Skip to contentSkip to conversation

netfilter: bridge: remove BRNF_STATE_BRIDGED flag

PATCH RESEND In review Florian Westphal 2 participants 2 messages Mar 10, 2015 – Mar 16, 2015

Subsystems (3)
Networking netfilter Ethernet bridge
lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH RESEND 1/1 netfilter: bridge: remove BRNF_STATE_BRIDGED flag

Its not needed anymore since 2bf540b73ed5b
([NETFILTER]: bridge-netfilter: remove deferred hooks). Before this it was possible to have physoutdev set for locally generated packets -- this isn't the case anymore:

BRNF_STATE_BRIDGED flag is set when we assign nf_bridge->physoutdev, so physoutdev != NULL means BRNF_STATE_BRIDGED is set.
If physoutdev is NULL, then we are looking at locally-delivered and routed packet.

Signed-off-by Florian Westphal <[email protected]>

---
Resending it outside of rest-of-series because it has no dependencies.

 include/linux/netfilter_bridge.h | 1 -
 net/bridge/br_netfilter.c        | 9 ++++++---
 net/netfilter/xt_physdev.c       | 3 +--
 3 files changed, 7 insertions(+), 6 deletions(-)
include/linux/netfilter_bridge.h +0 -1
index de123d7..ed0d3bf 100644
@@ -19,7 +19,6 @@ enum nf_br_hook_priorities {
 
 #define BRNF_PKT_TYPE			0x01
 #define BRNF_BRIDGED_DNAT		0x02
-#define BRNF_BRIDGED			0x04
 #define BRNF_NF_BRIDGE_PREROUTING	0x08
 #define BRNF_8021Q			0x10
 #define BRNF_PPPoE			0x20
net/bridge/br_netfilter.c +6 -3
index f92a1f7..787d878 100644
@@ -736,8 +736,6 @@ static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
 	if (pf == NFPROTO_IPV4 && br_parse_ip_options(skb))
 		return NF_DROP;
 
-	/* The physdev module checks on this */
-	nf_bridge->mask |= BRNF_BRIDGED;
 	nf_bridge->physoutdev = skb->dev;
 	if (pf == NFPROTO_IPV4)
 		skb->protocol = htons(ETH_P_IP);

@@ -857,7 +855,12 @@ static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
 	struct net_device *realoutdev = bridge_parent(skb->dev);
 	u_int8_t pf;
 
-	if (!nf_bridge || !(nf_bridge->mask & BRNF_BRIDGED))
+	/* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
+	 * on a bridge, but was delivered locally and is now being routed:
+	 *
+	 * POST_ROUTING was already invoked from the ip stack.
+	 */
+	if (!nf_bridge || !nf_bridge->physoutdev)
 		return NF_ACCEPT;
 
 	if (!realoutdev)
net/netfilter/xt_physdev.c +1 -2
index f440f57..50a5204 100644
@@ -56,8 +56,7 @@ physdev_mt(const struct sk_buff *skb, struct xt_action_param *par)
 
 	/* This only makes sense in the FORWARD and POSTROUTING chains */
 	if ((info->bitmask & XT_PHYSDEV_OP_BRIDGED) &&
-	    (!!(nf_bridge->mask & BRNF_BRIDGED) ^
-	    !(info->invert & XT_PHYSDEV_OP_BRIDGED)))
+	    (!!nf_bridge->physoutdev ^ !(info->invert & XT_PHYSDEV_OP_BRIDGED)))
 		return false;
 
 	if ((info->bitmask & XT_PHYSDEV_OP_ISIN &&
Signature · 1 line
2.0.5
maintainer
#1 lore raw
Quoted 9 lines · Florian Westphal

Its not needed anymore since 2bf540b73ed5b
([NETFILTER]: bridge-netfilter: remove deferred hooks). Before this it was possible to have physoutdev set for locally generated packets -- this isn't the case anymore:

BRNF_STATE_BRIDGED flag is set when we assign nf_bridge->physoutdev, so physoutdev != NULL means BRNF_STATE_BRIDGED is set.
If physoutdev is NULL, then we are looking at locally-delivered and routed packet.

Also applied, thanks.