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_ulong[] = "%lu\n";
  35 static const char fmt_u64[] = "%llu\n";
  36
  37 static inline int dev_isalive(const struct net_device *dev)
  38 {
  39         return dev->reg_state <= NETREG_REGISTERED;
  40 }
  41
  42 /* use same locking rules as GIF* ioctl's */
  43 static ssize_t netdev_show(const struct device *dev,
  44                            struct device_attribute *attr, char *buf,
  45                            ssize_t (*format)(const struct net_device *, char *))
  46 {
  47         struct net_device *ndev = to_net_dev(dev);
  48         ssize_t ret = -EINVAL;
  49
  50         read_lock(&dev_base_lock);
  51         if (dev_isalive(ndev))
  52                 ret = (*format)(ndev, buf);
  53         read_unlock(&dev_base_lock);
  54
  55         return ret;
  56 }
  57
  58 /* generate a show function for simple field */
  59 #define NETDEVICE_SHOW(field, format_string)                            \
  60 static ssize_t format_##field(const struct net_device *dev, char *buf)  \
  61 {                                                                       \
  62         return sprintf(buf, format_string, dev->field);                 \
  63 }                                                                       \
  64 static ssize_t field##_show(struct device *dev,                         \
  65                             struct device_attribute *attr, char *buf)   \
  66 {                                                                       \
  67         return netdev_show(dev, attr, buf, format_##field);             \
  68 }                                                                       \
  69
  70 #define NETDEVICE_SHOW_RO(field, format_string)                         \
  71 NETDEVICE_SHOW(field, format_string);                                   \
  72 static DEVICE_ATTR_RO(field)
  73
  74 #define NETDEVICE_SHOW_RW(field, format_string)                         \
  75 NETDEVICE_SHOW(field, format_string);                                   \
  76 static DEVICE_ATTR_RW(field)
  77
  78 /* use same locking and permission rules as SIF* ioctl's */
  79 static ssize_t netdev_store(struct device *dev, struct device_attribute *attr,
  80                             const char *buf, size_t len,
  81                             int (*set)(struct net_device *, unsigned long))
  82 {
  83         struct net_device *netdev = to_net_dev(dev);
  84         struct net *net = dev_net(netdev);
  85         unsigned long new;
  86         int ret = -EINVAL;
  87
  88         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
  89                 return -EPERM;
  90
  91         ret = kstrtoul(buf, 0, &new);
  92         if (ret)
  93                 goto err;
  94
  95         if (!rtnl_trylock())
  96                 return restart_syscall();
  97
  98         if (dev_isalive(netdev)) {
  99                 if ((ret = (*set)(netdev, new)) == 0)
 100                         ret = len;
 101         }
 102         rtnl_unlock();
 103  err:
 104         return ret;
 105 }
 106
 107 NETDEVICE_SHOW_RO(dev_id, fmt_hex);
 108 NETDEVICE_SHOW_RO(dev_port, fmt_dec);
 109 NETDEVICE_SHOW_RO(addr_assign_type, fmt_dec);
 110 NETDEVICE_SHOW_RO(addr_len, fmt_dec);
 111 NETDEVICE_SHOW_RO(ifindex, fmt_dec);
 112 NETDEVICE_SHOW_RO(type, fmt_dec);
 113 NETDEVICE_SHOW_RO(link_mode, fmt_dec);
 114
 115 static ssize_t iflink_show(struct device *dev, struct device_attribute *attr,
 116                            char *buf)
 117 {
 118         struct net_device *ndev = to_net_dev(dev);
 119
 120         return sprintf(buf, fmt_dec, dev_get_iflink(ndev));
 121 }
 122 static DEVICE_ATTR_RO(iflink);
 123
 124 static ssize_t format_name_assign_type(const struct net_device *dev, char *buf)
 125 {
 126         return sprintf(buf, fmt_dec, dev->name_assign_type);
 127 }
 128
 129 static ssize_t name_assign_type_show(struct device *dev,
 130                                      struct device_attribute *attr,
 131                                      char *buf)
 132 {
 133         struct net_device *ndev = to_net_dev(dev);
 134         ssize_t ret = -EINVAL;
 135
 136         if (ndev->name_assign_type != NET_NAME_UNKNOWN)
 137                 ret = netdev_show(dev, attr, buf, format_name_assign_type);
 138
 139         return ret;
 140 }
 141 static DEVICE_ATTR_RO(name_assign_type);
 142
 143 /* use same locking rules as GIFHWADDR ioctl's */
 144 static ssize_t address_show(struct device *dev, struct device_attribute *attr,
 145                             char *buf)
 146 {
 147         struct net_device *ndev = to_net_dev(dev);
 148         ssize_t ret = -EINVAL;
 149
 150         read_lock(&dev_base_lock);
 151         if (dev_isalive(ndev))
 152                 ret = sysfs_format_mac(buf, ndev->dev_addr, ndev->addr_len);
 153         read_unlock(&dev_base_lock);
 154         return ret;
 155 }
 156 static DEVICE_ATTR_RO(address);
 157
 158 static ssize_t broadcast_show(struct device *dev,
 159                               struct device_attribute *attr, char *buf)
 160 {
 161         struct net_device *ndev = to_net_dev(dev);
 162         if (dev_isalive(ndev))
 163                 return sysfs_format_mac(buf, ndev->broadcast, ndev->addr_len);
 164         return -EINVAL;
 165 }
 166 static DEVICE_ATTR_RO(broadcast);
 167
 168 static int change_carrier(struct net_device *dev, unsigned long new_carrier)
 169 {
 170         if (!netif_running(dev))
 171                 return -EINVAL;
 172         return dev_change_carrier(dev, (bool) new_carrier);
 173 }
 174
 175 static ssize_t carrier_store(struct device *dev, struct device_attribute *attr,
 176                              const char *buf, size_t len)
 177 {
 178         return netdev_store(dev, attr, buf, len, change_carrier);
 179 }
 180
 181 static ssize_t carrier_show(struct device *dev,
 182                             struct device_attribute *attr, char *buf)
 183 {
 184         struct net_device *netdev = to_net_dev(dev);
 185         if (netif_running(netdev)) {
 186                 return sprintf(buf, fmt_dec, !!netif_carrier_ok(netdev));
 187         }
 188         return -EINVAL;
 189 }
 190 static DEVICE_ATTR_RW(carrier);
 191
 192 static ssize_t speed_show(struct device *dev,
 193                           struct device_attribute *attr, char *buf)
 194 {
 195         struct net_device *netdev = to_net_dev(dev);
 196         int ret = -EINVAL;
 197
 198         if (!rtnl_trylock())
 199                 return restart_syscall();
 200
 201         if (netif_running(netdev)) {
 202                 struct ethtool_cmd cmd;
 203                 if (!__ethtool_get_settings(netdev, &cmd))
 204                         ret = sprintf(buf, fmt_dec, ethtool_cmd_speed(&cmd));
 205         }
 206         rtnl_unlock();
 207         return ret;
 208 }
 209 static DEVICE_ATTR_RO(speed);
 210
 211 static ssize_t duplex_show(struct device *dev,
 212                            struct device_attribute *attr, char *buf)
 213 {
 214         struct net_device *netdev = to_net_dev(dev);
 215         int ret = -EINVAL;
 216
 217         if (!rtnl_trylock())
 218                 return restart_syscall();
 219
 220         if (netif_running(netdev)) {
 221                 struct ethtool_cmd cmd;
 222                 if (!__ethtool_get_settings(netdev, &cmd)) {
 223                         const char *duplex;
 224                         switch (cmd.duplex) {
 225                         case DUPLEX_HALF:
 226                                 duplex = "half";
 227                                 break;
 228                         case DUPLEX_FULL:
 229                                 duplex = "full";
 230                                 break;
 231                         default:
 232                                 duplex = "unknown";
 233                                 break;
 234                         }
 235                         ret = sprintf(buf, "%s\n", duplex);
 236                 }
 237         }
 238         rtnl_unlock();
 239         return ret;
 240 }
 241 static DEVICE_ATTR_RO(duplex);
 242
 243 static ssize_t dormant_show(struct device *dev,
 244                             struct device_attribute *attr, char *buf)
 245 {
 246         struct net_device *netdev = to_net_dev(dev);
 247
 248         if (netif_running(netdev))
 249                 return sprintf(buf, fmt_dec, !!netif_dormant(netdev));
 250
 251         return -EINVAL;
 252 }
 253 static DEVICE_ATTR_RO(dormant);
 254
 255 static const char *const operstates[] = {
 256         "unknown",
 257         "notpresent", /* currently unused */
 258         "down",
 259         "lowerlayerdown",
 260         "testing", /* currently unused */
 261         "dormant",
 262         "up"
 263 };
 264
 265 static ssize_t operstate_show(struct device *dev,
 266                               struct device_attribute *attr, char *buf)
 267 {
 268         const struct net_device *netdev = to_net_dev(dev);
 269         unsigned char operstate;
 270
 271         read_lock(&dev_base_lock);
 272         operstate = netdev->operstate;
 273         if (!netif_running(netdev))
 274                 operstate = IF_OPER_DOWN;
 275         read_unlock(&dev_base_lock);
 276
 277         if (operstate >= ARRAY_SIZE(operstates))
 278                 return -EINVAL; /* should not happen */
 279
 280         return sprintf(buf, "%s\n", operstates[operstate]);
 281 }
 282 static DEVICE_ATTR_RO(operstate);
 283
 284 static ssize_t carrier_changes_show(struct device *dev,
 285                                     struct device_attribute *attr,
 286                                     char *buf)
 287 {
 288         struct net_device *netdev = to_net_dev(dev);
 289         return sprintf(buf, fmt_dec,
 290                        atomic_read(&netdev->carrier_changes));
 291 }
 292 static DEVICE_ATTR_RO(carrier_changes);
 293
 294 /* read-write attributes */
 295
 296 static int change_mtu(struct net_device *dev, unsigned long new_mtu)
 297 {
 298         return dev_set_mtu(dev, (int) new_mtu);
 299 }
 300
 301 static ssize_t mtu_store(struct device *dev, struct device_attribute *attr,
 302                          const char *buf, size_t len)
 303 {
 304         return netdev_store(dev, attr, buf, len, change_mtu);
 305 }
 306 NETDEVICE_SHOW_RW(mtu, fmt_dec);
 307
 308 static int change_flags(struct net_device *dev, unsigned long new_flags)
 309 {
 310         return dev_change_flags(dev, (unsigned int) new_flags);
 311 }
 312
 313 static ssize_t flags_store(struct device *dev, struct device_attribute *attr,
 314                            const char *buf, size_t len)
 315 {
 316         return netdev_store(dev, attr, buf, len, change_flags);
 317 }
 318 NETDEVICE_SHOW_RW(flags, fmt_hex);
 319
 320 static int change_tx_queue_len(struct net_device *dev, unsigned long new_len)
 321 {
 322         dev->tx_queue_len = new_len;
 323         return 0;
 324 }
 325
 326 static ssize_t tx_queue_len_store(struct device *dev,
 327                                   struct device_attribute *attr,
 328                                   const char *buf, size_t len)
 329 {
 330         if (!capable(CAP_NET_ADMIN))
 331                 return -EPERM;
 332
 333         return netdev_store(dev, attr, buf, len, change_tx_queue_len);
 334 }
 335 NETDEVICE_SHOW_RW(tx_queue_len, fmt_ulong);
 336
 337 static int change_gro_flush_timeout(struct net_device *dev, unsigned long val)
 338 {
 339         dev->gro_flush_timeout = val;
 340         return 0;
 341 }
 342
 343 static ssize_t gro_flush_timeout_store(struct device *dev,
 344                                   struct device_attribute *attr,
 345                                   const char *buf, size_t len)
 346 {
 347         if (!capable(CAP_NET_ADMIN))
 348                 return -EPERM;
 349
 350         return netdev_store(dev, attr, buf, len, change_gro_flush_timeout);
 351 }
 352 NETDEVICE_SHOW_RW(gro_flush_timeout, fmt_ulong);
 353
 354 static ssize_t ifalias_store(struct device *dev, struct device_attribute *attr,
 355                              const char *buf, size_t len)
 356 {
 357         struct net_device *netdev = to_net_dev(dev);
 358         struct net *net = dev_net(netdev);
 359         size_t count = len;
 360         ssize_t ret;
 361
 362         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 363                 return -EPERM;
 364
 365         /* ignore trailing newline */
 366         if (len >  0 && buf[len - 1] == '\n')
 367                 --count;
 368
 369         if (!rtnl_trylock())
 370                 return restart_syscall();
 371         ret = dev_set_alias(netdev, buf, count);
 372         rtnl_unlock();
 373
 374         return ret < 0 ? ret : len;
 375 }
 376
 377 static ssize_t ifalias_show(struct device *dev,
 378                             struct device_attribute *attr, char *buf)
 379 {
 380         const struct net_device *netdev = to_net_dev(dev);
 381         ssize_t ret = 0;
 382
 383         if (!rtnl_trylock())
 384                 return restart_syscall();
 385         if (netdev->ifalias)
 386                 ret = sprintf(buf, "%s\n", netdev->ifalias);
 387         rtnl_unlock();
 388         return ret;
 389 }
 390 static DEVICE_ATTR_RW(ifalias);
 391
 392 static int change_group(struct net_device *dev, unsigned long new_group)
 393 {
 394         dev_set_group(dev, (int) new_group);
 395         return 0;
 396 }
 397
 398 static ssize_t group_store(struct device *dev, struct device_attribute *attr,
 399                            const char *buf, size_t len)
 400 {
 401         return netdev_store(dev, attr, buf, len, change_group);
 402 }
 403 NETDEVICE_SHOW(group, fmt_dec);
 404 static DEVICE_ATTR(netdev_group, S_IRUGO | S_IWUSR, group_show, group_store);
 405
 406 static int change_proto_down(struct net_device *dev, unsigned long proto_down)
 407 {
 408         return dev_change_proto_down(dev, (bool) proto_down);
 409 }
 410
 411 static ssize_t proto_down_store(struct device *dev,
 412                                 struct device_attribute *attr,
 413                                 const char *buf, size_t len)
 414 {
 415         return netdev_store(dev, attr, buf, len, change_proto_down);
 416 }
 417 NETDEVICE_SHOW_RW(proto_down, fmt_dec);
 418
 419 static ssize_t phys_port_id_show(struct device *dev,
 420                                  struct device_attribute *attr, char *buf)
 421 {
 422         struct net_device *netdev = to_net_dev(dev);
 423         ssize_t ret = -EINVAL;
 424
 425         if (!rtnl_trylock())
 426                 return restart_syscall();
 427
 428         if (dev_isalive(netdev)) {
 429                 struct netdev_phys_item_id ppid;
 430
 431                 ret = dev_get_phys_port_id(netdev, &ppid);
 432                 if (!ret)
 433                         ret = sprintf(buf, "%*phN\n", ppid.id_len, ppid.id);
 434         }
 435         rtnl_unlock();
 436
 437         return ret;
 438 }
 439 static DEVICE_ATTR_RO(phys_port_id);
 440
 441 static ssize_t phys_port_name_show(struct device *dev,
 442                                    struct device_attribute *attr, char *buf)
 443 {
 444         struct net_device *netdev = to_net_dev(dev);
 445         ssize_t ret = -EINVAL;
 446
 447         if (!rtnl_trylock())
 448                 return restart_syscall();
 449
 450         if (dev_isalive(netdev)) {
 451                 char name[IFNAMSIZ];
 452
 453                 ret = dev_get_phys_port_name(netdev, name, sizeof(name));
 454                 if (!ret)
 455                         ret = sprintf(buf, "%s\n", name);
 456         }
 457         rtnl_unlock();
 458
 459         return ret;
 460 }
 461 static DEVICE_ATTR_RO(phys_port_name);
 462
 463 static ssize_t phys_switch_id_show(struct device *dev,
 464                                    struct device_attribute *attr, char *buf)
 465 {
 466         struct net_device *netdev = to_net_dev(dev);
 467         ssize_t ret = -EINVAL;
 468
 469         if (!rtnl_trylock())
 470                 return restart_syscall();
 471
 472         if (dev_isalive(netdev)) {
 473                 struct switchdev_attr attr = {
 474                         .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
 475                         .flags = SWITCHDEV_F_NO_RECURSE,
 476                 };
 477
 478                 ret = switchdev_port_attr_get(netdev, &attr);
 479                 if (!ret)
 480                         ret = sprintf(buf, "%*phN\n", attr.u.ppid.id_len,
 481                                       attr.u.ppid.id);
 482         }
 483         rtnl_unlock();
 484
 485         return ret;
 486 }
 487 static DEVICE_ATTR_RO(phys_switch_id);
 488
 489 static struct attribute *net_class_attrs[] = {
 490         &dev_attr_netdev_group.attr,
 491         &dev_attr_type.attr,
 492         &dev_attr_dev_id.attr,
 493         &dev_attr_dev_port.attr,
 494         &dev_attr_iflink.attr,
 495         &dev_attr_ifindex.attr,
 496         &dev_attr_name_assign_type.attr,
 497         &dev_attr_addr_assign_type.attr,
 498         &dev_attr_addr_len.attr,
 499         &dev_attr_link_mode.attr,
 500         &dev_attr_address.attr,
 501         &dev_attr_broadcast.attr,
 502         &dev_attr_speed.attr,
 503         &dev_attr_duplex.attr,
 504         &dev_attr_dormant.attr,
 505         &dev_attr_operstate.attr,
 506         &dev_attr_carrier_changes.attr,
 507         &dev_attr_ifalias.attr,
 508         &dev_attr_carrier.attr,
 509         &dev_attr_mtu.attr,
 510         &dev_attr_flags.attr,
 511         &dev_attr_tx_queue_len.attr,
 512         &dev_attr_gro_flush_timeout.attr,
 513         &dev_attr_phys_port_id.attr,
 514         &dev_attr_phys_port_name.attr,
 515         &dev_attr_phys_switch_id.attr,
 516         &dev_attr_proto_down.attr,
 517         NULL,
 518 };
 519 ATTRIBUTE_GROUPS(net_class);
 520
 521 /* Show a given an attribute in the statistics group */
 522 static ssize_t netstat_show(const struct device *d,
 523                             struct device_attribute *attr, char *buf,
 524                             unsigned long offset)
 525 {
 526         struct net_device *dev = to_net_dev(d);
 527         ssize_t ret = -EINVAL;
 528
 529         WARN_ON(offset > sizeof(struct rtnl_link_stats64) ||
 530                         offset % sizeof(u64) != 0);
 531
 532         read_lock(&dev_base_lock);
 533         if (dev_isalive(dev)) {
 534                 struct rtnl_link_stats64 temp;
 535                 const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp);
 536
 537                 ret = sprintf(buf, fmt_u64, *(u64 *)(((u8 *) stats) + offset));
 538         }
 539         read_unlock(&dev_base_lock);
 540         return ret;
 541 }
 542
 543 /* generate a read-only statistics attribute */
 544 #define NETSTAT_ENTRY(name)                                             \
 545 static ssize_t name##_show(struct device *d,                            \
 546                            struct device_attribute *attr, char *buf)    \
 547 {                                                                       \
 548         return netstat_show(d, attr, buf,                               \
 549                             offsetof(struct rtnl_link_stats64, name));  \
 550 }                                                                       \
 551 static DEVICE_ATTR_RO(name)
 552
 553 NETSTAT_ENTRY(rx_packets);
 554 NETSTAT_ENTRY(tx_packets);
 555 NETSTAT_ENTRY(rx_bytes);
 556 NETSTAT_ENTRY(tx_bytes);
 557 NETSTAT_ENTRY(rx_errors);
 558 NETSTAT_ENTRY(tx_errors);
 559 NETSTAT_ENTRY(rx_dropped);
 560 NETSTAT_ENTRY(tx_dropped);
 561 NETSTAT_ENTRY(multicast);
 562 NETSTAT_ENTRY(collisions);
 563 NETSTAT_ENTRY(rx_length_errors);
 564 NETSTAT_ENTRY(rx_over_errors);
 565 NETSTAT_ENTRY(rx_crc_errors);
 566 NETSTAT_ENTRY(rx_frame_errors);
 567 NETSTAT_ENTRY(rx_fifo_errors);
 568 NETSTAT_ENTRY(rx_missed_errors);
 569 NETSTAT_ENTRY(tx_aborted_errors);
 570 NETSTAT_ENTRY(tx_carrier_errors);
 571 NETSTAT_ENTRY(tx_fifo_errors);
 572 NETSTAT_ENTRY(tx_heartbeat_errors);
 573 NETSTAT_ENTRY(tx_window_errors);
 574 NETSTAT_ENTRY(rx_compressed);
 575 NETSTAT_ENTRY(tx_compressed);
 576
 577 static struct attribute *netstat_attrs[] = {
 578         &dev_attr_rx_packets.attr,
 579         &dev_attr_tx_packets.attr,
 580         &dev_attr_rx_bytes.attr,
 581         &dev_attr_tx_bytes.attr,
 582         &dev_attr_rx_errors.attr,
 583         &dev_attr_tx_errors.attr,
 584         &dev_attr_rx_dropped.attr,
 585         &dev_attr_tx_dropped.attr,
 586         &dev_attr_multicast.attr,
 587         &dev_attr_collisions.attr,
 588         &dev_attr_rx_length_errors.attr,
 589         &dev_attr_rx_over_errors.attr,
 590         &dev_attr_rx_crc_errors.attr,
 591         &dev_attr_rx_frame_errors.attr,
 592         &dev_attr_rx_fifo_errors.attr,
 593         &dev_attr_rx_missed_errors.attr,
 594         &dev_attr_tx_aborted_errors.attr,
 595         &dev_attr_tx_carrier_errors.attr,
 596         &dev_attr_tx_fifo_errors.attr,
 597         &dev_attr_tx_heartbeat_errors.attr,
 598         &dev_attr_tx_window_errors.attr,
 599         &dev_attr_rx_compressed.attr,
 600         &dev_attr_tx_compressed.attr,
 601         NULL
 602 };
 603
 604
 605 static struct attribute_group netstat_group = {
 606         .name  = "statistics",
 607         .attrs  = netstat_attrs,
 608 };
 609
 610 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
 611 static struct attribute *wireless_attrs[] = {
 612         NULL
 613 };
 614
 615 static struct attribute_group wireless_group = {
 616         .name = "wireless",
 617         .attrs = wireless_attrs,
 618 };
 619 #endif
 620
 621 #else /* CONFIG_SYSFS */
 622 #define net_class_groups        NULL
 623 #endif /* CONFIG_SYSFS */
 624
 625 #ifdef CONFIG_SYSFS
 626 #define to_rx_queue_attr(_attr) container_of(_attr,             \
 627     struct rx_queue_attribute, attr)
 628
 629 #define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj)
 630
 631 static ssize_t rx_queue_attr_show(struct kobject *kobj, struct attribute *attr,
 632                                   char *buf)
 633 {
 634         struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
 635         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 636
 637         if (!attribute->show)
 638                 return -EIO;
 639
 640         return attribute->show(queue, attribute, buf);
 641 }
 642
 643 static ssize_t rx_queue_attr_store(struct kobject *kobj, struct attribute *attr,
 644                                    const char *buf, size_t count)
 645 {
 646         struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
 647         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 648
 649         if (!attribute->store)
 650                 return -EIO;
 651
 652         return attribute->store(queue, attribute, buf, count);
 653 }
 654
 655 static const struct sysfs_ops rx_queue_sysfs_ops = {
 656         .show = rx_queue_attr_show,
 657         .store = rx_queue_attr_store,
 658 };
 659
 660 #ifdef CONFIG_RPS
 661 static ssize_t show_rps_map(struct netdev_rx_queue *queue,
 662                             struct rx_queue_attribute *attribute, char *buf)
 663 {
 664         struct rps_map *map;
 665         cpumask_var_t mask;
 666         int i, len;
 667
 668         if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
 669                 return -ENOMEM;
 670
 671         rcu_read_lock();
 672         map = rcu_dereference(queue->rps_map);
 673         if (map)
 674                 for (i = 0; i < map->len; i++)
 675                         cpumask_set_cpu(map->cpus[i], mask);
 676
 677         len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
 678         rcu_read_unlock();
 679         free_cpumask_var(mask);
 680
 681         return len < PAGE_SIZE ? len : -EINVAL;
 682 }
 683
 684 static ssize_t store_rps_map(struct netdev_rx_queue *queue,
 685                       struct rx_queue_attribute *attribute,
 686                       const char *buf, size_t len)
 687 {
 688         struct rps_map *old_map, *map;
 689         cpumask_var_t mask;
 690         int err, cpu, i;
 691         static DEFINE_MUTEX(rps_map_mutex);
 692
 693         if (!capable(CAP_NET_ADMIN))
 694                 return -EPERM;
 695
 696         if (!alloc_cpumask_var(&mask, GFP_KERNEL))
 697                 return -ENOMEM;
 698
 699         err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
 700         if (err) {
 701                 free_cpumask_var(mask);
 702                 return err;
 703         }
 704
 705         map = kzalloc(max_t(unsigned int,
 706             RPS_MAP_SIZE(cpumask_weight(mask)), L1_CACHE_BYTES),
 707             GFP_KERNEL);
 708         if (!map) {
 709                 free_cpumask_var(mask);
 710                 return -ENOMEM;
 711         }
 712
 713         i = 0;
 714         for_each_cpu_and(cpu, mask, cpu_online_mask)
 715                 map->cpus[i++] = cpu;
 716
 717         if (i)
 718                 map->len = i;
 719         else {
 720                 kfree(map);
 721                 map = NULL;
 722         }
 723
 724         mutex_lock(&rps_map_mutex);
 725         old_map = rcu_dereference_protected(queue->rps_map,
 726                                             mutex_is_locked(&rps_map_mutex));
 727         rcu_assign_pointer(queue->rps_map, map);
 728
 729         if (map)
 730                 static_key_slow_inc(&rps_needed);
 731         if (old_map)
 732                 static_key_slow_dec(&rps_needed);
 733
 734         mutex_unlock(&rps_map_mutex);
 735
 736         if (old_map)
 737                 kfree_rcu(old_map, rcu);
 738
 739         free_cpumask_var(mask);
 740         return len;
 741 }
 742
 743 static ssize_t show_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
 744                                            struct rx_queue_attribute *attr,
 745                                            char *buf)
 746 {
 747         struct rps_dev_flow_table *flow_table;
 748         unsigned long val = 0;
 749
 750         rcu_read_lock();
 751         flow_table = rcu_dereference(queue->rps_flow_table);
 752         if (flow_table)
 753                 val = (unsigned long)flow_table->mask + 1;
 754         rcu_read_unlock();
 755
 756         return sprintf(buf, "%lu\n", val);
 757 }
 758
 759 static void rps_dev_flow_table_release(struct rcu_head *rcu)
 760 {
 761         struct rps_dev_flow_table *table = container_of(rcu,
 762             struct rps_dev_flow_table, rcu);
 763         vfree(table);
 764 }
 765
 766 static ssize_t store_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
 767                                      struct rx_queue_attribute *attr,
 768                                      const char *buf, size_t len)
 769 {
 770         unsigned long mask, count;
 771         struct rps_dev_flow_table *table, *old_table;
 772         static DEFINE_SPINLOCK(rps_dev_flow_lock);
 773         int rc;
 774
 775         if (!capable(CAP_NET_ADMIN))
 776                 return -EPERM;
 777
 778         rc = kstrtoul(buf, 0, &count);
 779         if (rc < 0)
 780                 return rc;
 781
 782         if (count) {
 783                 mask = count - 1;
 784                 /* mask = roundup_pow_of_two(count) - 1;
 785                  * without overflows...
 786                  */
 787                 while ((mask | (mask >> 1)) != mask)
 788                         mask |= (mask >> 1);
 789                 /* On 64 bit arches, must check mask fits in table->mask (u32),
 790                  * and on 32bit arches, must check
 791                  * RPS_DEV_FLOW_TABLE_SIZE(mask + 1) doesn't overflow.
 792                  */
 793 #if BITS_PER_LONG > 32
 794                 if (mask > (unsigned long)(u32)mask)
 795                         return -EINVAL;
 796 #else
 797                 if (mask > (ULONG_MAX - RPS_DEV_FLOW_TABLE_SIZE(1))
 798                                 / sizeof(struct rps_dev_flow)) {
 799                         /* Enforce a limit to prevent overflow */
 800                         return -EINVAL;
 801                 }
 802 #endif
 803                 table = vmalloc(RPS_DEV_FLOW_TABLE_SIZE(mask + 1));
 804                 if (!table)
 805                         return -ENOMEM;
 806
 807                 table->mask = mask;
 808                 for (count = 0; count <= mask; count++)
 809                         table->flows[count].cpu = RPS_NO_CPU;
 810         } else
 811                 table = NULL;
 812
 813         spin_lock(&rps_dev_flow_lock);
 814         old_table = rcu_dereference_protected(queue->rps_flow_table,
 815                                               lockdep_is_held(&rps_dev_flow_lock));
 816         rcu_assign_pointer(queue->rps_flow_table, table);
 817         spin_unlock(&rps_dev_flow_lock);
 818
 819         if (old_table)
 820                 call_rcu(&old_table->rcu, rps_dev_flow_table_release);
 821
 822         return len;
 823 }
 824
 825 static struct rx_queue_attribute rps_cpus_attribute =
 826         __ATTR(rps_cpus, S_IRUGO | S_IWUSR, show_rps_map, store_rps_map);
 827
 828
 829 static struct rx_queue_attribute rps_dev_flow_table_cnt_attribute =
 830         __ATTR(rps_flow_cnt, S_IRUGO | S_IWUSR,
 831             show_rps_dev_flow_table_cnt, store_rps_dev_flow_table_cnt);
 832 #endif /* CONFIG_RPS */
 833
 834 static struct attribute *rx_queue_default_attrs[] = {
 835 #ifdef CONFIG_RPS
 836         &rps_cpus_attribute.attr,
 837         &rps_dev_flow_table_cnt_attribute.attr,
 838 #endif
 839         NULL
 840 };
 841
 842 static void rx_queue_release(struct kobject *kobj)
 843 {
 844         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 845 #ifdef CONFIG_RPS
 846         struct rps_map *map;
 847         struct rps_dev_flow_table *flow_table;
 848
 849
 850         map = rcu_dereference_protected(queue->rps_map, 1);
 851         if (map) {
 852                 RCU_INIT_POINTER(queue->rps_map, NULL);
 853                 kfree_rcu(map, rcu);
 854         }
 855
 856         flow_table = rcu_dereference_protected(queue->rps_flow_table, 1);
 857         if (flow_table) {
 858                 RCU_INIT_POINTER(queue->rps_flow_table, NULL);
 859                 call_rcu(&flow_table->rcu, rps_dev_flow_table_release);
 860         }
 861 #endif
 862
 863         memset(kobj, 0, sizeof(*kobj));
 864         dev_put(queue->dev);
 865 }
 866
 867 static const void *rx_queue_namespace(struct kobject *kobj)
 868 {
 869         struct netdev_rx_queue *queue = to_rx_queue(kobj);
 870         struct device *dev = &queue->dev->dev;
 871         const void *ns = NULL;
 872
 873         if (dev->class && dev->class->ns_type)
 874                 ns = dev->class->namespace(dev);
 875
 876         return ns;
 877 }
 878
 879 static struct kobj_type rx_queue_ktype = {
 880         .sysfs_ops = &rx_queue_sysfs_ops,
 881         .release = rx_queue_release,
 882         .default_attrs = rx_queue_default_attrs,
 883         .namespace = rx_queue_namespace
 884 };
 885
 886 static int rx_queue_add_kobject(struct net_device *dev, int index)
 887 {
 888         struct netdev_rx_queue *queue = dev->_rx + index;
 889         struct kobject *kobj = &queue->kobj;
 890         int error = 0;
 891
 892         kobj->kset = dev->queues_kset;
 893         error = kobject_init_and_add(kobj, &rx_queue_ktype, NULL,
 894             "rx-%u", index);
 895         if (error)
 896                 goto exit;
 897
 898         if (dev->sysfs_rx_queue_group) {
 899                 error = sysfs_create_group(kobj, dev->sysfs_rx_queue_group);
 900                 if (error)
 901                         goto exit;
 902         }
 903
 904         kobject_uevent(kobj, KOBJ_ADD);
 905         dev_hold(queue->dev);
 906
 907         return error;
 908 exit:
 909         kobject_put(kobj);
 910         return error;
 911 }
 912 #endif /* CONFIG_SYSFS */
 913
 914 int
 915 net_rx_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
 916 {
 917 #ifdef CONFIG_SYSFS
 918         int i;
 919         int error = 0;
 920
 921 #ifndef CONFIG_RPS
 922         if (!dev->sysfs_rx_queue_group)
 923                 return 0;
 924 #endif
 925         for (i = old_num; i < new_num; i++) {
 926                 error = rx_queue_add_kobject(dev, i);
 927                 if (error) {
 928                         new_num = old_num;
 929                         break;
 930                 }
 931         }
 932
 933         while (--i >= new_num) {
 934                 if (dev->sysfs_rx_queue_group)
 935                         sysfs_remove_group(&dev->_rx[i].kobj,
 936                                            dev->sysfs_rx_queue_group);
 937                 kobject_put(&dev->_rx[i].kobj);
 938         }
 939
 940         return error;
 941 #else
 942         return 0;
 943 #endif
 944 }
 945
 946 #ifdef CONFIG_SYSFS
 947 /*
 948  * netdev_queue sysfs structures and functions.
 949  */
 950 struct netdev_queue_attribute {
 951         struct attribute attr;
 952         ssize_t (*show)(struct netdev_queue *queue,
 953             struct netdev_queue_attribute *attr, char *buf);
 954         ssize_t (*store)(struct netdev_queue *queue,
 955             struct netdev_queue_attribute *attr, const char *buf, size_t len);
 956 };
 957 #define to_netdev_queue_attr(_attr) container_of(_attr,         \
 958     struct netdev_queue_attribute, attr)
 959
 960 #define to_netdev_queue(obj) container_of(obj, struct netdev_queue, kobj)
 961
 962 static ssize_t netdev_queue_attr_show(struct kobject *kobj,
 963                                       struct attribute *attr, char *buf)
 964 {
 965         struct netdev_queue_attribute *attribute = to_netdev_queue_attr(attr);
 966         struct netdev_queue *queue = to_netdev_queue(kobj);
 967
 968         if (!attribute->show)
 969                 return -EIO;
 970
 971         return attribute->show(queue, attribute, buf);
 972 }
 973
 974 static ssize_t netdev_queue_attr_store(struct kobject *kobj,
 975                                        struct attribute *attr,
 976                                        const char *buf, size_t count)
 977 {
 978         struct netdev_queue_attribute *attribute = to_netdev_queue_attr(attr);
 979         struct netdev_queue *queue = to_netdev_queue(kobj);
 980
 981         if (!attribute->store)
 982                 return -EIO;
 983
 984         return attribute->store(queue, attribute, buf, count);
 985 }
 986
 987 static const struct sysfs_ops netdev_queue_sysfs_ops = {
 988         .show = netdev_queue_attr_show,
 989         .store = netdev_queue_attr_store,
 990 };
 991
 992 static ssize_t show_trans_timeout(struct netdev_queue *queue,
 993                                   struct netdev_queue_attribute *attribute,
 994                                   char *buf)
 995 {
 996         unsigned long trans_timeout;
 997
 998         spin_lock_irq(&queue->_xmit_lock);
 999         trans_timeout = queue->trans_timeout;
1000         spin_unlock_irq(&queue->_xmit_lock);
1001
1002         return sprintf(buf, "%lu", trans_timeout);
1003 }
1004
1005 #ifdef CONFIG_XPS
1006 static unsigned int get_netdev_queue_index(struct netdev_queue *queue)
1007 {
1008         struct net_device *dev = queue->dev;
1009         unsigned int i;
1010
1011         i = queue - dev->_tx;
1012         BUG_ON(i >= dev->num_tx_queues);
1013
1014         return i;
1015 }
1016
1017 static ssize_t show_tx_maxrate(struct netdev_queue *queue,
1018                                struct netdev_queue_attribute *attribute,
1019                                char *buf)
1020 {
1021         return sprintf(buf, "%lu\n", queue->tx_maxrate);
1022 }
1023
1024 static ssize_t set_tx_maxrate(struct netdev_queue *queue,
1025                               struct netdev_queue_attribute *attribute,
1026                               const char *buf, size_t len)
1027 {
1028         struct net_device *dev = queue->dev;
1029         int err, index = get_netdev_queue_index(queue);
1030         u32 rate = 0;
1031
1032         err = kstrtou32(buf, 10, &rate);
1033         if (err < 0)
1034                 return err;
1035
1036         if (!rtnl_trylock())
1037                 return restart_syscall();
1038
1039         err = -EOPNOTSUPP;
1040         if (dev->netdev_ops->ndo_set_tx_maxrate)
1041                 err = dev->netdev_ops->ndo_set_tx_maxrate(dev, index, rate);
1042
1043         rtnl_unlock();
1044         if (!err) {
1045                 queue->tx_maxrate = rate;
1046                 return len;
1047         }
1048         return err;
1049 }
1050
1051 static struct netdev_queue_attribute queue_tx_maxrate =
1052         __ATTR(tx_maxrate, S_IRUGO | S_IWUSR,
1053                show_tx_maxrate, set_tx_maxrate);
1054 #endif
1055
1056 static struct netdev_queue_attribute queue_trans_timeout =
1057         __ATTR(tx_timeout, S_IRUGO, show_trans_timeout, NULL);
1058
1059 #ifdef CONFIG_BQL
1060 /*
1061  * Byte queue limits sysfs structures and functions.
1062  */
1063 static ssize_t bql_show(char *buf, unsigned int value)
1064 {
1065         return sprintf(buf, "%u\n", value);
1066 }
1067
1068 static ssize_t bql_set(const char *buf, const size_t count,
1069                        unsigned int *pvalue)
1070 {
1071         unsigned int value;
1072         int err;
1073
1074         if (!strcmp(buf, "max") || !strcmp(buf, "max\n"))
1075                 value = DQL_MAX_LIMIT;
1076         else {
1077                 err = kstrtouint(buf, 10, &value);
1078                 if (err < 0)
1079                         return err;
1080                 if (value > DQL_MAX_LIMIT)
1081                         return -EINVAL;
1082         }
1083
1084         *pvalue = value;
1085
1086         return count;
1087 }
1088
1089 static ssize_t bql_show_hold_time(struct netdev_queue *queue,
1090                                   struct netdev_queue_attribute *attr,
1091                                   char *buf)
1092 {
1093         struct dql *dql = &queue->dql;
1094
1095         return sprintf(buf, "%u\n", jiffies_to_msecs(dql->slack_hold_time));
1096 }
1097
1098 static ssize_t bql_set_hold_time(struct netdev_queue *queue,
1099                                  struct netdev_queue_attribute *attribute,
1100                                  const char *buf, size_t len)
1101 {
1102         struct dql *dql = &queue->dql;
1103         unsigned int value;
1104         int err;
1105
1106         err = kstrtouint(buf, 10, &value);
1107         if (err < 0)
1108                 return err;
1109
1110         dql->slack_hold_time = msecs_to_jiffies(value);
1111
1112         return len;
1113 }
1114
1115 static struct netdev_queue_attribute bql_hold_time_attribute =
1116         __ATTR(hold_time, S_IRUGO | S_IWUSR, bql_show_hold_time,
1117             bql_set_hold_time);
1118
1119 static ssize_t bql_show_inflight(struct netdev_queue *queue,
1120                                  struct netdev_queue_attribute *attr,
1121                                  char *buf)
1122 {
1123         struct dql *dql = &queue->dql;
1124
1125         return sprintf(buf, "%u\n", dql->num_queued - dql->num_completed);
1126 }
1127
1128 static struct netdev_queue_attribute bql_inflight_attribute =
1129         __ATTR(inflight, S_IRUGO, bql_show_inflight, NULL);
1130
1131 #define BQL_ATTR(NAME, FIELD)                                           \
1132 static ssize_t bql_show_ ## NAME(struct netdev_queue *queue,            \
1133                                  struct netdev_queue_attribute *attr,   \
1134                                  char *buf)                             \
1135 {                                                                       \
1136         return bql_show(buf, queue->dql.FIELD);                         \
1137 }                                                                       \
1138                                                                         \
1139 static ssize_t bql_set_ ## NAME(struct netdev_queue *queue,             \
1140                                 struct netdev_queue_attribute *attr,    \
1141                                 const char *buf, size_t len)            \
1142 {                                                                       \
1143         return bql_set(buf, len, &queue->dql.FIELD);                    \
1144 }                                                                       \
1145                                                                         \
1146 static struct netdev_queue_attribute bql_ ## NAME ## _attribute =       \
1147         __ATTR(NAME, S_IRUGO | S_IWUSR, bql_show_ ## NAME,              \
1148             bql_set_ ## NAME);
1149
1150 BQL_ATTR(limit, limit)
1151 BQL_ATTR(limit_max, max_limit)
1152 BQL_ATTR(limit_min, min_limit)
1153
1154 static struct attribute *dql_attrs[] = {
1155         &bql_limit_attribute.attr,
1156         &bql_limit_max_attribute.attr,
1157         &bql_limit_min_attribute.attr,
1158         &bql_hold_time_attribute.attr,
1159         &bql_inflight_attribute.attr,
1160         NULL
1161 };
1162
1163 static struct attribute_group dql_group = {
1164         .name  = "byte_queue_limits",
1165         .attrs  = dql_attrs,
1166 };
1167 #endif /* CONFIG_BQL */
1168
1169 #ifdef CONFIG_XPS
1170 static ssize_t show_xps_map(struct netdev_queue *queue,
1171                             struct netdev_queue_attribute *attribute, char *buf)
1172 {
1173         struct net_device *dev = queue->dev;
1174         struct xps_dev_maps *dev_maps;
1175         cpumask_var_t mask;
1176         unsigned long index;
1177         int i, len;
1178
1179         if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
1180                 return -ENOMEM;
1181
1182         index = get_netdev_queue_index(queue);
1183
1184         rcu_read_lock();
1185         dev_maps = rcu_dereference(dev->xps_maps);
1186         if (dev_maps) {
1187                 for_each_possible_cpu(i) {
1188                         struct xps_map *map =
1189                             rcu_dereference(dev_maps->cpu_map[i]);
1190                         if (map) {
1191                                 int j;
1192                                 for (j = 0; j < map->len; j++) {
1193                                         if (map->queues[j] == index) {
1194                                                 cpumask_set_cpu(i, mask);
1195                                                 break;
1196                                         }
1197                                 }
1198                         }
1199                 }
1200         }
1201         rcu_read_unlock();
1202
1203         len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
1204         free_cpumask_var(mask);
1205         return len < PAGE_SIZE ? len : -EINVAL;
1206 }
1207
1208 static ssize_t store_xps_map(struct netdev_queue *queue,
1209                       struct netdev_queue_attribute *attribute,
1210                       const char *buf, size_t len)
1211 {
1212         struct net_device *dev = queue->dev;
1213         unsigned long index;
1214         cpumask_var_t mask;
1215         int err;
1216
1217         if (!capable(CAP_NET_ADMIN))
1218                 return -EPERM;
1219
1220         if (!alloc_cpumask_var(&mask, GFP_KERNEL))
1221                 return -ENOMEM;
1222
1223         index = get_netdev_queue_index(queue);
1224
1225         err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
1226         if (err) {
1227                 free_cpumask_var(mask);
1228                 return err;
1229         }
1230
1231         err = netif_set_xps_queue(dev, mask, index);
1232
1233         free_cpumask_var(mask);
1234
1235         return err ? : len;
1236 }
1237
1238 static struct netdev_queue_attribute xps_cpus_attribute =
1239     __ATTR(xps_cpus, S_IRUGO | S_IWUSR, show_xps_map, store_xps_map);
1240 #endif /* CONFIG_XPS */
1241
1242 static struct attribute *netdev_queue_default_attrs[] = {
1243         &queue_trans_timeout.attr,
1244 #ifdef CONFIG_XPS
1245         &xps_cpus_attribute.attr,
1246         &queue_tx_maxrate.attr,
1247 #endif
1248         NULL
1249 };
1250
1251 static void netdev_queue_release(struct kobject *kobj)
1252 {
1253         struct netdev_queue *queue = to_netdev_queue(kobj);
1254
1255         memset(kobj, 0, sizeof(*kobj));
1256         dev_put(queue->dev);
1257 }
1258
1259 static const void *netdev_queue_namespace(struct kobject *kobj)
1260 {
1261         struct netdev_queue *queue = to_netdev_queue(kobj);
1262         struct device *dev = &queue->dev->dev;
1263         const void *ns = NULL;
1264
1265         if (dev->class && dev->class->ns_type)
1266                 ns = dev->class->namespace(dev);
1267
1268         return ns;
1269 }
1270
1271 static struct kobj_type netdev_queue_ktype = {
1272         .sysfs_ops = &netdev_queue_sysfs_ops,
1273         .release = netdev_queue_release,
1274         .default_attrs = netdev_queue_default_attrs,
1275         .namespace = netdev_queue_namespace,
1276 };
1277
1278 static int netdev_queue_add_kobject(struct net_device *dev, int index)
1279 {
1280         struct netdev_queue *queue = dev->_tx + index;
1281         struct kobject *kobj = &queue->kobj;
1282         int error = 0;
1283
1284         kobj->kset = dev->queues_kset;
1285         error = kobject_init_and_add(kobj, &netdev_queue_ktype, NULL,
1286             "tx-%u", index);
1287         if (error)
1288                 goto exit;
1289
1290 #ifdef CONFIG_BQL
1291         error = sysfs_create_group(kobj, &dql_group);
1292         if (error)
1293                 goto exit;
1294 #endif
1295
1296         kobject_uevent(kobj, KOBJ_ADD);
1297         dev_hold(queue->dev);
1298
1299         return 0;
1300 exit:
1301         kobject_put(kobj);
1302         return error;
1303 }
1304 #endif /* CONFIG_SYSFS */
1305
1306 int
1307 netdev_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
1308 {
1309 #ifdef CONFIG_SYSFS
1310         int i;
1311         int error = 0;
1312
1313         for (i = old_num; i < new_num; i++) {
1314                 error = netdev_queue_add_kobject(dev, i);
1315                 if (error) {
1316                         new_num = old_num;
1317                         break;
1318                 }
1319         }
1320
1321         while (--i >= new_num) {
1322                 struct netdev_queue *queue = dev->_tx + i;
1323
1324 #ifdef CONFIG_BQL
1325                 sysfs_remove_group(&queue->kobj, &dql_group);
1326 #endif
1327                 kobject_put(&queue->kobj);
1328         }
1329
1330         return error;
1331 #else
1332         return 0;
1333 #endif /* CONFIG_SYSFS */
1334 }
1335
1336 static int register_queue_kobjects(struct net_device *dev)
1337 {
1338         int error = 0, txq = 0, rxq = 0, real_rx = 0, real_tx = 0;
1339
1340 #ifdef CONFIG_SYSFS
1341         dev->queues_kset = kset_create_and_add("queues",
1342             NULL, &dev->dev.kobj);
1343         if (!dev->queues_kset)
1344                 return -ENOMEM;
1345         real_rx = dev->real_num_rx_queues;
1346 #endif
1347         real_tx = dev->real_num_tx_queues;
1348
1349         error = net_rx_queue_update_kobjects(dev, 0, real_rx);
1350         if (error)
1351                 goto error;
1352         rxq = real_rx;
1353
1354         error = netdev_queue_update_kobjects(dev, 0, real_tx);
1355         if (error)
1356                 goto error;
1357         txq = real_tx;
1358
1359         return 0;
1360
1361 error:
1362         netdev_queue_update_kobjects(dev, txq, 0);
1363         net_rx_queue_update_kobjects(dev, rxq, 0);
1364         return error;
1365 }
1366
1367 static void remove_queue_kobjects(struct net_device *dev)
1368 {
1369         int real_rx = 0, real_tx = 0;
1370
1371 #ifdef CONFIG_SYSFS
1372         real_rx = dev->real_num_rx_queues;
1373 #endif
1374         real_tx = dev->real_num_tx_queues;
1375
1376         net_rx_queue_update_kobjects(dev, real_rx, 0);
1377         netdev_queue_update_kobjects(dev, real_tx, 0);
1378 #ifdef CONFIG_SYSFS
1379         kset_unregister(dev->queues_kset);
1380 #endif
1381 }
1382
1383 static bool net_current_may_mount(void)
1384 {
1385         struct net *net = current->nsproxy->net_ns;
1386
1387         return ns_capable(net->user_ns, CAP_SYS_ADMIN);
1388 }
1389
1390 static void *net_grab_current_ns(void)
1391 {
1392         struct net *ns = current->nsproxy->net_ns;
1393 #ifdef CONFIG_NET_NS
1394         if (ns)
1395                 atomic_inc(&ns->passive);
1396 #endif
1397         return ns;
1398 }
1399
1400 static const void *net_initial_ns(void)
1401 {
1402         return &init_net;
1403 }
1404
1405 static const void *net_netlink_ns(struct sock *sk)
1406 {
1407         return sock_net(sk);
1408 }
1409
1410 struct kobj_ns_type_operations net_ns_type_operations = {
1411         .type = KOBJ_NS_TYPE_NET,
1412         .current_may_mount = net_current_may_mount,
1413         .grab_current_ns = net_grab_current_ns,
1414         .netlink_ns = net_netlink_ns,
1415         .initial_ns = net_initial_ns,
1416         .drop_ns = net_drop_ns,
1417 };
1418 EXPORT_SYMBOL_GPL(net_ns_type_operations);
1419
1420 static int netdev_uevent(struct device *d, struct kobj_uevent_env *env)
1421 {
1422         struct net_device *dev = to_net_dev(d);
1423         int retval;
1424
1425         /* pass interface to uevent. */
1426         retval = add_uevent_var(env, "INTERFACE=%s", dev->name);
1427         if (retval)
1428                 goto exit;
1429
1430         /* pass ifindex to uevent.
1431          * ifindex is useful as it won't change (interface name may change)
1432          * and is what RtNetlink uses natively. */
1433         retval = add_uevent_var(env, "IFINDEX=%d", dev->ifindex);
1434
1435 exit:
1436         return retval;
1437 }
1438
1439 /*
1440  *      netdev_release -- destroy and free a dead device.
1441  *      Called when last reference to device kobject is gone.
1442  */
1443 static void netdev_release(struct device *d)
1444 {
1445         struct net_device *dev = to_net_dev(d);
1446
1447         BUG_ON(dev->reg_state != NETREG_RELEASED);
1448
1449         kfree(dev->ifalias);
1450         netdev_freemem(dev);
1451 }
1452
1453 static const void *net_namespace(struct device *d)
1454 {
1455         struct net_device *dev;
1456         dev = container_of(d, struct net_device, dev);
1457         return dev_net(dev);
1458 }
1459
1460 static struct class net_class = {
1461         .name = "net",
1462         .dev_release = netdev_release,
1463         .dev_groups = net_class_groups,
1464         .dev_uevent = netdev_uevent,
1465         .ns_type = &net_ns_type_operations,
1466         .namespace = net_namespace,
1467 };
1468
1469 #ifdef CONFIG_OF_NET
1470 static int of_dev_node_match(struct device *dev, const void *data)
1471 {
1472         int ret = 0;
1473
1474         if (dev->parent)
1475                 ret = dev->parent->of_node == data;
1476
1477         return ret == 0 ? dev->of_node == data : ret;
1478 }
1479
1480 /*
1481  * of_find_net_device_by_node - lookup the net device for the device node
1482  * @np: OF device node
1483  *
1484  * Looks up the net_device structure corresponding with the device node.
1485  * If successful, returns a pointer to the net_device with the embedded
1486  * struct device refcount incremented by one, or NULL on failure. The
1487  * refcount must be dropped when done with the net_device.
1488  */
1489 struct net_device *of_find_net_device_by_node(struct device_node *np)
1490 {
1491         struct device *dev;
1492
1493         dev = class_find_device(&net_class, NULL, np, of_dev_node_match);
1494         if (!dev)
1495                 return NULL;
1496
1497         return to_net_dev(dev);
1498 }
1499 EXPORT_SYMBOL(of_find_net_device_by_node);
1500 #endif
1501
1502 /* Delete sysfs entries but hold kobject reference until after all
1503  * netdev references are gone.
1504  */
1505 void netdev_unregister_kobject(struct net_device *ndev)
1506 {
1507         struct device *dev = &(ndev->dev);
1508
1509         kobject_get(&dev->kobj);
1510
1511         remove_queue_kobjects(ndev);
1512
1513         pm_runtime_set_memalloc_noio(dev, false);
1514
1515         device_del(dev);
1516 }
1517
1518 /* Create sysfs entries for network device. */
1519 int netdev_register_kobject(struct net_device *ndev)
1520 {
1521         struct device *dev = &(ndev->dev);
1522         const struct attribute_group **groups = ndev->sysfs_groups;
1523         int error = 0;
1524
1525         device_initialize(dev);
1526         dev->class = &net_class;
1527         dev->platform_data = ndev;
1528         dev->groups = groups;
1529
1530         dev_set_name(dev, "%s", ndev->name);
1531
1532 #ifdef CONFIG_SYSFS
1533         /* Allow for a device specific group */
1534         if (*groups)
1535                 groups++;
1536
1537         *groups++ = &netstat_group;
1538
1539 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
1540         if (ndev->ieee80211_ptr)
1541                 *groups++ = &wireless_group;
1542 #if IS_ENABLED(CONFIG_WIRELESS_EXT)
1543         else if (ndev->wireless_handlers)
1544                 *groups++ = &wireless_group;
1545 #endif
1546 #endif
1547 #endif /* CONFIG_SYSFS */
1548
1549         error = device_add(dev);
1550         if (error)
1551                 return error;
1552
1553         error = register_queue_kobjects(ndev);
1554         if (error) {
1555                 device_del(dev);
1556                 return error;
1557         }
1558
1559         pm_runtime_set_memalloc_noio(dev, true);
1560
1561         return error;
1562 }
1563
1564 int netdev_class_create_file_ns(struct class_attribute *class_attr,
1565                                 const void *ns)
1566 {
1567         return class_create_file_ns(&net_class, class_attr, ns);
1568 }
1569 EXPORT_SYMBOL(netdev_class_create_file_ns);
1570
1571 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
1572                                  const void *ns)
1573 {
1574         class_remove_file_ns(&net_class, class_attr, ns);
1575 }
1576 EXPORT_SYMBOL(netdev_class_remove_file_ns);
1577
1578 int __init netdev_kobject_init(void)
1579 {
1580         kobj_ns_type_register(&net_ns_type_operations);
1581         return class_register(&net_class);
1582 }