Merge branch 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6.git] / net / core / netpoll.c
index df30feb..06be243 100644 (file)
@@ -9,6 +9,7 @@
  * Copyright (C) 2002  Red Hat, Inc.
  */
 
+#include <linux/moduleparam.h>
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/string.h>
@@ -21,6 +22,7 @@
 #include <linux/delay.h>
 #include <linux/rcupdate.h>
 #include <linux/workqueue.h>
+#include <linux/slab.h>
 #include <net/tcp.h>
 #include <net/udp.h>
 #include <asm/unaligned.h>
@@ -33,7 +35,6 @@
 
 #define MAX_UDP_CHUNK 1460
 #define MAX_SKBS 32
-#define MAX_QUEUE_DEPTH (MAX_SKBS / 2)
 
 static struct sk_buff_head skb_pool;
 
@@ -50,6 +51,9 @@ static atomic_t trapped;
 static void zap_completion_queue(void);
 static void arp_reply(struct sk_buff *skb);
 
+static unsigned int carrier_timeout = 4;
+module_param(carrier_timeout, uint, 0644);
+
 static void queue_process(struct work_struct *work)
 {
        struct netpoll_info *npinfo =
@@ -71,8 +75,7 @@ static void queue_process(struct work_struct *work)
 
                local_irq_save(flags);
                __netif_tx_lock(txq, smp_processor_id());
-               if (netif_tx_queue_stopped(txq) ||
-                   netif_tx_queue_frozen(txq) ||
+               if (netif_tx_queue_frozen_or_stopped(txq) ||
                    ops->ndo_start_xmit(skb, dev) != NETDEV_TX_OK) {
                        skb_queue_head(&npinfo->txq, skb);
                        __netif_tx_unlock(txq);
@@ -174,9 +177,8 @@ static void service_arp_queue(struct netpoll_info *npi)
        }
 }
 
-void netpoll_poll(struct netpoll *np)
+void netpoll_poll_dev(struct net_device *dev)
 {
-       struct net_device *dev = np->dev;
        const struct net_device_ops *ops;
 
        if (!dev || !netif_running(dev))
@@ -191,10 +193,28 @@ void netpoll_poll(struct netpoll *np)
 
        poll_napi(dev);
 
+       if (dev->priv_flags & IFF_SLAVE) {
+               if (dev->npinfo) {
+                       struct net_device *bond_dev = dev->master;
+                       struct sk_buff *skb;
+                       while ((skb = skb_dequeue(&dev->npinfo->arp_tx))) {
+                               skb->dev = bond_dev;
+                               skb_queue_tail(&bond_dev->npinfo->arp_tx, skb);
+                       }
+               }
+       }
+
        service_arp_queue(dev->npinfo);
 
        zap_completion_queue();
 }
+EXPORT_SYMBOL(netpoll_poll_dev);
+
+void netpoll_poll(struct netpoll *np)
+{
+       netpoll_poll_dev(np->dev);
+}
+EXPORT_SYMBOL(netpoll_poll);
 
 static void refill_skbs(void)
 {
@@ -277,12 +297,13 @@ static int netpoll_owner_active(struct net_device *dev)
        return 0;
 }
 
-static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
+void netpoll_send_skb_on_dev(struct netpoll *np, struct sk_buff *skb,
+                            struct net_device *dev)
 {
        int status = NETDEV_TX_BUSY;
        unsigned long tries;
-       struct net_device *dev = np->dev;
        const struct net_device_ops *ops = dev->netdev_ops;
+       /* It is up to the caller to keep npinfo alive. */
        struct netpoll_info *npinfo = np->dev->npinfo;
 
        if (!npinfo || !netif_running(dev) || !netif_device_present(dev)) {
@@ -319,6 +340,11 @@ static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
 
                        udelay(USEC_PER_POLL);
                }
+
+               WARN_ONCE(!irqs_disabled(),
+                       "netpoll_send_skb(): %s enabled interrupts in poll (%pF)\n",
+                       dev->name, ops->ndo_start_xmit);
+
                local_irq_restore(flags);
        }
 
@@ -327,6 +353,7 @@ static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
                schedule_delayed_work(&npinfo->tx_work,0);
        }
 }
+EXPORT_SYMBOL(netpoll_send_skb_on_dev);
 
 void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
 {
@@ -388,6 +415,7 @@ void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
 
        netpoll_send_skb(np, skb);
 }
+EXPORT_SYMBOL(netpoll_send_udp);
 
 static void arp_reply(struct sk_buff *skb)
 {
@@ -398,11 +426,24 @@ static void arp_reply(struct sk_buff *skb)
        __be32 sip, tip;
        unsigned char *sha;
        struct sk_buff *send_skb;
-       struct netpoll *np = NULL;
+       struct netpoll *np, *tmp;
+       unsigned long flags;
+       int hits = 0;
+
+       if (list_empty(&npinfo->rx_np))
+               return;
+
+       /* Before checking the packet, we do some early
+          inspection whether this is interesting at all */
+       spin_lock_irqsave(&npinfo->rx_lock, flags);
+       list_for_each_entry_safe(np, tmp, &npinfo->rx_np, rx) {
+               if (np->dev == skb->dev)
+                       hits++;
+       }
+       spin_unlock_irqrestore(&npinfo->rx_lock, flags);
 
-       if (npinfo->rx_np && npinfo->rx_np->dev == skb->dev)
-               np = npinfo->rx_np;
-       if (!np)
+       /* No netpoll struct is using this dev */
+       if (!hits)
                return;
 
        /* No arp on this interface */
@@ -428,77 +469,91 @@ static void arp_reply(struct sk_buff *skb)
        arp_ptr += skb->dev->addr_len;
        memcpy(&sip, arp_ptr, 4);
        arp_ptr += 4;
-       /* if we actually cared about dst hw addr, it would get copied here */
+       /* If we actually cared about dst hw addr,
+          it would get copied here */
        arp_ptr += skb->dev->addr_len;
        memcpy(&tip, arp_ptr, 4);
 
        /* Should we ignore arp? */
-       if (tip != np->local_ip ||
-           ipv4_is_loopback(tip) || ipv4_is_multicast(tip))
+       if (ipv4_is_loopback(tip) || ipv4_is_multicast(tip))
                return;
 
        size = arp_hdr_len(skb->dev);
-       send_skb = find_skb(np, size + LL_ALLOCATED_SPACE(np->dev),
-                           LL_RESERVED_SPACE(np->dev));
-
-       if (!send_skb)
-               return;
 
-       skb_reset_network_header(send_skb);
-       arp = (struct arphdr *) skb_put(send_skb, size);
-       send_skb->dev = skb->dev;
-       send_skb->protocol = htons(ETH_P_ARP);
+       spin_lock_irqsave(&npinfo->rx_lock, flags);
+       list_for_each_entry_safe(np, tmp, &npinfo->rx_np, rx) {
+               if (tip != np->local_ip)
+                       continue;
 
-       /* Fill the device header for the ARP frame */
-       if (dev_hard_header(send_skb, skb->dev, ptype,
-                           sha, np->dev->dev_addr,
-                           send_skb->len) < 0) {
-               kfree_skb(send_skb);
-               return;
-       }
+               send_skb = find_skb(np, size + LL_ALLOCATED_SPACE(np->dev),
+                                   LL_RESERVED_SPACE(np->dev));
+               if (!send_skb)
+                       continue;
 
-       /*
-        * Fill out the arp protocol part.
-        *
-        * we only support ethernet device type,
-        * which (according to RFC 1390) should always equal 1 (Ethernet).
-        */
+               skb_reset_network_header(send_skb);
+               arp = (struct arphdr *) skb_put(send_skb, size);
+               send_skb->dev = skb->dev;
+               send_skb->protocol = htons(ETH_P_ARP);
 
-       arp->ar_hrd = htons(np->dev->type);
-       arp->ar_pro = htons(ETH_P_IP);
-       arp->ar_hln = np->dev->addr_len;
-       arp->ar_pln = 4;
-       arp->ar_op = htons(type);
+               /* Fill the device header for the ARP frame */
+               if (dev_hard_header(send_skb, skb->dev, ptype,
+                                   sha, np->dev->dev_addr,
+                                   send_skb->len) < 0) {
+                       kfree_skb(send_skb);
+                       continue;
+               }
 
-       arp_ptr=(unsigned char *)(arp + 1);
-       memcpy(arp_ptr, np->dev->dev_addr, np->dev->addr_len);
-       arp_ptr += np->dev->addr_len;
-       memcpy(arp_ptr, &tip, 4);
-       arp_ptr += 4;
-       memcpy(arp_ptr, sha, np->dev->addr_len);
-       arp_ptr += np->dev->addr_len;
-       memcpy(arp_ptr, &sip, 4);
+               /*
+                * Fill out the arp protocol part.
+                *
+                * we only support ethernet device type,
+                * which (according to RFC 1390) should
+                * always equal 1 (Ethernet).
+                */
 
-       netpoll_send_skb(np, send_skb);
+               arp->ar_hrd = htons(np->dev->type);
+               arp->ar_pro = htons(ETH_P_IP);
+               arp->ar_hln = np->dev->addr_len;
+               arp->ar_pln = 4;
+               arp->ar_op = htons(type);
+
+               arp_ptr = (unsigned char *)(arp + 1);
+               memcpy(arp_ptr, np->dev->dev_addr, np->dev->addr_len);
+               arp_ptr += np->dev->addr_len;
+               memcpy(arp_ptr, &tip, 4);
+               arp_ptr += 4;
+               memcpy(arp_ptr, sha, np->dev->addr_len);
+               arp_ptr += np->dev->addr_len;
+               memcpy(arp_ptr, &sip, 4);
+
+               netpoll_send_skb(np, send_skb);
+
+               /* If there are several rx_hooks for the same address,
+                  we're fine by sending a single reply */
+               break;
+       }
+       spin_unlock_irqrestore(&npinfo->rx_lock, flags);
 }
 
 int __netpoll_rx(struct sk_buff *skb)
 {
        int proto, len, ulen;
+       int hits = 0;
        struct iphdr *iph;
        struct udphdr *uh;
-       struct netpoll_info *npi = skb->dev->npinfo;
-       struct netpoll *np = npi->rx_np;
+       struct netpoll_info *npinfo = skb->dev->npinfo;
+       struct netpoll *np, *tmp;
 
-       if (!np)
+       if (list_empty(&npinfo->rx_np))
                goto out;
+
        if (skb->dev->type != ARPHRD_ETHER)
                goto out;
 
        /* check if netpoll clients need ARP */
        if (skb->protocol == htons(ETH_P_ARP) &&
            atomic_read(&trapped)) {
-               skb_queue_tail(&npi->arp_tx, skb);
+               skb_queue_tail(&npinfo->arp_tx, skb);
                return 1;
        }
 
@@ -542,16 +597,23 @@ int __netpoll_rx(struct sk_buff *skb)
                goto out;
        if (checksum_udp(skb, uh, ulen, iph->saddr, iph->daddr))
                goto out;
-       if (np->local_ip && np->local_ip != iph->daddr)
-               goto out;
-       if (np->remote_ip && np->remote_ip != iph->saddr)
-               goto out;
-       if (np->local_port && np->local_port != ntohs(uh->dest))
-               goto out;
 
-       np->rx_hook(np, ntohs(uh->source),
-                   (char *)(uh+1),
-                   ulen - sizeof(struct udphdr));
+       list_for_each_entry_safe(np, tmp, &npinfo->rx_np, rx) {
+               if (np->local_ip && np->local_ip != iph->daddr)
+                       continue;
+               if (np->remote_ip && np->remote_ip != iph->saddr)
+                       continue;
+               if (np->local_port && np->local_port != ntohs(uh->dest))
+                       continue;
+
+               np->rx_hook(np, ntohs(uh->source),
+                              (char *)(uh+1),
+                              ulen - sizeof(struct udphdr));
+               hits++;
+       }
+
+       if (!hits)
+               goto out;
 
        kfree_skb(skb);
        return 1;
@@ -571,7 +633,7 @@ void netpoll_print_options(struct netpoll *np)
                         np->name, np->local_port);
        printk(KERN_INFO "%s: local IP %pI4\n",
                         np->name, &np->local_ip);
-       printk(KERN_INFO "%s: interface %s\n",
+       printk(KERN_INFO "%s: interface '%s'\n",
                         np->name, np->dev_name);
        printk(KERN_INFO "%s: remote port %d\n",
                         np->name, np->remote_port);
@@ -580,6 +642,7 @@ void netpoll_print_options(struct netpoll *np)
        printk(KERN_INFO "%s: remote ethernet address %pM\n",
                         np->name, np->remote_mac);
 }
+EXPORT_SYMBOL(netpoll_print_options);
 
 int netpoll_parse_options(struct netpoll *np, char *opt)
 {
@@ -618,6 +681,9 @@ int netpoll_parse_options(struct netpoll *np, char *opt)
                if ((delim = strchr(cur, '@')) == NULL)
                        goto parse_failed;
                *delim = 0;
+               if (*cur == ' ' || *cur == '\t')
+                       printk(KERN_INFO "%s: warning: whitespace"
+                                       "is not allowed\n", np->name);
                np->remote_port = simple_strtol(cur, NULL, 10);
                cur = delim;
        }
@@ -665,37 +731,37 @@ int netpoll_parse_options(struct netpoll *np, char *opt)
        return 0;
 
  parse_failed:
-       printk(KERN_INFO "%s: couldn't parse config at %s!\n",
+       printk(KERN_INFO "%s: couldn't parse config at '%s'!\n",
               np->name, cur);
        return -1;
 }
+EXPORT_SYMBOL(netpoll_parse_options);
 
-int netpoll_setup(struct netpoll *np)
+int __netpoll_setup(struct netpoll *np)
 {
-       struct net_device *ndev = NULL;
-       struct in_device *in_dev;
+       struct net_device *ndev = np->dev;
        struct netpoll_info *npinfo;
+       const struct net_device_ops *ops;
        unsigned long flags;
        int err;
 
-       if (np->dev_name)
-               ndev = dev_get_by_name(&init_net, np->dev_name);
-       if (!ndev) {
-               printk(KERN_ERR "%s: %s doesn't exist, aborting.\n",
+       if ((ndev->priv_flags & IFF_DISABLE_NETPOLL) ||
+           !ndev->netdev_ops->ndo_poll_controller) {
+               printk(KERN_ERR "%s: %s doesn't support polling, aborting.\n",
                       np->name, np->dev_name);
-               return -ENODEV;
+               err = -ENOTSUPP;
+               goto out;
        }
 
-       np->dev = ndev;
        if (!ndev->npinfo) {
                npinfo = kmalloc(sizeof(*npinfo), GFP_KERNEL);
                if (!npinfo) {
                        err = -ENOMEM;
-                       goto release;
+                       goto out;
                }
 
                npinfo->rx_flags = 0;
-               npinfo->rx_np = NULL;
+               INIT_LIST_HEAD(&npinfo->rx_np);
 
                spin_lock_init(&npinfo->rx_lock);
                skb_queue_head_init(&npinfo->arp_tx);
@@ -703,16 +769,51 @@ int netpoll_setup(struct netpoll *np)
                INIT_DELAYED_WORK(&npinfo->tx_work, queue_process);
 
                atomic_set(&npinfo->refcnt, 1);
+
+               ops = np->dev->netdev_ops;
+               if (ops->ndo_netpoll_setup) {
+                       err = ops->ndo_netpoll_setup(ndev, npinfo);
+                       if (err)
+                               goto free_npinfo;
+               }
        } else {
                npinfo = ndev->npinfo;
                atomic_inc(&npinfo->refcnt);
        }
 
-       if (!ndev->netdev_ops->ndo_poll_controller) {
-               printk(KERN_ERR "%s: %s doesn't support polling, aborting.\n",
+       npinfo->netpoll = np;
+
+       if (np->rx_hook) {
+               spin_lock_irqsave(&npinfo->rx_lock, flags);
+               npinfo->rx_flags |= NETPOLL_RX_ENABLED;
+               list_add_tail(&np->rx, &npinfo->rx_np);
+               spin_unlock_irqrestore(&npinfo->rx_lock, flags);
+       }
+
+       /* last thing to do is link it to the net device structure */
+       rcu_assign_pointer(ndev->npinfo, npinfo);
+
+       return 0;
+
+free_npinfo:
+       kfree(npinfo);
+out:
+       return err;
+}
+EXPORT_SYMBOL_GPL(__netpoll_setup);
+
+int netpoll_setup(struct netpoll *np)
+{
+       struct net_device *ndev = NULL;
+       struct in_device *in_dev;
+       int err;
+
+       if (np->dev_name)
+               ndev = dev_get_by_name(&init_net, np->dev_name);
+       if (!ndev) {
+               printk(KERN_ERR "%s: %s doesn't exist, aborting.\n",
                       np->name, np->dev_name);
-               err = -ENOTSUPP;
-               goto release;
+               return -ENODEV;
        }
 
        if (!netif_running(ndev)) {
@@ -728,11 +829,11 @@ int netpoll_setup(struct netpoll *np)
                if (err) {
                        printk(KERN_ERR "%s: failed to open %s\n",
                               np->name, ndev->name);
-                       goto release;
+                       goto put;
                }
 
                atleast = jiffies + HZ/10;
-               atmost = jiffies + 4*HZ;
+               atmost = jiffies + carrier_timeout * HZ;
                while (!netif_carrier_ok(ndev)) {
                        if (time_after(jiffies, atmost)) {
                                printk(KERN_NOTICE
@@ -765,7 +866,7 @@ int netpoll_setup(struct netpoll *np)
                        printk(KERN_ERR "%s: no IP address for %s, aborting\n",
                               np->name, np->dev_name);
                        err = -EDESTADDRREQ;
-                       goto release;
+                       goto put;
                }
 
                np->local_ip = in_dev->ifa_list->ifa_local;
@@ -773,31 +874,25 @@ int netpoll_setup(struct netpoll *np)
                printk(KERN_INFO "%s: local IP %pI4\n", np->name, &np->local_ip);
        }
 
-       if (np->rx_hook) {
-               spin_lock_irqsave(&npinfo->rx_lock, flags);
-               npinfo->rx_flags |= NETPOLL_RX_ENABLED;
-               npinfo->rx_np = np;
-               spin_unlock_irqrestore(&npinfo->rx_lock, flags);
-       }
+       np->dev = ndev;
 
        /* fill up the skb queue */
        refill_skbs();
 
-       /* last thing to do is link it to the net device structure */
-       ndev->npinfo = npinfo;
+       rtnl_lock();
+       err = __netpoll_setup(np);
+       rtnl_unlock();
 
-       /* avoid racing with NAPI reading npinfo */
-       synchronize_rcu();
+       if (err)
+               goto put;
 
        return 0;
 
- release:
-       if (!ndev->npinfo)
-               kfree(npinfo);
-       np->dev = NULL;
+put:
        dev_put(ndev);
        return err;
 }
+EXPORT_SYMBOL(netpoll_setup);
 
 static int __init netpoll_init(void)
 {
@@ -806,43 +901,65 @@ static int __init netpoll_init(void)
 }
 core_initcall(netpoll_init);
 
-void netpoll_cleanup(struct netpoll *np)
+void __netpoll_cleanup(struct netpoll *np)
 {
        struct netpoll_info *npinfo;
        unsigned long flags;
 
-       if (np->dev) {
-               npinfo = np->dev->npinfo;
-               if (npinfo) {
-                       if (npinfo->rx_np == np) {
-                               spin_lock_irqsave(&npinfo->rx_lock, flags);
-                               npinfo->rx_np = NULL;
-                               npinfo->rx_flags &= ~NETPOLL_RX_ENABLED;
-                               spin_unlock_irqrestore(&npinfo->rx_lock, flags);
-                       }
+       npinfo = np->dev->npinfo;
+       if (!npinfo)
+               return;
+
+       if (!list_empty(&npinfo->rx_np)) {
+               spin_lock_irqsave(&npinfo->rx_lock, flags);
+               list_del(&np->rx);
+               if (list_empty(&npinfo->rx_np))
+                       npinfo->rx_flags &= ~NETPOLL_RX_ENABLED;
+               spin_unlock_irqrestore(&npinfo->rx_lock, flags);
+       }
 
-                       if (atomic_dec_and_test(&npinfo->refcnt)) {
-                               skb_queue_purge(&npinfo->arp_tx);
-                               skb_queue_purge(&npinfo->txq);
-                               cancel_rearming_delayed_work(&npinfo->tx_work);
+       if (atomic_dec_and_test(&npinfo->refcnt)) {
+               const struct net_device_ops *ops;
 
-                               /* clean after last, unfinished work */
-                               __skb_queue_purge(&npinfo->txq);
-                               kfree(npinfo);
-                               np->dev->npinfo = NULL;
-                       }
-               }
+               ops = np->dev->netdev_ops;
+               if (ops->ndo_netpoll_cleanup)
+                       ops->ndo_netpoll_cleanup(np->dev);
 
-               dev_put(np->dev);
+               rcu_assign_pointer(np->dev->npinfo, NULL);
+
+               /* avoid racing with NAPI reading npinfo */
+               synchronize_rcu_bh();
+
+               skb_queue_purge(&npinfo->arp_tx);
+               skb_queue_purge(&npinfo->txq);
+               cancel_delayed_work_sync(&npinfo->tx_work);
+
+               /* clean after last, unfinished work */
+               __skb_queue_purge(&npinfo->txq);
+               kfree(npinfo);
        }
+}
+EXPORT_SYMBOL_GPL(__netpoll_cleanup);
+
+void netpoll_cleanup(struct netpoll *np)
+{
+       if (!np->dev)
+               return;
+
+       rtnl_lock();
+       __netpoll_cleanup(np);
+       rtnl_unlock();
 
+       dev_put(np->dev);
        np->dev = NULL;
 }
+EXPORT_SYMBOL(netpoll_cleanup);
 
 int netpoll_trap(void)
 {
        return atomic_read(&trapped);
 }
+EXPORT_SYMBOL(netpoll_trap);
 
 void netpoll_set_trap(int trap)
 {
@@ -851,12 +968,4 @@ void netpoll_set_trap(int trap)
        else
                atomic_dec(&trapped);
 }
-
 EXPORT_SYMBOL(netpoll_set_trap);
-EXPORT_SYMBOL(netpoll_trap);
-EXPORT_SYMBOL(netpoll_print_options);
-EXPORT_SYMBOL(netpoll_parse_options);
-EXPORT_SYMBOL(netpoll_setup);
-EXPORT_SYMBOL(netpoll_cleanup);
-EXPORT_SYMBOL(netpoll_send_udp);
-EXPORT_SYMBOL(netpoll_poll);