forked from mirrors/linux
		
	net_sched: transform qdisc running bit into a seqcount
Instead of using a single bit (__QDISC___STATE_RUNNING) in sch->__state, use a seqcount. This adds lockdep support, but more importantly it will allow us to sample qdisc/class statistics without having to grab qdisc root lock. Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Cong Wang <xiyou.wangcong@gmail.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
		
							parent
							
								
									64151ae36e
								
							
						
					
					
						commit
						f9eb8aea2a
					
				
					 10 changed files with 32 additions and 16 deletions
				
			
		|  | @ -4610,6 +4610,7 @@ static int bond_check_params(struct bond_params *params) | |||
| static struct lock_class_key bonding_netdev_xmit_lock_key; | ||||
| static struct lock_class_key bonding_netdev_addr_lock_key; | ||||
| static struct lock_class_key bonding_tx_busylock_key; | ||||
| static struct lock_class_key bonding_qdisc_running_key; | ||||
| 
 | ||||
| static void bond_set_lockdep_class_one(struct net_device *dev, | ||||
| 				       struct netdev_queue *txq, | ||||
|  | @ -4625,6 +4626,7 @@ static void bond_set_lockdep_class(struct net_device *dev) | |||
| 			  &bonding_netdev_addr_lock_key); | ||||
| 	netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); | ||||
| 	dev->qdisc_tx_busylock = &bonding_tx_busylock_key; | ||||
| 	dev->qdisc_running_key = &bonding_qdisc_running_key; | ||||
| } | ||||
| 
 | ||||
| /* Called from registration process */ | ||||
|  |  | |||
|  | @ -1313,9 +1313,12 @@ ppp_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats64) | |||
| } | ||||
| 
 | ||||
| static struct lock_class_key ppp_tx_busylock; | ||||
| static struct lock_class_key ppp_qdisc_running_key; | ||||
| 
 | ||||
| static int ppp_dev_init(struct net_device *dev) | ||||
| { | ||||
| 	dev->qdisc_tx_busylock = &ppp_tx_busylock; | ||||
| 	dev->qdisc_running_key = &ppp_qdisc_running_key; | ||||
| 	return 0; | ||||
| } | ||||
| 
 | ||||
|  |  | |||
|  | @ -1577,6 +1577,7 @@ static const struct team_option team_options[] = { | |||
| static struct lock_class_key team_netdev_xmit_lock_key; | ||||
| static struct lock_class_key team_netdev_addr_lock_key; | ||||
| static struct lock_class_key team_tx_busylock_key; | ||||
| static struct lock_class_key team_qdisc_running_key; | ||||
| 
 | ||||
| static void team_set_lockdep_class_one(struct net_device *dev, | ||||
| 				       struct netdev_queue *txq, | ||||
|  | @ -1590,6 +1591,7 @@ static void team_set_lockdep_class(struct net_device *dev) | |||
| 	lockdep_set_class(&dev->addr_list_lock, &team_netdev_addr_lock_key); | ||||
| 	netdev_for_each_tx_queue(dev, team_set_lockdep_class_one, NULL); | ||||
| 	dev->qdisc_tx_busylock = &team_tx_busylock_key; | ||||
| 	dev->qdisc_running_key = &team_qdisc_running_key; | ||||
| } | ||||
| 
 | ||||
| static int team_init(struct net_device *dev) | ||||
|  |  | |||
|  | @ -1862,6 +1862,7 @@ struct net_device { | |||
| #endif | ||||
| 	struct phy_device	*phydev; | ||||
| 	struct lock_class_key	*qdisc_tx_busylock; | ||||
| 	struct lock_class_key	*qdisc_running_key; | ||||
| 	bool			proto_down; | ||||
| }; | ||||
| #define to_net_dev(d) container_of(d, struct net_device, dev) | ||||
|  |  | |||
|  | @ -29,13 +29,6 @@ enum qdisc_state_t { | |||
| 	__QDISC_STATE_THROTTLED, | ||||
| }; | ||||
| 
 | ||||
| /*
 | ||||
|  * following bits are only changed while qdisc lock is held | ||||
|  */ | ||||
| enum qdisc___state_t { | ||||
| 	__QDISC___STATE_RUNNING = 1, | ||||
| }; | ||||
| 
 | ||||
| struct qdisc_size_table { | ||||
| 	struct rcu_head		rcu; | ||||
| 	struct list_head	list; | ||||
|  | @ -93,7 +86,7 @@ struct Qdisc { | |||
| 	unsigned long		state; | ||||
| 	struct sk_buff_head	q; | ||||
| 	struct gnet_stats_basic_packed bstats; | ||||
| 	unsigned int		__state; | ||||
| 	seqcount_t		running; | ||||
| 	struct gnet_stats_queue	qstats; | ||||
| 	struct rcu_head		rcu_head; | ||||
| 	int			padded; | ||||
|  | @ -104,20 +97,20 @@ struct Qdisc { | |||
| 
 | ||||
| static inline bool qdisc_is_running(const struct Qdisc *qdisc) | ||||
| { | ||||
| 	return (qdisc->__state & __QDISC___STATE_RUNNING) ? true : false; | ||||
| 	return (raw_read_seqcount(&qdisc->running) & 1) ? true : false; | ||||
| } | ||||
| 
 | ||||
| static inline bool qdisc_run_begin(struct Qdisc *qdisc) | ||||
| { | ||||
| 	if (qdisc_is_running(qdisc)) | ||||
| 		return false; | ||||
| 	qdisc->__state |= __QDISC___STATE_RUNNING; | ||||
| 	write_seqcount_begin(&qdisc->running); | ||||
| 	return true; | ||||
| } | ||||
| 
 | ||||
| static inline void qdisc_run_end(struct Qdisc *qdisc) | ||||
| { | ||||
| 	qdisc->__state &= ~__QDISC___STATE_RUNNING; | ||||
| 	write_seqcount_end(&qdisc->running); | ||||
| } | ||||
| 
 | ||||
| static inline bool qdisc_may_bulk(const struct Qdisc *qdisc) | ||||
|  |  | |||
|  | @ -629,6 +629,7 @@ static netdev_tx_t bt_xmit(struct sk_buff *skb, struct net_device *netdev) | |||
| 
 | ||||
| static struct lock_class_key bt_tx_busylock; | ||||
| static struct lock_class_key bt_netdev_xmit_lock_key; | ||||
| static struct lock_class_key bt_qdisc_running_key; | ||||
| 
 | ||||
| static void bt_set_lockdep_class_one(struct net_device *dev, | ||||
| 				     struct netdev_queue *txq, | ||||
|  | @ -641,6 +642,7 @@ static int bt_dev_init(struct net_device *dev) | |||
| { | ||||
| 	netdev_for_each_tx_queue(dev, bt_set_lockdep_class_one, NULL); | ||||
| 	dev->qdisc_tx_busylock = &bt_tx_busylock; | ||||
| 	dev->qdisc_running_key = &bt_qdisc_running_key; | ||||
| 
 | ||||
| 	return 0; | ||||
| } | ||||
|  |  | |||
|  | @ -3075,7 +3075,7 @@ static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q, | |||
| 	/*
 | ||||
| 	 * Heuristic to force contended enqueues to serialize on a | ||||
| 	 * separate lock before trying to get qdisc main lock. | ||||
| 	 * This permits __QDISC___STATE_RUNNING owner to get the lock more | ||||
| 	 * This permits qdisc->running owner to get the lock more | ||||
| 	 * often and dequeue packets faster. | ||||
| 	 */ | ||||
| 	contended = qdisc_is_running(q); | ||||
|  |  | |||
|  | @ -60,6 +60,7 @@ static struct header_ops lowpan_header_ops = { | |||
| 
 | ||||
| static struct lock_class_key lowpan_tx_busylock; | ||||
| static struct lock_class_key lowpan_netdev_xmit_lock_key; | ||||
| static struct lock_class_key lowpan_qdisc_running_key; | ||||
| 
 | ||||
| static void lowpan_set_lockdep_class_one(struct net_device *ldev, | ||||
| 					 struct netdev_queue *txq, | ||||
|  | @ -73,6 +74,8 @@ static int lowpan_dev_init(struct net_device *ldev) | |||
| { | ||||
| 	netdev_for_each_tx_queue(ldev, lowpan_set_lockdep_class_one, NULL); | ||||
| 	ldev->qdisc_tx_busylock = &lowpan_tx_busylock; | ||||
| 	ldev->qdisc_running_key = &lowpan_qdisc_running_key; | ||||
| 
 | ||||
| 	return 0; | ||||
| } | ||||
| 
 | ||||
|  |  | |||
|  | @ -68,6 +68,8 @@ static inline struct l2tp_eth_net *l2tp_eth_pernet(struct net *net) | |||
| } | ||||
| 
 | ||||
| static struct lock_class_key l2tp_eth_tx_busylock; | ||||
| static struct lock_class_key l2tp_qdisc_running_key; | ||||
| 
 | ||||
| static int l2tp_eth_dev_init(struct net_device *dev) | ||||
| { | ||||
| 	struct l2tp_eth *priv = netdev_priv(dev); | ||||
|  | @ -76,6 +78,8 @@ static int l2tp_eth_dev_init(struct net_device *dev) | |||
| 	eth_hw_addr_random(dev); | ||||
| 	eth_broadcast_addr(dev->broadcast); | ||||
| 	dev->qdisc_tx_busylock = &l2tp_eth_tx_busylock; | ||||
| 	dev->qdisc_running_key = &l2tp_qdisc_running_key; | ||||
| 
 | ||||
| 	return 0; | ||||
| } | ||||
| 
 | ||||
|  |  | |||
|  | @ -110,7 +110,7 @@ static struct sk_buff *dequeue_skb(struct Qdisc *q, bool *validate, | |||
| 
 | ||||
| /*
 | ||||
|  * Transmit possibly several skbs, and handle the return status as | ||||
|  * required. Holding the __QDISC___STATE_RUNNING bit guarantees that | ||||
|  * required. Owning running seqcount bit guarantees that | ||||
|  * only one CPU can execute this function. | ||||
|  * | ||||
|  * Returns to the caller: | ||||
|  | @ -137,10 +137,10 @@ int sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q, | |||
| 
 | ||||
| 		HARD_TX_UNLOCK(dev, txq); | ||||
| 	} else { | ||||
| 		spin_lock(root_lock); | ||||
| 		spin_lock_nested(root_lock, SINGLE_DEPTH_NESTING); | ||||
| 		return qdisc_qlen(q); | ||||
| 	} | ||||
| 	spin_lock(root_lock); | ||||
| 	spin_lock_nested(root_lock, SINGLE_DEPTH_NESTING); | ||||
| 
 | ||||
| 	if (dev_xmit_complete(ret)) { | ||||
| 		/* Driver sent out skb successfully or skb was consumed */ | ||||
|  | @ -163,7 +163,7 @@ int sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q, | |||
| /*
 | ||||
|  * NOTE: Called under qdisc_lock(q) with locally disabled BH. | ||||
|  * | ||||
|  * __QDISC___STATE_RUNNING guarantees only one CPU can process | ||||
|  * running seqcount guarantees only one CPU can process | ||||
|  * this qdisc at a time. qdisc_lock(q) serializes queue accesses for | ||||
|  * this queue. | ||||
|  * | ||||
|  | @ -379,6 +379,7 @@ struct Qdisc noop_qdisc = { | |||
| 	.list		=	LIST_HEAD_INIT(noop_qdisc.list), | ||||
| 	.q.lock		=	__SPIN_LOCK_UNLOCKED(noop_qdisc.q.lock), | ||||
| 	.dev_queue	=	&noop_netdev_queue, | ||||
| 	.running	=	SEQCNT_ZERO(noop_qdisc.running), | ||||
| 	.busylock	=	__SPIN_LOCK_UNLOCKED(noop_qdisc.busylock), | ||||
| }; | ||||
| EXPORT_SYMBOL(noop_qdisc); | ||||
|  | @ -537,6 +538,7 @@ struct Qdisc_ops pfifo_fast_ops __read_mostly = { | |||
| EXPORT_SYMBOL(pfifo_fast_ops); | ||||
| 
 | ||||
| static struct lock_class_key qdisc_tx_busylock; | ||||
| static struct lock_class_key qdisc_running_key; | ||||
| 
 | ||||
| struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue, | ||||
| 			  const struct Qdisc_ops *ops) | ||||
|  | @ -570,6 +572,10 @@ struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue, | |||
| 	lockdep_set_class(&sch->busylock, | ||||
| 			  dev->qdisc_tx_busylock ?: &qdisc_tx_busylock); | ||||
| 
 | ||||
| 	seqcount_init(&sch->running); | ||||
| 	lockdep_set_class(&sch->running, | ||||
| 			  dev->qdisc_running_key ?: &qdisc_running_key); | ||||
| 
 | ||||
| 	sch->ops = ops; | ||||
| 	sch->enqueue = ops->enqueue; | ||||
| 	sch->dequeue = ops->dequeue; | ||||
|  |  | |||
		Loading…
	
		Reference in a new issue
	
	 Eric Dumazet
						Eric Dumazet