Skip to contentSkip to conversation

add vlan detection on RTNET receive path

PATCH In review François Legal 2 participants 3 messages Jan 13 – Jan 14

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 add vlan detection on RTNET receive path

This patch allows classifying the frame in the right protocol (skb->protocol) when receiving an ethernet frame with a VLAN tag.

This allows to make raw PF_PACKET sockets be transparent to VLAN tagging when filtering on ethertype.

|    MAC Dest @    |     MAC Src @     | EtherType | VLAN TPID | VLAN TCI | Payload |
| 11:22:33:44:55:66 | FF:FF:FF:FF:FF:FF |     8100       |      0806     |    201B     |        ...    |

With this patch and CONFIG_VLAN_8021Q set in kconfig, the packet above will be classified as ARP (skb->protocol = 0806)

The behaviour here is somewhat different from upstream as we dont strip the tag from the frame (for performance reason) as upstream does it and return the VLAN information as out of band (with recvmsg).

--

kernel/drivers/net/stack/eth.c +14 -1
index 427687f..9b3ad94 100644
@@ -23,6 +23,9 @@
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/skbuff.h>
+#ifdef CONFIG_VLAN_8021Q
+#include <linux/if_vlan.h>
+#endif
 
 #include <rtdev.h>
 #include <rtnet_internal.h>
@@ -107,8 +110,18 @@
 			skb->pkt_type = PACKET_OTHERHOST;
 	}
 
-	if (ntohs(eth->h_proto) >= 1536)
+	if (ntohs(eth->h_proto) >= 1536) {
+#ifdef CONFIG_VLAN_8021Q
+		if (eth->h_proto == htons (ETH_P_8021Q)) {
+			struct vlan_hdr * vlan = (struct vlan_hdr *) (skb->data);
+			return vlan->h_vlan_encapsulated_proto;
+		} else {
+			return eth->h_proto;
+		}
+#else
 		return eth->h_proto;
+#endif
+	}
 
 	rawp = skb->data;
 
#1 lore raw
Quoted 8 lines · François Legal

This patch allows classifying the frame in the right protocol (skb->protocol) when receiving an ethernet frame with a VLAN tag.

This allows to make raw PF_PACKET sockets be transparent to VLAN tagging when filtering on ethertype.

|    MAC Dest @    |     MAC Src @     | EtherType | VLAN TPID | VLAN TCI | Payload |
| 11:22:33:44:55:66 | FF:FF:FF:FF:FF:FF |     8100       |      0806     |    201B     |        ...    |

With this patch and CONFIG_VLAN_8021Q set in kconfig, the packet above will be classified as ARP (skb->protocol = 0806)

We generally avoided to bind RTnet settings to regular network stack configs, just to control them independently and avoid surprises on the RTnet side when changing something on the Linux part.

Quoted 3 lines

The behaviour here is somewhat different from upstream as we dont strip the tag from the frame (for performance reason) as upstream does it and return the VLAN information as out of band (with recvmsg).

So, what is the envisioned use case with RTnet? How does the RX-side handling help when not doing anything special for TX? We should document that as well.

Jan

Quoted 37 lines

--

kernel/drivers/net/stack/eth.c +14 -1
index 427687f..9b3ad94 100644
@@ -23,6 +23,9 @@
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/skbuff.h>
+#ifdef CONFIG_VLAN_8021Q
+#include <linux/if_vlan.h>
+#endif
 
 #include <rtdev.h>
 #include <rtnet_internal.h>
@@ -107,8 +110,18 @@
 			skb->pkt_type = PACKET_OTHERHOST;
 	}
 
-	if (ntohs(eth->h_proto) >= 1536)
+	if (ntohs(eth->h_proto) >= 1536) {
+#ifdef CONFIG_VLAN_8021Q
+		if (eth->h_proto == htons (ETH_P_8021Q)) {
+			struct vlan_hdr * vlan = (struct vlan_hdr *) (skb->data);
+			return vlan->h_vlan_encapsulated_proto;
+		} else {
+			return eth->h_proto;
+		}
+#else
 		return eth->h_proto;
+#endif
+	}
 
 	rawp = skb->data;
 
Signature · 2 lines
Siemens AG, Foundational Technologies
Linux Expert Center
#2 lore raw
Quoted 5 lines · Jan Kiszka
Quoted 4 lines · François Legal

This patch allows classifying the frame in the right protocol (skb->protocol) when receiving an ethernet frame with a VLAN tag.

This allows to make raw PF_PACKET sockets be transparent to VLAN tagging when filtering on ethertype.

Quoted 9 lines
Quoted 4 lines
|    MAC Dest @    |     MAC Src @     | EtherType | VLAN TPID | VLAN TCI | Payload |
| 11:22:33:44:55:66 | FF:FF:FF:FF:FF:FF |     8100       |      0806     |    201B     |        ...    |

With this patch and CONFIG_VLAN_8021Q set in kconfig, the packet above will be classified as ARP (skb->protocol = 0806)

We generally avoided to bind RTnet settings to regular network stack configs, just to control them independently and avoid surprises on the RTnet side when changing something on the Linux part.

Ok. I'll submit a new version of the patch

Quoted 8 lines
Quoted 3 lines

The behaviour here is somewhat different from upstream as we dont strip the tag from the frame (for performance reason) as upstream does it and return the VLAN information as out of band (with recvmsg).

So, what is the envisioned use case with RTnet? How does the RX-side handling help when not doing anything special for TX? We should document that as well.

I use it in a data recorder app, where I want to capture some ethertype traffic wether or not this ethertype is in a specific vlan or not.

François

Quoted 43 lines

Jan

Quoted 27 lines

--

kernel/drivers/net/stack/eth.c +8 -1
index 427687f..9b3ad94 100644
@@ -23,6 +23,9 @@
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/skbuff.h>
+#ifdef CONFIG_VLAN_8021Q
+#include <linux/if_vlan.h>
+#endif
 
 #include <rtdev.h>
 #include <rtnet_internal.h>
@@ -107,8 +110,18 @@
 			skb->pkt_type = PACKET_OTHERHOST;
 	}
 
-	if (ntohs(eth->h_proto) >= 1536)
+	if (ntohs(eth->h_proto) >= 1536) {
+#ifdef CONFIG_VLAN_8021Q
+		if (eth->h_proto == htons (ETH_P_8021Q)) {
+			struct vlan_hdr * vlan = (struct vlan_hdr *) (skb->data);
+			return vlan->h_vlan_encapsulated_proto;

kernel/drivers/net/stack/eth.c line 127

Quoted 10 lines
kernel/drivers/net/stack/eth.c +6 -0
+		} else {
+			return eth->h_proto;
+		}
+#else
 		return eth->h_proto;
+#endif
+	}
 
 	rawp = skb->data;
 

--
Siemens AG, Foundational Technologies
Linux Expert Center