1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void synchronize_net(void); 74 void netdev_set_default_ethtool_ops(struct net_device *dev, 75 const struct ethtool_ops *ops); 76 77 /* Backlog congestion levels */ 78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 79 #define NET_RX_DROP 1 /* packet dropped */ 80 81 #define MAX_NEST_DEV 8 82 83 /* 84 * Transmit return codes: transmit return codes originate from three different 85 * namespaces: 86 * 87 * - qdisc return codes 88 * - driver transmit return codes 89 * - errno values 90 * 91 * Drivers are allowed to return any one of those in their hard_start_xmit() 92 * function. Real network devices commonly used with qdiscs should only return 93 * the driver transmit return codes though - when qdiscs are used, the actual 94 * transmission happens asynchronously, so the value is not propagated to 95 * higher layers. Virtual network devices transmit synchronously; in this case 96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 97 * others are propagated to higher layers. 98 */ 99 100 /* qdisc ->enqueue() return codes. */ 101 #define NET_XMIT_SUCCESS 0x00 102 #define NET_XMIT_DROP 0x01 /* skb dropped */ 103 #define NET_XMIT_CN 0x02 /* congestion notification */ 104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 105 106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 107 * indicates that the device will soon be dropping packets, or already drops 108 * some packets of the same priority; prompting us to send less aggressively. */ 109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 111 112 /* Driver transmit return codes */ 113 #define NETDEV_TX_MASK 0xf0 114 115 enum netdev_tx { 116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 117 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 119 }; 120 typedef enum netdev_tx netdev_tx_t; 121 122 /* 123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 125 */ 126 static inline bool dev_xmit_complete(int rc) 127 { 128 /* 129 * Positive cases with an skb consumed by a driver: 130 * - successful transmission (rc == NETDEV_TX_OK) 131 * - error while transmitting (rc < 0) 132 * - error while queueing to a different device (rc & NET_XMIT_MASK) 133 */ 134 if (likely(rc < NET_XMIT_MASK)) 135 return true; 136 137 return false; 138 } 139 140 /* 141 * Compute the worst-case header length according to the protocols 142 * used. 143 */ 144 145 #if defined(CONFIG_HYPERV_NET) 146 # define LL_MAX_HEADER 128 147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 148 # if defined(CONFIG_MAC80211_MESH) 149 # define LL_MAX_HEADER 128 150 # else 151 # define LL_MAX_HEADER 96 152 # endif 153 #else 154 # define LL_MAX_HEADER 32 155 #endif 156 157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 159 #define MAX_HEADER LL_MAX_HEADER 160 #else 161 #define MAX_HEADER (LL_MAX_HEADER + 48) 162 #endif 163 164 /* 165 * Old network device statistics. Fields are native words 166 * (unsigned long) so they can be read and written atomically. 167 */ 168 169 struct net_device_stats { 170 unsigned long rx_packets; 171 unsigned long tx_packets; 172 unsigned long rx_bytes; 173 unsigned long tx_bytes; 174 unsigned long rx_errors; 175 unsigned long tx_errors; 176 unsigned long rx_dropped; 177 unsigned long tx_dropped; 178 unsigned long multicast; 179 unsigned long collisions; 180 unsigned long rx_length_errors; 181 unsigned long rx_over_errors; 182 unsigned long rx_crc_errors; 183 unsigned long rx_frame_errors; 184 unsigned long rx_fifo_errors; 185 unsigned long rx_missed_errors; 186 unsigned long tx_aborted_errors; 187 unsigned long tx_carrier_errors; 188 unsigned long tx_fifo_errors; 189 unsigned long tx_heartbeat_errors; 190 unsigned long tx_window_errors; 191 unsigned long rx_compressed; 192 unsigned long tx_compressed; 193 }; 194 195 196 #include <linux/cache.h> 197 #include <linux/skbuff.h> 198 199 #ifdef CONFIG_RPS 200 #include <linux/static_key.h> 201 extern struct static_key_false rps_needed; 202 extern struct static_key_false rfs_needed; 203 #endif 204 205 struct neighbour; 206 struct neigh_parms; 207 struct sk_buff; 208 209 struct netdev_hw_addr { 210 struct list_head list; 211 unsigned char addr[MAX_ADDR_LEN]; 212 unsigned char type; 213 #define NETDEV_HW_ADDR_T_LAN 1 214 #define NETDEV_HW_ADDR_T_SAN 2 215 #define NETDEV_HW_ADDR_T_UNICAST 3 216 #define NETDEV_HW_ADDR_T_MULTICAST 4 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 }; 351 352 enum { 353 NAPI_STATE_SCHED, /* Poll is scheduled */ 354 NAPI_STATE_MISSED, /* reschedule a napi */ 355 NAPI_STATE_DISABLE, /* Disable pending */ 356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 357 NAPI_STATE_LISTED, /* NAPI added to system lists */ 358 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 359 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 360 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 361 }; 362 363 enum { 364 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 365 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 366 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 367 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 368 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 369 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 370 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 371 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 372 }; 373 374 enum gro_result { 375 GRO_MERGED, 376 GRO_MERGED_FREE, 377 GRO_HELD, 378 GRO_NORMAL, 379 GRO_DROP, 380 GRO_CONSUMED, 381 }; 382 typedef enum gro_result gro_result_t; 383 384 /* 385 * enum rx_handler_result - Possible return values for rx_handlers. 386 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 387 * further. 388 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 389 * case skb->dev was changed by rx_handler. 390 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 391 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 392 * 393 * rx_handlers are functions called from inside __netif_receive_skb(), to do 394 * special processing of the skb, prior to delivery to protocol handlers. 395 * 396 * Currently, a net_device can only have a single rx_handler registered. Trying 397 * to register a second rx_handler will return -EBUSY. 398 * 399 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 400 * To unregister a rx_handler on a net_device, use 401 * netdev_rx_handler_unregister(). 402 * 403 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 404 * do with the skb. 405 * 406 * If the rx_handler consumed the skb in some way, it should return 407 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 408 * the skb to be delivered in some other way. 409 * 410 * If the rx_handler changed skb->dev, to divert the skb to another 411 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 412 * new device will be called if it exists. 413 * 414 * If the rx_handler decides the skb should be ignored, it should return 415 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 416 * are registered on exact device (ptype->dev == skb->dev). 417 * 418 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 419 * delivered, it should return RX_HANDLER_PASS. 420 * 421 * A device without a registered rx_handler will behave as if rx_handler 422 * returned RX_HANDLER_PASS. 423 */ 424 425 enum rx_handler_result { 426 RX_HANDLER_CONSUMED, 427 RX_HANDLER_ANOTHER, 428 RX_HANDLER_EXACT, 429 RX_HANDLER_PASS, 430 }; 431 typedef enum rx_handler_result rx_handler_result_t; 432 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 433 434 void __napi_schedule(struct napi_struct *n); 435 void __napi_schedule_irqoff(struct napi_struct *n); 436 437 static inline bool napi_disable_pending(struct napi_struct *n) 438 { 439 return test_bit(NAPI_STATE_DISABLE, &n->state); 440 } 441 442 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 443 { 444 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 445 } 446 447 bool napi_schedule_prep(struct napi_struct *n); 448 449 /** 450 * napi_schedule - schedule NAPI poll 451 * @n: NAPI context 452 * 453 * Schedule NAPI poll routine to be called if it is not already 454 * running. 455 */ 456 static inline void napi_schedule(struct napi_struct *n) 457 { 458 if (napi_schedule_prep(n)) 459 __napi_schedule(n); 460 } 461 462 /** 463 * napi_schedule_irqoff - schedule NAPI poll 464 * @n: NAPI context 465 * 466 * Variant of napi_schedule(), assuming hard irqs are masked. 467 */ 468 static inline void napi_schedule_irqoff(struct napi_struct *n) 469 { 470 if (napi_schedule_prep(n)) 471 __napi_schedule_irqoff(n); 472 } 473 474 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 475 static inline bool napi_reschedule(struct napi_struct *napi) 476 { 477 if (napi_schedule_prep(napi)) { 478 __napi_schedule(napi); 479 return true; 480 } 481 return false; 482 } 483 484 bool napi_complete_done(struct napi_struct *n, int work_done); 485 /** 486 * napi_complete - NAPI processing complete 487 * @n: NAPI context 488 * 489 * Mark NAPI processing as complete. 490 * Consider using napi_complete_done() instead. 491 * Return false if device should avoid rearming interrupts. 492 */ 493 static inline bool napi_complete(struct napi_struct *n) 494 { 495 return napi_complete_done(n, 0); 496 } 497 498 /** 499 * napi_disable - prevent NAPI from scheduling 500 * @n: NAPI context 501 * 502 * Stop NAPI from being scheduled on this context. 503 * Waits till any outstanding processing completes. 504 */ 505 void napi_disable(struct napi_struct *n); 506 507 /** 508 * napi_enable - enable NAPI scheduling 509 * @n: NAPI context 510 * 511 * Resume NAPI from being scheduled on this context. 512 * Must be paired with napi_disable. 513 */ 514 static inline void napi_enable(struct napi_struct *n) 515 { 516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 517 smp_mb__before_atomic(); 518 clear_bit(NAPI_STATE_SCHED, &n->state); 519 clear_bit(NAPI_STATE_NPSVC, &n->state); 520 } 521 522 /** 523 * napi_synchronize - wait until NAPI is not running 524 * @n: NAPI context 525 * 526 * Wait until NAPI is done being scheduled on this context. 527 * Waits till any outstanding processing completes but 528 * does not disable future activations. 529 */ 530 static inline void napi_synchronize(const struct napi_struct *n) 531 { 532 if (IS_ENABLED(CONFIG_SMP)) 533 while (test_bit(NAPI_STATE_SCHED, &n->state)) 534 msleep(1); 535 else 536 barrier(); 537 } 538 539 /** 540 * napi_if_scheduled_mark_missed - if napi is running, set the 541 * NAPIF_STATE_MISSED 542 * @n: NAPI context 543 * 544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 545 * NAPI is scheduled. 546 **/ 547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 548 { 549 unsigned long val, new; 550 551 do { 552 val = READ_ONCE(n->state); 553 if (val & NAPIF_STATE_DISABLE) 554 return true; 555 556 if (!(val & NAPIF_STATE_SCHED)) 557 return false; 558 559 new = val | NAPIF_STATE_MISSED; 560 } while (cmpxchg(&n->state, val, new) != val); 561 562 return true; 563 } 564 565 enum netdev_queue_state_t { 566 __QUEUE_STATE_DRV_XOFF, 567 __QUEUE_STATE_STACK_XOFF, 568 __QUEUE_STATE_FROZEN, 569 }; 570 571 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 572 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 573 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 574 575 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 579 QUEUE_STATE_FROZEN) 580 581 /* 582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 583 * netif_tx_* functions below are used to manipulate this flag. The 584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 585 * queue independently. The netif_xmit_*stopped functions below are called 586 * to check if the queue has been stopped by the driver or stack (either 587 * of the XOFF bits are set in the state). Drivers should not need to call 588 * netif_xmit*stopped functions, they should only be using netif_tx_*. 589 */ 590 591 struct netdev_queue { 592 /* 593 * read-mostly part 594 */ 595 struct net_device *dev; 596 struct Qdisc __rcu *qdisc; 597 struct Qdisc *qdisc_sleeping; 598 #ifdef CONFIG_SYSFS 599 struct kobject kobj; 600 #endif 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 602 int numa_node; 603 #endif 604 unsigned long tx_maxrate; 605 /* 606 * Number of TX timeouts for this queue 607 * (/sys/class/net/DEV/Q/trans_timeout) 608 */ 609 unsigned long trans_timeout; 610 611 /* Subordinate device that the queue has been assigned to */ 612 struct net_device *sb_dev; 613 #ifdef CONFIG_XDP_SOCKETS 614 struct xsk_buff_pool *pool; 615 #endif 616 /* 617 * write-mostly part 618 */ 619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 620 int xmit_lock_owner; 621 /* 622 * Time (in jiffies) of last Tx 623 */ 624 unsigned long trans_start; 625 626 unsigned long state; 627 628 #ifdef CONFIG_BQL 629 struct dql dql; 630 #endif 631 } ____cacheline_aligned_in_smp; 632 633 extern int sysctl_fb_tunnels_only_for_init_net; 634 extern int sysctl_devconf_inherit_init_net; 635 636 /* 637 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 638 * == 1 : For initns only 639 * == 2 : For none. 640 */ 641 static inline bool net_has_fallback_tunnels(const struct net *net) 642 { 643 return !IS_ENABLED(CONFIG_SYSCTL) || 644 !sysctl_fb_tunnels_only_for_init_net || 645 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 646 } 647 648 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 649 { 650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 651 return q->numa_node; 652 #else 653 return NUMA_NO_NODE; 654 #endif 655 } 656 657 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 658 { 659 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 660 q->numa_node = node; 661 #endif 662 } 663 664 #ifdef CONFIG_RPS 665 /* 666 * This structure holds an RPS map which can be of variable length. The 667 * map is an array of CPUs. 668 */ 669 struct rps_map { 670 unsigned int len; 671 struct rcu_head rcu; 672 u16 cpus[]; 673 }; 674 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 675 676 /* 677 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 678 * tail pointer for that CPU's input queue at the time of last enqueue, and 679 * a hardware filter index. 680 */ 681 struct rps_dev_flow { 682 u16 cpu; 683 u16 filter; 684 unsigned int last_qtail; 685 }; 686 #define RPS_NO_FILTER 0xffff 687 688 /* 689 * The rps_dev_flow_table structure contains a table of flow mappings. 690 */ 691 struct rps_dev_flow_table { 692 unsigned int mask; 693 struct rcu_head rcu; 694 struct rps_dev_flow flows[]; 695 }; 696 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 697 ((_num) * sizeof(struct rps_dev_flow))) 698 699 /* 700 * The rps_sock_flow_table contains mappings of flows to the last CPU 701 * on which they were processed by the application (set in recvmsg). 702 * Each entry is a 32bit value. Upper part is the high-order bits 703 * of flow hash, lower part is CPU number. 704 * rps_cpu_mask is used to partition the space, depending on number of 705 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 706 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 707 * meaning we use 32-6=26 bits for the hash. 708 */ 709 struct rps_sock_flow_table { 710 u32 mask; 711 712 u32 ents[] ____cacheline_aligned_in_smp; 713 }; 714 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 715 716 #define RPS_NO_CPU 0xffff 717 718 extern u32 rps_cpu_mask; 719 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 720 721 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 722 u32 hash) 723 { 724 if (table && hash) { 725 unsigned int index = hash & table->mask; 726 u32 val = hash & ~rps_cpu_mask; 727 728 /* We only give a hint, preemption can change CPU under us */ 729 val |= raw_smp_processor_id(); 730 731 if (table->ents[index] != val) 732 table->ents[index] = val; 733 } 734 } 735 736 #ifdef CONFIG_RFS_ACCEL 737 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 738 u16 filter_id); 739 #endif 740 #endif /* CONFIG_RPS */ 741 742 /* This structure contains an instance of an RX queue. */ 743 struct netdev_rx_queue { 744 #ifdef CONFIG_RPS 745 struct rps_map __rcu *rps_map; 746 struct rps_dev_flow_table __rcu *rps_flow_table; 747 #endif 748 struct kobject kobj; 749 struct net_device *dev; 750 struct xdp_rxq_info xdp_rxq; 751 #ifdef CONFIG_XDP_SOCKETS 752 struct xsk_buff_pool *pool; 753 #endif 754 } ____cacheline_aligned_in_smp; 755 756 /* 757 * RX queue sysfs structures and functions. 758 */ 759 struct rx_queue_attribute { 760 struct attribute attr; 761 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 762 ssize_t (*store)(struct netdev_rx_queue *queue, 763 const char *buf, size_t len); 764 }; 765 766 #ifdef CONFIG_XPS 767 /* 768 * This structure holds an XPS map which can be of variable length. The 769 * map is an array of queues. 770 */ 771 struct xps_map { 772 unsigned int len; 773 unsigned int alloc_len; 774 struct rcu_head rcu; 775 u16 queues[]; 776 }; 777 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 778 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 779 - sizeof(struct xps_map)) / sizeof(u16)) 780 781 /* 782 * This structure holds all XPS maps for device. Maps are indexed by CPU. 783 */ 784 struct xps_dev_maps { 785 struct rcu_head rcu; 786 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 787 }; 788 789 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 790 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 791 792 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 793 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 794 795 #endif /* CONFIG_XPS */ 796 797 #define TC_MAX_QUEUE 16 798 #define TC_BITMASK 15 799 /* HW offloaded queuing disciplines txq count and offset maps */ 800 struct netdev_tc_txq { 801 u16 count; 802 u16 offset; 803 }; 804 805 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 806 /* 807 * This structure is to hold information about the device 808 * configured to run FCoE protocol stack. 809 */ 810 struct netdev_fcoe_hbainfo { 811 char manufacturer[64]; 812 char serial_number[64]; 813 char hardware_version[64]; 814 char driver_version[64]; 815 char optionrom_version[64]; 816 char firmware_version[64]; 817 char model[256]; 818 char model_description[256]; 819 }; 820 #endif 821 822 #define MAX_PHYS_ITEM_ID_LEN 32 823 824 /* This structure holds a unique identifier to identify some 825 * physical item (port for example) used by a netdevice. 826 */ 827 struct netdev_phys_item_id { 828 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 829 unsigned char id_len; 830 }; 831 832 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 833 struct netdev_phys_item_id *b) 834 { 835 return a->id_len == b->id_len && 836 memcmp(a->id, b->id, a->id_len) == 0; 837 } 838 839 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 840 struct sk_buff *skb, 841 struct net_device *sb_dev); 842 843 enum tc_setup_type { 844 TC_SETUP_QDISC_MQPRIO, 845 TC_SETUP_CLSU32, 846 TC_SETUP_CLSFLOWER, 847 TC_SETUP_CLSMATCHALL, 848 TC_SETUP_CLSBPF, 849 TC_SETUP_BLOCK, 850 TC_SETUP_QDISC_CBS, 851 TC_SETUP_QDISC_RED, 852 TC_SETUP_QDISC_PRIO, 853 TC_SETUP_QDISC_MQ, 854 TC_SETUP_QDISC_ETF, 855 TC_SETUP_ROOT_QDISC, 856 TC_SETUP_QDISC_GRED, 857 TC_SETUP_QDISC_TAPRIO, 858 TC_SETUP_FT, 859 TC_SETUP_QDISC_ETS, 860 TC_SETUP_QDISC_TBF, 861 TC_SETUP_QDISC_FIFO, 862 }; 863 864 /* These structures hold the attributes of bpf state that are being passed 865 * to the netdevice through the bpf op. 866 */ 867 enum bpf_netdev_command { 868 /* Set or clear a bpf program used in the earliest stages of packet 869 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 870 * is responsible for calling bpf_prog_put on any old progs that are 871 * stored. In case of error, the callee need not release the new prog 872 * reference, but on success it takes ownership and must bpf_prog_put 873 * when it is no longer used. 874 */ 875 XDP_SETUP_PROG, 876 XDP_SETUP_PROG_HW, 877 /* BPF program for offload callbacks, invoked at program load time. */ 878 BPF_OFFLOAD_MAP_ALLOC, 879 BPF_OFFLOAD_MAP_FREE, 880 XDP_SETUP_XSK_POOL, 881 }; 882 883 struct bpf_prog_offload_ops; 884 struct netlink_ext_ack; 885 struct xdp_umem; 886 struct xdp_dev_bulk_queue; 887 struct bpf_xdp_link; 888 889 enum bpf_xdp_mode { 890 XDP_MODE_SKB = 0, 891 XDP_MODE_DRV = 1, 892 XDP_MODE_HW = 2, 893 __MAX_XDP_MODE 894 }; 895 896 struct bpf_xdp_entity { 897 struct bpf_prog *prog; 898 struct bpf_xdp_link *link; 899 }; 900 901 struct netdev_bpf { 902 enum bpf_netdev_command command; 903 union { 904 /* XDP_SETUP_PROG */ 905 struct { 906 u32 flags; 907 struct bpf_prog *prog; 908 struct netlink_ext_ack *extack; 909 }; 910 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 911 struct { 912 struct bpf_offloaded_map *offmap; 913 }; 914 /* XDP_SETUP_XSK_POOL */ 915 struct { 916 struct xsk_buff_pool *pool; 917 u16 queue_id; 918 } xsk; 919 }; 920 }; 921 922 /* Flags for ndo_xsk_wakeup. */ 923 #define XDP_WAKEUP_RX (1 << 0) 924 #define XDP_WAKEUP_TX (1 << 1) 925 926 #ifdef CONFIG_XFRM_OFFLOAD 927 struct xfrmdev_ops { 928 int (*xdo_dev_state_add) (struct xfrm_state *x); 929 void (*xdo_dev_state_delete) (struct xfrm_state *x); 930 void (*xdo_dev_state_free) (struct xfrm_state *x); 931 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 932 struct xfrm_state *x); 933 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 934 }; 935 #endif 936 937 struct dev_ifalias { 938 struct rcu_head rcuhead; 939 char ifalias[]; 940 }; 941 942 struct devlink; 943 struct tlsdev_ops; 944 945 struct netdev_name_node { 946 struct hlist_node hlist; 947 struct list_head list; 948 struct net_device *dev; 949 const char *name; 950 }; 951 952 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 953 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 954 955 struct netdev_net_notifier { 956 struct list_head list; 957 struct notifier_block *nb; 958 }; 959 960 /* 961 * This structure defines the management hooks for network devices. 962 * The following hooks can be defined; unless noted otherwise, they are 963 * optional and can be filled with a null pointer. 964 * 965 * int (*ndo_init)(struct net_device *dev); 966 * This function is called once when a network device is registered. 967 * The network device can use this for any late stage initialization 968 * or semantic validation. It can fail with an error code which will 969 * be propagated back to register_netdev. 970 * 971 * void (*ndo_uninit)(struct net_device *dev); 972 * This function is called when device is unregistered or when registration 973 * fails. It is not called if init fails. 974 * 975 * int (*ndo_open)(struct net_device *dev); 976 * This function is called when a network device transitions to the up 977 * state. 978 * 979 * int (*ndo_stop)(struct net_device *dev); 980 * This function is called when a network device transitions to the down 981 * state. 982 * 983 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 984 * struct net_device *dev); 985 * Called when a packet needs to be transmitted. 986 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 987 * the queue before that can happen; it's for obsolete devices and weird 988 * corner cases, but the stack really does a non-trivial amount 989 * of useless work if you return NETDEV_TX_BUSY. 990 * Required; cannot be NULL. 991 * 992 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 993 * struct net_device *dev 994 * netdev_features_t features); 995 * Called by core transmit path to determine if device is capable of 996 * performing offload operations on a given packet. This is to give 997 * the device an opportunity to implement any restrictions that cannot 998 * be otherwise expressed by feature flags. The check is called with 999 * the set of features that the stack has calculated and it returns 1000 * those the driver believes to be appropriate. 1001 * 1002 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1003 * struct net_device *sb_dev); 1004 * Called to decide which queue to use when device supports multiple 1005 * transmit queues. 1006 * 1007 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1008 * This function is called to allow device receiver to make 1009 * changes to configuration when multicast or promiscuous is enabled. 1010 * 1011 * void (*ndo_set_rx_mode)(struct net_device *dev); 1012 * This function is called device changes address list filtering. 1013 * If driver handles unicast address filtering, it should set 1014 * IFF_UNICAST_FLT in its priv_flags. 1015 * 1016 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1017 * This function is called when the Media Access Control address 1018 * needs to be changed. If this interface is not defined, the 1019 * MAC address can not be changed. 1020 * 1021 * int (*ndo_validate_addr)(struct net_device *dev); 1022 * Test if Media Access Control address is valid for the device. 1023 * 1024 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1025 * Called when a user requests an ioctl which can't be handled by 1026 * the generic interface code. If not defined ioctls return 1027 * not supported error code. 1028 * 1029 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1030 * Used to set network devices bus interface parameters. This interface 1031 * is retained for legacy reasons; new devices should use the bus 1032 * interface (PCI) for low level management. 1033 * 1034 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1035 * Called when a user wants to change the Maximum Transfer Unit 1036 * of a device. 1037 * 1038 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1039 * Callback used when the transmitter has not made any progress 1040 * for dev->watchdog ticks. 1041 * 1042 * void (*ndo_get_stats64)(struct net_device *dev, 1043 * struct rtnl_link_stats64 *storage); 1044 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1045 * Called when a user wants to get the network device usage 1046 * statistics. Drivers must do one of the following: 1047 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1048 * rtnl_link_stats64 structure passed by the caller. 1049 * 2. Define @ndo_get_stats to update a net_device_stats structure 1050 * (which should normally be dev->stats) and return a pointer to 1051 * it. The structure may be changed asynchronously only if each 1052 * field is written atomically. 1053 * 3. Update dev->stats asynchronously and atomically, and define 1054 * neither operation. 1055 * 1056 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1057 * Return true if this device supports offload stats of this attr_id. 1058 * 1059 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1060 * void *attr_data) 1061 * Get statistics for offload operations by attr_id. Write it into the 1062 * attr_data pointer. 1063 * 1064 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1065 * If device supports VLAN filtering this function is called when a 1066 * VLAN id is registered. 1067 * 1068 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1069 * If device supports VLAN filtering this function is called when a 1070 * VLAN id is unregistered. 1071 * 1072 * void (*ndo_poll_controller)(struct net_device *dev); 1073 * 1074 * SR-IOV management functions. 1075 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1076 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1077 * u8 qos, __be16 proto); 1078 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1079 * int max_tx_rate); 1080 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1081 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1082 * int (*ndo_get_vf_config)(struct net_device *dev, 1083 * int vf, struct ifla_vf_info *ivf); 1084 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1085 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1086 * struct nlattr *port[]); 1087 * 1088 * Enable or disable the VF ability to query its RSS Redirection Table and 1089 * Hash Key. This is needed since on some devices VF share this information 1090 * with PF and querying it may introduce a theoretical security risk. 1091 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1092 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1093 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1094 * void *type_data); 1095 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1096 * This is always called from the stack with the rtnl lock held and netif 1097 * tx queues stopped. This allows the netdevice to perform queue 1098 * management safely. 1099 * 1100 * Fiber Channel over Ethernet (FCoE) offload functions. 1101 * int (*ndo_fcoe_enable)(struct net_device *dev); 1102 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1103 * so the underlying device can perform whatever needed configuration or 1104 * initialization to support acceleration of FCoE traffic. 1105 * 1106 * int (*ndo_fcoe_disable)(struct net_device *dev); 1107 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1108 * so the underlying device can perform whatever needed clean-ups to 1109 * stop supporting acceleration of FCoE traffic. 1110 * 1111 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1112 * struct scatterlist *sgl, unsigned int sgc); 1113 * Called when the FCoE Initiator wants to initialize an I/O that 1114 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1115 * perform necessary setup and returns 1 to indicate the device is set up 1116 * successfully to perform DDP on this I/O, otherwise this returns 0. 1117 * 1118 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1119 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1120 * indicated by the FC exchange id 'xid', so the underlying device can 1121 * clean up and reuse resources for later DDP requests. 1122 * 1123 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1124 * struct scatterlist *sgl, unsigned int sgc); 1125 * Called when the FCoE Target wants to initialize an I/O that 1126 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1127 * perform necessary setup and returns 1 to indicate the device is set up 1128 * successfully to perform DDP on this I/O, otherwise this returns 0. 1129 * 1130 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1131 * struct netdev_fcoe_hbainfo *hbainfo); 1132 * Called when the FCoE Protocol stack wants information on the underlying 1133 * device. This information is utilized by the FCoE protocol stack to 1134 * register attributes with Fiber Channel management service as per the 1135 * FC-GS Fabric Device Management Information(FDMI) specification. 1136 * 1137 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1138 * Called when the underlying device wants to override default World Wide 1139 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1140 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1141 * protocol stack to use. 1142 * 1143 * RFS acceleration. 1144 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1145 * u16 rxq_index, u32 flow_id); 1146 * Set hardware filter for RFS. rxq_index is the target queue index; 1147 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1148 * Return the filter ID on success, or a negative error code. 1149 * 1150 * Slave management functions (for bridge, bonding, etc). 1151 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1152 * Called to make another netdev an underling. 1153 * 1154 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1155 * Called to release previously enslaved netdev. 1156 * 1157 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1158 * struct sk_buff *skb, 1159 * bool all_slaves); 1160 * Get the xmit slave of master device. If all_slaves is true, function 1161 * assume all the slaves can transmit. 1162 * 1163 * Feature/offload setting functions. 1164 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1165 * netdev_features_t features); 1166 * Adjusts the requested feature flags according to device-specific 1167 * constraints, and returns the resulting flags. Must not modify 1168 * the device state. 1169 * 1170 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1171 * Called to update device configuration to new features. Passed 1172 * feature set might be less than what was returned by ndo_fix_features()). 1173 * Must return >0 or -errno if it changed dev->features itself. 1174 * 1175 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1176 * struct net_device *dev, 1177 * const unsigned char *addr, u16 vid, u16 flags, 1178 * struct netlink_ext_ack *extack); 1179 * Adds an FDB entry to dev for addr. 1180 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1181 * struct net_device *dev, 1182 * const unsigned char *addr, u16 vid) 1183 * Deletes the FDB entry from dev coresponding to addr. 1184 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1185 * struct net_device *dev, struct net_device *filter_dev, 1186 * int *idx) 1187 * Used to add FDB entries to dump requests. Implementers should add 1188 * entries to skb and update idx with the number of entries. 1189 * 1190 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1191 * u16 flags, struct netlink_ext_ack *extack) 1192 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1193 * struct net_device *dev, u32 filter_mask, 1194 * int nlflags) 1195 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1196 * u16 flags); 1197 * 1198 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1199 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1200 * which do not represent real hardware may define this to allow their 1201 * userspace components to manage their virtual carrier state. Devices 1202 * that determine carrier state from physical hardware properties (eg 1203 * network cables) or protocol-dependent mechanisms (eg 1204 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1205 * 1206 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1207 * struct netdev_phys_item_id *ppid); 1208 * Called to get ID of physical port of this device. If driver does 1209 * not implement this, it is assumed that the hw is not able to have 1210 * multiple net devices on single physical port. 1211 * 1212 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1213 * struct netdev_phys_item_id *ppid) 1214 * Called to get the parent ID of the physical port of this device. 1215 * 1216 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1217 * struct udp_tunnel_info *ti); 1218 * Called by UDP tunnel to notify a driver about the UDP port and socket 1219 * address family that a UDP tunnel is listnening to. It is called only 1220 * when a new port starts listening. The operation is protected by the 1221 * RTNL. 1222 * 1223 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1224 * struct udp_tunnel_info *ti); 1225 * Called by UDP tunnel to notify the driver about a UDP port and socket 1226 * address family that the UDP tunnel is not listening to anymore. The 1227 * operation is protected by the RTNL. 1228 * 1229 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1230 * struct net_device *dev) 1231 * Called by upper layer devices to accelerate switching or other 1232 * station functionality into hardware. 'pdev is the lowerdev 1233 * to use for the offload and 'dev' is the net device that will 1234 * back the offload. Returns a pointer to the private structure 1235 * the upper layer will maintain. 1236 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1237 * Called by upper layer device to delete the station created 1238 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1239 * the station and priv is the structure returned by the add 1240 * operation. 1241 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1242 * int queue_index, u32 maxrate); 1243 * Called when a user wants to set a max-rate limitation of specific 1244 * TX queue. 1245 * int (*ndo_get_iflink)(const struct net_device *dev); 1246 * Called to get the iflink value of this device. 1247 * void (*ndo_change_proto_down)(struct net_device *dev, 1248 * bool proto_down); 1249 * This function is used to pass protocol port error state information 1250 * to the switch driver. The switch driver can react to the proto_down 1251 * by doing a phys down on the associated switch port. 1252 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1253 * This function is used to get egress tunnel information for given skb. 1254 * This is useful for retrieving outer tunnel header parameters while 1255 * sampling packet. 1256 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1257 * This function is used to specify the headroom that the skb must 1258 * consider when allocation skb during packet reception. Setting 1259 * appropriate rx headroom value allows avoiding skb head copy on 1260 * forward. Setting a negative value resets the rx headroom to the 1261 * default value. 1262 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1263 * This function is used to set or query state related to XDP on the 1264 * netdevice and manage BPF offload. See definition of 1265 * enum bpf_netdev_command for details. 1266 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1267 * u32 flags); 1268 * This function is used to submit @n XDP packets for transmit on a 1269 * netdevice. Returns number of frames successfully transmitted, frames 1270 * that got dropped are freed/returned via xdp_return_frame(). 1271 * Returns negative number, means general error invoking ndo, meaning 1272 * no frames were xmit'ed and core-caller will free all frames. 1273 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1274 * This function is used to wake up the softirq, ksoftirqd or kthread 1275 * responsible for sending and/or receiving packets on a specific 1276 * queue id bound to an AF_XDP socket. The flags field specifies if 1277 * only RX, only Tx, or both should be woken up using the flags 1278 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1279 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1280 * Get devlink port instance associated with a given netdev. 1281 * Called with a reference on the netdevice and devlink locks only, 1282 * rtnl_lock is not held. 1283 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1284 * int cmd); 1285 * Add, change, delete or get information on an IPv4 tunnel. 1286 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1287 * If a device is paired with a peer device, return the peer instance. 1288 * The caller must be under RCU read context. 1289 */ 1290 struct net_device_ops { 1291 int (*ndo_init)(struct net_device *dev); 1292 void (*ndo_uninit)(struct net_device *dev); 1293 int (*ndo_open)(struct net_device *dev); 1294 int (*ndo_stop)(struct net_device *dev); 1295 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1296 struct net_device *dev); 1297 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1298 struct net_device *dev, 1299 netdev_features_t features); 1300 u16 (*ndo_select_queue)(struct net_device *dev, 1301 struct sk_buff *skb, 1302 struct net_device *sb_dev); 1303 void (*ndo_change_rx_flags)(struct net_device *dev, 1304 int flags); 1305 void (*ndo_set_rx_mode)(struct net_device *dev); 1306 int (*ndo_set_mac_address)(struct net_device *dev, 1307 void *addr); 1308 int (*ndo_validate_addr)(struct net_device *dev); 1309 int (*ndo_do_ioctl)(struct net_device *dev, 1310 struct ifreq *ifr, int cmd); 1311 int (*ndo_set_config)(struct net_device *dev, 1312 struct ifmap *map); 1313 int (*ndo_change_mtu)(struct net_device *dev, 1314 int new_mtu); 1315 int (*ndo_neigh_setup)(struct net_device *dev, 1316 struct neigh_parms *); 1317 void (*ndo_tx_timeout) (struct net_device *dev, 1318 unsigned int txqueue); 1319 1320 void (*ndo_get_stats64)(struct net_device *dev, 1321 struct rtnl_link_stats64 *storage); 1322 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1323 int (*ndo_get_offload_stats)(int attr_id, 1324 const struct net_device *dev, 1325 void *attr_data); 1326 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1327 1328 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1329 __be16 proto, u16 vid); 1330 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1331 __be16 proto, u16 vid); 1332 #ifdef CONFIG_NET_POLL_CONTROLLER 1333 void (*ndo_poll_controller)(struct net_device *dev); 1334 int (*ndo_netpoll_setup)(struct net_device *dev, 1335 struct netpoll_info *info); 1336 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1337 #endif 1338 int (*ndo_set_vf_mac)(struct net_device *dev, 1339 int queue, u8 *mac); 1340 int (*ndo_set_vf_vlan)(struct net_device *dev, 1341 int queue, u16 vlan, 1342 u8 qos, __be16 proto); 1343 int (*ndo_set_vf_rate)(struct net_device *dev, 1344 int vf, int min_tx_rate, 1345 int max_tx_rate); 1346 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1347 int vf, bool setting); 1348 int (*ndo_set_vf_trust)(struct net_device *dev, 1349 int vf, bool setting); 1350 int (*ndo_get_vf_config)(struct net_device *dev, 1351 int vf, 1352 struct ifla_vf_info *ivf); 1353 int (*ndo_set_vf_link_state)(struct net_device *dev, 1354 int vf, int link_state); 1355 int (*ndo_get_vf_stats)(struct net_device *dev, 1356 int vf, 1357 struct ifla_vf_stats 1358 *vf_stats); 1359 int (*ndo_set_vf_port)(struct net_device *dev, 1360 int vf, 1361 struct nlattr *port[]); 1362 int (*ndo_get_vf_port)(struct net_device *dev, 1363 int vf, struct sk_buff *skb); 1364 int (*ndo_get_vf_guid)(struct net_device *dev, 1365 int vf, 1366 struct ifla_vf_guid *node_guid, 1367 struct ifla_vf_guid *port_guid); 1368 int (*ndo_set_vf_guid)(struct net_device *dev, 1369 int vf, u64 guid, 1370 int guid_type); 1371 int (*ndo_set_vf_rss_query_en)( 1372 struct net_device *dev, 1373 int vf, bool setting); 1374 int (*ndo_setup_tc)(struct net_device *dev, 1375 enum tc_setup_type type, 1376 void *type_data); 1377 #if IS_ENABLED(CONFIG_FCOE) 1378 int (*ndo_fcoe_enable)(struct net_device *dev); 1379 int (*ndo_fcoe_disable)(struct net_device *dev); 1380 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1381 u16 xid, 1382 struct scatterlist *sgl, 1383 unsigned int sgc); 1384 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1385 u16 xid); 1386 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1387 u16 xid, 1388 struct scatterlist *sgl, 1389 unsigned int sgc); 1390 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1391 struct netdev_fcoe_hbainfo *hbainfo); 1392 #endif 1393 1394 #if IS_ENABLED(CONFIG_LIBFCOE) 1395 #define NETDEV_FCOE_WWNN 0 1396 #define NETDEV_FCOE_WWPN 1 1397 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1398 u64 *wwn, int type); 1399 #endif 1400 1401 #ifdef CONFIG_RFS_ACCEL 1402 int (*ndo_rx_flow_steer)(struct net_device *dev, 1403 const struct sk_buff *skb, 1404 u16 rxq_index, 1405 u32 flow_id); 1406 #endif 1407 int (*ndo_add_slave)(struct net_device *dev, 1408 struct net_device *slave_dev, 1409 struct netlink_ext_ack *extack); 1410 int (*ndo_del_slave)(struct net_device *dev, 1411 struct net_device *slave_dev); 1412 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1413 struct sk_buff *skb, 1414 bool all_slaves); 1415 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1416 netdev_features_t features); 1417 int (*ndo_set_features)(struct net_device *dev, 1418 netdev_features_t features); 1419 int (*ndo_neigh_construct)(struct net_device *dev, 1420 struct neighbour *n); 1421 void (*ndo_neigh_destroy)(struct net_device *dev, 1422 struct neighbour *n); 1423 1424 int (*ndo_fdb_add)(struct ndmsg *ndm, 1425 struct nlattr *tb[], 1426 struct net_device *dev, 1427 const unsigned char *addr, 1428 u16 vid, 1429 u16 flags, 1430 struct netlink_ext_ack *extack); 1431 int (*ndo_fdb_del)(struct ndmsg *ndm, 1432 struct nlattr *tb[], 1433 struct net_device *dev, 1434 const unsigned char *addr, 1435 u16 vid); 1436 int (*ndo_fdb_dump)(struct sk_buff *skb, 1437 struct netlink_callback *cb, 1438 struct net_device *dev, 1439 struct net_device *filter_dev, 1440 int *idx); 1441 int (*ndo_fdb_get)(struct sk_buff *skb, 1442 struct nlattr *tb[], 1443 struct net_device *dev, 1444 const unsigned char *addr, 1445 u16 vid, u32 portid, u32 seq, 1446 struct netlink_ext_ack *extack); 1447 int (*ndo_bridge_setlink)(struct net_device *dev, 1448 struct nlmsghdr *nlh, 1449 u16 flags, 1450 struct netlink_ext_ack *extack); 1451 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1452 u32 pid, u32 seq, 1453 struct net_device *dev, 1454 u32 filter_mask, 1455 int nlflags); 1456 int (*ndo_bridge_dellink)(struct net_device *dev, 1457 struct nlmsghdr *nlh, 1458 u16 flags); 1459 int (*ndo_change_carrier)(struct net_device *dev, 1460 bool new_carrier); 1461 int (*ndo_get_phys_port_id)(struct net_device *dev, 1462 struct netdev_phys_item_id *ppid); 1463 int (*ndo_get_port_parent_id)(struct net_device *dev, 1464 struct netdev_phys_item_id *ppid); 1465 int (*ndo_get_phys_port_name)(struct net_device *dev, 1466 char *name, size_t len); 1467 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1468 struct udp_tunnel_info *ti); 1469 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1470 struct udp_tunnel_info *ti); 1471 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1472 struct net_device *dev); 1473 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1474 void *priv); 1475 1476 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1477 int queue_index, 1478 u32 maxrate); 1479 int (*ndo_get_iflink)(const struct net_device *dev); 1480 int (*ndo_change_proto_down)(struct net_device *dev, 1481 bool proto_down); 1482 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1483 struct sk_buff *skb); 1484 void (*ndo_set_rx_headroom)(struct net_device *dev, 1485 int needed_headroom); 1486 int (*ndo_bpf)(struct net_device *dev, 1487 struct netdev_bpf *bpf); 1488 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1489 struct xdp_frame **xdp, 1490 u32 flags); 1491 int (*ndo_xsk_wakeup)(struct net_device *dev, 1492 u32 queue_id, u32 flags); 1493 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1494 int (*ndo_tunnel_ctl)(struct net_device *dev, 1495 struct ip_tunnel_parm *p, int cmd); 1496 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1497 }; 1498 1499 /** 1500 * enum net_device_priv_flags - &struct net_device priv_flags 1501 * 1502 * These are the &struct net_device, they are only set internally 1503 * by drivers and used in the kernel. These flags are invisible to 1504 * userspace; this means that the order of these flags can change 1505 * during any kernel release. 1506 * 1507 * You should have a pretty good reason to be extending these flags. 1508 * 1509 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1510 * @IFF_EBRIDGE: Ethernet bridging device 1511 * @IFF_BONDING: bonding master or slave 1512 * @IFF_ISATAP: ISATAP interface (RFC4214) 1513 * @IFF_WAN_HDLC: WAN HDLC device 1514 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1515 * release skb->dst 1516 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1517 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1518 * @IFF_MACVLAN_PORT: device used as macvlan port 1519 * @IFF_BRIDGE_PORT: device used as bridge port 1520 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1521 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1522 * @IFF_UNICAST_FLT: Supports unicast filtering 1523 * @IFF_TEAM_PORT: device used as team port 1524 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1525 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1526 * change when it's running 1527 * @IFF_MACVLAN: Macvlan device 1528 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1529 * underlying stacked devices 1530 * @IFF_L3MDEV_MASTER: device is an L3 master device 1531 * @IFF_NO_QUEUE: device can run without qdisc attached 1532 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1533 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1534 * @IFF_TEAM: device is a team device 1535 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1536 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1537 * entity (i.e. the master device for bridged veth) 1538 * @IFF_MACSEC: device is a MACsec device 1539 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1540 * @IFF_FAILOVER: device is a failover master device 1541 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1542 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1543 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1544 */ 1545 enum netdev_priv_flags { 1546 IFF_802_1Q_VLAN = 1<<0, 1547 IFF_EBRIDGE = 1<<1, 1548 IFF_BONDING = 1<<2, 1549 IFF_ISATAP = 1<<3, 1550 IFF_WAN_HDLC = 1<<4, 1551 IFF_XMIT_DST_RELEASE = 1<<5, 1552 IFF_DONT_BRIDGE = 1<<6, 1553 IFF_DISABLE_NETPOLL = 1<<7, 1554 IFF_MACVLAN_PORT = 1<<8, 1555 IFF_BRIDGE_PORT = 1<<9, 1556 IFF_OVS_DATAPATH = 1<<10, 1557 IFF_TX_SKB_SHARING = 1<<11, 1558 IFF_UNICAST_FLT = 1<<12, 1559 IFF_TEAM_PORT = 1<<13, 1560 IFF_SUPP_NOFCS = 1<<14, 1561 IFF_LIVE_ADDR_CHANGE = 1<<15, 1562 IFF_MACVLAN = 1<<16, 1563 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1564 IFF_L3MDEV_MASTER = 1<<18, 1565 IFF_NO_QUEUE = 1<<19, 1566 IFF_OPENVSWITCH = 1<<20, 1567 IFF_L3MDEV_SLAVE = 1<<21, 1568 IFF_TEAM = 1<<22, 1569 IFF_RXFH_CONFIGURED = 1<<23, 1570 IFF_PHONY_HEADROOM = 1<<24, 1571 IFF_MACSEC = 1<<25, 1572 IFF_NO_RX_HANDLER = 1<<26, 1573 IFF_FAILOVER = 1<<27, 1574 IFF_FAILOVER_SLAVE = 1<<28, 1575 IFF_L3MDEV_RX_HANDLER = 1<<29, 1576 IFF_LIVE_RENAME_OK = 1<<30, 1577 }; 1578 1579 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1580 #define IFF_EBRIDGE IFF_EBRIDGE 1581 #define IFF_BONDING IFF_BONDING 1582 #define IFF_ISATAP IFF_ISATAP 1583 #define IFF_WAN_HDLC IFF_WAN_HDLC 1584 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1585 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1586 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1587 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1588 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1589 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1590 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1591 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1592 #define IFF_TEAM_PORT IFF_TEAM_PORT 1593 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1594 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1595 #define IFF_MACVLAN IFF_MACVLAN 1596 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1597 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1598 #define IFF_NO_QUEUE IFF_NO_QUEUE 1599 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1600 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1601 #define IFF_TEAM IFF_TEAM 1602 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1603 #define IFF_MACSEC IFF_MACSEC 1604 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1605 #define IFF_FAILOVER IFF_FAILOVER 1606 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1607 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1608 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1609 1610 /** 1611 * struct net_device - The DEVICE structure. 1612 * 1613 * Actually, this whole structure is a big mistake. It mixes I/O 1614 * data with strictly "high-level" data, and it has to know about 1615 * almost every data structure used in the INET module. 1616 * 1617 * @name: This is the first field of the "visible" part of this structure 1618 * (i.e. as seen by users in the "Space.c" file). It is the name 1619 * of the interface. 1620 * 1621 * @name_node: Name hashlist node 1622 * @ifalias: SNMP alias 1623 * @mem_end: Shared memory end 1624 * @mem_start: Shared memory start 1625 * @base_addr: Device I/O address 1626 * @irq: Device IRQ number 1627 * 1628 * @state: Generic network queuing layer state, see netdev_state_t 1629 * @dev_list: The global list of network devices 1630 * @napi_list: List entry used for polling NAPI devices 1631 * @unreg_list: List entry when we are unregistering the 1632 * device; see the function unregister_netdev 1633 * @close_list: List entry used when we are closing the device 1634 * @ptype_all: Device-specific packet handlers for all protocols 1635 * @ptype_specific: Device-specific, protocol-specific packet handlers 1636 * 1637 * @adj_list: Directly linked devices, like slaves for bonding 1638 * @features: Currently active device features 1639 * @hw_features: User-changeable features 1640 * 1641 * @wanted_features: User-requested features 1642 * @vlan_features: Mask of features inheritable by VLAN devices 1643 * 1644 * @hw_enc_features: Mask of features inherited by encapsulating devices 1645 * This field indicates what encapsulation 1646 * offloads the hardware is capable of doing, 1647 * and drivers will need to set them appropriately. 1648 * 1649 * @mpls_features: Mask of features inheritable by MPLS 1650 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1651 * 1652 * @ifindex: interface index 1653 * @group: The group the device belongs to 1654 * 1655 * @stats: Statistics struct, which was left as a legacy, use 1656 * rtnl_link_stats64 instead 1657 * 1658 * @rx_dropped: Dropped packets by core network, 1659 * do not use this in drivers 1660 * @tx_dropped: Dropped packets by core network, 1661 * do not use this in drivers 1662 * @rx_nohandler: nohandler dropped packets by core network on 1663 * inactive devices, do not use this in drivers 1664 * @carrier_up_count: Number of times the carrier has been up 1665 * @carrier_down_count: Number of times the carrier has been down 1666 * 1667 * @wireless_handlers: List of functions to handle Wireless Extensions, 1668 * instead of ioctl, 1669 * see <net/iw_handler.h> for details. 1670 * @wireless_data: Instance data managed by the core of wireless extensions 1671 * 1672 * @netdev_ops: Includes several pointers to callbacks, 1673 * if one wants to override the ndo_*() functions 1674 * @ethtool_ops: Management operations 1675 * @l3mdev_ops: Layer 3 master device operations 1676 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1677 * discovery handling. Necessary for e.g. 6LoWPAN. 1678 * @xfrmdev_ops: Transformation offload operations 1679 * @tlsdev_ops: Transport Layer Security offload operations 1680 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1681 * of Layer 2 headers. 1682 * 1683 * @flags: Interface flags (a la BSD) 1684 * @priv_flags: Like 'flags' but invisible to userspace, 1685 * see if.h for the definitions 1686 * @gflags: Global flags ( kept as legacy ) 1687 * @padded: How much padding added by alloc_netdev() 1688 * @operstate: RFC2863 operstate 1689 * @link_mode: Mapping policy to operstate 1690 * @if_port: Selectable AUI, TP, ... 1691 * @dma: DMA channel 1692 * @mtu: Interface MTU value 1693 * @min_mtu: Interface Minimum MTU value 1694 * @max_mtu: Interface Maximum MTU value 1695 * @type: Interface hardware type 1696 * @hard_header_len: Maximum hardware header length. 1697 * @min_header_len: Minimum hardware header length 1698 * 1699 * @needed_headroom: Extra headroom the hardware may need, but not in all 1700 * cases can this be guaranteed 1701 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1702 * cases can this be guaranteed. Some cases also use 1703 * LL_MAX_HEADER instead to allocate the skb 1704 * 1705 * interface address info: 1706 * 1707 * @perm_addr: Permanent hw address 1708 * @addr_assign_type: Hw address assignment type 1709 * @addr_len: Hardware address length 1710 * @upper_level: Maximum depth level of upper devices. 1711 * @lower_level: Maximum depth level of lower devices. 1712 * @neigh_priv_len: Used in neigh_alloc() 1713 * @dev_id: Used to differentiate devices that share 1714 * the same link layer address 1715 * @dev_port: Used to differentiate devices that share 1716 * the same function 1717 * @addr_list_lock: XXX: need comments on this one 1718 * @name_assign_type: network interface name assignment type 1719 * @uc_promisc: Counter that indicates promiscuous mode 1720 * has been enabled due to the need to listen to 1721 * additional unicast addresses in a device that 1722 * does not implement ndo_set_rx_mode() 1723 * @uc: unicast mac addresses 1724 * @mc: multicast mac addresses 1725 * @dev_addrs: list of device hw addresses 1726 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1727 * @promiscuity: Number of times the NIC is told to work in 1728 * promiscuous mode; if it becomes 0 the NIC will 1729 * exit promiscuous mode 1730 * @allmulti: Counter, enables or disables allmulticast mode 1731 * 1732 * @vlan_info: VLAN info 1733 * @dsa_ptr: dsa specific data 1734 * @tipc_ptr: TIPC specific data 1735 * @atalk_ptr: AppleTalk link 1736 * @ip_ptr: IPv4 specific data 1737 * @dn_ptr: DECnet specific data 1738 * @ip6_ptr: IPv6 specific data 1739 * @ax25_ptr: AX.25 specific data 1740 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1741 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1742 * device struct 1743 * @mpls_ptr: mpls_dev struct pointer 1744 * 1745 * @dev_addr: Hw address (before bcast, 1746 * because most packets are unicast) 1747 * 1748 * @_rx: Array of RX queues 1749 * @num_rx_queues: Number of RX queues 1750 * allocated at register_netdev() time 1751 * @real_num_rx_queues: Number of RX queues currently active in device 1752 * @xdp_prog: XDP sockets filter program pointer 1753 * @gro_flush_timeout: timeout for GRO layer in NAPI 1754 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1755 * allow to avoid NIC hard IRQ, on busy queues. 1756 * 1757 * @rx_handler: handler for received packets 1758 * @rx_handler_data: XXX: need comments on this one 1759 * @miniq_ingress: ingress/clsact qdisc specific data for 1760 * ingress processing 1761 * @ingress_queue: XXX: need comments on this one 1762 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1763 * @broadcast: hw bcast address 1764 * 1765 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1766 * indexed by RX queue number. Assigned by driver. 1767 * This must only be set if the ndo_rx_flow_steer 1768 * operation is defined 1769 * @index_hlist: Device index hash chain 1770 * 1771 * @_tx: Array of TX queues 1772 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1773 * @real_num_tx_queues: Number of TX queues currently active in device 1774 * @qdisc: Root qdisc from userspace point of view 1775 * @tx_queue_len: Max frames per queue allowed 1776 * @tx_global_lock: XXX: need comments on this one 1777 * @xdp_bulkq: XDP device bulk queue 1778 * @xps_cpus_map: all CPUs map for XPS device 1779 * @xps_rxqs_map: all RXQs map for XPS device 1780 * 1781 * @xps_maps: XXX: need comments on this one 1782 * @miniq_egress: clsact qdisc specific data for 1783 * egress processing 1784 * @qdisc_hash: qdisc hash table 1785 * @watchdog_timeo: Represents the timeout that is used by 1786 * the watchdog (see dev_watchdog()) 1787 * @watchdog_timer: List of timers 1788 * 1789 * @proto_down_reason: reason a netdev interface is held down 1790 * @pcpu_refcnt: Number of references to this device 1791 * @todo_list: Delayed register/unregister 1792 * @link_watch_list: XXX: need comments on this one 1793 * 1794 * @reg_state: Register/unregister state machine 1795 * @dismantle: Device is going to be freed 1796 * @rtnl_link_state: This enum represents the phases of creating 1797 * a new link 1798 * 1799 * @needs_free_netdev: Should unregister perform free_netdev? 1800 * @priv_destructor: Called from unregister 1801 * @npinfo: XXX: need comments on this one 1802 * @nd_net: Network namespace this network device is inside 1803 * 1804 * @ml_priv: Mid-layer private 1805 * @lstats: Loopback statistics 1806 * @tstats: Tunnel statistics 1807 * @dstats: Dummy statistics 1808 * @vstats: Virtual ethernet statistics 1809 * 1810 * @garp_port: GARP 1811 * @mrp_port: MRP 1812 * 1813 * @dev: Class/net/name entry 1814 * @sysfs_groups: Space for optional device, statistics and wireless 1815 * sysfs groups 1816 * 1817 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1818 * @rtnl_link_ops: Rtnl_link_ops 1819 * 1820 * @gso_max_size: Maximum size of generic segmentation offload 1821 * @gso_max_segs: Maximum number of segments that can be passed to the 1822 * NIC for GSO 1823 * 1824 * @dcbnl_ops: Data Center Bridging netlink ops 1825 * @num_tc: Number of traffic classes in the net device 1826 * @tc_to_txq: XXX: need comments on this one 1827 * @prio_tc_map: XXX: need comments on this one 1828 * 1829 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1830 * 1831 * @priomap: XXX: need comments on this one 1832 * @phydev: Physical device may attach itself 1833 * for hardware timestamping 1834 * @sfp_bus: attached &struct sfp_bus structure. 1835 * 1836 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1837 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1838 * 1839 * @proto_down: protocol port state information can be sent to the 1840 * switch driver and used to set the phys state of the 1841 * switch port. 1842 * 1843 * @wol_enabled: Wake-on-LAN is enabled 1844 * 1845 * @net_notifier_list: List of per-net netdev notifier block 1846 * that follow this device when it is moved 1847 * to another network namespace. 1848 * 1849 * @macsec_ops: MACsec offloading ops 1850 * 1851 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1852 * offload capabilities of the device 1853 * @udp_tunnel_nic: UDP tunnel offload state 1854 * @xdp_state: stores info on attached XDP BPF programs 1855 * 1856 * @nested_level: Used as as a parameter of spin_lock_nested() of 1857 * dev->addr_list_lock. 1858 * @unlink_list: As netif_addr_lock() can be called recursively, 1859 * keep a list of interfaces to be deleted. 1860 * 1861 * FIXME: cleanup struct net_device such that network protocol info 1862 * moves out. 1863 */ 1864 1865 struct net_device { 1866 char name[IFNAMSIZ]; 1867 struct netdev_name_node *name_node; 1868 struct dev_ifalias __rcu *ifalias; 1869 /* 1870 * I/O specific fields 1871 * FIXME: Merge these and struct ifmap into one 1872 */ 1873 unsigned long mem_end; 1874 unsigned long mem_start; 1875 unsigned long base_addr; 1876 int irq; 1877 1878 /* 1879 * Some hardware also needs these fields (state,dev_list, 1880 * napi_list,unreg_list,close_list) but they are not 1881 * part of the usual set specified in Space.c. 1882 */ 1883 1884 unsigned long state; 1885 1886 struct list_head dev_list; 1887 struct list_head napi_list; 1888 struct list_head unreg_list; 1889 struct list_head close_list; 1890 struct list_head ptype_all; 1891 struct list_head ptype_specific; 1892 1893 struct { 1894 struct list_head upper; 1895 struct list_head lower; 1896 } adj_list; 1897 1898 netdev_features_t features; 1899 netdev_features_t hw_features; 1900 netdev_features_t wanted_features; 1901 netdev_features_t vlan_features; 1902 netdev_features_t hw_enc_features; 1903 netdev_features_t mpls_features; 1904 netdev_features_t gso_partial_features; 1905 1906 int ifindex; 1907 int group; 1908 1909 struct net_device_stats stats; 1910 1911 atomic_long_t rx_dropped; 1912 atomic_long_t tx_dropped; 1913 atomic_long_t rx_nohandler; 1914 1915 /* Stats to monitor link on/off, flapping */ 1916 atomic_t carrier_up_count; 1917 atomic_t carrier_down_count; 1918 1919 #ifdef CONFIG_WIRELESS_EXT 1920 const struct iw_handler_def *wireless_handlers; 1921 struct iw_public_data *wireless_data; 1922 #endif 1923 const struct net_device_ops *netdev_ops; 1924 const struct ethtool_ops *ethtool_ops; 1925 #ifdef CONFIG_NET_L3_MASTER_DEV 1926 const struct l3mdev_ops *l3mdev_ops; 1927 #endif 1928 #if IS_ENABLED(CONFIG_IPV6) 1929 const struct ndisc_ops *ndisc_ops; 1930 #endif 1931 1932 #ifdef CONFIG_XFRM_OFFLOAD 1933 const struct xfrmdev_ops *xfrmdev_ops; 1934 #endif 1935 1936 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1937 const struct tlsdev_ops *tlsdev_ops; 1938 #endif 1939 1940 const struct header_ops *header_ops; 1941 1942 unsigned int flags; 1943 unsigned int priv_flags; 1944 1945 unsigned short gflags; 1946 unsigned short padded; 1947 1948 unsigned char operstate; 1949 unsigned char link_mode; 1950 1951 unsigned char if_port; 1952 unsigned char dma; 1953 1954 /* Note : dev->mtu is often read without holding a lock. 1955 * Writers usually hold RTNL. 1956 * It is recommended to use READ_ONCE() to annotate the reads, 1957 * and to use WRITE_ONCE() to annotate the writes. 1958 */ 1959 unsigned int mtu; 1960 unsigned int min_mtu; 1961 unsigned int max_mtu; 1962 unsigned short type; 1963 unsigned short hard_header_len; 1964 unsigned char min_header_len; 1965 unsigned char name_assign_type; 1966 1967 unsigned short needed_headroom; 1968 unsigned short needed_tailroom; 1969 1970 /* Interface address info. */ 1971 unsigned char perm_addr[MAX_ADDR_LEN]; 1972 unsigned char addr_assign_type; 1973 unsigned char addr_len; 1974 unsigned char upper_level; 1975 unsigned char lower_level; 1976 1977 unsigned short neigh_priv_len; 1978 unsigned short dev_id; 1979 unsigned short dev_port; 1980 spinlock_t addr_list_lock; 1981 1982 struct netdev_hw_addr_list uc; 1983 struct netdev_hw_addr_list mc; 1984 struct netdev_hw_addr_list dev_addrs; 1985 1986 #ifdef CONFIG_SYSFS 1987 struct kset *queues_kset; 1988 #endif 1989 #ifdef CONFIG_LOCKDEP 1990 struct list_head unlink_list; 1991 #endif 1992 unsigned int promiscuity; 1993 unsigned int allmulti; 1994 bool uc_promisc; 1995 #ifdef CONFIG_LOCKDEP 1996 unsigned char nested_level; 1997 #endif 1998 1999 2000 /* Protocol-specific pointers */ 2001 2002 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2003 struct vlan_info __rcu *vlan_info; 2004 #endif 2005 #if IS_ENABLED(CONFIG_NET_DSA) 2006 struct dsa_port *dsa_ptr; 2007 #endif 2008 #if IS_ENABLED(CONFIG_TIPC) 2009 struct tipc_bearer __rcu *tipc_ptr; 2010 #endif 2011 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 2012 void *atalk_ptr; 2013 #endif 2014 struct in_device __rcu *ip_ptr; 2015 #if IS_ENABLED(CONFIG_DECNET) 2016 struct dn_dev __rcu *dn_ptr; 2017 #endif 2018 struct inet6_dev __rcu *ip6_ptr; 2019 #if IS_ENABLED(CONFIG_AX25) 2020 void *ax25_ptr; 2021 #endif 2022 struct wireless_dev *ieee80211_ptr; 2023 struct wpan_dev *ieee802154_ptr; 2024 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2025 struct mpls_dev __rcu *mpls_ptr; 2026 #endif 2027 2028 /* 2029 * Cache lines mostly used on receive path (including eth_type_trans()) 2030 */ 2031 /* Interface address info used in eth_type_trans() */ 2032 unsigned char *dev_addr; 2033 2034 struct netdev_rx_queue *_rx; 2035 unsigned int num_rx_queues; 2036 unsigned int real_num_rx_queues; 2037 2038 struct bpf_prog __rcu *xdp_prog; 2039 unsigned long gro_flush_timeout; 2040 int napi_defer_hard_irqs; 2041 rx_handler_func_t __rcu *rx_handler; 2042 void __rcu *rx_handler_data; 2043 2044 #ifdef CONFIG_NET_CLS_ACT 2045 struct mini_Qdisc __rcu *miniq_ingress; 2046 #endif 2047 struct netdev_queue __rcu *ingress_queue; 2048 #ifdef CONFIG_NETFILTER_INGRESS 2049 struct nf_hook_entries __rcu *nf_hooks_ingress; 2050 #endif 2051 2052 unsigned char broadcast[MAX_ADDR_LEN]; 2053 #ifdef CONFIG_RFS_ACCEL 2054 struct cpu_rmap *rx_cpu_rmap; 2055 #endif 2056 struct hlist_node index_hlist; 2057 2058 /* 2059 * Cache lines mostly used on transmit path 2060 */ 2061 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2062 unsigned int num_tx_queues; 2063 unsigned int real_num_tx_queues; 2064 struct Qdisc *qdisc; 2065 unsigned int tx_queue_len; 2066 spinlock_t tx_global_lock; 2067 2068 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2069 2070 #ifdef CONFIG_XPS 2071 struct xps_dev_maps __rcu *xps_cpus_map; 2072 struct xps_dev_maps __rcu *xps_rxqs_map; 2073 #endif 2074 #ifdef CONFIG_NET_CLS_ACT 2075 struct mini_Qdisc __rcu *miniq_egress; 2076 #endif 2077 2078 #ifdef CONFIG_NET_SCHED 2079 DECLARE_HASHTABLE (qdisc_hash, 4); 2080 #endif 2081 /* These may be needed for future network-power-down code. */ 2082 struct timer_list watchdog_timer; 2083 int watchdog_timeo; 2084 2085 u32 proto_down_reason; 2086 2087 struct list_head todo_list; 2088 int __percpu *pcpu_refcnt; 2089 2090 struct list_head link_watch_list; 2091 2092 enum { NETREG_UNINITIALIZED=0, 2093 NETREG_REGISTERED, /* completed register_netdevice */ 2094 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2095 NETREG_UNREGISTERED, /* completed unregister todo */ 2096 NETREG_RELEASED, /* called free_netdev */ 2097 NETREG_DUMMY, /* dummy device for NAPI poll */ 2098 } reg_state:8; 2099 2100 bool dismantle; 2101 2102 enum { 2103 RTNL_LINK_INITIALIZED, 2104 RTNL_LINK_INITIALIZING, 2105 } rtnl_link_state:16; 2106 2107 bool needs_free_netdev; 2108 void (*priv_destructor)(struct net_device *dev); 2109 2110 #ifdef CONFIG_NETPOLL 2111 struct netpoll_info __rcu *npinfo; 2112 #endif 2113 2114 possible_net_t nd_net; 2115 2116 /* mid-layer private */ 2117 union { 2118 void *ml_priv; 2119 struct pcpu_lstats __percpu *lstats; 2120 struct pcpu_sw_netstats __percpu *tstats; 2121 struct pcpu_dstats __percpu *dstats; 2122 }; 2123 2124 #if IS_ENABLED(CONFIG_GARP) 2125 struct garp_port __rcu *garp_port; 2126 #endif 2127 #if IS_ENABLED(CONFIG_MRP) 2128 struct mrp_port __rcu *mrp_port; 2129 #endif 2130 2131 struct device dev; 2132 const struct attribute_group *sysfs_groups[4]; 2133 const struct attribute_group *sysfs_rx_queue_group; 2134 2135 const struct rtnl_link_ops *rtnl_link_ops; 2136 2137 /* for setting kernel sock attribute on TCP connection setup */ 2138 #define GSO_MAX_SIZE 65536 2139 unsigned int gso_max_size; 2140 #define GSO_MAX_SEGS 65535 2141 u16 gso_max_segs; 2142 2143 #ifdef CONFIG_DCB 2144 const struct dcbnl_rtnl_ops *dcbnl_ops; 2145 #endif 2146 s16 num_tc; 2147 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2148 u8 prio_tc_map[TC_BITMASK + 1]; 2149 2150 #if IS_ENABLED(CONFIG_FCOE) 2151 unsigned int fcoe_ddp_xid; 2152 #endif 2153 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2154 struct netprio_map __rcu *priomap; 2155 #endif 2156 struct phy_device *phydev; 2157 struct sfp_bus *sfp_bus; 2158 struct lock_class_key *qdisc_tx_busylock; 2159 struct lock_class_key *qdisc_running_key; 2160 bool proto_down; 2161 unsigned wol_enabled:1; 2162 2163 struct list_head net_notifier_list; 2164 2165 #if IS_ENABLED(CONFIG_MACSEC) 2166 /* MACsec management functions */ 2167 const struct macsec_ops *macsec_ops; 2168 #endif 2169 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2170 struct udp_tunnel_nic *udp_tunnel_nic; 2171 2172 /* protected by rtnl_lock */ 2173 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2174 }; 2175 #define to_net_dev(d) container_of(d, struct net_device, dev) 2176 2177 static inline bool netif_elide_gro(const struct net_device *dev) 2178 { 2179 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2180 return true; 2181 return false; 2182 } 2183 2184 #define NETDEV_ALIGN 32 2185 2186 static inline 2187 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2188 { 2189 return dev->prio_tc_map[prio & TC_BITMASK]; 2190 } 2191 2192 static inline 2193 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2194 { 2195 if (tc >= dev->num_tc) 2196 return -EINVAL; 2197 2198 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2199 return 0; 2200 } 2201 2202 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2203 void netdev_reset_tc(struct net_device *dev); 2204 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2205 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2206 2207 static inline 2208 int netdev_get_num_tc(struct net_device *dev) 2209 { 2210 return dev->num_tc; 2211 } 2212 2213 static inline void net_prefetch(void *p) 2214 { 2215 prefetch(p); 2216 #if L1_CACHE_BYTES < 128 2217 prefetch((u8 *)p + L1_CACHE_BYTES); 2218 #endif 2219 } 2220 2221 static inline void net_prefetchw(void *p) 2222 { 2223 prefetchw(p); 2224 #if L1_CACHE_BYTES < 128 2225 prefetchw((u8 *)p + L1_CACHE_BYTES); 2226 #endif 2227 } 2228 2229 void netdev_unbind_sb_channel(struct net_device *dev, 2230 struct net_device *sb_dev); 2231 int netdev_bind_sb_channel_queue(struct net_device *dev, 2232 struct net_device *sb_dev, 2233 u8 tc, u16 count, u16 offset); 2234 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2235 static inline int netdev_get_sb_channel(struct net_device *dev) 2236 { 2237 return max_t(int, -dev->num_tc, 0); 2238 } 2239 2240 static inline 2241 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2242 unsigned int index) 2243 { 2244 return &dev->_tx[index]; 2245 } 2246 2247 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2248 const struct sk_buff *skb) 2249 { 2250 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2251 } 2252 2253 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2254 void (*f)(struct net_device *, 2255 struct netdev_queue *, 2256 void *), 2257 void *arg) 2258 { 2259 unsigned int i; 2260 2261 for (i = 0; i < dev->num_tx_queues; i++) 2262 f(dev, &dev->_tx[i], arg); 2263 } 2264 2265 #define netdev_lockdep_set_classes(dev) \ 2266 { \ 2267 static struct lock_class_key qdisc_tx_busylock_key; \ 2268 static struct lock_class_key qdisc_running_key; \ 2269 static struct lock_class_key qdisc_xmit_lock_key; \ 2270 static struct lock_class_key dev_addr_list_lock_key; \ 2271 unsigned int i; \ 2272 \ 2273 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2274 (dev)->qdisc_running_key = &qdisc_running_key; \ 2275 lockdep_set_class(&(dev)->addr_list_lock, \ 2276 &dev_addr_list_lock_key); \ 2277 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2278 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2279 &qdisc_xmit_lock_key); \ 2280 } 2281 2282 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2283 struct net_device *sb_dev); 2284 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2285 struct sk_buff *skb, 2286 struct net_device *sb_dev); 2287 2288 /* returns the headroom that the master device needs to take in account 2289 * when forwarding to this dev 2290 */ 2291 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2292 { 2293 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2294 } 2295 2296 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2297 { 2298 if (dev->netdev_ops->ndo_set_rx_headroom) 2299 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2300 } 2301 2302 /* set the device rx headroom to the dev's default */ 2303 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2304 { 2305 netdev_set_rx_headroom(dev, -1); 2306 } 2307 2308 /* 2309 * Net namespace inlines 2310 */ 2311 static inline 2312 struct net *dev_net(const struct net_device *dev) 2313 { 2314 return read_pnet(&dev->nd_net); 2315 } 2316 2317 static inline 2318 void dev_net_set(struct net_device *dev, struct net *net) 2319 { 2320 write_pnet(&dev->nd_net, net); 2321 } 2322 2323 /** 2324 * netdev_priv - access network device private data 2325 * @dev: network device 2326 * 2327 * Get network device private data 2328 */ 2329 static inline void *netdev_priv(const struct net_device *dev) 2330 { 2331 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2332 } 2333 2334 /* Set the sysfs physical device reference for the network logical device 2335 * if set prior to registration will cause a symlink during initialization. 2336 */ 2337 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2338 2339 /* Set the sysfs device type for the network logical device to allow 2340 * fine-grained identification of different network device types. For 2341 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2342 */ 2343 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2344 2345 /* Default NAPI poll() weight 2346 * Device drivers are strongly advised to not use bigger value 2347 */ 2348 #define NAPI_POLL_WEIGHT 64 2349 2350 /** 2351 * netif_napi_add - initialize a NAPI context 2352 * @dev: network device 2353 * @napi: NAPI context 2354 * @poll: polling function 2355 * @weight: default weight 2356 * 2357 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2358 * *any* of the other NAPI-related functions. 2359 */ 2360 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2361 int (*poll)(struct napi_struct *, int), int weight); 2362 2363 /** 2364 * netif_tx_napi_add - initialize a NAPI context 2365 * @dev: network device 2366 * @napi: NAPI context 2367 * @poll: polling function 2368 * @weight: default weight 2369 * 2370 * This variant of netif_napi_add() should be used from drivers using NAPI 2371 * to exclusively poll a TX queue. 2372 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2373 */ 2374 static inline void netif_tx_napi_add(struct net_device *dev, 2375 struct napi_struct *napi, 2376 int (*poll)(struct napi_struct *, int), 2377 int weight) 2378 { 2379 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2380 netif_napi_add(dev, napi, poll, weight); 2381 } 2382 2383 /** 2384 * __netif_napi_del - remove a NAPI context 2385 * @napi: NAPI context 2386 * 2387 * Warning: caller must observe RCU grace period before freeing memory 2388 * containing @napi. Drivers might want to call this helper to combine 2389 * all the needed RCU grace periods into a single one. 2390 */ 2391 void __netif_napi_del(struct napi_struct *napi); 2392 2393 /** 2394 * netif_napi_del - remove a NAPI context 2395 * @napi: NAPI context 2396 * 2397 * netif_napi_del() removes a NAPI context from the network device NAPI list 2398 */ 2399 static inline void netif_napi_del(struct napi_struct *napi) 2400 { 2401 __netif_napi_del(napi); 2402 synchronize_net(); 2403 } 2404 2405 struct napi_gro_cb { 2406 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2407 void *frag0; 2408 2409 /* Length of frag0. */ 2410 unsigned int frag0_len; 2411 2412 /* This indicates where we are processing relative to skb->data. */ 2413 int data_offset; 2414 2415 /* This is non-zero if the packet cannot be merged with the new skb. */ 2416 u16 flush; 2417 2418 /* Save the IP ID here and check when we get to the transport layer */ 2419 u16 flush_id; 2420 2421 /* Number of segments aggregated. */ 2422 u16 count; 2423 2424 /* Start offset for remote checksum offload */ 2425 u16 gro_remcsum_start; 2426 2427 /* jiffies when first packet was created/queued */ 2428 unsigned long age; 2429 2430 /* Used in ipv6_gro_receive() and foo-over-udp */ 2431 u16 proto; 2432 2433 /* This is non-zero if the packet may be of the same flow. */ 2434 u8 same_flow:1; 2435 2436 /* Used in tunnel GRO receive */ 2437 u8 encap_mark:1; 2438 2439 /* GRO checksum is valid */ 2440 u8 csum_valid:1; 2441 2442 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2443 u8 csum_cnt:3; 2444 2445 /* Free the skb? */ 2446 u8 free:2; 2447 #define NAPI_GRO_FREE 1 2448 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2449 2450 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2451 u8 is_ipv6:1; 2452 2453 /* Used in GRE, set in fou/gue_gro_receive */ 2454 u8 is_fou:1; 2455 2456 /* Used to determine if flush_id can be ignored */ 2457 u8 is_atomic:1; 2458 2459 /* Number of gro_receive callbacks this packet already went through */ 2460 u8 recursion_counter:4; 2461 2462 /* GRO is done by frag_list pointer chaining. */ 2463 u8 is_flist:1; 2464 2465 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2466 __wsum csum; 2467 2468 /* used in skb_gro_receive() slow path */ 2469 struct sk_buff *last; 2470 }; 2471 2472 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2473 2474 #define GRO_RECURSION_LIMIT 15 2475 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2476 { 2477 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2478 } 2479 2480 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2481 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2482 struct list_head *head, 2483 struct sk_buff *skb) 2484 { 2485 if (unlikely(gro_recursion_inc_test(skb))) { 2486 NAPI_GRO_CB(skb)->flush |= 1; 2487 return NULL; 2488 } 2489 2490 return cb(head, skb); 2491 } 2492 2493 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2494 struct sk_buff *); 2495 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2496 struct sock *sk, 2497 struct list_head *head, 2498 struct sk_buff *skb) 2499 { 2500 if (unlikely(gro_recursion_inc_test(skb))) { 2501 NAPI_GRO_CB(skb)->flush |= 1; 2502 return NULL; 2503 } 2504 2505 return cb(sk, head, skb); 2506 } 2507 2508 struct packet_type { 2509 __be16 type; /* This is really htons(ether_type). */ 2510 bool ignore_outgoing; 2511 struct net_device *dev; /* NULL is wildcarded here */ 2512 int (*func) (struct sk_buff *, 2513 struct net_device *, 2514 struct packet_type *, 2515 struct net_device *); 2516 void (*list_func) (struct list_head *, 2517 struct packet_type *, 2518 struct net_device *); 2519 bool (*id_match)(struct packet_type *ptype, 2520 struct sock *sk); 2521 void *af_packet_priv; 2522 struct list_head list; 2523 }; 2524 2525 struct offload_callbacks { 2526 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2527 netdev_features_t features); 2528 struct sk_buff *(*gro_receive)(struct list_head *head, 2529 struct sk_buff *skb); 2530 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2531 }; 2532 2533 struct packet_offload { 2534 __be16 type; /* This is really htons(ether_type). */ 2535 u16 priority; 2536 struct offload_callbacks callbacks; 2537 struct list_head list; 2538 }; 2539 2540 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2541 struct pcpu_sw_netstats { 2542 u64 rx_packets; 2543 u64 rx_bytes; 2544 u64 tx_packets; 2545 u64 tx_bytes; 2546 struct u64_stats_sync syncp; 2547 } __aligned(4 * sizeof(u64)); 2548 2549 struct pcpu_lstats { 2550 u64_stats_t packets; 2551 u64_stats_t bytes; 2552 struct u64_stats_sync syncp; 2553 } __aligned(2 * sizeof(u64)); 2554 2555 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2556 2557 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2558 { 2559 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2560 2561 u64_stats_update_begin(&tstats->syncp); 2562 tstats->rx_bytes += len; 2563 tstats->rx_packets++; 2564 u64_stats_update_end(&tstats->syncp); 2565 } 2566 2567 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2568 unsigned int packets, 2569 unsigned int len) 2570 { 2571 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2572 2573 u64_stats_update_begin(&tstats->syncp); 2574 tstats->tx_bytes += len; 2575 tstats->tx_packets += packets; 2576 u64_stats_update_end(&tstats->syncp); 2577 } 2578 2579 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2580 { 2581 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2582 2583 u64_stats_update_begin(&lstats->syncp); 2584 u64_stats_add(&lstats->bytes, len); 2585 u64_stats_inc(&lstats->packets); 2586 u64_stats_update_end(&lstats->syncp); 2587 } 2588 2589 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2590 ({ \ 2591 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2592 if (pcpu_stats) { \ 2593 int __cpu; \ 2594 for_each_possible_cpu(__cpu) { \ 2595 typeof(type) *stat; \ 2596 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2597 u64_stats_init(&stat->syncp); \ 2598 } \ 2599 } \ 2600 pcpu_stats; \ 2601 }) 2602 2603 #define netdev_alloc_pcpu_stats(type) \ 2604 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2605 2606 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2607 ({ \ 2608 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2609 if (pcpu_stats) { \ 2610 int __cpu; \ 2611 for_each_possible_cpu(__cpu) { \ 2612 typeof(type) *stat; \ 2613 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2614 u64_stats_init(&stat->syncp); \ 2615 } \ 2616 } \ 2617 pcpu_stats; \ 2618 }) 2619 2620 enum netdev_lag_tx_type { 2621 NETDEV_LAG_TX_TYPE_UNKNOWN, 2622 NETDEV_LAG_TX_TYPE_RANDOM, 2623 NETDEV_LAG_TX_TYPE_BROADCAST, 2624 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2625 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2626 NETDEV_LAG_TX_TYPE_HASH, 2627 }; 2628 2629 enum netdev_lag_hash { 2630 NETDEV_LAG_HASH_NONE, 2631 NETDEV_LAG_HASH_L2, 2632 NETDEV_LAG_HASH_L34, 2633 NETDEV_LAG_HASH_L23, 2634 NETDEV_LAG_HASH_E23, 2635 NETDEV_LAG_HASH_E34, 2636 NETDEV_LAG_HASH_UNKNOWN, 2637 }; 2638 2639 struct netdev_lag_upper_info { 2640 enum netdev_lag_tx_type tx_type; 2641 enum netdev_lag_hash hash_type; 2642 }; 2643 2644 struct netdev_lag_lower_state_info { 2645 u8 link_up : 1, 2646 tx_enabled : 1; 2647 }; 2648 2649 #include <linux/notifier.h> 2650 2651 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2652 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2653 * adding new types. 2654 */ 2655 enum netdev_cmd { 2656 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2657 NETDEV_DOWN, 2658 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2659 detected a hardware crash and restarted 2660 - we can use this eg to kick tcp sessions 2661 once done */ 2662 NETDEV_CHANGE, /* Notify device state change */ 2663 NETDEV_REGISTER, 2664 NETDEV_UNREGISTER, 2665 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2666 NETDEV_CHANGEADDR, /* notify after the address change */ 2667 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2668 NETDEV_GOING_DOWN, 2669 NETDEV_CHANGENAME, 2670 NETDEV_FEAT_CHANGE, 2671 NETDEV_BONDING_FAILOVER, 2672 NETDEV_PRE_UP, 2673 NETDEV_PRE_TYPE_CHANGE, 2674 NETDEV_POST_TYPE_CHANGE, 2675 NETDEV_POST_INIT, 2676 NETDEV_RELEASE, 2677 NETDEV_NOTIFY_PEERS, 2678 NETDEV_JOIN, 2679 NETDEV_CHANGEUPPER, 2680 NETDEV_RESEND_IGMP, 2681 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2682 NETDEV_CHANGEINFODATA, 2683 NETDEV_BONDING_INFO, 2684 NETDEV_PRECHANGEUPPER, 2685 NETDEV_CHANGELOWERSTATE, 2686 NETDEV_UDP_TUNNEL_PUSH_INFO, 2687 NETDEV_UDP_TUNNEL_DROP_INFO, 2688 NETDEV_CHANGE_TX_QUEUE_LEN, 2689 NETDEV_CVLAN_FILTER_PUSH_INFO, 2690 NETDEV_CVLAN_FILTER_DROP_INFO, 2691 NETDEV_SVLAN_FILTER_PUSH_INFO, 2692 NETDEV_SVLAN_FILTER_DROP_INFO, 2693 }; 2694 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2695 2696 int register_netdevice_notifier(struct notifier_block *nb); 2697 int unregister_netdevice_notifier(struct notifier_block *nb); 2698 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2699 int unregister_netdevice_notifier_net(struct net *net, 2700 struct notifier_block *nb); 2701 int register_netdevice_notifier_dev_net(struct net_device *dev, 2702 struct notifier_block *nb, 2703 struct netdev_net_notifier *nn); 2704 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2705 struct notifier_block *nb, 2706 struct netdev_net_notifier *nn); 2707 2708 struct netdev_notifier_info { 2709 struct net_device *dev; 2710 struct netlink_ext_ack *extack; 2711 }; 2712 2713 struct netdev_notifier_info_ext { 2714 struct netdev_notifier_info info; /* must be first */ 2715 union { 2716 u32 mtu; 2717 } ext; 2718 }; 2719 2720 struct netdev_notifier_change_info { 2721 struct netdev_notifier_info info; /* must be first */ 2722 unsigned int flags_changed; 2723 }; 2724 2725 struct netdev_notifier_changeupper_info { 2726 struct netdev_notifier_info info; /* must be first */ 2727 struct net_device *upper_dev; /* new upper dev */ 2728 bool master; /* is upper dev master */ 2729 bool linking; /* is the notification for link or unlink */ 2730 void *upper_info; /* upper dev info */ 2731 }; 2732 2733 struct netdev_notifier_changelowerstate_info { 2734 struct netdev_notifier_info info; /* must be first */ 2735 void *lower_state_info; /* is lower dev state */ 2736 }; 2737 2738 struct netdev_notifier_pre_changeaddr_info { 2739 struct netdev_notifier_info info; /* must be first */ 2740 const unsigned char *dev_addr; 2741 }; 2742 2743 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2744 struct net_device *dev) 2745 { 2746 info->dev = dev; 2747 info->extack = NULL; 2748 } 2749 2750 static inline struct net_device * 2751 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2752 { 2753 return info->dev; 2754 } 2755 2756 static inline struct netlink_ext_ack * 2757 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2758 { 2759 return info->extack; 2760 } 2761 2762 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2763 2764 2765 extern rwlock_t dev_base_lock; /* Device list lock */ 2766 2767 #define for_each_netdev(net, d) \ 2768 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2769 #define for_each_netdev_reverse(net, d) \ 2770 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2771 #define for_each_netdev_rcu(net, d) \ 2772 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2773 #define for_each_netdev_safe(net, d, n) \ 2774 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2775 #define for_each_netdev_continue(net, d) \ 2776 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2777 #define for_each_netdev_continue_reverse(net, d) \ 2778 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2779 dev_list) 2780 #define for_each_netdev_continue_rcu(net, d) \ 2781 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2782 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2783 for_each_netdev_rcu(&init_net, slave) \ 2784 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2785 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2786 2787 static inline struct net_device *next_net_device(struct net_device *dev) 2788 { 2789 struct list_head *lh; 2790 struct net *net; 2791 2792 net = dev_net(dev); 2793 lh = dev->dev_list.next; 2794 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2795 } 2796 2797 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2798 { 2799 struct list_head *lh; 2800 struct net *net; 2801 2802 net = dev_net(dev); 2803 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2804 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2805 } 2806 2807 static inline struct net_device *first_net_device(struct net *net) 2808 { 2809 return list_empty(&net->dev_base_head) ? NULL : 2810 net_device_entry(net->dev_base_head.next); 2811 } 2812 2813 static inline struct net_device *first_net_device_rcu(struct net *net) 2814 { 2815 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2816 2817 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2818 } 2819 2820 int netdev_boot_setup_check(struct net_device *dev); 2821 unsigned long netdev_boot_base(const char *prefix, int unit); 2822 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2823 const char *hwaddr); 2824 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2825 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2826 void dev_add_pack(struct packet_type *pt); 2827 void dev_remove_pack(struct packet_type *pt); 2828 void __dev_remove_pack(struct packet_type *pt); 2829 void dev_add_offload(struct packet_offload *po); 2830 void dev_remove_offload(struct packet_offload *po); 2831 2832 int dev_get_iflink(const struct net_device *dev); 2833 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2834 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2835 unsigned short mask); 2836 struct net_device *dev_get_by_name(struct net *net, const char *name); 2837 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2838 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2839 int dev_alloc_name(struct net_device *dev, const char *name); 2840 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2841 void dev_close(struct net_device *dev); 2842 void dev_close_many(struct list_head *head, bool unlink); 2843 void dev_disable_lro(struct net_device *dev); 2844 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2845 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2846 struct net_device *sb_dev); 2847 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2848 struct net_device *sb_dev); 2849 int dev_queue_xmit(struct sk_buff *skb); 2850 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2851 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2852 int register_netdevice(struct net_device *dev); 2853 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2854 void unregister_netdevice_many(struct list_head *head); 2855 static inline void unregister_netdevice(struct net_device *dev) 2856 { 2857 unregister_netdevice_queue(dev, NULL); 2858 } 2859 2860 int netdev_refcnt_read(const struct net_device *dev); 2861 void free_netdev(struct net_device *dev); 2862 void netdev_freemem(struct net_device *dev); 2863 int init_dummy_netdev(struct net_device *dev); 2864 2865 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2866 struct sk_buff *skb, 2867 bool all_slaves); 2868 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2869 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2870 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2871 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2872 int netdev_get_name(struct net *net, char *name, int ifindex); 2873 int dev_restart(struct net_device *dev); 2874 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2875 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2876 2877 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2878 { 2879 return NAPI_GRO_CB(skb)->data_offset; 2880 } 2881 2882 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2883 { 2884 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2885 } 2886 2887 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2888 { 2889 NAPI_GRO_CB(skb)->data_offset += len; 2890 } 2891 2892 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2893 unsigned int offset) 2894 { 2895 return NAPI_GRO_CB(skb)->frag0 + offset; 2896 } 2897 2898 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2899 { 2900 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2901 } 2902 2903 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2904 { 2905 NAPI_GRO_CB(skb)->frag0 = NULL; 2906 NAPI_GRO_CB(skb)->frag0_len = 0; 2907 } 2908 2909 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2910 unsigned int offset) 2911 { 2912 if (!pskb_may_pull(skb, hlen)) 2913 return NULL; 2914 2915 skb_gro_frag0_invalidate(skb); 2916 return skb->data + offset; 2917 } 2918 2919 static inline void *skb_gro_network_header(struct sk_buff *skb) 2920 { 2921 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2922 skb_network_offset(skb); 2923 } 2924 2925 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2926 const void *start, unsigned int len) 2927 { 2928 if (NAPI_GRO_CB(skb)->csum_valid) 2929 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2930 csum_partial(start, len, 0)); 2931 } 2932 2933 /* GRO checksum functions. These are logical equivalents of the normal 2934 * checksum functions (in skbuff.h) except that they operate on the GRO 2935 * offsets and fields in sk_buff. 2936 */ 2937 2938 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2939 2940 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2941 { 2942 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2943 } 2944 2945 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2946 bool zero_okay, 2947 __sum16 check) 2948 { 2949 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2950 skb_checksum_start_offset(skb) < 2951 skb_gro_offset(skb)) && 2952 !skb_at_gro_remcsum_start(skb) && 2953 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2954 (!zero_okay || check)); 2955 } 2956 2957 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2958 __wsum psum) 2959 { 2960 if (NAPI_GRO_CB(skb)->csum_valid && 2961 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2962 return 0; 2963 2964 NAPI_GRO_CB(skb)->csum = psum; 2965 2966 return __skb_gro_checksum_complete(skb); 2967 } 2968 2969 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2970 { 2971 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2972 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2973 NAPI_GRO_CB(skb)->csum_cnt--; 2974 } else { 2975 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2976 * verified a new top level checksum or an encapsulated one 2977 * during GRO. This saves work if we fallback to normal path. 2978 */ 2979 __skb_incr_checksum_unnecessary(skb); 2980 } 2981 } 2982 2983 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2984 compute_pseudo) \ 2985 ({ \ 2986 __sum16 __ret = 0; \ 2987 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2988 __ret = __skb_gro_checksum_validate_complete(skb, \ 2989 compute_pseudo(skb, proto)); \ 2990 if (!__ret) \ 2991 skb_gro_incr_csum_unnecessary(skb); \ 2992 __ret; \ 2993 }) 2994 2995 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2996 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2997 2998 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2999 compute_pseudo) \ 3000 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 3001 3002 #define skb_gro_checksum_simple_validate(skb) \ 3003 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 3004 3005 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3006 { 3007 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3008 !NAPI_GRO_CB(skb)->csum_valid); 3009 } 3010 3011 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3012 __wsum pseudo) 3013 { 3014 NAPI_GRO_CB(skb)->csum = ~pseudo; 3015 NAPI_GRO_CB(skb)->csum_valid = 1; 3016 } 3017 3018 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3019 do { \ 3020 if (__skb_gro_checksum_convert_check(skb)) \ 3021 __skb_gro_checksum_convert(skb, \ 3022 compute_pseudo(skb, proto)); \ 3023 } while (0) 3024 3025 struct gro_remcsum { 3026 int offset; 3027 __wsum delta; 3028 }; 3029 3030 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3031 { 3032 grc->offset = 0; 3033 grc->delta = 0; 3034 } 3035 3036 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3037 unsigned int off, size_t hdrlen, 3038 int start, int offset, 3039 struct gro_remcsum *grc, 3040 bool nopartial) 3041 { 3042 __wsum delta; 3043 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3044 3045 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3046 3047 if (!nopartial) { 3048 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3049 return ptr; 3050 } 3051 3052 ptr = skb_gro_header_fast(skb, off); 3053 if (skb_gro_header_hard(skb, off + plen)) { 3054 ptr = skb_gro_header_slow(skb, off + plen, off); 3055 if (!ptr) 3056 return NULL; 3057 } 3058 3059 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3060 start, offset); 3061 3062 /* Adjust skb->csum since we changed the packet */ 3063 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3064 3065 grc->offset = off + hdrlen + offset; 3066 grc->delta = delta; 3067 3068 return ptr; 3069 } 3070 3071 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3072 struct gro_remcsum *grc) 3073 { 3074 void *ptr; 3075 size_t plen = grc->offset + sizeof(u16); 3076 3077 if (!grc->delta) 3078 return; 3079 3080 ptr = skb_gro_header_fast(skb, grc->offset); 3081 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3082 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3083 if (!ptr) 3084 return; 3085 } 3086 3087 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3088 } 3089 3090 #ifdef CONFIG_XFRM_OFFLOAD 3091 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3092 { 3093 if (PTR_ERR(pp) != -EINPROGRESS) 3094 NAPI_GRO_CB(skb)->flush |= flush; 3095 } 3096 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3097 struct sk_buff *pp, 3098 int flush, 3099 struct gro_remcsum *grc) 3100 { 3101 if (PTR_ERR(pp) != -EINPROGRESS) { 3102 NAPI_GRO_CB(skb)->flush |= flush; 3103 skb_gro_remcsum_cleanup(skb, grc); 3104 skb->remcsum_offload = 0; 3105 } 3106 } 3107 #else 3108 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3109 { 3110 NAPI_GRO_CB(skb)->flush |= flush; 3111 } 3112 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3113 struct sk_buff *pp, 3114 int flush, 3115 struct gro_remcsum *grc) 3116 { 3117 NAPI_GRO_CB(skb)->flush |= flush; 3118 skb_gro_remcsum_cleanup(skb, grc); 3119 skb->remcsum_offload = 0; 3120 } 3121 #endif 3122 3123 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3124 unsigned short type, 3125 const void *daddr, const void *saddr, 3126 unsigned int len) 3127 { 3128 if (!dev->header_ops || !dev->header_ops->create) 3129 return 0; 3130 3131 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3132 } 3133 3134 static inline int dev_parse_header(const struct sk_buff *skb, 3135 unsigned char *haddr) 3136 { 3137 const struct net_device *dev = skb->dev; 3138 3139 if (!dev->header_ops || !dev->header_ops->parse) 3140 return 0; 3141 return dev->header_ops->parse(skb, haddr); 3142 } 3143 3144 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3145 { 3146 const struct net_device *dev = skb->dev; 3147 3148 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3149 return 0; 3150 return dev->header_ops->parse_protocol(skb); 3151 } 3152 3153 /* ll_header must have at least hard_header_len allocated */ 3154 static inline bool dev_validate_header(const struct net_device *dev, 3155 char *ll_header, int len) 3156 { 3157 if (likely(len >= dev->hard_header_len)) 3158 return true; 3159 if (len < dev->min_header_len) 3160 return false; 3161 3162 if (capable(CAP_SYS_RAWIO)) { 3163 memset(ll_header + len, 0, dev->hard_header_len - len); 3164 return true; 3165 } 3166 3167 if (dev->header_ops && dev->header_ops->validate) 3168 return dev->header_ops->validate(ll_header, len); 3169 3170 return false; 3171 } 3172 3173 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3174 int len, int size); 3175 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3176 static inline int unregister_gifconf(unsigned int family) 3177 { 3178 return register_gifconf(family, NULL); 3179 } 3180 3181 #ifdef CONFIG_NET_FLOW_LIMIT 3182 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3183 struct sd_flow_limit { 3184 u64 count; 3185 unsigned int num_buckets; 3186 unsigned int history_head; 3187 u16 history[FLOW_LIMIT_HISTORY]; 3188 u8 buckets[]; 3189 }; 3190 3191 extern int netdev_flow_limit_table_len; 3192 #endif /* CONFIG_NET_FLOW_LIMIT */ 3193 3194 /* 3195 * Incoming packets are placed on per-CPU queues 3196 */ 3197 struct softnet_data { 3198 struct list_head poll_list; 3199 struct sk_buff_head process_queue; 3200 3201 /* stats */ 3202 unsigned int processed; 3203 unsigned int time_squeeze; 3204 unsigned int received_rps; 3205 #ifdef CONFIG_RPS 3206 struct softnet_data *rps_ipi_list; 3207 #endif 3208 #ifdef CONFIG_NET_FLOW_LIMIT 3209 struct sd_flow_limit __rcu *flow_limit; 3210 #endif 3211 struct Qdisc *output_queue; 3212 struct Qdisc **output_queue_tailp; 3213 struct sk_buff *completion_queue; 3214 #ifdef CONFIG_XFRM_OFFLOAD 3215 struct sk_buff_head xfrm_backlog; 3216 #endif 3217 /* written and read only by owning cpu: */ 3218 struct { 3219 u16 recursion; 3220 u8 more; 3221 } xmit; 3222 #ifdef CONFIG_RPS 3223 /* input_queue_head should be written by cpu owning this struct, 3224 * and only read by other cpus. Worth using a cache line. 3225 */ 3226 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3227 3228 /* Elements below can be accessed between CPUs for RPS/RFS */ 3229 call_single_data_t csd ____cacheline_aligned_in_smp; 3230 struct softnet_data *rps_ipi_next; 3231 unsigned int cpu; 3232 unsigned int input_queue_tail; 3233 #endif 3234 unsigned int dropped; 3235 struct sk_buff_head input_pkt_queue; 3236 struct napi_struct backlog; 3237 3238 }; 3239 3240 static inline void input_queue_head_incr(struct softnet_data *sd) 3241 { 3242 #ifdef CONFIG_RPS 3243 sd->input_queue_head++; 3244 #endif 3245 } 3246 3247 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3248 unsigned int *qtail) 3249 { 3250 #ifdef CONFIG_RPS 3251 *qtail = ++sd->input_queue_tail; 3252 #endif 3253 } 3254 3255 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3256 3257 static inline int dev_recursion_level(void) 3258 { 3259 return this_cpu_read(softnet_data.xmit.recursion); 3260 } 3261 3262 #define XMIT_RECURSION_LIMIT 8 3263 static inline bool dev_xmit_recursion(void) 3264 { 3265 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3266 XMIT_RECURSION_LIMIT); 3267 } 3268 3269 static inline void dev_xmit_recursion_inc(void) 3270 { 3271 __this_cpu_inc(softnet_data.xmit.recursion); 3272 } 3273 3274 static inline void dev_xmit_recursion_dec(void) 3275 { 3276 __this_cpu_dec(softnet_data.xmit.recursion); 3277 } 3278 3279 void __netif_schedule(struct Qdisc *q); 3280 void netif_schedule_queue(struct netdev_queue *txq); 3281 3282 static inline void netif_tx_schedule_all(struct net_device *dev) 3283 { 3284 unsigned int i; 3285 3286 for (i = 0; i < dev->num_tx_queues; i++) 3287 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3288 } 3289 3290 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3291 { 3292 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3293 } 3294 3295 /** 3296 * netif_start_queue - allow transmit 3297 * @dev: network device 3298 * 3299 * Allow upper layers to call the device hard_start_xmit routine. 3300 */ 3301 static inline void netif_start_queue(struct net_device *dev) 3302 { 3303 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3304 } 3305 3306 static inline void netif_tx_start_all_queues(struct net_device *dev) 3307 { 3308 unsigned int i; 3309 3310 for (i = 0; i < dev->num_tx_queues; i++) { 3311 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3312 netif_tx_start_queue(txq); 3313 } 3314 } 3315 3316 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3317 3318 /** 3319 * netif_wake_queue - restart transmit 3320 * @dev: network device 3321 * 3322 * Allow upper layers to call the device hard_start_xmit routine. 3323 * Used for flow control when transmit resources are available. 3324 */ 3325 static inline void netif_wake_queue(struct net_device *dev) 3326 { 3327 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3328 } 3329 3330 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3331 { 3332 unsigned int i; 3333 3334 for (i = 0; i < dev->num_tx_queues; i++) { 3335 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3336 netif_tx_wake_queue(txq); 3337 } 3338 } 3339 3340 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3341 { 3342 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3343 } 3344 3345 /** 3346 * netif_stop_queue - stop transmitted packets 3347 * @dev: network device 3348 * 3349 * Stop upper layers calling the device hard_start_xmit routine. 3350 * Used for flow control when transmit resources are unavailable. 3351 */ 3352 static inline void netif_stop_queue(struct net_device *dev) 3353 { 3354 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3355 } 3356 3357 void netif_tx_stop_all_queues(struct net_device *dev); 3358 3359 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3360 { 3361 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3362 } 3363 3364 /** 3365 * netif_queue_stopped - test if transmit queue is flowblocked 3366 * @dev: network device 3367 * 3368 * Test if transmit queue on device is currently unable to send. 3369 */ 3370 static inline bool netif_queue_stopped(const struct net_device *dev) 3371 { 3372 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3373 } 3374 3375 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3376 { 3377 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3378 } 3379 3380 static inline bool 3381 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3382 { 3383 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3384 } 3385 3386 static inline bool 3387 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3388 { 3389 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3390 } 3391 3392 /** 3393 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3394 * @dev_queue: pointer to transmit queue 3395 * 3396 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3397 * to give appropriate hint to the CPU. 3398 */ 3399 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3400 { 3401 #ifdef CONFIG_BQL 3402 prefetchw(&dev_queue->dql.num_queued); 3403 #endif 3404 } 3405 3406 /** 3407 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3408 * @dev_queue: pointer to transmit queue 3409 * 3410 * BQL enabled drivers might use this helper in their TX completion path, 3411 * to give appropriate hint to the CPU. 3412 */ 3413 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3414 { 3415 #ifdef CONFIG_BQL 3416 prefetchw(&dev_queue->dql.limit); 3417 #endif 3418 } 3419 3420 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3421 unsigned int bytes) 3422 { 3423 #ifdef CONFIG_BQL 3424 dql_queued(&dev_queue->dql, bytes); 3425 3426 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3427 return; 3428 3429 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3430 3431 /* 3432 * The XOFF flag must be set before checking the dql_avail below, 3433 * because in netdev_tx_completed_queue we update the dql_completed 3434 * before checking the XOFF flag. 3435 */ 3436 smp_mb(); 3437 3438 /* check again in case another CPU has just made room avail */ 3439 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3440 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3441 #endif 3442 } 3443 3444 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3445 * that they should not test BQL status themselves. 3446 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3447 * skb of a batch. 3448 * Returns true if the doorbell must be used to kick the NIC. 3449 */ 3450 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3451 unsigned int bytes, 3452 bool xmit_more) 3453 { 3454 if (xmit_more) { 3455 #ifdef CONFIG_BQL 3456 dql_queued(&dev_queue->dql, bytes); 3457 #endif 3458 return netif_tx_queue_stopped(dev_queue); 3459 } 3460 netdev_tx_sent_queue(dev_queue, bytes); 3461 return true; 3462 } 3463 3464 /** 3465 * netdev_sent_queue - report the number of bytes queued to hardware 3466 * @dev: network device 3467 * @bytes: number of bytes queued to the hardware device queue 3468 * 3469 * Report the number of bytes queued for sending/completion to the network 3470 * device hardware queue. @bytes should be a good approximation and should 3471 * exactly match netdev_completed_queue() @bytes 3472 */ 3473 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3474 { 3475 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3476 } 3477 3478 static inline bool __netdev_sent_queue(struct net_device *dev, 3479 unsigned int bytes, 3480 bool xmit_more) 3481 { 3482 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3483 xmit_more); 3484 } 3485 3486 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3487 unsigned int pkts, unsigned int bytes) 3488 { 3489 #ifdef CONFIG_BQL 3490 if (unlikely(!bytes)) 3491 return; 3492 3493 dql_completed(&dev_queue->dql, bytes); 3494 3495 /* 3496 * Without the memory barrier there is a small possiblity that 3497 * netdev_tx_sent_queue will miss the update and cause the queue to 3498 * be stopped forever 3499 */ 3500 smp_mb(); 3501 3502 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3503 return; 3504 3505 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3506 netif_schedule_queue(dev_queue); 3507 #endif 3508 } 3509 3510 /** 3511 * netdev_completed_queue - report bytes and packets completed by device 3512 * @dev: network device 3513 * @pkts: actual number of packets sent over the medium 3514 * @bytes: actual number of bytes sent over the medium 3515 * 3516 * Report the number of bytes and packets transmitted by the network device 3517 * hardware queue over the physical medium, @bytes must exactly match the 3518 * @bytes amount passed to netdev_sent_queue() 3519 */ 3520 static inline void netdev_completed_queue(struct net_device *dev, 3521 unsigned int pkts, unsigned int bytes) 3522 { 3523 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3524 } 3525 3526 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3527 { 3528 #ifdef CONFIG_BQL 3529 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3530 dql_reset(&q->dql); 3531 #endif 3532 } 3533 3534 /** 3535 * netdev_reset_queue - reset the packets and bytes count of a network device 3536 * @dev_queue: network device 3537 * 3538 * Reset the bytes and packet count of a network device and clear the 3539 * software flow control OFF bit for this network device 3540 */ 3541 static inline void netdev_reset_queue(struct net_device *dev_queue) 3542 { 3543 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3544 } 3545 3546 /** 3547 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3548 * @dev: network device 3549 * @queue_index: given tx queue index 3550 * 3551 * Returns 0 if given tx queue index >= number of device tx queues, 3552 * otherwise returns the originally passed tx queue index. 3553 */ 3554 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3555 { 3556 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3557 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3558 dev->name, queue_index, 3559 dev->real_num_tx_queues); 3560 return 0; 3561 } 3562 3563 return queue_index; 3564 } 3565 3566 /** 3567 * netif_running - test if up 3568 * @dev: network device 3569 * 3570 * Test if the device has been brought up. 3571 */ 3572 static inline bool netif_running(const struct net_device *dev) 3573 { 3574 return test_bit(__LINK_STATE_START, &dev->state); 3575 } 3576 3577 /* 3578 * Routines to manage the subqueues on a device. We only need start, 3579 * stop, and a check if it's stopped. All other device management is 3580 * done at the overall netdevice level. 3581 * Also test the device if we're multiqueue. 3582 */ 3583 3584 /** 3585 * netif_start_subqueue - allow sending packets on subqueue 3586 * @dev: network device 3587 * @queue_index: sub queue index 3588 * 3589 * Start individual transmit queue of a device with multiple transmit queues. 3590 */ 3591 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3592 { 3593 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3594 3595 netif_tx_start_queue(txq); 3596 } 3597 3598 /** 3599 * netif_stop_subqueue - stop sending packets on subqueue 3600 * @dev: network device 3601 * @queue_index: sub queue index 3602 * 3603 * Stop individual transmit queue of a device with multiple transmit queues. 3604 */ 3605 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3606 { 3607 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3608 netif_tx_stop_queue(txq); 3609 } 3610 3611 /** 3612 * netif_subqueue_stopped - test status of subqueue 3613 * @dev: network device 3614 * @queue_index: sub queue index 3615 * 3616 * Check individual transmit queue of a device with multiple transmit queues. 3617 */ 3618 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3619 u16 queue_index) 3620 { 3621 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3622 3623 return netif_tx_queue_stopped(txq); 3624 } 3625 3626 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3627 struct sk_buff *skb) 3628 { 3629 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3630 } 3631 3632 /** 3633 * netif_wake_subqueue - allow sending packets on subqueue 3634 * @dev: network device 3635 * @queue_index: sub queue index 3636 * 3637 * Resume individual transmit queue of a device with multiple transmit queues. 3638 */ 3639 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3640 { 3641 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3642 3643 netif_tx_wake_queue(txq); 3644 } 3645 3646 #ifdef CONFIG_XPS 3647 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3648 u16 index); 3649 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3650 u16 index, bool is_rxqs_map); 3651 3652 /** 3653 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3654 * @j: CPU/Rx queue index 3655 * @mask: bitmask of all cpus/rx queues 3656 * @nr_bits: number of bits in the bitmask 3657 * 3658 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3659 */ 3660 static inline bool netif_attr_test_mask(unsigned long j, 3661 const unsigned long *mask, 3662 unsigned int nr_bits) 3663 { 3664 cpu_max_bits_warn(j, nr_bits); 3665 return test_bit(j, mask); 3666 } 3667 3668 /** 3669 * netif_attr_test_online - Test for online CPU/Rx queue 3670 * @j: CPU/Rx queue index 3671 * @online_mask: bitmask for CPUs/Rx queues that are online 3672 * @nr_bits: number of bits in the bitmask 3673 * 3674 * Returns true if a CPU/Rx queue is online. 3675 */ 3676 static inline bool netif_attr_test_online(unsigned long j, 3677 const unsigned long *online_mask, 3678 unsigned int nr_bits) 3679 { 3680 cpu_max_bits_warn(j, nr_bits); 3681 3682 if (online_mask) 3683 return test_bit(j, online_mask); 3684 3685 return (j < nr_bits); 3686 } 3687 3688 /** 3689 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3690 * @n: CPU/Rx queue index 3691 * @srcp: the cpumask/Rx queue mask pointer 3692 * @nr_bits: number of bits in the bitmask 3693 * 3694 * Returns >= nr_bits if no further CPUs/Rx queues set. 3695 */ 3696 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3697 unsigned int nr_bits) 3698 { 3699 /* -1 is a legal arg here. */ 3700 if (n != -1) 3701 cpu_max_bits_warn(n, nr_bits); 3702 3703 if (srcp) 3704 return find_next_bit(srcp, nr_bits, n + 1); 3705 3706 return n + 1; 3707 } 3708 3709 /** 3710 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3711 * @n: CPU/Rx queue index 3712 * @src1p: the first CPUs/Rx queues mask pointer 3713 * @src2p: the second CPUs/Rx queues mask pointer 3714 * @nr_bits: number of bits in the bitmask 3715 * 3716 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3717 */ 3718 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3719 const unsigned long *src2p, 3720 unsigned int nr_bits) 3721 { 3722 /* -1 is a legal arg here. */ 3723 if (n != -1) 3724 cpu_max_bits_warn(n, nr_bits); 3725 3726 if (src1p && src2p) 3727 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3728 else if (src1p) 3729 return find_next_bit(src1p, nr_bits, n + 1); 3730 else if (src2p) 3731 return find_next_bit(src2p, nr_bits, n + 1); 3732 3733 return n + 1; 3734 } 3735 #else 3736 static inline int netif_set_xps_queue(struct net_device *dev, 3737 const struct cpumask *mask, 3738 u16 index) 3739 { 3740 return 0; 3741 } 3742 3743 static inline int __netif_set_xps_queue(struct net_device *dev, 3744 const unsigned long *mask, 3745 u16 index, bool is_rxqs_map) 3746 { 3747 return 0; 3748 } 3749 #endif 3750 3751 /** 3752 * netif_is_multiqueue - test if device has multiple transmit queues 3753 * @dev: network device 3754 * 3755 * Check if device has multiple transmit queues 3756 */ 3757 static inline bool netif_is_multiqueue(const struct net_device *dev) 3758 { 3759 return dev->num_tx_queues > 1; 3760 } 3761 3762 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3763 3764 #ifdef CONFIG_SYSFS 3765 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3766 #else 3767 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3768 unsigned int rxqs) 3769 { 3770 dev->real_num_rx_queues = rxqs; 3771 return 0; 3772 } 3773 #endif 3774 3775 static inline struct netdev_rx_queue * 3776 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3777 { 3778 return dev->_rx + rxq; 3779 } 3780 3781 #ifdef CONFIG_SYSFS 3782 static inline unsigned int get_netdev_rx_queue_index( 3783 struct netdev_rx_queue *queue) 3784 { 3785 struct net_device *dev = queue->dev; 3786 int index = queue - dev->_rx; 3787 3788 BUG_ON(index >= dev->num_rx_queues); 3789 return index; 3790 } 3791 #endif 3792 3793 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3794 int netif_get_num_default_rss_queues(void); 3795 3796 enum skb_free_reason { 3797 SKB_REASON_CONSUMED, 3798 SKB_REASON_DROPPED, 3799 }; 3800 3801 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3802 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3803 3804 /* 3805 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3806 * interrupt context or with hardware interrupts being disabled. 3807 * (in_irq() || irqs_disabled()) 3808 * 3809 * We provide four helpers that can be used in following contexts : 3810 * 3811 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3812 * replacing kfree_skb(skb) 3813 * 3814 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3815 * Typically used in place of consume_skb(skb) in TX completion path 3816 * 3817 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3818 * replacing kfree_skb(skb) 3819 * 3820 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3821 * and consumed a packet. Used in place of consume_skb(skb) 3822 */ 3823 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3824 { 3825 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3826 } 3827 3828 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3829 { 3830 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3831 } 3832 3833 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3834 { 3835 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3836 } 3837 3838 static inline void dev_consume_skb_any(struct sk_buff *skb) 3839 { 3840 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3841 } 3842 3843 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3844 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3845 int netif_rx(struct sk_buff *skb); 3846 int netif_rx_ni(struct sk_buff *skb); 3847 int netif_rx_any_context(struct sk_buff *skb); 3848 int netif_receive_skb(struct sk_buff *skb); 3849 int netif_receive_skb_core(struct sk_buff *skb); 3850 void netif_receive_skb_list(struct list_head *head); 3851 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3852 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3853 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3854 gro_result_t napi_gro_frags(struct napi_struct *napi); 3855 struct packet_offload *gro_find_receive_by_type(__be16 type); 3856 struct packet_offload *gro_find_complete_by_type(__be16 type); 3857 3858 static inline void napi_free_frags(struct napi_struct *napi) 3859 { 3860 kfree_skb(napi->skb); 3861 napi->skb = NULL; 3862 } 3863 3864 bool netdev_is_rx_handler_busy(struct net_device *dev); 3865 int netdev_rx_handler_register(struct net_device *dev, 3866 rx_handler_func_t *rx_handler, 3867 void *rx_handler_data); 3868 void netdev_rx_handler_unregister(struct net_device *dev); 3869 3870 bool dev_valid_name(const char *name); 3871 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3872 bool *need_copyout); 3873 int dev_ifconf(struct net *net, struct ifconf *, int); 3874 int dev_ethtool(struct net *net, struct ifreq *); 3875 unsigned int dev_get_flags(const struct net_device *); 3876 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3877 struct netlink_ext_ack *extack); 3878 int dev_change_flags(struct net_device *dev, unsigned int flags, 3879 struct netlink_ext_ack *extack); 3880 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3881 unsigned int gchanges); 3882 int dev_change_name(struct net_device *, const char *); 3883 int dev_set_alias(struct net_device *, const char *, size_t); 3884 int dev_get_alias(const struct net_device *, char *, size_t); 3885 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3886 int __dev_set_mtu(struct net_device *, int); 3887 int dev_validate_mtu(struct net_device *dev, int mtu, 3888 struct netlink_ext_ack *extack); 3889 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3890 struct netlink_ext_ack *extack); 3891 int dev_set_mtu(struct net_device *, int); 3892 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3893 void dev_set_group(struct net_device *, int); 3894 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3895 struct netlink_ext_ack *extack); 3896 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3897 struct netlink_ext_ack *extack); 3898 int dev_change_carrier(struct net_device *, bool new_carrier); 3899 int dev_get_phys_port_id(struct net_device *dev, 3900 struct netdev_phys_item_id *ppid); 3901 int dev_get_phys_port_name(struct net_device *dev, 3902 char *name, size_t len); 3903 int dev_get_port_parent_id(struct net_device *dev, 3904 struct netdev_phys_item_id *ppid, bool recurse); 3905 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3906 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3907 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3908 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3909 u32 value); 3910 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3911 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3912 struct netdev_queue *txq, int *ret); 3913 3914 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3915 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3916 int fd, int expected_fd, u32 flags); 3917 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3918 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3919 3920 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3921 3922 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3923 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3924 bool is_skb_forwardable(const struct net_device *dev, 3925 const struct sk_buff *skb); 3926 3927 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3928 struct sk_buff *skb) 3929 { 3930 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3931 unlikely(!is_skb_forwardable(dev, skb))) { 3932 atomic_long_inc(&dev->rx_dropped); 3933 kfree_skb(skb); 3934 return NET_RX_DROP; 3935 } 3936 3937 skb_scrub_packet(skb, true); 3938 skb->priority = 0; 3939 return 0; 3940 } 3941 3942 bool dev_nit_active(struct net_device *dev); 3943 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3944 3945 extern int netdev_budget; 3946 extern unsigned int netdev_budget_usecs; 3947 3948 /* Called by rtnetlink.c:rtnl_unlock() */ 3949 void netdev_run_todo(void); 3950 3951 /** 3952 * dev_put - release reference to device 3953 * @dev: network device 3954 * 3955 * Release reference to device to allow it to be freed. 3956 */ 3957 static inline void dev_put(struct net_device *dev) 3958 { 3959 this_cpu_dec(*dev->pcpu_refcnt); 3960 } 3961 3962 /** 3963 * dev_hold - get reference to device 3964 * @dev: network device 3965 * 3966 * Hold reference to device to keep it from being freed. 3967 */ 3968 static inline void dev_hold(struct net_device *dev) 3969 { 3970 this_cpu_inc(*dev->pcpu_refcnt); 3971 } 3972 3973 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3974 * and _off may be called from IRQ context, but it is caller 3975 * who is responsible for serialization of these calls. 3976 * 3977 * The name carrier is inappropriate, these functions should really be 3978 * called netif_lowerlayer_*() because they represent the state of any 3979 * kind of lower layer not just hardware media. 3980 */ 3981 3982 void linkwatch_init_dev(struct net_device *dev); 3983 void linkwatch_fire_event(struct net_device *dev); 3984 void linkwatch_forget_dev(struct net_device *dev); 3985 3986 /** 3987 * netif_carrier_ok - test if carrier present 3988 * @dev: network device 3989 * 3990 * Check if carrier is present on device 3991 */ 3992 static inline bool netif_carrier_ok(const struct net_device *dev) 3993 { 3994 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3995 } 3996 3997 unsigned long dev_trans_start(struct net_device *dev); 3998 3999 void __netdev_watchdog_up(struct net_device *dev); 4000 4001 void netif_carrier_on(struct net_device *dev); 4002 4003 void netif_carrier_off(struct net_device *dev); 4004 4005 /** 4006 * netif_dormant_on - mark device as dormant. 4007 * @dev: network device 4008 * 4009 * Mark device as dormant (as per RFC2863). 4010 * 4011 * The dormant state indicates that the relevant interface is not 4012 * actually in a condition to pass packets (i.e., it is not 'up') but is 4013 * in a "pending" state, waiting for some external event. For "on- 4014 * demand" interfaces, this new state identifies the situation where the 4015 * interface is waiting for events to place it in the up state. 4016 */ 4017 static inline void netif_dormant_on(struct net_device *dev) 4018 { 4019 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4020 linkwatch_fire_event(dev); 4021 } 4022 4023 /** 4024 * netif_dormant_off - set device as not dormant. 4025 * @dev: network device 4026 * 4027 * Device is not in dormant state. 4028 */ 4029 static inline void netif_dormant_off(struct net_device *dev) 4030 { 4031 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4032 linkwatch_fire_event(dev); 4033 } 4034 4035 /** 4036 * netif_dormant - test if device is dormant 4037 * @dev: network device 4038 * 4039 * Check if device is dormant. 4040 */ 4041 static inline bool netif_dormant(const struct net_device *dev) 4042 { 4043 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4044 } 4045 4046 4047 /** 4048 * netif_testing_on - mark device as under test. 4049 * @dev: network device 4050 * 4051 * Mark device as under test (as per RFC2863). 4052 * 4053 * The testing state indicates that some test(s) must be performed on 4054 * the interface. After completion, of the test, the interface state 4055 * will change to up, dormant, or down, as appropriate. 4056 */ 4057 static inline void netif_testing_on(struct net_device *dev) 4058 { 4059 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4060 linkwatch_fire_event(dev); 4061 } 4062 4063 /** 4064 * netif_testing_off - set device as not under test. 4065 * @dev: network device 4066 * 4067 * Device is not in testing state. 4068 */ 4069 static inline void netif_testing_off(struct net_device *dev) 4070 { 4071 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4072 linkwatch_fire_event(dev); 4073 } 4074 4075 /** 4076 * netif_testing - test if device is under test 4077 * @dev: network device 4078 * 4079 * Check if device is under test 4080 */ 4081 static inline bool netif_testing(const struct net_device *dev) 4082 { 4083 return test_bit(__LINK_STATE_TESTING, &dev->state); 4084 } 4085 4086 4087 /** 4088 * netif_oper_up - test if device is operational 4089 * @dev: network device 4090 * 4091 * Check if carrier is operational 4092 */ 4093 static inline bool netif_oper_up(const struct net_device *dev) 4094 { 4095 return (dev->operstate == IF_OPER_UP || 4096 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4097 } 4098 4099 /** 4100 * netif_device_present - is device available or removed 4101 * @dev: network device 4102 * 4103 * Check if device has not been removed from system. 4104 */ 4105 static inline bool netif_device_present(struct net_device *dev) 4106 { 4107 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4108 } 4109 4110 void netif_device_detach(struct net_device *dev); 4111 4112 void netif_device_attach(struct net_device *dev); 4113 4114 /* 4115 * Network interface message level settings 4116 */ 4117 4118 enum { 4119 NETIF_MSG_DRV_BIT, 4120 NETIF_MSG_PROBE_BIT, 4121 NETIF_MSG_LINK_BIT, 4122 NETIF_MSG_TIMER_BIT, 4123 NETIF_MSG_IFDOWN_BIT, 4124 NETIF_MSG_IFUP_BIT, 4125 NETIF_MSG_RX_ERR_BIT, 4126 NETIF_MSG_TX_ERR_BIT, 4127 NETIF_MSG_TX_QUEUED_BIT, 4128 NETIF_MSG_INTR_BIT, 4129 NETIF_MSG_TX_DONE_BIT, 4130 NETIF_MSG_RX_STATUS_BIT, 4131 NETIF_MSG_PKTDATA_BIT, 4132 NETIF_MSG_HW_BIT, 4133 NETIF_MSG_WOL_BIT, 4134 4135 /* When you add a new bit above, update netif_msg_class_names array 4136 * in net/ethtool/common.c 4137 */ 4138 NETIF_MSG_CLASS_COUNT, 4139 }; 4140 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4141 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4142 4143 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4144 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4145 4146 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4147 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4148 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4149 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4150 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4151 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4152 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4153 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4154 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4155 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4156 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4157 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4158 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4159 #define NETIF_MSG_HW __NETIF_MSG(HW) 4160 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4161 4162 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4163 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4164 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4165 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4166 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4167 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4168 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4169 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4170 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4171 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4172 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4173 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4174 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4175 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4176 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4177 4178 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4179 { 4180 /* use default */ 4181 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4182 return default_msg_enable_bits; 4183 if (debug_value == 0) /* no output */ 4184 return 0; 4185 /* set low N bits */ 4186 return (1U << debug_value) - 1; 4187 } 4188 4189 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4190 { 4191 spin_lock(&txq->_xmit_lock); 4192 txq->xmit_lock_owner = cpu; 4193 } 4194 4195 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4196 { 4197 __acquire(&txq->_xmit_lock); 4198 return true; 4199 } 4200 4201 static inline void __netif_tx_release(struct netdev_queue *txq) 4202 { 4203 __release(&txq->_xmit_lock); 4204 } 4205 4206 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4207 { 4208 spin_lock_bh(&txq->_xmit_lock); 4209 txq->xmit_lock_owner = smp_processor_id(); 4210 } 4211 4212 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4213 { 4214 bool ok = spin_trylock(&txq->_xmit_lock); 4215 if (likely(ok)) 4216 txq->xmit_lock_owner = smp_processor_id(); 4217 return ok; 4218 } 4219 4220 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4221 { 4222 txq->xmit_lock_owner = -1; 4223 spin_unlock(&txq->_xmit_lock); 4224 } 4225 4226 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4227 { 4228 txq->xmit_lock_owner = -1; 4229 spin_unlock_bh(&txq->_xmit_lock); 4230 } 4231 4232 static inline void txq_trans_update(struct netdev_queue *txq) 4233 { 4234 if (txq->xmit_lock_owner != -1) 4235 txq->trans_start = jiffies; 4236 } 4237 4238 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4239 static inline void netif_trans_update(struct net_device *dev) 4240 { 4241 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4242 4243 if (txq->trans_start != jiffies) 4244 txq->trans_start = jiffies; 4245 } 4246 4247 /** 4248 * netif_tx_lock - grab network device transmit lock 4249 * @dev: network device 4250 * 4251 * Get network device transmit lock 4252 */ 4253 static inline void netif_tx_lock(struct net_device *dev) 4254 { 4255 unsigned int i; 4256 int cpu; 4257 4258 spin_lock(&dev->tx_global_lock); 4259 cpu = smp_processor_id(); 4260 for (i = 0; i < dev->num_tx_queues; i++) { 4261 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4262 4263 /* We are the only thread of execution doing a 4264 * freeze, but we have to grab the _xmit_lock in 4265 * order to synchronize with threads which are in 4266 * the ->hard_start_xmit() handler and already 4267 * checked the frozen bit. 4268 */ 4269 __netif_tx_lock(txq, cpu); 4270 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4271 __netif_tx_unlock(txq); 4272 } 4273 } 4274 4275 static inline void netif_tx_lock_bh(struct net_device *dev) 4276 { 4277 local_bh_disable(); 4278 netif_tx_lock(dev); 4279 } 4280 4281 static inline void netif_tx_unlock(struct net_device *dev) 4282 { 4283 unsigned int i; 4284 4285 for (i = 0; i < dev->num_tx_queues; i++) { 4286 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4287 4288 /* No need to grab the _xmit_lock here. If the 4289 * queue is not stopped for another reason, we 4290 * force a schedule. 4291 */ 4292 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4293 netif_schedule_queue(txq); 4294 } 4295 spin_unlock(&dev->tx_global_lock); 4296 } 4297 4298 static inline void netif_tx_unlock_bh(struct net_device *dev) 4299 { 4300 netif_tx_unlock(dev); 4301 local_bh_enable(); 4302 } 4303 4304 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4305 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4306 __netif_tx_lock(txq, cpu); \ 4307 } else { \ 4308 __netif_tx_acquire(txq); \ 4309 } \ 4310 } 4311 4312 #define HARD_TX_TRYLOCK(dev, txq) \ 4313 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4314 __netif_tx_trylock(txq) : \ 4315 __netif_tx_acquire(txq)) 4316 4317 #define HARD_TX_UNLOCK(dev, txq) { \ 4318 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4319 __netif_tx_unlock(txq); \ 4320 } else { \ 4321 __netif_tx_release(txq); \ 4322 } \ 4323 } 4324 4325 static inline void netif_tx_disable(struct net_device *dev) 4326 { 4327 unsigned int i; 4328 int cpu; 4329 4330 local_bh_disable(); 4331 cpu = smp_processor_id(); 4332 for (i = 0; i < dev->num_tx_queues; i++) { 4333 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4334 4335 __netif_tx_lock(txq, cpu); 4336 netif_tx_stop_queue(txq); 4337 __netif_tx_unlock(txq); 4338 } 4339 local_bh_enable(); 4340 } 4341 4342 static inline void netif_addr_lock(struct net_device *dev) 4343 { 4344 unsigned char nest_level = 0; 4345 4346 #ifdef CONFIG_LOCKDEP 4347 nest_level = dev->nested_level; 4348 #endif 4349 spin_lock_nested(&dev->addr_list_lock, nest_level); 4350 } 4351 4352 static inline void netif_addr_lock_bh(struct net_device *dev) 4353 { 4354 unsigned char nest_level = 0; 4355 4356 #ifdef CONFIG_LOCKDEP 4357 nest_level = dev->nested_level; 4358 #endif 4359 local_bh_disable(); 4360 spin_lock_nested(&dev->addr_list_lock, nest_level); 4361 } 4362 4363 static inline void netif_addr_unlock(struct net_device *dev) 4364 { 4365 spin_unlock(&dev->addr_list_lock); 4366 } 4367 4368 static inline void netif_addr_unlock_bh(struct net_device *dev) 4369 { 4370 spin_unlock_bh(&dev->addr_list_lock); 4371 } 4372 4373 /* 4374 * dev_addrs walker. Should be used only for read access. Call with 4375 * rcu_read_lock held. 4376 */ 4377 #define for_each_dev_addr(dev, ha) \ 4378 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4379 4380 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4381 4382 void ether_setup(struct net_device *dev); 4383 4384 /* Support for loadable net-drivers */ 4385 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4386 unsigned char name_assign_type, 4387 void (*setup)(struct net_device *), 4388 unsigned int txqs, unsigned int rxqs); 4389 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4390 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4391 4392 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4393 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4394 count) 4395 4396 int register_netdev(struct net_device *dev); 4397 void unregister_netdev(struct net_device *dev); 4398 4399 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4400 4401 /* General hardware address lists handling functions */ 4402 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4403 struct netdev_hw_addr_list *from_list, int addr_len); 4404 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4405 struct netdev_hw_addr_list *from_list, int addr_len); 4406 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4407 struct net_device *dev, 4408 int (*sync)(struct net_device *, const unsigned char *), 4409 int (*unsync)(struct net_device *, 4410 const unsigned char *)); 4411 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4412 struct net_device *dev, 4413 int (*sync)(struct net_device *, 4414 const unsigned char *, int), 4415 int (*unsync)(struct net_device *, 4416 const unsigned char *, int)); 4417 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4418 struct net_device *dev, 4419 int (*unsync)(struct net_device *, 4420 const unsigned char *, int)); 4421 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4422 struct net_device *dev, 4423 int (*unsync)(struct net_device *, 4424 const unsigned char *)); 4425 void __hw_addr_init(struct netdev_hw_addr_list *list); 4426 4427 /* Functions used for device addresses handling */ 4428 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4429 unsigned char addr_type); 4430 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4431 unsigned char addr_type); 4432 void dev_addr_flush(struct net_device *dev); 4433 int dev_addr_init(struct net_device *dev); 4434 4435 /* Functions used for unicast addresses handling */ 4436 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4437 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4438 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4439 int dev_uc_sync(struct net_device *to, struct net_device *from); 4440 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4441 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4442 void dev_uc_flush(struct net_device *dev); 4443 void dev_uc_init(struct net_device *dev); 4444 4445 /** 4446 * __dev_uc_sync - Synchonize device's unicast list 4447 * @dev: device to sync 4448 * @sync: function to call if address should be added 4449 * @unsync: function to call if address should be removed 4450 * 4451 * Add newly added addresses to the interface, and release 4452 * addresses that have been deleted. 4453 */ 4454 static inline int __dev_uc_sync(struct net_device *dev, 4455 int (*sync)(struct net_device *, 4456 const unsigned char *), 4457 int (*unsync)(struct net_device *, 4458 const unsigned char *)) 4459 { 4460 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4461 } 4462 4463 /** 4464 * __dev_uc_unsync - Remove synchronized addresses from device 4465 * @dev: device to sync 4466 * @unsync: function to call if address should be removed 4467 * 4468 * Remove all addresses that were added to the device by dev_uc_sync(). 4469 */ 4470 static inline void __dev_uc_unsync(struct net_device *dev, 4471 int (*unsync)(struct net_device *, 4472 const unsigned char *)) 4473 { 4474 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4475 } 4476 4477 /* Functions used for multicast addresses handling */ 4478 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4479 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4480 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4481 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4482 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4483 int dev_mc_sync(struct net_device *to, struct net_device *from); 4484 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4485 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4486 void dev_mc_flush(struct net_device *dev); 4487 void dev_mc_init(struct net_device *dev); 4488 4489 /** 4490 * __dev_mc_sync - Synchonize device's multicast list 4491 * @dev: device to sync 4492 * @sync: function to call if address should be added 4493 * @unsync: function to call if address should be removed 4494 * 4495 * Add newly added addresses to the interface, and release 4496 * addresses that have been deleted. 4497 */ 4498 static inline int __dev_mc_sync(struct net_device *dev, 4499 int (*sync)(struct net_device *, 4500 const unsigned char *), 4501 int (*unsync)(struct net_device *, 4502 const unsigned char *)) 4503 { 4504 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4505 } 4506 4507 /** 4508 * __dev_mc_unsync - Remove synchronized addresses from device 4509 * @dev: device to sync 4510 * @unsync: function to call if address should be removed 4511 * 4512 * Remove all addresses that were added to the device by dev_mc_sync(). 4513 */ 4514 static inline void __dev_mc_unsync(struct net_device *dev, 4515 int (*unsync)(struct net_device *, 4516 const unsigned char *)) 4517 { 4518 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4519 } 4520 4521 /* Functions used for secondary unicast and multicast support */ 4522 void dev_set_rx_mode(struct net_device *dev); 4523 void __dev_set_rx_mode(struct net_device *dev); 4524 int dev_set_promiscuity(struct net_device *dev, int inc); 4525 int dev_set_allmulti(struct net_device *dev, int inc); 4526 void netdev_state_change(struct net_device *dev); 4527 void netdev_notify_peers(struct net_device *dev); 4528 void netdev_features_change(struct net_device *dev); 4529 /* Load a device via the kmod */ 4530 void dev_load(struct net *net, const char *name); 4531 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4532 struct rtnl_link_stats64 *storage); 4533 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4534 const struct net_device_stats *netdev_stats); 4535 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4536 const struct pcpu_sw_netstats __percpu *netstats); 4537 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4538 4539 extern int netdev_max_backlog; 4540 extern int netdev_tstamp_prequeue; 4541 extern int weight_p; 4542 extern int dev_weight_rx_bias; 4543 extern int dev_weight_tx_bias; 4544 extern int dev_rx_weight; 4545 extern int dev_tx_weight; 4546 extern int gro_normal_batch; 4547 4548 enum { 4549 NESTED_SYNC_IMM_BIT, 4550 NESTED_SYNC_TODO_BIT, 4551 }; 4552 4553 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4554 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4555 4556 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4557 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4558 4559 struct netdev_nested_priv { 4560 unsigned char flags; 4561 void *data; 4562 }; 4563 4564 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4565 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4566 struct list_head **iter); 4567 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4568 struct list_head **iter); 4569 4570 #ifdef CONFIG_LOCKDEP 4571 static LIST_HEAD(net_unlink_list); 4572 4573 static inline void net_unlink_todo(struct net_device *dev) 4574 { 4575 if (list_empty(&dev->unlink_list)) 4576 list_add_tail(&dev->unlink_list, &net_unlink_list); 4577 } 4578 #endif 4579 4580 /* iterate through upper list, must be called under RCU read lock */ 4581 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4582 for (iter = &(dev)->adj_list.upper, \ 4583 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4584 updev; \ 4585 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4586 4587 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4588 int (*fn)(struct net_device *upper_dev, 4589 struct netdev_nested_priv *priv), 4590 struct netdev_nested_priv *priv); 4591 4592 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4593 struct net_device *upper_dev); 4594 4595 bool netdev_has_any_upper_dev(struct net_device *dev); 4596 4597 void *netdev_lower_get_next_private(struct net_device *dev, 4598 struct list_head **iter); 4599 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4600 struct list_head **iter); 4601 4602 #define netdev_for_each_lower_private(dev, priv, iter) \ 4603 for (iter = (dev)->adj_list.lower.next, \ 4604 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4605 priv; \ 4606 priv = netdev_lower_get_next_private(dev, &(iter))) 4607 4608 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4609 for (iter = &(dev)->adj_list.lower, \ 4610 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4611 priv; \ 4612 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4613 4614 void *netdev_lower_get_next(struct net_device *dev, 4615 struct list_head **iter); 4616 4617 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4618 for (iter = (dev)->adj_list.lower.next, \ 4619 ldev = netdev_lower_get_next(dev, &(iter)); \ 4620 ldev; \ 4621 ldev = netdev_lower_get_next(dev, &(iter))) 4622 4623 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4624 struct list_head **iter); 4625 int netdev_walk_all_lower_dev(struct net_device *dev, 4626 int (*fn)(struct net_device *lower_dev, 4627 struct netdev_nested_priv *priv), 4628 struct netdev_nested_priv *priv); 4629 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4630 int (*fn)(struct net_device *lower_dev, 4631 struct netdev_nested_priv *priv), 4632 struct netdev_nested_priv *priv); 4633 4634 void *netdev_adjacent_get_private(struct list_head *adj_list); 4635 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4636 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4637 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4638 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4639 struct netlink_ext_ack *extack); 4640 int netdev_master_upper_dev_link(struct net_device *dev, 4641 struct net_device *upper_dev, 4642 void *upper_priv, void *upper_info, 4643 struct netlink_ext_ack *extack); 4644 void netdev_upper_dev_unlink(struct net_device *dev, 4645 struct net_device *upper_dev); 4646 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4647 struct net_device *new_dev, 4648 struct net_device *dev, 4649 struct netlink_ext_ack *extack); 4650 void netdev_adjacent_change_commit(struct net_device *old_dev, 4651 struct net_device *new_dev, 4652 struct net_device *dev); 4653 void netdev_adjacent_change_abort(struct net_device *old_dev, 4654 struct net_device *new_dev, 4655 struct net_device *dev); 4656 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4657 void *netdev_lower_dev_get_private(struct net_device *dev, 4658 struct net_device *lower_dev); 4659 void netdev_lower_state_changed(struct net_device *lower_dev, 4660 void *lower_state_info); 4661 4662 /* RSS keys are 40 or 52 bytes long */ 4663 #define NETDEV_RSS_KEY_LEN 52 4664 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4665 void netdev_rss_key_fill(void *buffer, size_t len); 4666 4667 int skb_checksum_help(struct sk_buff *skb); 4668 int skb_crc32c_csum_help(struct sk_buff *skb); 4669 int skb_csum_hwoffload_help(struct sk_buff *skb, 4670 const netdev_features_t features); 4671 4672 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4673 netdev_features_t features, bool tx_path); 4674 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4675 netdev_features_t features); 4676 4677 struct netdev_bonding_info { 4678 ifslave slave; 4679 ifbond master; 4680 }; 4681 4682 struct netdev_notifier_bonding_info { 4683 struct netdev_notifier_info info; /* must be first */ 4684 struct netdev_bonding_info bonding_info; 4685 }; 4686 4687 void netdev_bonding_info_change(struct net_device *dev, 4688 struct netdev_bonding_info *bonding_info); 4689 4690 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4691 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4692 #else 4693 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4694 const void *data) 4695 { 4696 } 4697 #endif 4698 4699 static inline 4700 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4701 { 4702 return __skb_gso_segment(skb, features, true); 4703 } 4704 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4705 4706 static inline bool can_checksum_protocol(netdev_features_t features, 4707 __be16 protocol) 4708 { 4709 if (protocol == htons(ETH_P_FCOE)) 4710 return !!(features & NETIF_F_FCOE_CRC); 4711 4712 /* Assume this is an IP checksum (not SCTP CRC) */ 4713 4714 if (features & NETIF_F_HW_CSUM) { 4715 /* Can checksum everything */ 4716 return true; 4717 } 4718 4719 switch (protocol) { 4720 case htons(ETH_P_IP): 4721 return !!(features & NETIF_F_IP_CSUM); 4722 case htons(ETH_P_IPV6): 4723 return !!(features & NETIF_F_IPV6_CSUM); 4724 default: 4725 return false; 4726 } 4727 } 4728 4729 #ifdef CONFIG_BUG 4730 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4731 #else 4732 static inline void netdev_rx_csum_fault(struct net_device *dev, 4733 struct sk_buff *skb) 4734 { 4735 } 4736 #endif 4737 /* rx skb timestamps */ 4738 void net_enable_timestamp(void); 4739 void net_disable_timestamp(void); 4740 4741 #ifdef CONFIG_PROC_FS 4742 int __init dev_proc_init(void); 4743 #else 4744 #define dev_proc_init() 0 4745 #endif 4746 4747 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4748 struct sk_buff *skb, struct net_device *dev, 4749 bool more) 4750 { 4751 __this_cpu_write(softnet_data.xmit.more, more); 4752 return ops->ndo_start_xmit(skb, dev); 4753 } 4754 4755 static inline bool netdev_xmit_more(void) 4756 { 4757 return __this_cpu_read(softnet_data.xmit.more); 4758 } 4759 4760 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4761 struct netdev_queue *txq, bool more) 4762 { 4763 const struct net_device_ops *ops = dev->netdev_ops; 4764 netdev_tx_t rc; 4765 4766 rc = __netdev_start_xmit(ops, skb, dev, more); 4767 if (rc == NETDEV_TX_OK) 4768 txq_trans_update(txq); 4769 4770 return rc; 4771 } 4772 4773 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4774 const void *ns); 4775 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4776 const void *ns); 4777 4778 extern const struct kobj_ns_type_operations net_ns_type_operations; 4779 4780 const char *netdev_drivername(const struct net_device *dev); 4781 4782 void linkwatch_run_queue(void); 4783 4784 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4785 netdev_features_t f2) 4786 { 4787 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4788 if (f1 & NETIF_F_HW_CSUM) 4789 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4790 else 4791 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4792 } 4793 4794 return f1 & f2; 4795 } 4796 4797 static inline netdev_features_t netdev_get_wanted_features( 4798 struct net_device *dev) 4799 { 4800 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4801 } 4802 netdev_features_t netdev_increment_features(netdev_features_t all, 4803 netdev_features_t one, netdev_features_t mask); 4804 4805 /* Allow TSO being used on stacked device : 4806 * Performing the GSO segmentation before last device 4807 * is a performance improvement. 4808 */ 4809 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4810 netdev_features_t mask) 4811 { 4812 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4813 } 4814 4815 int __netdev_update_features(struct net_device *dev); 4816 void netdev_update_features(struct net_device *dev); 4817 void netdev_change_features(struct net_device *dev); 4818 4819 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4820 struct net_device *dev); 4821 4822 netdev_features_t passthru_features_check(struct sk_buff *skb, 4823 struct net_device *dev, 4824 netdev_features_t features); 4825 netdev_features_t netif_skb_features(struct sk_buff *skb); 4826 4827 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4828 { 4829 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4830 4831 /* check flags correspondence */ 4832 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4833 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4834 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4835 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4836 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4837 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4838 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4839 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4840 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4841 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4842 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4843 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4844 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4845 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4846 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4847 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4848 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4849 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4850 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4851 4852 return (features & feature) == feature; 4853 } 4854 4855 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4856 { 4857 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4858 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4859 } 4860 4861 static inline bool netif_needs_gso(struct sk_buff *skb, 4862 netdev_features_t features) 4863 { 4864 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4865 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4866 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4867 } 4868 4869 static inline void netif_set_gso_max_size(struct net_device *dev, 4870 unsigned int size) 4871 { 4872 dev->gso_max_size = size; 4873 } 4874 4875 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4876 int pulled_hlen, u16 mac_offset, 4877 int mac_len) 4878 { 4879 skb->protocol = protocol; 4880 skb->encapsulation = 1; 4881 skb_push(skb, pulled_hlen); 4882 skb_reset_transport_header(skb); 4883 skb->mac_header = mac_offset; 4884 skb->network_header = skb->mac_header + mac_len; 4885 skb->mac_len = mac_len; 4886 } 4887 4888 static inline bool netif_is_macsec(const struct net_device *dev) 4889 { 4890 return dev->priv_flags & IFF_MACSEC; 4891 } 4892 4893 static inline bool netif_is_macvlan(const struct net_device *dev) 4894 { 4895 return dev->priv_flags & IFF_MACVLAN; 4896 } 4897 4898 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4899 { 4900 return dev->priv_flags & IFF_MACVLAN_PORT; 4901 } 4902 4903 static inline bool netif_is_bond_master(const struct net_device *dev) 4904 { 4905 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4906 } 4907 4908 static inline bool netif_is_bond_slave(const struct net_device *dev) 4909 { 4910 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4911 } 4912 4913 static inline bool netif_supports_nofcs(struct net_device *dev) 4914 { 4915 return dev->priv_flags & IFF_SUPP_NOFCS; 4916 } 4917 4918 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4919 { 4920 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4921 } 4922 4923 static inline bool netif_is_l3_master(const struct net_device *dev) 4924 { 4925 return dev->priv_flags & IFF_L3MDEV_MASTER; 4926 } 4927 4928 static inline bool netif_is_l3_slave(const struct net_device *dev) 4929 { 4930 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4931 } 4932 4933 static inline bool netif_is_bridge_master(const struct net_device *dev) 4934 { 4935 return dev->priv_flags & IFF_EBRIDGE; 4936 } 4937 4938 static inline bool netif_is_bridge_port(const struct net_device *dev) 4939 { 4940 return dev->priv_flags & IFF_BRIDGE_PORT; 4941 } 4942 4943 static inline bool netif_is_ovs_master(const struct net_device *dev) 4944 { 4945 return dev->priv_flags & IFF_OPENVSWITCH; 4946 } 4947 4948 static inline bool netif_is_ovs_port(const struct net_device *dev) 4949 { 4950 return dev->priv_flags & IFF_OVS_DATAPATH; 4951 } 4952 4953 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4954 { 4955 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4956 } 4957 4958 static inline bool netif_is_team_master(const struct net_device *dev) 4959 { 4960 return dev->priv_flags & IFF_TEAM; 4961 } 4962 4963 static inline bool netif_is_team_port(const struct net_device *dev) 4964 { 4965 return dev->priv_flags & IFF_TEAM_PORT; 4966 } 4967 4968 static inline bool netif_is_lag_master(const struct net_device *dev) 4969 { 4970 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4971 } 4972 4973 static inline bool netif_is_lag_port(const struct net_device *dev) 4974 { 4975 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4976 } 4977 4978 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4979 { 4980 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4981 } 4982 4983 static inline bool netif_is_failover(const struct net_device *dev) 4984 { 4985 return dev->priv_flags & IFF_FAILOVER; 4986 } 4987 4988 static inline bool netif_is_failover_slave(const struct net_device *dev) 4989 { 4990 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4991 } 4992 4993 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4994 static inline void netif_keep_dst(struct net_device *dev) 4995 { 4996 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4997 } 4998 4999 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5000 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5001 { 5002 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5003 return dev->priv_flags & IFF_MACSEC; 5004 } 5005 5006 extern struct pernet_operations __net_initdata loopback_net_ops; 5007 5008 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5009 5010 /* netdev_printk helpers, similar to dev_printk */ 5011 5012 static inline const char *netdev_name(const struct net_device *dev) 5013 { 5014 if (!dev->name[0] || strchr(dev->name, '%')) 5015 return "(unnamed net_device)"; 5016 return dev->name; 5017 } 5018 5019 static inline bool netdev_unregistering(const struct net_device *dev) 5020 { 5021 return dev->reg_state == NETREG_UNREGISTERING; 5022 } 5023 5024 static inline const char *netdev_reg_state(const struct net_device *dev) 5025 { 5026 switch (dev->reg_state) { 5027 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5028 case NETREG_REGISTERED: return ""; 5029 case NETREG_UNREGISTERING: return " (unregistering)"; 5030 case NETREG_UNREGISTERED: return " (unregistered)"; 5031 case NETREG_RELEASED: return " (released)"; 5032 case NETREG_DUMMY: return " (dummy)"; 5033 } 5034 5035 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5036 return " (unknown)"; 5037 } 5038 5039 __printf(3, 4) __cold 5040 void netdev_printk(const char *level, const struct net_device *dev, 5041 const char *format, ...); 5042 __printf(2, 3) __cold 5043 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5044 __printf(2, 3) __cold 5045 void netdev_alert(const struct net_device *dev, const char *format, ...); 5046 __printf(2, 3) __cold 5047 void netdev_crit(const struct net_device *dev, const char *format, ...); 5048 __printf(2, 3) __cold 5049 void netdev_err(const struct net_device *dev, const char *format, ...); 5050 __printf(2, 3) __cold 5051 void netdev_warn(const struct net_device *dev, const char *format, ...); 5052 __printf(2, 3) __cold 5053 void netdev_notice(const struct net_device *dev, const char *format, ...); 5054 __printf(2, 3) __cold 5055 void netdev_info(const struct net_device *dev, const char *format, ...); 5056 5057 #define netdev_level_once(level, dev, fmt, ...) \ 5058 do { \ 5059 static bool __print_once __read_mostly; \ 5060 \ 5061 if (!__print_once) { \ 5062 __print_once = true; \ 5063 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5064 } \ 5065 } while (0) 5066 5067 #define netdev_emerg_once(dev, fmt, ...) \ 5068 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5069 #define netdev_alert_once(dev, fmt, ...) \ 5070 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5071 #define netdev_crit_once(dev, fmt, ...) \ 5072 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5073 #define netdev_err_once(dev, fmt, ...) \ 5074 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5075 #define netdev_warn_once(dev, fmt, ...) \ 5076 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5077 #define netdev_notice_once(dev, fmt, ...) \ 5078 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5079 #define netdev_info_once(dev, fmt, ...) \ 5080 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5081 5082 #define MODULE_ALIAS_NETDEV(device) \ 5083 MODULE_ALIAS("netdev-" device) 5084 5085 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5086 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5087 #define netdev_dbg(__dev, format, args...) \ 5088 do { \ 5089 dynamic_netdev_dbg(__dev, format, ##args); \ 5090 } while (0) 5091 #elif defined(DEBUG) 5092 #define netdev_dbg(__dev, format, args...) \ 5093 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5094 #else 5095 #define netdev_dbg(__dev, format, args...) \ 5096 ({ \ 5097 if (0) \ 5098 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5099 }) 5100 #endif 5101 5102 #if defined(VERBOSE_DEBUG) 5103 #define netdev_vdbg netdev_dbg 5104 #else 5105 5106 #define netdev_vdbg(dev, format, args...) \ 5107 ({ \ 5108 if (0) \ 5109 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5110 0; \ 5111 }) 5112 #endif 5113 5114 /* 5115 * netdev_WARN() acts like dev_printk(), but with the key difference 5116 * of using a WARN/WARN_ON to get the message out, including the 5117 * file/line information and a backtrace. 5118 */ 5119 #define netdev_WARN(dev, format, args...) \ 5120 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5121 netdev_reg_state(dev), ##args) 5122 5123 #define netdev_WARN_ONCE(dev, format, args...) \ 5124 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5125 netdev_reg_state(dev), ##args) 5126 5127 /* netif printk helpers, similar to netdev_printk */ 5128 5129 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5130 do { \ 5131 if (netif_msg_##type(priv)) \ 5132 netdev_printk(level, (dev), fmt, ##args); \ 5133 } while (0) 5134 5135 #define netif_level(level, priv, type, dev, fmt, args...) \ 5136 do { \ 5137 if (netif_msg_##type(priv)) \ 5138 netdev_##level(dev, fmt, ##args); \ 5139 } while (0) 5140 5141 #define netif_emerg(priv, type, dev, fmt, args...) \ 5142 netif_level(emerg, priv, type, dev, fmt, ##args) 5143 #define netif_alert(priv, type, dev, fmt, args...) \ 5144 netif_level(alert, priv, type, dev, fmt, ##args) 5145 #define netif_crit(priv, type, dev, fmt, args...) \ 5146 netif_level(crit, priv, type, dev, fmt, ##args) 5147 #define netif_err(priv, type, dev, fmt, args...) \ 5148 netif_level(err, priv, type, dev, fmt, ##args) 5149 #define netif_warn(priv, type, dev, fmt, args...) \ 5150 netif_level(warn, priv, type, dev, fmt, ##args) 5151 #define netif_notice(priv, type, dev, fmt, args...) \ 5152 netif_level(notice, priv, type, dev, fmt, ##args) 5153 #define netif_info(priv, type, dev, fmt, args...) \ 5154 netif_level(info, priv, type, dev, fmt, ##args) 5155 5156 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5157 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5158 #define netif_dbg(priv, type, netdev, format, args...) \ 5159 do { \ 5160 if (netif_msg_##type(priv)) \ 5161 dynamic_netdev_dbg(netdev, format, ##args); \ 5162 } while (0) 5163 #elif defined(DEBUG) 5164 #define netif_dbg(priv, type, dev, format, args...) \ 5165 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5166 #else 5167 #define netif_dbg(priv, type, dev, format, args...) \ 5168 ({ \ 5169 if (0) \ 5170 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5171 0; \ 5172 }) 5173 #endif 5174 5175 /* if @cond then downgrade to debug, else print at @level */ 5176 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5177 do { \ 5178 if (cond) \ 5179 netif_dbg(priv, type, netdev, fmt, ##args); \ 5180 else \ 5181 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5182 } while (0) 5183 5184 #if defined(VERBOSE_DEBUG) 5185 #define netif_vdbg netif_dbg 5186 #else 5187 #define netif_vdbg(priv, type, dev, format, args...) \ 5188 ({ \ 5189 if (0) \ 5190 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5191 0; \ 5192 }) 5193 #endif 5194 5195 /* 5196 * The list of packet types we will receive (as opposed to discard) 5197 * and the routines to invoke. 5198 * 5199 * Why 16. Because with 16 the only overlap we get on a hash of the 5200 * low nibble of the protocol value is RARP/SNAP/X.25. 5201 * 5202 * 0800 IP 5203 * 0001 802.3 5204 * 0002 AX.25 5205 * 0004 802.2 5206 * 8035 RARP 5207 * 0005 SNAP 5208 * 0805 X.25 5209 * 0806 ARP 5210 * 8137 IPX 5211 * 0009 Localtalk 5212 * 86DD IPv6 5213 */ 5214 #define PTYPE_HASH_SIZE (16) 5215 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5216 5217 extern struct net_device *blackhole_netdev; 5218 5219 #endif /* _LINUX_NETDEVICE_H */ 5220