Skip to contentSkip to conversation

: MACVLAN driver

PATCH In review Patrick McHardy 6 participants 12 messages Jun 19, 2007 – Jun 21, 2007

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover : MACVLAN driver

These patches add a rewritten version of Ben's macvlan driver, which allows to create virtual ethernet devices. The underlying device is put in promiscous mode and packets are demuxed based on MAC address. It behaves similar to bridge devices in that packets are visible on the real device before delivery to the macvlan driver. The driver supports all features of the underlying device except VLAN acceleration, this is currently very hard to support. I might look into this in the future.

The main downside of this driver is that it adds another hook in netif_receive_skb, unfortunately that is unavoidable. When not compiled in the hook vanishes of course.

Usage is simple:

# ip link add link eth0 type macvlan

Will create a macvlan0 device with a random MAC address on top of eth0. No iproute patches are required.

Please apply, thanks.

 MAINTAINERS                |    6 +
 drivers/net/Kconfig        |   10 +
 drivers/net/Makefile       |    1 +
 drivers/net/macvlan.c      |  468 ++++++++++++++++++++++++++++++++++++++++++++
 include/linux/if_macvlan.h |    9 +
 include/linux/netdevice.h  |    2 +
 net/core/dev.c             |   26 +++
 net/ethernet/eth.c         |    3 +
 8 files changed, 525 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/macvlan.c
 create mode 100644 include/linux/if_macvlan.h

Patrick McHardy (2):
[ETHERNET]: Validate new address in eth_mac_addr [NET]: Add MACVLAN driver

#1 lore raw

PATCH 1/2 : Validate new address in eth_mac_addr

[ETHERNET]: Validate new address in eth_mac_addr

---
commit c92ac42cd3badaf120e97223ac4798abf5226424
tree 824d080a4748fcd6f025bbe7b9466feee46e3303
parent 45da27ba265dba3c740c45d47f584c30d7066f82
author Patrick McHardy <[email protected]> Mon, 18 Jun 2007 16:15:32 +0200 committer Patrick McHardy <[email protected]> Mon, 18 Jun 2007 16:15:32 +0200

 net/ethernet/eth.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
net/ethernet/eth.c +3 -0
index 0ac2524..9298a2c 100644
@@ -266,8 +266,11 @@ void eth_header_cache_update(struct hh_cache *hh, struct net_device *dev,
 static int eth_mac_addr(struct net_device *dev, void *p)
 {
 	struct sockaddr *addr = p;
+
 	if (netif_running(dev))
 		return -EBUSY;
+	if (!is_valid_ether_addr(addr->sa_data))
+		return -EINVAL;
 	memcpy(dev->dev_addr, addr->sa_data, dev->addr_len);
 	return 0;
 }
#2 lore raw

PATCH 2/2 : Add MACVLAN driver

In reply to Patrick McHardy

[NET]: Add MACVLAN driver

Add macvlan driver, which allows to create virtual ethernet devices based on MAC address.

---
commit 890e2ae4ef5599ee34f280af4882f97c2dcfcb7b
tree a41a8564a8add18a2095a18d4a3f1a2ce0c25997
parent c92ac42cd3badaf120e97223ac4798abf5226424
author Patrick McHardy <[email protected]> Tue, 19 Jun 2007 15:01:06 +0200 committer Patrick McHardy <[email protected]> Tue, 19 Jun 2007 15:01:06 +0200

 MAINTAINERS                |    6 +
 drivers/net/Kconfig        |   10 +
 drivers/net/Makefile       |    1 
 drivers/net/macvlan.c      |  468 ++++++++++++++++++++++++++++++++++++++++++++
 include/linux/if_macvlan.h |    9 +
 include/linux/netdevice.h  |    2 
 net/core/dev.c             |   26 ++
 7 files changed, 522 insertions(+), 0 deletions(-)
MAINTAINERS +6 -0
index 4c715a7..7d61360 100644
@@ -2320,6 +2320,12 @@ W:	http://linuxwireless.org/
 T:	git kernel.org:/pub/scm/linux/kernel/git/jbenc/mac80211.git
 S:	Maintained
 
+MACVLAN DRIVER
+P:	Patrick McHardy
+M:	[email protected]
+L:	[email protected]
+S:	Maintained
+
 MARVELL YUKON / SYSKONNECT DRIVER
 P:	Mirko Lindner
 M: 	[email protected]
drivers/net/Kconfig +10 -0
index 7d57f4a..6c3c1f0 100644
@@ -77,6 +77,16 @@ config BONDING
 	  To compile this driver as a module, choose M here: the module
 	  will be called bonding.
 
+config MACVLAN
+	tristate "MAC-VLAN support (EXPERIMENTAL)"
+	depends on EXPERIMENTAL
+	---help---
+	  This allows one to create virtual interfaces that map packets to
+	  or from specific MAC addresses to a particular interface.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called macvlan.
+
 config EQUALIZER
 	tristate "EQL (serial line load balancing) support"
 	---help---
drivers/net/Makefile +1 -0
index a77affa..fc10e37 100644
@@ -125,6 +125,7 @@ obj-$(CONFIG_SLHC) += slhc.o
 
 obj-$(CONFIG_DUMMY) += dummy.o
 obj-$(CONFIG_IFB) += ifb.o
+obj-$(CONFIG_MACVLAN) += macvlan.o
 obj-$(CONFIG_DE600) += de600.o
 obj-$(CONFIG_DE620) += de620.o
 obj-$(CONFIG_LANCE) += lance.o
drivers/net/macvlan.c new file +468 -0

468 lines not on this page · raw message

include/linux/if_macvlan.h new file +9 -0
new file mode 100644
index 0000000..0d9d7ea
@@ -0,0 +1,9 @@
+#ifndef _LINUX_IF_MACVLAN_H
+#define _LINUX_IF_MACVLAN_H
+
+#ifdef __KERNEL__
+
+extern struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *);
+
+#endif /* __KERNEL__ */
+#endif /* _LINUX_IF_MACVLAN_H */
include/linux/netdevice.h +2 -0
index e7913ee..868140d 100644
@@ -535,6 +535,8 @@ struct net_device
 
 	/* bridge stuff */
 	struct net_bridge_port	*br_port;
+	/* macvlan */
+	struct macvlan_port	*macvlan_port;
 
 	/* class/net/name entry */
 	struct device		dev;
net/core/dev.c +26 -0
index 2609062..5974e5b 100644
@@ -98,6 +98,7 @@
 #include <linux/seq_file.h>
 #include <linux/stat.h>
 #include <linux/if_bridge.h>
+#include <linux/if_macvlan.h>
 #include <net/dst.h>
 #include <net/pkt_sched.h>
 #include <net/checksum.h>
@@ -1793,6 +1794,28 @@ static inline struct sk_buff *handle_bridge(struct sk_buff *skb,
 #define handle_bridge(skb, pt_prev, ret, orig_dev)	(skb)
 #endif
 
+#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE)
+struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *skb) __read_mostly;
+EXPORT_SYMBOL_GPL(macvlan_handle_frame_hook);
+
+static inline struct sk_buff *handle_macvlan(struct sk_buff *skb,
+					     struct packet_type **pt_prev,
+					     int *ret,
+					     struct net_device *orig_dev)
+{
+	if (skb->dev->macvlan_port == NULL)
+		return skb;
+
+	if (*pt_prev) {
+		*ret = deliver_skb(skb, *pt_prev, orig_dev);
+		*pt_prev = NULL;
+	}
+	return macvlan_handle_frame_hook(skb);
+}
+#else
+#define handle_macvlan(skb, pt_prev, ret, orig_dev)	(skb)
+#endif
+
 #ifdef CONFIG_NET_CLS_ACT
 /* TODO: Maybe we should just force sch_ingress to be compiled in
  * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions
@@ -1900,6 +1923,9 @@ ncls:
 	skb = handle_bridge(skb, &pt_prev, &ret, orig_dev);
 	if (!skb)
 		goto out;
+	skb = handle_macvlan(skb, &pt_prev, &ret, orig_dev);
+	if (!skb)
+		goto out;
 
 	type = skb->protocol;
 	list_for_each_entry_rcu(ptype, &ptype_base[ntohs(type)&15], list) {
#3 lore raw
In reply to Patrick McHardy
Quoted 15 lines · On Tue, 19 Jun 2007 15:08:41 +0200 (MEST) Patrick McHardy <[email protected]>…

These patches add a rewritten version of Ben's macvlan driver, which allows to create virtual ethernet devices. The underlying device is put in promiscous mode and packets are demuxed based on MAC address. It behaves similar to bridge devices in that packets are visible on the real device before delivery to the macvlan driver. The driver supports all features of the underlying device except VLAN acceleration, this is currently very hard to support. I might look into this in the future.

The main downside of this driver is that it adds another hook in netif_receive_skb, unfortunately that is unavoidable. When not compiled in the hook vanishes of course.

Usage is simple:

Quoted 8 lines
# ip link add link eth0 type macvlan

Will create a macvlan0 device with a random MAC address on top of eth0. No iproute patches are required.

Please apply, thanks.

Looks good. I have some changes to allow devices with multiple MAC addresses (never finished). This device could use that.

Signature · 1 line
Stephen Hemminger <[email protected]>
#4 lore raw
Quoted 2 lines · Stephen Hemminger

Looks good. I have some changes to allow devices with multiple MAC addresses (never finished). This device could use that.

Stephen,

Is this patch available somewhere? I'd be interested in taking a look at it.

- Chris

#5 lore raw
In reply to Stephen Hemminger
Quoted 2 lines · Stephen Hemminger

Looks good. I have some changes to allow devices with multiple MAC addresses (never finished). This device could use that.

How would the driver do that? I was thinking about using dev_add_mc, but wasn't sure if that would work with all drivers for non-multicast addresses.

#6 lore raw
Quoted 9 lines · On Tue, 19 Jun 2007 23:14:21 +0200 Patrick McHardy <[email protected]> wrote:
Quoted 2 lines · Stephen Hemminger

Looks good. I have some changes to allow devices with multiple MAC addresses (never finished). This device could use that.

How would the driver do that? I was thinking about using dev_add_mc, but wasn't sure if that would work with all drivers for non-multicast addresses.

Basically, add a new dev_add_addr hook to netdevice.
Drivers that can do it, would add the hook. If driver didn't have hook, then core would put device into promiscuous. Most devices with multicast allow non-multicast addresses, so it could be a simple extension. Rather than doing it for all devices, wanted to go in one by one and test.

Signature · 1 line
Stephen Hemminger <[email protected]>
maintainer
#7 lore raw
In reply to Patrick McHardy

From: Patrick McHardy <[email protected]>
Date: Tue, 19 Jun 2007 23:14:21 +0200

Quoted 8 lines
Quoted 2 lines · Stephen Hemminger

Looks good. I have some changes to allow devices with multiple MAC addresses (never finished). This device could use that.

How would the driver do that? I was thinking about using dev_add_mc, but wasn't sure if that would work with all drivers for non-multicast addresses.

This is actually a real issue for virtualization, and many if not all current generation ethernet chips support programming several unicast ethernet addresses in the MAC.

Networking switches in domain0 on virtualization hosts use this feature to support seperate MACs per guest node, and if the chip doesn't support this the chip is put into promiscuous mode.

We don't have any clean interfaces by which to do this MAC programming, and we do need something for it soon.

#8 lore raw
Quoted 18 lines · David Miller

From: Patrick McHardy <[email protected]>
Date: Tue, 19 Jun 2007 23:14:21 +0200

Quoted 9 lines
Quoted 3 lines · Stephen Hemminger

Looks good. I have some changes to allow devices with multiple MAC addresses (never finished). This device could use that.

How would the driver do that? I was thinking about using dev_add_mc, but wasn't sure if that would work with all drivers for non-multicast addresses.

This is actually a real issue for virtualization, and many if not all current generation ethernet chips support programming several unicast ethernet addresses in the MAC.

Quoted 8 lines

Networking switches in domain0 on virtualization hosts use this feature to support seperate MACs per guest node, and if the chip doesn't support this the chip is put into promiscuous mode.

We don't have any clean interfaces by which to do this MAC programming, and we do need something for it soon.

Stephen's suggestion sounds reasonable and easily implementable, I'll see if I can come up with a patch.

#9 lore raw
In reply to David Miller
Quoted 11 lines · David Miller

This is actually a real issue for virtualization, and many if not all current generation ethernet chips support programming several unicast ethernet addresses in the MAC.

Networking switches in domain0 on virtualization hosts use this feature to support seperate MACs per guest node, and if the chip doesn't support this the chip is put into promiscuous mode.

We don't have any clean interfaces by which to do this MAC programming, and we do need something for it soon.

Yep, that's been on my long term wish list for a while, as well.

Overall I would like to see a more flexible way of allowing the net stack to learn each NIC's RX filter capabilities, and exploiting them. Plenty of NICs, even 100Mbps ones, support RX filter management that allows scanning for $hw_limit unicast addresses, before having to put the hardware into promisc mode.

Jeff

#10 lore raw
Quoted 16 lines · Jeff Garzik
Quoted 11 lines · David Miller

This is actually a real issue for virtualization, and many if not all current generation ethernet chips support programming several unicast ethernet addresses in the MAC.

Networking switches in domain0 on virtualization hosts use this feature to support seperate MACs per guest node, and if the chip doesn't support this the chip is put into promiscuous mode.

We don't have any clean interfaces by which to do this MAC programming, and we do need something for it soon.

Yep, that's been on my long term wish list for a while, as well.

Quoted 7 lines

Overall I would like to see a more flexible way of allowing the net stack to learn each NIC's RX filter capabilities, and exploiting them. Plenty of NICs, even 100Mbps ones, support RX filter management that allows scanning for $hw_limit unicast addresses, before having to put the hardware into promisc mode.

Jeff

So how do we manage mac address to RX queue association? Ideally we should extend ethtool to do that rather than inventing a new tool - Is that acceptable?

-Prafulla

#11 lore raw
Quoted 23 lines · Prafulla Deuskar
Quoted 22 lines · Jeff Garzik
Quoted 11 lines · David Miller

This is actually a real issue for virtualization, and many if not all current generation ethernet chips support programming several unicast ethernet addresses in the MAC.

Networking switches in domain0 on virtualization hosts use this feature to support seperate MACs per guest node, and if the chip doesn't support this the chip is put into promiscuous mode.

We don't have any clean interfaces by which to do this MAC programming, and we do need something for it soon.

Yep, that's been on my long term wish list for a while, as well.

Overall I would like to see a more flexible way of allowing the net stack to learn each NIC's RX filter capabilities, and exploiting them. Plenty of NICs, even 100Mbps ones, support RX filter management that allows scanning for $hw_limit unicast addresses, before having to put the hardware into promisc mode.

Quoted 3 lines
Quoted 1 line

Jeff

So how do we manage mac address to RX queue association?

This is not about multiple RX queues but filtering multiple unicast addresses without going to promiscous mode. The addresses are used by something outside the driver, like macvlan.