New virtual ethernet driver submitting...

New virtual ethernet driver submitting...

[Katsuyuki Yumoto]

Hello,

I've already written new two drivers of virtual ethernet. These
aggregate(bundle) plural physical or virtual ethernet devices to
single.

1. Link aggregation driver based on IEEE 802.3ad spec.
2. Link redundancy driver based on my original specification (but
   simple and easy to use).

These can be accessed via my WWW site.
http://www.st.rim.or.jp/~yumo
 -> veth-0.4.0.tar.gz (link aggregation)
 -> lr-0.4.0.tar.gz (link redundancy)

I think these are useful for improving performance and availability of
Linux machines. I'd like to submit those, but there are some issues.

- They need a small patch for network core codes(see below), which
  does intercept all the packets incoming via physical interfaces.

  Also, especially link aggregation driver need to get property of
  physical ethernet devices correctly by standard method. -- E.g. link
  status (up or down), link speed and duplex mode. However at this
  time point, each ethernet driver does not report such property by
  standard method. Therefore my small patch introduces such common data
  structure.

- Patch for 2.2.x contains also physical device driver module locking
  code using module reference count. (Patch for 2.4.x does not contain
  it because 2.4.x already has such function.)

  Purpose of this additional patch is avoiding fault because of
  deletion of physical driver module by user before virtual ethernet
  driver release them.

I'd like comments or advices to submit them.

Regards,
Yumo (Katsuyuki Yumoto)

----- Patch for 2.4.0 -----
diff -u -N -r linux.b/include/linux/if_ether.h linux/include/linux/if_ether.h
--- linux.b/include/linux/if_ether.h	Sat Oct 28 03:03:14 2000
+++ linux/include/linux/if_ether.h	Tue May  1 12:23:44 2001
@@ -82,6 +82,7 @@
 #define ETH_P_CONTROL	0x0016		/* Card specific control frames */
 #define ETH_P_IRDA	0x0017		/* Linux-IrDA			*/
 #define ETH_P_ECONET	0x0018		/* Acorn Econet			*/
+#define ETH_P_VETH	0x0019		/* Linux Link Aggregation	*/
 
 /*
  *	This is an Ethernet frame header.
diff -u -N -r linux.b/include/linux/netdevice.h linux/include/linux/netdevice.h
--- linux.b/include/linux/netdevice.h	Fri Jan  5 07:51:20 2001
+++ linux/include/linux/netdevice.h	Tue May  1 12:23:44 2001
@@ -121,6 +121,12 @@
 	/* for cslip etc */
 	unsigned long	rx_compressed;
 	unsigned long	tx_compressed;
+
+	/* for Ethernet devices */
+	unsigned long	media;
+	unsigned long	duplex_mode;
+	int		link_stat;
+
 };
 
 
@@ -132,7 +138,24 @@
         IF_PORT_AUI,
         IF_PORT_100BASET,
         IF_PORT_100BASETX,
-        IF_PORT_100BASEFX
+        IF_PORT_100BASEFX,
+	IF_PORT_1000BASET,
+	IF_PORT_1000BASESX,
+	IF_PORT_1000BASELX
+};
+
+/* Ethernet device link status */
+enum {
+	IF_PORT_LINK_UNKNOWN = 0,
+	IF_PORT_LINK_DOWN,
+	IF_PORT_LINK_UP
+};
+
+/* Ethernet device duplex mode status */
+enum {
+	IF_PORT_DUPLEX_UNKNOWN = 0,
+	IF_PORT_DUPLEX_HALF,
+	IF_PORT_DUPLEX_FULL
 };
 
 #ifdef __KERNEL__
@@ -320,6 +343,9 @@
 	struct Qdisc		*qdisc_list;
 	struct Qdisc		*qdisc_ingress;
 	unsigned long		tx_queue_len;	/* Max frames per queue allowed */
+
+	/* Intercepted ptype */
+	unsigned short		intercepted_ptype;		
 
 	/* hard_start_xmit synchronizer */
 	spinlock_t		xmit_lock;
diff -u -N -r linux.b/net/core/dev.c linux/net/core/dev.c
--- linux.b/net/core/dev.c	Tue Dec 12 06:29:35 2000
+++ linux/net/core/dev.c	Tue May  1 12:23:44 2001
@@ -1346,9 +1346,20 @@
 		{
 			struct packet_type *ptype, *pt_prev;
 			unsigned short type = skb->protocol;
+			unsigned short i_ptype;
 
 			pt_prev = NULL;
-			for (ptype = ptype_all; ptype; ptype = ptype->next) {
+			i_ptype = skb->dev->intercepted_ptype;
+			if (i_ptype){
+				for (ptype = ptype_base[i_ptype&15]; ptype != NULL; ptype = ptype->next) {
+					if (ntohs(ptype->type) == i_ptype) {
+						/* Remember the current last to do */
+						pt_prev=ptype;
+					}
+				}
+			}
+
+			for (ptype = ptype_all; !i_ptype && ptype; ptype = ptype->next) {
 				if (!ptype->dev || ptype->dev == skb->dev) {
 					if (pt_prev) {
 						if (!pt_prev->data) {
@@ -1379,7 +1390,7 @@
 			}
 #endif
 
-			for (ptype=ptype_base[ntohs(type)&15];ptype;ptype=ptype->next) {
+			for (ptype=ptype_base[ntohs(type)&15]; !i_ptype && ptype;ptype=ptype->next) {
 				if (ptype->type == type &&
 				    (!ptype->dev || ptype->dev == skb->dev)) {
 					if (pt_prev) {
@@ -2341,6 +2352,7 @@
 			}
 		}
 		dev->next = NULL;
+		dev->intercepted_ptype = 0;
 		write_lock_bh(&dev_base_lock);
 		*dp = dev;
 		dev_hold(dev);
----- End fo patch for 2.4.0---

----- Patch for 2.2.19 -----
diff -u -N -r linux.b/include/linux/if_ether.h linux/include/linux/if_ether.h
--- linux.b/include/linux/if_ether.h	Mon Mar 26 01:31:03 2001
+++ linux/include/linux/if_ether.h	Fri Oct 12 07:28:13 2001
@@ -79,6 +79,7 @@
 #define ETH_P_MOBITEX	0x0015		/* Mobitex (kaz@cafe.net)	*/
 #define ETH_P_CONTROL	0x0016		/* Card specific control frames */
 #define ETH_P_IRDA	0x0017		/* Linux/IR			*/
+#define ETH_P_VETH	0x0018		/* Linux Link Aggregation	*/
 
 /*
  *	This is an Ethernet frame header.
diff -u -N -r linux.b/include/linux/module.h linux/include/linux/module.h
--- linux.b/include/linux/module.h	Mon Mar 26 01:31:04 2001
+++ linux/include/linux/module.h	Fri Oct 12 07:30:30 2001
@@ -146,6 +146,7 @@
 static inline unsigned long get_module_symbol(char *A, char *B) { return 0; };
 #else
 extern unsigned long get_module_symbol(char *, char *);
+extern struct module *find_module(const char *);
 #endif
 #if defined(MODULE) && !defined(__GENKSYMS__)
 
diff -u -N -r linux.b/include/linux/netdevice.h linux/include/linux/netdevice.h
--- linux.b/include/linux/netdevice.h	Mon Mar 26 01:31:03 2001
+++ linux/include/linux/netdevice.h	Fri Oct 12 07:28:13 2001
@@ -105,6 +105,12 @@
 	/* for cslip etc */
 	unsigned long	rx_compressed;
 	unsigned long	tx_compressed;
+
+	/* for Ethernet devices */
+	unsigned long	media;
+	unsigned long	duplex_mode;
+	int		link_stat;
+
 };
 
 #ifdef CONFIG_NET_FASTROUTE
@@ -125,7 +131,24 @@
         IF_PORT_AUI,
         IF_PORT_100BASET,
         IF_PORT_100BASETX,
-        IF_PORT_100BASEFX
+        IF_PORT_100BASEFX,
+	IF_PORT_1000BASET,
+	IF_PORT_1000BASESX,
+	IF_PORT_1000BASELX
+};
+
+/* Ethernet device link status */
+enum {
+	IF_PORT_LINK_UNKNOWN = 0,
+	IF_PORT_LINK_DOWN,
+	IF_PORT_LINK_UP
+};
+
+/* Ethernet device duplex mode status */
+enum {
+	IF_PORT_DUPLEX_UNKNOWN = 0,
+	IF_PORT_DUPLEX_HALF,
+	IF_PORT_DUPLEX_FULL
 };
 
 #ifdef __KERNEL__
@@ -269,6 +292,9 @@
 	struct Qdisc		*qdisc_sleeping;
 	struct Qdisc		*qdisc_list;
 	unsigned long		tx_queue_len;	/* Max frames per queue allowed */
+
+	/* Intercepted ptype */
+	unsigned short		intercepted_ptype;		
 
 	/* Bridge stuff */
 	int			bridge_port_id;		
diff -u -N -r linux.b/kernel/ksyms.c linux/kernel/ksyms.c
--- linux.b/kernel/ksyms.c	Mon Mar 26 01:31:02 2001
+++ linux/kernel/ksyms.c	Fri Oct 12 07:28:13 2001
@@ -81,6 +81,7 @@
 
 #ifdef CONFIG_MODULES
 EXPORT_SYMBOL(get_module_symbol);
+EXPORT_SYMBOL(find_module);
 #endif
 EXPORT_SYMBOL(get_options);
 
diff -u -N -r linux.b/kernel/module.c linux/kernel/module.c
--- linux.b/kernel/module.c	Mon Mar 26 01:31:02 2001
+++ linux/kernel/module.c	Fri Oct 12 07:28:13 2001
@@ -50,7 +50,7 @@
 
 static long get_mod_name(const char *user_name, char **buf);
 static void put_mod_name(char *buf);
-static struct module *find_module(const char *name);
+/*struct module *find_module(const char *name);*/
 static void free_module(struct module *, int tag_freed);
 
 
@@ -754,7 +754,7 @@
  * Look for a module by name, ignoring modules marked for deletion.
  */
 
-static struct module *
+struct module *
 find_module(const char *name)
 {
 	struct module *mod;
diff -u -N -r linux.b/net/core/dev.c linux/net/core/dev.c
--- linux.b/net/core/dev.c	Mon Mar 26 01:37:41 2001
+++ linux/net/core/dev.c	Fri Oct 12 07:28:13 2001
@@ -903,6 +903,7 @@
 	while (!skb_queue_empty(&backlog)) 
 	{
 		struct sk_buff * skb;
+		unsigned short i_ptype;
 
 		/* Give chance to other bottom halves to run */
 		if (jiffies - start_time > 1)
@@ -986,7 +987,17 @@
 		 */
 
 		pt_prev = NULL;
-		for (ptype = ptype_all; ptype!=NULL; ptype=ptype->next)
+		i_ptype = skb->dev->intercepted_ptype;
+		if (i_ptype){
+			for (ptype = ptype_base[i_ptype&15]; ptype != NULL; ptype = ptype->next) {
+				if (ntohs(ptype->type) == i_ptype) {
+					/* Remember the current last to do */
+					pt_prev=ptype;
+				}
+			}
+		}
+
+		for (ptype = ptype_all; !i_ptype && ptype!=NULL; ptype=ptype->next)
 		{
 			if (!ptype->dev || ptype->dev == skb->dev) {
 				if(pt_prev)
@@ -999,7 +1010,7 @@
 			}
 		}
 
-		for (ptype = ptype_base[ntohs(type)&15]; ptype != NULL; ptype = ptype->next) 
+		for (ptype = ptype_base[ntohs(type)&15]; !i_ptype && ptype != NULL; ptype = ptype->next) 
 		{
 			if (ptype->type == type && (!ptype->dev || ptype->dev==skb->dev))
 			{
@@ -1809,6 +1820,7 @@
 				return -EEXIST;
 		}
 		dev->next = NULL;
+		dev->intercepted_ptype = 0;
 		*dp = dev;
 #ifdef CONFIG_NET_DIVERT
 		ret=alloc_divert_blk(dev);
----- End fo patch for 2.2.19---

Re: New virtual ethernet driver submitting...

[David S. Miller]

   From: Katsuyuki Yumoto <yumoto@jpn.hp.com>
   Date: Wed, 17 Oct 2001 11:47:07 +0900
   
   I've already written new two drivers of virtual ethernet. These
   aggregate(bundle) plural physical or virtual ethernet devices to
   single.

How is this different to, or more beneficial than, what
drivers/net/bonding.c is doing?

Franks a lot,
David S. Miller
davem@redhat.com

Re: New virtual ethernet driver submitting...

[Katsuyuki Yumoto]

>>>>> "David" == David S Miller <davem@redhat.com> writes:

>    From: Katsuyuki Yumoto <yumoto@jpn.hp.com>
>    Date: Wed, 17 Oct 2001 11:47:07 +0900
   
>    I've already written new two drivers of virtual ethernet. These
>    aggregate(bundle) plural physical or virtual ethernet devices to
>    single.

> How is this different to, or more beneficial than, what
> drivers/net/bonding.c is doing?

Yes. I think what to want to do is same between bonding.c and veth
(link aggregation).  But veth has link aggretation maintenance
functionality(LACP) and TCP/UDP port number based distributing
algorithm. Of course, such things are optional features...

On the other hand, purpose of lr(link redundancy) is different of
yours and veth. It's a new link redundant technology rather than
aggregation.

By the way, it seems that your code don't have incoming packet handler
code, right? My drivers contains packet handler for incoming
packets. If your incoming packet handling method is better than mine,
I'd like to follow it. (I'm not sure its method yet.)

Regards,
Yumo (Katsuyuki Yumoto)

Re: New virtual ethernet driver submitting...

[Andi Kleen]

In article <200110170417.f9H4H9g23739@hpujffg8.jpn.hp.com>,
Katsuyuki Yumoto <yumoto@jpn.hp.com> writes:
>>>>>> "David" == David S Miller <davem@redhat.com> writes:

>> From: Katsuyuki Yumoto <yumoto@jpn.hp.com>
>> Date: Wed, 17 Oct 2001 11:47:07 +0900
   
>> I've already written new two drivers of virtual ethernet. These
>> aggregate(bundle) plural physical or virtual ethernet devices to
>> single.

>> How is this different to, or more beneficial than, what
>> drivers/net/bonding.c is doing?

> Yes. I think what to want to do is same between bonding.c and veth
> (link aggregation).  But veth has link aggretation maintenance
> functionality(LACP) and TCP/UDP port number based distributing
> algorithm. Of course, such things are optional features...

This is the fifth link bundling implementation for linux now
(TEQL, bonding, EQL, multipath routing are all in tree and do similar things)

TEQL, EQL do not know about streams, so they're not too useful. There
are some patches for the bonding device to add probalistic stream balancing,
which will likely make it equivalent to your code.

I personally prefer the fourth because it is the cleanest and most flexible.

> On the other hand, purpose of lr(link redundancy) is different of
> yours and veth. It's a new link redundant technology rather than
> aggregation.

> By the way, it seems that your code don't have incoming packet handler
> code, right? My drivers contains packet handler for incoming
> packets. If your incoming packet handling method is better than mine,
> I'd like to follow it. (I'm not sure its method yet.)

multipath routing supports load balancing for incoming connections using
the arp filter. Basically it'll filter arp responses based on the outgoing
filter and this leads to an automatic distribution of incoming traffic.

It distributes routes (IP<->IP,TOS) pairs, not L3 streams however; but that
seems to be sufficient for most usages. It's also very easy to do 802.3ad on 
top of it.

If you wanted to do it for streams I think it would be better to use
the existing infrastructure for it -- this is netfilter connection
tracking -- and make it generate multiple routes.
It also provides failover BTW with some support from user space (e.g. an
OSPF daemon) 

-Andi