 0287587884
			
		
	
	
	0287587884
	
	
	
		
			
			Testing xmit_more support with netperf and connected UDP sockets, I found strange dst refcount false sharing. Current handling of IFF_XMIT_DST_RELEASE is not optimal. Dropping dst in validate_xmit_skb() is certainly too late in case packet was queued by cpu X but dequeued by cpu Y The logical point to take care of drop/force is in __dev_queue_xmit() before even taking qdisc lock. As Julian Anastasov pointed out, need for skb_dst() might come from some packet schedulers or classifiers. This patch adds new helper to cleanly express needs of various drivers or qdiscs/classifiers. Drivers that need skb_dst() in their ndo_start_xmit() should call following helper in their setup instead of the prior : dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; -> netif_keep_dst(dev); Instead of using a single bit, we use two bits, one being eventually rebuilt in bonding/team drivers. The other one, is permanent and blocks IFF_XMIT_DST_RELEASE being rebuilt in bonding/team. Eventually, we could add something smarter later. Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Julian Anastasov <ja@ssi.bg> Signed-off-by: David S. Miller <davem@davemloft.net>
		
			
				
	
	
		
			535 lines
		
	
	
	
		
			12 KiB
			
		
	
	
	
		
			C
		
	
	
	
	
	
			
		
		
	
	
			535 lines
		
	
	
	
		
			12 KiB
			
		
	
	
	
		
			C
		
	
	
	
	
	
| /* net/sched/sch_teql.c	"True" (or "trivial") link equalizer.
 | |
|  *
 | |
|  *		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.
 | |
|  *
 | |
|  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 | |
|  */
 | |
| 
 | |
| #include <linux/module.h>
 | |
| #include <linux/types.h>
 | |
| #include <linux/kernel.h>
 | |
| #include <linux/slab.h>
 | |
| #include <linux/string.h>
 | |
| #include <linux/errno.h>
 | |
| #include <linux/if_arp.h>
 | |
| #include <linux/netdevice.h>
 | |
| #include <linux/init.h>
 | |
| #include <linux/skbuff.h>
 | |
| #include <linux/moduleparam.h>
 | |
| #include <net/dst.h>
 | |
| #include <net/neighbour.h>
 | |
| #include <net/pkt_sched.h>
 | |
| 
 | |
| /*
 | |
|    How to setup it.
 | |
|    ----------------
 | |
| 
 | |
|    After loading this module you will find a new device teqlN
 | |
|    and new qdisc with the same name. To join a slave to the equalizer
 | |
|    you should just set this qdisc on a device f.e.
 | |
| 
 | |
|    # tc qdisc add dev eth0 root teql0
 | |
|    # tc qdisc add dev eth1 root teql0
 | |
| 
 | |
|    That's all. Full PnP 8)
 | |
| 
 | |
|    Applicability.
 | |
|    --------------
 | |
| 
 | |
|    1. Slave devices MUST be active devices, i.e., they must raise the tbusy
 | |
|       signal and generate EOI events. If you want to equalize virtual devices
 | |
|       like tunnels, use a normal eql device.
 | |
|    2. This device puts no limitations on physical slave characteristics
 | |
|       f.e. it will equalize 9600baud line and 100Mb ethernet perfectly :-)
 | |
|       Certainly, large difference in link speeds will make the resulting
 | |
|       eqalized link unusable, because of huge packet reordering.
 | |
|       I estimate an upper useful difference as ~10 times.
 | |
|    3. If the slave requires address resolution, only protocols using
 | |
|       neighbour cache (IPv4/IPv6) will work over the equalized link.
 | |
|       Other protocols are still allowed to use the slave device directly,
 | |
|       which will not break load balancing, though native slave
 | |
|       traffic will have the highest priority.  */
 | |
| 
 | |
| struct teql_master {
 | |
| 	struct Qdisc_ops qops;
 | |
| 	struct net_device *dev;
 | |
| 	struct Qdisc *slaves;
 | |
| 	struct list_head master_list;
 | |
| 	unsigned long	tx_bytes;
 | |
| 	unsigned long	tx_packets;
 | |
| 	unsigned long	tx_errors;
 | |
| 	unsigned long	tx_dropped;
 | |
| };
 | |
| 
 | |
| struct teql_sched_data {
 | |
| 	struct Qdisc *next;
 | |
| 	struct teql_master *m;
 | |
| 	struct sk_buff_head q;
 | |
| };
 | |
| 
 | |
| #define NEXT_SLAVE(q) (((struct teql_sched_data *)qdisc_priv(q))->next)
 | |
| 
 | |
| #define FMASK (IFF_BROADCAST | IFF_POINTOPOINT)
 | |
| 
 | |
| /* "teql*" qdisc routines */
 | |
| 
 | |
| static int
 | |
| teql_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 | |
| {
 | |
| 	struct net_device *dev = qdisc_dev(sch);
 | |
| 	struct teql_sched_data *q = qdisc_priv(sch);
 | |
| 
 | |
| 	if (q->q.qlen < dev->tx_queue_len) {
 | |
| 		__skb_queue_tail(&q->q, skb);
 | |
| 		return NET_XMIT_SUCCESS;
 | |
| 	}
 | |
| 
 | |
| 	return qdisc_drop(skb, sch);
 | |
| }
 | |
| 
 | |
| static struct sk_buff *
 | |
| teql_dequeue(struct Qdisc *sch)
 | |
| {
 | |
| 	struct teql_sched_data *dat = qdisc_priv(sch);
 | |
| 	struct netdev_queue *dat_queue;
 | |
| 	struct sk_buff *skb;
 | |
| 	struct Qdisc *q;
 | |
| 
 | |
| 	skb = __skb_dequeue(&dat->q);
 | |
| 	dat_queue = netdev_get_tx_queue(dat->m->dev, 0);
 | |
| 	q = rcu_dereference_bh(dat_queue->qdisc);
 | |
| 
 | |
| 	if (skb == NULL) {
 | |
| 		struct net_device *m = qdisc_dev(q);
 | |
| 		if (m) {
 | |
| 			dat->m->slaves = sch;
 | |
| 			netif_wake_queue(m);
 | |
| 		}
 | |
| 	} else {
 | |
| 		qdisc_bstats_update(sch, skb);
 | |
| 	}
 | |
| 	sch->q.qlen = dat->q.qlen + q->q.qlen;
 | |
| 	return skb;
 | |
| }
 | |
| 
 | |
| static struct sk_buff *
 | |
| teql_peek(struct Qdisc *sch)
 | |
| {
 | |
| 	/* teql is meant to be used as root qdisc */
 | |
| 	return NULL;
 | |
| }
 | |
| 
 | |
| static inline void
 | |
| teql_neigh_release(struct neighbour *n)
 | |
| {
 | |
| 	if (n)
 | |
| 		neigh_release(n);
 | |
| }
 | |
| 
 | |
| static void
 | |
| teql_reset(struct Qdisc *sch)
 | |
| {
 | |
| 	struct teql_sched_data *dat = qdisc_priv(sch);
 | |
| 
 | |
| 	skb_queue_purge(&dat->q);
 | |
| 	sch->q.qlen = 0;
 | |
| }
 | |
| 
 | |
| static void
 | |
| teql_destroy(struct Qdisc *sch)
 | |
| {
 | |
| 	struct Qdisc *q, *prev;
 | |
| 	struct teql_sched_data *dat = qdisc_priv(sch);
 | |
| 	struct teql_master *master = dat->m;
 | |
| 
 | |
| 	prev = master->slaves;
 | |
| 	if (prev) {
 | |
| 		do {
 | |
| 			q = NEXT_SLAVE(prev);
 | |
| 			if (q == sch) {
 | |
| 				NEXT_SLAVE(prev) = NEXT_SLAVE(q);
 | |
| 				if (q == master->slaves) {
 | |
| 					master->slaves = NEXT_SLAVE(q);
 | |
| 					if (q == master->slaves) {
 | |
| 						struct netdev_queue *txq;
 | |
| 						spinlock_t *root_lock;
 | |
| 
 | |
| 						txq = netdev_get_tx_queue(master->dev, 0);
 | |
| 						master->slaves = NULL;
 | |
| 
 | |
| 						root_lock = qdisc_root_sleeping_lock(rtnl_dereference(txq->qdisc));
 | |
| 						spin_lock_bh(root_lock);
 | |
| 						qdisc_reset(rtnl_dereference(txq->qdisc));
 | |
| 						spin_unlock_bh(root_lock);
 | |
| 					}
 | |
| 				}
 | |
| 				skb_queue_purge(&dat->q);
 | |
| 				break;
 | |
| 			}
 | |
| 
 | |
| 		} while ((prev = q) != master->slaves);
 | |
| 	}
 | |
| }
 | |
| 
 | |
| static int teql_qdisc_init(struct Qdisc *sch, struct nlattr *opt)
 | |
| {
 | |
| 	struct net_device *dev = qdisc_dev(sch);
 | |
| 	struct teql_master *m = (struct teql_master *)sch->ops;
 | |
| 	struct teql_sched_data *q = qdisc_priv(sch);
 | |
| 
 | |
| 	if (dev->hard_header_len > m->dev->hard_header_len)
 | |
| 		return -EINVAL;
 | |
| 
 | |
| 	if (m->dev == dev)
 | |
| 		return -ELOOP;
 | |
| 
 | |
| 	q->m = m;
 | |
| 
 | |
| 	skb_queue_head_init(&q->q);
 | |
| 
 | |
| 	if (m->slaves) {
 | |
| 		if (m->dev->flags & IFF_UP) {
 | |
| 			if ((m->dev->flags & IFF_POINTOPOINT &&
 | |
| 			     !(dev->flags & IFF_POINTOPOINT)) ||
 | |
| 			    (m->dev->flags & IFF_BROADCAST &&
 | |
| 			     !(dev->flags & IFF_BROADCAST)) ||
 | |
| 			    (m->dev->flags & IFF_MULTICAST &&
 | |
| 			     !(dev->flags & IFF_MULTICAST)) ||
 | |
| 			    dev->mtu < m->dev->mtu)
 | |
| 				return -EINVAL;
 | |
| 		} else {
 | |
| 			if (!(dev->flags&IFF_POINTOPOINT))
 | |
| 				m->dev->flags &= ~IFF_POINTOPOINT;
 | |
| 			if (!(dev->flags&IFF_BROADCAST))
 | |
| 				m->dev->flags &= ~IFF_BROADCAST;
 | |
| 			if (!(dev->flags&IFF_MULTICAST))
 | |
| 				m->dev->flags &= ~IFF_MULTICAST;
 | |
| 			if (dev->mtu < m->dev->mtu)
 | |
| 				m->dev->mtu = dev->mtu;
 | |
| 		}
 | |
| 		q->next = NEXT_SLAVE(m->slaves);
 | |
| 		NEXT_SLAVE(m->slaves) = sch;
 | |
| 	} else {
 | |
| 		q->next = sch;
 | |
| 		m->slaves = sch;
 | |
| 		m->dev->mtu = dev->mtu;
 | |
| 		m->dev->flags = (m->dev->flags&~FMASK)|(dev->flags&FMASK);
 | |
| 	}
 | |
| 	return 0;
 | |
| }
 | |
| 
 | |
| 
 | |
| static int
 | |
| __teql_resolve(struct sk_buff *skb, struct sk_buff *skb_res,
 | |
| 	       struct net_device *dev, struct netdev_queue *txq,
 | |
| 	       struct dst_entry *dst)
 | |
| {
 | |
| 	struct neighbour *n;
 | |
| 	int err = 0;
 | |
| 
 | |
| 	n = dst_neigh_lookup_skb(dst, skb);
 | |
| 	if (!n)
 | |
| 		return -ENOENT;
 | |
| 
 | |
| 	if (dst->dev != dev) {
 | |
| 		struct neighbour *mn;
 | |
| 
 | |
| 		mn = __neigh_lookup_errno(n->tbl, n->primary_key, dev);
 | |
| 		neigh_release(n);
 | |
| 		if (IS_ERR(mn))
 | |
| 			return PTR_ERR(mn);
 | |
| 		n = mn;
 | |
| 	}
 | |
| 
 | |
| 	if (neigh_event_send(n, skb_res) == 0) {
 | |
| 		int err;
 | |
| 		char haddr[MAX_ADDR_LEN];
 | |
| 
 | |
| 		neigh_ha_snapshot(haddr, n, dev);
 | |
| 		err = dev_hard_header(skb, dev, ntohs(skb->protocol), haddr,
 | |
| 				      NULL, skb->len);
 | |
| 
 | |
| 		if (err < 0)
 | |
| 			err = -EINVAL;
 | |
| 	} else {
 | |
| 		err = (skb_res == NULL) ? -EAGAIN : 1;
 | |
| 	}
 | |
| 	neigh_release(n);
 | |
| 	return err;
 | |
| }
 | |
| 
 | |
| static inline int teql_resolve(struct sk_buff *skb,
 | |
| 			       struct sk_buff *skb_res,
 | |
| 			       struct net_device *dev,
 | |
| 			       struct netdev_queue *txq)
 | |
| {
 | |
| 	struct dst_entry *dst = skb_dst(skb);
 | |
| 	int res;
 | |
| 
 | |
| 	if (rcu_access_pointer(txq->qdisc) == &noop_qdisc)
 | |
| 		return -ENODEV;
 | |
| 
 | |
| 	if (!dev->header_ops || !dst)
 | |
| 		return 0;
 | |
| 
 | |
| 	rcu_read_lock();
 | |
| 	res = __teql_resolve(skb, skb_res, dev, txq, dst);
 | |
| 	rcu_read_unlock();
 | |
| 
 | |
| 	return res;
 | |
| }
 | |
| 
 | |
| static netdev_tx_t teql_master_xmit(struct sk_buff *skb, struct net_device *dev)
 | |
| {
 | |
| 	struct teql_master *master = netdev_priv(dev);
 | |
| 	struct Qdisc *start, *q;
 | |
| 	int busy;
 | |
| 	int nores;
 | |
| 	int subq = skb_get_queue_mapping(skb);
 | |
| 	struct sk_buff *skb_res = NULL;
 | |
| 
 | |
| 	start = master->slaves;
 | |
| 
 | |
| restart:
 | |
| 	nores = 0;
 | |
| 	busy = 0;
 | |
| 
 | |
| 	q = start;
 | |
| 	if (!q)
 | |
| 		goto drop;
 | |
| 
 | |
| 	do {
 | |
| 		struct net_device *slave = qdisc_dev(q);
 | |
| 		struct netdev_queue *slave_txq = netdev_get_tx_queue(slave, 0);
 | |
| 
 | |
| 		if (slave_txq->qdisc_sleeping != q)
 | |
| 			continue;
 | |
| 		if (netif_xmit_stopped(netdev_get_tx_queue(slave, subq)) ||
 | |
| 		    !netif_running(slave)) {
 | |
| 			busy = 1;
 | |
| 			continue;
 | |
| 		}
 | |
| 
 | |
| 		switch (teql_resolve(skb, skb_res, slave, slave_txq)) {
 | |
| 		case 0:
 | |
| 			if (__netif_tx_trylock(slave_txq)) {
 | |
| 				unsigned int length = qdisc_pkt_len(skb);
 | |
| 
 | |
| 				if (!netif_xmit_frozen_or_stopped(slave_txq) &&
 | |
| 				    netdev_start_xmit(skb, slave, slave_txq, false) ==
 | |
| 				    NETDEV_TX_OK) {
 | |
| 					__netif_tx_unlock(slave_txq);
 | |
| 					master->slaves = NEXT_SLAVE(q);
 | |
| 					netif_wake_queue(dev);
 | |
| 					master->tx_packets++;
 | |
| 					master->tx_bytes += length;
 | |
| 					return NETDEV_TX_OK;
 | |
| 				}
 | |
| 				__netif_tx_unlock(slave_txq);
 | |
| 			}
 | |
| 			if (netif_xmit_stopped(netdev_get_tx_queue(dev, 0)))
 | |
| 				busy = 1;
 | |
| 			break;
 | |
| 		case 1:
 | |
| 			master->slaves = NEXT_SLAVE(q);
 | |
| 			return NETDEV_TX_OK;
 | |
| 		default:
 | |
| 			nores = 1;
 | |
| 			break;
 | |
| 		}
 | |
| 		__skb_pull(skb, skb_network_offset(skb));
 | |
| 	} while ((q = NEXT_SLAVE(q)) != start);
 | |
| 
 | |
| 	if (nores && skb_res == NULL) {
 | |
| 		skb_res = skb;
 | |
| 		goto restart;
 | |
| 	}
 | |
| 
 | |
| 	if (busy) {
 | |
| 		netif_stop_queue(dev);
 | |
| 		return NETDEV_TX_BUSY;
 | |
| 	}
 | |
| 	master->tx_errors++;
 | |
| 
 | |
| drop:
 | |
| 	master->tx_dropped++;
 | |
| 	dev_kfree_skb(skb);
 | |
| 	return NETDEV_TX_OK;
 | |
| }
 | |
| 
 | |
| static int teql_master_open(struct net_device *dev)
 | |
| {
 | |
| 	struct Qdisc *q;
 | |
| 	struct teql_master *m = netdev_priv(dev);
 | |
| 	int mtu = 0xFFFE;
 | |
| 	unsigned int flags = IFF_NOARP | IFF_MULTICAST;
 | |
| 
 | |
| 	if (m->slaves == NULL)
 | |
| 		return -EUNATCH;
 | |
| 
 | |
| 	flags = FMASK;
 | |
| 
 | |
| 	q = m->slaves;
 | |
| 	do {
 | |
| 		struct net_device *slave = qdisc_dev(q);
 | |
| 
 | |
| 		if (slave == NULL)
 | |
| 			return -EUNATCH;
 | |
| 
 | |
| 		if (slave->mtu < mtu)
 | |
| 			mtu = slave->mtu;
 | |
| 		if (slave->hard_header_len > LL_MAX_HEADER)
 | |
| 			return -EINVAL;
 | |
| 
 | |
| 		/* If all the slaves are BROADCAST, master is BROADCAST
 | |
| 		   If all the slaves are PtP, master is PtP
 | |
| 		   Otherwise, master is NBMA.
 | |
| 		 */
 | |
| 		if (!(slave->flags&IFF_POINTOPOINT))
 | |
| 			flags &= ~IFF_POINTOPOINT;
 | |
| 		if (!(slave->flags&IFF_BROADCAST))
 | |
| 			flags &= ~IFF_BROADCAST;
 | |
| 		if (!(slave->flags&IFF_MULTICAST))
 | |
| 			flags &= ~IFF_MULTICAST;
 | |
| 	} while ((q = NEXT_SLAVE(q)) != m->slaves);
 | |
| 
 | |
| 	m->dev->mtu = mtu;
 | |
| 	m->dev->flags = (m->dev->flags&~FMASK) | flags;
 | |
| 	netif_start_queue(m->dev);
 | |
| 	return 0;
 | |
| }
 | |
| 
 | |
| static int teql_master_close(struct net_device *dev)
 | |
| {
 | |
| 	netif_stop_queue(dev);
 | |
| 	return 0;
 | |
| }
 | |
| 
 | |
| static struct rtnl_link_stats64 *teql_master_stats64(struct net_device *dev,
 | |
| 						     struct rtnl_link_stats64 *stats)
 | |
| {
 | |
| 	struct teql_master *m = netdev_priv(dev);
 | |
| 
 | |
| 	stats->tx_packets	= m->tx_packets;
 | |
| 	stats->tx_bytes		= m->tx_bytes;
 | |
| 	stats->tx_errors	= m->tx_errors;
 | |
| 	stats->tx_dropped	= m->tx_dropped;
 | |
| 	return stats;
 | |
| }
 | |
| 
 | |
| static int teql_master_mtu(struct net_device *dev, int new_mtu)
 | |
| {
 | |
| 	struct teql_master *m = netdev_priv(dev);
 | |
| 	struct Qdisc *q;
 | |
| 
 | |
| 	if (new_mtu < 68)
 | |
| 		return -EINVAL;
 | |
| 
 | |
| 	q = m->slaves;
 | |
| 	if (q) {
 | |
| 		do {
 | |
| 			if (new_mtu > qdisc_dev(q)->mtu)
 | |
| 				return -EINVAL;
 | |
| 		} while ((q = NEXT_SLAVE(q)) != m->slaves);
 | |
| 	}
 | |
| 
 | |
| 	dev->mtu = new_mtu;
 | |
| 	return 0;
 | |
| }
 | |
| 
 | |
| static const struct net_device_ops teql_netdev_ops = {
 | |
| 	.ndo_open	= teql_master_open,
 | |
| 	.ndo_stop	= teql_master_close,
 | |
| 	.ndo_start_xmit	= teql_master_xmit,
 | |
| 	.ndo_get_stats64 = teql_master_stats64,
 | |
| 	.ndo_change_mtu	= teql_master_mtu,
 | |
| };
 | |
| 
 | |
| static __init void teql_master_setup(struct net_device *dev)
 | |
| {
 | |
| 	struct teql_master *master = netdev_priv(dev);
 | |
| 	struct Qdisc_ops *ops = &master->qops;
 | |
| 
 | |
| 	master->dev	= dev;
 | |
| 	ops->priv_size  = sizeof(struct teql_sched_data);
 | |
| 
 | |
| 	ops->enqueue	=	teql_enqueue;
 | |
| 	ops->dequeue	=	teql_dequeue;
 | |
| 	ops->peek	=	teql_peek;
 | |
| 	ops->init	=	teql_qdisc_init;
 | |
| 	ops->reset	=	teql_reset;
 | |
| 	ops->destroy	=	teql_destroy;
 | |
| 	ops->owner	=	THIS_MODULE;
 | |
| 
 | |
| 	dev->netdev_ops =       &teql_netdev_ops;
 | |
| 	dev->type		= ARPHRD_VOID;
 | |
| 	dev->mtu		= 1500;
 | |
| 	dev->tx_queue_len	= 100;
 | |
| 	dev->flags		= IFF_NOARP;
 | |
| 	dev->hard_header_len	= LL_MAX_HEADER;
 | |
| 	netif_keep_dst(dev);
 | |
| }
 | |
| 
 | |
| static LIST_HEAD(master_dev_list);
 | |
| static int max_equalizers = 1;
 | |
| module_param(max_equalizers, int, 0);
 | |
| MODULE_PARM_DESC(max_equalizers, "Max number of link equalizers");
 | |
| 
 | |
| static int __init teql_init(void)
 | |
| {
 | |
| 	int i;
 | |
| 	int err = -ENODEV;
 | |
| 
 | |
| 	for (i = 0; i < max_equalizers; i++) {
 | |
| 		struct net_device *dev;
 | |
| 		struct teql_master *master;
 | |
| 
 | |
| 		dev = alloc_netdev(sizeof(struct teql_master), "teql%d",
 | |
| 				   NET_NAME_UNKNOWN, teql_master_setup);
 | |
| 		if (!dev) {
 | |
| 			err = -ENOMEM;
 | |
| 			break;
 | |
| 		}
 | |
| 
 | |
| 		if ((err = register_netdev(dev))) {
 | |
| 			free_netdev(dev);
 | |
| 			break;
 | |
| 		}
 | |
| 
 | |
| 		master = netdev_priv(dev);
 | |
| 
 | |
| 		strlcpy(master->qops.id, dev->name, IFNAMSIZ);
 | |
| 		err = register_qdisc(&master->qops);
 | |
| 
 | |
| 		if (err) {
 | |
| 			unregister_netdev(dev);
 | |
| 			free_netdev(dev);
 | |
| 			break;
 | |
| 		}
 | |
| 
 | |
| 		list_add_tail(&master->master_list, &master_dev_list);
 | |
| 	}
 | |
| 	return i ? 0 : err;
 | |
| }
 | |
| 
 | |
| static void __exit teql_exit(void)
 | |
| {
 | |
| 	struct teql_master *master, *nxt;
 | |
| 
 | |
| 	list_for_each_entry_safe(master, nxt, &master_dev_list, master_list) {
 | |
| 
 | |
| 		list_del(&master->master_list);
 | |
| 
 | |
| 		unregister_qdisc(&master->qops);
 | |
| 		unregister_netdev(master->dev);
 | |
| 		free_netdev(master->dev);
 | |
| 	}
 | |
| }
 | |
| 
 | |
| module_init(teql_init);
 | |
| module_exit(teql_exit);
 | |
| 
 | |
| MODULE_LICENSE("GPL");
 |