Skip to contentSkip to conversation

netfilter: arp_tables: fix IEEE1394 ARP payload mangling

PATCH Superseded Florian Westphal 2 participants 3 messages Apr 17 – Apr 18

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH 1/1 netfilter: arp_tables: fix IEEE1394 ARP payload mangling

sashiko.dev noticed that similar bug pattern exists in arpt_mangle:
"IEEE1394 ARP payloads omit the target hardware address, advancing arpptr by hln after the source IP address skips over the actual target IP address."

Apply similar fix: check dev->type. If we're asked to mangle what doesn't exist, drop the packet.

Fixes 6752c8db8e0c ("firewire net, ipv4 arp: Extend hardware address and remove driver-level packet inspection.")

---
Collides with a inflight patch.
I'll rebase or discard depending on what netdev@ does.

 net/ipv4/netfilter/arpt_mangle.c | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)
net/ipv4/netfilter/arpt_mangle.c +22 -1
index a4e07e5e9c11..5a3560e1b59b 100644
@@ -13,6 +13,7 @@ static unsigned int
 target(struct sk_buff *skb, const struct xt_action_param *par)
 {
 	const struct arpt_mangle *mangle = par->targinfo;
+	bool has_tgt_devaddr = true;
 	const struct arphdr *arp;
 	unsigned char *arpptr;
 	int pln, hln;
@@ -39,13 +40,33 @@ target(struct sk_buff *skb, const struct xt_action_param *par)
 		memcpy(arpptr, &mangle->u_s.src_ip, pln);
 	}
 	arpptr += pln;
+
+	if (IS_ENABLED(CONFIG_FIREWIRE_NET)) {
+		const struct net_device *dev = skb->dev;
+
+		if (!dev) {
+			/* can't munge without arphrd type. */
+			if (mangle->flags & (ARPT_MANGLE_TDEV|ARPT_MANGLE_TIP))
+				return NF_DROP;
+			return mangle->target;
+		}
+
+		if (dev->type == ARPHRD_IEEE1394)
+			has_tgt_devaddr = false;
+	}
+
 	if (mangle->flags & ARPT_MANGLE_TDEV) {
+		if (!has_tgt_devaddr)
+			return NF_DROP;
+
 		if (ARPT_DEV_ADDR_LEN_MAX < hln ||
 		   (arpptr + hln > skb_tail_pointer(skb)))
 			return NF_DROP;
 		memcpy(arpptr, mangle->tgt_devaddr, hln);
 	}
-	arpptr += hln;
+	if (has_tgt_devaddr)
+		arpptr += hln;
+
 	if (mangle->flags & ARPT_MANGLE_TIP) {
 		if (ARPT_MANGLE_ADDR_LEN_MAX < pln ||
 		   (arpptr + pln > skb_tail_pointer(skb)))
Signature · 1 line
2.52.0
#1 lore raw

Hi Florian,

Quoted 7 lines · Florian Westphal

sashiko.dev noticed that similar bug pattern exists in arpt_mangle:
"IEEE1394 ARP payloads omit the target hardware address, advancing arpptr by hln after the source IP address skips over the actual target IP address."

Apply similar fix: check dev->type. If we're asked to mangle what doesn't exist, drop the packet.

I included a fix for this in:

https://patchwork.kernel.org/project/netdevbpf/patch/[email protected]/

I forgot to mangle the patch title though to:

netfilter: arp_tables: fix IEEE1394 ARP payload parsing

Quoted 4 lines
Fixes 6752c8db8e0c ("firewire net, ipv4 arp: Extend hardware address and remove driver-level packet inspection.")

---
Collides with a inflight patch.

Are you referring to the patch I made?

Quoted 56 lines

I'll rebase or discard depending on what netdev@ does.

 net/ipv4/netfilter/arpt_mangle.c | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)
net/ipv4/netfilter/arpt_mangle.c +22 -1
index a4e07e5e9c11..5a3560e1b59b 100644
@@ -13,6 +13,7 @@ static unsigned int
 target(struct sk_buff *skb, const struct xt_action_param *par)
 {
 	const struct arpt_mangle *mangle = par->targinfo;
+	bool has_tgt_devaddr = true;
 	const struct arphdr *arp;
 	unsigned char *arpptr;
 	int pln, hln;
@@ -39,13 +40,33 @@ target(struct sk_buff *skb, const struct xt_action_param *par)
 		memcpy(arpptr, &mangle->u_s.src_ip, pln);
 	}
 	arpptr += pln;
+
+	if (IS_ENABLED(CONFIG_FIREWIRE_NET)) {
+		const struct net_device *dev = skb->dev;
+
+		if (!dev) {
+			/* can't munge without arphrd type. */
+			if (mangle->flags & (ARPT_MANGLE_TDEV|ARPT_MANGLE_TIP))
+				return NF_DROP;
+			return mangle->target;
+		}
+
+		if (dev->type == ARPHRD_IEEE1394)
+			has_tgt_devaddr = false;
+	}
+
 	if (mangle->flags & ARPT_MANGLE_TDEV) {
+		if (!has_tgt_devaddr)
+			return NF_DROP;
+
 		if (ARPT_DEV_ADDR_LEN_MAX < hln ||
 		   (arpptr + hln > skb_tail_pointer(skb)))
 			return NF_DROP;
 		memcpy(arpptr, mangle->tgt_devaddr, hln);
 	}
-	arpptr += hln;
+	if (has_tgt_devaddr)
+		arpptr += hln;
+
 	if (mangle->flags & ARPT_MANGLE_TIP) {
 		if (ARPT_MANGLE_ADDR_LEN_MAX < pln ||
 		   (arpptr + pln > skb_tail_pointer(skb)))
Signature · 1 line
2.52.0
author maintainer
#2 lore raw
Quoted 8 lines · Pablo Neira Ayuso

Hi Florian,

Quoted 5 lines · Florian Westphal

sashiko.dev noticed that similar bug pattern exists in arpt_mangle:
"IEEE1394 ARP payloads omit the target hardware address, advancing arpptr by hln after the source IP address skips over the actual target IP address."

Quoted 6 lines
Quoted 2 lines

Apply similar fix: check dev->type. If we're asked to mangle what doesn't exist, drop the packet.

I included a fix for this in:

https://patchwork.kernel.org/project/netdevbpf/patch/[email protected]/

I saw that *after* I had made this patch.

Quoted 3 lines
Quoted 1 line

Collides with a inflight patch.

Are you referring to the patch I made?

Yes.