linux 2.6.16.38 w/ vs2.0.3-rc1
[linux-2.6.git] / net / sched / sch_generic.c
index 74d4a1d..0326752 100644 (file)
    The idea is the following:
    - enqueue, dequeue are serialized via top level device
      spinlock dev->queue_lock.
-   - tree walking is protected by read_lock_bh(qdisc_tree_lock)
+   - tree walking is protected by read_lock(qdisc_tree_lock)
      and this lock is used only in process context.
-   - updates to tree are made under rtnl semaphore or
-     from softirq context (__qdisc_destroy rcu-callback)
-     hence this lock needs local bh disabling.
+   - updates to tree are made only under rtnl semaphore,
+     hence this lock may be made without local bh disabling.
 
    qdisc_tree_lock must be grabbed BEFORE dev->queue_lock!
  */
@@ -58,23 +57,23 @@ DEFINE_RWLOCK(qdisc_tree_lock);
 
 void qdisc_lock_tree(struct net_device *dev)
 {
-       write_lock_bh(&qdisc_tree_lock);
+       write_lock(&qdisc_tree_lock);
        spin_lock_bh(&dev->queue_lock);
 }
 
 void qdisc_unlock_tree(struct net_device *dev)
 {
        spin_unlock_bh(&dev->queue_lock);
-       write_unlock_bh(&qdisc_tree_lock);
+       write_unlock(&qdisc_tree_lock);
 }
 
 /* 
    dev->queue_lock serializes queue accesses for this device
    AND dev->qdisc pointer itself.
 
-   netif_tx_lock serializes accesses to device driver.
+   dev->xmit_lock serializes accesses to device driver.
 
-   dev->queue_lock and netif_tx_lock are mutually exclusive,
+   dev->queue_lock and dev->xmit_lock are mutually exclusive,
    if one is grabbed, another must be free.
  */
 
@@ -90,17 +89,14 @@ void qdisc_unlock_tree(struct net_device *dev)
    NOTE: Called under dev->queue_lock with locally disabled BH.
 */
 
-static inline int qdisc_restart(struct net_device *dev)
+int qdisc_restart(struct net_device *dev)
 {
        struct Qdisc *q = dev->qdisc;
        struct sk_buff *skb;
 
        /* Dequeue packet */
-       if (((skb = dev->gso_skb)) || ((skb = q->dequeue(q)))) {
+       if ((skb = q->dequeue(q)) != NULL) {
                unsigned nolock = (dev->features & NETIF_F_LLTX);
-
-               dev->gso_skb = NULL;
-
                /*
                 * When the driver has LLTX set it does its own locking
                 * in start_xmit. No need to add additional overhead by
@@ -111,7 +107,7 @@ static inline int qdisc_restart(struct net_device *dev)
                 * will be requeued.
                 */
                if (!nolock) {
-                       if (!netif_tx_trylock(dev)) {
+                       if (!spin_trylock(&dev->xmit_lock)) {
                        collision:
                                /* So, someone grabbed the driver. */
                                
@@ -129,6 +125,8 @@ static inline int qdisc_restart(struct net_device *dev)
                                __get_cpu_var(netdev_rx_stat).cpu_collision++;
                                goto requeue;
                        }
+                       /* Remember that the driver is grabbed by us. */
+                       dev->xmit_lock_owner = smp_processor_id();
                }
                
                {
@@ -137,11 +135,14 @@ static inline int qdisc_restart(struct net_device *dev)
 
                        if (!netif_queue_stopped(dev)) {
                                int ret;
+                               if (netdev_nit)
+                                       dev_queue_xmit_nit(skb, dev);
 
-                               ret = dev_hard_start_xmit(skb, dev);
+                               ret = dev->hard_start_xmit(skb, dev);
                                if (ret == NETDEV_TX_OK) { 
                                        if (!nolock) {
-                                               netif_tx_unlock(dev);
+                                               dev->xmit_lock_owner = -1;
+                                               spin_unlock(&dev->xmit_lock);
                                        }
                                        spin_lock(&dev->queue_lock);
                                        return -1;
@@ -155,7 +156,8 @@ static inline int qdisc_restart(struct net_device *dev)
                        /* NETDEV_TX_BUSY - we need to requeue */
                        /* Release the driver */
                        if (!nolock) { 
-                               netif_tx_unlock(dev);
+                               dev->xmit_lock_owner = -1;
+                               spin_unlock(&dev->xmit_lock);
                        } 
                        spin_lock(&dev->queue_lock);
                        q = dev->qdisc;
@@ -172,10 +174,7 @@ static inline int qdisc_restart(struct net_device *dev)
                 */
 
 requeue:
-               if (skb->next)
-                       dev->gso_skb = skb;
-               else
-                       q->ops->requeue(skb, q);
+               q->ops->requeue(skb, q);
                netif_schedule(dev);
                return 1;
        }
@@ -183,39 +182,25 @@ requeue:
        return q->q.qlen;
 }
 
-void __qdisc_run(struct net_device *dev)
-{
-       if (unlikely(dev->qdisc == &noop_qdisc))
-               goto out;
-
-       while (qdisc_restart(dev) < 0 && !netif_queue_stopped(dev))
-               /* NOTHING */;
-
-out:
-       clear_bit(__LINK_STATE_QDISC_RUNNING, &dev->state);
-}
-
 static void dev_watchdog(unsigned long arg)
 {
        struct net_device *dev = (struct net_device *)arg;
 
-       netif_tx_lock(dev);
+       spin_lock(&dev->xmit_lock);
        if (dev->qdisc != &noop_qdisc) {
                if (netif_device_present(dev) &&
                    netif_running(dev) &&
                    netif_carrier_ok(dev)) {
                        if (netif_queue_stopped(dev) &&
-                           time_after(jiffies, dev->trans_start + dev->watchdog_timeo)) {
-
-                               printk(KERN_INFO "NETDEV WATCHDOG: %s: transmit timed out\n",
-                                      dev->name);
+                           (jiffies - dev->trans_start) > dev->watchdog_timeo) {
+                               printk(KERN_INFO "NETDEV WATCHDOG: %s: transmit timed out\n", dev->name);
                                dev->tx_timeout(dev);
                        }
                        if (!mod_timer(&dev->watchdog_timer, jiffies + dev->watchdog_timeo))
                                dev_hold(dev);
                }
        }
-       netif_tx_unlock(dev);
+       spin_unlock(&dev->xmit_lock);
 
        dev_put(dev);
 }
@@ -239,17 +224,17 @@ void __netdev_watchdog_up(struct net_device *dev)
 
 static void dev_watchdog_up(struct net_device *dev)
 {
-       netif_tx_lock_bh(dev);
+       spin_lock_bh(&dev->xmit_lock);
        __netdev_watchdog_up(dev);
-       netif_tx_unlock_bh(dev);
+       spin_unlock_bh(&dev->xmit_lock);
 }
 
 static void dev_watchdog_down(struct net_device *dev)
 {
-       netif_tx_lock_bh(dev);
+       spin_lock_bh(&dev->xmit_lock);
        if (del_timer(&dev->watchdog_timer))
-               dev_put(dev);
-       netif_tx_unlock_bh(dev);
+               __dev_put(dev);
+       spin_unlock_bh(&dev->xmit_lock);
 }
 
 void netif_carrier_on(struct net_device *dev)
@@ -487,20 +472,6 @@ void qdisc_reset(struct Qdisc *qdisc)
 static void __qdisc_destroy(struct rcu_head *head)
 {
        struct Qdisc *qdisc = container_of(head, struct Qdisc, q_rcu);
-       struct Qdisc_ops  *ops = qdisc->ops;
-
-#ifdef CONFIG_NET_ESTIMATOR
-       gen_kill_estimator(&qdisc->bstats, &qdisc->rate_est);
-#endif
-       write_lock(&qdisc_tree_lock);
-       if (ops->reset)
-               ops->reset(qdisc);
-       if (ops->destroy)
-               ops->destroy(qdisc);
-       write_unlock(&qdisc_tree_lock);
-       module_put(ops->owner);
-
-       dev_put(qdisc->dev);
        kfree((char *) qdisc - qdisc->padded);
 }
 
@@ -508,32 +479,23 @@ static void __qdisc_destroy(struct rcu_head *head)
 
 void qdisc_destroy(struct Qdisc *qdisc)
 {
-       struct list_head cql = LIST_HEAD_INIT(cql);
-       struct Qdisc *cq, *q, *n;
+       struct Qdisc_ops  *ops = qdisc->ops;
 
        if (qdisc->flags & TCQ_F_BUILTIN ||
-               !atomic_dec_and_test(&qdisc->refcnt))
+           !atomic_dec_and_test(&qdisc->refcnt))
                return;
 
-       if (!list_empty(&qdisc->list)) {
-               if (qdisc->ops->cl_ops == NULL)
-                       list_del(&qdisc->list);
-               else
-                       list_move(&qdisc->list, &cql);
-       }
-
-       /* unlink inner qdiscs from dev->qdisc_list immediately */
-       list_for_each_entry(cq, &cql, list)
-               list_for_each_entry_safe(q, n, &qdisc->dev->qdisc_list, list)
-                       if (TC_H_MAJ(q->parent) == TC_H_MAJ(cq->handle)) {
-                               if (q->ops->cl_ops == NULL)
-                                       list_del_init(&q->list);
-                               else
-                                       list_move_tail(&q->list, &cql);
-                       }
-       list_for_each_entry_safe(cq, n, &cql, list)
-               list_del_init(&cq->list);
+       list_del(&qdisc->list);
+#ifdef CONFIG_NET_ESTIMATOR
+       gen_kill_estimator(&qdisc->bstats, &qdisc->rate_est);
+#endif
+       if (ops->reset)
+               ops->reset(qdisc);
+       if (ops->destroy)
+               ops->destroy(qdisc);
 
+       module_put(ops->owner);
+       dev_put(qdisc->dev);
        call_rcu(&qdisc->q_rcu, __qdisc_destroy);
 }
 
@@ -553,15 +515,15 @@ void dev_activate(struct net_device *dev)
                                printk(KERN_INFO "%s: activation failed\n", dev->name);
                                return;
                        }
-                       write_lock_bh(&qdisc_tree_lock);
+                       write_lock(&qdisc_tree_lock);
                        list_add_tail(&qdisc->list, &dev->qdisc_list);
-                       write_unlock_bh(&qdisc_tree_lock);
+                       write_unlock(&qdisc_tree_lock);
                } else {
                        qdisc =  &noqueue_qdisc;
                }
-               write_lock_bh(&qdisc_tree_lock);
+               write_lock(&qdisc_tree_lock);
                dev->qdisc_sleeping = qdisc;
-               write_unlock_bh(&qdisc_tree_lock);
+               write_unlock(&qdisc_tree_lock);
        }
 
        if (!netif_carrier_ok(dev))
@@ -591,17 +553,10 @@ void dev_deactivate(struct net_device *dev)
 
        dev_watchdog_down(dev);
 
-       /* Wait for outstanding dev_queue_xmit calls. */
-       synchronize_rcu();
-
-       /* Wait for outstanding qdisc_run calls. */
-       while (test_bit(__LINK_STATE_QDISC_RUNNING, &dev->state))
+       while (test_bit(__LINK_STATE_SCHED, &dev->state))
                yield();
 
-       if (dev->gso_skb) {
-               kfree_skb(dev->gso_skb);
-               dev->gso_skb = NULL;
-       }
+       spin_unlock_wait(&dev->xmit_lock);
 }
 
 void dev_init_scheduler(struct net_device *dev)
@@ -643,5 +598,6 @@ EXPORT_SYMBOL(qdisc_create_dflt);
 EXPORT_SYMBOL(qdisc_alloc);
 EXPORT_SYMBOL(qdisc_destroy);
 EXPORT_SYMBOL(qdisc_reset);
+EXPORT_SYMBOL(qdisc_restart);
 EXPORT_SYMBOL(qdisc_lock_tree);
 EXPORT_SYMBOL(qdisc_unlock_tree);