[netdrvr bonding] add 802.3ad support

Contributed by Intel, with updates by
Jay Vosburgh @ IBM (bonding maintainer)

drivers/net/bonding/Makefile

+#
+# Makefile for the Ethernet Bonding driver
+#
+
+O_TARGET := bonding.o
+
+obj-y	 := bond_main.o \
+            bond_3ad.o  \
+	    bond_alb.o
+
+obj-m	 := $(O_TARGET)
+
+include $(TOPDIR)/Rules.make
+

drivers/net/bonding/bond_alb.h

+/*
+ * Copyright(c) 1999 - 2003 Intel Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
+ *
+ * The full GNU General Public License is included in this distribution in the
+ * file called LICENSE.
+ */
+
+#ifndef __BOND_ALB_H__
+#define __BOND_ALB_H__
+
+#include <linux/if_ether.h>
+
+struct bonding;
+struct slave;
+
+#define BOND_ALB_INFO(bond)   ((bond)->alb_info)
+#define SLAVE_TLB_INFO(slave) ((slave)->tlb_info)
+
+struct tlb_client_info {
+	struct slave *tx_slave;	/* A pointer to slave used for transmiting
+				 * packets to a Client that the Hash function
+				 * gave this entry index.
+				 */
+	u32 tx_bytes;		/* Each Client acumulates the BytesTx that
+				 * were tranmitted to it, and after each
+				 * CallBack the LoadHistory is devided
+				 * by the balance interval
+				 */
+	u32 load_history;	/* This field contains the amount of Bytes
+				 * that were transmitted to this client by
+				 * the server on the previous balance
+				 * interval in Bps.
+				 */
+	u32 next;		/* The next Hash table entry index, assigned
+				 * to use the same adapter for transmit.
+				 */
+	u32 prev;		/* The previous Hash table entry index,
+				 * assigned to use the same
+				 */
+};
+
+/* -------------------------------------------------------------------------
+ * struct rlb_client_info contains all info related to a specific rx client
+ * connection. This is the Clients Hash Table entry struct
+ * -------------------------------------------------------------------------
+ */
+struct rlb_client_info {
+	u32 ip_src;		/* the server IP address */
+	u32 ip_dst;		/* the client IP address */
+	u8  mac_dst[ETH_ALEN];	/* the client MAC address */
+	u32 next;		/* The next Hash table entry index */
+	u32 prev;		/* The previous Hash table entry index */
+	u8  assigned;		/* checking whether this entry is assigned */
+	u8  ntt;		/* flag - need to transmit client info */
+	struct slave *slave;	/* the slave assigned to this client */
+};
+
+struct tlb_slave_info {
+	u32 head;	/* Index to the head of the bi-directional clients
+			 * hash table entries list. The entries in the list
+			 * are the entries that were assigned to use this
+			 * slave for transmit.
+			 */
+	u32 load;	/* Each slave sums the loadHistory of all clients
+			 * assigned to it
+			 */
+};
+
+struct alb_bond_info {
+	struct timer_list	alb_timer;
+	struct tlb_client_info	*tx_hashtbl; /* Dynamically allocated */
+	spinlock_t		tx_hashtbl_lock;
+	u32			unbalanced_load;
+	int			tx_rebalance_counter;
+	int			lp_counter;
+	/* -------- rlb parameters -------- */
+	int rlb_enabled;
+	struct packet_type	rlb_pkt_type;
+	struct rlb_client_info	*rx_hashtbl;	/* Receive hash table */
+	spinlock_t		rx_hashtbl_lock;
+	u32			rx_hashtbl_head;
+	u8			rx_ntt;	/* flag - need to transmit
+					 * to all rx clients
+					 */
+	struct slave		*next_rx_slave;/* next slave to be assigned
+						* to a new rx client for
+						*/
+	u32			rlb_interval_counter;
+	u8			primary_is_promisc;	   /* boolean */
+	u32			rlb_promisc_timeout_counter;/* counts primary
+							     * promiscuity time
+							     */
+	u32			rlb_update_delay_counter;
+	u32			rlb_update_retry_counter;/* counter of retries
+							  * of client update
+							  */
+	u8			rlb_rebalance;	/* flag - indicates that the
+						 * rx traffic should be
+						 * rebalanced
+						 */
+};
+
+int bond_alb_initialize(struct bonding *bond, int rlb_enabled);
+void bond_alb_deinitialize(struct bonding *bond);
+int bond_alb_init_slave(struct bonding *bond, struct slave *slave);
+void bond_alb_deinit_slave(struct bonding *bond, struct slave *slave);
+void bond_alb_handle_link_change(struct bonding *bond, struct slave *slave, char link);
+void bond_alb_assign_current_slave(struct bonding *bond, struct slave *new_slave);
+int bond_alb_xmit(struct sk_buff *skb, struct net_device *dev);
+void bond_alb_monitor(struct bonding *bond);
+
+#endif /* __BOND_ALB_H__ */
+

drivers/net/bonding/bonding.h

+/*
+ * Bond several ethernet interfaces into a Cisco, running 'Etherchannel'.
+ *
+ * Portions are (c) Copyright 1995 Simon "Guru Aleph-Null" Janes
+ * NCM: Network and Communications Management, Inc.
+ *
+ * BUT, I'm the one who modified it for ethernet, so:
+ * (c) Copyright 1999, Thomas Davis, tadavis@lbl.gov
+ *
+ *	This software may be used and distributed according to the terms
+ *	of the GNU Public License, incorporated herein by reference.
+ * 
+ *
+ * 2003/03/18 - Amir Noam <amir.noam at intel dot com>,
+ *		Tsippy Mendelson <tsippy.mendelson at intel dot com> and
+ *		Shmulik Hen <shmulik.hen at intel dot com>
+ *	- Added support for IEEE 802.3ad Dynamic link aggregation mode.
+ *
+ * 2003/05/01 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
+ *		Amir Noam <amir.noam at intel dot com>
+ *	- Code beautification and style changes (mainly in comments).
+ *
+ * 2003/05/01 - Shmulik Hen <shmulik.hen at intel dot com>
+ *	- Added support for Transmit load balancing mode.
+ */
+ 
+#ifndef _LINUX_BONDING_H
+#define _LINUX_BONDING_H
+
+#include <linux/timer.h>
+#include <linux/proc_fs.h>
+#include "bond_3ad.h"
+#include "bond_alb.h"
+
+#ifdef BONDING_DEBUG
+
+// use this like so: BOND_PRINT_DBG(("foo = %d, bar = %d", foo, bar));
+#define BOND_PRINT_DBG(X)                                     \
+do {                                                          \
+	printk(KERN_DEBUG "%s (%d)", __FUNCTION__, __LINE__); \
+	printk X;                                             \
+	printk("\n");                                         \
+} while(0)
+
+#else
+#define BOND_PRINT_DBG(X)
+#endif /* BONDING_DEBUG */
+
+#define IS_UP(dev)  ((((dev)->flags & (IFF_UP)) == (IFF_UP)) && \
+		     (netif_running(dev) && netif_carrier_ok(dev)))
+
+/* Checks whether the dev is ready for transmit. We do not check netif_running
+ * since a device can be stopped by the driver for short periods of time for
+ * maintainance. dev_queue_xmit() handles this by queing the packet until the
+ * the dev is running again. Keeping packets ordering requires sticking the
+ * same dev as much as possible
+ */
+#define SLAVE_IS_OK(slave) \
+		     ((((slave)->dev->flags & (IFF_UP)) == (IFF_UP)) && \
+		     netif_carrier_ok((slave)->dev) && \
+		     ((slave)->link == BOND_LINK_UP) && \
+		     ((slave)->state == BOND_STATE_ACTIVE))
+
+
+typedef struct slave {
+	struct slave *next;
+	struct slave *prev;
+	struct net_device *dev;
+	short  delay;
+	unsigned long jiffies;
+	char   link;    /* one of BOND_LINK_XXXX */
+	char   state;   /* one of BOND_STATE_XXXX */
+	unsigned short original_flags;
+	u32 link_failure_count;
+	u16    speed;
+	u8     duplex;
+	u8     perm_hwaddr[ETH_ALEN];
+	struct ad_slave_info ad_info; /* HUGE - better to dynamically alloc */
+	struct tlb_slave_info tlb_info;
+} slave_t;
+
+/*
+ * Here are the locking policies for the two bonding locks:
+ *
+ * 1) Get bond->lock when reading/writing slave list.
+ * 2) Get bond->ptrlock when reading/writing bond->current_slave.
+ *    (It is unnecessary when the write-lock is put with bond->lock.)
+ * 3) When we lock with bond->ptrlock, we must lock with bond->lock
+ *    beforehand.
+ */
+typedef struct bonding {
+	slave_t *next;
+	slave_t *prev;
+	slave_t *current_slave;
+	slave_t *primary_slave;
+	slave_t *current_arp_slave;
+	__s32 slave_cnt;
+	rwlock_t lock;
+	rwlock_t ptrlock;
+	struct timer_list mii_timer;
+	struct timer_list arp_timer;
+	struct net_device_stats *stats;
+#ifdef CONFIG_PROC_FS
+	struct proc_dir_entry *bond_proc_dir;
+	struct proc_dir_entry *bond_proc_info_file;
+#endif /* CONFIG_PROC_FS */
+	struct bonding *next_bond;
+	struct net_device *device;
+	struct dev_mc_list *mc_list;
+	unsigned short flags;
+	struct ad_bond_info ad_info;
+	struct alb_bond_info alb_info;
+} bonding_t;
+
+/* Forward declarations */
+void bond_set_slave_active_flags(slave_t *slave);
+void bond_set_slave_inactive_flags(slave_t *slave);
+
+/**
+ * These functions can be used for iterating the slave list
+ * (which is circular)
+ * Caller must hold bond lock for read
+ */
+extern inline struct slave*
+bond_get_first_slave(struct bonding *bond)
+{
+	/* if there are no slaves return NULL */
+	if (bond->next == (slave_t *)bond) {
+		return NULL;
+	}
+	return bond->next;
+}
+
+/**
+ * Caller must hold bond lock for read
+ */
+extern inline struct slave*
+bond_get_next_slave(struct bonding *bond, struct slave *slave)
+{
+	/* If we have reached the last slave return NULL */
+	if (slave->next == bond->next) {
+		return NULL;
+	}
+	return slave->next;
+}
+
+/**
+ * Returns NULL if the net_device does not belong to any of the bond's slaves
+ *
+ * Caller must hold bond lock for read
+ */
+extern inline struct slave*
+bond_get_slave_by_dev(struct bonding *bond, struct net_device *slave_dev)
+{
+	struct slave *our_slave = bond->next;
+
+	/* check if the list of slaves is empty */
+	if (our_slave == (slave_t *)bond) {
+		return NULL;
+	}
+
+	for (; our_slave; our_slave = bond_get_next_slave(bond, our_slave)) {
+		if (our_slave->dev == slave_dev) {
+			break;
+		}
+	}
+	return our_slave;
+}
+
+extern inline struct bonding*
+bond_get_bond_by_slave(struct slave *slave)
+{
+	if (!slave || !slave->dev->master) {
+		return NULL;
+	}
+
+	return (struct bonding *)(slave->dev->master->priv);
+}
+
+#endif /* _LINUX_BONDING_H */
+

include/linux/if_bonding.h

@@ -11,18 +11,38 @@
  *	This software may be used and distributed according to the terms
  *	of the GNU Public License, incorporated herein by reference.
  * 
+ * 2003/03/18 - Amir Noam <amir.noam at intel dot com>
+ *	- Added support for getting slave's speed and duplex via ethtool.
+ *	  Needed for 802.3ad and other future modes.
+ * 
+ * 2003/03/18 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
+ *		Shmulik Hen <shmulik.hen at intel dot com>
+ *	- Enable support of modes that need to use the unique mac address of
+ *	  each slave.
+ *
+ * 2003/03/18 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
+ *		Amir Noam <amir.noam at intel dot com>
+ *	- Moved driver's private data types to bonding.h
+ *
+ * 2003/03/18 - Amir Noam <amir.noam at intel dot com>,
+ *		Tsippy Mendelson <tsippy.mendelson at intel dot com> and
+ *		Shmulik Hen <shmulik.hen at intel dot com>
+ *	- Added support for IEEE 802.3ad Dynamic link aggregation mode.
+ *
+ * 2003/05/01 - Amir Noam <amir.noam at intel dot com>
+ *	- Added ABI version control to restore compatibility between
+ *	  new/old ifenslave and new/old bonding.
  */
 
 #ifndef _LINUX_IF_BONDING_H
 #define _LINUX_IF_BONDING_H
 
-#ifdef __KERNEL__
-#include <linux/timer.h>
 #include <linux/if.h>
-#include <linux/proc_fs.h>
-#endif /* __KERNEL__ */
-
 #include <linux/types.h>
+#include <linux/if_ether.h>
+
+/* userland - kernel ABI version (2003/05/08) */
+#define BOND_ABI_VERSION 1
 
 /*
  * We can remove these ioctl definitions in 2.5.  People should use the
@@ -41,6 +61,9 @@
 #define BOND_MODE_ACTIVEBACKUP	1
 #define BOND_MODE_XOR		2
 #define BOND_MODE_BROADCAST	3
+#define BOND_MODE_8023AD        4
+#define BOND_MODE_TLB           5
+#define BOND_MODE_ALB		6 /* TLB + RLB (receive load balancing) */
 
 /* each slave's link has 4 states */
 #define BOND_LINK_UP    0           /* link is up and running */
@@ -58,11 +81,6 @@
 #define BOND_MULTICAST_ACTIVE   1
 #define BOND_MULTICAST_ALL      2
 
-struct bond_parm_tbl {
-	char *modename;
-	int mode;
-};
-
 typedef struct ifbond {
 	__s32 bond_mode;
 	__s32 num_slaves;
@@ -78,52 +96,15 @@ typedef struct ifslave
 	__u32  link_failure_count;
 } ifslave;
 
-#ifdef __KERNEL__
-typedef struct slave {
-	struct slave *next;
-	struct slave *prev;
-	struct net_device *dev;
-	short  delay;
-	unsigned long jiffies;	
-	char   link;    /* one of BOND_LINK_XXXX */
-	char   state;   /* one of BOND_STATE_XXXX */
-	unsigned short original_flags;
-	u32 link_failure_count;
-} slave_t;
-
-/*
- * Here are the locking policies for the two bonding locks:
- *
- * 1) Get bond->lock when reading/writing slave list.
- * 2) Get bond->ptrlock when reading/writing bond->current_slave.
- *    (It is unnecessary when the write-lock is put with bond->lock.)
- * 3) When we lock with bond->ptrlock, we must lock with bond->lock
- *    beforehand.
- */
-typedef struct bonding {
-	slave_t *next;
-	slave_t *prev;
-	slave_t *current_slave;
-	slave_t *primary_slave;
-	slave_t *current_arp_slave;
-	__s32 slave_cnt;
-	rwlock_t lock;
-	rwlock_t ptrlock;
-	struct timer_list mii_timer;
-	struct timer_list arp_timer;
-	struct net_device_stats *stats;
-#ifdef CONFIG_PROC_FS
-	struct proc_dir_entry *bond_proc_dir;
-	struct proc_dir_entry *bond_proc_info_file;
-#endif /* CONFIG_PROC_FS */
-	struct bonding *next_bond;
-	struct net_device *device;
-	struct dev_mc_list *mc_list;
-	unsigned short flags;
-} bonding_t;
-#endif /* __KERNEL__ */
+struct ad_info {
+	__u16 aggregator_id;
+	__u16 ports;
+	__u16 actor_key;
+	__u16 partner_key;
+	__u8 partner_system[ETH_ALEN];
+};
 
-#endif /* _LINUX_BOND_H */
+#endif /* _LINUX_IF_BONDING_H */
 
 /*
  * Local variables:

include/linux/if_vlan.h

@@ -148,6 +148,7 @@ static inline int __vlan_hwaccel_rx(struct sk_buff *skb,
 {
 	struct net_device_stats *stats;
 
+	skb->real_dev = skb->dev;
 	skb->dev = grp->vlan_devices[vlan_tag & VLAN_VID_MASK];
 	if (skb->dev == NULL) {
 		kfree_skb(skb);

include/linux/skbuff.h

@@ -190,6 +190,7 @@ struct sk_buff {
 	struct sock		*sk;
 	struct timeval		stamp;
 	struct net_device	*dev;
+	struct net_device	*real_dev;
 
 	union {
 		struct tcphdr	*th;

net/core/dev.c

@@ -1385,8 +1385,10 @@ static __inline__ void skb_bond(struct sk_buff *skb)
 {
 	struct net_device *dev = skb->dev;
 
-	if (dev->master)
+	if (dev->master) {
+		skb->real_dev = skb->dev;
 		skb->dev = dev->master;
+	}
 }
 
 static void net_tx_action(struct softirq_action *h)

net/core/skbuff.c

@@ -271,6 +271,7 @@ struct sk_buff *skb_clone(struct sk_buff *skb, int gfp_mask)
 	n->sk = NULL;
 	C(stamp);
 	C(dev);
+	C(real_dev);
 	C(h);
 	C(nh);
 	C(mac);
@@ -334,6 +335,7 @@ static void copy_skb_header(struct sk_buff *new, const struct sk_buff *old)
 	new->list	= NULL;
 	new->sk		= NULL;
 	new->dev	= old->dev;
+	new->real_dev	= old->real_dev;
 	new->priority	= old->priority;
 	new->protocol	= old->protocol;
 	new->dst	= dst_clone(old->dst);