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/sched/signal.h> 19 #include <linux/nsproxy.h> 20 #include <net/sock.h> 21 #include <net/net_namespace.h> 22 #include <linux/rtnetlink.h> 23 #include <linux/vmalloc.h> 24 #include <linux/export.h> 25 #include <linux/jiffies.h> 26 #include <linux/pm_runtime.h> 27 #include <linux/of.h> 28 #include <linux/of_net.h> 29 30 #include "net-sysfs.h" 31 32 #ifdef CONFIG_SYSFS 33 static const char fmt_hex[] = "%#x\n"; 34 static const char fmt_dec[] = "%d\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 ret = (*set)(netdev, new); 101 if (ret == 0) 102 ret = len; 103 } 104 rtnl_unlock(); 105 err: 106 return ret; 107 } 108 109 NETDEVICE_SHOW_RO(dev_id, fmt_hex); 110 NETDEVICE_SHOW_RO(dev_port, fmt_dec); 111 NETDEVICE_SHOW_RO(addr_assign_type, fmt_dec); 112 NETDEVICE_SHOW_RO(addr_len, 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 iflink_show(struct device *dev, struct device_attribute *attr, 118 char *buf) 119 { 120 struct net_device *ndev = to_net_dev(dev); 121 122 return sprintf(buf, fmt_dec, dev_get_iflink(ndev)); 123 } 124 static DEVICE_ATTR_RO(iflink); 125 126 static ssize_t format_name_assign_type(const struct net_device *dev, char *buf) 127 { 128 return sprintf(buf, fmt_dec, dev->name_assign_type); 129 } 130 131 static ssize_t name_assign_type_show(struct device *dev, 132 struct device_attribute *attr, 133 char *buf) 134 { 135 struct net_device *ndev = to_net_dev(dev); 136 ssize_t ret = -EINVAL; 137 138 if (ndev->name_assign_type != NET_NAME_UNKNOWN) 139 ret = netdev_show(dev, attr, buf, format_name_assign_type); 140 141 return ret; 142 } 143 static DEVICE_ATTR_RO(name_assign_type); 144 145 /* use same locking rules as GIFHWADDR ioctl's */ 146 static ssize_t address_show(struct device *dev, struct device_attribute *attr, 147 char *buf) 148 { 149 struct net_device *ndev = to_net_dev(dev); 150 ssize_t ret = -EINVAL; 151 152 read_lock(&dev_base_lock); 153 if (dev_isalive(ndev)) 154 ret = sysfs_format_mac(buf, ndev->dev_addr, ndev->addr_len); 155 read_unlock(&dev_base_lock); 156 return ret; 157 } 158 static DEVICE_ATTR_RO(address); 159 160 static ssize_t broadcast_show(struct device *dev, 161 struct device_attribute *attr, char *buf) 162 { 163 struct net_device *ndev = to_net_dev(dev); 164 165 if (dev_isalive(ndev)) 166 return sysfs_format_mac(buf, ndev->broadcast, ndev->addr_len); 167 return -EINVAL; 168 } 169 static DEVICE_ATTR_RO(broadcast); 170 171 static int change_carrier(struct net_device *dev, unsigned long new_carrier) 172 { 173 if (!netif_running(dev)) 174 return -EINVAL; 175 return dev_change_carrier(dev, (bool)new_carrier); 176 } 177 178 static ssize_t carrier_store(struct device *dev, struct device_attribute *attr, 179 const char *buf, size_t len) 180 { 181 return netdev_store(dev, attr, buf, len, change_carrier); 182 } 183 184 static ssize_t carrier_show(struct device *dev, 185 struct device_attribute *attr, char *buf) 186 { 187 struct net_device *netdev = to_net_dev(dev); 188 189 if (netif_running(netdev)) 190 return sprintf(buf, fmt_dec, !!netif_carrier_ok(netdev)); 191 192 return -EINVAL; 193 } 194 static DEVICE_ATTR_RW(carrier); 195 196 static ssize_t speed_show(struct device *dev, 197 struct device_attribute *attr, char *buf) 198 { 199 struct net_device *netdev = to_net_dev(dev); 200 int ret = -EINVAL; 201 202 if (!rtnl_trylock()) 203 return restart_syscall(); 204 205 if (netif_running(netdev)) { 206 struct ethtool_link_ksettings cmd; 207 208 if (!__ethtool_get_link_ksettings(netdev, &cmd)) 209 ret = sprintf(buf, fmt_dec, cmd.base.speed); 210 } 211 rtnl_unlock(); 212 return ret; 213 } 214 static DEVICE_ATTR_RO(speed); 215 216 static ssize_t duplex_show(struct device *dev, 217 struct device_attribute *attr, char *buf) 218 { 219 struct net_device *netdev = to_net_dev(dev); 220 int ret = -EINVAL; 221 222 if (!rtnl_trylock()) 223 return restart_syscall(); 224 225 if (netif_running(netdev)) { 226 struct ethtool_link_ksettings cmd; 227 228 if (!__ethtool_get_link_ksettings(netdev, &cmd)) { 229 const char *duplex; 230 231 switch (cmd.base.duplex) { 232 case DUPLEX_HALF: 233 duplex = "half"; 234 break; 235 case DUPLEX_FULL: 236 duplex = "full"; 237 break; 238 default: 239 duplex = "unknown"; 240 break; 241 } 242 ret = sprintf(buf, "%s\n", duplex); 243 } 244 } 245 rtnl_unlock(); 246 return ret; 247 } 248 static DEVICE_ATTR_RO(duplex); 249 250 static ssize_t dormant_show(struct device *dev, 251 struct device_attribute *attr, char *buf) 252 { 253 struct net_device *netdev = to_net_dev(dev); 254 255 if (netif_running(netdev)) 256 return sprintf(buf, fmt_dec, !!netif_dormant(netdev)); 257 258 return -EINVAL; 259 } 260 static DEVICE_ATTR_RO(dormant); 261 262 static const char *const operstates[] = { 263 "unknown", 264 "notpresent", /* currently unused */ 265 "down", 266 "lowerlayerdown", 267 "testing", /* currently unused */ 268 "dormant", 269 "up" 270 }; 271 272 static ssize_t operstate_show(struct device *dev, 273 struct device_attribute *attr, char *buf) 274 { 275 const struct net_device *netdev = to_net_dev(dev); 276 unsigned char operstate; 277 278 read_lock(&dev_base_lock); 279 operstate = netdev->operstate; 280 if (!netif_running(netdev)) 281 operstate = IF_OPER_DOWN; 282 read_unlock(&dev_base_lock); 283 284 if (operstate >= ARRAY_SIZE(operstates)) 285 return -EINVAL; /* should not happen */ 286 287 return sprintf(buf, "%s\n", operstates[operstate]); 288 } 289 static DEVICE_ATTR_RO(operstate); 290 291 static ssize_t carrier_changes_show(struct device *dev, 292 struct device_attribute *attr, 293 char *buf) 294 { 295 struct net_device *netdev = to_net_dev(dev); 296 297 return sprintf(buf, fmt_dec, 298 atomic_read(&netdev->carrier_changes)); 299 } 300 static DEVICE_ATTR_RO(carrier_changes); 301 302 /* read-write attributes */ 303 304 static int change_mtu(struct net_device *dev, unsigned long new_mtu) 305 { 306 return dev_set_mtu(dev, (int)new_mtu); 307 } 308 309 static ssize_t mtu_store(struct device *dev, struct device_attribute *attr, 310 const char *buf, size_t len) 311 { 312 return netdev_store(dev, attr, buf, len, change_mtu); 313 } 314 NETDEVICE_SHOW_RW(mtu, fmt_dec); 315 316 static int change_flags(struct net_device *dev, unsigned long new_flags) 317 { 318 return dev_change_flags(dev, (unsigned int)new_flags); 319 } 320 321 static ssize_t flags_store(struct device *dev, struct device_attribute *attr, 322 const char *buf, size_t len) 323 { 324 return netdev_store(dev, attr, buf, len, change_flags); 325 } 326 NETDEVICE_SHOW_RW(flags, fmt_hex); 327 328 static int change_tx_queue_len(struct net_device *dev, unsigned long new_len) 329 { 330 unsigned int orig_len = dev->tx_queue_len; 331 int res; 332 333 if (new_len != (unsigned int)new_len) 334 return -ERANGE; 335 336 if (new_len != orig_len) { 337 dev->tx_queue_len = new_len; 338 res = call_netdevice_notifiers(NETDEV_CHANGE_TX_QUEUE_LEN, dev); 339 res = notifier_to_errno(res); 340 if (res) { 341 netdev_err(dev, 342 "refused to change device tx_queue_len\n"); 343 dev->tx_queue_len = orig_len; 344 return -EFAULT; 345 } 346 } 347 348 return 0; 349 } 350 351 static ssize_t tx_queue_len_store(struct device *dev, 352 struct device_attribute *attr, 353 const char *buf, size_t len) 354 { 355 if (!capable(CAP_NET_ADMIN)) 356 return -EPERM; 357 358 return netdev_store(dev, attr, buf, len, change_tx_queue_len); 359 } 360 NETDEVICE_SHOW_RW(tx_queue_len, fmt_dec); 361 362 static int change_gro_flush_timeout(struct net_device *dev, unsigned long val) 363 { 364 dev->gro_flush_timeout = val; 365 return 0; 366 } 367 368 static ssize_t gro_flush_timeout_store(struct device *dev, 369 struct device_attribute *attr, 370 const char *buf, size_t len) 371 { 372 if (!capable(CAP_NET_ADMIN)) 373 return -EPERM; 374 375 return netdev_store(dev, attr, buf, len, change_gro_flush_timeout); 376 } 377 NETDEVICE_SHOW_RW(gro_flush_timeout, fmt_ulong); 378 379 static ssize_t ifalias_store(struct device *dev, struct device_attribute *attr, 380 const char *buf, size_t len) 381 { 382 struct net_device *netdev = to_net_dev(dev); 383 struct net *net = dev_net(netdev); 384 size_t count = len; 385 ssize_t ret; 386 387 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 388 return -EPERM; 389 390 /* ignore trailing newline */ 391 if (len > 0 && buf[len - 1] == '\n') 392 --count; 393 394 ret = dev_set_alias(netdev, buf, count); 395 396 return ret < 0 ? ret : len; 397 } 398 399 static ssize_t ifalias_show(struct device *dev, 400 struct device_attribute *attr, char *buf) 401 { 402 const struct net_device *netdev = to_net_dev(dev); 403 char tmp[IFALIASZ]; 404 ssize_t ret = 0; 405 406 ret = dev_get_alias(netdev, tmp, sizeof(tmp)); 407 if (ret > 0) 408 ret = sprintf(buf, "%s\n", tmp); 409 return ret; 410 } 411 static DEVICE_ATTR_RW(ifalias); 412 413 static int change_group(struct net_device *dev, unsigned long new_group) 414 { 415 dev_set_group(dev, (int)new_group); 416 return 0; 417 } 418 419 static ssize_t group_store(struct device *dev, struct device_attribute *attr, 420 const char *buf, size_t len) 421 { 422 return netdev_store(dev, attr, buf, len, change_group); 423 } 424 NETDEVICE_SHOW(group, fmt_dec); 425 static DEVICE_ATTR(netdev_group, S_IRUGO | S_IWUSR, group_show, group_store); 426 427 static int change_proto_down(struct net_device *dev, unsigned long proto_down) 428 { 429 return dev_change_proto_down(dev, (bool)proto_down); 430 } 431 432 static ssize_t proto_down_store(struct device *dev, 433 struct device_attribute *attr, 434 const char *buf, size_t len) 435 { 436 return netdev_store(dev, attr, buf, len, change_proto_down); 437 } 438 NETDEVICE_SHOW_RW(proto_down, fmt_dec); 439 440 static ssize_t phys_port_id_show(struct device *dev, 441 struct device_attribute *attr, char *buf) 442 { 443 struct net_device *netdev = to_net_dev(dev); 444 ssize_t ret = -EINVAL; 445 446 if (!rtnl_trylock()) 447 return restart_syscall(); 448 449 if (dev_isalive(netdev)) { 450 struct netdev_phys_item_id ppid; 451 452 ret = dev_get_phys_port_id(netdev, &ppid); 453 if (!ret) 454 ret = sprintf(buf, "%*phN\n", ppid.id_len, ppid.id); 455 } 456 rtnl_unlock(); 457 458 return ret; 459 } 460 static DEVICE_ATTR_RO(phys_port_id); 461 462 static ssize_t phys_port_name_show(struct device *dev, 463 struct device_attribute *attr, char *buf) 464 { 465 struct net_device *netdev = to_net_dev(dev); 466 ssize_t ret = -EINVAL; 467 468 if (!rtnl_trylock()) 469 return restart_syscall(); 470 471 if (dev_isalive(netdev)) { 472 char name[IFNAMSIZ]; 473 474 ret = dev_get_phys_port_name(netdev, name, sizeof(name)); 475 if (!ret) 476 ret = sprintf(buf, "%s\n", name); 477 } 478 rtnl_unlock(); 479 480 return ret; 481 } 482 static DEVICE_ATTR_RO(phys_port_name); 483 484 static ssize_t phys_switch_id_show(struct device *dev, 485 struct device_attribute *attr, char *buf) 486 { 487 struct net_device *netdev = to_net_dev(dev); 488 ssize_t ret = -EINVAL; 489 490 if (!rtnl_trylock()) 491 return restart_syscall(); 492 493 if (dev_isalive(netdev)) { 494 struct switchdev_attr attr = { 495 .orig_dev = netdev, 496 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID, 497 .flags = SWITCHDEV_F_NO_RECURSE, 498 }; 499 500 ret = switchdev_port_attr_get(netdev, &attr); 501 if (!ret) 502 ret = sprintf(buf, "%*phN\n", attr.u.ppid.id_len, 503 attr.u.ppid.id); 504 } 505 rtnl_unlock(); 506 507 return ret; 508 } 509 static DEVICE_ATTR_RO(phys_switch_id); 510 511 static struct attribute *net_class_attrs[] __ro_after_init = { 512 &dev_attr_netdev_group.attr, 513 &dev_attr_type.attr, 514 &dev_attr_dev_id.attr, 515 &dev_attr_dev_port.attr, 516 &dev_attr_iflink.attr, 517 &dev_attr_ifindex.attr, 518 &dev_attr_name_assign_type.attr, 519 &dev_attr_addr_assign_type.attr, 520 &dev_attr_addr_len.attr, 521 &dev_attr_link_mode.attr, 522 &dev_attr_address.attr, 523 &dev_attr_broadcast.attr, 524 &dev_attr_speed.attr, 525 &dev_attr_duplex.attr, 526 &dev_attr_dormant.attr, 527 &dev_attr_operstate.attr, 528 &dev_attr_carrier_changes.attr, 529 &dev_attr_ifalias.attr, 530 &dev_attr_carrier.attr, 531 &dev_attr_mtu.attr, 532 &dev_attr_flags.attr, 533 &dev_attr_tx_queue_len.attr, 534 &dev_attr_gro_flush_timeout.attr, 535 &dev_attr_phys_port_id.attr, 536 &dev_attr_phys_port_name.attr, 537 &dev_attr_phys_switch_id.attr, 538 &dev_attr_proto_down.attr, 539 NULL, 540 }; 541 ATTRIBUTE_GROUPS(net_class); 542 543 /* Show a given an attribute in the statistics group */ 544 static ssize_t netstat_show(const struct device *d, 545 struct device_attribute *attr, char *buf, 546 unsigned long offset) 547 { 548 struct net_device *dev = to_net_dev(d); 549 ssize_t ret = -EINVAL; 550 551 WARN_ON(offset > sizeof(struct rtnl_link_stats64) || 552 offset % sizeof(u64) != 0); 553 554 read_lock(&dev_base_lock); 555 if (dev_isalive(dev)) { 556 struct rtnl_link_stats64 temp; 557 const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp); 558 559 ret = sprintf(buf, fmt_u64, *(u64 *)(((u8 *)stats) + offset)); 560 } 561 read_unlock(&dev_base_lock); 562 return ret; 563 } 564 565 /* generate a read-only statistics attribute */ 566 #define NETSTAT_ENTRY(name) \ 567 static ssize_t name##_show(struct device *d, \ 568 struct device_attribute *attr, char *buf) \ 569 { \ 570 return netstat_show(d, attr, buf, \ 571 offsetof(struct rtnl_link_stats64, name)); \ 572 } \ 573 static DEVICE_ATTR_RO(name) 574 575 NETSTAT_ENTRY(rx_packets); 576 NETSTAT_ENTRY(tx_packets); 577 NETSTAT_ENTRY(rx_bytes); 578 NETSTAT_ENTRY(tx_bytes); 579 NETSTAT_ENTRY(rx_errors); 580 NETSTAT_ENTRY(tx_errors); 581 NETSTAT_ENTRY(rx_dropped); 582 NETSTAT_ENTRY(tx_dropped); 583 NETSTAT_ENTRY(multicast); 584 NETSTAT_ENTRY(collisions); 585 NETSTAT_ENTRY(rx_length_errors); 586 NETSTAT_ENTRY(rx_over_errors); 587 NETSTAT_ENTRY(rx_crc_errors); 588 NETSTAT_ENTRY(rx_frame_errors); 589 NETSTAT_ENTRY(rx_fifo_errors); 590 NETSTAT_ENTRY(rx_missed_errors); 591 NETSTAT_ENTRY(tx_aborted_errors); 592 NETSTAT_ENTRY(tx_carrier_errors); 593 NETSTAT_ENTRY(tx_fifo_errors); 594 NETSTAT_ENTRY(tx_heartbeat_errors); 595 NETSTAT_ENTRY(tx_window_errors); 596 NETSTAT_ENTRY(rx_compressed); 597 NETSTAT_ENTRY(tx_compressed); 598 NETSTAT_ENTRY(rx_nohandler); 599 600 static struct attribute *netstat_attrs[] __ro_after_init = { 601 &dev_attr_rx_packets.attr, 602 &dev_attr_tx_packets.attr, 603 &dev_attr_rx_bytes.attr, 604 &dev_attr_tx_bytes.attr, 605 &dev_attr_rx_errors.attr, 606 &dev_attr_tx_errors.attr, 607 &dev_attr_rx_dropped.attr, 608 &dev_attr_tx_dropped.attr, 609 &dev_attr_multicast.attr, 610 &dev_attr_collisions.attr, 611 &dev_attr_rx_length_errors.attr, 612 &dev_attr_rx_over_errors.attr, 613 &dev_attr_rx_crc_errors.attr, 614 &dev_attr_rx_frame_errors.attr, 615 &dev_attr_rx_fifo_errors.attr, 616 &dev_attr_rx_missed_errors.attr, 617 &dev_attr_tx_aborted_errors.attr, 618 &dev_attr_tx_carrier_errors.attr, 619 &dev_attr_tx_fifo_errors.attr, 620 &dev_attr_tx_heartbeat_errors.attr, 621 &dev_attr_tx_window_errors.attr, 622 &dev_attr_rx_compressed.attr, 623 &dev_attr_tx_compressed.attr, 624 &dev_attr_rx_nohandler.attr, 625 NULL 626 }; 627 628 static const struct attribute_group netstat_group = { 629 .name = "statistics", 630 .attrs = netstat_attrs, 631 }; 632 633 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211) 634 static struct attribute *wireless_attrs[] = { 635 NULL 636 }; 637 638 static const struct attribute_group wireless_group = { 639 .name = "wireless", 640 .attrs = wireless_attrs, 641 }; 642 #endif 643 644 #else /* CONFIG_SYSFS */ 645 #define net_class_groups NULL 646 #endif /* CONFIG_SYSFS */ 647 648 #ifdef CONFIG_SYSFS 649 #define to_rx_queue_attr(_attr) \ 650 container_of(_attr, struct rx_queue_attribute, attr) 651 652 #define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj) 653 654 static ssize_t rx_queue_attr_show(struct kobject *kobj, struct attribute *attr, 655 char *buf) 656 { 657 const struct rx_queue_attribute *attribute = to_rx_queue_attr(attr); 658 struct netdev_rx_queue *queue = to_rx_queue(kobj); 659 660 if (!attribute->show) 661 return -EIO; 662 663 return attribute->show(queue, buf); 664 } 665 666 static ssize_t rx_queue_attr_store(struct kobject *kobj, struct attribute *attr, 667 const char *buf, size_t count) 668 { 669 const struct rx_queue_attribute *attribute = to_rx_queue_attr(attr); 670 struct netdev_rx_queue *queue = to_rx_queue(kobj); 671 672 if (!attribute->store) 673 return -EIO; 674 675 return attribute->store(queue, buf, count); 676 } 677 678 static const struct sysfs_ops rx_queue_sysfs_ops = { 679 .show = rx_queue_attr_show, 680 .store = rx_queue_attr_store, 681 }; 682 683 #ifdef CONFIG_RPS 684 static ssize_t show_rps_map(struct netdev_rx_queue *queue, char *buf) 685 { 686 struct rps_map *map; 687 cpumask_var_t mask; 688 int i, len; 689 690 if (!zalloc_cpumask_var(&mask, GFP_KERNEL)) 691 return -ENOMEM; 692 693 rcu_read_lock(); 694 map = rcu_dereference(queue->rps_map); 695 if (map) 696 for (i = 0; i < map->len; i++) 697 cpumask_set_cpu(map->cpus[i], mask); 698 699 len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask)); 700 rcu_read_unlock(); 701 free_cpumask_var(mask); 702 703 return len < PAGE_SIZE ? len : -EINVAL; 704 } 705 706 static ssize_t store_rps_map(struct netdev_rx_queue *queue, 707 const char *buf, size_t len) 708 { 709 struct rps_map *old_map, *map; 710 cpumask_var_t mask; 711 int err, cpu, i; 712 static DEFINE_MUTEX(rps_map_mutex); 713 714 if (!capable(CAP_NET_ADMIN)) 715 return -EPERM; 716 717 if (!alloc_cpumask_var(&mask, GFP_KERNEL)) 718 return -ENOMEM; 719 720 err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits); 721 if (err) { 722 free_cpumask_var(mask); 723 return err; 724 } 725 726 map = kzalloc(max_t(unsigned int, 727 RPS_MAP_SIZE(cpumask_weight(mask)), L1_CACHE_BYTES), 728 GFP_KERNEL); 729 if (!map) { 730 free_cpumask_var(mask); 731 return -ENOMEM; 732 } 733 734 i = 0; 735 for_each_cpu_and(cpu, mask, cpu_online_mask) 736 map->cpus[i++] = cpu; 737 738 if (i) { 739 map->len = i; 740 } else { 741 kfree(map); 742 map = NULL; 743 } 744 745 mutex_lock(&rps_map_mutex); 746 old_map = rcu_dereference_protected(queue->rps_map, 747 mutex_is_locked(&rps_map_mutex)); 748 rcu_assign_pointer(queue->rps_map, map); 749 750 if (map) 751 static_key_slow_inc(&rps_needed); 752 if (old_map) 753 static_key_slow_dec(&rps_needed); 754 755 mutex_unlock(&rps_map_mutex); 756 757 if (old_map) 758 kfree_rcu(old_map, rcu); 759 760 free_cpumask_var(mask); 761 return len; 762 } 763 764 static ssize_t show_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue, 765 char *buf) 766 { 767 struct rps_dev_flow_table *flow_table; 768 unsigned long val = 0; 769 770 rcu_read_lock(); 771 flow_table = rcu_dereference(queue->rps_flow_table); 772 if (flow_table) 773 val = (unsigned long)flow_table->mask + 1; 774 rcu_read_unlock(); 775 776 return sprintf(buf, "%lu\n", val); 777 } 778 779 static void rps_dev_flow_table_release(struct rcu_head *rcu) 780 { 781 struct rps_dev_flow_table *table = container_of(rcu, 782 struct rps_dev_flow_table, rcu); 783 vfree(table); 784 } 785 786 static ssize_t store_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue, 787 const char *buf, size_t len) 788 { 789 unsigned long mask, count; 790 struct rps_dev_flow_table *table, *old_table; 791 static DEFINE_SPINLOCK(rps_dev_flow_lock); 792 int rc; 793 794 if (!capable(CAP_NET_ADMIN)) 795 return -EPERM; 796 797 rc = kstrtoul(buf, 0, &count); 798 if (rc < 0) 799 return rc; 800 801 if (count) { 802 mask = count - 1; 803 /* mask = roundup_pow_of_two(count) - 1; 804 * without overflows... 805 */ 806 while ((mask | (mask >> 1)) != mask) 807 mask |= (mask >> 1); 808 /* On 64 bit arches, must check mask fits in table->mask (u32), 809 * and on 32bit arches, must check 810 * RPS_DEV_FLOW_TABLE_SIZE(mask + 1) doesn't overflow. 811 */ 812 #if BITS_PER_LONG > 32 813 if (mask > (unsigned long)(u32)mask) 814 return -EINVAL; 815 #else 816 if (mask > (ULONG_MAX - RPS_DEV_FLOW_TABLE_SIZE(1)) 817 / sizeof(struct rps_dev_flow)) { 818 /* Enforce a limit to prevent overflow */ 819 return -EINVAL; 820 } 821 #endif 822 table = vmalloc(RPS_DEV_FLOW_TABLE_SIZE(mask + 1)); 823 if (!table) 824 return -ENOMEM; 825 826 table->mask = mask; 827 for (count = 0; count <= mask; count++) 828 table->flows[count].cpu = RPS_NO_CPU; 829 } else { 830 table = NULL; 831 } 832 833 spin_lock(&rps_dev_flow_lock); 834 old_table = rcu_dereference_protected(queue->rps_flow_table, 835 lockdep_is_held(&rps_dev_flow_lock)); 836 rcu_assign_pointer(queue->rps_flow_table, table); 837 spin_unlock(&rps_dev_flow_lock); 838 839 if (old_table) 840 call_rcu(&old_table->rcu, rps_dev_flow_table_release); 841 842 return len; 843 } 844 845 static struct rx_queue_attribute rps_cpus_attribute __ro_after_init 846 = __ATTR(rps_cpus, S_IRUGO | S_IWUSR, show_rps_map, store_rps_map); 847 848 static struct rx_queue_attribute rps_dev_flow_table_cnt_attribute __ro_after_init 849 = __ATTR(rps_flow_cnt, S_IRUGO | S_IWUSR, 850 show_rps_dev_flow_table_cnt, store_rps_dev_flow_table_cnt); 851 #endif /* CONFIG_RPS */ 852 853 static struct attribute *rx_queue_default_attrs[] __ro_after_init = { 854 #ifdef CONFIG_RPS 855 &rps_cpus_attribute.attr, 856 &rps_dev_flow_table_cnt_attribute.attr, 857 #endif 858 NULL 859 }; 860 861 static void rx_queue_release(struct kobject *kobj) 862 { 863 struct netdev_rx_queue *queue = to_rx_queue(kobj); 864 #ifdef CONFIG_RPS 865 struct rps_map *map; 866 struct rps_dev_flow_table *flow_table; 867 868 map = rcu_dereference_protected(queue->rps_map, 1); 869 if (map) { 870 RCU_INIT_POINTER(queue->rps_map, NULL); 871 kfree_rcu(map, rcu); 872 } 873 874 flow_table = rcu_dereference_protected(queue->rps_flow_table, 1); 875 if (flow_table) { 876 RCU_INIT_POINTER(queue->rps_flow_table, NULL); 877 call_rcu(&flow_table->rcu, rps_dev_flow_table_release); 878 } 879 #endif 880 881 memset(kobj, 0, sizeof(*kobj)); 882 dev_put(queue->dev); 883 } 884 885 static const void *rx_queue_namespace(struct kobject *kobj) 886 { 887 struct netdev_rx_queue *queue = to_rx_queue(kobj); 888 struct device *dev = &queue->dev->dev; 889 const void *ns = NULL; 890 891 if (dev->class && dev->class->ns_type) 892 ns = dev->class->namespace(dev); 893 894 return ns; 895 } 896 897 static struct kobj_type rx_queue_ktype __ro_after_init = { 898 .sysfs_ops = &rx_queue_sysfs_ops, 899 .release = rx_queue_release, 900 .default_attrs = rx_queue_default_attrs, 901 .namespace = rx_queue_namespace 902 }; 903 904 static int rx_queue_add_kobject(struct net_device *dev, int index) 905 { 906 struct netdev_rx_queue *queue = dev->_rx + index; 907 struct kobject *kobj = &queue->kobj; 908 int error = 0; 909 910 kobj->kset = dev->queues_kset; 911 error = kobject_init_and_add(kobj, &rx_queue_ktype, NULL, 912 "rx-%u", index); 913 if (error) 914 return error; 915 916 if (dev->sysfs_rx_queue_group) { 917 error = sysfs_create_group(kobj, dev->sysfs_rx_queue_group); 918 if (error) { 919 kobject_put(kobj); 920 return error; 921 } 922 } 923 924 kobject_uevent(kobj, KOBJ_ADD); 925 dev_hold(queue->dev); 926 927 return error; 928 } 929 #endif /* CONFIG_SYSFS */ 930 931 int 932 net_rx_queue_update_kobjects(struct net_device *dev, int old_num, int new_num) 933 { 934 #ifdef CONFIG_SYSFS 935 int i; 936 int error = 0; 937 938 #ifndef CONFIG_RPS 939 if (!dev->sysfs_rx_queue_group) 940 return 0; 941 #endif 942 for (i = old_num; i < new_num; i++) { 943 error = rx_queue_add_kobject(dev, i); 944 if (error) { 945 new_num = old_num; 946 break; 947 } 948 } 949 950 while (--i >= new_num) { 951 struct kobject *kobj = &dev->_rx[i].kobj; 952 953 if (!atomic_read(&dev_net(dev)->count)) 954 kobj->uevent_suppress = 1; 955 if (dev->sysfs_rx_queue_group) 956 sysfs_remove_group(kobj, dev->sysfs_rx_queue_group); 957 kobject_put(kobj); 958 } 959 960 return error; 961 #else 962 return 0; 963 #endif 964 } 965 966 #ifdef CONFIG_SYSFS 967 /* 968 * netdev_queue sysfs structures and functions. 969 */ 970 struct netdev_queue_attribute { 971 struct attribute attr; 972 ssize_t (*show)(struct netdev_queue *queue, char *buf); 973 ssize_t (*store)(struct netdev_queue *queue, 974 const char *buf, size_t len); 975 }; 976 #define to_netdev_queue_attr(_attr) \ 977 container_of(_attr, struct netdev_queue_attribute, attr) 978 979 #define to_netdev_queue(obj) container_of(obj, struct netdev_queue, kobj) 980 981 static ssize_t netdev_queue_attr_show(struct kobject *kobj, 982 struct attribute *attr, char *buf) 983 { 984 const struct netdev_queue_attribute *attribute 985 = to_netdev_queue_attr(attr); 986 struct netdev_queue *queue = to_netdev_queue(kobj); 987 988 if (!attribute->show) 989 return -EIO; 990 991 return attribute->show(queue, buf); 992 } 993 994 static ssize_t netdev_queue_attr_store(struct kobject *kobj, 995 struct attribute *attr, 996 const char *buf, size_t count) 997 { 998 const struct netdev_queue_attribute *attribute 999 = to_netdev_queue_attr(attr); 1000 struct netdev_queue *queue = to_netdev_queue(kobj); 1001 1002 if (!attribute->store) 1003 return -EIO; 1004 1005 return attribute->store(queue, buf, count); 1006 } 1007 1008 static const struct sysfs_ops netdev_queue_sysfs_ops = { 1009 .show = netdev_queue_attr_show, 1010 .store = netdev_queue_attr_store, 1011 }; 1012 1013 static ssize_t tx_timeout_show(struct netdev_queue *queue, char *buf) 1014 { 1015 unsigned long trans_timeout; 1016 1017 spin_lock_irq(&queue->_xmit_lock); 1018 trans_timeout = queue->trans_timeout; 1019 spin_unlock_irq(&queue->_xmit_lock); 1020 1021 return sprintf(buf, "%lu", trans_timeout); 1022 } 1023 1024 static unsigned int get_netdev_queue_index(struct netdev_queue *queue) 1025 { 1026 struct net_device *dev = queue->dev; 1027 unsigned int i; 1028 1029 i = queue - dev->_tx; 1030 BUG_ON(i >= dev->num_tx_queues); 1031 1032 return i; 1033 } 1034 1035 static ssize_t traffic_class_show(struct netdev_queue *queue, 1036 char *buf) 1037 { 1038 struct net_device *dev = queue->dev; 1039 int index = get_netdev_queue_index(queue); 1040 int tc = netdev_txq_to_tc(dev, index); 1041 1042 if (tc < 0) 1043 return -EINVAL; 1044 1045 return sprintf(buf, "%u\n", tc); 1046 } 1047 1048 #ifdef CONFIG_XPS 1049 static ssize_t tx_maxrate_show(struct netdev_queue *queue, 1050 char *buf) 1051 { 1052 return sprintf(buf, "%lu\n", queue->tx_maxrate); 1053 } 1054 1055 static ssize_t tx_maxrate_store(struct netdev_queue *queue, 1056 const char *buf, size_t len) 1057 { 1058 struct net_device *dev = queue->dev; 1059 int err, index = get_netdev_queue_index(queue); 1060 u32 rate = 0; 1061 1062 err = kstrtou32(buf, 10, &rate); 1063 if (err < 0) 1064 return err; 1065 1066 if (!rtnl_trylock()) 1067 return restart_syscall(); 1068 1069 err = -EOPNOTSUPP; 1070 if (dev->netdev_ops->ndo_set_tx_maxrate) 1071 err = dev->netdev_ops->ndo_set_tx_maxrate(dev, index, rate); 1072 1073 rtnl_unlock(); 1074 if (!err) { 1075 queue->tx_maxrate = rate; 1076 return len; 1077 } 1078 return err; 1079 } 1080 1081 static struct netdev_queue_attribute queue_tx_maxrate __ro_after_init 1082 = __ATTR_RW(tx_maxrate); 1083 #endif 1084 1085 static struct netdev_queue_attribute queue_trans_timeout __ro_after_init 1086 = __ATTR_RO(tx_timeout); 1087 1088 static struct netdev_queue_attribute queue_traffic_class __ro_after_init 1089 = __ATTR_RO(traffic_class); 1090 1091 #ifdef CONFIG_BQL 1092 /* 1093 * Byte queue limits sysfs structures and functions. 1094 */ 1095 static ssize_t bql_show(char *buf, unsigned int value) 1096 { 1097 return sprintf(buf, "%u\n", value); 1098 } 1099 1100 static ssize_t bql_set(const char *buf, const size_t count, 1101 unsigned int *pvalue) 1102 { 1103 unsigned int value; 1104 int err; 1105 1106 if (!strcmp(buf, "max") || !strcmp(buf, "max\n")) { 1107 value = DQL_MAX_LIMIT; 1108 } else { 1109 err = kstrtouint(buf, 10, &value); 1110 if (err < 0) 1111 return err; 1112 if (value > DQL_MAX_LIMIT) 1113 return -EINVAL; 1114 } 1115 1116 *pvalue = value; 1117 1118 return count; 1119 } 1120 1121 static ssize_t bql_show_hold_time(struct netdev_queue *queue, 1122 char *buf) 1123 { 1124 struct dql *dql = &queue->dql; 1125 1126 return sprintf(buf, "%u\n", jiffies_to_msecs(dql->slack_hold_time)); 1127 } 1128 1129 static ssize_t bql_set_hold_time(struct netdev_queue *queue, 1130 const char *buf, size_t len) 1131 { 1132 struct dql *dql = &queue->dql; 1133 unsigned int value; 1134 int err; 1135 1136 err = kstrtouint(buf, 10, &value); 1137 if (err < 0) 1138 return err; 1139 1140 dql->slack_hold_time = msecs_to_jiffies(value); 1141 1142 return len; 1143 } 1144 1145 static struct netdev_queue_attribute bql_hold_time_attribute __ro_after_init 1146 = __ATTR(hold_time, S_IRUGO | S_IWUSR, 1147 bql_show_hold_time, bql_set_hold_time); 1148 1149 static ssize_t bql_show_inflight(struct netdev_queue *queue, 1150 char *buf) 1151 { 1152 struct dql *dql = &queue->dql; 1153 1154 return sprintf(buf, "%u\n", dql->num_queued - dql->num_completed); 1155 } 1156 1157 static struct netdev_queue_attribute bql_inflight_attribute __ro_after_init = 1158 __ATTR(inflight, S_IRUGO, bql_show_inflight, NULL); 1159 1160 #define BQL_ATTR(NAME, FIELD) \ 1161 static ssize_t bql_show_ ## NAME(struct netdev_queue *queue, \ 1162 char *buf) \ 1163 { \ 1164 return bql_show(buf, queue->dql.FIELD); \ 1165 } \ 1166 \ 1167 static ssize_t bql_set_ ## NAME(struct netdev_queue *queue, \ 1168 const char *buf, size_t len) \ 1169 { \ 1170 return bql_set(buf, len, &queue->dql.FIELD); \ 1171 } \ 1172 \ 1173 static struct netdev_queue_attribute bql_ ## NAME ## _attribute __ro_after_init \ 1174 = __ATTR(NAME, S_IRUGO | S_IWUSR, \ 1175 bql_show_ ## NAME, bql_set_ ## NAME) 1176 1177 BQL_ATTR(limit, limit); 1178 BQL_ATTR(limit_max, max_limit); 1179 BQL_ATTR(limit_min, min_limit); 1180 1181 static struct attribute *dql_attrs[] __ro_after_init = { 1182 &bql_limit_attribute.attr, 1183 &bql_limit_max_attribute.attr, 1184 &bql_limit_min_attribute.attr, 1185 &bql_hold_time_attribute.attr, 1186 &bql_inflight_attribute.attr, 1187 NULL 1188 }; 1189 1190 static const struct attribute_group dql_group = { 1191 .name = "byte_queue_limits", 1192 .attrs = dql_attrs, 1193 }; 1194 #endif /* CONFIG_BQL */ 1195 1196 #ifdef CONFIG_XPS 1197 static ssize_t xps_cpus_show(struct netdev_queue *queue, 1198 char *buf) 1199 { 1200 struct net_device *dev = queue->dev; 1201 int cpu, len, num_tc = 1, tc = 0; 1202 struct xps_dev_maps *dev_maps; 1203 cpumask_var_t mask; 1204 unsigned long index; 1205 1206 if (!zalloc_cpumask_var(&mask, GFP_KERNEL)) 1207 return -ENOMEM; 1208 1209 index = get_netdev_queue_index(queue); 1210 1211 if (dev->num_tc) { 1212 num_tc = dev->num_tc; 1213 tc = netdev_txq_to_tc(dev, index); 1214 if (tc < 0) 1215 return -EINVAL; 1216 } 1217 1218 rcu_read_lock(); 1219 dev_maps = rcu_dereference(dev->xps_maps); 1220 if (dev_maps) { 1221 for_each_possible_cpu(cpu) { 1222 int i, tci = cpu * num_tc + tc; 1223 struct xps_map *map; 1224 1225 map = rcu_dereference(dev_maps->cpu_map[tci]); 1226 if (!map) 1227 continue; 1228 1229 for (i = map->len; i--;) { 1230 if (map->queues[i] == index) { 1231 cpumask_set_cpu(cpu, mask); 1232 break; 1233 } 1234 } 1235 } 1236 } 1237 rcu_read_unlock(); 1238 1239 len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask)); 1240 free_cpumask_var(mask); 1241 return len < PAGE_SIZE ? len : -EINVAL; 1242 } 1243 1244 static ssize_t xps_cpus_store(struct netdev_queue *queue, 1245 const char *buf, size_t len) 1246 { 1247 struct net_device *dev = queue->dev; 1248 unsigned long index; 1249 cpumask_var_t mask; 1250 int err; 1251 1252 if (!capable(CAP_NET_ADMIN)) 1253 return -EPERM; 1254 1255 if (!alloc_cpumask_var(&mask, GFP_KERNEL)) 1256 return -ENOMEM; 1257 1258 index = get_netdev_queue_index(queue); 1259 1260 err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits); 1261 if (err) { 1262 free_cpumask_var(mask); 1263 return err; 1264 } 1265 1266 err = netif_set_xps_queue(dev, mask, index); 1267 1268 free_cpumask_var(mask); 1269 1270 return err ? : len; 1271 } 1272 1273 static struct netdev_queue_attribute xps_cpus_attribute __ro_after_init 1274 = __ATTR_RW(xps_cpus); 1275 #endif /* CONFIG_XPS */ 1276 1277 static struct attribute *netdev_queue_default_attrs[] __ro_after_init = { 1278 &queue_trans_timeout.attr, 1279 &queue_traffic_class.attr, 1280 #ifdef CONFIG_XPS 1281 &xps_cpus_attribute.attr, 1282 &queue_tx_maxrate.attr, 1283 #endif 1284 NULL 1285 }; 1286 1287 static void netdev_queue_release(struct kobject *kobj) 1288 { 1289 struct netdev_queue *queue = to_netdev_queue(kobj); 1290 1291 memset(kobj, 0, sizeof(*kobj)); 1292 dev_put(queue->dev); 1293 } 1294 1295 static const void *netdev_queue_namespace(struct kobject *kobj) 1296 { 1297 struct netdev_queue *queue = to_netdev_queue(kobj); 1298 struct device *dev = &queue->dev->dev; 1299 const void *ns = NULL; 1300 1301 if (dev->class && dev->class->ns_type) 1302 ns = dev->class->namespace(dev); 1303 1304 return ns; 1305 } 1306 1307 static struct kobj_type netdev_queue_ktype __ro_after_init = { 1308 .sysfs_ops = &netdev_queue_sysfs_ops, 1309 .release = netdev_queue_release, 1310 .default_attrs = netdev_queue_default_attrs, 1311 .namespace = netdev_queue_namespace, 1312 }; 1313 1314 static int netdev_queue_add_kobject(struct net_device *dev, int index) 1315 { 1316 struct netdev_queue *queue = dev->_tx + index; 1317 struct kobject *kobj = &queue->kobj; 1318 int error = 0; 1319 1320 kobj->kset = dev->queues_kset; 1321 error = kobject_init_and_add(kobj, &netdev_queue_ktype, NULL, 1322 "tx-%u", index); 1323 if (error) 1324 return error; 1325 1326 #ifdef CONFIG_BQL 1327 error = sysfs_create_group(kobj, &dql_group); 1328 if (error) { 1329 kobject_put(kobj); 1330 return error; 1331 } 1332 #endif 1333 1334 kobject_uevent(kobj, KOBJ_ADD); 1335 dev_hold(queue->dev); 1336 1337 return 0; 1338 } 1339 #endif /* CONFIG_SYSFS */ 1340 1341 int 1342 netdev_queue_update_kobjects(struct net_device *dev, int old_num, int new_num) 1343 { 1344 #ifdef CONFIG_SYSFS 1345 int i; 1346 int error = 0; 1347 1348 for (i = old_num; i < new_num; i++) { 1349 error = netdev_queue_add_kobject(dev, i); 1350 if (error) { 1351 new_num = old_num; 1352 break; 1353 } 1354 } 1355 1356 while (--i >= new_num) { 1357 struct netdev_queue *queue = dev->_tx + i; 1358 1359 if (!atomic_read(&dev_net(dev)->count)) 1360 queue->kobj.uevent_suppress = 1; 1361 #ifdef CONFIG_BQL 1362 sysfs_remove_group(&queue->kobj, &dql_group); 1363 #endif 1364 kobject_put(&queue->kobj); 1365 } 1366 1367 return error; 1368 #else 1369 return 0; 1370 #endif /* CONFIG_SYSFS */ 1371 } 1372 1373 static int register_queue_kobjects(struct net_device *dev) 1374 { 1375 int error = 0, txq = 0, rxq = 0, real_rx = 0, real_tx = 0; 1376 1377 #ifdef CONFIG_SYSFS 1378 dev->queues_kset = kset_create_and_add("queues", 1379 NULL, &dev->dev.kobj); 1380 if (!dev->queues_kset) 1381 return -ENOMEM; 1382 real_rx = dev->real_num_rx_queues; 1383 #endif 1384 real_tx = dev->real_num_tx_queues; 1385 1386 error = net_rx_queue_update_kobjects(dev, 0, real_rx); 1387 if (error) 1388 goto error; 1389 rxq = real_rx; 1390 1391 error = netdev_queue_update_kobjects(dev, 0, real_tx); 1392 if (error) 1393 goto error; 1394 txq = real_tx; 1395 1396 return 0; 1397 1398 error: 1399 netdev_queue_update_kobjects(dev, txq, 0); 1400 net_rx_queue_update_kobjects(dev, rxq, 0); 1401 return error; 1402 } 1403 1404 static void remove_queue_kobjects(struct net_device *dev) 1405 { 1406 int real_rx = 0, real_tx = 0; 1407 1408 #ifdef CONFIG_SYSFS 1409 real_rx = dev->real_num_rx_queues; 1410 #endif 1411 real_tx = dev->real_num_tx_queues; 1412 1413 net_rx_queue_update_kobjects(dev, real_rx, 0); 1414 netdev_queue_update_kobjects(dev, real_tx, 0); 1415 #ifdef CONFIG_SYSFS 1416 kset_unregister(dev->queues_kset); 1417 #endif 1418 } 1419 1420 static bool net_current_may_mount(void) 1421 { 1422 struct net *net = current->nsproxy->net_ns; 1423 1424 return ns_capable(net->user_ns, CAP_SYS_ADMIN); 1425 } 1426 1427 static void *net_grab_current_ns(void) 1428 { 1429 struct net *ns = current->nsproxy->net_ns; 1430 #ifdef CONFIG_NET_NS 1431 if (ns) 1432 refcount_inc(&ns->passive); 1433 #endif 1434 return ns; 1435 } 1436 1437 static const void *net_initial_ns(void) 1438 { 1439 return &init_net; 1440 } 1441 1442 static const void *net_netlink_ns(struct sock *sk) 1443 { 1444 return sock_net(sk); 1445 } 1446 1447 const struct kobj_ns_type_operations net_ns_type_operations = { 1448 .type = KOBJ_NS_TYPE_NET, 1449 .current_may_mount = net_current_may_mount, 1450 .grab_current_ns = net_grab_current_ns, 1451 .netlink_ns = net_netlink_ns, 1452 .initial_ns = net_initial_ns, 1453 .drop_ns = net_drop_ns, 1454 }; 1455 EXPORT_SYMBOL_GPL(net_ns_type_operations); 1456 1457 static int netdev_uevent(struct device *d, struct kobj_uevent_env *env) 1458 { 1459 struct net_device *dev = to_net_dev(d); 1460 int retval; 1461 1462 /* pass interface to uevent. */ 1463 retval = add_uevent_var(env, "INTERFACE=%s", dev->name); 1464 if (retval) 1465 goto exit; 1466 1467 /* pass ifindex to uevent. 1468 * ifindex is useful as it won't change (interface name may change) 1469 * and is what RtNetlink uses natively. 1470 */ 1471 retval = add_uevent_var(env, "IFINDEX=%d", dev->ifindex); 1472 1473 exit: 1474 return retval; 1475 } 1476 1477 /* 1478 * netdev_release -- destroy and free a dead device. 1479 * Called when last reference to device kobject is gone. 1480 */ 1481 static void netdev_release(struct device *d) 1482 { 1483 struct net_device *dev = to_net_dev(d); 1484 1485 BUG_ON(dev->reg_state != NETREG_RELEASED); 1486 1487 /* no need to wait for rcu grace period: 1488 * device is dead and about to be freed. 1489 */ 1490 kfree(rcu_access_pointer(dev->ifalias)); 1491 netdev_freemem(dev); 1492 } 1493 1494 static const void *net_namespace(struct device *d) 1495 { 1496 struct net_device *dev = to_net_dev(d); 1497 1498 return dev_net(dev); 1499 } 1500 1501 static struct class net_class __ro_after_init = { 1502 .name = "net", 1503 .dev_release = netdev_release, 1504 .dev_groups = net_class_groups, 1505 .dev_uevent = netdev_uevent, 1506 .ns_type = &net_ns_type_operations, 1507 .namespace = net_namespace, 1508 }; 1509 1510 #ifdef CONFIG_OF_NET 1511 static int of_dev_node_match(struct device *dev, const void *data) 1512 { 1513 int ret = 0; 1514 1515 if (dev->parent) 1516 ret = dev->parent->of_node == data; 1517 1518 return ret == 0 ? dev->of_node == data : ret; 1519 } 1520 1521 /* 1522 * of_find_net_device_by_node - lookup the net device for the device node 1523 * @np: OF device node 1524 * 1525 * Looks up the net_device structure corresponding with the device node. 1526 * If successful, returns a pointer to the net_device with the embedded 1527 * struct device refcount incremented by one, or NULL on failure. The 1528 * refcount must be dropped when done with the net_device. 1529 */ 1530 struct net_device *of_find_net_device_by_node(struct device_node *np) 1531 { 1532 struct device *dev; 1533 1534 dev = class_find_device(&net_class, NULL, np, of_dev_node_match); 1535 if (!dev) 1536 return NULL; 1537 1538 return to_net_dev(dev); 1539 } 1540 EXPORT_SYMBOL(of_find_net_device_by_node); 1541 #endif 1542 1543 /* Delete sysfs entries but hold kobject reference until after all 1544 * netdev references are gone. 1545 */ 1546 void netdev_unregister_kobject(struct net_device *ndev) 1547 { 1548 struct device *dev = &ndev->dev; 1549 1550 if (!atomic_read(&dev_net(ndev)->count)) 1551 dev_set_uevent_suppress(dev, 1); 1552 1553 kobject_get(&dev->kobj); 1554 1555 remove_queue_kobjects(ndev); 1556 1557 pm_runtime_set_memalloc_noio(dev, false); 1558 1559 device_del(dev); 1560 } 1561 1562 /* Create sysfs entries for network device. */ 1563 int netdev_register_kobject(struct net_device *ndev) 1564 { 1565 struct device *dev = &ndev->dev; 1566 const struct attribute_group **groups = ndev->sysfs_groups; 1567 int error = 0; 1568 1569 device_initialize(dev); 1570 dev->class = &net_class; 1571 dev->platform_data = ndev; 1572 dev->groups = groups; 1573 1574 dev_set_name(dev, "%s", ndev->name); 1575 1576 #ifdef CONFIG_SYSFS 1577 /* Allow for a device specific group */ 1578 if (*groups) 1579 groups++; 1580 1581 *groups++ = &netstat_group; 1582 1583 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211) 1584 if (ndev->ieee80211_ptr) 1585 *groups++ = &wireless_group; 1586 #if IS_ENABLED(CONFIG_WIRELESS_EXT) 1587 else if (ndev->wireless_handlers) 1588 *groups++ = &wireless_group; 1589 #endif 1590 #endif 1591 #endif /* CONFIG_SYSFS */ 1592 1593 error = device_add(dev); 1594 if (error) 1595 return error; 1596 1597 error = register_queue_kobjects(ndev); 1598 if (error) { 1599 device_del(dev); 1600 return error; 1601 } 1602 1603 pm_runtime_set_memalloc_noio(dev, true); 1604 1605 return error; 1606 } 1607 1608 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 1609 const void *ns) 1610 { 1611 return class_create_file_ns(&net_class, class_attr, ns); 1612 } 1613 EXPORT_SYMBOL(netdev_class_create_file_ns); 1614 1615 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 1616 const void *ns) 1617 { 1618 class_remove_file_ns(&net_class, class_attr, ns); 1619 } 1620 EXPORT_SYMBOL(netdev_class_remove_file_ns); 1621 1622 int __init netdev_kobject_init(void) 1623 { 1624 kobj_ns_type_register(&net_ns_type_operations); 1625 return class_register(&net_class); 1626 } 1627