Staging: Octeon Ethernet: Rewrite transmit code.
authorDavid Daney <ddaney@caviumnetworks.com>
Thu, 7 Jan 2010 19:05:03 +0000 (11:05 -0800)
committerRalf Baechle <ralf@linux-mips.org>
Sat, 27 Feb 2010 11:53:07 +0000 (12:53 +0100)
Stop the queue if too many packets are queued.  Restart it from a high
resolution timer.

Rearrange and simplify locking and SKB freeing code

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
To: linux-mips@linux-mips.org
To: gregkh@suse.de
Patchwork: http://patchwork.linux-mips.org/patch/843/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
drivers/staging/octeon/Kconfig
drivers/staging/octeon/ethernet-tx.c
drivers/staging/octeon/ethernet-tx.h
drivers/staging/octeon/ethernet.c
drivers/staging/octeon/octeon-ethernet.h

index 638ad6b3589107d6f6f146e64eb6f4011d425597..579b8f129e6e5ddd146454b7c09b74e1764f4f69 100644 (file)
@@ -3,6 +3,7 @@ config OCTEON_ETHERNET
        depends on CPU_CAVIUM_OCTEON
        select PHYLIB
        select MDIO_OCTEON
+       select HIGH_RES_TIMERS
        help
          This driver supports the builtin ethernet ports on Cavium
          Networks' products in the Octeon family. This driver supports the
index e5695d964d9aea83dc7cf24d21fd022f1c0d44e2..05b58f8b58fd97bfb5dd20d7048612920e9d572b 100644 (file)
 #define GET_SKBUFF_QOS(skb) 0
 #endif
 
+
+static inline int32_t cvm_oct_adjust_skb_to_free(int32_t skb_to_free, int fau)
+{
+       int32_t undo;
+       undo = skb_to_free > 0 ? MAX_SKB_TO_FREE : skb_to_free + MAX_SKB_TO_FREE;
+       if (undo > 0)
+               cvmx_fau_atomic_add32(fau, -undo);
+       skb_to_free = -skb_to_free > MAX_SKB_TO_FREE ? MAX_SKB_TO_FREE : -skb_to_free;
+       return skb_to_free;
+}
+
+void cvm_oct_free_tx_skbs(struct octeon_ethernet *priv)
+{
+       int32_t skb_to_free;
+       int qos, queues_per_port;
+       queues_per_port = cvmx_pko_get_num_queues(priv->port);
+       /* Drain any pending packets in the free list */
+       for (qos = 0; qos < queues_per_port; qos++) {
+               if (skb_queue_len(&priv->tx_free_list[qos]) == 0)
+                       continue;
+               skb_to_free = cvmx_fau_fetch_and_add32(priv->fau+qos*4, MAX_SKB_TO_FREE);
+               skb_to_free = cvm_oct_adjust_skb_to_free(skb_to_free, priv->fau+qos*4);
+
+               while (skb_to_free > 0) {
+                       dev_kfree_skb_any(skb_dequeue(&priv->tx_free_list[qos]));
+                       skb_to_free--;
+               }
+       }
+}
+
+enum hrtimer_restart cvm_oct_restart_tx(struct hrtimer *timer)
+{
+       struct octeon_ethernet *priv = container_of(timer, struct octeon_ethernet, tx_restart_timer);
+       struct net_device *dev = cvm_oct_device[priv->port];
+
+       cvm_oct_free_tx_skbs(priv);
+
+       if (netif_queue_stopped(dev))
+               netif_wake_queue(dev);
+
+       return HRTIMER_NORESTART;
+}
+
 /**
  * Packet transmit
  *
@@ -77,13 +120,13 @@ int cvm_oct_xmit(struct sk_buff *skb, struct net_device *dev)
        union cvmx_buf_ptr hw_buffer;
        uint64_t old_scratch;
        uint64_t old_scratch2;
-       int dropped;
        int qos;
-       int queue_it_up;
+       enum {QUEUE_CORE, QUEUE_HW, QUEUE_DROP} queue_type;
        struct octeon_ethernet *priv = netdev_priv(dev);
+       struct sk_buff *to_free_list;
        int32_t skb_to_free;
-       int32_t undo;
        int32_t buffers_to_free;
+       unsigned long flags;
 #if REUSE_SKBUFFS_WITHOUT_FREE
        unsigned char *fpa_head;
 #endif
@@ -94,9 +137,6 @@ int cvm_oct_xmit(struct sk_buff *skb, struct net_device *dev)
         */
        prefetch(priv);
 
-       /* Start off assuming no drop */
-       dropped = 0;
-
        /*
         * The check on CVMX_PKO_QUEUES_PER_PORT_* is designed to
         * completely remove "qos" in the event neither interface
@@ -268,9 +308,9 @@ int cvm_oct_xmit(struct sk_buff *skb, struct net_device *dev)
        skb->tc_verd = 0;
 #endif /* CONFIG_NET_CLS_ACT */
 #endif /* CONFIG_NET_SCHED */
+#endif /* REUSE_SKBUFFS_WITHOUT_FREE */
 
 dont_put_skbuff_in_hw:
-#endif /* REUSE_SKBUFFS_WITHOUT_FREE */
 
        /* Check if we can use the hardware checksumming */
        if (USE_HW_TCPUDP_CHECKSUM && (skb->protocol == htons(ETH_P_IP)) &&
@@ -295,18 +335,7 @@ dont_put_skbuff_in_hw:
                    cvmx_fau_fetch_and_add32(FAU_NUM_PACKET_BUFFERS_TO_FREE, 0);
        }
 
-       /*
-        * We try to claim MAX_SKB_TO_FREE buffers.  If there were not
-        * that many available, we have to un-claim (undo) any that
-        * were in excess.  If skb_to_free is positive we will free
-        * that many buffers.
-        */
-       undo = skb_to_free > 0 ?
-               MAX_SKB_TO_FREE : skb_to_free + MAX_SKB_TO_FREE;
-       if (undo > 0)
-               cvmx_fau_atomic_add32(priv->fau+qos*4, -undo);
-       skb_to_free = -skb_to_free > MAX_SKB_TO_FREE ?
-               MAX_SKB_TO_FREE : -skb_to_free;
+       skb_to_free = cvm_oct_adjust_skb_to_free(skb_to_free, priv->fau+qos*4);
 
        /*
         * If we're sending faster than the receive can free them then
@@ -317,60 +346,83 @@ dont_put_skbuff_in_hw:
                pko_command.s.reg0 = priv->fau + qos * 4;
        }
 
-       cvmx_pko_send_packet_prepare(priv->port, priv->queue + qos,
-                                    CVMX_PKO_LOCK_CMD_QUEUE);
+       if (pko_command.s.dontfree)
+               queue_type = QUEUE_CORE;
+       else
+               queue_type = QUEUE_HW;
+
+       spin_lock_irqsave(&priv->tx_free_list[qos].lock, flags);
 
        /* Drop this packet if we have too many already queued to the HW */
-       if (unlikely
-           (skb_queue_len(&priv->tx_free_list[qos]) >= MAX_OUT_QUEUE_DEPTH)) {
-               /*
-                  DEBUGPRINT("%s: Tx dropped. Too many queued\n", dev->name);
-                */
-               dropped = 1;
+       if (unlikely(skb_queue_len(&priv->tx_free_list[qos]) >= MAX_OUT_QUEUE_DEPTH)) {
+               if (dev->tx_queue_len != 0) {
+                       /* Drop the lock when notifying the core.  */
+                       spin_unlock_irqrestore(&priv->tx_free_list[qos].lock, flags);
+                       netif_stop_queue(dev);
+                       hrtimer_start(&priv->tx_restart_timer,
+                                     priv->tx_restart_interval, HRTIMER_MODE_REL);
+                       spin_lock_irqsave(&priv->tx_free_list[qos].lock, flags);
+
+               } else {
+                       /* If not using normal queueing.  */
+                       queue_type = QUEUE_DROP;
+                       goto skip_xmit;
+               }
        }
+
+       cvmx_pko_send_packet_prepare(priv->port, priv->queue + qos,
+                                    CVMX_PKO_LOCK_NONE);
+
        /* Send the packet to the output queue */
-       else if (unlikely
-                (cvmx_pko_send_packet_finish
-                 (priv->port, priv->queue + qos, pko_command, hw_buffer,
-                  CVMX_PKO_LOCK_CMD_QUEUE))) {
+       if (unlikely(cvmx_pko_send_packet_finish(priv->port,
+                                                priv->queue + qos,
+                                                pko_command, hw_buffer,
+                                                CVMX_PKO_LOCK_NONE))) {
                DEBUGPRINT("%s: Failed to send the packet\n", dev->name);
-               dropped = 1;
+               queue_type = QUEUE_DROP;
        }
+skip_xmit:
+       to_free_list = NULL;
 
-       if (USE_ASYNC_IOBDMA) {
-               /* Restore the scratch area */
-               cvmx_scratch_write64(CVMX_SCR_SCRATCH, old_scratch);
-               cvmx_scratch_write64(CVMX_SCR_SCRATCH + 8, old_scratch2);
+       switch (queue_type) {
+       case QUEUE_DROP:
+               skb->next = to_free_list;
+               to_free_list = skb;
+               priv->stats.tx_dropped++;
+               break;
+       case QUEUE_HW:
+               cvmx_fau_atomic_add32(FAU_NUM_PACKET_BUFFERS_TO_FREE, -1);
+               break;
+       case QUEUE_CORE:
+               __skb_queue_tail(&priv->tx_free_list[qos], skb);
+               break;
+       default:
+               BUG();
        }
 
-       queue_it_up = 0;
-       if (unlikely(dropped)) {
-               dev_kfree_skb_any(skb);
-               priv->stats.tx_dropped++;
-       } else {
-               if (USE_SKBUFFS_IN_HW) {
-                       /* Put this packet on the queue to be freed later */
-                       if (pko_command.s.dontfree)
-                               queue_it_up = 1;
-                       else
-                               cvmx_fau_atomic_add32
-                                   (FAU_NUM_PACKET_BUFFERS_TO_FREE, -1);
-               } else {
-                       /* Put this packet on the queue to be freed later */
-                       queue_it_up = 1;
-               }
+       while (skb_to_free > 0) {
+               struct sk_buff *t = __skb_dequeue(&priv->tx_free_list[qos]);
+               t->next = to_free_list;
+               to_free_list = t;
+               skb_to_free--;
        }
 
-       if (queue_it_up) {
-               spin_lock(&priv->tx_free_list[qos].lock);
-               __skb_queue_tail(&priv->tx_free_list[qos], skb);
-               cvm_oct_free_tx_skbs(priv, skb_to_free, qos, 0);
-               spin_unlock(&priv->tx_free_list[qos].lock);
-       } else {
-               cvm_oct_free_tx_skbs(priv, skb_to_free, qos, 1);
+       spin_unlock_irqrestore(&priv->tx_free_list[qos].lock, flags);
+
+       /* Do the actual freeing outside of the lock. */
+       while (to_free_list) {
+               struct sk_buff *t = to_free_list;
+               to_free_list = to_free_list->next;
+               dev_kfree_skb_any(t);
        }
 
-       return 0;
+       if (USE_ASYNC_IOBDMA) {
+               /* Restore the scratch area */
+               cvmx_scratch_write64(CVMX_SCR_SCRATCH, old_scratch);
+               cvmx_scratch_write64(CVMX_SCR_SCRATCH + 8, old_scratch2);
+       }
+
+       return NETDEV_TX_OK;
 }
 
 /**
index c0bebf750bc059acc8d4a8a52b4ab3451e81250d..b628d8c8421d0bb66afa11f995572ba31a666ba5 100644 (file)
@@ -30,28 +30,5 @@ int cvm_oct_xmit_pow(struct sk_buff *skb, struct net_device *dev);
 int cvm_oct_transmit_qos(struct net_device *dev, void *work_queue_entry,
                         int do_free, int qos);
 void cvm_oct_tx_shutdown(struct net_device *dev);
-
-/**
- * Free dead transmit skbs.
- *
- * @priv:              The driver data
- * @skb_to_free:       The number of SKBs to free (free none if negative).
- * @qos:               The queue to free from.
- * @take_lock:         If true, acquire the skb list lock.
- */
-static inline void cvm_oct_free_tx_skbs(struct octeon_ethernet *priv,
-                                       int skb_to_free,
-                                       int qos, int take_lock)
-{
-       /* Free skbuffs not in use by the hardware.  */
-       if (skb_to_free > 0) {
-               if (take_lock)
-                       spin_lock(&priv->tx_free_list[qos].lock);
-               while (skb_to_free > 0) {
-                       dev_kfree_skb(__skb_dequeue(&priv->tx_free_list[qos]));
-                       skb_to_free--;
-               }
-               if (take_lock)
-                       spin_unlock(&priv->tx_free_list[qos].lock);
-       }
-}
+void cvm_oct_free_tx_skbs(struct octeon_ethernet *priv);
+enum hrtimer_restart cvm_oct_restart_tx(struct hrtimer *timer);
index 4e054262a005887777879624ddfe408a0917b8a8..973178a80c937f2a93bc655f6fb029a1102e1495 100644 (file)
@@ -131,50 +131,29 @@ struct net_device *cvm_oct_device[TOTAL_NUMBER_OF_PORTS];
  */
 static void cvm_do_timer(unsigned long arg)
 {
-       int32_t skb_to_free, undo;
-       int queues_per_port;
-       int qos;
-       struct octeon_ethernet *priv;
        static int port;
-
-       if (port >= CVMX_PIP_NUM_INPUT_PORTS) {
+       if (port < CVMX_PIP_NUM_INPUT_PORTS) {
+               if (cvm_oct_device[port]) {
+                       struct octeon_ethernet *priv = netdev_priv(cvm_oct_device[port]);
+                       if (priv->poll)
+                               priv->poll(cvm_oct_device[port]);
+                       cvm_oct_free_tx_skbs(priv);
+                       cvm_oct_device[port]->netdev_ops->ndo_get_stats(cvm_oct_device[port]);
+               }
+               port++;
                /*
-                * All ports have been polled. Start the next
-                * iteration through the ports in one second.
+                * Poll the next port in a 50th of a second.  This
+                * spreads the polling of ports out a little bit.
                 */
+               mod_timer(&cvm_oct_poll_timer, jiffies + HZ/50);
+       } else {
                port = 0;
+               /*
+                * All ports have been polled. Start the next iteration through
+                * the ports in one second.
+                */
                mod_timer(&cvm_oct_poll_timer, jiffies + HZ);
-               return;
        }
-       if (!cvm_oct_device[port])
-               goto out;
-
-       priv = netdev_priv(cvm_oct_device[port]);
-       if (priv->poll)
-               priv->poll(cvm_oct_device[port]);
-
-       queues_per_port = cvmx_pko_get_num_queues(port);
-       /* Drain any pending packets in the free list */
-       for (qos = 0; qos < queues_per_port; qos++) {
-               if (skb_queue_len(&priv->tx_free_list[qos]) == 0)
-                       continue;
-               skb_to_free = cvmx_fau_fetch_and_add32(priv->fau + qos * 4,
-                                                      MAX_SKB_TO_FREE);
-               undo = skb_to_free > 0 ?
-                       MAX_SKB_TO_FREE : skb_to_free + MAX_SKB_TO_FREE;
-               if (undo > 0)
-                       cvmx_fau_atomic_add32(priv->fau+qos*4, -undo);
-               skb_to_free = -skb_to_free > MAX_SKB_TO_FREE ?
-                       MAX_SKB_TO_FREE : -skb_to_free;
-               cvm_oct_free_tx_skbs(priv, skb_to_free, qos, 1);
-       }
-       cvm_oct_device[port]->netdev_ops->ndo_get_stats(cvm_oct_device[port]);
-
-out:
-       port++;
-       /* Poll the next port in a 50th of a second.
-          This spreads the polling of ports out a little bit */
-       mod_timer(&cvm_oct_poll_timer, jiffies + HZ / 50);
 }
 
 /**
@@ -678,6 +657,18 @@ static int __init cvm_oct_init_module(void)
                        /* Initialize the device private structure. */
                        struct octeon_ethernet *priv = netdev_priv(dev);
 
+                       hrtimer_init(&priv->tx_restart_timer,
+                                    CLOCK_MONOTONIC,
+                                    HRTIMER_MODE_REL);
+                       priv->tx_restart_timer.function = cvm_oct_restart_tx;
+
+                       /*
+                        * Default for 10GE 5000nS enough time to
+                        * transmit about 100 64byte packtes.  1GE
+                        * interfaces will get 50000nS below.
+                        */
+                       priv->tx_restart_interval = ktime_set(0, 5000);
+
                        dev->netdev_ops = &cvm_oct_pow_netdev_ops;
                        priv->imode = CVMX_HELPER_INTERFACE_MODE_DISABLED;
                        priv->port = CVMX_PIP_NUM_INPUT_PORTS;
@@ -757,6 +748,7 @@ static int __init cvm_oct_init_module(void)
 
                        case CVMX_HELPER_INTERFACE_MODE_SGMII:
                                dev->netdev_ops = &cvm_oct_sgmii_netdev_ops;
+                               priv->tx_restart_interval = ktime_set(0, 50000);
                                strcpy(dev->name, "eth%d");
                                break;
 
@@ -768,6 +760,7 @@ static int __init cvm_oct_init_module(void)
                        case CVMX_HELPER_INTERFACE_MODE_RGMII:
                        case CVMX_HELPER_INTERFACE_MODE_GMII:
                                dev->netdev_ops = &cvm_oct_rgmii_netdev_ops;
+                               priv->tx_restart_interval = ktime_set(0, 50000);
                                strcpy(dev->name, "eth%d");
                                break;
                        }
index 208da27bc02d6c2a236575847b30501b076129f1..203c6a920af51409241232a068daebed417c5d04 100644 (file)
@@ -31,6 +31,8 @@
 #ifndef OCTEON_ETHERNET_H
 #define OCTEON_ETHERNET_H
 
+#include <linux/hrtimer.h>
+
 /**
  * This is the definition of the Ethernet driver's private
  * driver state stored in netdev_priv(dev).
@@ -57,6 +59,8 @@ struct octeon_ethernet {
        uint64_t link_info;
        /* Called periodically to check link status */
        void (*poll) (struct net_device *dev);
+       struct hrtimer          tx_restart_timer;
+       ktime_t                 tx_restart_interval;
 };
 
 /**