net: Expose all network devices in a namespaces in sysfs
[safe/jmp/linux-2.6] / net / core / net-sysfs.c
index 47a6fce..99e7052 100644 (file)
@@ -2,7 +2,7 @@
  * net-sysfs.c - network device class and attributes
  *
  * Copyright (c) 2003 Stephen Hemminger <shemminger@osdl.org>
- * 
+ *
  *     This program is free software; you can redistribute it and/or
  *     modify it under the terms of the GNU General Public License
  *     as published by the Free Software Foundation; either version
  */
 
 #include <linux/capability.h>
-#include <linux/config.h>
 #include <linux/kernel.h>
 #include <linux/netdevice.h>
 #include <linux/if_arp.h>
+#include <linux/slab.h>
+#include <linux/nsproxy.h>
 #include <net/sock.h>
+#include <net/net_namespace.h>
 #include <linux/rtnetlink.h>
 #include <linux/wireless.h>
-#include <net/iw_handler.h>
+#include <linux/vmalloc.h>
+#include <net/wext.h>
 
-#define to_class_dev(obj) container_of(obj,struct class_device,kobj)
-#define to_net_dev(class) container_of(class, struct net_device, class_dev)
+#include "net-sysfs.h"
 
+#ifdef CONFIG_SYSFS
 static const char fmt_hex[] = "%#x\n";
 static const char fmt_long_hex[] = "%#lx\n";
 static const char fmt_dec[] = "%d\n";
 static const char fmt_ulong[] = "%lu\n";
 
-static inline int dev_isalive(const struct net_device *dev) 
+static inline int dev_isalive(const struct net_device *dev)
 {
        return dev->reg_state <= NETREG_REGISTERED;
 }
 
 /* use same locking rules as GIF* ioctl's */
-static ssize_t netdev_show(const struct class_device *cd, char *buf,
+static ssize_t netdev_show(const struct device *dev,
+                          struct device_attribute *attr, char *buf,
                           ssize_t (*format)(const struct net_device *, char *))
 {
-       struct net_device *net = to_net_dev(cd);
+       struct net_device *net = to_net_dev(dev);
        ssize_t ret = -EINVAL;
 
        read_lock(&dev_base_lock);
@@ -53,14 +57,15 @@ static ssize_t format_##field(const struct net_device *net, char *buf)      \
 {                                                                      \
        return sprintf(buf, format_string, net->field);                 \
 }                                                                      \
-static ssize_t show_##field(struct class_device *cd, char *buf)                \
+static ssize_t show_##field(struct device *dev,                                \
+                           struct device_attribute *attr, char *buf)   \
 {                                                                      \
-       return netdev_show(cd, buf, format_##field);                    \
+       return netdev_show(dev, attr, buf, format_##field);             \
 }
 
 
 /* use same locking and permission rules as SIF* ioctl's */
-static ssize_t netdev_store(struct class_device *dev,
+static ssize_t netdev_store(struct device *dev, struct device_attribute *attr,
                            const char *buf, size_t len,
                            int (*set)(struct net_device *, unsigned long))
 {
@@ -76,7 +81,9 @@ static ssize_t netdev_store(struct class_device *dev,
        if (endp == buf)
                goto err;
 
-       rtnl_lock();
+       if (!rtnl_trylock())
+               return restart_syscall();
+
        if (dev_isalive(net)) {
                if ((ret = (*set)(net, new)) == 0)
                        ret = len;
@@ -86,6 +93,7 @@ static ssize_t netdev_store(struct class_device *dev,
        return ret;
 }
 
+NETDEVICE_SHOW(dev_id, fmt_hex);
 NETDEVICE_SHOW(addr_len, fmt_dec);
 NETDEVICE_SHOW(iflink, fmt_dec);
 NETDEVICE_SHOW(ifindex, fmt_dec);
@@ -94,38 +102,30 @@ NETDEVICE_SHOW(type, fmt_dec);
 NETDEVICE_SHOW(link_mode, fmt_dec);
 
 /* use same locking rules as GIFHWADDR ioctl's */
-static ssize_t format_addr(char *buf, const unsigned char *addr, int len)
-{
-       int i;
-       char *cp = buf;
-
-       for (i = 0; i < len; i++)
-               cp += sprintf(cp, "%02x%c", addr[i],
-                             i == (len - 1) ? '\n' : ':');
-       return cp - buf;
-}
-
-static ssize_t show_address(struct class_device *dev, char *buf)
+static ssize_t show_address(struct device *dev, struct device_attribute *attr,
+                           char *buf)
 {
        struct net_device *net = to_net_dev(dev);
        ssize_t ret = -EINVAL;
 
        read_lock(&dev_base_lock);
        if (dev_isalive(net))
-           ret = format_addr(buf, net->dev_addr, net->addr_len);
+               ret = sysfs_format_mac(buf, net->dev_addr, net->addr_len);
        read_unlock(&dev_base_lock);
        return ret;
 }
 
-static ssize_t show_broadcast(struct class_device *dev, char *buf)
+static ssize_t show_broadcast(struct device *dev,
+                           struct device_attribute *attr, char *buf)
 {
        struct net_device *net = to_net_dev(dev);
        if (dev_isalive(net))
-               return format_addr(buf, net->broadcast, net->addr_len);
+               return sysfs_format_mac(buf, net->broadcast, net->addr_len);
        return -EINVAL;
 }
 
-static ssize_t show_carrier(struct class_device *dev, char *buf)
+static ssize_t show_carrier(struct device *dev,
+                           struct device_attribute *attr, char *buf)
 {
        struct net_device *netdev = to_net_dev(dev);
        if (netif_running(netdev)) {
@@ -134,7 +134,50 @@ static ssize_t show_carrier(struct class_device *dev, char *buf)
        return -EINVAL;
 }
 
-static ssize_t show_dormant(struct class_device *dev, char *buf)
+static ssize_t show_speed(struct device *dev,
+                         struct device_attribute *attr, char *buf)
+{
+       struct net_device *netdev = to_net_dev(dev);
+       int ret = -EINVAL;
+
+       if (!rtnl_trylock())
+               return restart_syscall();
+
+       if (netif_running(netdev) &&
+           netdev->ethtool_ops &&
+           netdev->ethtool_ops->get_settings) {
+               struct ethtool_cmd cmd = { ETHTOOL_GSET };
+
+               if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
+                       ret = sprintf(buf, fmt_dec, ethtool_cmd_speed(&cmd));
+       }
+       rtnl_unlock();
+       return ret;
+}
+
+static ssize_t show_duplex(struct device *dev,
+                          struct device_attribute *attr, char *buf)
+{
+       struct net_device *netdev = to_net_dev(dev);
+       int ret = -EINVAL;
+
+       if (!rtnl_trylock())
+               return restart_syscall();
+
+       if (netif_running(netdev) &&
+           netdev->ethtool_ops &&
+           netdev->ethtool_ops->get_settings) {
+               struct ethtool_cmd cmd = { ETHTOOL_GSET };
+
+               if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
+                       ret = sprintf(buf, "%s\n", cmd.duplex ? "full" : "half");
+       }
+       rtnl_unlock();
+       return ret;
+}
+
+static ssize_t show_dormant(struct device *dev,
+                           struct device_attribute *attr, char *buf)
 {
        struct net_device *netdev = to_net_dev(dev);
 
@@ -144,7 +187,7 @@ static ssize_t show_dormant(struct class_device *dev, char *buf)
        return -EINVAL;
 }
 
-static const char *operstates[] = {
+static const char *const operstates[] = {
        "unknown",
        "notpresent", /* currently unused */
        "down",
@@ -154,7 +197,8 @@ static const char *operstates[] = {
        "up"
 };
 
-static ssize_t show_operstate(struct class_device *dev, char *buf)
+static ssize_t show_operstate(struct device *dev,
+                             struct device_attribute *attr, char *buf)
 {
        const struct net_device *netdev = to_net_dev(dev);
        unsigned char operstate;
@@ -179,9 +223,10 @@ static int change_mtu(struct net_device *net, unsigned long new_mtu)
        return dev_set_mtu(net, (int) new_mtu);
 }
 
-static ssize_t store_mtu(struct class_device *dev, const char *buf, size_t len)
+static ssize_t store_mtu(struct device *dev, struct device_attribute *attr,
+                        const char *buf, size_t len)
 {
-       return netdev_store(dev, buf, len, change_mtu);
+       return netdev_store(dev, attr, buf, len, change_mtu);
 }
 
 NETDEVICE_SHOW(flags, fmt_hex);
@@ -191,9 +236,10 @@ static int change_flags(struct net_device *net, unsigned long new_flags)
        return dev_change_flags(net, (unsigned) new_flags);
 }
 
-static ssize_t store_flags(struct class_device *dev, const char *buf, size_t len)
+static ssize_t store_flags(struct device *dev, struct device_attribute *attr,
+                          const char *buf, size_t len)
 {
-       return netdev_store(dev, buf, len, change_flags);
+       return netdev_store(dev, attr, buf, len, change_flags);
 }
 
 NETDEVICE_SHOW(tx_queue_len, fmt_ulong);
@@ -204,26 +250,53 @@ static int change_tx_queue_len(struct net_device *net, unsigned long new_len)
        return 0;
 }
 
-static ssize_t store_tx_queue_len(struct class_device *dev, const char *buf, size_t len)
+static ssize_t store_tx_queue_len(struct device *dev,
+                                 struct device_attribute *attr,
+                                 const char *buf, size_t len)
 {
-       return netdev_store(dev, buf, len, change_tx_queue_len);
+       return netdev_store(dev, attr, buf, len, change_tx_queue_len);
 }
 
-NETDEVICE_SHOW(weight, fmt_dec);
-
-static int change_weight(struct net_device *net, unsigned long new_weight)
+static ssize_t store_ifalias(struct device *dev, struct device_attribute *attr,
+                            const char *buf, size_t len)
 {
-       net->weight = new_weight;
-       return 0;
+       struct net_device *netdev = to_net_dev(dev);
+       size_t count = len;
+       ssize_t ret;
+
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+
+       /* ignore trailing newline */
+       if (len >  0 && buf[len - 1] == '\n')
+               --count;
+
+       if (!rtnl_trylock())
+               return restart_syscall();
+       ret = dev_set_alias(netdev, buf, count);
+       rtnl_unlock();
+
+       return ret < 0 ? ret : len;
 }
 
-static ssize_t store_weight(struct class_device *dev, const char *buf, size_t len)
+static ssize_t show_ifalias(struct device *dev,
+                           struct device_attribute *attr, char *buf)
 {
-       return netdev_store(dev, buf, len, change_weight);
+       const struct net_device *netdev = to_net_dev(dev);
+       ssize_t ret = 0;
+
+       if (!rtnl_trylock())
+               return restart_syscall();
+       if (netdev->ifalias)
+               ret = sprintf(buf, "%s\n", netdev->ifalias);
+       rtnl_unlock();
+       return ret;
 }
 
-static struct class_device_attribute net_class_attributes[] = {
+static struct device_attribute net_class_attributes[] = {
        __ATTR(addr_len, S_IRUGO, show_addr_len, NULL),
+       __ATTR(dev_id, S_IRUGO, show_dev_id, NULL),
+       __ATTR(ifalias, S_IRUGO | S_IWUSR, show_ifalias, store_ifalias),
        __ATTR(iflink, S_IRUGO, show_iflink, NULL),
        __ATTR(ifindex, S_IRUGO, show_ifindex, NULL),
        __ATTR(features, S_IRUGO, show_features, NULL),
@@ -232,46 +305,47 @@ static struct class_device_attribute net_class_attributes[] = {
        __ATTR(address, S_IRUGO, show_address, NULL),
        __ATTR(broadcast, S_IRUGO, show_broadcast, NULL),
        __ATTR(carrier, S_IRUGO, show_carrier, NULL),
+       __ATTR(speed, S_IRUGO, show_speed, NULL),
+       __ATTR(duplex, S_IRUGO, show_duplex, NULL),
        __ATTR(dormant, S_IRUGO, show_dormant, NULL),
        __ATTR(operstate, S_IRUGO, show_operstate, NULL),
        __ATTR(mtu, S_IRUGO | S_IWUSR, show_mtu, store_mtu),
        __ATTR(flags, S_IRUGO | S_IWUSR, show_flags, store_flags),
        __ATTR(tx_queue_len, S_IRUGO | S_IWUSR, show_tx_queue_len,
               store_tx_queue_len),
-       __ATTR(weight, S_IRUGO | S_IWUSR, show_weight, store_weight),
        {}
 };
 
 /* Show a given an attribute in the statistics group */
-static ssize_t netstat_show(const struct class_device *cd, char *buf, 
+static ssize_t netstat_show(const struct device *d,
+                           struct device_attribute *attr, char *buf,
                            unsigned long offset)
 {
-       struct net_device *dev = to_net_dev(cd);
-       struct net_device_stats *stats;
+       struct net_device *dev = to_net_dev(d);
        ssize_t ret = -EINVAL;
 
-       if (offset > sizeof(struct net_device_stats) ||
-           offset % sizeof(unsigned long) != 0)
-               WARN_ON(1);
+       WARN_ON(offset > sizeof(struct net_device_stats) ||
+                       offset % sizeof(unsigned long) != 0);
 
        read_lock(&dev_base_lock);
-       if (dev_isalive(dev) && dev->get_stats &&
-           (stats = (*dev->get_stats)(dev))) 
+       if (dev_isalive(dev)) {
+               const struct net_device_stats *stats = dev_get_stats(dev);
                ret = sprintf(buf, fmt_ulong,
                              *(unsigned long *)(((u8 *) stats) + offset));
-
+       }
        read_unlock(&dev_base_lock);
        return ret;
 }
 
 /* generate a read-only statistics attribute */
 #define NETSTAT_ENTRY(name)                                            \
-static ssize_t show_##name(struct class_device *cd, char *buf)                 \
+static ssize_t show_##name(struct device *d,                           \
+                          struct device_attribute *attr, char *buf)    \
 {                                                                      \
-       return netstat_show(cd, buf,                                    \
+       return netstat_show(d, attr, buf,                               \
                            offsetof(struct net_device_stats, name));   \
 }                                                                      \
-static CLASS_DEVICE_ATTR(name, S_IRUGO, show_##name, NULL)
+static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL)
 
 NETSTAT_ENTRY(rx_packets);
 NETSTAT_ENTRY(tx_packets);
@@ -298,29 +372,29 @@ NETSTAT_ENTRY(rx_compressed);
 NETSTAT_ENTRY(tx_compressed);
 
 static struct attribute *netstat_attrs[] = {
-       &class_device_attr_rx_packets.attr,
-       &class_device_attr_tx_packets.attr,
-       &class_device_attr_rx_bytes.attr,
-       &class_device_attr_tx_bytes.attr,
-       &class_device_attr_rx_errors.attr,
-       &class_device_attr_tx_errors.attr,
-       &class_device_attr_rx_dropped.attr,
-       &class_device_attr_tx_dropped.attr,
-       &class_device_attr_multicast.attr,
-       &class_device_attr_collisions.attr,
-       &class_device_attr_rx_length_errors.attr,
-       &class_device_attr_rx_over_errors.attr,
-       &class_device_attr_rx_crc_errors.attr,
-       &class_device_attr_rx_frame_errors.attr,
-       &class_device_attr_rx_fifo_errors.attr,
-       &class_device_attr_rx_missed_errors.attr,
-       &class_device_attr_tx_aborted_errors.attr,
-       &class_device_attr_tx_carrier_errors.attr,
-       &class_device_attr_tx_fifo_errors.attr,
-       &class_device_attr_tx_heartbeat_errors.attr,
-       &class_device_attr_tx_window_errors.attr,
-       &class_device_attr_rx_compressed.attr,
-       &class_device_attr_tx_compressed.attr,
+       &dev_attr_rx_packets.attr,
+       &dev_attr_tx_packets.attr,
+       &dev_attr_rx_bytes.attr,
+       &dev_attr_tx_bytes.attr,
+       &dev_attr_rx_errors.attr,
+       &dev_attr_tx_errors.attr,
+       &dev_attr_rx_dropped.attr,
+       &dev_attr_tx_dropped.attr,
+       &dev_attr_multicast.attr,
+       &dev_attr_collisions.attr,
+       &dev_attr_rx_length_errors.attr,
+       &dev_attr_rx_over_errors.attr,
+       &dev_attr_rx_crc_errors.attr,
+       &dev_attr_rx_frame_errors.attr,
+       &dev_attr_rx_fifo_errors.attr,
+       &dev_attr_rx_missed_errors.attr,
+       &dev_attr_tx_aborted_errors.attr,
+       &dev_attr_tx_carrier_errors.attr,
+       &dev_attr_tx_fifo_errors.attr,
+       &dev_attr_tx_heartbeat_errors.attr,
+       &dev_attr_tx_window_errors.attr,
+       &dev_attr_rx_compressed.attr,
+       &dev_attr_tx_compressed.attr,
        NULL
 };
 
@@ -330,27 +404,24 @@ static struct attribute_group netstat_group = {
        .attrs  = netstat_attrs,
 };
 
-#ifdef WIRELESS_EXT
+#ifdef CONFIG_WIRELESS_EXT_SYSFS
 /* helper function that does all the locking etc for wireless stats */
-static ssize_t wireless_show(struct class_device *cd, char *buf,
+static ssize_t wireless_show(struct device *d, char *buf,
                             ssize_t (*format)(const struct iw_statistics *,
                                               char *))
 {
-       struct net_device *dev = to_net_dev(cd);
-       const struct iw_statistics *iw = NULL;
+       struct net_device *dev = to_net_dev(d);
+       const struct iw_statistics *iw;
        ssize_t ret = -EINVAL;
-       
-       read_lock(&dev_base_lock);
+
+       if (!rtnl_trylock())
+               return restart_syscall();
        if (dev_isalive(dev)) {
-               if(dev->wireless_handlers &&
-                  dev->wireless_handlers->get_wireless_stats)
-                       iw = dev->wireless_handlers->get_wireless_stats(dev);
-               else if (dev->get_wireless_stats)
-                       iw = dev->get_wireless_stats(dev);
-               if (iw != NULL)
+               iw = get_wireless_stats(dev);
+               if (iw)
                        ret = (*format)(iw, buf);
        }
-       read_unlock(&dev_base_lock);
+       rtnl_unlock();
 
        return ret;
 }
@@ -361,11 +432,12 @@ static ssize_t format_iw_##name(const struct iw_statistics *iw, char *buf) \
 {                                                                      \
        return sprintf(buf, format_string, iw->field);                  \
 }                                                                      \
-static ssize_t show_iw_##name(struct class_device *cd, char *buf)      \
+static ssize_t show_iw_##name(struct device *d,                                \
+                             struct device_attribute *attr, char *buf) \
 {                                                                      \
-       return wireless_show(cd, buf, format_iw_##name);                \
+       return wireless_show(d, buf, format_iw_##name);                 \
 }                                                                      \
-static CLASS_DEVICE_ATTR(name, S_IRUGO, show_iw_##name, NULL)
+static DEVICE_ATTR(name, S_IRUGO, show_iw_##name, NULL)
 
 WIRELESS_SHOW(status, status, fmt_hex);
 WIRELESS_SHOW(link, qual.qual, fmt_dec);
@@ -379,16 +451,16 @@ WIRELESS_SHOW(retries, discard.retries, fmt_dec);
 WIRELESS_SHOW(beacon, miss.beacon, fmt_dec);
 
 static struct attribute *wireless_attrs[] = {
-       &class_device_attr_status.attr,
-       &class_device_attr_link.attr,
-       &class_device_attr_level.attr,
-       &class_device_attr_noise.attr,
-       &class_device_attr_nwid.attr,
-       &class_device_attr_crypt.attr,
-       &class_device_attr_fragment.attr,
-       &class_device_attr_retries.attr,
-       &class_device_attr_misc.attr,
-       &class_device_attr_beacon.attr,
+       &dev_attr_status.attr,
+       &dev_attr_link.attr,
+       &dev_attr_level.attr,
+       &dev_attr_noise.attr,
+       &dev_attr_nwid.attr,
+       &dev_attr_crypt.attr,
+       &dev_attr_fragment.attr,
+       &dev_attr_retries.attr,
+       &dev_attr_misc.attr,
+       &dev_attr_beacon.attr,
        NULL
 };
 
@@ -397,84 +469,471 @@ static struct attribute_group wireless_group = {
        .attrs = wireless_attrs,
 };
 #endif
+#endif /* CONFIG_SYSFS */
+
+#ifdef CONFIG_RPS
+/*
+ * RX queue sysfs structures and functions.
+ */
+struct rx_queue_attribute {
+       struct attribute attr;
+       ssize_t (*show)(struct netdev_rx_queue *queue,
+           struct rx_queue_attribute *attr, char *buf);
+       ssize_t (*store)(struct netdev_rx_queue *queue,
+           struct rx_queue_attribute *attr, const char *buf, size_t len);
+};
+#define to_rx_queue_attr(_attr) container_of(_attr,            \
+    struct rx_queue_attribute, attr)
+
+#define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj)
+
+static ssize_t rx_queue_attr_show(struct kobject *kobj, struct attribute *attr,
+                                 char *buf)
+{
+       struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
+       struct netdev_rx_queue *queue = to_rx_queue(kobj);
+
+       if (!attribute->show)
+               return -EIO;
+
+       return attribute->show(queue, attribute, buf);
+}
+
+static ssize_t rx_queue_attr_store(struct kobject *kobj, struct attribute *attr,
+                                  const char *buf, size_t count)
+{
+       struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
+       struct netdev_rx_queue *queue = to_rx_queue(kobj);
+
+       if (!attribute->store)
+               return -EIO;
+
+       return attribute->store(queue, attribute, buf, count);
+}
+
+static struct sysfs_ops rx_queue_sysfs_ops = {
+       .show = rx_queue_attr_show,
+       .store = rx_queue_attr_store,
+};
+
+static ssize_t show_rps_map(struct netdev_rx_queue *queue,
+                           struct rx_queue_attribute *attribute, char *buf)
+{
+       struct rps_map *map;
+       cpumask_var_t mask;
+       size_t len = 0;
+       int i;
+
+       if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
+               return -ENOMEM;
+
+       rcu_read_lock();
+       map = rcu_dereference(queue->rps_map);
+       if (map)
+               for (i = 0; i < map->len; i++)
+                       cpumask_set_cpu(map->cpus[i], mask);
+
+       len += cpumask_scnprintf(buf + len, PAGE_SIZE, mask);
+       if (PAGE_SIZE - len < 3) {
+               rcu_read_unlock();
+               free_cpumask_var(mask);
+               return -EINVAL;
+       }
+       rcu_read_unlock();
+
+       free_cpumask_var(mask);
+       len += sprintf(buf + len, "\n");
+       return len;
+}
+
+static void rps_map_release(struct rcu_head *rcu)
+{
+       struct rps_map *map = container_of(rcu, struct rps_map, rcu);
+
+       kfree(map);
+}
+
+static ssize_t store_rps_map(struct netdev_rx_queue *queue,
+                     struct rx_queue_attribute *attribute,
+                     const char *buf, size_t len)
+{
+       struct rps_map *old_map, *map;
+       cpumask_var_t mask;
+       int err, cpu, i;
+       static DEFINE_SPINLOCK(rps_map_lock);
+
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+
+       if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+               return -ENOMEM;
+
+       err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
+       if (err) {
+               free_cpumask_var(mask);
+               return err;
+       }
+
+       map = kzalloc(max_t(unsigned,
+           RPS_MAP_SIZE(cpumask_weight(mask)), L1_CACHE_BYTES),
+           GFP_KERNEL);
+       if (!map) {
+               free_cpumask_var(mask);
+               return -ENOMEM;
+       }
+
+       i = 0;
+       for_each_cpu_and(cpu, mask, cpu_online_mask)
+               map->cpus[i++] = cpu;
+
+       if (i)
+               map->len = i;
+       else {
+               kfree(map);
+               map = NULL;
+       }
+
+       spin_lock(&rps_map_lock);
+       old_map = queue->rps_map;
+       rcu_assign_pointer(queue->rps_map, map);
+       spin_unlock(&rps_map_lock);
+
+       if (old_map)
+               call_rcu(&old_map->rcu, rps_map_release);
+
+       free_cpumask_var(mask);
+       return len;
+}
+
+static ssize_t show_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
+                                          struct rx_queue_attribute *attr,
+                                          char *buf)
+{
+       struct rps_dev_flow_table *flow_table;
+       unsigned int val = 0;
+
+       rcu_read_lock();
+       flow_table = rcu_dereference(queue->rps_flow_table);
+       if (flow_table)
+               val = flow_table->mask + 1;
+       rcu_read_unlock();
+
+       return sprintf(buf, "%u\n", val);
+}
+
+static void rps_dev_flow_table_release_work(struct work_struct *work)
+{
+       struct rps_dev_flow_table *table = container_of(work,
+           struct rps_dev_flow_table, free_work);
+
+       vfree(table);
+}
+
+static void rps_dev_flow_table_release(struct rcu_head *rcu)
+{
+       struct rps_dev_flow_table *table = container_of(rcu,
+           struct rps_dev_flow_table, rcu);
+
+       INIT_WORK(&table->free_work, rps_dev_flow_table_release_work);
+       schedule_work(&table->free_work);
+}
+
+static ssize_t store_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
+                                    struct rx_queue_attribute *attr,
+                                    const char *buf, size_t len)
+{
+       unsigned int count;
+       char *endp;
+       struct rps_dev_flow_table *table, *old_table;
+       static DEFINE_SPINLOCK(rps_dev_flow_lock);
+
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+
+       count = simple_strtoul(buf, &endp, 0);
+       if (endp == buf)
+               return -EINVAL;
+
+       if (count) {
+               int i;
+
+               if (count > 1<<30) {
+                       /* Enforce a limit to prevent overflow */
+                       return -EINVAL;
+               }
+               count = roundup_pow_of_two(count);
+               table = vmalloc(RPS_DEV_FLOW_TABLE_SIZE(count));
+               if (!table)
+                       return -ENOMEM;
+
+               table->mask = count - 1;
+               for (i = 0; i < count; i++)
+                       table->flows[i].cpu = RPS_NO_CPU;
+       } else
+               table = NULL;
+
+       spin_lock(&rps_dev_flow_lock);
+       old_table = queue->rps_flow_table;
+       rcu_assign_pointer(queue->rps_flow_table, table);
+       spin_unlock(&rps_dev_flow_lock);
+
+       if (old_table)
+               call_rcu(&old_table->rcu, rps_dev_flow_table_release);
+
+       return len;
+}
+
+static struct rx_queue_attribute rps_cpus_attribute =
+       __ATTR(rps_cpus, S_IRUGO | S_IWUSR, show_rps_map, store_rps_map);
+
+
+static struct rx_queue_attribute rps_dev_flow_table_cnt_attribute =
+       __ATTR(rps_flow_cnt, S_IRUGO | S_IWUSR,
+           show_rps_dev_flow_table_cnt, store_rps_dev_flow_table_cnt);
+
+static struct attribute *rx_queue_default_attrs[] = {
+       &rps_cpus_attribute.attr,
+       &rps_dev_flow_table_cnt_attribute.attr,
+       NULL
+};
+
+static void rx_queue_release(struct kobject *kobj)
+{
+       struct netdev_rx_queue *queue = to_rx_queue(kobj);
+       struct netdev_rx_queue *first = queue->first;
+
+       if (queue->rps_map)
+               call_rcu(&queue->rps_map->rcu, rps_map_release);
+
+       if (queue->rps_flow_table)
+               call_rcu(&queue->rps_flow_table->rcu,
+                   rps_dev_flow_table_release);
+
+       if (atomic_dec_and_test(&first->count))
+               kfree(first);
+}
+
+static struct kobj_type rx_queue_ktype = {
+       .sysfs_ops = &rx_queue_sysfs_ops,
+       .release = rx_queue_release,
+       .default_attrs = rx_queue_default_attrs,
+};
+
+static int rx_queue_add_kobject(struct net_device *net, int index)
+{
+       struct netdev_rx_queue *queue = net->_rx + index;
+       struct kobject *kobj = &queue->kobj;
+       int error = 0;
+
+       kobj->kset = net->queues_kset;
+       error = kobject_init_and_add(kobj, &rx_queue_ktype, NULL,
+           "rx-%u", index);
+       if (error) {
+               kobject_put(kobj);
+               return error;
+       }
+
+       kobject_uevent(kobj, KOBJ_ADD);
+
+       return error;
+}
+
+static int rx_queue_register_kobjects(struct net_device *net)
+{
+       int i;
+       int error = 0;
+
+       net->queues_kset = kset_create_and_add("queues",
+           NULL, &net->dev.kobj);
+       if (!net->queues_kset)
+               return -ENOMEM;
+       for (i = 0; i < net->num_rx_queues; i++) {
+               error = rx_queue_add_kobject(net, i);
+               if (error)
+                       break;
+       }
+
+       if (error)
+               while (--i >= 0)
+                       kobject_put(&net->_rx[i].kobj);
+
+       return error;
+}
+
+static void rx_queue_remove_kobjects(struct net_device *net)
+{
+       int i;
+
+       for (i = 0; i < net->num_rx_queues; i++)
+               kobject_put(&net->_rx[i].kobj);
+       kset_unregister(net->queues_kset);
+}
+#endif /* CONFIG_RPS */
+
+static const void *net_current_ns(void)
+{
+       return current->nsproxy->net_ns;
+}
+
+static const void *net_initial_ns(void)
+{
+       return &init_net;
+}
+
+static const void *net_netlink_ns(struct sock *sk)
+{
+       return sock_net(sk);
+}
+
+static struct kobj_ns_type_operations net_ns_type_operations = {
+       .type = KOBJ_NS_TYPE_NET,
+       .current_ns = net_current_ns,
+       .netlink_ns = net_netlink_ns,
+       .initial_ns = net_initial_ns,
+};
+
+static void net_kobj_ns_exit(struct net *net)
+{
+       kobj_ns_exit(KOBJ_NS_TYPE_NET, net);
+}
+
+static struct pernet_operations kobj_net_ops = {
+       .exit = net_kobj_ns_exit,
+};
+
 
 #ifdef CONFIG_HOTPLUG
-static int netdev_uevent(struct class_device *cd, char **envp,
-                        int num_envp, char *buf, int size)
+static int netdev_uevent(struct device *d, struct kobj_uevent_env *env)
 {
-       struct net_device *dev = to_net_dev(cd);
-       int i = 0;
-       int n;
+       struct net_device *dev = to_net_dev(d);
+       int retval;
 
        /* pass interface to uevent. */
-       envp[i++] = buf;
-       n = snprintf(buf, size, "INTERFACE=%s", dev->name) + 1;
-       buf += n;
-       size -= n;
+       retval = add_uevent_var(env, "INTERFACE=%s", dev->name);
+       if (retval)
+               goto exit;
 
-       if ((size <= 0) || (i >= num_envp))
-               return -ENOMEM;
+       /* pass ifindex to uevent.
+        * ifindex is useful as it won't change (interface name may change)
+        * and is what RtNetlink uses natively. */
+       retval = add_uevent_var(env, "IFINDEX=%d", dev->ifindex);
 
-       envp[i] = NULL;
-       return 0;
+exit:
+       return retval;
 }
 #endif
 
 /*
- *     netdev_release -- destroy and free a dead device. 
- *     Called when last reference to class_device kobject is gone.
+ *     netdev_release -- destroy and free a dead device.
+ *     Called when last reference to device kobject is gone.
  */
-static void netdev_release(struct class_device *cd)
+static void netdev_release(struct device *d)
 {
-       struct net_device *dev 
-               = container_of(cd, struct net_device, class_dev);
+       struct net_device *dev = to_net_dev(d);
 
        BUG_ON(dev->reg_state != NETREG_RELEASED);
 
+       kfree(dev->ifalias);
        kfree((char *)dev - dev->padded);
 }
 
+static const void *net_namespace(struct device *d)
+{
+       struct net_device *dev;
+       dev = container_of(d, struct net_device, dev);
+       return dev_net(dev);
+}
+
 static struct class net_class = {
        .name = "net",
-       .release = netdev_release,
-       .class_dev_attrs = net_class_attributes,
+       .dev_release = netdev_release,
+#ifdef CONFIG_SYSFS
+       .dev_attrs = net_class_attributes,
+#endif /* CONFIG_SYSFS */
 #ifdef CONFIG_HOTPLUG
-       .uevent = netdev_uevent,
+       .dev_uevent = netdev_uevent,
 #endif
+       .ns_type = &net_ns_type_operations,
+       .namespace = net_namespace,
 };
 
-void netdev_unregister_sysfs(struct net_device * net)
+/* Delete sysfs entries but hold kobject reference until after all
+ * netdev references are gone.
+ */
+void netdev_unregister_kobject(struct net_device * net)
 {
-       class_device_del(&(net->class_dev));
+       struct device *dev = &(net->dev);
+
+       kobject_get(&dev->kobj);
+
+#ifdef CONFIG_RPS
+       rx_queue_remove_kobjects(net);
+#endif
+
+       device_del(dev);
 }
 
 /* Create sysfs entries for network device. */
-int netdev_register_sysfs(struct net_device *net)
+int netdev_register_kobject(struct net_device *net)
 {
-       struct class_device *class_dev = &(net->class_dev);
-       struct attribute_group **groups = net->sysfs_groups;
+       struct device *dev = &(net->dev);
+       const struct attribute_group **groups = net->sysfs_groups;
+       int error = 0;
 
-       class_device_initialize(class_dev);
-       class_dev->class = &net_class;
-       class_dev->class_data = net;
-       class_dev->groups = groups;
+       device_initialize(dev);
+       dev->class = &net_class;
+       dev->platform_data = net;
+       dev->groups = groups;
 
-       BUILD_BUG_ON(BUS_ID_SIZE < IFNAMSIZ);
-       strlcpy(class_dev->class_id, net->name, BUS_ID_SIZE);
+       dev_set_name(dev, "%s", net->name);
 
-       if (net->get_stats)
-               *groups++ = &netstat_group;
+#ifdef CONFIG_SYSFS
+       /* Allow for a device specific group */
+       if (*groups)
+               groups++;
 
-#ifdef WIRELESS_EXT
-       if (net->get_wireless_stats
-           || (net->wireless_handlers && net->wireless_handlers->get_wireless_stats))
+       *groups++ = &netstat_group;
+#ifdef CONFIG_WIRELESS_EXT_SYSFS
+       if (net->ieee80211_ptr)
+               *groups++ = &wireless_group;
+#ifdef CONFIG_WIRELESS_EXT
+       else if (net->wireless_handlers)
                *groups++ = &wireless_group;
 #endif
+#endif
+#endif /* CONFIG_SYSFS */
+
+       error = device_add(dev);
+       if (error)
+               return error;
+
+#ifdef CONFIG_RPS
+       error = rx_queue_register_kobjects(net);
+       if (error) {
+               device_del(dev);
+               return error;
+       }
+#endif
+
+       return error;
+}
+
+int netdev_class_create_file(struct class_attribute *class_attr)
+{
+       return class_create_file(&net_class, class_attr);
+}
 
-       return class_device_add(class_dev);
+void netdev_class_remove_file(struct class_attribute *class_attr)
+{
+       class_remove_file(&net_class, class_attr);
 }
 
-int netdev_sysfs_init(void)
+EXPORT_SYMBOL(netdev_class_create_file);
+EXPORT_SYMBOL(netdev_class_remove_file);
+
+int netdev_kobject_init(void)
 {
+       kobj_ns_type_register(&net_ns_type_operations);
+       register_pernet_subsys(&kobj_net_ops);
        return class_register(&net_class);
 }