]> nv-tegra.nvidia Code Review - linux-2.6.git/blobdiff - net/core/dev.c
[NET_SCHED]: ingress: switch back to using ingress_lock
[linux-2.6.git] / net / core / dev.c
index 582db646cc54a4bdb360f24daf7251a11855638a..c8f5ea9aea81c81c352b2778269d2525c0ef0b46 100644 (file)
  */
 
 static DEFINE_SPINLOCK(ptype_lock);
-static struct list_head ptype_base[16];        /* 16 way hashed list */
-static struct list_head ptype_all;             /* Taps */
+static struct list_head ptype_base[16] __read_mostly;  /* 16 way hashed list */
+static struct list_head ptype_all __read_mostly;       /* Taps */
 
 #ifdef CONFIG_NET_DMA
 static struct dma_client *net_dma_client;
@@ -817,7 +817,6 @@ static int default_rebuild_header(struct sk_buff *skb)
        return 1;
 }
 
-
 /**
  *     dev_open        - prepare an interface for use.
  *     @dev:   device to open
@@ -1066,18 +1065,18 @@ static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
                           set by sender, so that the second statement is
                           just protection against buggy protocols.
                         */
-                       skb2->mac.raw = skb2->data;
+                       skb_reset_mac_header(skb2);
 
-                       if (skb2->nh.raw < skb2->data ||
-                           skb2->nh.raw > skb2->tail) {
+                       if (skb_network_header(skb2) < skb2->data ||
+                           skb2->network_header > skb2->tail) {
                                if (net_ratelimit())
                                        printk(KERN_CRIT "protocol %04x is "
                                               "buggy, dev %s\n",
                                               skb2->protocol, dev->name);
-                               skb2->nh.raw = skb2->data;
+                               skb_reset_network_header(skb2);
                        }
 
-                       skb2->h.raw = skb2->nh.raw;
+                       skb2->transport_header = skb2->network_header;
                        skb2->pkt_type = PACKET_OUTGOING;
                        ptype->func(skb2, skb->dev, ptype, skb->dev);
                }
@@ -1156,7 +1155,7 @@ EXPORT_SYMBOL(netif_device_attach);
 int skb_checksum_help(struct sk_buff *skb)
 {
        __wsum csum;
-       int ret = 0, offset = skb->h.raw - skb->data;
+       int ret = 0, offset;
 
        if (skb->ip_summed == CHECKSUM_COMPLETE)
                goto out_set_summed;
@@ -1172,15 +1171,16 @@ int skb_checksum_help(struct sk_buff *skb)
                        goto out;
        }
 
+       offset = skb->csum_start - skb_headroom(skb);
        BUG_ON(offset > (int)skb->len);
        csum = skb_checksum(skb, offset, skb->len-offset, 0);
 
-       offset = skb->tail - skb->h.raw;
+       offset = skb_headlen(skb) - offset;
        BUG_ON(offset <= 0);
        BUG_ON(skb->csum_offset + 2 > offset);
 
-       *(__sum16*)(skb->h.raw + skb->csum_offset) = csum_fold(csum);
-
+       *(__sum16 *)(skb->head + skb->csum_start + skb->csum_offset) =
+               csum_fold(csum);
 out_set_summed:
        skb->ip_summed = CHECKSUM_NONE;
 out:
@@ -1206,8 +1206,8 @@ struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features)
 
        BUG_ON(skb_shinfo(skb)->frag_list);
 
-       skb->mac.raw = skb->data;
-       skb->mac_len = skb->nh.raw - skb->data;
+       skb_reset_mac_header(skb);
+       skb->mac_len = skb->network_header - skb->mac_header;
        __skb_pull(skb, skb->mac_len);
 
        if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
@@ -1224,7 +1224,8 @@ struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features)
                                segs = ERR_PTR(err);
                                if (err || skb_gso_ok(skb, features))
                                        break;
-                               __skb_push(skb, skb->data - skb->nh.raw);
+                               __skb_push(skb, (skb->data -
+                                                skb_network_header(skb)));
                        }
                        segs = ptype->gso_segment(skb, features);
                        break;
@@ -1232,7 +1233,7 @@ struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features)
        }
        rcu_read_unlock();
 
-       __skb_push(skb, skb->data - skb->mac.raw);
+       __skb_push(skb, skb->data - skb_mac_header(skb));
 
        return segs;
 }
@@ -1431,12 +1432,16 @@ int dev_queue_xmit(struct sk_buff *skb)
        /* If packet is not checksummed and device does not support
         * checksumming for this protocol, complete checksumming here.
         */
-       if (skb->ip_summed == CHECKSUM_PARTIAL &&
-           (!(dev->features & NETIF_F_GEN_CSUM) &&
-            (!(dev->features & NETIF_F_IP_CSUM) ||
-             skb->protocol != htons(ETH_P_IP))))
-               if (skb_checksum_help(skb))
-                       goto out_kfree_skb;
+       if (skb->ip_summed == CHECKSUM_PARTIAL) {
+               skb_set_transport_header(skb, skb->csum_start -
+                                             skb_headroom(skb));
+
+               if (!(dev->features & NETIF_F_GEN_CSUM) &&
+                   (!(dev->features & NETIF_F_IP_CSUM) ||
+                    skb->protocol != htons(ETH_P_IP)))
+                       if (skb_checksum_help(skb))
+                               goto out_kfree_skb;
+       }
 
 gso:
        spin_lock_prefetch(&dev->queue_lock);
@@ -1532,9 +1537,9 @@ out:
                        Receiver routines
   =======================================================================*/
 
-int netdev_max_backlog = 1000;
-int netdev_budget = 300;
-int weight_p = 64;            /* old backlog weight */
+int netdev_max_backlog __read_mostly = 1000;
+int netdev_budget __read_mostly = 300;
+int weight_p __read_mostly = 64;            /* old backlog weight */
 
 DEFINE_PER_CPU(struct netif_rx_stats, netdev_rx_stat) = { 0, };
 
@@ -1673,40 +1678,46 @@ static void net_tx_action(struct softirq_action *h)
        }
 }
 
-static __inline__ int deliver_skb(struct sk_buff *skb,
-                                 struct packet_type *pt_prev,
-                                 struct net_device *orig_dev)
+static inline int deliver_skb(struct sk_buff *skb,
+                             struct packet_type *pt_prev,
+                             struct net_device *orig_dev)
 {
        atomic_inc(&skb->users);
        return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
 }
 
 #if defined(CONFIG_BRIDGE) || defined (CONFIG_BRIDGE_MODULE)
-int (*br_handle_frame_hook)(struct net_bridge_port *p, struct sk_buff **pskb);
+/* These hooks defined here for ATM */
 struct net_bridge;
 struct net_bridge_fdb_entry *(*br_fdb_get_hook)(struct net_bridge *br,
                                                unsigned char *addr);
-void (*br_fdb_put_hook)(struct net_bridge_fdb_entry *ent);
+void (*br_fdb_put_hook)(struct net_bridge_fdb_entry *ent) __read_mostly;
 
-static __inline__ int handle_bridge(struct sk_buff **pskb,
-                                   struct packet_type **pt_prev, int *ret,
-                                   struct net_device *orig_dev)
+/*
+ * If bridge module is loaded call bridging hook.
+ *  returns NULL if packet was consumed.
+ */
+struct sk_buff *(*br_handle_frame_hook)(struct net_bridge_port *p,
+                                       struct sk_buff *skb) __read_mostly;
+static inline struct sk_buff *handle_bridge(struct sk_buff *skb,
+                                           struct packet_type **pt_prev, int *ret,
+                                           struct net_device *orig_dev)
 {
        struct net_bridge_port *port;
 
-       if ((*pskb)->pkt_type == PACKET_LOOPBACK ||
-           (port = rcu_dereference((*pskb)->dev->br_port)) == NULL)
-               return 0;
+       if (skb->pkt_type == PACKET_LOOPBACK ||
+           (port = rcu_dereference(skb->dev->br_port)) == NULL)
+               return skb;
 
        if (*pt_prev) {
-               *ret = deliver_skb(*pskb, *pt_prev, orig_dev);
+               *ret = deliver_skb(skb, *pt_prev, orig_dev);
                *pt_prev = NULL;
        }
 
-       return br_handle_frame_hook(port, pskb);
+       return br_handle_frame_hook(port, skb);
 }
 #else
-#define handle_bridge(skb, pt_prev, ret, orig_dev)     (0)
+#define handle_bridge(skb, pt_prev, ret, orig_dev)     (skb)
 #endif
 
 #ifdef CONFIG_NET_CLS_ACT
@@ -1736,10 +1747,10 @@ static int ing_filter(struct sk_buff *skb)
 
                skb->tc_verd = SET_TC_AT(skb->tc_verd,AT_INGRESS);
 
-               spin_lock(&dev->queue_lock);
+               spin_lock(&dev->ingress_lock);
                if ((q = dev->qdisc_ingress) != NULL)
                        result = q->enqueue(skb, q);
-               spin_unlock(&dev->queue_lock);
+               spin_unlock(&dev->ingress_lock);
 
        }
 
@@ -1771,8 +1782,9 @@ int netif_receive_skb(struct sk_buff *skb)
 
        __get_cpu_var(netdev_rx_stat).total++;
 
-       skb->h.raw = skb->nh.raw = skb->data;
-       skb->mac_len = skb->nh.raw - skb->mac.raw;
+       skb_reset_network_header(skb);
+       skb_reset_transport_header(skb);
+       skb->mac_len = skb->network_header - skb->mac_header;
 
        pt_prev = NULL;
 
@@ -1812,7 +1824,8 @@ int netif_receive_skb(struct sk_buff *skb)
 ncls:
 #endif
 
-       if (handle_bridge(&skb, &pt_prev, &ret, orig_dev))
+       skb = handle_bridge(skb, &pt_prev, &ret, orig_dev);
+       if (!skb)
                goto out;
 
        type = skb->protocol;
@@ -2065,7 +2078,7 @@ static int dev_ifconf(char __user *arg)
  *     This is invoked by the /proc filesystem handler to display a device
  *     in detail.
  */
-static __inline__ struct net_device *dev_get_idx(loff_t pos)
+static struct net_device *dev_get_idx(loff_t pos)
 {
        struct net_device *dev;
        loff_t i;
@@ -2094,9 +2107,9 @@ void dev_seq_stop(struct seq_file *seq, void *v)
 
 static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev)
 {
-       if (dev->get_stats) {
-               struct net_device_stats *stats = dev->get_stats(dev);
+       struct net_device_stats *stats = dev->get_stats(dev);
 
+       if (stats) {
                seq_printf(seq, "%6s:%8lu %7lu %4lu %4lu %4lu %5lu %10lu %9lu "
                                "%8lu %7lu %4lu %4lu %4lu %5lu %7lu %10lu\n",
                           dev->name, stats->rx_bytes, stats->rx_packets,
@@ -2174,7 +2187,7 @@ static int softnet_seq_show(struct seq_file *seq, void *v)
        return 0;
 }
 
-static struct seq_operations dev_seq_ops = {
+static const struct seq_operations dev_seq_ops = {
        .start = dev_seq_start,
        .next  = dev_seq_next,
        .stop  = dev_seq_stop,
@@ -2194,7 +2207,7 @@ static const struct file_operations dev_seq_fops = {
        .release = seq_release,
 };
 
-static struct seq_operations softnet_seq_ops = {
+static const struct seq_operations softnet_seq_ops = {
        .start = softnet_seq_start,
        .next  = softnet_seq_next,
        .stop  = softnet_seq_stop,
@@ -2214,6 +2227,135 @@ static const struct file_operations softnet_seq_fops = {
        .release = seq_release,
 };
 
+static void *ptype_get_idx(loff_t pos)
+{
+       struct packet_type *pt = NULL;
+       loff_t i = 0;
+       int t;
+
+       list_for_each_entry_rcu(pt, &ptype_all, list) {
+               if (i == pos)
+                       return pt;
+               ++i;
+       }
+
+       for (t = 0; t < 16; t++) {
+               list_for_each_entry_rcu(pt, &ptype_base[t], list) {
+                       if (i == pos)
+                               return pt;
+                       ++i;
+               }
+       }
+       return NULL;
+}
+
+static void *ptype_seq_start(struct seq_file *seq, loff_t *pos)
+{
+       rcu_read_lock();
+       return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN;
+}
+
+static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+       struct packet_type *pt;
+       struct list_head *nxt;
+       int hash;
+
+       ++*pos;
+       if (v == SEQ_START_TOKEN)
+               return ptype_get_idx(0);
+
+       pt = v;
+       nxt = pt->list.next;
+       if (pt->type == htons(ETH_P_ALL)) {
+               if (nxt != &ptype_all)
+                       goto found;
+               hash = 0;
+               nxt = ptype_base[0].next;
+       } else
+               hash = ntohs(pt->type) & 15;
+
+       while (nxt == &ptype_base[hash]) {
+               if (++hash >= 16)
+                       return NULL;
+               nxt = ptype_base[hash].next;
+       }
+found:
+       return list_entry(nxt, struct packet_type, list);
+}
+
+static void ptype_seq_stop(struct seq_file *seq, void *v)
+{
+       rcu_read_unlock();
+}
+
+static void ptype_seq_decode(struct seq_file *seq, void *sym)
+{
+#ifdef CONFIG_KALLSYMS
+       unsigned long offset = 0, symsize;
+       const char *symname;
+       char *modname;
+       char namebuf[128];
+
+       symname = kallsyms_lookup((unsigned long)sym, &symsize, &offset,
+                                 &modname, namebuf);
+
+       if (symname) {
+               char *delim = ":";
+
+               if (!modname)
+                       modname = delim = "";
+               seq_printf(seq, "%s%s%s%s+0x%lx", delim, modname, delim,
+                          symname, offset);
+               return;
+       }
+#endif
+
+       seq_printf(seq, "[%p]", sym);
+}
+
+static int ptype_seq_show(struct seq_file *seq, void *v)
+{
+       struct packet_type *pt = v;
+
+       if (v == SEQ_START_TOKEN)
+               seq_puts(seq, "Type Device      Function\n");
+       else {
+               if (pt->type == htons(ETH_P_ALL))
+                       seq_puts(seq, "ALL ");
+               else
+                       seq_printf(seq, "%04x", ntohs(pt->type));
+
+               seq_printf(seq, " %-8s ",
+                          pt->dev ? pt->dev->name : "");
+               ptype_seq_decode(seq,  pt->func);
+               seq_putc(seq, '\n');
+       }
+
+       return 0;
+}
+
+static const struct seq_operations ptype_seq_ops = {
+       .start = ptype_seq_start,
+       .next  = ptype_seq_next,
+       .stop  = ptype_seq_stop,
+       .show  = ptype_seq_show,
+};
+
+static int ptype_seq_open(struct inode *inode, struct file *file)
+{
+       return seq_open(file, &ptype_seq_ops);
+}
+
+static const struct file_operations ptype_seq_fops = {
+       .owner   = THIS_MODULE,
+       .open    = ptype_seq_open,
+       .read    = seq_read,
+       .llseek  = seq_lseek,
+       .release = seq_release,
+};
+
+
 #ifdef CONFIG_WIRELESS_EXT
 extern int wireless_proc_init(void);
 #else
@@ -2228,6 +2370,9 @@ static int __init dev_proc_init(void)
                goto out;
        if (!proc_net_fops_create("softnet_stat", S_IRUGO, &softnet_seq_fops))
                goto out_dev;
+       if (!proc_net_fops_create("ptype", S_IRUGO, &ptype_seq_fops))
+               goto out_dev2;
+
        if (wireless_proc_init())
                goto out_softnet;
        rc = 0;
@@ -2235,6 +2380,8 @@ out:
        return rc;
 out_softnet:
        proc_net_remove("softnet_stat");
+out_dev2:
+       proc_net_remove("ptype");
 out_dev:
        proc_net_remove("dev");
        goto out;
@@ -2836,7 +2983,7 @@ static int dev_boot_phase = 1;
 static DEFINE_SPINLOCK(net_todo_list_lock);
 static struct list_head net_todo_list = LIST_HEAD_INIT(net_todo_list);
 
-static inline void net_set_todo(struct net_device *dev)
+static void net_set_todo(struct net_device *dev)
 {
        spin_lock(&net_todo_list_lock);
        list_add_tail(&dev->todo_list, &net_todo_list);
@@ -3146,6 +3293,13 @@ out:
        mutex_unlock(&net_todo_run_mutex);
 }
 
+static struct net_device_stats *maybe_internal_stats(struct net_device *dev)
+{
+       if (dev->features & NETIF_F_INTERNAL_STATS)
+               return &dev->stats;
+       return NULL;
+}
+
 /**
  *     alloc_netdev - allocate network device
  *     @sizeof_priv:   size of private data to allocate space for
@@ -3181,6 +3335,7 @@ struct net_device *alloc_netdev(int sizeof_priv, const char *name,
        if (sizeof_priv)
                dev->priv = netdev_priv(dev);
 
+       dev->get_stats = maybe_internal_stats;
        setup(dev);
        strcpy(dev->name, name);
        return dev;