Skip to contentSkip to conversation

net: fix skb->protocol not updated after VLAN network header adjustment

PATCH In review [email protected] 4 participants 10 messages Aug 5 – Aug 7

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

cover net: fix skb->protocol not updated after VLAN network header adjustment

From: Wei Fang <[email protected]>

After calling skb_set_network_header() to skip past the VLAN tag, skb->protocol still points to the outer VLAN EtherType. If skb_probe_transport_header() is then called without updating skb->protocol to the inner protocol, the flow dissector will misparse the inner header as a VLAN header and fail to set transport_header, leaving it at its uninitialized sentinel value (~0U).

This series fixes two instances of this bug in packet_parse_headers() and tap_get_user_xdp() by assigning the return value of vlan_get_protocol_and_depth() to skb->protocol after skb_set_network_header().

Wei Fang (2):
net: packet: fix skb->protocol not updated after VLAN network header
adjustment
net: tap: fix skb->protocol not updated after VLAN network header
adjustment

 drivers/net/tap.c      | 12 +++++++++---
 net/packet/af_packet.c | 12 +++++++++---
 2 files changed, 18 insertions(+), 6 deletions(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH 1/2 net: packet: fix skb->protocol not updated after VLAN network header adjustment

From: Wei Fang <[email protected]>

In packet_parse_headers(), when processing a VLAN-tagged frame on a SOCK_RAW AF_PACKET socket, skb_set_network_header() is called with the depth returned by vlan_get_protocol_and_depth() to advance network_header past the VLAN tag to the inner protocol header. However, skb->protocol was not updated to reflect the inner EtherType resolved by vlan_get_protocol_and_depth(), leaving it pointing to the outer VLAN EtherType (e.g. ETH_P_8021Q).

This mismatch causes skb_probe_transport_header() to invoke the flow dissector with proto=ETH_P_8021Q but nhoff already pointing past the VLAN tag to the inner header. The dissector interprets the inner header bytes as a VLAN header, fails to find a recognizable encapsulated protocol, and returns false. Consequently, transport_header is never set and remains at its uninitialized sentinel value (~0U = 0xFFFF).

Any subsequent code that calls skb_transport_header() or udp_hdr() on such an skb will dereference a pointer 65535 bytes past skb->head, potentially corrupting arbitrary kernel memory.

Save the return value of vlan_get_protocol_and_depth(), which already resolves the inner EtherType, and assign it to skb->protocol after skb_set_network_header(). This keeps skb->protocol and network_header consistent when skb_probe_transport_header() is called, allowing the flow dissector to correctly identify the transport layer header.

Fixes dfed913e8b55 ("net/af_packet: add VLAN support for AF_PACKET SOCK_RAW GSO")
Assisted-by WChat:claude-opus-4-8
---
 net/packet/af_packet.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
net/packet/af_packet.c +9 -3
index 0e1355be89f6..a62d445047c0 100644
@@ -1945,9 +1945,15 @@ static void packet_parse_headers(struct sk_buff *skb, struct socket *sock)
 
 	/* Move network header to the right position for VLAN tagged packets */
 	if (likely(skb->dev->type == ARPHRD_ETHER) &&
-	    eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	    eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb, skb->protocol,
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}
 
 	skb_probe_transport_header(skb);
 }
Signature · 1 line
2.34.1
#2 lore raw

Re: [PATCH net 1/2] net: packet: fix skb->protocol not updated after VLAN network header adjustment

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [High] Overwriting `skb->protocol` with the inner protocol while leaving the VLAN tag inline in the…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Overwriting skb->protocol with the inner protocol while leaving the VLAN tag inline in the payload hides the VLAN tag from the network stack and drivers. This breaks packet transmission for VLAN-tagged packets sent via AF_PACKET SOCK_RAW. --

--- Patch [1]: [PATCH net 1/2] net: packet: fix skb->protocol not updated after VLAN network header adjustment --- commit f62ea124b860e7404cd95296234dc5505cb676c4
Author: Wei Fang <[email protected]>

net: packet: fix skb->protocol not updated after VLAN network header adjustment

This commit updates skb->protocol with the inner protocol resolved by vlan_get_protocol_and_depth() for VLAN-tagged frames sent via AF_PACKET SOCK_RAW. This allows the flow dissector to correctly identify the inner transport header instead of failing and leaving the transport header unset.

Quoted 8 lines
net/packet/af_packet.c +0 -0
index 0e1355be89f64..a62d445047c0e 100644
@@ -1945,9 +1945,15 @@ static void packet_parse_headers(struct sk_buff *skb, struct socket *sock)
 
 	/* Move network header to the right position for VLAN tagged packets */
 	if (likely(skb->dev->type == ARPHRD_ETHER) &&

net/packet/af_packet.c static void packet_parse_headers(struct sk_buff *skb, struct socket *sock) · line 1954

Quoted 10 lines
net/packet/af_packet.c +7 -3
-	    eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	    eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb, skb->protocol,
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;

[Severity: High]
Does overwriting skb->protocol here hide the inline VLAN tag from the network stack and drivers?

Since the VLAN tag is left inline in the MAC payload, updating skb->protocol to the inner protocol (e.g. ETH_P_IP) causes eth_type_vlan(skb->protocol) to become false.

This can lead to vlan_features_check() being bypassed, which might allow the stack to send VLAN packets to a driver with TSO enabled even if the hardware doesn't support it.

Additionally, drivers like ixgbe that check for ETH_P_8021Q might fail to recognize the inline software VLAN tag, potentially causing issues like inserting duplicate VLAN tags when DCB is enabled or breaking egress TC filters.

Quoted 5 lines
net/packet/af_packet.c +2 -0
+		}
+	}
 
 	skb_probe_transport_header(skb);
 }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#3 lore raw

PATCH 2/2 net: tap: fix skb->protocol not updated after VLAN network header adjustment

In reply to [email protected]

From: Wei Fang <[email protected]>

In tap_get_user_xdp(), when processing a VLAN-tagged frame, skb_set_network_header() is called with the depth returned by vlan_get_protocol_and_depth() to advance network_header past the VLAN tag to the inner protocol header. However, skb->protocol was not updated to reflect the inner EtherType, leaving it pointing to the outer VLAN EtherType (e.g. ETH_P_8021Q).

This mismatch has two consequences. First, skb_probe_transport_header() is called after the VLAN adjustment with proto=ETH_P_8021Q but nhoff already pointing past the VLAN tag to the inner header. The flow dissector interprets the inner header bytes as a VLAN header, fails to find a recognizable encapsulated protocol, and returns false. Consequently, transport_header is never set and remains at its uninitialized sentinel value (~0U), causing any subsequent skb_transport_header() or udp_hdr() call to dereference a pointer 65535 bytes past skb->head, potentially corrupting arbitrary kernel memory. Second, TC egress and eBPF programs that inspect skb->protocol directly (e.g. bpf_skb_net_base_len(), bpf_skb_net_grow(), __bpf_redirect_neigh()) will see ETH_P_8021Q instead of the inner protocol and behave incorrectly.

Save the return value of vlan_get_protocol_and_depth(), which already resolves the inner EtherType, and assign it to skb->protocol after skb_set_network_header(). This keeps skb->protocol and network_header consistent for all subsequent processing.

Fixes 8c76e77f9069 ("tap: call skb_probe_transport_header after setting skb->dev")
Assisted-by WChat:claude-opus-4-8
---
 drivers/net/tap.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
drivers/net/tap.c +9 -3
index fae115915c8e..afcc4919bd04 100644
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp)
 	}
 
 	/* Move network header to the right position for VLAN tagged packets */
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb, skb->protocol,
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}
 
 	rcu_read_lock();
 	tap = rcu_dereference(q->tap);
Signature · 1 line
2.34.1
maintainer
#4 lore raw

Re: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment

Quoted 45 lines · wei.fang@ wrote:

From: Wei Fang <[email protected]>

In tap_get_user_xdp(), when processing a VLAN-tagged frame, skb_set_network_header() is called with the depth returned by vlan_get_protocol_and_depth() to advance network_header past the VLAN tag to the inner protocol header. However, skb->protocol was not updated to reflect the inner EtherType, leaving it pointing to the outer VLAN EtherType (e.g. ETH_P_8021Q).

This mismatch has two consequences. First, skb_probe_transport_header() is called after the VLAN adjustment with proto=ETH_P_8021Q but nhoff already pointing past the VLAN tag to the inner header. The flow dissector interprets the inner header bytes as a VLAN header, fails to find a recognizable encapsulated protocol, and returns false. Consequently, transport_header is never set and remains at its uninitialized sentinel value (~0U), causing any subsequent skb_transport_header() or udp_hdr() call to dereference a pointer 65535 bytes past skb->head, potentially corrupting arbitrary kernel memory. Second, TC egress and eBPF programs that inspect skb->protocol directly (e.g. bpf_skb_net_base_len(), bpf_skb_net_grow(), __bpf_redirect_neigh()) will see ETH_P_8021Q instead of the inner protocol and behave incorrectly.

Save the return value of vlan_get_protocol_and_depth(), which already resolves the inner EtherType, and assign it to skb->protocol after skb_set_network_header(). This keeps skb->protocol and network_header consistent for all subsequent processing.

Fixes 8c76e77f9069 ("tap: call skb_probe_transport_header after setting skb->dev")
Assisted-by WChat:claude-opus-4-8
---
 drivers/net/tap.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
drivers/net/tap.c +0 -2
index fae115915c8e..afcc4919bd04 100644
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp)
 	}
 
 	/* Move network header to the right position for VLAN tagged packets */
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp) · line 1092

Quoted 10 lines
drivers/net/tap.c +9 -1
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb, skb->protocol,
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}

Does the same apply to the same call in tap_get_user?

And in general to other callers of vlan_get_protocol_and_depth, including through wrapper skb_network_protocol?

Quoted 6 lines
drivers/net/tap.c +0 -0
 	rcu_read_lock();
 	tap = rcu_dereference(q->tap);
Signature · 1 line
2.34.1
#5 lore raw

RE: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment

Quoted 42 lines
Quoted 21 lines

In tap_get_user_xdp(), when processing a VLAN-tagged frame, skb_set_network_header() is called with the depth returned by vlan_get_protocol_and_depth() to advance network_header past the VLAN tag to the inner protocol header. However, skb->protocol was not updated to reflect the inner EtherType, leaving it pointing to the outer VLAN EtherType (e.g. ETH_P_8021Q).

This mismatch has two consequences. First, skb_probe_transport_header() is called after the VLAN adjustment with proto=ETH_P_8021Q but nhoff already pointing past the VLAN tag to the inner header. The flow dissector interprets the inner header bytes as a VLAN header, fails to find a recognizable encapsulated protocol, and returns false. Consequently, transport_header is never set and remains at its uninitialized sentinel value (~0U), causing any subsequent skb_transport_header() or udp_hdr() call to dereference a pointer 65535 bytes past skb->head, potentially corrupting arbitrary kernel memory. Second, TC egress and eBPF programs that inspect skb->protocol directly (e.g. bpf_skb_net_base_len(), bpf_skb_net_grow(), __bpf_redirect_neigh()) will see ETH_P_8021Q instead of the inner protocol and behave incorrectly.

Quoted 6 lines

Save the return value of vlan_get_protocol_and_depth(), which already resolves the inner EtherType, and assign it to skb->protocol after skb_set_network_header(). This keeps skb->protocol and network_header consistent for all subsequent processing.

Fixes 8c76e77f9069 ("tap: call skb_probe_transport_header after setting

skb->dev")

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q,

Quoted 11 lines
Assisted-by WChat:claude-opus-4-8
---
 drivers/net/tap.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
drivers/net/tap.c +0 -0
index fae115915c8e..afcc4919bd04 100644
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct tap_queue *q,

struct xdp_buff *xdp)

Quoted 2 lines
drivers/net/tap.c +0 -0
 	}
 
Quoted 15 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp) · line 1092

Quoted 13 lines
drivers/net/tap.c +9 -3
 	/* Move network header to the right position for VLAN tagged packets */
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb, skb->protocol,
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}

Does the same apply to the same call in tap_get_user?

The situation of tap_get_user() is different, skb_probe_transport_header() is called before the VLAN adjustment block. So I think transport_header should be correct. The only concern is whether skb->protocol needs to be updated after calling skb_set_network_header().

I'm not sure as I am not fairly familiar with the tap driver. I added this patch because Sashiko reported that the tap driver has the same issue as af_packet.

Quoted 3 lines

And in general to other callers of vlan_get_protocol_and_depth, including through wrapper skb_network_protocol?

I don't think this issue exists elsewhere. The issue arises because skb_probe_transport_header() is called after skb_set_network_header(), and at this point, skb->protocol and network_header are not synchronized ( In __skb_flow_dissect(), nhoff = skb_network_offset(skb) but proto is ETH_P_8021Q or ETH_P_8021AD)), causing skb_probe_transport_header() to fail to set transport_header correctly.

Perhaps the correct approach would be to restore the original skb->protocol value after skb_probe_transport_header(), maintaining consistency with the previous behavior; otherwise, it might introduce new issues.

#6 lore raw

RE: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment

Quoted 44 lines
Quoted 3 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue

Quoted 3 lines
drivers/net/tap.c +0 -0
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct tap_queue

*q,

Quoted 4 lines

struct xdp_buff *xdp)

Quoted 3 lines
drivers/net/tap.c +0 -0
 	}
 
 	/* Move network header to the right position for VLAN tagged

packets */

Quoted 5 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp)

Quoted 5 lines
drivers/net/tap.c +2 -3
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb,

skb->protocol,

Quoted 9 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp) · line 1092

Quoted 7 lines
drivers/net/tap.c +7 -0
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}

Does the same apply to the same call in tap_get_user?

The situation of tap_get_user() is different, skb_probe_transport_header() is called before the VLAN adjustment block. So I think transport_header should be correct. The only concern is whether skb->protocol needs to be updated after calling skb_set_network_header().

I'm not sure as I am not fairly familiar with the tap driver. I added this patch because Sashiko reported that the tap driver has the same issue as af_packet.

Quoted 3 lines

And in general to other callers of vlan_get_protocol_and_depth, including through wrapper skb_network_protocol?

I don't think this issue exists elsewhere. The issue arises because skb_probe_transport_header() is called after skb_set_network_header(), and at this point, skb->protocol and network_header are not synchronized ( In __skb_flow_dissect(), nhoff = skb_network_offset(skb) but proto is ETH_P_8021Q or ETH_P_8021AD)), causing skb_probe_transport_header() to fail to set transport_header correctly.

Quoted 3 lines

Perhaps the correct approach would be to restore the original skb->protocol value after skb_probe_transport_header(), maintaining consistency with the previous behavior; otherwise, it might introduce new issues.

For AF_PACKET, it has been confirmed that skb->protocol does not need to be restored to its initial value; otherwise, the egress tc flower for protocol ip will not match the packet. A known issue is that the packets cannot match the egress TC flower rules for protocol 802.1Q, but this issue exists before this series. This is likely a limitation of using AF_PACKET to send packets.

maintainer
#7 lore raw

RE: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment

Quoted 48 lines · Wei Fang
Quoted 44 lines
Quoted 3 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue

Quoted 3 lines
drivers/net/tap.c +0 -0
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct tap_queue

*q,

Quoted 4 lines

struct xdp_buff *xdp)

Quoted 3 lines
drivers/net/tap.c +0 -0
 	}
 
 	/* Move network header to the right position for VLAN tagged

packets */

Quoted 5 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp)

Quoted 5 lines
drivers/net/tap.c +2 -3
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb,

skb->protocol,

Quoted 9 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp) · line 1092

Quoted 7 lines
drivers/net/tap.c +7 -0
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}

Does the same apply to the same call in tap_get_user?

The situation of tap_get_user() is different, skb_probe_transport_header() is called before the VLAN adjustment block. So I think transport_header should be correct. The only concern is whether skb->protocol needs to be updated after calling skb_set_network_header().

I'm not sure as I am not fairly familiar with the tap driver. I added this patch because Sashiko reported that the tap driver has the same issue as af_packet.

Quoted 3 lines

And in general to other callers of vlan_get_protocol_and_depth, including through wrapper skb_network_protocol?

I don't think this issue exists elsewhere. The issue arises because skb_probe_transport_header() is called after skb_set_network_header(), and at this point, skb->protocol and network_header are not synchronized ( In __skb_flow_dissect(), nhoff = skb_network_offset(skb) but proto is ETH_P_8021Q or ETH_P_8021AD)), causing skb_probe_transport_header() to fail to set transport_header correctly.

Quoted 3 lines

Perhaps the correct approach would be to restore the original skb->protocol value after skb_probe_transport_header(), maintaining consistency with the previous behavior; otherwise, it might introduce new issues.

Quoted 5 lines

For AF_PACKET, it has been confirmed that skb->protocol does not need to be restored to its initial value; otherwise, the egress tc flower for protocol ip will not match the packet. A known issue is that the packets cannot match the egress TC flower rules for protocol 802.1Q, but this issue exists before this series. This is likely a limitation of using AF_PACKET to send packets.

Actually this update of network header to start of the IP header may have been a mistake. If userspace inserts a VLAN packet, that is what should enter the stack.

But for this specific issue: would it make sense to just move skb_probe_transport_header before that adjustment?

#8 lore raw

RE: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment

Quoted 61 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

drivers/net/tap.c static int tap_get_user_xdp(struct

Quoted 3 lines
drivers/net/tap.c +0 -0
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct

tap_queue

Quoted 22 lines
Quoted 22 lines

*q,

Quoted 4 lines

struct xdp_buff *xdp)

Quoted 3 lines
drivers/net/tap.c +0 -0
 	}
 
 	/* Move network header to the right position for VLAN tagged

packets */

Quoted 5 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp)

Quoted 5 lines
drivers/net/tap.c +2 -3
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb,

skb->protocol,

Quoted 9 lines

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp) · line 1092

Quoted 7 lines
drivers/net/tap.c +7 -0
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}

Does the same apply to the same call in tap_get_user?

Quoted 7 lines
Quoted 7 lines

The situation of tap_get_user() is different, skb_probe_transport_header() is called before the VLAN adjustment block. So I think transport_header should be correct. The only concern is whether skb->protocol needs to be updated after calling skb_set_network_header().

I'm not sure as I am not fairly familiar with the tap driver. I added this patch because Sashiko reported that the tap driver has the same issue as

af_packet.

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

And in general to other callers of vlan_get_protocol_and_depth, including through wrapper skb_network_protocol?

I don't think this issue exists elsewhere. The issue arises because skb_probe_transport_header() is called after skb_set_network_header(),

and

Quoted 3 lines
Quoted 3 lines

at this point, skb->protocol and network_header are not synchronized ( In __skb_flow_dissect(), nhoff = skb_network_offset(skb) but proto is ETH_P_8021Q or ETH_P_8021AD)), causing skb_probe_transport_header()

to

Quoted 3 lines
Quoted 3 lines

fail to set transport_header correctly.

Perhaps the correct approach would be to restore the original

skb->protocol

Quoted 1 line
Quoted 1 line

value after skb_probe_transport_header(), maintaining consistency with

the

Quoted 5 lines
Quoted 1 line

previous behavior; otherwise, it might introduce new issues.

For AF_PACKET, it has been confirmed that skb->protocol does not need to be restored to its initial value; otherwise, the egress tc flower for protocol ip will not match the packet. A known issue is that the packets cannot match the

egress

Quoted 2 lines

TC flower rules for protocol 802.1Q, but this issue exists before this series. This is likely a limitation of using AF_PACKET to send packets.

Quoted 6 lines

Actually this update of network header to start of the IP header may have been a mistake. If userspace inserts a VLAN packet, that is what should enter the stack.

But for this specific issue: would it make sense to just move skb_probe_transport_header before that adjustment?

I think it is feasible. The transport_header will be set correctly, and other behaviors will remain consistent with before, without introducing an new issues.

#9 lore raw

Re: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment

In reply to [email protected]
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Incomplete fix: skb->protocol is not updated after network_header adjustment in tap_get_user(). --

--- Patch [2]: [PATCH net 2/2] net: tap: fix skb->protocol not updated after VLAN network header adjustment --- commit 6866d2a8bbb104d45013a4b8fcd7c1800086ae28
Author: Wei Fang <[email protected]>

net: tap: fix skb->protocol not updated after VLAN network header adjustment

This commit fixes an issue in tap_get_user_xdp() where skb->protocol was not updated after advancing the network header past the VLAN tag. It sets the protocol to the inner EtherType to keep skb->protocol and network_header consistent for subsequent processing.

Quoted 10 lines
drivers/net/tap.c +0 -2
index fae115915c8ef..afcc4919bd042 100644
@@ -1081,9 +1081,15 @@ static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp)
 	}
 
 	/* Move network header to the right position for VLAN tagged packets */
-	if (eth_type_vlan(skb->protocol) &&
-	    vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)

drivers/net/tap.c static int tap_get_user_xdp(struct tap_queue *q, struct xdp_buff *xdp) · line 1092

Quoted 10 lines
drivers/net/tap.c +9 -1
-		skb_set_network_header(skb, depth);
+	if (eth_type_vlan(skb->protocol)) {
+		__be16 proto = vlan_get_protocol_and_depth(skb, skb->protocol,
+							   &depth);
+
+		if (proto != 0) {
+			skb_set_network_header(skb, depth);
+			skb->protocol = proto;
+		}
+	}

[Severity: High]
This isn't a bug introduced by this patch, but does tap_get_user() also need this same fix for non-XDP transmission paths?

Looking at tap_get_user() in drivers/net/tap.c, it contains the identical logic that was patched here:

drivers/net/tap.c:tap_get_user() {
    ...
    /* Move network header to the right position for VLAN tagged packets */
    if (eth_type_vlan(skb->protocol) &&
        vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0)
        skb_set_network_header(skb, depth);
    ...
}

Without updating skb->protocol here as well, will TC egress and eBPF programs still read skb->protocol as ETH_P_8021Q while network_header points to the inner IP header?

If a user or guest VM writes a VLAN-tagged packet to a TAP file descriptor using standard (non-XDP) I/O, could this allow them to craft an IP header that causes traffic classification and security policies to misinterpret the packet?

Quoted 3 lines
drivers/net/tap.c +0 -0
 	rcu_read_lock();
 	tap = rcu_dereference(q->tap);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2