net/core/net-sysfs.c

   1 /*
   2  * net-sysfs.c - network device class and attributes
   3  *
   4  * Copyright (c) 2003 Stephen Hemminger <shemminger@osdl.org>
   5  *
   6  *      This program is free software; you can redistribute it and/or
   7  *      modify it under the terms of the GNU General Public License
   8  *      as published by the Free Software Foundation; either version
   9  *      2 of the License, or (at your option) any later version.
  10  */
  11
  12 #include <linux/capability.h>
  13 #include <linux/kernel.h>
  14 #include <linux/netdevice.h>
  15 #include <net/switchdev.h>
  16 #include <linux/if_arp.h>
  17 #include <linux/slab.h>
  18 #include <linux/nsproxy.h>
  19 #include <net/sock.h>
  20 #include <net/net_namespace.h>
  21 #include <linux/rtnetlink.h>
  22 #include <linux/vmalloc.h>
  23 #include <linux/export.h>
  24 #include <linux/jiffies.h>
  25 #include <linux/pm_runtime.h>
  26 #include <linux/of.h>
  27
  28 #include "net-sysfs.h"
  29
  30 #ifdef CONFIG_SYSFS
  31 static const char fmt_hex[] = "%#x\n";
  32 static const char fmt_long_hex[] = "%#lx\n";
  33 static const char fmt_dec[] = "%d\n";
  34 static const char fmt_udec[] = "%u\n";
  35 static const char fmt_ulong[] = "%lu\n";
  36 static const char fmt_u64[] = "%llu\n";
  37
  38 static inline int dev_isalive(const struct net_device *dev)
  39 {
  40         return dev->reg_state <= NETREG_REGISTERED;
  41 }
  42
  43 /* use same locking rules as GIF* ioctl's */
  44 static ssize_t netdev_show(const struct device *dev,
  45                            struct device_attribute *attr, char *buf,
  46                            ssize_t (*format)(const struct net_device *, char *))
  47 {
  48         struct net_device *ndev = to_net_dev(dev);
  49         ssize_t ret = -EINVAL;
  50
  51         read_lock(&dev_base_lock);
  52         if (dev_isalive(ndev))
  53                 ret = (*format)(ndev, buf);
  54         read_unlock(&dev_base_lock);
  55
  56         return ret;
  57 }
  58
  59 /* generate a show function for simple field */
  60 #define NETDEVICE_SHOW(field, format_string)                            \
  61 static ssize_t format_##field(const struct net_device *dev, char *buf)  \
  62 {                                                                       \
  63         return sprintf(buf, format_string, dev->field);                 \
  64 }                                                                       \
  65 static ssize_t field##_show(struct device *dev,                         \
  66                             struct device_attribute *attr, char *buf)   \
  67 {                                                                       \
  68         return netdev_show(dev, attr, buf, format_##field);             \
  69 }                                                                       \
  70
  71 #define NETDEVICE_SHOW_RO(field, format_string)                         \
  72 NETDEVICE_SHOW(field, format_string);                                   \
  73 static DEVICE_ATTR_RO(field)
  74
  75 #define NETDEVICE_SHOW_RW(field, format_string)                         \
  76 NETDEVICE_SHOW(field, format_string);                                   \
  77 static DEVICE_ATTR_RW(field)
  78
  79 /* use same locking and permission rules as SIF* ioctl's */
  80 static ssize_t netdev_store(struct device *dev, struct device_attribute *attr,
  81                             const char *buf, size_t len,
  82                             int (*set)(struct net_device *, unsigned long))
  83 {
  84         struct net_device *netdev = to_net_dev(dev);
  85         struct net *net = dev_net(netdev);
  86         unsigned long new;
  87         int ret = -EINVAL;
  88
  89         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
  90                 return -EPERM;
  91
  92         ret = kstrtoul(buf, 0, &new);
  93         if (ret)
  94                 goto err;
  95
  96         if (!rtnl_trylock())
  97                 return restart_syscall();
  98
  99         if (dev_isalive(netdev)) {
 100                 if ((ret = (*set)(netdev, new)) == 0)
 101                         ret = len;
 102         }
 103         rtnl_unlock();
 104  err:
 105         return ret;
 106 }
 107
 108 NETDEVICE_SHOW_RO(dev_id, fmt_hex);
 109 NETDEVICE_SHOW_RO(dev_port, fmt_dec);
 110 NETDEVICE_SHOW_RO(addr_assign_type, fmt_dec);
 111 NETDEVICE_SHOW_RO(addr_len, fmt_dec);
 112 NETDEVICE_SHOW_RO(iflink, fmt_dec);
 113 NETDEVICE_SHOW_RO(ifindex, fmt_dec);
 114 NETDEVICE_SHOW_RO(type, fmt_dec);
 115 NETDEVICE_SHOW_RO(link_mode, fmt_dec);
 116
 117 static ssize_t format_name_assign_type(const struct net_device *dev, char *buf)
 118 {
 119         return sprintf(buf, fmt_dec, dev->name_assign_type);
 120 }
 121
 122 static ssize_t name_assign_type_show(struct device *dev,
 123                                      struct device_attribute *attr,
 124                                      char *buf)
 125 {
 126         struct net_device *ndev = to_net_dev(dev);
 127         ssize_t ret = -EINVAL;
 128
 129         if (ndev->name_assign_type != NET_NAME_UNKNOWN)
 130                 ret = netdev_show(dev, attr, buf, format_name_assign_type);
 131
 132         return ret;
 133 }
 134 static DEVICE_ATTR_RO(name_assign_type);
 135
 136 /* use same locking rules as GIFHWADDR ioctl's */
 137 static ssize_t address_show(struct device *dev, struct device_attribute *attr,
 138                             char *buf)
 139 {
 140         struct net_device *ndev = to_net_dev(dev);
 141         ssize_t ret = -EINVAL;
 142
 143         read_lock(&dev_base_lock);
 144         if (dev_isalive(ndev))
 145                 ret = sysfs_format_mac(buf, ndev->dev_addr, ndev->addr_len);
 146         read_unlock(&dev_base_lock);
 147         return ret;
 148 }
 149 static DEVICE_ATTR_RO(address);
 150
 151 static ssize_t broadcast_show(struct device *dev,
 152                               struct device_attribute *attr, char *buf)
 153 {
 154         struct net_device *ndev = to_net_dev(dev);
 155         if (dev_isalive(ndev))
 156                 return sysfs_format_mac(buf, ndev->broadcast, ndev->addr_len);
 157         return -EINVAL;
 158 }
 159 static DEVICE_ATTR_RO(broadcast);
 160
 161 static int change_carrier(struct net_device *dev, unsigned long new_carrier)
 162 {
 163         if (!netif_running(dev))
 164                 return -EINVAL;
 165         return dev_change_carrier(dev, (bool) new_carrier);
 166 }
 167
 168 static ssize_t carrier_store(struct device *dev, struct device_attribute *attr,
 169                              const char *buf, size_t len)
 170 {
 171         return netdev_store(dev, attr, buf, len, change_carrier);
 172 }
 173
 174 static ssize_t carrier_show(struct device *dev,
 175                             struct device_attribute *attr, char *buf)
 176 {
 177         struct net_device *netdev = to_net_dev(dev);
 178         if (netif_running(netdev)) {
 179                 return sprintf(buf, fmt_dec, !!netif_carrier_ok(netdev));
 180         }
 181         return -EINVAL;
 182 }
 183 static DEVICE_ATTR_RW(carrier);
 184
 185 static ssize_t speed_show(struct device *dev,
 186                           struct device_attribute *attr, char *buf)
 187 {
 188         struct net_device *netdev = to_net_dev(dev);
 189         int ret = -EINVAL;
 190
 191         if (!rtnl_trylock())
 192                 return restart_syscall();
 193
 194         if (netif_running(netdev)) {
 195                 struct ethtool_cmd cmd;
 196                 if (!__ethtool_get_settings(netdev, &cmd))
 197                         ret = sprintf(buf, fmt_udec, ethtool_cmd_speed(&cmd));
 198         }
 199         rtnl_unlock();
 200         return ret;
 201 }
 202 static DEVICE_ATTR_RO(speed);
 203
 204 static ssize_t duplex_show(struct device *dev,
 205                            struct device_attribute *attr, char *buf)
 206 {
 207         struct net_device *netdev = to_net_dev(dev);
 208         int ret = -EINVAL;
 209
 210         if (!rtnl_trylock())
 211                 return restart_syscall();
 212
 213         if (netif_running(netdev)) {
 214                 struct ethtool_cmd cmd;
 215                 if (!__ethtool_get_settings(netdev, &cmd)) {
 216                         const char *duplex;
 217                         switch (cmd.duplex) {
 218                         case DUPLEX_HALF:
 219                                 duplex = "half";
 220                                 break;
 221                         case DUPLEX_FULL:
 222                                 duplex = "full";
 223                                 break;
 224                         default:
 225                                 duplex = "unknown";
 226                                 break;
 227                         }
 228                         ret = sprintf(buf, "%s\n", duplex);
 229                 }
 230         }
 231         rtnl_unlock();
 232         return ret;
 233 }
 234 static DEVICE_ATTR_RO(duplex);
 235
 236 static ssize_t dormant_show(struct device *dev,
 237                             struct device_attribute *attr, char *buf)
 238 {
 239         struct net_device *netdev = to_net_dev(dev);
 240
 241         if (netif_running(netdev))
 242                 return sprintf(buf, fmt_dec, !!netif_dormant(netdev));
 243
 244         return -EINVAL;
 245 }
 246 static DEVICE_ATTR_RO(dormant);
 247
 248 static const char *const operstates[] = {
 249         "unknown",
 250         "notpresent", /* currently unused */
 251         "down",
 252         "lowerlayerdown",
 253         "testing", /* currently unused */
 254         "dormant",
 255         "up"
 256 };
 257
 258 static ssize_t operstate_show(struct device *dev,
 259                               struct device_attribute *attr, char *buf)
 260 {
 261         const struct net_device *netdev = to_net_dev(dev);
 262         unsigned char operstate;
 263
 264         read_lock(&dev_base_lock);
 265         operstate = netdev->operstate;
 266         if (!netif_running(netdev))
 267                 operstate = IF_OPER_DOWN;
 268         read_unlock(&dev_base_lock);
 269
 270         if (operstate >= ARRAY_SIZE(operstates))
 271                 return -EINVAL; /* should not happen */
 272
 273         return sprintf(buf, "%s\n", operstates[operstate]);
 274 }
 275 static DEVICE_ATTR_RO(operstate);
 276
 277 static ssize_t carrier_changes_show(struct device *dev,
 278                                     struct device_attribute *attr,
 279                                     char *buf)
 280 {
 281         struct net_device *netdev = to_net_dev(dev);
 282         return sprintf(buf, fmt_dec,
 283                        atomic_read(&netdev->carrier_changes));
 284 }
 285 static DEVICE_ATTR_RO(carrier_changes);
 286
 287 /* read-write attributes */
 288
 289 static int change_mtu(struct net_device *dev, unsigned long new_mtu)
 290 {
 291         return dev_set_mtu(dev, (int) new_mtu);
 292 }
 293
 294 static ssize_t mtu_store(struct device *dev, struct device_attribute *attr,
 295                          const char *buf, size_t len)
 296 {
 297         return netdev_store(dev, attr, buf, len, change_mtu);
 298 }
 299 NETDEVICE_SHOW_RW(mtu, fmt_dec);
 300
 301 static int change_flags(struct net_device *dev, unsigned long new_flags)
 302 {
 303         return dev_change_flags(dev, (unsigned int) new_flags);
 304 }
 305
 306 static ssize_t flags_store(struct device *dev, struct device_attribute *attr,
 307                            const char *buf, size_t len)
 308 {
 309         return netdev_store(dev, attr, buf, len, change_flags);
 310 }
 311 NETDEVICE_SHOW_RW(flags, fmt_hex);
 312
 313 static int change_tx_queue_len(struct net_device *dev, unsigned long new_len)
 314 {
 315         dev->tx_queue_len = new_len;
 316         return 0;
 317 }
 318
 319 static ssize_t tx_queue_len_store(struct device *dev,
 320                                   struct device_attribute *attr,
 321                                   const char *buf, size_t len)
 322 {
 323         if (!capable(CAP_NET_ADMIN))
 324                 return -EPERM;
 325
 326         return netdev_store(dev, attr, buf, len, change_tx_queue_len);
 327 }
 328 NETDEVICE_SHOW_RW(tx_queue_len, fmt_ulong);
 329
 330 static int change_gro_flush_timeout(struct net_device *dev, unsigned long val)
 331 {
 332         dev->gro_flush_timeout = val;
 333         return 0;
 334 }
 335
 336 static ssize_t gro_flush_timeout_store(struct device *dev,
 337                                   struct device_attribute *attr,
 338                                   const char *buf, size_t len)
 339 {
 340         if (!capable(CAP_NET_ADMIN))
 341                 return -EPERM;
 342
 343         return netdev_store(dev, attr, buf, len, change_gro_flush_timeout);
 344 }
 345 NETDEVICE_SHOW_RW(gro_flush_timeout, fmt_ulong);
 346
 347 static ssize_t ifalias_store(struct device *dev, struct device_attribute *attr,
 348                              const char *buf, size_t len)
 349 {
 350         struct net_device *netdev = to_net_dev(dev);
 351         struct net *net = dev_net(netdev);
 352         size_t count = len;
 353         ssize_t ret;
 354
 355         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 356                 return -EPERM;
 357
 358         /* ignore trailing newline */
 359         if (len >  0 && buf[len - 1] == '\n')
 360                 --count;
 361
 362         if (!rtnl_trylock())
 363                 return restart_syscall();
 364         ret = dev_set_alias(netdev, buf, count);
 365         rtnl_unlock();
 366
 367         return ret < 0 ? ret : len;
 368 }
 369
 370 static ssize_t ifalias_show(struct device *dev,
 371                             struct device_attribute *attr, char *buf)
 372 {
 373         const struct net_device *netdev = to_net_dev(dev);
 374         ssize_t ret = 0;
 375
 376         if (!rtnl_trylock())
 377                 return restart_syscall();
 378         if (netdev->ifalias)
 379                 ret = sprintf(buf, "%s\n", netdev->ifalias);
 380         rtnl_unlock();
 381         return ret;
 382 }
 383 static DEVICE_ATTR_RW(ifalias);
 384
 385 static int change_group(struct net_device *dev, unsigned long new_group)
 386 {
 387         dev_set_group(dev, (int) new_group);
 388         return 0;
 389 }
 390
 391 static ssize_t group_store(struct device *dev, struct device_attribute *attr,
 392                            const char *buf, size_t len)
 393 {
 394         return netdev_store(dev, attr, buf, len, change_group);
 395 }
 396 NETDEVICE_SHOW(group, fmt_dec);
 397 static DEVICE_ATTR(netdev_group, S_IRUGO | S_IWUSR, group_show, group_store);
 398
 399 static ssize_t phys_port_id_show(struct device *dev,
 400                                  struct device_attribute *attr, char *buf)
 401 {
 402         struct net_device *netdev = to_net_dev(dev);
 403         ssize_t ret = -EINVAL;
 404
 405         if (!rtnl_trylock())
 406                 return restart_syscall();
 407
 408         if (dev_isalive(netdev)) {
 409                 struct netdev_phys_item_id ppid;
 410
 411                 ret = dev_get_phys_port_id(netdev, &ppid);
 412                 if (!ret)
 413                         ret = sprintf(buf, "%*phN\n", ppid.id_len, ppid.id);
 414         }
 415         rtnl_unlock();
 416
 417         return ret;
 418 }
 419 static DEVICE_ATTR_RO(phys_port_id);
 420
 421 static ssize_t phys_switch_id_show(struct device *dev,
 422                                    struct device_attribute *attr, char *buf)
 423 {
 424         struct net_device *netdev = to_net_dev(dev);
 425         ssize_t ret = -EINVAL;
 426
 427         if (!rtnl_trylock())
 428                 return restart_syscall();
 429
 430         if (dev_isalive(netdev)) {
 431                 struct netdev_phys_item_id ppid;
 432
 433                 ret = netdev_switch_parent_id_get(netdev, &ppid);
 434                 if (!ret)
 435                         ret = sprintf(buf, "%*phN\n", ppid.id_len, ppid.id);
 436         }
 437         rtnl_unlock();
 438
 439         return ret;
 440 }
 441 static DEVICE_ATTR_RO(phys_switch_id);
 442
 443 static struct attribute *net_class_attrs[] = {
 444         &dev_attr_netdev_group.attr,
 445         &dev_attr_type.attr,
 446         &dev_attr_dev_id.attr,
 447         &dev_attr_dev_port.attr,
 448         &dev_attr_iflink.attr,
 449         &dev_attr_ifindex.attr,
 450         &dev_attr_name_assign_type.attr,
 451         &dev_attr_addr_assign_type.attr,
 452         &dev_attr_addr_len.attr,
 453         &dev_attr_link_mode.attr,
 454         &dev_attr_address.attr,
 455         &dev_attr_broadcast.attr,
 456         &dev_attr_speed.attr,
 457         &dev_attr_duplex.attr,
 458         &dev_attr_dormant.attr,
 459         &dev_attr_operstate.attr,
 460         &dev_attr_carrier_changes.attr,
 461         &dev_attr_ifalias.attr,
 462         &dev_attr_carrier.attr,
 463         &dev_attr_mtu.attr,
 464         &dev_attr_flags.attr,
 465         &dev_attr_tx_queue_len.attr,
 466         &dev_attr_gro_flush_timeout.attr,
 467         &dev_attr_phys_port_id.attr,
 468         &dev_attr_phys_switch_id.attr,
 469         NULL,
 470 };
 471 ATTRIBUTE_GROUPS(net_class);
 472
 473 /* Show a given an attribute in the statistics group */
 474 static ssize_t netstat_show(const struct device *d,
 475                             struct device_attribute *attr, char *buf,
 476                             unsigned long offset)
 477 {
 478         struct net_device *dev = to_net_dev(d);
 479         ssize_t ret = -EINVAL;
 480
 481         WARN_ON(offset > sizeof(struct rtnl_link_stats64) ||
 482                         offset % sizeof(u64) != 0);
 483
 484         read_lock(&dev_base_lock);
 485         if (dev_isalive(dev)) {
 486                 struct rtnl_link_stats64 temp;
 487                 const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp);
 488
 489                 ret = sprintf(buf, fmt_u64, *(u64 *)(((u8 *) stats) + offset));
 490         }
 491         read_unlock(&dev_base_lock);
 492         return ret;
 493 }
 494
 495 /* generate a read-only statistics attribute */
 496 #define NETSTAT_ENTRY(name)                                             \
 497 static ssize_t name##_show(struct device *d,                            \
 498                            struct device_attribute *attr, char *buf)    \
 499 {                                                                       \
 500         return netstat_show(d, attr, buf,                               \
 501                             offsetof(struct rtnl_link_stats64, name));  \
 502 }                                                                       \
 503 static DEVICE_ATTR_RO(name)
 504
 505 NETSTAT_ENTRY(rx_packets);
 506 NETSTAT_ENTRY(tx_packets);
 507 NETSTAT_ENTRY(rx_bytes);
 508 NETSTAT_ENTRY(tx_bytes);
 509 NETSTAT_ENTRY(rx_errors);
 510 NETSTAT_ENTRY(tx_errors);
 511 NETSTAT_ENTRY(rx_dropped);
 512 NETSTAT_ENTRY(tx_dropped);
 513 NETSTAT_ENTRY(multicast);
 514 NETSTAT_ENTRY(collisions);
 515 NETSTAT_ENTRY(rx_length_errors);
 516 NETSTAT_ENTRY(rx_over_errors);
 517 NETSTAT_ENTRY(rx_crc_errors);
 518 NETSTAT_ENTRY(rx_frame_errors);
 519 NETSTAT_ENTRY(rx_fifo_errors);
 520 NETSTAT_ENTRY(rx_missed_errors);
 521 NETSTAT_ENTRY(tx_aborted_errors);
 522 NETSTAT_ENTRY(tx_carrier_errors);
 523 NETSTAT_ENTRY(tx_fifo_errors);
 524 NETSTAT_ENTRY(tx_heartbeat_errors);
 525 NETSTAT_ENTRY(tx_window_errors);
 526 NETSTAT_ENTRY(rx_compressed);
 527 NETSTAT_ENTRY(tx_compressed);
 528
 529 static struct attribute *netstat_attrs[] = {
 530         &dev_attr_rx_packets.attr,
 531         &dev_attr_tx_packets.attr,
 532         &dev_attr_rx_bytes.attr,
 533         &dev_attr_tx_bytes.attr,
 534         &dev_attr_rx_errors.attr,
 535         &dev_attr_tx_errors.attr,
 536         &dev_attr_rx_dropped.attr,
 537         &dev_attr_tx_dropped.attr,
 538         &dev_attr_multicast.attr,
 539         &dev_attr_collisions.attr,
 540         &dev_attr_rx_length_errors.attr,
 541         &dev_attr_rx_over_errors.attr,
 542         &dev_attr_rx_crc_errors.attr,
 543         &dev_attr_rx_frame_errors.attr,
 544         &dev_attr_rx_fifo_errors.attr,
 545         &dev_attr_rx_missed_errors.attr,
 546         &dev_attr_tx_aborted_errors.attr,
 547         &dev_attr_tx_carrier_errors.attr,
 548         &dev_attr_tx_fifo_errors.attr,
 549         &dev_attr_tx_heartbeat_errors.attr,
 550         &dev_attr_tx_window_errors.attr,
 551         &dev_attr_rx_compressed.attr,
 552         &dev_attr_tx_compressed.attr,
 553         NULL
 554 };
 555
 556
 557 static struct attribute_group netstat_group = {
 558         .name  = "statistics",
 559         .attrs  = netstat_attrs,
 560 };
 561
 562 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
 563 static struct attribute *wireless_attrs[] = {
 564         NULL
 565 };
 566
 567 static struct attribute_group wireless_group = {
 568         .name = "wireless",
 569         .attrs = wireless_attrs,
 570 };
 571 #endif
 572
 573 #else /* CONFIG_SYSFS */
 574 #define net_class_groups        NULL
 575 #endif /* CONFIG_SYSFS */
 576
 577 #ifdef CONFIG_SYSFS
 578 #define to_rx_queue_attr(_attr) container_of(_attr,             \
 579     struct rx_queue_attribute, attr)
 580
 581 #define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj)
 582
 583 static ssize_t rx_queue_attr_show(struct kobject *kobj, struct attribute *attr,
 584                                   char *buf)
 585 {
 586         struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
 587         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 588
 589         if (!attribute->show)
 590                 return -EIO;
 591
 592         return attribute->show(queue, attribute, buf);
 593 }
 594
 595 static ssize_t rx_queue_attr_store(struct kobject *kobj, struct attribute *attr,
 596                                    const char *buf, size_t count)
 597 {
 598         struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
 599         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 600
 601         if (!attribute->store)
 602                 return -EIO;
 603
 604         return attribute->store(queue, attribute, buf, count);
 605 }
 606
 607 static const struct sysfs_ops rx_queue_sysfs_ops = {
 608         .show = rx_queue_attr_show,
 609         .store = rx_queue_attr_store,
 610 };
 611
 612 #ifdef CONFIG_RPS
 613 static ssize_t show_rps_map(struct netdev_rx_queue *queue,
 614                             struct rx_queue_attribute *attribute, char *buf)
 615 {
 616         struct rps_map *map;
 617         cpumask_var_t mask;
 618         int i, len;
 619
 620         if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
 621                 return -ENOMEM;
 622
 623         rcu_read_lock();
 624         map = rcu_dereference(queue->rps_map);
 625         if (map)
 626                 for (i = 0; i < map->len; i++)
 627                         cpumask_set_cpu(map->cpus[i], mask);
 628
 629         len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
 630         rcu_read_unlock();
 631         free_cpumask_var(mask);
 632
 633         return len < PAGE_SIZE ? len : -EINVAL;
 634 }
 635
 636 static ssize_t store_rps_map(struct netdev_rx_queue *queue,
 637                       struct rx_queue_attribute *attribute,
 638                       const char *buf, size_t len)
 639 {
 640         struct rps_map *old_map, *map;
 641         cpumask_var_t mask;
 642         int err, cpu, i;
 643         static DEFINE_SPINLOCK(rps_map_lock);
 644
 645         if (!capable(CAP_NET_ADMIN))
 646                 return -EPERM;
 647
 648         if (!alloc_cpumask_var(&mask, GFP_KERNEL))
 649                 return -ENOMEM;
 650
 651         err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
 652         if (err) {
 653                 free_cpumask_var(mask);
 654                 return err;
 655         }
 656
 657         map = kzalloc(max_t(unsigned int,
 658             RPS_MAP_SIZE(cpumask_weight(mask)), L1_CACHE_BYTES),
 659             GFP_KERNEL);
 660         if (!map) {
 661                 free_cpumask_var(mask);
 662                 return -ENOMEM;
 663         }
 664
 665         i = 0;
 666         for_each_cpu_and(cpu, mask, cpu_online_mask)
 667                 map->cpus[i++] = cpu;
 668
 669         if (i)
 670                 map->len = i;
 671         else {
 672                 kfree(map);
 673                 map = NULL;
 674         }
 675
 676         spin_lock(&rps_map_lock);
 677         old_map = rcu_dereference_protected(queue->rps_map,
 678                                             lockdep_is_held(&rps_map_lock));
 679         rcu_assign_pointer(queue->rps_map, map);
 680         spin_unlock(&rps_map_lock);
 681
 682         if (map)
 683                 static_key_slow_inc(&rps_needed);
 684         if (old_map) {
 685                 kfree_rcu(old_map, rcu);
 686                 static_key_slow_dec(&rps_needed);
 687         }
 688         free_cpumask_var(mask);
 689         return len;
 690 }
 691
 692 static ssize_t show_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
 693                                            struct rx_queue_attribute *attr,
 694                                            char *buf)
 695 {
 696         struct rps_dev_flow_table *flow_table;
 697         unsigned long val = 0;
 698
 699         rcu_read_lock();
 700         flow_table = rcu_dereference(queue->rps_flow_table);
 701         if (flow_table)
 702                 val = (unsigned long)flow_table->mask + 1;
 703         rcu_read_unlock();
 704
 705         return sprintf(buf, "%lu\n", val);
 706 }
 707
 708 static void rps_dev_flow_table_release(struct rcu_head *rcu)
 709 {
 710         struct rps_dev_flow_table *table = container_of(rcu,
 711             struct rps_dev_flow_table, rcu);
 712         vfree(table);
 713 }
 714
 715 static ssize_t store_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
 716                                      struct rx_queue_attribute *attr,
 717                                      const char *buf, size_t len)
 718 {
 719         unsigned long mask, count;
 720         struct rps_dev_flow_table *table, *old_table;
 721         static DEFINE_SPINLOCK(rps_dev_flow_lock);
 722         int rc;
 723
 724         if (!capable(CAP_NET_ADMIN))
 725                 return -EPERM;
 726
 727         rc = kstrtoul(buf, 0, &count);
 728         if (rc < 0)
 729                 return rc;
 730
 731         if (count) {
 732                 mask = count - 1;
 733                 /* mask = roundup_pow_of_two(count) - 1;
 734                  * without overflows...
 735                  */
 736                 while ((mask | (mask >> 1)) != mask)
 737                         mask |= (mask >> 1);
 738                 /* On 64 bit arches, must check mask fits in table->mask (u32),
 739                  * and on 32bit arches, must check
 740                  * RPS_DEV_FLOW_TABLE_SIZE(mask + 1) doesn't overflow.
 741                  */
 742 #if BITS_PER_LONG > 32
 743                 if (mask > (unsigned long)(u32)mask)
 744                         return -EINVAL;
 745 #else
 746                 if (mask > (ULONG_MAX - RPS_DEV_FLOW_TABLE_SIZE(1))
 747                                 / sizeof(struct rps_dev_flow)) {
 748                         /* Enforce a limit to prevent overflow */
 749                         return -EINVAL;
 750                 }
 751 #endif
 752                 table = vmalloc(RPS_DEV_FLOW_TABLE_SIZE(mask + 1));
 753                 if (!table)
 754                         return -ENOMEM;
 755
 756                 table->mask = mask;
 757                 for (count = 0; count <= mask; count++)
 758                         table->flows[count].cpu = RPS_NO_CPU;
 759         } else
 760                 table = NULL;
 761
 762         spin_lock(&rps_dev_flow_lock);
 763         old_table = rcu_dereference_protected(queue->rps_flow_table,
 764                                               lockdep_is_held(&rps_dev_flow_lock));
 765         rcu_assign_pointer(queue->rps_flow_table, table);
 766         spin_unlock(&rps_dev_flow_lock);
 767
 768         if (old_table)
 769                 call_rcu(&old_table->rcu, rps_dev_flow_table_release);
 770
 771         return len;
 772 }
 773
 774 static struct rx_queue_attribute rps_cpus_attribute =
 775         __ATTR(rps_cpus, S_IRUGO | S_IWUSR, show_rps_map, store_rps_map);
 776
 777
 778 static struct rx_queue_attribute rps_dev_flow_table_cnt_attribute =
 779         __ATTR(rps_flow_cnt, S_IRUGO | S_IWUSR,
 780             show_rps_dev_flow_table_cnt, store_rps_dev_flow_table_cnt);
 781 #endif /* CONFIG_RPS */
 782
 783 static struct attribute *rx_queue_default_attrs[] = {
 784 #ifdef CONFIG_RPS
 785         &rps_cpus_attribute.attr,
 786         &rps_dev_flow_table_cnt_attribute.attr,
 787 #endif
 788         NULL
 789 };
 790
 791 static void rx_queue_release(struct kobject *kobj)
 792 {
 793         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 794 #ifdef CONFIG_RPS
 795         struct rps_map *map;
 796         struct rps_dev_flow_table *flow_table;
 797
 798
 799         map = rcu_dereference_protected(queue->rps_map, 1);
 800         if (map) {
 801                 RCU_INIT_POINTER(queue->rps_map, NULL);
 802                 kfree_rcu(map, rcu);
 803         }
 804
 805         flow_table = rcu_dereference_protected(queue->rps_flow_table, 1);
 806         if (flow_table) {
 807                 RCU_INIT_POINTER(queue->rps_flow_table, NULL);
 808                 call_rcu(&flow_table->rcu, rps_dev_flow_table_release);
 809         }
 810 #endif
 811
 812         memset(kobj, 0, sizeof(*kobj));
 813         dev_put(queue->dev);
 814 }
 815
 816 static const void *rx_queue_namespace(struct kobject *kobj)
 817 {
 818         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 819         struct device *dev = &queue->dev->dev;
 820         const void *ns = NULL;
 821
 822         if (dev->class && dev->class->ns_type)
 823                 ns = dev->class->namespace(dev);
 824
 825         return ns;
 826 }
 827
 828 static struct kobj_type rx_queue_ktype = {
 829         .sysfs_ops = &rx_queue_sysfs_ops,
 830         .release = rx_queue_release,
 831         .default_attrs = rx_queue_default_attrs,
 832         .namespace = rx_queue_namespace
 833 };
 834
 835 static int rx_queue_add_kobject(struct net_device *dev, int index)
 836 {
 837         struct netdev_rx_queue *queue = dev->_rx + index;
 838         struct kobject *kobj = &queue->kobj;
 839         int error = 0;
 840
 841         kobj->kset = dev->queues_kset;
 842         error = kobject_init_and_add(kobj, &rx_queue_ktype, NULL,
 843             "rx-%u", index);
 844         if (error)
 845                 goto exit;
 846
 847         if (dev->sysfs_rx_queue_group) {
 848                 error = sysfs_create_group(kobj, dev->sysfs_rx_queue_group);
 849                 if (error)
 850                         goto exit;
 851         }
 852
 853         kobject_uevent(kobj, KOBJ_ADD);
 854         dev_hold(queue->dev);
 855
 856         return error;
 857 exit:
 858         kobject_put(kobj);
 859         return error;
 860 }
 861 #endif /* CONFIG_SYSFS */
 862
 863 int
 864 net_rx_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
 865 {
 866 #ifdef CONFIG_SYSFS
 867         int i;
 868         int error = 0;
 869
 870 #ifndef CONFIG_RPS
 871         if (!dev->sysfs_rx_queue_group)
 872                 return 0;
 873 #endif
 874         for (i = old_num; i < new_num; i++) {
 875                 error = rx_queue_add_kobject(dev, i);
 876                 if (error) {
 877                         new_num = old_num;
 878                         break;
 879                 }
 880         }
 881
 882         while (--i >= new_num) {
 883                 if (dev->sysfs_rx_queue_group)
 884                         sysfs_remove_group(&dev->_rx[i].kobj,
 885                                            dev->sysfs_rx_queue_group);
 886                 kobject_put(&dev->_rx[i].kobj);
 887         }
 888
 889         return error;
 890 #else
 891         return 0;
 892 #endif
 893 }
 894
 895 #ifdef CONFIG_SYSFS
 896 /*
 897  * netdev_queue sysfs structures and functions.
 898  */
 899 struct netdev_queue_attribute {
 900         struct attribute attr;
 901         ssize_t (*show)(struct netdev_queue *queue,
 902             struct netdev_queue_attribute *attr, char *buf);
 903         ssize_t (*store)(struct netdev_queue *queue,
 904             struct netdev_queue_attribute *attr, const char *buf, size_t len);
 905 };
 906 #define to_netdev_queue_attr(_attr) container_of(_attr,         \
 907     struct netdev_queue_attribute, attr)
 908
 909 #define to_netdev_queue(obj) container_of(obj, struct netdev_queue, kobj)
 910
 911 static ssize_t netdev_queue_attr_show(struct kobject *kobj,
 912                                       struct attribute *attr, char *buf)
 913 {
 914         struct netdev_queue_attribute *attribute = to_netdev_queue_attr(attr);
 915         struct netdev_queue *queue = to_netdev_queue(kobj);
 916
 917         if (!attribute->show)
 918                 return -EIO;
 919
 920         return attribute->show(queue, attribute, buf);
 921 }
 922
 923 static ssize_t netdev_queue_attr_store(struct kobject *kobj,
 924                                        struct attribute *attr,
 925                                        const char *buf, size_t count)
 926 {
 927         struct netdev_queue_attribute *attribute = to_netdev_queue_attr(attr);
 928         struct netdev_queue *queue = to_netdev_queue(kobj);
 929
 930         if (!attribute->store)
 931                 return -EIO;
 932
 933         return attribute->store(queue, attribute, buf, count);
 934 }
 935
 936 static const struct sysfs_ops netdev_queue_sysfs_ops = {
 937         .show = netdev_queue_attr_show,
 938         .store = netdev_queue_attr_store,
 939 };
 940
 941 static ssize_t show_trans_timeout(struct netdev_queue *queue,
 942                                   struct netdev_queue_attribute *attribute,
 943                                   char *buf)
 944 {
 945         unsigned long trans_timeout;
 946
 947         spin_lock_irq(&queue->_xmit_lock);
 948         trans_timeout = queue->trans_timeout;
 949         spin_unlock_irq(&queue->_xmit_lock);
 950
 951         return sprintf(buf, "%lu", trans_timeout);
 952 }
 953
 954 static struct netdev_queue_attribute queue_trans_timeout =
 955         __ATTR(tx_timeout, S_IRUGO, show_trans_timeout, NULL);
 956
 957 #ifdef CONFIG_BQL
 958 /*
 959  * Byte queue limits sysfs structures and functions.
 960  */
 961 static ssize_t bql_show(char *buf, unsigned int value)
 962 {
 963         return sprintf(buf, "%u\n", value);
 964 }
 965
 966 static ssize_t bql_set(const char *buf, const size_t count,
 967                        unsigned int *pvalue)
 968 {
 969         unsigned int value;
 970         int err;
 971
 972         if (!strcmp(buf, "max") || !strcmp(buf, "max\n"))
 973                 value = DQL_MAX_LIMIT;
 974         else {
 975                 err = kstrtouint(buf, 10, &value);
 976                 if (err < 0)
 977                         return err;
 978                 if (value > DQL_MAX_LIMIT)
 979                         return -EINVAL;
 980         }
 981
 982         *pvalue = value;
 983
 984         return count;
 985 }
 986
 987 static ssize_t bql_show_hold_time(struct netdev_queue *queue,
 988                                   struct netdev_queue_attribute *attr,
 989                                   char *buf)
 990 {
 991         struct dql *dql = &queue->dql;
 992
 993         return sprintf(buf, "%u\n", jiffies_to_msecs(dql->slack_hold_time));
 994 }
 995
 996 static ssize_t bql_set_hold_time(struct netdev_queue *queue,
 997                                  struct netdev_queue_attribute *attribute,
 998                                  const char *buf, size_t len)
 999 {
1000         struct dql *dql = &queue->dql;
1001         unsigned int value;
1002         int err;
1003
1004         err = kstrtouint(buf, 10, &value);
1005         if (err < 0)
1006                 return err;
1007
1008         dql->slack_hold_time = msecs_to_jiffies(value);
1009
1010         return len;
1011 }
1012
1013 static struct netdev_queue_attribute bql_hold_time_attribute =
1014         __ATTR(hold_time, S_IRUGO | S_IWUSR, bql_show_hold_time,
1015             bql_set_hold_time);
1016
1017 static ssize_t bql_show_inflight(struct netdev_queue *queue,
1018                                  struct netdev_queue_attribute *attr,
1019                                  char *buf)
1020 {
1021         struct dql *dql = &queue->dql;
1022
1023         return sprintf(buf, "%u\n", dql->num_queued - dql->num_completed);
1024 }
1025
1026 static struct netdev_queue_attribute bql_inflight_attribute =
1027         __ATTR(inflight, S_IRUGO, bql_show_inflight, NULL);
1028
1029 #define BQL_ATTR(NAME, FIELD)                                           \
1030 static ssize_t bql_show_ ## NAME(struct netdev_queue *queue,            \
1031                                  struct netdev_queue_attribute *attr,   \
1032                                  char *buf)                             \
1033 {                                                                       \
1034         return bql_show(buf, queue->dql.FIELD);                         \
1035 }                                                                       \
1036                                                                         \
1037 static ssize_t bql_set_ ## NAME(struct netdev_queue *queue,             \
1038                                 struct netdev_queue_attribute *attr,    \
1039                                 const char *buf, size_t len)            \
1040 {                                                                       \
1041         return bql_set(buf, len, &queue->dql.FIELD);                    \
1042 }                                                                       \
1043                                                                         \
1044 static struct netdev_queue_attribute bql_ ## NAME ## _attribute =       \
1045         __ATTR(NAME, S_IRUGO | S_IWUSR, bql_show_ ## NAME,              \
1046             bql_set_ ## NAME);
1047
1048 BQL_ATTR(limit, limit)
1049 BQL_ATTR(limit_max, max_limit)
1050 BQL_ATTR(limit_min, min_limit)
1051
1052 static struct attribute *dql_attrs[] = {
1053         &bql_limit_attribute.attr,
1054         &bql_limit_max_attribute.attr,
1055         &bql_limit_min_attribute.attr,
1056         &bql_hold_time_attribute.attr,
1057         &bql_inflight_attribute.attr,
1058         NULL
1059 };
1060
1061 static struct attribute_group dql_group = {
1062         .name  = "byte_queue_limits",
1063         .attrs  = dql_attrs,
1064 };
1065 #endif /* CONFIG_BQL */
1066
1067 #ifdef CONFIG_XPS
1068 static unsigned int get_netdev_queue_index(struct netdev_queue *queue)
1069 {
1070         struct net_device *dev = queue->dev;
1071         unsigned int i;
1072
1073         i = queue - dev->_tx;
1074         BUG_ON(i >= dev->num_tx_queues);
1075
1076         return i;
1077 }
1078
1079
1080 static ssize_t show_xps_map(struct netdev_queue *queue,
1081                             struct netdev_queue_attribute *attribute, char *buf)
1082 {
1083         struct net_device *dev = queue->dev;
1084         struct xps_dev_maps *dev_maps;
1085         cpumask_var_t mask;
1086         unsigned long index;
1087         int i, len;
1088
1089         if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
1090                 return -ENOMEM;
1091
1092         index = get_netdev_queue_index(queue);
1093
1094         rcu_read_lock();
1095         dev_maps = rcu_dereference(dev->xps_maps);
1096         if (dev_maps) {
1097                 for_each_possible_cpu(i) {
1098                         struct xps_map *map =
1099                             rcu_dereference(dev_maps->cpu_map[i]);
1100                         if (map) {
1101                                 int j;
1102                                 for (j = 0; j < map->len; j++) {
1103                                         if (map->queues[j] == index) {
1104                                                 cpumask_set_cpu(i, mask);
1105                                                 break;
1106                                         }
1107                                 }
1108                         }
1109                 }
1110         }
1111         rcu_read_unlock();
1112
1113         len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
1114         free_cpumask_var(mask);
1115         return len < PAGE_SIZE ? len : -EINVAL;
1116 }
1117
1118 static ssize_t store_xps_map(struct netdev_queue *queue,
1119                       struct netdev_queue_attribute *attribute,
1120                       const char *buf, size_t len)
1121 {
1122         struct net_device *dev = queue->dev;
1123         unsigned long index;
1124         cpumask_var_t mask;
1125         int err;
1126
1127         if (!capable(CAP_NET_ADMIN))
1128                 return -EPERM;
1129
1130         if (!alloc_cpumask_var(&mask, GFP_KERNEL))
1131                 return -ENOMEM;
1132
1133         index = get_netdev_queue_index(queue);
1134
1135         err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
1136         if (err) {
1137                 free_cpumask_var(mask);
1138                 return err;
1139         }
1140
1141         err = netif_set_xps_queue(dev, mask, index);
1142
1143         free_cpumask_var(mask);
1144
1145         return err ? : len;
1146 }
1147
1148 static struct netdev_queue_attribute xps_cpus_attribute =
1149     __ATTR(xps_cpus, S_IRUGO | S_IWUSR, show_xps_map, store_xps_map);
1150 #endif /* CONFIG_XPS */
1151
1152 static struct attribute *netdev_queue_default_attrs[] = {
1153         &queue_trans_timeout.attr,
1154 #ifdef CONFIG_XPS
1155         &xps_cpus_attribute.attr,
1156 #endif
1157         NULL
1158 };
1159
1160 static void netdev_queue_release(struct kobject *kobj)
1161 {
1162         struct netdev_queue *queue = to_netdev_queue(kobj);
1163
1164         memset(kobj, 0, sizeof(*kobj));
1165         dev_put(queue->dev);
1166 }
1167
1168 static const void *netdev_queue_namespace(struct kobject *kobj)
1169 {
1170         struct netdev_queue *queue = to_netdev_queue(kobj);
1171         struct device *dev = &queue->dev->dev;
1172         const void *ns = NULL;
1173
1174         if (dev->class && dev->class->ns_type)
1175                 ns = dev->class->namespace(dev);
1176
1177         return ns;
1178 }
1179
1180 static struct kobj_type netdev_queue_ktype = {
1181         .sysfs_ops = &netdev_queue_sysfs_ops,
1182         .release = netdev_queue_release,
1183         .default_attrs = netdev_queue_default_attrs,
1184         .namespace = netdev_queue_namespace,
1185 };
1186
1187 static int netdev_queue_add_kobject(struct net_device *dev, int index)
1188 {
1189         struct netdev_queue *queue = dev->_tx + index;
1190         struct kobject *kobj = &queue->kobj;
1191         int error = 0;
1192
1193         kobj->kset = dev->queues_kset;
1194         error = kobject_init_and_add(kobj, &netdev_queue_ktype, NULL,
1195             "tx-%u", index);
1196         if (error)
1197                 goto exit;
1198
1199 #ifdef CONFIG_BQL
1200         error = sysfs_create_group(kobj, &dql_group);
1201         if (error)
1202                 goto exit;
1203 #endif
1204
1205         kobject_uevent(kobj, KOBJ_ADD);
1206         dev_hold(queue->dev);
1207
1208         return 0;
1209 exit:
1210         kobject_put(kobj);
1211         return error;
1212 }
1213 #endif /* CONFIG_SYSFS */
1214
1215 int
1216 netdev_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
1217 {
1218 #ifdef CONFIG_SYSFS
1219         int i;
1220         int error = 0;
1221
1222         for (i = old_num; i < new_num; i++) {
1223                 error = netdev_queue_add_kobject(dev, i);
1224                 if (error) {
1225                         new_num = old_num;
1226                         break;
1227                 }
1228         }
1229
1230         while (--i >= new_num) {
1231                 struct netdev_queue *queue = dev->_tx + i;
1232
1233 #ifdef CONFIG_BQL
1234                 sysfs_remove_group(&queue->kobj, &dql_group);
1235 #endif
1236                 kobject_put(&queue->kobj);
1237         }
1238
1239         return error;
1240 #else
1241         return 0;
1242 #endif /* CONFIG_SYSFS */
1243 }
1244
1245 static int register_queue_kobjects(struct net_device *dev)
1246 {
1247         int error = 0, txq = 0, rxq = 0, real_rx = 0, real_tx = 0;
1248
1249 #ifdef CONFIG_SYSFS
1250         dev->queues_kset = kset_create_and_add("queues",
1251             NULL, &dev->dev.kobj);
1252         if (!dev->queues_kset)
1253                 return -ENOMEM;
1254         real_rx = dev->real_num_rx_queues;
1255 #endif
1256         real_tx = dev->real_num_tx_queues;
1257
1258         error = net_rx_queue_update_kobjects(dev, 0, real_rx);
1259         if (error)
1260                 goto error;
1261         rxq = real_rx;
1262
1263         error = netdev_queue_update_kobjects(dev, 0, real_tx);
1264         if (error)
1265                 goto error;
1266         txq = real_tx;
1267
1268         return 0;
1269
1270 error:
1271         netdev_queue_update_kobjects(dev, txq, 0);
1272         net_rx_queue_update_kobjects(dev, rxq, 0);
1273         return error;
1274 }
1275
1276 static void remove_queue_kobjects(struct net_device *dev)
1277 {
1278         int real_rx = 0, real_tx = 0;
1279
1280 #ifdef CONFIG_SYSFS
1281         real_rx = dev->real_num_rx_queues;
1282 #endif
1283         real_tx = dev->real_num_tx_queues;
1284
1285         net_rx_queue_update_kobjects(dev, real_rx, 0);
1286         netdev_queue_update_kobjects(dev, real_tx, 0);
1287 #ifdef CONFIG_SYSFS
1288         kset_unregister(dev->queues_kset);
1289 #endif
1290 }
1291
1292 static bool net_current_may_mount(void)
1293 {
1294         struct net *net = current->nsproxy->net_ns;
1295
1296         return ns_capable(net->user_ns, CAP_SYS_ADMIN);
1297 }
1298
1299 static void *net_grab_current_ns(void)
1300 {
1301         struct net *ns = current->nsproxy->net_ns;
1302 #ifdef CONFIG_NET_NS
1303         if (ns)
1304                 atomic_inc(&ns->passive);
1305 #endif
1306         return ns;
1307 }
1308
1309 static const void *net_initial_ns(void)
1310 {
1311         return &init_net;
1312 }
1313
1314 static const void *net_netlink_ns(struct sock *sk)
1315 {
1316         return sock_net(sk);
1317 }
1318
1319 struct kobj_ns_type_operations net_ns_type_operations = {
1320         .type = KOBJ_NS_TYPE_NET,
1321         .current_may_mount = net_current_may_mount,
1322         .grab_current_ns = net_grab_current_ns,
1323         .netlink_ns = net_netlink_ns,
1324         .initial_ns = net_initial_ns,
1325         .drop_ns = net_drop_ns,
1326 };
1327 EXPORT_SYMBOL_GPL(net_ns_type_operations);
1328
1329 static int netdev_uevent(struct device *d, struct kobj_uevent_env *env)
1330 {
1331         struct net_device *dev = to_net_dev(d);
1332         int retval;
1333
1334         /* pass interface to uevent. */
1335         retval = add_uevent_var(env, "INTERFACE=%s", dev->name);
1336         if (retval)
1337                 goto exit;
1338
1339         /* pass ifindex to uevent.
1340          * ifindex is useful as it won't change (interface name may change)
1341          * and is what RtNetlink uses natively. */
1342         retval = add_uevent_var(env, "IFINDEX=%d", dev->ifindex);
1343
1344 exit:
1345         return retval;
1346 }
1347
1348 /*
1349  *      netdev_release -- destroy and free a dead device.
1350  *      Called when last reference to device kobject is gone.
1351  */
1352 static void netdev_release(struct device *d)
1353 {
1354         struct net_device *dev = to_net_dev(d);
1355
1356         BUG_ON(dev->reg_state != NETREG_RELEASED);
1357
1358         kfree(dev->ifalias);
1359         netdev_freemem(dev);
1360 }
1361
1362 static const void *net_namespace(struct device *d)
1363 {
1364         struct net_device *dev;
1365         dev = container_of(d, struct net_device, dev);
1366         return dev_net(dev);
1367 }
1368
1369 static struct class net_class = {
1370         .name = "net",
1371         .dev_release = netdev_release,
1372         .dev_groups = net_class_groups,
1373         .dev_uevent = netdev_uevent,
1374         .ns_type = &net_ns_type_operations,
1375         .namespace = net_namespace,
1376 };
1377
1378 #ifdef CONFIG_OF_NET
1379 static int of_dev_node_match(struct device *dev, const void *data)
1380 {
1381         int ret = 0;
1382
1383         if (dev->parent)
1384                 ret = dev->parent->of_node == data;
1385
1386         return ret == 0 ? dev->of_node == data : ret;
1387 }
1388
1389 struct net_device *of_find_net_device_by_node(struct device_node *np)
1390 {
1391         struct device *dev;
1392
1393         dev = class_find_device(&net_class, NULL, np, of_dev_node_match);
1394         if (!dev)
1395                 return NULL;
1396
1397         return to_net_dev(dev);
1398 }
1399 EXPORT_SYMBOL(of_find_net_device_by_node);
1400 #endif
1401
1402 /* Delete sysfs entries but hold kobject reference until after all
1403  * netdev references are gone.
1404  */
1405 void netdev_unregister_kobject(struct net_device *ndev)
1406 {
1407         struct device *dev = &(ndev->dev);
1408
1409         kobject_get(&dev->kobj);
1410
1411         remove_queue_kobjects(ndev);
1412
1413         pm_runtime_set_memalloc_noio(dev, false);
1414
1415         device_del(dev);
1416 }
1417
1418 /* Create sysfs entries for network device. */
1419 int netdev_register_kobject(struct net_device *ndev)
1420 {
1421         struct device *dev = &(ndev->dev);
1422         const struct attribute_group **groups = ndev->sysfs_groups;
1423         int error = 0;
1424
1425         device_initialize(dev);
1426         dev->class = &net_class;
1427         dev->platform_data = ndev;
1428         dev->groups = groups;
1429
1430         dev_set_name(dev, "%s", ndev->name);
1431
1432 #ifdef CONFIG_SYSFS
1433         /* Allow for a device specific group */
1434         if (*groups)
1435                 groups++;
1436
1437         *groups++ = &netstat_group;
1438
1439 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
1440         if (ndev->ieee80211_ptr)
1441                 *groups++ = &wireless_group;
1442 #if IS_ENABLED(CONFIG_WIRELESS_EXT)
1443         else if (ndev->wireless_handlers)
1444                 *groups++ = &wireless_group;
1445 #endif
1446 #endif
1447 #endif /* CONFIG_SYSFS */
1448
1449         error = device_add(dev);
1450         if (error)
1451                 return error;
1452
1453         error = register_queue_kobjects(ndev);
1454         if (error) {
1455                 device_del(dev);
1456                 return error;
1457         }
1458
1459         pm_runtime_set_memalloc_noio(dev, true);
1460
1461         return error;
1462 }
1463
1464 int netdev_class_create_file_ns(struct class_attribute *class_attr,
1465                                 const void *ns)
1466 {
1467         return class_create_file_ns(&net_class, class_attr, ns);
1468 }
1469 EXPORT_SYMBOL(netdev_class_create_file_ns);
1470
1471 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
1472                                  const void *ns)
1473 {
1474         class_remove_file_ns(&net_class, class_attr, ns);
1475 }
1476 EXPORT_SYMBOL(netdev_class_remove_file_ns);
1477
1478 int __init netdev_kobject_init(void)
1479 {
1480         kobj_ns_type_register(&net_ns_type_operations);
1481         return class_register(&net_class);
1482 }