=off [$swp2]=off [br0]=on )

If it's not too much, maybe these could be called "flags_pass1", etc. Again, it was a bit hard to digest on first read.

More like flags_pass_fail, but since its the flooding flags, maybe flood_patternN would be better?

This works.

#23 lore raw

RFC PATCH 8/13 net: bridge: avoid classifying unknown multicast as mrouters_only

In reply to Joachim Wiberg

Unknown multicast, MAC/IPv4/IPv6, should always be flooded according to the per-port mcast_flood setting, as well as to detected and configured mcast_router ports.

This patch drops the mrouters_only classifier of unknown IP multicast and moves the flow handling from br_multicast_flood() to br_flood(). This in turn means br_flood() must know about multicast router ports.

---
 net/bridge/br_forward.c   | 11 +++++++++++
 net/bridge/br_multicast.c |  6 +-----
 2 files changed, 12 insertions(+), 5 deletions(-)
net/bridge/br_forward.c +11 -0
index 02bb620d3b8d..ab5b97a8c12e 100644
@@ -199,9 +199,15 @@ static struct net_bridge_port *maybe_deliver(
 void br_flood(struct net_bridge *br, struct sk_buff *skb,
 	      enum br_pkt_type pkt_type, bool local_rcv, bool local_orig)
 {
+	struct net_bridge_mcast *brmctx = &br->multicast_ctx;
+	struct net_bridge_port *rport = NULL;
 	struct net_bridge_port *prev = NULL;
+	struct hlist_node *rp = NULL;
 	struct net_bridge_port *p;
 
+	if (pkt_type == BR_PKT_MULTICAST)
+		rp = br_multicast_get_first_rport_node(brmctx, skb);
+
 	list_for_each_entry_rcu(p, &br->port_list, list) {
 		/* Do not flood unicast traffic to ports that turn it off, nor
 		 * other traffic if flood off, except for traffic we originate
@@ -212,6 +218,11 @@ void br_flood(struct net_bridge *br, struct sk_buff *skb,
 				continue;
 			break;
 		case BR_PKT_MULTICAST:
+			rport = br_multicast_rport_from_node_skb(rp, skb);
+			if (rport == p) {
+				rp = rcu_dereference(hlist_next_rcu(rp));
+				break;
+			}
 			if (!(p->flags & BR_MCAST_FLOOD) && skb->dev != br->dev)
 				continue;
 			break;
net/bridge/br_multicast.c +1 -5
index db4f2641d1cd..c57e3bbb00ad 100644
@@ -3643,9 +3643,7 @@ static int br_multicast_ipv4_rcv(struct net_bridge_mcast *brmctx,
 	err = ip_mc_check_igmp(skb);
 
 	if (err == -ENOMSG) {
-		if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr)) {
-			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
-		} else if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
+		if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
 			if (ip_hdr(skb)->protocol == IPPROTO_PIM)
 				br_multicast_pim(brmctx, pmctx, skb);
 		} else if (ipv4_is_all_snoopers(ip_hdr(skb)->daddr)) {
@@ -3712,8 +3710,6 @@ static int br_multicast_ipv6_rcv(struct net_bridge_mcast *brmctx,
 	err = ipv6_mc_check_mld(skb);
 
 	if (err == -ENOMSG || err == -ENODATA) {
-		if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
-			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
 		if (err == -ENODATA &&
 		    ipv6_addr_is_all_snoopers(&ipv6_hdr(skb)->daddr))
 			br_ip6_multicast_mrd_rcv(brmctx, pmctx, skb);
Signature · 1 line
2.25.1
#24 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Quoted 8 lines · Joachim Wiberg

Unknown multicast, MAC/IPv4/IPv6, should always be flooded according to the per-port mcast_flood setting, as well as to detected and configured mcast_router ports.

This patch drops the mrouters_only classifier of unknown IP multicast and moves the flow handling from br_multicast_flood() to br_flood(). This in turn means br_flood() must know about multicast router ports.

Quoted 6 lines
---
 net/bridge/br_forward.c   | 11 +++++++++++
 net/bridge/br_multicast.c |  6 +-----
 2 files changed, 12 insertions(+), 5 deletions(-)

If you'd like to flood unknown mcast traffic when a router is present please add a new option which defaults to the current state (disabled).

net/bridge/br_forward.c static struct net_bridge_port *maybe_deliver( · line 202

Quoted 9 lines
net/bridge/br_forward.c +1 -0
index 02bb620d3b8d..ab5b97a8c12e 100644
@@ -199,9 +199,15 @@ static struct net_bridge_port *maybe_deliver(
 void br_flood(struct net_bridge *br, struct sk_buff *skb,
 	      enum br_pkt_type pkt_type, bool local_rcv, bool local_orig)
 {
+	struct net_bridge_mcast *brmctx = &br->multicast_ctx;

Note this breaks per-vlan mcast. You have to use the inferred mctx.

Quoted 47 lines
net/bridge/br_forward.c +10 -0
+	struct net_bridge_port *rport = NULL;
 	struct net_bridge_port *prev = NULL;
+	struct hlist_node *rp = NULL;
 	struct net_bridge_port *p;
 
+	if (pkt_type == BR_PKT_MULTICAST)
+		rp = br_multicast_get_first_rport_node(brmctx, skb);
+
 	list_for_each_entry_rcu(p, &br->port_list, list) {
 		/* Do not flood unicast traffic to ports that turn it off, nor
 		 * other traffic if flood off, except for traffic we originate
@@ -212,6 +218,11 @@ void br_flood(struct net_bridge *br, struct sk_buff *skb,
 				continue;
 			break;
 		case BR_PKT_MULTICAST:
+			rport = br_multicast_rport_from_node_skb(rp, skb);
+			if (rport == p) {
+				rp = rcu_dereference(hlist_next_rcu(rp));
+				break;
+			}
 			if (!(p->flags & BR_MCAST_FLOOD) && skb->dev != br->dev)
 				continue;
 			break;
net/bridge/br_multicast.c +1 -5
index db4f2641d1cd..c57e3bbb00ad 100644
@@ -3643,9 +3643,7 @@ static int br_multicast_ipv4_rcv(struct net_bridge_mcast *brmctx,
 	err = ip_mc_check_igmp(skb);
 
 	if (err == -ENOMSG) {
-		if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr)) {
-			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
-		} else if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
+		if (pim_ipv4_all_pim_routers(ip_hdr(skb)->daddr)) {
 			if (ip_hdr(skb)->protocol == IPPROTO_PIM)
 				br_multicast_pim(brmctx, pmctx, skb);
 		} else if (ipv4_is_all_snoopers(ip_hdr(skb)->daddr)) {
@@ -3712,8 +3710,6 @@ static int br_multicast_ipv6_rcv(struct net_bridge_mcast *brmctx,
 	err = ipv6_mc_check_mld(skb);
 
 	if (err == -ENOMSG || err == -ENODATA) {
-		if (!ipv6_addr_is_ll_all_nodes(&ipv6_hdr(skb)->daddr))
-			BR_INPUT_SKB_CB(skb)->mrouters_only = 1;
 		if (err == -ENODATA &&
 		    ipv6_addr_is_all_snoopers(&ipv6_hdr(skb)->daddr))
 			br_ip6_multicast_mrd_rcv(brmctx, pmctx, skb);
#25 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Hi Nik,

and thank you for taking the time to respond!

Quoted 4 lines · Nikolay Aleksandrov
Quoted 3 lines · Joachim Wiberg

Unknown multicast, MAC/IPv4/IPv6, should always be flooded according to the per-port mcast_flood setting, as well as to detected and configured mcast_router ports.

I realize I should've included a reference to RFC4541 here. Will add that in the non-RFC patch.

Quoted 5 lines
Quoted 3 lines

This patch drops the mrouters_only classifier of unknown IP multicast and moves the flow handling from br_multicast_flood() to br_flood(). This in turn means br_flood() must know about multicast router ports.

If you'd like to flood unknown mcast traffic when a router is present please add a new option which defaults to the current state (disabled).

I don't think we have to add another option, because according to the snooping RFC[1], section 2.1.2 Data Forwarding Rules:

"3) [..] If a switch receives an unregistered packet, it must forward
that packet on all ports to which an IGMP[2] router is attached. A switch may default to forwarding unregistered packets on all ports. Switches that do not forward unregistered packets to all ports must include a configuration option to force the flooding of unregistered packets on specified ports. [..]"

From this I'd like to argue that our current behavior in the bridge is wrong. To me it's clear that, since we have a confiugration option, we should forward unknown IP multicast to all MCAST_FLOOD ports (as well as the router ports).

Also, and more critically, the current behavior of offloaded switches do forwarding like this already. So there is a discrepancy currently between how the bridge forwards unknown multicast and how any underlying switchcore does it.

Sure, we'll break bridge behavior slightly by forwarding to more ports than previous (until the group becomes known/registered), but we'd be standards compliant, and the behavior can still be controlled per-port.

[1]: https://www.rfc-editor.org/rfc/rfc4541.html#section-2.1.2
[2]: Section 3 goes on to explain how this is similar also for MLD

Quoted 10 lines

net/bridge/br_forward.c static struct net_bridge_port *maybe_deliver( · line 202

Quoted 9 lines
net/bridge/br_forward.c +1 -0
index 02bb620d3b8d..ab5b97a8c12e 100644
@@ -199,9 +199,15 @@ static struct net_bridge_port *maybe_deliver(
 void br_flood(struct net_bridge *br, struct sk_buff *skb,
 	      enum br_pkt_type pkt_type, bool local_rcv, bool local_orig)
 {
+	struct net_bridge_mcast *brmctx = &br->multicast_ctx;

Note this breaks per-vlan mcast. You have to use the inferred mctx.

Thank you, this was one of the things I was really unsure about since the introduction of per-VLAN support. I'll extend the prototype and include the brmctx from br_handle_frame_finish(). Thanks!

Best regards
/Joachim

#26 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Quoted 30 lines · Joachim Wiberg

Hi Nik,

and thank you for taking the time to respond!

Quoted 4 lines · Nikolay Aleksandrov
Quoted 3 lines · Joachim Wiberg

Unknown multicast, MAC/IPv4/IPv6, should always be flooded according to the per-port mcast_flood setting, as well as to detected and configured mcast_router ports.

I realize I should've included a reference to RFC4541 here. Will add that in the non-RFC patch.

Quoted 5 lines
Quoted 3 lines

This patch drops the mrouters_only classifier of unknown IP multicast and moves the flow handling from br_multicast_flood() to br_flood(). This in turn means br_flood() must know about multicast router ports.

If you'd like to flood unknown mcast traffic when a router is present please add a new option which defaults to the current state (disabled).

I don't think we have to add another option, because according to the snooping RFC[1], section 2.1.2 Data Forwarding Rules:

"3) [..] If a switch receives an unregistered packet, it must forward
that packet on all ports to which an IGMP[2] router is attached. A switch may default to forwarding unregistered packets on all ports. Switches that do not forward unregistered packets to all ports must include a configuration option to force the flooding of unregistered packets on specified ports. [..]"

Quoted 4 lines

From this I'd like to argue that our current behavior in the bridge is wrong. To me it's clear that, since we have a confiugration option, we should forward unknown IP multicast to all MCAST_FLOOD ports (as well as the router ports).

Definitely not wrong. In fact:
"Switches that do not forward unregistered packets to all ports must
include a configuration option to force the flooding of unregistered packets on specified ports. [..]"

is already implemented because the admin can mark any port as a router and enable flooding to it.

Quoted 6 lines

Also, and more critically, the current behavior of offloaded switches do forwarding like this already. So there is a discrepancy currently between how the bridge forwards unknown multicast and how any underlying switchcore does it.

Quoted 7 lines

Sure, we'll break bridge behavior slightly by forwarding to more ports than previous (until the group becomes known/registered), but we'd be standards compliant, and the behavior can still be controlled per-port.

[1]: https://www.rfc-editor.org/rfc/rfc4541.html#section-2.1.2
[2]: Section 3 goes on to explain how this is similar also for MLD

RFC4541 is only recommending, it's not a mandatory behaviour. This default has been placed for a very long time and a lot of users and tests take it into consideration. We cannot break such assumptions and start suddenly flooding packets, but we can leave it up to the admin or distribution/network software to configure it as default.

Quoted 17 lines
Quoted 10 lines

net/bridge/br_forward.c static struct net_bridge_port *maybe_deliver( · line 202

Quoted 9 lines
net/bridge/br_forward.c +1 -0
index 02bb620d3b8d..ab5b97a8c12e 100644
@@ -199,9 +199,15 @@ static struct net_bridge_port *maybe_deliver(
 void br_flood(struct net_bridge *br, struct sk_buff *skb,
 	      enum br_pkt_type pkt_type, bool local_rcv, bool local_orig)
 {
+	struct net_bridge_mcast *brmctx = &br->multicast_ctx;

Note this breaks per-vlan mcast. You have to use the inferred mctx.

Thank you, this was one of the things I was really unsure about since the introduction of per-VLAN support. I'll extend the prototype and include the brmctx from br_handle_frame_finish(). Thanks!

Best regards
/Joachim

#27 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Quoted 6 lines · Nikolay Aleksandrov
Quoted 5 lines · Joachim Wiberg

[snip]
From this I'd like to argue that our current behavior in the bridge is wrong. To me it's clear that, since we have a confiugration option, we should forward unknown IP multicast to all MCAST_FLOOD ports (as well as the router ports).

Quoted 6 lines

Definitely not wrong. In fact:
"Switches that do not forward unregistered packets to all ports must
include a configuration option to force the flooding of unregistered packets on specified ports. [..]"
is already implemented because the admin can mark any port as a router and enable flooding to it.

Hmm, I understand your point (here and below), and won't drive this point further. Instead I'll pick up on what you said in your first reply ... (below, last)

Btw, thank you for taking the time to reply and explain your standpoint, really helps my understanding of how we can develop the bridge further, without breaking userspace! :)

Quoted 4 lines
Quoted 1 line

[1]: https://www.rfc-editor.org/rfc/rfc4541.html#section-2.1.2

RFC4541 is only recommending, it's not a mandatory behaviour. This default has been placed for a very long time and a lot of users and tests take it into consideration.

Noted.

Quoted 3 lines

We cannot break such assumptions and start suddenly flooding packets, but we can leave it up to the admin or distribution/network software to configure it as default.

So, if I add a bridge flag, default off as you mentioned out earlier, which changes the default behavior of MCAST_FLOOD, then you'd be OK with that? Something cheeky like this perhaps:

    if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr))
       	BR_INPUT_SKB_CB(skb)->mrouters_only = !br_opt_get(br, BROPT_MCAST_FLOOD_RFC4541);

Best regards
/Joachim

#28 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Quoted 33 lines · Joachim Wiberg
Quoted 6 lines · Nikolay Aleksandrov
Quoted 5 lines · Joachim Wiberg

[snip]
From this I'd like to argue that our current behavior in the bridge is wrong. To me it's clear that, since we have a confiugration option, we should forward unknown IP multicast to all MCAST_FLOOD ports (as well as the router ports).

Quoted 6 lines

Definitely not wrong. In fact:
"Switches that do not forward unregistered packets to all ports must
include a configuration option to force the flooding of unregistered packets on specified ports. [..]"
is already implemented because the admin can mark any port as a router and enable flooding to it.

Hmm, I understand your point (here and below), and won't drive this point further. Instead I'll pick up on what you said in your first reply ... (below, last)

Btw, thank you for taking the time to reply and explain your standpoint, really helps my understanding of how we can develop the bridge further, without breaking userspace! :)

Quoted 4 lines
Quoted 1 line

[1]: https://www.rfc-editor.org/rfc/rfc4541.html#section-2.1.2

RFC4541 is only recommending, it's not a mandatory behaviour. This default has been placed for a very long time and a lot of users and tests take it into consideration.

Noted.

Quoted 3 lines

We cannot break such assumptions and start suddenly flooding packets, but we can leave it up to the admin or distribution/network software to configure it as default.

Quoted 6 lines

So, if I add a bridge flag, default off as you mentioned out earlier, which changes the default behavior of MCAST_FLOOD, then you'd be OK with that? Something cheeky like this perhaps:

    if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr))
       	BR_INPUT_SKB_CB(skb)->mrouters_only = !br_opt_get(br, BROPT_MCAST_FLOOD_RFC4541);

Exactly! And that is exactly what I had in mind when I wrote it. :)

Thanks,
Nik

Quoted 4 lines

Best regards
/Joachim

#29 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Quoted 34 lines · Nikolay Aleksandrov
Quoted 33 lines · Joachim Wiberg
Quoted 6 lines · Nikolay Aleksandrov
Quoted 5 lines · Joachim Wiberg

[snip]
From this I'd like to argue that our current behavior in the bridge is wrong. To me it's clear that, since we have a confiugration option, we should forward unknown IP multicast to all MCAST_FLOOD ports (as well as the router ports).

Quoted 6 lines

Definitely not wrong. In fact:
"Switches that do not forward unregistered packets to all ports must
include a configuration option to force the flooding of unregistered packets on specified ports. [..]"
is already implemented because the admin can mark any port as a router and enable flooding to it.

Hmm, I understand your point (here and below), and won't drive this point further. Instead I'll pick up on what you said in your first reply ... (below, last)

Btw, thank you for taking the time to reply and explain your standpoint, really helps my understanding of how we can develop the bridge further, without breaking userspace! :)

Quoted 4 lines
Quoted 1 line

[1]: https://www.rfc-editor.org/rfc/rfc4541.html#section-2.1.2

RFC4541 is only recommending, it's not a mandatory behaviour. This default has been placed for a very long time and a lot of users and tests take it into consideration.

Noted.

Quoted 3 lines

We cannot break such assumptions and start suddenly flooding packets, but we can leave it up to the admin or distribution/network software to configure it as default.

Quoted 9 lines
Quoted 6 lines

So, if I add a bridge flag, default off as you mentioned out earlier, which changes the default behavior of MCAST_FLOOD, then you'd be OK with that? Something cheeky like this perhaps:

    if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr))
       	BR_INPUT_SKB_CB(skb)->mrouters_only = !br_opt_get(br, BROPT_MCAST_FLOOD_RFC4541);

Exactly! And that is exactly what I had in mind when I wrote it. :)

Just please use a different option name that better suggests what it does.

Quoted 8 lines

Thanks,
Nik

Quoted 4 lines

Best regards
/Joachim

#30 lore raw

Re: [PATCH RFC net-next 08/13] net: bridge: avoid classifying unknown multicast as mrouters_only

Quoted 8 lines · Nikolay Aleksandrov
Quoted 7 lines · Nikolay Aleksandrov
Quoted 5 lines · Joachim Wiberg

So, if I add a bridge flag, default off as you mentioned out earlier, which changes the default behavior of MCAST_FLOOD, then you'd be OK with that? Something cheeky like this perhaps:

    if (!ipv4_is_local_multicast(ip_hdr(skb)->daddr))
       	BR_INPUT_SKB_CB(skb)->mrouters_only = !br_opt_get(br, BROPT_MCAST_FLOOD_RFC4541);

Exactly! And that is exactly what I had in mind when I wrote it. :)

Awesome, thank you! :)

Quoted 1 line

Just please use a different option name that better suggests what it does.

Heh, yeah spent a good while with my colleague (Tobias) thinking about how to name this one. I'll see what I can come up with, but whatever shows up in the next patch iteration will be very open for discussion.

Cheers
/Joachim

#31 lore raw

RFC PATCH 9/13 selftests: forwarding: rename test groups for next bridge mdb tests

In reply to Joachim Wiberg

Rename test groups to PASS and FAIL, respectively, for upcoming changes to test suite.

---
 .../selftests/net/forwarding/bridge_mdb.sh     | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)
tools/testing/selftests/net/forwarding/bridge_mdb.sh +12 -6
index b1ba6876dd86..c0b84b7d4364 100755
@@ -7,9 +7,15 @@
 ALL_TESTS="mdb_add_del_test"
 NUM_NETIFS=2
 
-TEST_GROUP_IP4="225.1.2.3"
-TEST_GROUP_IP6="ff02::42"
-TEST_GROUP_MAC="01:00:01:c0:ff:ee"
+PASS_GRP_IP4="225.1.2.3"
+FAIL_GRP_IP4="225.1.2.4"
+
+PASS_GRP_MAC="01:00:01:c0:ff:ee"
+FAIL_GRP_MAC="01:00:01:c0:ff:ef"
+
+PASS_GRP_IP6="ff02::42"
+FAIL_GRP_IP6="ff02::43"
+
 
 source lib.sh
 
@@ -88,9 +94,9 @@ do_mdb_add_del()
 
 mdb_add_del_test()
 {
-	do_mdb_add_del $TEST_GROUP_MAC permanent
-	do_mdb_add_del $TEST_GROUP_IP4
-	do_mdb_add_del $TEST_GROUP_IP6
+	do_mdb_add_del $PASS_GRP_MAC permanent
+	do_mdb_add_del $PASS_GRP_IP4
+	do_mdb_add_del $PASS_GRP_IP6
 }
 
 trap cleanup EXIT
Signature · 1 line
2.25.1
maintainer
#32 lore raw

Re: [PATCH RFC net-next 09/13] selftests: forwarding: rename test groups for next bridge mdb tests

Quoted 13 lines · Joachim Wiberg

Rename test groups to PASS and FAIL, respectively, for upcoming changes to test suite.

---
 .../selftests/net/forwarding/bridge_mdb.sh     | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)
tools/testing/selftests/net/forwarding/bridge_mdb.sh +0 -0
index b1ba6876dd86..c0b84b7d4364 100755

tools/testing/selftests/net/forwarding/bridge_mdb.sh line 11

Quoted 8 lines
tools/testing/selftests/net/forwarding/bridge_mdb.sh +2 -3
@@ -7,9 +7,15 @@
 ALL_TESTS="mdb_add_del_test"
 NUM_NETIFS=2
 
-TEST_GROUP_IP4="225.1.2.3"
-TEST_GROUP_IP6="ff02::42"
-TEST_GROUP_MAC="01:00:01:c0:ff:ee"
+PASS_GRP_IP4="225.1.2.3"
+FAIL_GRP_IP4="225.1.2.4"

This is more than just the advertised rename, the fail groups are new and not used in this change.

Quoted 26 lines
tools/testing/selftests/net/forwarding/bridge_mdb.sh +10 -3
+
+PASS_GRP_MAC="01:00:01:c0:ff:ee"
+FAIL_GRP_MAC="01:00:01:c0:ff:ef"
+
+PASS_GRP_IP6="ff02::42"
+FAIL_GRP_IP6="ff02::43"
+
 
 source lib.sh
 
@@ -88,9 +94,9 @@ do_mdb_add_del()
 
 mdb_add_del_test()
 {
-	do_mdb_add_del $TEST_GROUP_MAC permanent
-	do_mdb_add_del $TEST_GROUP_IP4
-	do_mdb_add_del $TEST_GROUP_IP6
+	do_mdb_add_del $PASS_GRP_MAC permanent
+	do_mdb_add_del $PASS_GRP_IP4
+	do_mdb_add_del $PASS_GRP_IP6
 }
 
 trap cleanup EXIT
Signature · 1 line
2.25.1
#33 lore raw

Re: [PATCH RFC net-next 09/13] selftests: forwarding: rename test groups for next bridge mdb tests

Quoted 13 lines · Vladimir Oltean

tools/testing/selftests/net/forwarding/bridge_mdb.sh line 17

Quoted 10 lines · Joachim Wiberg
tools/testing/selftests/net/forwarding/bridge_mdb.sh +7 -3
-TEST_GROUP_IP4="225.1.2.3"
-TEST_GROUP_IP6="ff02::42"
-TEST_GROUP_MAC="01:00:01:c0:ff:ee"
+PASS_GRP_IP4="225.1.2.3"
+FAIL_GRP_IP4="225.1.2.4"
+PASS_GRP_MAC="01:00:01:c0:ff:ee"
+FAIL_GRP_MAC="01:00:01:c0:ff:ef"
+
+PASS_GRP_IP6="ff02::42"
+FAIL_GRP_IP6="ff02::43"

This is more than just the advertised rename, the fail groups are new and not used in this change.

Yeah I rushed the set out to get feedback on the overall take, sorry about this. I'll see to fixing this for the non-RFC drop. Thanks!

#34 lore raw

RFC PATCH 10/13 selftests: forwarding: verify flooding of unknown multicast

In reply to Joachim Wiberg

This patch adds forwarding test to bridge_mdb.sh, using groups and group names that make little sense right now.

Ensure flooding of unknown multicast works as intended after the initial grace period when we don't know if we are the querier or if someone else on the LAN is. Hence the reduced timers.

---
 .../selftests/net/forwarding/bridge_mdb.sh    | 135 +++++++++++++++++-
 1 file changed, 131 insertions(+), 4 deletions(-)
tools/testing/selftests/net/forwarding/bridge_mdb.sh +131 -4
index c0b84b7d4364..6de98c59a620 100755
@@ -3,22 +3,50 @@
 #
 # Verify that adding host mdb entries work as intended for all types of
 # multicast filters: ipv4, ipv6, and mac
+#
+# Verify forwarding (default flooding behavior) to all ports of unknown
+# multicast: MAC, IPv4, IPv6.
+#
+# Note: this test completely disables IPv6 auto-configuration to avoid
+#       any type of dynamic behavior outside of MLD and IGMP protocols.
+#       Static IPv6 addresses are used to ensure consistent behavior,
+#       even in the startup phase when multicast snooping is enabled.
+#
 
-ALL_TESTS="mdb_add_del_test"
-NUM_NETIFS=2
+ALL_TESTS="mdb_add_del_test mdb_compat_fwd_test"
+NUM_NETIFS=4
 
+SRC_PORT="1234"
+DST_PORT="4321"
+
+SRC_ADDR_IP4="1.2.3.4"
 PASS_GRP_IP4="225.1.2.3"
 FAIL_GRP_IP4="225.1.2.4"
 
+SRC_ADDR_MAC="00:de:ad:be:ef:00"
 PASS_GRP_MAC="01:00:01:c0:ff:ee"
 FAIL_GRP_MAC="01:00:01:c0:ff:ef"
 
+PASS_PKT_MAC="$PASS_GRP_MAC $SRC_ADDR_MAC 00:04 48:45:4c:4f"
+FAIL_PKT_MAC="$FAIL_GRP_MAC $SRC_ADDR_MAC 00:04 46:41:49:4c"
+
+PASS_PKT_IP4="01:00:5e:01:02:03 $SRC_ADDR_MAC 08:00 45:00 00:20 c2:10 00:00 ff 11 12:b2 01:02:03:04 e1:01:02:03 04:d2 10:e1 00:0c 6e:84 48:45:4c:4f"
+FAIL_PKT_IP4="01:00:5e:01:02:04 $SRC_ADDR_MAC 08:00 45:00 00:20 dc:e4 00:00 ff 11 f7:dc 01:02:03:04 e1:01:02:04 04:d2 10:e1 00:0c 73:8a 46:41:49:4c"
+
+SRC_ADDR_IP6="ff2e::42"
 PASS_GRP_IP6="ff02::42"
 FAIL_GRP_IP6="ff02::43"
 
+PASS_PKT_IP6="33 33 00 00 00 42 36 1e b4 04 cd e8 86 dd 60 00 01 01 00 08 11 ff ff 2e 00 00 00 00 00 00 00 00 00 00 00 00 00 42 ff 02 00 00 00 00 00 00 00 00 00 00 00 00 00 42 04 d2 10 e1 00 08 eb 75"
+FAIL_PKT_IP6="33 33 00 00 00 43 36 1e b4 04 cd e8 86 dd 60 00 01 01 00 08 11 ff ff 2e 00 00 00 00 00 00 00 00 00 00 00 00 00 42 ff 02 00 00 00 00 00 00 00 00 00 00 00 00 00 43 04 d2 10 e1 00 08 eb 74"
+
+# Disable promisc to ensure we only receive $TEST_GROUP*
+export TCPDUMP_EXTRA_FLAGS="-pl"
 
 source lib.sh
 
+require_command tcpdump
+
 h1_create()
 {
 	simple_if_init $h1 192.0.2.1/24 2001:db8:1::1/64
@@ -29,20 +57,51 @@ h1_destroy()
 	simple_if_fini $h1 192.0.2.1/24 2001:db8:1::1/64
 }
 
+h2_create()
+{
+	simple_if_init $h2 192.0.2.2/24 2001:db8:1::2/64
+}
+
+h2_destroy()
+{
+	simple_if_fini $h2 192.0.2.2/24 2001:db8:1::2/64
+}
+
 switch_create()
 {
-	# Enable multicast filtering
-	ip link add dev br0 type bridge mcast_snooping 1
+	# Enable multicast filtering w/ querier, reduce query response
+	# and startup interval to speed up test a bit.
+	ip link add dev br0 type bridge mcast_snooping 1 \
+	   mcast_startup_query_interval 400 mcast_query_response_interval 200
+
+	# Set static IPv6 address before we enable the multicast querier
+	# function.  This, along with disabling IPv6 address auto config
+	# (previously), ensures correct forwarding according to the MDB
+	# even when per-port flooding is disabled, *after* the initial
+	# startup phase when the bridge floods all multicast (according
+	# to its per-port mcast_flood settings.
+	ip addr add 2001:db8:1::42/64 dev br0
+	ip link set br0 type bridge mcast_querier 1
 
 	ip link set dev $swp1 master br0
+	ip link set dev $swp2 master br0
 
 	ip link set dev br0 up
 	ip link set dev $swp1 up
+	ip link set dev $swp2 up
+
+	# Initial delay, when bridge floods all mcast, is set to 200
+	# above (2 sec.)  We wait 3 sec to handle the case when a single
+	# strict fwd test is run directly after the initial setup, e.g.,
+	# TESTS=mdb_ip6_fwd_test
+	sleep 3
 }
 
 switch_destroy()
 {
+	ip link set dev $swp2 down
 	ip link set dev $swp1 down
+
 	ip link del dev br0
 }
 
@@ -51,9 +110,17 @@ setup_prepare()
 	h1=${NETIFS[p1]}
 	swp1=${NETIFS[p2]}
 
+	swp2=${NETIFS[p3]}
+	h2=${NETIFS[p4]}
+
+	# Disable all IPv6 autoconfiguration during test, we want
+	# full control of when MLD queries start etc.
+	sysctl_set net.ipv6.conf.default.accept_ra 0
 	vrf_prepare
 
 	h1_create
+	h2_create
+
 	switch_create
 }
 
@@ -62,9 +129,12 @@ cleanup()
 	pre_cleanup
 
 	switch_destroy
+
+	h2_destroy
 	h1_destroy
 
 	vrf_cleanup
+	sysctl_restore net.ipv6.conf.default.accept_ra
 }
 
 do_mdb_add_del()
@@ -99,6 +169,63 @@ mdb_add_del_test()
 	do_mdb_add_del $PASS_GRP_IP6
 }
 
+do_compat_fwd()
+{
+	port=$1
+	RET=0
+
+	# Ensure default settings, regardless of test start order
+	bridge link set dev "$swp1" mcast_flood on
+	bridge link set dev "$swp2" mcast_flood on
+	bridge link set dev "br0"   mcast_flood on self
+
+	tcpdump_start "$port"
+
+	$MZ -q $h1 "$PASS_PKT_MAC"
+	$MZ -q $h1 "$FAIL_PKT_MAC"
+
+	$MZ -q $h1 "$PASS_PKT_IP4"
+	$MZ -q $h1 "$FAIL_PKT_IP4"
+
+	$MZ -q $h1 "$PASS_PKT_IP6"
+	$MZ -q $h1 "$FAIL_PKT_IP6"
+
+	sleep 1
+	tcpdump_stop "$port"
+
+	tcpdump_show "$port" |grep -q "${SRC_ADDR_MAC} > ${PASS_GRP_MAC}"
+	check_err $? "Failed forwarding multicast group $PASS_GRP_MAC from $h1 to port $port"
+
+	tcpdump_show "$port" |grep -q "${SRC_ADDR_MAC} > ${FAIL_GRP_MAC}"
+	check_err $? "Failed forwarding multicast group ${FAIL_GRP_MAC} from $h1 to port $port"
+
+	tcpdump_show "$port" |grep -q "${SRC_ADDR_IP4}.${SRC_PORT} > ${PASS_GRP_IP4}.${DST_PORT}"
+	check_err $? "Failed forwarding multicast group $PASS_GRP_IP4 from $h1 to port $port"
+
+	tcpdump_show "$port" |grep -q "${SRC_ADDR_IP4}.${SRC_PORT} > ${FAIL_GRP_IP4}.${DST_PORT}"
+	check_err $? "Failed forwarding multicast group ${FAIL_GRP_IP4} from $h1 to port $port"
+
+	tcpdump_show "$port" |grep -q "${SRC_ADDR_IP6}.${SRC_PORT} > ${PASS_GRP_IP6}.${DST_PORT}"
+	check_err $? "Failed forwarding multicast group $PASS_GRP_IP6 from $h1 to port $port"
+
+	tcpdump_show "$port" |grep -q "${SRC_ADDR_IP6}.${SRC_PORT} > ${FAIL_GRP_IP6}.${DST_PORT}"
+	check_err $? "Failed forwarding multicast group ${FAIL_GRP_IP6} from $h1 to port $port"
+
+	log_test "MDB forward unknown MAC/IPv4/IPv6 multicast to port $port"
+	tcpdump_cleanup "$port"
+}
+
+#
+# Verify default behavior, unknown multicast is flooded, to both
+# regular bridge ports and the bridge itself (also a port).
+#
+mdb_compat_fwd_test()
+{
+	do_compat_fwd "$h2"
+	do_compat_fwd "br0"
+}
+}
+
 trap cleanup EXIT
 
 setup_prepare
Signature · 1 line
2.25.1
#35 lore raw

RFC PATCH 11/13 selftests: forwarding: verify strict mdb fwd of known multicast

In reply to Joachim Wiberg

When mcast_flood is disabled forwarding of multicast should strictly follow the mdb, and mcast_router ports, dedicated test added later.

This patch updates bridge_mdb.sh with MAC, IPv4 and IPv6 strict MDB forwarding tests. The bulk of the work is done by do_mdb_fwd(); one MC packet to a known group is verified to reach its destination port, and one MC packet to an unknown group is verified to not be forwarded.

---
 .../selftests/net/forwarding/bridge_mdb.sh    | 88 ++++++++++++++++++-
 1 file changed, 87 insertions(+), 1 deletion(-)
tools/testing/selftests/net/forwarding/bridge_mdb.sh +87 -1
index 6de98c59a620..4e3bb950263f 100755
@@ -7,13 +7,16 @@
 # Verify forwarding (default flooding behavior) to all ports of unknown
 # multicast: MAC, IPv4, IPv6.
 #
+# Verify selective multicast forwarding (strict mdb), when bridge port
+# mcast_flood is disabled, of known MAC, IPv4, IPv6 traffic.
+#
 # Note: this test completely disables IPv6 auto-configuration to avoid
 #       any type of dynamic behavior outside of MLD and IGMP protocols.
 #       Static IPv6 addresses are used to ensure consistent behavior,
 #       even in the startup phase when multicast snooping is enabled.
 #
 
-ALL_TESTS="mdb_add_del_test mdb_compat_fwd_test"
+ALL_TESTS="mdb_add_del_test mdb_compat_fwd_test mdb_mac_fwd_test mdb_ip4_fwd_test mdb_ip6_fwd_test"
 NUM_NETIFS=4
 
 SRC_PORT="1234"
@@ -224,6 +227,89 @@ mdb_compat_fwd_test()
 	do_compat_fwd "$h2"
 	do_compat_fwd "br0"
 }
+
+do_mdb_fwd()
+{
+	type=$1
+	port=$2
+	swp=$port
+	src=$3
+	pass_grp=$4
+	fail_grp=$5
+	pass_pkt=$6
+	fail_pkt=$7
+	RET=0
+
+	if [ "$type" = "MAC" ]; then
+		flag="permanent"
+	else
+		flag=""
+		spt=".$SRC_PORT"
+		dpt=".$DST_PORT"
+	fi
+	if [ "$port" = "$h2" ]; then
+		swp=$swp2
+	fi
+
+	# Disable flooding of unknown multicast, strict MDB forwarding
+	bridge link set dev "$swp1" mcast_flood off
+	bridge link set dev "$swp2" mcast_flood off
+	bridge link set dev "br0"   mcast_flood off self
+
+	# Static filter only to this port
+	bridge mdb add dev br0 port "$swp" grp "$pass_grp" $flag
+	check_err $? "Failed adding $type multicast group $pass_grp to bridge port $swp"
+
+	tcpdump_start "$port"
+
+	# Real data we're expecting
+	$MZ -q "$h1" "$pass_pkt"
+	# This should not pass
+	$MZ -q "$h1" "$fail_pkt"
+
+	sleep 1
+	tcpdump_stop "$port"
+
+	tcpdump_show "$port" |grep -q "$src$spt > $pass_grp$dpt"
+	check_err $? "Failed forwarding $type multicast $pass_grp from $h1 to port $port"
+
+	tcpdump_show "$port" |grep -q "$src$spt > $fail_grp$dpt"
+	check_err_fail 1 $? "Received $type multicast group $fail_grp from $h1 to port $port"
+
+	bridge mdb del dev br0 port "$swp" grp "$pass_grp"
+
+	log_test "MDB forward $type multicast to bridge port $port"
+	tcpdump_cleanup "$port"
+}
+
+#
+# Forwarding of known MAC multicast according to mdb, verify blocking
+# unknown MAC multicast (flood off)
+#
+mdb_mac_fwd_test()
+{
+	do_mdb_fwd MAC "br0" $SRC_ADDR_MAC $PASS_GRP_MAC $FAIL_GRP_MAC "$PASS_PKT_MAC" "$FAIL_PKT_MAC"
+	do_mdb_fwd MAC "$h2" $SRC_ADDR_MAC $PASS_GRP_MAC $FAIL_GRP_MAC "$PASS_PKT_MAC" "$FAIL_PKT_MAC"
+}
+
+#
+# Forwarding of known IPv4 UDP multicast according to mdb, verify
+# blocking unknown IPv4 UDP multicast (flood off)
+#
+mdb_ip4_fwd_test()
+{
+	do_mdb_fwd IPv4 br0 $SRC_ADDR_IP4 $PASS_GRP_IP4 $FAIL_GRP_IP4 "$PASS_PKT_IP4" "$FAIL_PKT_IP4"
+	do_mdb_fwd IPv4 $h2 $SRC_ADDR_IP4 $PASS_GRP_IP4 $FAIL_GRP_IP4 "$PASS_PKT_IP4" "$FAIL_PKT_IP4"
+}
+
+#
+# Forwarding of known IPv6 UDP multicast according to mdb, verify
+# blocking unknown IPv6 UDP multicast (flood off)
+#
+mdb_ip6_fwd_test()
+{
+	do_mdb_fwd IPv6 br0 $SRC_ADDR_IP6 $PASS_GRP_IP6 $FAIL_GRP_IP6 "$PASS_PKT_IP6" "$FAIL_PKT_IP6"
+	do_mdb_fwd IPv6 $h2 $SRC_ADDR_IP6 $PASS_GRP_IP6 $FAIL_GRP_IP6 "$PASS_PKT_IP6" "$FAIL_PKT_IP6"
 }
 
 trap cleanup EXIT
Signature · 1 line
2.25.1
#36 lore raw

RFC PATCH 12/13 selftests: forwarding: verify strict filtering doesn't leak

In reply to Joachim Wiberg

Add a another veth pair to get a dedicated canary port. This to help us verify that multicast is received by only a subset of the ports in the MDB filter.

---
 .../selftests/net/forwarding/bridge_mdb.sh    | 32 ++++++++++++++++++-
 1 file changed, 31 insertions(+), 1 deletion(-)
tools/testing/selftests/net/forwarding/bridge_mdb.sh +31 -1
index 4e3bb950263f..137bc79fd677 100755
@@ -17,7 +17,7 @@
 #
 
 ALL_TESTS="mdb_add_del_test mdb_compat_fwd_test mdb_mac_fwd_test mdb_ip4_fwd_test mdb_ip6_fwd_test"
-NUM_NETIFS=4
+NUM_NETIFS=6
 
 SRC_PORT="1234"
 DST_PORT="4321"
@@ -70,6 +70,16 @@ h2_destroy()
 	simple_if_fini $h2 192.0.2.2/24 2001:db8:1::2/64
 }
 
+h3_create()
+{
+	simple_if_init $h3 192.0.2.3/24 2001:db8:1::3/64
+}
+
+h3_destroy()
+{
+	simple_if_fini $h3 192.0.2.3/24 2001:db8:1::3/64
+}
+
 switch_create()
 {
 	# Enable multicast filtering w/ querier, reduce query response
@@ -88,10 +98,12 @@ switch_create()
 
 	ip link set dev $swp1 master br0
 	ip link set dev $swp2 master br0
+	ip link set dev $swp3 master br0
 
 	ip link set dev br0 up
 	ip link set dev $swp1 up
 	ip link set dev $swp2 up
+	ip link set dev $swp3 up
 
 	# Initial delay, when bridge floods all mcast, is set to 200
 	# above (2 sec.)  We wait 3 sec to handle the case when a single
@@ -102,6 +114,7 @@ switch_create()
 
 switch_destroy()
 {
+	ip link set dev $swp3 down
 	ip link set dev $swp2 down
 	ip link set dev $swp1 down
 
@@ -116,6 +129,9 @@ setup_prepare()
 	swp2=${NETIFS[p3]}
 	h2=${NETIFS[p4]}
 
+	swp3=${NETIFS[p5]}
+	h3=${NETIFS[p6]}
+
 	# Disable all IPv6 autoconfiguration during test, we want
 	# full control of when MLD queries start etc.
 	sysctl_set net.ipv6.conf.default.accept_ra 0
@@ -123,6 +139,7 @@ setup_prepare()
 
 	h1_create
 	h2_create
+	h3_create
 
 	switch_create
 }
@@ -133,6 +150,7 @@ cleanup()
 
 	switch_destroy
 
+	h3_destroy
 	h2_destroy
 	h1_destroy
 
@@ -249,11 +267,15 @@ do_mdb_fwd()
 	fi
 	if [ "$port" = "$h2" ]; then
 		swp=$swp2
+		nop="$h3"
+	else
+		nop="$h2"
 	fi
 
 	# Disable flooding of unknown multicast, strict MDB forwarding
 	bridge link set dev "$swp1" mcast_flood off
 	bridge link set dev "$swp2" mcast_flood off
+	bridge link set dev "$swp3" mcast_flood off
 	bridge link set dev "br0"   mcast_flood off self
 
 	# Static filter only to this port
@@ -261,6 +283,7 @@ do_mdb_fwd()
 	check_err $? "Failed adding $type multicast group $pass_grp to bridge port $swp"
 
 	tcpdump_start "$port"
+	tcpdump_start "$nop"
 
 	# Real data we're expecting
 	$MZ -q "$h1" "$pass_pkt"
@@ -268,6 +291,7 @@ do_mdb_fwd()
 	$MZ -q "$h1" "$fail_pkt"
 
 	sleep 1
+	tcpdump_stop "$nop"
 	tcpdump_stop "$port"
 
 	tcpdump_show "$port" |grep -q "$src$spt > $pass_grp$dpt"
@@ -276,6 +300,12 @@ do_mdb_fwd()
 	tcpdump_show "$port" |grep -q "$src$spt > $fail_grp$dpt"
 	check_err_fail 1 $? "Received $type multicast group $fail_grp from $h1 to port $port"
 
+	# Verify we don't get multicast to the canary port
+	tcpdump_show "$nop" |grep -q "$src$spt > $pass_grp$dpt"
+	check_err_fail 1 $? "Received $type multicast group $pass_grp from $h1 to port $nop"
+	tcpdump_show "$nop" |grep -q "$src$spt > $fail_grp$dpt"
+	check_err_fail 1 $? "Received $type multicast group $fail_grp from $h1 to port $nop"
+
 	bridge mdb del dev br0 port "$swp" grp "$pass_grp"
 
 	log_test "MDB forward $type multicast to bridge port $port"
Signature · 1 line
2.25.1
#37 lore raw

RFC PATCH 13/13 selftests: forwarding: verify flood of known mc on mcast_router port

In reply to Joachim Wiberg

This test verifies that both known (in mdb) and unknown IP multicast is forwarded to a mcast_router port.

---
 .../selftests/net/forwarding/bridge_mdb.sh    | 54 ++++++++++++++++++-
 1 file changed, 53 insertions(+), 1 deletion(-)
tools/testing/selftests/net/forwarding/bridge_mdb.sh +53 -1
index 137bc79fd677..3fd7d68bca09 100755
@@ -10,13 +10,16 @@
 # Verify selective multicast forwarding (strict mdb), when bridge port
 # mcast_flood is disabled, of known MAC, IPv4, IPv6 traffic.
 #
+# Verify flooding towards mcast_router ports of known IP multicast.
+#
 # Note: this test completely disables IPv6 auto-configuration to avoid
 #       any type of dynamic behavior outside of MLD and IGMP protocols.
 #       Static IPv6 addresses are used to ensure consistent behavior,
 #       even in the startup phase when multicast snooping is enabled.
 #
 
-ALL_TESTS="mdb_add_del_test mdb_compat_fwd_test mdb_mac_fwd_test mdb_ip4_fwd_test mdb_ip6_fwd_test"
+ALL_TESTS="mdb_add_del_test mdb_compat_fwd_test mdb_rport_fwd_test \
+	   mdb_mac_fwd_test mdb_ip4_fwd_test mdb_ip6_fwd_test"
 NUM_NETIFS=6
 
 SRC_PORT="1234"
@@ -246,6 +249,55 @@ mdb_compat_fwd_test()
 	do_compat_fwd "br0"
 }
 
+#
+# Verify fwd of IP multicast to router ports.  A detected multicast
+# router should always receive both known and unknown multicast.
+#
+mdb_rport_fwd_test()
+{
+	pass_grp=$PASS_GRP_IP4
+	fail_grp=$FAIL_GRP_IP4
+	pass_pkt=$PASS_PKT_IP4
+	fail_pkt=$FAIL_PKT_IP4
+	decoy="br0"
+	port=$h1
+	type="IPv4"
+
+	# Disable flooding of unknown multicast, strict MDB forwarding
+	bridge link set dev "$swp1" mcast_flood off
+	bridge link set dev "$swp2" mcast_flood off
+	bridge link set dev "br0"   mcast_flood off self
+
+	# Let h2 act as a multicast router
+	ip link set dev "$swp1" type bridge_slave mcast_router 2
+
+	# Static filter only to this decoy port
+	bridge mdb add dev br0 port $decoy grp "$pass_grp"
+	check_err $? "Failed adding multicast group $pass_grp to bridge port $decoy"
+
+	tcpdump_start "$port"
+
+	# Real data we're expecting
+	$MZ -q "$h2" "$pass_pkt"
+	# This should not pass
+	$MZ -q "$h2" "$fail_pkt"
+
+	sleep 1
+	tcpdump_stop "$port"
+
+	tcpdump_show "$port" |grep -q "$src$spt > $pass_grp$dpt"
+	check_err $? "Failed forwarding $type multicast $pass_grp from $h2 to port $port"
+
+	tcpdump_show "$port" |grep -q "$src$spt > $fail_grp$dpt"
+	check_err $? "Failed forwarding $type multicast $fail_grp from $h2 to port $port"
+
+	bridge mdb del dev br0 port br0 grp "$pass_grp"
+	ip link set dev "$swp1" type bridge_slave mcast_router 1
+
+	log_test "MDB forward all $type multicast to multicast router on $port"
+	tcpdump_cleanup "$port"
+}
+
 do_mdb_fwd()
 {
 	type=$1
Signature · 1 line
2.25.1