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, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 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 #include <asm/local.h> 32 33 #include <linux/percpu.h> 34 #include <linux/rculist.h> 35 #include <linux/workqueue.h> 36 #include <linux/dynamic_queue_limits.h> 37 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 #include <linux/rbtree.h> 52 #include <net/net_trackers.h> 53 #include <net/net_debug.h> 54 #include <linux/android_kabi.h> 55 56 struct netpoll_info; 57 struct device; 58 struct ethtool_ops; 59 struct phy_device; 60 struct dsa_port; 61 struct ip_tunnel_parm; 62 struct macsec_context; 63 struct macsec_ops; 64 struct netdev_name_node; 65 struct sd_flow_limit; 66 struct sfp_bus; 67 /* 802.11 specific */ 68 struct wireless_dev; 69 /* 802.15.4 specific */ 70 struct wpan_dev; 71 struct mpls_dev; 72 /* UDP Tunnel offloads */ 73 struct udp_tunnel_info; 74 struct udp_tunnel_nic_info; 75 struct udp_tunnel_nic; 76 struct bpf_prog; 77 struct xdp_buff; 78 79 void synchronize_net(void); 80 void netdev_set_default_ethtool_ops(struct net_device *dev, 81 const struct ethtool_ops *ops); 82 83 /* Backlog congestion levels */ 84 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 85 #define NET_RX_DROP 1 /* packet dropped */ 86 87 #define MAX_NEST_DEV 8 88 89 /* 90 * Transmit return codes: transmit return codes originate from three different 91 * namespaces: 92 * 93 * - qdisc return codes 94 * - driver transmit return codes 95 * - errno values 96 * 97 * Drivers are allowed to return any one of those in their hard_start_xmit() 98 * function. Real network devices commonly used with qdiscs should only return 99 * the driver transmit return codes though - when qdiscs are used, the actual 100 * transmission happens asynchronously, so the value is not propagated to 101 * higher layers. Virtual network devices transmit synchronously; in this case 102 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 103 * others are propagated to higher layers. 104 */ 105 106 /* qdisc ->enqueue() return codes. */ 107 #define NET_XMIT_SUCCESS 0x00 108 #define NET_XMIT_DROP 0x01 /* skb dropped */ 109 #define NET_XMIT_CN 0x02 /* congestion notification */ 110 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 111 112 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 113 * indicates that the device will soon be dropping packets, or already drops 114 * some packets of the same priority; prompting us to send less aggressively. */ 115 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 116 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 117 118 /* Driver transmit return codes */ 119 #define NETDEV_TX_MASK 0xf0 120 121 enum netdev_tx { 122 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 123 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 124 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 125 }; 126 typedef enum netdev_tx netdev_tx_t; 127 128 /* 129 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 130 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 131 */ dev_xmit_complete(int rc)132 static inline bool dev_xmit_complete(int rc) 133 { 134 /* 135 * Positive cases with an skb consumed by a driver: 136 * - successful transmission (rc == NETDEV_TX_OK) 137 * - error while transmitting (rc < 0) 138 * - error while queueing to a different device (rc & NET_XMIT_MASK) 139 */ 140 if (likely(rc < NET_XMIT_MASK)) 141 return true; 142 143 return false; 144 } 145 146 /* 147 * Compute the worst-case header length according to the protocols 148 * used. 149 */ 150 151 #if defined(CONFIG_HYPERV_NET) 152 # define LL_MAX_HEADER 128 153 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 154 # if defined(CONFIG_MAC80211_MESH) 155 # define LL_MAX_HEADER 128 156 # else 157 # define LL_MAX_HEADER 96 158 # endif 159 #else 160 # define LL_MAX_HEADER 32 161 #endif 162 163 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 164 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 165 #define MAX_HEADER LL_MAX_HEADER 166 #else 167 #define MAX_HEADER (LL_MAX_HEADER + 48) 168 #endif 169 170 /* 171 * Old network device statistics. Fields are native words 172 * (unsigned long) so they can be read and written atomically. 173 */ 174 175 #define NET_DEV_STAT(FIELD) \ 176 union { \ 177 unsigned long FIELD; \ 178 atomic_long_t __##FIELD; \ 179 } 180 181 struct net_device_stats { 182 NET_DEV_STAT(rx_packets); 183 NET_DEV_STAT(tx_packets); 184 NET_DEV_STAT(rx_bytes); 185 NET_DEV_STAT(tx_bytes); 186 NET_DEV_STAT(rx_errors); 187 NET_DEV_STAT(tx_errors); 188 NET_DEV_STAT(rx_dropped); 189 NET_DEV_STAT(tx_dropped); 190 NET_DEV_STAT(multicast); 191 NET_DEV_STAT(collisions); 192 NET_DEV_STAT(rx_length_errors); 193 NET_DEV_STAT(rx_over_errors); 194 NET_DEV_STAT(rx_crc_errors); 195 NET_DEV_STAT(rx_frame_errors); 196 NET_DEV_STAT(rx_fifo_errors); 197 NET_DEV_STAT(rx_missed_errors); 198 NET_DEV_STAT(tx_aborted_errors); 199 NET_DEV_STAT(tx_carrier_errors); 200 NET_DEV_STAT(tx_fifo_errors); 201 NET_DEV_STAT(tx_heartbeat_errors); 202 NET_DEV_STAT(tx_window_errors); 203 NET_DEV_STAT(rx_compressed); 204 NET_DEV_STAT(tx_compressed); 205 }; 206 #undef NET_DEV_STAT 207 208 /* per-cpu stats, allocated on demand. 209 * Try to fit them in a single cache line, for dev_get_stats() sake. 210 */ 211 struct net_device_core_stats { 212 unsigned long rx_dropped; 213 unsigned long tx_dropped; 214 unsigned long rx_nohandler; 215 unsigned long rx_otherhost_dropped; 216 } __aligned(4 * sizeof(unsigned long)); 217 218 #include <linux/cache.h> 219 #include <linux/skbuff.h> 220 221 #ifdef CONFIG_RPS 222 #include <linux/static_key.h> 223 extern struct static_key_false rps_needed; 224 extern struct static_key_false rfs_needed; 225 #endif 226 227 struct neighbour; 228 struct neigh_parms; 229 struct sk_buff; 230 231 struct netdev_hw_addr { 232 struct list_head list; 233 struct rb_node node; 234 unsigned char addr[MAX_ADDR_LEN]; 235 unsigned char type; 236 #define NETDEV_HW_ADDR_T_LAN 1 237 #define NETDEV_HW_ADDR_T_SAN 2 238 #define NETDEV_HW_ADDR_T_UNICAST 3 239 #define NETDEV_HW_ADDR_T_MULTICAST 4 240 bool global_use; 241 int sync_cnt; 242 int refcount; 243 int synced; 244 struct rcu_head rcu_head; 245 }; 246 247 struct netdev_hw_addr_list { 248 struct list_head list; 249 int count; 250 251 /* Auxiliary tree for faster lookup on addition and deletion */ 252 struct rb_root tree; 253 }; 254 255 #define netdev_hw_addr_list_count(l) ((l)->count) 256 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 257 #define netdev_hw_addr_list_for_each(ha, l) \ 258 list_for_each_entry(ha, &(l)->list, list) 259 260 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 261 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 262 #define netdev_for_each_uc_addr(ha, dev) \ 263 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 264 #define netdev_for_each_synced_uc_addr(_ha, _dev) \ 265 netdev_for_each_uc_addr((_ha), (_dev)) \ 266 if ((_ha)->sync_cnt) 267 268 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 269 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 270 #define netdev_for_each_mc_addr(ha, dev) \ 271 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 272 #define netdev_for_each_synced_mc_addr(_ha, _dev) \ 273 netdev_for_each_mc_addr((_ha), (_dev)) \ 274 if ((_ha)->sync_cnt) 275 276 struct hh_cache { 277 unsigned int hh_len; 278 seqlock_t hh_lock; 279 280 /* cached hardware header; allow for machine alignment needs. */ 281 #define HH_DATA_MOD 16 282 #define HH_DATA_OFF(__len) \ 283 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 284 #define HH_DATA_ALIGN(__len) \ 285 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 286 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 287 }; 288 289 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 290 * Alternative is: 291 * dev->hard_header_len ? (dev->hard_header_len + 292 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 293 * 294 * We could use other alignment values, but we must maintain the 295 * relationship HH alignment <= LL alignment. 296 */ 297 #define LL_RESERVED_SPACE(dev) \ 298 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 299 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 300 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 301 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 302 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 303 304 struct header_ops { 305 int (*create) (struct sk_buff *skb, struct net_device *dev, 306 unsigned short type, const void *daddr, 307 const void *saddr, unsigned int len); 308 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 309 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 310 void (*cache_update)(struct hh_cache *hh, 311 const struct net_device *dev, 312 const unsigned char *haddr); 313 bool (*validate)(const char *ll_header, unsigned int len); 314 __be16 (*parse_protocol)(const struct sk_buff *skb); 315 316 ANDROID_KABI_RESERVE(1); 317 ANDROID_KABI_RESERVE(2); 318 }; 319 320 /* These flag bits are private to the generic network queueing 321 * layer; they may not be explicitly referenced by any other 322 * code. 323 */ 324 325 enum netdev_state_t { 326 __LINK_STATE_START, 327 __LINK_STATE_PRESENT, 328 __LINK_STATE_NOCARRIER, 329 __LINK_STATE_LINKWATCH_PENDING, 330 __LINK_STATE_DORMANT, 331 __LINK_STATE_TESTING, 332 }; 333 334 struct gro_list { 335 struct list_head list; 336 int count; 337 }; 338 339 /* 340 * size of gro hash buckets, must less than bit number of 341 * napi_struct::gro_bitmask 342 */ 343 #define GRO_HASH_BUCKETS 8 344 345 /* 346 * Structure for NAPI scheduling similar to tasklet but with weighting 347 */ 348 struct napi_struct { 349 /* The poll_list must only be managed by the entity which 350 * changes the state of the NAPI_STATE_SCHED bit. This means 351 * whoever atomically sets that bit can add this napi_struct 352 * to the per-CPU poll_list, and whoever clears that bit 353 * can remove from the list right before clearing the bit. 354 */ 355 struct list_head poll_list; 356 357 unsigned long state; 358 int weight; 359 int defer_hard_irqs_count; 360 unsigned long gro_bitmask; 361 int (*poll)(struct napi_struct *, int); 362 #ifdef CONFIG_NETPOLL 363 int poll_owner; 364 #endif 365 struct net_device *dev; 366 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 367 struct sk_buff *skb; 368 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 369 int rx_count; /* length of rx_list */ 370 struct hrtimer timer; 371 struct list_head dev_list; 372 struct hlist_node napi_hash_node; 373 unsigned int napi_id; 374 struct task_struct *thread; 375 376 ANDROID_KABI_RESERVE(1); 377 ANDROID_KABI_RESERVE(2); 378 ANDROID_KABI_RESERVE(3); 379 ANDROID_KABI_RESERVE(4); 380 }; 381 382 enum { 383 NAPI_STATE_SCHED, /* Poll is scheduled */ 384 NAPI_STATE_MISSED, /* reschedule a napi */ 385 NAPI_STATE_DISABLE, /* Disable pending */ 386 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 387 NAPI_STATE_LISTED, /* NAPI added to system lists */ 388 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 389 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 390 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 391 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 392 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 393 }; 394 395 enum { 396 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 397 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 398 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 399 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 400 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 401 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 402 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 403 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 404 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 405 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 406 }; 407 408 enum gro_result { 409 GRO_MERGED, 410 GRO_MERGED_FREE, 411 GRO_HELD, 412 GRO_NORMAL, 413 GRO_CONSUMED, 414 }; 415 typedef enum gro_result gro_result_t; 416 417 /* 418 * enum rx_handler_result - Possible return values for rx_handlers. 419 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 420 * further. 421 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 422 * case skb->dev was changed by rx_handler. 423 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 424 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 425 * 426 * rx_handlers are functions called from inside __netif_receive_skb(), to do 427 * special processing of the skb, prior to delivery to protocol handlers. 428 * 429 * Currently, a net_device can only have a single rx_handler registered. Trying 430 * to register a second rx_handler will return -EBUSY. 431 * 432 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 433 * To unregister a rx_handler on a net_device, use 434 * netdev_rx_handler_unregister(). 435 * 436 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 437 * do with the skb. 438 * 439 * If the rx_handler consumed the skb in some way, it should return 440 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 441 * the skb to be delivered in some other way. 442 * 443 * If the rx_handler changed skb->dev, to divert the skb to another 444 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 445 * new device will be called if it exists. 446 * 447 * If the rx_handler decides the skb should be ignored, it should return 448 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 449 * are registered on exact device (ptype->dev == skb->dev). 450 * 451 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 452 * delivered, it should return RX_HANDLER_PASS. 453 * 454 * A device without a registered rx_handler will behave as if rx_handler 455 * returned RX_HANDLER_PASS. 456 */ 457 458 enum rx_handler_result { 459 RX_HANDLER_CONSUMED, 460 RX_HANDLER_ANOTHER, 461 RX_HANDLER_EXACT, 462 RX_HANDLER_PASS, 463 }; 464 typedef enum rx_handler_result rx_handler_result_t; 465 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 466 467 void __napi_schedule(struct napi_struct *n); 468 void __napi_schedule_irqoff(struct napi_struct *n); 469 napi_disable_pending(struct napi_struct * n)470 static inline bool napi_disable_pending(struct napi_struct *n) 471 { 472 return test_bit(NAPI_STATE_DISABLE, &n->state); 473 } 474 napi_prefer_busy_poll(struct napi_struct * n)475 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 476 { 477 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 478 } 479 480 bool napi_schedule_prep(struct napi_struct *n); 481 482 /** 483 * napi_schedule - schedule NAPI poll 484 * @n: NAPI context 485 * 486 * Schedule NAPI poll routine to be called if it is not already 487 * running. 488 */ napi_schedule(struct napi_struct * n)489 static inline void napi_schedule(struct napi_struct *n) 490 { 491 if (napi_schedule_prep(n)) 492 __napi_schedule(n); 493 } 494 495 /** 496 * napi_schedule_irqoff - schedule NAPI poll 497 * @n: NAPI context 498 * 499 * Variant of napi_schedule(), assuming hard irqs are masked. 500 */ napi_schedule_irqoff(struct napi_struct * n)501 static inline void napi_schedule_irqoff(struct napi_struct *n) 502 { 503 if (napi_schedule_prep(n)) 504 __napi_schedule_irqoff(n); 505 } 506 507 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ napi_reschedule(struct napi_struct * napi)508 static inline bool napi_reschedule(struct napi_struct *napi) 509 { 510 if (napi_schedule_prep(napi)) { 511 __napi_schedule(napi); 512 return true; 513 } 514 return false; 515 } 516 517 bool napi_complete_done(struct napi_struct *n, int work_done); 518 /** 519 * napi_complete - NAPI processing complete 520 * @n: NAPI context 521 * 522 * Mark NAPI processing as complete. 523 * Consider using napi_complete_done() instead. 524 * Return false if device should avoid rearming interrupts. 525 */ napi_complete(struct napi_struct * n)526 static inline bool napi_complete(struct napi_struct *n) 527 { 528 return napi_complete_done(n, 0); 529 } 530 531 int dev_set_threaded(struct net_device *dev, bool threaded); 532 533 /** 534 * napi_disable - prevent NAPI from scheduling 535 * @n: NAPI context 536 * 537 * Stop NAPI from being scheduled on this context. 538 * Waits till any outstanding processing completes. 539 */ 540 void napi_disable(struct napi_struct *n); 541 542 void napi_enable(struct napi_struct *n); 543 544 /** 545 * napi_synchronize - wait until NAPI is not running 546 * @n: NAPI context 547 * 548 * Wait until NAPI is done being scheduled on this context. 549 * Waits till any outstanding processing completes but 550 * does not disable future activations. 551 */ napi_synchronize(const struct napi_struct * n)552 static inline void napi_synchronize(const struct napi_struct *n) 553 { 554 if (IS_ENABLED(CONFIG_SMP)) 555 while (test_bit(NAPI_STATE_SCHED, &n->state)) 556 msleep(1); 557 else 558 barrier(); 559 } 560 561 /** 562 * napi_if_scheduled_mark_missed - if napi is running, set the 563 * NAPIF_STATE_MISSED 564 * @n: NAPI context 565 * 566 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 567 * NAPI is scheduled. 568 **/ napi_if_scheduled_mark_missed(struct napi_struct * n)569 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 570 { 571 unsigned long val, new; 572 573 val = READ_ONCE(n->state); 574 do { 575 if (val & NAPIF_STATE_DISABLE) 576 return true; 577 578 if (!(val & NAPIF_STATE_SCHED)) 579 return false; 580 581 new = val | NAPIF_STATE_MISSED; 582 } while (!try_cmpxchg(&n->state, &val, new)); 583 584 return true; 585 } 586 587 enum netdev_queue_state_t { 588 __QUEUE_STATE_DRV_XOFF, 589 __QUEUE_STATE_STACK_XOFF, 590 __QUEUE_STATE_FROZEN, 591 }; 592 593 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 594 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 595 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 596 597 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 598 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 599 QUEUE_STATE_FROZEN) 600 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 601 QUEUE_STATE_FROZEN) 602 603 /* 604 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 605 * netif_tx_* functions below are used to manipulate this flag. The 606 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 607 * queue independently. The netif_xmit_*stopped functions below are called 608 * to check if the queue has been stopped by the driver or stack (either 609 * of the XOFF bits are set in the state). Drivers should not need to call 610 * netif_xmit*stopped functions, they should only be using netif_tx_*. 611 */ 612 613 struct netdev_queue { 614 /* 615 * read-mostly part 616 */ 617 struct net_device *dev; 618 netdevice_tracker dev_tracker; 619 620 struct Qdisc __rcu *qdisc; 621 struct Qdisc __rcu *qdisc_sleeping; 622 #ifdef CONFIG_SYSFS 623 struct kobject kobj; 624 #endif 625 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 626 int numa_node; 627 #endif 628 unsigned long tx_maxrate; 629 /* 630 * Number of TX timeouts for this queue 631 * (/sys/class/net/DEV/Q/trans_timeout) 632 */ 633 atomic_long_t trans_timeout; 634 635 /* Subordinate device that the queue has been assigned to */ 636 struct net_device *sb_dev; 637 #ifdef CONFIG_XDP_SOCKETS 638 struct xsk_buff_pool *pool; 639 #endif 640 /* 641 * write-mostly part 642 */ 643 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 644 int xmit_lock_owner; 645 /* 646 * Time (in jiffies) of last Tx 647 */ 648 unsigned long trans_start; 649 650 unsigned long state; 651 652 #ifdef CONFIG_BQL 653 struct dql dql; 654 #endif 655 656 ANDROID_KABI_RESERVE(1); 657 ANDROID_KABI_RESERVE(2); 658 ANDROID_KABI_RESERVE(3); 659 ANDROID_KABI_RESERVE(4); 660 } ____cacheline_aligned_in_smp; 661 662 extern int sysctl_fb_tunnels_only_for_init_net; 663 extern int sysctl_devconf_inherit_init_net; 664 665 /* 666 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 667 * == 1 : For initns only 668 * == 2 : For none. 669 */ net_has_fallback_tunnels(const struct net * net)670 static inline bool net_has_fallback_tunnels(const struct net *net) 671 { 672 #if IS_ENABLED(CONFIG_SYSCTL) 673 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 674 675 return !fb_tunnels_only_for_init_net || 676 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 677 #else 678 return true; 679 #endif 680 } 681 net_inherit_devconf(void)682 static inline int net_inherit_devconf(void) 683 { 684 #if IS_ENABLED(CONFIG_SYSCTL) 685 return READ_ONCE(sysctl_devconf_inherit_init_net); 686 #else 687 return 0; 688 #endif 689 } 690 netdev_queue_numa_node_read(const struct netdev_queue * q)691 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 692 { 693 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 694 return q->numa_node; 695 #else 696 return NUMA_NO_NODE; 697 #endif 698 } 699 netdev_queue_numa_node_write(struct netdev_queue * q,int node)700 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 701 { 702 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 703 q->numa_node = node; 704 #endif 705 } 706 707 #ifdef CONFIG_RPS 708 /* 709 * This structure holds an RPS map which can be of variable length. The 710 * map is an array of CPUs. 711 */ 712 struct rps_map { 713 unsigned int len; 714 struct rcu_head rcu; 715 u16 cpus[]; 716 }; 717 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 718 719 /* 720 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 721 * tail pointer for that CPU's input queue at the time of last enqueue, and 722 * a hardware filter index. 723 */ 724 struct rps_dev_flow { 725 u16 cpu; 726 u16 filter; 727 unsigned int last_qtail; 728 }; 729 #define RPS_NO_FILTER 0xffff 730 731 /* 732 * The rps_dev_flow_table structure contains a table of flow mappings. 733 */ 734 struct rps_dev_flow_table { 735 unsigned int mask; 736 struct rcu_head rcu; 737 struct rps_dev_flow flows[]; 738 }; 739 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 740 ((_num) * sizeof(struct rps_dev_flow))) 741 742 /* 743 * The rps_sock_flow_table contains mappings of flows to the last CPU 744 * on which they were processed by the application (set in recvmsg). 745 * Each entry is a 32bit value. Upper part is the high-order bits 746 * of flow hash, lower part is CPU number. 747 * rps_cpu_mask is used to partition the space, depending on number of 748 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 749 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 750 * meaning we use 32-6=26 bits for the hash. 751 */ 752 struct rps_sock_flow_table { 753 u32 mask; 754 755 u32 ents[] ____cacheline_aligned_in_smp; 756 }; 757 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 758 759 #define RPS_NO_CPU 0xffff 760 761 extern u32 rps_cpu_mask; 762 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 763 rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)764 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 765 u32 hash) 766 { 767 if (table && hash) { 768 unsigned int index = hash & table->mask; 769 u32 val = hash & ~rps_cpu_mask; 770 771 /* We only give a hint, preemption can change CPU under us */ 772 val |= raw_smp_processor_id(); 773 774 /* The following WRITE_ONCE() is paired with the READ_ONCE() 775 * here, and another one in get_rps_cpu(). 776 */ 777 if (READ_ONCE(table->ents[index]) != val) 778 WRITE_ONCE(table->ents[index], val); 779 } 780 } 781 782 #ifdef CONFIG_RFS_ACCEL 783 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 784 u16 filter_id); 785 #endif 786 #endif /* CONFIG_RPS */ 787 788 /* This structure contains an instance of an RX queue. */ 789 struct netdev_rx_queue { 790 struct xdp_rxq_info xdp_rxq; 791 #ifdef CONFIG_RPS 792 struct rps_map __rcu *rps_map; 793 struct rps_dev_flow_table __rcu *rps_flow_table; 794 #endif 795 struct kobject kobj; 796 struct net_device *dev; 797 netdevice_tracker dev_tracker; 798 799 #ifdef CONFIG_XDP_SOCKETS 800 struct xsk_buff_pool *pool; 801 #endif 802 803 ANDROID_KABI_RESERVE(1); 804 ANDROID_KABI_RESERVE(2); 805 ANDROID_KABI_RESERVE(3); 806 ANDROID_KABI_RESERVE(4); 807 } ____cacheline_aligned_in_smp; 808 809 /* 810 * RX queue sysfs structures and functions. 811 */ 812 struct rx_queue_attribute { 813 struct attribute attr; 814 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 815 ssize_t (*store)(struct netdev_rx_queue *queue, 816 const char *buf, size_t len); 817 }; 818 819 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 820 enum xps_map_type { 821 XPS_CPUS = 0, 822 XPS_RXQS, 823 XPS_MAPS_MAX, 824 }; 825 826 #ifdef CONFIG_XPS 827 /* 828 * This structure holds an XPS map which can be of variable length. The 829 * map is an array of queues. 830 */ 831 struct xps_map { 832 unsigned int len; 833 unsigned int alloc_len; 834 struct rcu_head rcu; 835 u16 queues[]; 836 }; 837 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 838 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 839 - sizeof(struct xps_map)) / sizeof(u16)) 840 841 /* 842 * This structure holds all XPS maps for device. Maps are indexed by CPU. 843 * 844 * We keep track of the number of cpus/rxqs used when the struct is allocated, 845 * in nr_ids. This will help not accessing out-of-bound memory. 846 * 847 * We keep track of the number of traffic classes used when the struct is 848 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 849 * not crossing its upper bound, as the original dev->num_tc can be updated in 850 * the meantime. 851 */ 852 struct xps_dev_maps { 853 struct rcu_head rcu; 854 unsigned int nr_ids; 855 s16 num_tc; 856 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 857 }; 858 859 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 860 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 861 862 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 863 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 864 865 #endif /* CONFIG_XPS */ 866 867 #define TC_MAX_QUEUE 16 868 #define TC_BITMASK 15 869 /* HW offloaded queuing disciplines txq count and offset maps */ 870 struct netdev_tc_txq { 871 u16 count; 872 u16 offset; 873 }; 874 875 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 876 /* 877 * This structure is to hold information about the device 878 * configured to run FCoE protocol stack. 879 */ 880 struct netdev_fcoe_hbainfo { 881 char manufacturer[64]; 882 char serial_number[64]; 883 char hardware_version[64]; 884 char driver_version[64]; 885 char optionrom_version[64]; 886 char firmware_version[64]; 887 char model[256]; 888 char model_description[256]; 889 }; 890 #endif 891 892 #define MAX_PHYS_ITEM_ID_LEN 32 893 894 /* This structure holds a unique identifier to identify some 895 * physical item (port for example) used by a netdevice. 896 */ 897 struct netdev_phys_item_id { 898 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 899 unsigned char id_len; 900 }; 901 netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)902 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 903 struct netdev_phys_item_id *b) 904 { 905 return a->id_len == b->id_len && 906 memcmp(a->id, b->id, a->id_len) == 0; 907 } 908 909 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 910 struct sk_buff *skb, 911 struct net_device *sb_dev); 912 913 enum net_device_path_type { 914 DEV_PATH_ETHERNET = 0, 915 DEV_PATH_VLAN, 916 DEV_PATH_BRIDGE, 917 DEV_PATH_PPPOE, 918 DEV_PATH_DSA, 919 DEV_PATH_MTK_WDMA, 920 }; 921 922 struct net_device_path { 923 enum net_device_path_type type; 924 const struct net_device *dev; 925 union { 926 struct { 927 u16 id; 928 __be16 proto; 929 u8 h_dest[ETH_ALEN]; 930 } encap; 931 struct { 932 enum { 933 DEV_PATH_BR_VLAN_KEEP, 934 DEV_PATH_BR_VLAN_TAG, 935 DEV_PATH_BR_VLAN_UNTAG, 936 DEV_PATH_BR_VLAN_UNTAG_HW, 937 } vlan_mode; 938 u16 vlan_id; 939 __be16 vlan_proto; 940 } bridge; 941 struct { 942 int port; 943 u16 proto; 944 } dsa; 945 struct { 946 u8 wdma_idx; 947 u8 queue; 948 u16 wcid; 949 u8 bss; 950 } mtk_wdma; 951 }; 952 }; 953 954 #define NET_DEVICE_PATH_STACK_MAX 5 955 #define NET_DEVICE_PATH_VLAN_MAX 2 956 957 struct net_device_path_stack { 958 int num_paths; 959 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 960 }; 961 962 struct net_device_path_ctx { 963 const struct net_device *dev; 964 u8 daddr[ETH_ALEN]; 965 966 int num_vlans; 967 struct { 968 u16 id; 969 __be16 proto; 970 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 971 }; 972 973 enum tc_setup_type { 974 TC_QUERY_CAPS, 975 TC_SETUP_QDISC_MQPRIO, 976 TC_SETUP_CLSU32, 977 TC_SETUP_CLSFLOWER, 978 TC_SETUP_CLSMATCHALL, 979 TC_SETUP_CLSBPF, 980 TC_SETUP_BLOCK, 981 TC_SETUP_QDISC_CBS, 982 TC_SETUP_QDISC_RED, 983 TC_SETUP_QDISC_PRIO, 984 TC_SETUP_QDISC_MQ, 985 TC_SETUP_QDISC_ETF, 986 TC_SETUP_ROOT_QDISC, 987 TC_SETUP_QDISC_GRED, 988 TC_SETUP_QDISC_TAPRIO, 989 TC_SETUP_FT, 990 TC_SETUP_QDISC_ETS, 991 TC_SETUP_QDISC_TBF, 992 TC_SETUP_QDISC_FIFO, 993 TC_SETUP_QDISC_HTB, 994 TC_SETUP_ACT, 995 }; 996 997 /* These structures hold the attributes of bpf state that are being passed 998 * to the netdevice through the bpf op. 999 */ 1000 enum bpf_netdev_command { 1001 /* Set or clear a bpf program used in the earliest stages of packet 1002 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 1003 * is responsible for calling bpf_prog_put on any old progs that are 1004 * stored. In case of error, the callee need not release the new prog 1005 * reference, but on success it takes ownership and must bpf_prog_put 1006 * when it is no longer used. 1007 */ 1008 XDP_SETUP_PROG, 1009 XDP_SETUP_PROG_HW, 1010 /* BPF program for offload callbacks, invoked at program load time. */ 1011 BPF_OFFLOAD_MAP_ALLOC, 1012 BPF_OFFLOAD_MAP_FREE, 1013 XDP_SETUP_XSK_POOL, 1014 }; 1015 1016 struct bpf_prog_offload_ops; 1017 struct netlink_ext_ack; 1018 struct xdp_umem; 1019 struct xdp_dev_bulk_queue; 1020 struct bpf_xdp_link; 1021 1022 enum bpf_xdp_mode { 1023 XDP_MODE_SKB = 0, 1024 XDP_MODE_DRV = 1, 1025 XDP_MODE_HW = 2, 1026 __MAX_XDP_MODE 1027 }; 1028 1029 struct bpf_xdp_entity { 1030 struct bpf_prog *prog; 1031 struct bpf_xdp_link *link; 1032 }; 1033 1034 struct netdev_bpf { 1035 enum bpf_netdev_command command; 1036 union { 1037 /* XDP_SETUP_PROG */ 1038 struct { 1039 u32 flags; 1040 struct bpf_prog *prog; 1041 struct netlink_ext_ack *extack; 1042 }; 1043 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1044 struct { 1045 struct bpf_offloaded_map *offmap; 1046 }; 1047 /* XDP_SETUP_XSK_POOL */ 1048 struct { 1049 struct xsk_buff_pool *pool; 1050 u16 queue_id; 1051 } xsk; 1052 }; 1053 }; 1054 1055 /* Flags for ndo_xsk_wakeup. */ 1056 #define XDP_WAKEUP_RX (1 << 0) 1057 #define XDP_WAKEUP_TX (1 << 1) 1058 1059 #ifdef CONFIG_XFRM_OFFLOAD 1060 struct xfrmdev_ops { 1061 int (*xdo_dev_state_add) (struct xfrm_state *x); 1062 void (*xdo_dev_state_delete) (struct xfrm_state *x); 1063 void (*xdo_dev_state_free) (struct xfrm_state *x); 1064 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1065 struct xfrm_state *x); 1066 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1067 1068 ANDROID_KABI_RESERVE(1); 1069 ANDROID_KABI_RESERVE(2); 1070 ANDROID_KABI_RESERVE(3); 1071 ANDROID_KABI_RESERVE(4); 1072 }; 1073 #endif 1074 1075 struct dev_ifalias { 1076 struct rcu_head rcuhead; 1077 char ifalias[]; 1078 }; 1079 1080 struct devlink; 1081 struct tlsdev_ops; 1082 1083 struct netdev_net_notifier { 1084 struct list_head list; 1085 struct notifier_block *nb; 1086 }; 1087 1088 /* 1089 * This structure defines the management hooks for network devices. 1090 * The following hooks can be defined; unless noted otherwise, they are 1091 * optional and can be filled with a null pointer. 1092 * 1093 * int (*ndo_init)(struct net_device *dev); 1094 * This function is called once when a network device is registered. 1095 * The network device can use this for any late stage initialization 1096 * or semantic validation. It can fail with an error code which will 1097 * be propagated back to register_netdev. 1098 * 1099 * void (*ndo_uninit)(struct net_device *dev); 1100 * This function is called when device is unregistered or when registration 1101 * fails. It is not called if init fails. 1102 * 1103 * int (*ndo_open)(struct net_device *dev); 1104 * This function is called when a network device transitions to the up 1105 * state. 1106 * 1107 * int (*ndo_stop)(struct net_device *dev); 1108 * This function is called when a network device transitions to the down 1109 * state. 1110 * 1111 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1112 * struct net_device *dev); 1113 * Called when a packet needs to be transmitted. 1114 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1115 * the queue before that can happen; it's for obsolete devices and weird 1116 * corner cases, but the stack really does a non-trivial amount 1117 * of useless work if you return NETDEV_TX_BUSY. 1118 * Required; cannot be NULL. 1119 * 1120 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1121 * struct net_device *dev 1122 * netdev_features_t features); 1123 * Called by core transmit path to determine if device is capable of 1124 * performing offload operations on a given packet. This is to give 1125 * the device an opportunity to implement any restrictions that cannot 1126 * be otherwise expressed by feature flags. The check is called with 1127 * the set of features that the stack has calculated and it returns 1128 * those the driver believes to be appropriate. 1129 * 1130 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1131 * struct net_device *sb_dev); 1132 * Called to decide which queue to use when device supports multiple 1133 * transmit queues. 1134 * 1135 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1136 * This function is called to allow device receiver to make 1137 * changes to configuration when multicast or promiscuous is enabled. 1138 * 1139 * void (*ndo_set_rx_mode)(struct net_device *dev); 1140 * This function is called device changes address list filtering. 1141 * If driver handles unicast address filtering, it should set 1142 * IFF_UNICAST_FLT in its priv_flags. 1143 * 1144 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1145 * This function is called when the Media Access Control address 1146 * needs to be changed. If this interface is not defined, the 1147 * MAC address can not be changed. 1148 * 1149 * int (*ndo_validate_addr)(struct net_device *dev); 1150 * Test if Media Access Control address is valid for the device. 1151 * 1152 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1153 * Old-style ioctl entry point. This is used internally by the 1154 * appletalk and ieee802154 subsystems but is no longer called by 1155 * the device ioctl handler. 1156 * 1157 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1158 * Used by the bonding driver for its device specific ioctls: 1159 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1160 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1161 * 1162 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1163 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1164 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1165 * 1166 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1167 * Used to set network devices bus interface parameters. This interface 1168 * is retained for legacy reasons; new devices should use the bus 1169 * interface (PCI) for low level management. 1170 * 1171 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1172 * Called when a user wants to change the Maximum Transfer Unit 1173 * of a device. 1174 * 1175 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1176 * Callback used when the transmitter has not made any progress 1177 * for dev->watchdog ticks. 1178 * 1179 * void (*ndo_get_stats64)(struct net_device *dev, 1180 * struct rtnl_link_stats64 *storage); 1181 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1182 * Called when a user wants to get the network device usage 1183 * statistics. Drivers must do one of the following: 1184 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1185 * rtnl_link_stats64 structure passed by the caller. 1186 * 2. Define @ndo_get_stats to update a net_device_stats structure 1187 * (which should normally be dev->stats) and return a pointer to 1188 * it. The structure may be changed asynchronously only if each 1189 * field is written atomically. 1190 * 3. Update dev->stats asynchronously and atomically, and define 1191 * neither operation. 1192 * 1193 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1194 * Return true if this device supports offload stats of this attr_id. 1195 * 1196 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1197 * void *attr_data) 1198 * Get statistics for offload operations by attr_id. Write it into the 1199 * attr_data pointer. 1200 * 1201 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1202 * If device supports VLAN filtering this function is called when a 1203 * VLAN id is registered. 1204 * 1205 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1206 * If device supports VLAN filtering this function is called when a 1207 * VLAN id is unregistered. 1208 * 1209 * void (*ndo_poll_controller)(struct net_device *dev); 1210 * 1211 * SR-IOV management functions. 1212 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1213 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1214 * u8 qos, __be16 proto); 1215 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1216 * int max_tx_rate); 1217 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1218 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1219 * int (*ndo_get_vf_config)(struct net_device *dev, 1220 * int vf, struct ifla_vf_info *ivf); 1221 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1222 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1223 * struct nlattr *port[]); 1224 * 1225 * Enable or disable the VF ability to query its RSS Redirection Table and 1226 * Hash Key. This is needed since on some devices VF share this information 1227 * with PF and querying it may introduce a theoretical security risk. 1228 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1229 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1230 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1231 * void *type_data); 1232 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1233 * This is always called from the stack with the rtnl lock held and netif 1234 * tx queues stopped. This allows the netdevice to perform queue 1235 * management safely. 1236 * 1237 * Fiber Channel over Ethernet (FCoE) offload functions. 1238 * int (*ndo_fcoe_enable)(struct net_device *dev); 1239 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1240 * so the underlying device can perform whatever needed configuration or 1241 * initialization to support acceleration of FCoE traffic. 1242 * 1243 * int (*ndo_fcoe_disable)(struct net_device *dev); 1244 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1245 * so the underlying device can perform whatever needed clean-ups to 1246 * stop supporting acceleration of FCoE traffic. 1247 * 1248 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1249 * struct scatterlist *sgl, unsigned int sgc); 1250 * Called when the FCoE Initiator wants to initialize an I/O that 1251 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1252 * perform necessary setup and returns 1 to indicate the device is set up 1253 * successfully to perform DDP on this I/O, otherwise this returns 0. 1254 * 1255 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1256 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1257 * indicated by the FC exchange id 'xid', so the underlying device can 1258 * clean up and reuse resources for later DDP requests. 1259 * 1260 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1261 * struct scatterlist *sgl, unsigned int sgc); 1262 * Called when the FCoE Target wants to initialize an I/O that 1263 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1264 * perform necessary setup and returns 1 to indicate the device is set up 1265 * successfully to perform DDP on this I/O, otherwise this returns 0. 1266 * 1267 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1268 * struct netdev_fcoe_hbainfo *hbainfo); 1269 * Called when the FCoE Protocol stack wants information on the underlying 1270 * device. This information is utilized by the FCoE protocol stack to 1271 * register attributes with Fiber Channel management service as per the 1272 * FC-GS Fabric Device Management Information(FDMI) specification. 1273 * 1274 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1275 * Called when the underlying device wants to override default World Wide 1276 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1277 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1278 * protocol stack to use. 1279 * 1280 * RFS acceleration. 1281 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1282 * u16 rxq_index, u32 flow_id); 1283 * Set hardware filter for RFS. rxq_index is the target queue index; 1284 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1285 * Return the filter ID on success, or a negative error code. 1286 * 1287 * Slave management functions (for bridge, bonding, etc). 1288 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1289 * Called to make another netdev an underling. 1290 * 1291 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1292 * Called to release previously enslaved netdev. 1293 * 1294 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1295 * struct sk_buff *skb, 1296 * bool all_slaves); 1297 * Get the xmit slave of master device. If all_slaves is true, function 1298 * assume all the slaves can transmit. 1299 * 1300 * Feature/offload setting functions. 1301 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1302 * netdev_features_t features); 1303 * Adjusts the requested feature flags according to device-specific 1304 * constraints, and returns the resulting flags. Must not modify 1305 * the device state. 1306 * 1307 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1308 * Called to update device configuration to new features. Passed 1309 * feature set might be less than what was returned by ndo_fix_features()). 1310 * Must return >0 or -errno if it changed dev->features itself. 1311 * 1312 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1313 * struct net_device *dev, 1314 * const unsigned char *addr, u16 vid, u16 flags, 1315 * struct netlink_ext_ack *extack); 1316 * Adds an FDB entry to dev for addr. 1317 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1318 * struct net_device *dev, 1319 * const unsigned char *addr, u16 vid) 1320 * Deletes the FDB entry from dev coresponding to addr. 1321 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[], 1322 * struct net_device *dev, 1323 * u16 vid, 1324 * struct netlink_ext_ack *extack); 1325 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1326 * struct net_device *dev, struct net_device *filter_dev, 1327 * int *idx) 1328 * Used to add FDB entries to dump requests. Implementers should add 1329 * entries to skb and update idx with the number of entries. 1330 * 1331 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1332 * u16 flags, struct netlink_ext_ack *extack) 1333 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1334 * struct net_device *dev, u32 filter_mask, 1335 * int nlflags) 1336 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1337 * u16 flags); 1338 * 1339 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1340 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1341 * which do not represent real hardware may define this to allow their 1342 * userspace components to manage their virtual carrier state. Devices 1343 * that determine carrier state from physical hardware properties (eg 1344 * network cables) or protocol-dependent mechanisms (eg 1345 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1346 * 1347 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1348 * struct netdev_phys_item_id *ppid); 1349 * Called to get ID of physical port of this device. If driver does 1350 * not implement this, it is assumed that the hw is not able to have 1351 * multiple net devices on single physical port. 1352 * 1353 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1354 * struct netdev_phys_item_id *ppid) 1355 * Called to get the parent ID of the physical port of this device. 1356 * 1357 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1358 * struct net_device *dev) 1359 * Called by upper layer devices to accelerate switching or other 1360 * station functionality into hardware. 'pdev is the lowerdev 1361 * to use for the offload and 'dev' is the net device that will 1362 * back the offload. Returns a pointer to the private structure 1363 * the upper layer will maintain. 1364 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1365 * Called by upper layer device to delete the station created 1366 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1367 * the station and priv is the structure returned by the add 1368 * operation. 1369 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1370 * int queue_index, u32 maxrate); 1371 * Called when a user wants to set a max-rate limitation of specific 1372 * TX queue. 1373 * int (*ndo_get_iflink)(const struct net_device *dev); 1374 * Called to get the iflink value of this device. 1375 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1376 * This function is used to get egress tunnel information for given skb. 1377 * This is useful for retrieving outer tunnel header parameters while 1378 * sampling packet. 1379 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1380 * This function is used to specify the headroom that the skb must 1381 * consider when allocation skb during packet reception. Setting 1382 * appropriate rx headroom value allows avoiding skb head copy on 1383 * forward. Setting a negative value resets the rx headroom to the 1384 * default value. 1385 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1386 * This function is used to set or query state related to XDP on the 1387 * netdevice and manage BPF offload. See definition of 1388 * enum bpf_netdev_command for details. 1389 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1390 * u32 flags); 1391 * This function is used to submit @n XDP packets for transmit on a 1392 * netdevice. Returns number of frames successfully transmitted, frames 1393 * that got dropped are freed/returned via xdp_return_frame(). 1394 * Returns negative number, means general error invoking ndo, meaning 1395 * no frames were xmit'ed and core-caller will free all frames. 1396 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1397 * struct xdp_buff *xdp); 1398 * Get the xmit slave of master device based on the xdp_buff. 1399 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1400 * This function is used to wake up the softirq, ksoftirqd or kthread 1401 * responsible for sending and/or receiving packets on a specific 1402 * queue id bound to an AF_XDP socket. The flags field specifies if 1403 * only RX, only Tx, or both should be woken up using the flags 1404 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1405 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1406 * Get devlink port instance associated with a given netdev. 1407 * Called with a reference on the netdevice and devlink locks only, 1408 * rtnl_lock is not held. 1409 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1410 * int cmd); 1411 * Add, change, delete or get information on an IPv4 tunnel. 1412 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1413 * If a device is paired with a peer device, return the peer instance. 1414 * The caller must be under RCU read context. 1415 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1416 * Get the forwarding path to reach the real device from the HW destination address 1417 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1418 * const struct skb_shared_hwtstamps *hwtstamps, 1419 * bool cycles); 1420 * Get hardware timestamp based on normal/adjustable time or free running 1421 * cycle counter. This function is required if physical clock supports a 1422 * free running cycle counter. 1423 */ 1424 struct net_device_ops { 1425 int (*ndo_init)(struct net_device *dev); 1426 void (*ndo_uninit)(struct net_device *dev); 1427 int (*ndo_open)(struct net_device *dev); 1428 int (*ndo_stop)(struct net_device *dev); 1429 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1430 struct net_device *dev); 1431 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1432 struct net_device *dev, 1433 netdev_features_t features); 1434 u16 (*ndo_select_queue)(struct net_device *dev, 1435 struct sk_buff *skb, 1436 struct net_device *sb_dev); 1437 void (*ndo_change_rx_flags)(struct net_device *dev, 1438 int flags); 1439 void (*ndo_set_rx_mode)(struct net_device *dev); 1440 int (*ndo_set_mac_address)(struct net_device *dev, 1441 void *addr); 1442 int (*ndo_validate_addr)(struct net_device *dev); 1443 int (*ndo_do_ioctl)(struct net_device *dev, 1444 struct ifreq *ifr, int cmd); 1445 int (*ndo_eth_ioctl)(struct net_device *dev, 1446 struct ifreq *ifr, int cmd); 1447 int (*ndo_siocbond)(struct net_device *dev, 1448 struct ifreq *ifr, int cmd); 1449 int (*ndo_siocwandev)(struct net_device *dev, 1450 struct if_settings *ifs); 1451 int (*ndo_siocdevprivate)(struct net_device *dev, 1452 struct ifreq *ifr, 1453 void __user *data, int cmd); 1454 int (*ndo_set_config)(struct net_device *dev, 1455 struct ifmap *map); 1456 int (*ndo_change_mtu)(struct net_device *dev, 1457 int new_mtu); 1458 int (*ndo_neigh_setup)(struct net_device *dev, 1459 struct neigh_parms *); 1460 void (*ndo_tx_timeout) (struct net_device *dev, 1461 unsigned int txqueue); 1462 1463 void (*ndo_get_stats64)(struct net_device *dev, 1464 struct rtnl_link_stats64 *storage); 1465 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1466 int (*ndo_get_offload_stats)(int attr_id, 1467 const struct net_device *dev, 1468 void *attr_data); 1469 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1470 1471 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1472 __be16 proto, u16 vid); 1473 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1474 __be16 proto, u16 vid); 1475 #ifdef CONFIG_NET_POLL_CONTROLLER 1476 void (*ndo_poll_controller)(struct net_device *dev); 1477 int (*ndo_netpoll_setup)(struct net_device *dev, 1478 struct netpoll_info *info); 1479 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1480 #endif 1481 int (*ndo_set_vf_mac)(struct net_device *dev, 1482 int queue, u8 *mac); 1483 int (*ndo_set_vf_vlan)(struct net_device *dev, 1484 int queue, u16 vlan, 1485 u8 qos, __be16 proto); 1486 int (*ndo_set_vf_rate)(struct net_device *dev, 1487 int vf, int min_tx_rate, 1488 int max_tx_rate); 1489 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1490 int vf, bool setting); 1491 int (*ndo_set_vf_trust)(struct net_device *dev, 1492 int vf, bool setting); 1493 int (*ndo_get_vf_config)(struct net_device *dev, 1494 int vf, 1495 struct ifla_vf_info *ivf); 1496 int (*ndo_set_vf_link_state)(struct net_device *dev, 1497 int vf, int link_state); 1498 int (*ndo_get_vf_stats)(struct net_device *dev, 1499 int vf, 1500 struct ifla_vf_stats 1501 *vf_stats); 1502 int (*ndo_set_vf_port)(struct net_device *dev, 1503 int vf, 1504 struct nlattr *port[]); 1505 int (*ndo_get_vf_port)(struct net_device *dev, 1506 int vf, struct sk_buff *skb); 1507 int (*ndo_get_vf_guid)(struct net_device *dev, 1508 int vf, 1509 struct ifla_vf_guid *node_guid, 1510 struct ifla_vf_guid *port_guid); 1511 int (*ndo_set_vf_guid)(struct net_device *dev, 1512 int vf, u64 guid, 1513 int guid_type); 1514 int (*ndo_set_vf_rss_query_en)( 1515 struct net_device *dev, 1516 int vf, bool setting); 1517 int (*ndo_setup_tc)(struct net_device *dev, 1518 enum tc_setup_type type, 1519 void *type_data); 1520 #if IS_ENABLED(CONFIG_FCOE) 1521 int (*ndo_fcoe_enable)(struct net_device *dev); 1522 int (*ndo_fcoe_disable)(struct net_device *dev); 1523 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1524 u16 xid, 1525 struct scatterlist *sgl, 1526 unsigned int sgc); 1527 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1528 u16 xid); 1529 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1530 u16 xid, 1531 struct scatterlist *sgl, 1532 unsigned int sgc); 1533 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1534 struct netdev_fcoe_hbainfo *hbainfo); 1535 #endif 1536 1537 #if IS_ENABLED(CONFIG_LIBFCOE) 1538 #define NETDEV_FCOE_WWNN 0 1539 #define NETDEV_FCOE_WWPN 1 1540 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1541 u64 *wwn, int type); 1542 #endif 1543 1544 #ifdef CONFIG_RFS_ACCEL 1545 int (*ndo_rx_flow_steer)(struct net_device *dev, 1546 const struct sk_buff *skb, 1547 u16 rxq_index, 1548 u32 flow_id); 1549 #endif 1550 int (*ndo_add_slave)(struct net_device *dev, 1551 struct net_device *slave_dev, 1552 struct netlink_ext_ack *extack); 1553 int (*ndo_del_slave)(struct net_device *dev, 1554 struct net_device *slave_dev); 1555 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1556 struct sk_buff *skb, 1557 bool all_slaves); 1558 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1559 struct sock *sk); 1560 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1561 netdev_features_t features); 1562 int (*ndo_set_features)(struct net_device *dev, 1563 netdev_features_t features); 1564 int (*ndo_neigh_construct)(struct net_device *dev, 1565 struct neighbour *n); 1566 void (*ndo_neigh_destroy)(struct net_device *dev, 1567 struct neighbour *n); 1568 1569 int (*ndo_fdb_add)(struct ndmsg *ndm, 1570 struct nlattr *tb[], 1571 struct net_device *dev, 1572 const unsigned char *addr, 1573 u16 vid, 1574 u16 flags, 1575 struct netlink_ext_ack *extack); 1576 int (*ndo_fdb_del)(struct ndmsg *ndm, 1577 struct nlattr *tb[], 1578 struct net_device *dev, 1579 const unsigned char *addr, 1580 u16 vid, struct netlink_ext_ack *extack); 1581 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, 1582 struct nlattr *tb[], 1583 struct net_device *dev, 1584 u16 vid, 1585 struct netlink_ext_ack *extack); 1586 int (*ndo_fdb_dump)(struct sk_buff *skb, 1587 struct netlink_callback *cb, 1588 struct net_device *dev, 1589 struct net_device *filter_dev, 1590 int *idx); 1591 int (*ndo_fdb_get)(struct sk_buff *skb, 1592 struct nlattr *tb[], 1593 struct net_device *dev, 1594 const unsigned char *addr, 1595 u16 vid, u32 portid, u32 seq, 1596 struct netlink_ext_ack *extack); 1597 int (*ndo_bridge_setlink)(struct net_device *dev, 1598 struct nlmsghdr *nlh, 1599 u16 flags, 1600 struct netlink_ext_ack *extack); 1601 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1602 u32 pid, u32 seq, 1603 struct net_device *dev, 1604 u32 filter_mask, 1605 int nlflags); 1606 int (*ndo_bridge_dellink)(struct net_device *dev, 1607 struct nlmsghdr *nlh, 1608 u16 flags); 1609 int (*ndo_change_carrier)(struct net_device *dev, 1610 bool new_carrier); 1611 int (*ndo_get_phys_port_id)(struct net_device *dev, 1612 struct netdev_phys_item_id *ppid); 1613 int (*ndo_get_port_parent_id)(struct net_device *dev, 1614 struct netdev_phys_item_id *ppid); 1615 int (*ndo_get_phys_port_name)(struct net_device *dev, 1616 char *name, size_t len); 1617 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1618 struct net_device *dev); 1619 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1620 void *priv); 1621 1622 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1623 int queue_index, 1624 u32 maxrate); 1625 int (*ndo_get_iflink)(const struct net_device *dev); 1626 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1627 struct sk_buff *skb); 1628 void (*ndo_set_rx_headroom)(struct net_device *dev, 1629 int needed_headroom); 1630 int (*ndo_bpf)(struct net_device *dev, 1631 struct netdev_bpf *bpf); 1632 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1633 struct xdp_frame **xdp, 1634 u32 flags); 1635 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1636 struct xdp_buff *xdp); 1637 int (*ndo_xsk_wakeup)(struct net_device *dev, 1638 u32 queue_id, u32 flags); 1639 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1640 int (*ndo_tunnel_ctl)(struct net_device *dev, 1641 struct ip_tunnel_parm *p, int cmd); 1642 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1643 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1644 struct net_device_path *path); 1645 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1646 const struct skb_shared_hwtstamps *hwtstamps, 1647 bool cycles); 1648 1649 ANDROID_KABI_RESERVE(1); 1650 ANDROID_KABI_RESERVE(2); 1651 ANDROID_KABI_RESERVE(3); 1652 ANDROID_KABI_RESERVE(4); 1653 ANDROID_KABI_RESERVE(5); 1654 ANDROID_KABI_RESERVE(6); 1655 ANDROID_KABI_RESERVE(7); 1656 ANDROID_KABI_RESERVE(8); 1657 }; 1658 1659 /** 1660 * enum netdev_priv_flags - &struct net_device priv_flags 1661 * 1662 * These are the &struct net_device, they are only set internally 1663 * by drivers and used in the kernel. These flags are invisible to 1664 * userspace; this means that the order of these flags can change 1665 * during any kernel release. 1666 * 1667 * You should have a pretty good reason to be extending these flags. 1668 * 1669 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1670 * @IFF_EBRIDGE: Ethernet bridging device 1671 * @IFF_BONDING: bonding master or slave 1672 * @IFF_ISATAP: ISATAP interface (RFC4214) 1673 * @IFF_WAN_HDLC: WAN HDLC device 1674 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1675 * release skb->dst 1676 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1677 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1678 * @IFF_MACVLAN_PORT: device used as macvlan port 1679 * @IFF_BRIDGE_PORT: device used as bridge port 1680 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1681 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1682 * @IFF_UNICAST_FLT: Supports unicast filtering 1683 * @IFF_TEAM_PORT: device used as team port 1684 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1685 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1686 * change when it's running 1687 * @IFF_MACVLAN: Macvlan device 1688 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1689 * underlying stacked devices 1690 * @IFF_L3MDEV_MASTER: device is an L3 master device 1691 * @IFF_NO_QUEUE: device can run without qdisc attached 1692 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1693 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1694 * @IFF_TEAM: device is a team device 1695 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1696 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1697 * entity (i.e. the master device for bridged veth) 1698 * @IFF_MACSEC: device is a MACsec device 1699 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1700 * @IFF_FAILOVER: device is a failover master device 1701 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1702 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1703 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1704 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1705 * skb_headlen(skb) == 0 (data starts from frag0) 1706 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1707 */ 1708 enum netdev_priv_flags { 1709 IFF_802_1Q_VLAN = 1<<0, 1710 IFF_EBRIDGE = 1<<1, 1711 IFF_BONDING = 1<<2, 1712 IFF_ISATAP = 1<<3, 1713 IFF_WAN_HDLC = 1<<4, 1714 IFF_XMIT_DST_RELEASE = 1<<5, 1715 IFF_DONT_BRIDGE = 1<<6, 1716 IFF_DISABLE_NETPOLL = 1<<7, 1717 IFF_MACVLAN_PORT = 1<<8, 1718 IFF_BRIDGE_PORT = 1<<9, 1719 IFF_OVS_DATAPATH = 1<<10, 1720 IFF_TX_SKB_SHARING = 1<<11, 1721 IFF_UNICAST_FLT = 1<<12, 1722 IFF_TEAM_PORT = 1<<13, 1723 IFF_SUPP_NOFCS = 1<<14, 1724 IFF_LIVE_ADDR_CHANGE = 1<<15, 1725 IFF_MACVLAN = 1<<16, 1726 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1727 IFF_L3MDEV_MASTER = 1<<18, 1728 IFF_NO_QUEUE = 1<<19, 1729 IFF_OPENVSWITCH = 1<<20, 1730 IFF_L3MDEV_SLAVE = 1<<21, 1731 IFF_TEAM = 1<<22, 1732 IFF_RXFH_CONFIGURED = 1<<23, 1733 IFF_PHONY_HEADROOM = 1<<24, 1734 IFF_MACSEC = 1<<25, 1735 IFF_NO_RX_HANDLER = 1<<26, 1736 IFF_FAILOVER = 1<<27, 1737 IFF_FAILOVER_SLAVE = 1<<28, 1738 IFF_L3MDEV_RX_HANDLER = 1<<29, 1739 IFF_LIVE_RENAME_OK = 1<<30, 1740 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1741 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1742 }; 1743 1744 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1745 #define IFF_EBRIDGE IFF_EBRIDGE 1746 #define IFF_BONDING IFF_BONDING 1747 #define IFF_ISATAP IFF_ISATAP 1748 #define IFF_WAN_HDLC IFF_WAN_HDLC 1749 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1750 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1751 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1752 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1753 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1754 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1755 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1756 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1757 #define IFF_TEAM_PORT IFF_TEAM_PORT 1758 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1759 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1760 #define IFF_MACVLAN IFF_MACVLAN 1761 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1762 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1763 #define IFF_NO_QUEUE IFF_NO_QUEUE 1764 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1765 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1766 #define IFF_TEAM IFF_TEAM 1767 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1768 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1769 #define IFF_MACSEC IFF_MACSEC 1770 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1771 #define IFF_FAILOVER IFF_FAILOVER 1772 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1773 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1774 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1775 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1776 1777 /* Specifies the type of the struct net_device::ml_priv pointer */ 1778 enum netdev_ml_priv_type { 1779 ML_PRIV_NONE, 1780 ML_PRIV_CAN, 1781 }; 1782 1783 /** 1784 * struct net_device - The DEVICE structure. 1785 * 1786 * Actually, this whole structure is a big mistake. It mixes I/O 1787 * data with strictly "high-level" data, and it has to know about 1788 * almost every data structure used in the INET module. 1789 * 1790 * @name: This is the first field of the "visible" part of this structure 1791 * (i.e. as seen by users in the "Space.c" file). It is the name 1792 * of the interface. 1793 * 1794 * @name_node: Name hashlist node 1795 * @ifalias: SNMP alias 1796 * @mem_end: Shared memory end 1797 * @mem_start: Shared memory start 1798 * @base_addr: Device I/O address 1799 * @irq: Device IRQ number 1800 * 1801 * @state: Generic network queuing layer state, see netdev_state_t 1802 * @dev_list: The global list of network devices 1803 * @napi_list: List entry used for polling NAPI devices 1804 * @unreg_list: List entry when we are unregistering the 1805 * device; see the function unregister_netdev 1806 * @close_list: List entry used when we are closing the device 1807 * @ptype_all: Device-specific packet handlers for all protocols 1808 * @ptype_specific: Device-specific, protocol-specific packet handlers 1809 * 1810 * @adj_list: Directly linked devices, like slaves for bonding 1811 * @features: Currently active device features 1812 * @hw_features: User-changeable features 1813 * 1814 * @wanted_features: User-requested features 1815 * @vlan_features: Mask of features inheritable by VLAN devices 1816 * 1817 * @hw_enc_features: Mask of features inherited by encapsulating devices 1818 * This field indicates what encapsulation 1819 * offloads the hardware is capable of doing, 1820 * and drivers will need to set them appropriately. 1821 * 1822 * @mpls_features: Mask of features inheritable by MPLS 1823 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1824 * 1825 * @ifindex: interface index 1826 * @group: The group the device belongs to 1827 * 1828 * @stats: Statistics struct, which was left as a legacy, use 1829 * rtnl_link_stats64 instead 1830 * 1831 * @core_stats: core networking counters, 1832 * do not use this in drivers 1833 * @carrier_up_count: Number of times the carrier has been up 1834 * @carrier_down_count: Number of times the carrier has been down 1835 * 1836 * @wireless_handlers: List of functions to handle Wireless Extensions, 1837 * instead of ioctl, 1838 * see <net/iw_handler.h> for details. 1839 * @wireless_data: Instance data managed by the core of wireless extensions 1840 * 1841 * @netdev_ops: Includes several pointers to callbacks, 1842 * if one wants to override the ndo_*() functions 1843 * @ethtool_ops: Management operations 1844 * @l3mdev_ops: Layer 3 master device operations 1845 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1846 * discovery handling. Necessary for e.g. 6LoWPAN. 1847 * @xfrmdev_ops: Transformation offload operations 1848 * @tlsdev_ops: Transport Layer Security offload operations 1849 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1850 * of Layer 2 headers. 1851 * 1852 * @flags: Interface flags (a la BSD) 1853 * @priv_flags: Like 'flags' but invisible to userspace, 1854 * see if.h for the definitions 1855 * @gflags: Global flags ( kept as legacy ) 1856 * @padded: How much padding added by alloc_netdev() 1857 * @operstate: RFC2863 operstate 1858 * @link_mode: Mapping policy to operstate 1859 * @if_port: Selectable AUI, TP, ... 1860 * @dma: DMA channel 1861 * @mtu: Interface MTU value 1862 * @min_mtu: Interface Minimum MTU value 1863 * @max_mtu: Interface Maximum MTU value 1864 * @type: Interface hardware type 1865 * @hard_header_len: Maximum hardware header length. 1866 * @min_header_len: Minimum hardware header length 1867 * 1868 * @needed_headroom: Extra headroom the hardware may need, but not in all 1869 * cases can this be guaranteed 1870 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1871 * cases can this be guaranteed. Some cases also use 1872 * LL_MAX_HEADER instead to allocate the skb 1873 * 1874 * interface address info: 1875 * 1876 * @perm_addr: Permanent hw address 1877 * @addr_assign_type: Hw address assignment type 1878 * @addr_len: Hardware address length 1879 * @upper_level: Maximum depth level of upper devices. 1880 * @lower_level: Maximum depth level of lower devices. 1881 * @neigh_priv_len: Used in neigh_alloc() 1882 * @dev_id: Used to differentiate devices that share 1883 * the same link layer address 1884 * @dev_port: Used to differentiate devices that share 1885 * the same function 1886 * @addr_list_lock: XXX: need comments on this one 1887 * @name_assign_type: network interface name assignment type 1888 * @uc_promisc: Counter that indicates promiscuous mode 1889 * has been enabled due to the need to listen to 1890 * additional unicast addresses in a device that 1891 * does not implement ndo_set_rx_mode() 1892 * @uc: unicast mac addresses 1893 * @mc: multicast mac addresses 1894 * @dev_addrs: list of device hw addresses 1895 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1896 * @promiscuity: Number of times the NIC is told to work in 1897 * promiscuous mode; if it becomes 0 the NIC will 1898 * exit promiscuous mode 1899 * @allmulti: Counter, enables or disables allmulticast mode 1900 * 1901 * @vlan_info: VLAN info 1902 * @dsa_ptr: dsa specific data 1903 * @tipc_ptr: TIPC specific data 1904 * @atalk_ptr: AppleTalk link 1905 * @ip_ptr: IPv4 specific data 1906 * @ip6_ptr: IPv6 specific data 1907 * @ax25_ptr: AX.25 specific data 1908 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1909 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1910 * device struct 1911 * @mpls_ptr: mpls_dev struct pointer 1912 * @mctp_ptr: MCTP specific data 1913 * 1914 * @dev_addr: Hw address (before bcast, 1915 * because most packets are unicast) 1916 * 1917 * @_rx: Array of RX queues 1918 * @num_rx_queues: Number of RX queues 1919 * allocated at register_netdev() time 1920 * @real_num_rx_queues: Number of RX queues currently active in device 1921 * @xdp_prog: XDP sockets filter program pointer 1922 * @gro_flush_timeout: timeout for GRO layer in NAPI 1923 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1924 * allow to avoid NIC hard IRQ, on busy queues. 1925 * 1926 * @rx_handler: handler for received packets 1927 * @rx_handler_data: XXX: need comments on this one 1928 * @miniq_ingress: ingress/clsact qdisc specific data for 1929 * ingress processing 1930 * @ingress_queue: XXX: need comments on this one 1931 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1932 * @broadcast: hw bcast address 1933 * 1934 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1935 * indexed by RX queue number. Assigned by driver. 1936 * This must only be set if the ndo_rx_flow_steer 1937 * operation is defined 1938 * @index_hlist: Device index hash chain 1939 * 1940 * @_tx: Array of TX queues 1941 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1942 * @real_num_tx_queues: Number of TX queues currently active in device 1943 * @qdisc: Root qdisc from userspace point of view 1944 * @tx_queue_len: Max frames per queue allowed 1945 * @tx_global_lock: XXX: need comments on this one 1946 * @xdp_bulkq: XDP device bulk queue 1947 * @xps_maps: all CPUs/RXQs maps for XPS device 1948 * 1949 * @xps_maps: XXX: need comments on this one 1950 * @miniq_egress: clsact qdisc specific data for 1951 * egress processing 1952 * @nf_hooks_egress: netfilter hooks executed for egress packets 1953 * @qdisc_hash: qdisc hash table 1954 * @watchdog_timeo: Represents the timeout that is used by 1955 * the watchdog (see dev_watchdog()) 1956 * @watchdog_timer: List of timers 1957 * 1958 * @proto_down_reason: reason a netdev interface is held down 1959 * @pcpu_refcnt: Number of references to this device 1960 * @dev_refcnt: Number of references to this device 1961 * @refcnt_tracker: Tracker directory for tracked references to this device 1962 * @todo_list: Delayed register/unregister 1963 * @link_watch_list: XXX: need comments on this one 1964 * 1965 * @reg_state: Register/unregister state machine 1966 * @dismantle: Device is going to be freed 1967 * @rtnl_link_state: This enum represents the phases of creating 1968 * a new link 1969 * 1970 * @needs_free_netdev: Should unregister perform free_netdev? 1971 * @priv_destructor: Called from unregister 1972 * @npinfo: XXX: need comments on this one 1973 * @nd_net: Network namespace this network device is inside 1974 * 1975 * @ml_priv: Mid-layer private 1976 * @ml_priv_type: Mid-layer private type 1977 * @lstats: Loopback statistics 1978 * @tstats: Tunnel statistics 1979 * @dstats: Dummy statistics 1980 * @vstats: Virtual ethernet statistics 1981 * 1982 * @garp_port: GARP 1983 * @mrp_port: MRP 1984 * 1985 * @dm_private: Drop monitor private 1986 * 1987 * @dev: Class/net/name entry 1988 * @sysfs_groups: Space for optional device, statistics and wireless 1989 * sysfs groups 1990 * 1991 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1992 * @rtnl_link_ops: Rtnl_link_ops 1993 * 1994 * @gso_max_size: Maximum size of generic segmentation offload 1995 * @tso_max_size: Device (as in HW) limit on the max TSO request size 1996 * @gso_max_segs: Maximum number of segments that can be passed to the 1997 * NIC for GSO 1998 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 1999 * 2000 * @dcbnl_ops: Data Center Bridging netlink ops 2001 * @num_tc: Number of traffic classes in the net device 2002 * @tc_to_txq: XXX: need comments on this one 2003 * @prio_tc_map: XXX: need comments on this one 2004 * 2005 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2006 * 2007 * @priomap: XXX: need comments on this one 2008 * @phydev: Physical device may attach itself 2009 * for hardware timestamping 2010 * @sfp_bus: attached &struct sfp_bus structure. 2011 * 2012 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2013 * 2014 * @proto_down: protocol port state information can be sent to the 2015 * switch driver and used to set the phys state of the 2016 * switch port. 2017 * 2018 * @wol_enabled: Wake-on-LAN is enabled 2019 * 2020 * @threaded: napi threaded mode is enabled 2021 * 2022 * @net_notifier_list: List of per-net netdev notifier block 2023 * that follow this device when it is moved 2024 * to another network namespace. 2025 * 2026 * @macsec_ops: MACsec offloading ops 2027 * 2028 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2029 * offload capabilities of the device 2030 * @udp_tunnel_nic: UDP tunnel offload state 2031 * @xdp_state: stores info on attached XDP BPF programs 2032 * 2033 * @nested_level: Used as a parameter of spin_lock_nested() of 2034 * dev->addr_list_lock. 2035 * @unlink_list: As netif_addr_lock() can be called recursively, 2036 * keep a list of interfaces to be deleted. 2037 * @gro_max_size: Maximum size of aggregated packet in generic 2038 * receive offload (GRO) 2039 * 2040 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2041 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2042 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2043 * @dev_registered_tracker: tracker for reference held while 2044 * registered 2045 * @offload_xstats_l3: L3 HW stats for this netdevice. 2046 * 2047 * FIXME: cleanup struct net_device such that network protocol info 2048 * moves out. 2049 */ 2050 2051 struct net_device { 2052 char name[IFNAMSIZ]; 2053 struct netdev_name_node *name_node; 2054 struct dev_ifalias __rcu *ifalias; 2055 /* 2056 * I/O specific fields 2057 * FIXME: Merge these and struct ifmap into one 2058 */ 2059 unsigned long mem_end; 2060 unsigned long mem_start; 2061 unsigned long base_addr; 2062 2063 /* 2064 * Some hardware also needs these fields (state,dev_list, 2065 * napi_list,unreg_list,close_list) but they are not 2066 * part of the usual set specified in Space.c. 2067 */ 2068 2069 unsigned long state; 2070 2071 struct list_head dev_list; 2072 struct list_head napi_list; 2073 struct list_head unreg_list; 2074 struct list_head close_list; 2075 struct list_head ptype_all; 2076 struct list_head ptype_specific; 2077 2078 struct { 2079 struct list_head upper; 2080 struct list_head lower; 2081 } adj_list; 2082 2083 /* Read-mostly cache-line for fast-path access */ 2084 unsigned int flags; 2085 unsigned long long priv_flags; 2086 const struct net_device_ops *netdev_ops; 2087 int ifindex; 2088 unsigned short gflags; 2089 unsigned short hard_header_len; 2090 2091 /* Note : dev->mtu is often read without holding a lock. 2092 * Writers usually hold RTNL. 2093 * It is recommended to use READ_ONCE() to annotate the reads, 2094 * and to use WRITE_ONCE() to annotate the writes. 2095 */ 2096 unsigned int mtu; 2097 unsigned short needed_headroom; 2098 unsigned short needed_tailroom; 2099 2100 netdev_features_t features; 2101 netdev_features_t hw_features; 2102 netdev_features_t wanted_features; 2103 netdev_features_t vlan_features; 2104 netdev_features_t hw_enc_features; 2105 netdev_features_t mpls_features; 2106 netdev_features_t gso_partial_features; 2107 2108 unsigned int min_mtu; 2109 unsigned int max_mtu; 2110 unsigned short type; 2111 unsigned char min_header_len; 2112 unsigned char name_assign_type; 2113 2114 int group; 2115 2116 struct net_device_stats stats; /* not used by modern drivers */ 2117 2118 struct net_device_core_stats __percpu *core_stats; 2119 2120 /* Stats to monitor link on/off, flapping */ 2121 atomic_t carrier_up_count; 2122 atomic_t carrier_down_count; 2123 2124 /* Android KMI hack to allow vendors to have their own wifi changes in modules */ 2125 #ifdef __GENKSYMS__ 2126 void *wireless_handlers; 2127 void *wireless_data; 2128 #else 2129 const struct iw_handler_def *wireless_handlers; 2130 struct iw_public_data *wireless_data; 2131 #endif 2132 const struct ethtool_ops *ethtool_ops; 2133 #ifdef CONFIG_NET_L3_MASTER_DEV 2134 const struct l3mdev_ops *l3mdev_ops; 2135 #endif 2136 #if IS_ENABLED(CONFIG_IPV6) 2137 const struct ndisc_ops *ndisc_ops; 2138 #endif 2139 2140 #ifdef CONFIG_XFRM_OFFLOAD 2141 const struct xfrmdev_ops *xfrmdev_ops; 2142 #endif 2143 2144 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2145 const struct tlsdev_ops *tlsdev_ops; 2146 #endif 2147 2148 const struct header_ops *header_ops; 2149 2150 unsigned char operstate; 2151 unsigned char link_mode; 2152 2153 unsigned char if_port; 2154 unsigned char dma; 2155 2156 /* Interface address info. */ 2157 unsigned char perm_addr[MAX_ADDR_LEN]; 2158 unsigned char addr_assign_type; 2159 unsigned char addr_len; 2160 unsigned char upper_level; 2161 unsigned char lower_level; 2162 2163 unsigned short neigh_priv_len; 2164 unsigned short dev_id; 2165 unsigned short dev_port; 2166 unsigned short padded; 2167 2168 spinlock_t addr_list_lock; 2169 int irq; 2170 2171 struct netdev_hw_addr_list uc; 2172 struct netdev_hw_addr_list mc; 2173 struct netdev_hw_addr_list dev_addrs; 2174 2175 #ifdef CONFIG_SYSFS 2176 struct kset *queues_kset; 2177 #endif 2178 #ifdef CONFIG_LOCKDEP 2179 struct list_head unlink_list; 2180 #endif 2181 unsigned int promiscuity; 2182 unsigned int allmulti; 2183 bool uc_promisc; 2184 #ifdef CONFIG_LOCKDEP 2185 unsigned char nested_level; 2186 #endif 2187 2188 2189 /* Protocol-specific pointers */ 2190 2191 struct in_device __rcu *ip_ptr; 2192 struct inet6_dev __rcu *ip6_ptr; 2193 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2194 struct vlan_info __rcu *vlan_info; 2195 #endif 2196 #if IS_ENABLED(CONFIG_NET_DSA) 2197 struct dsa_port *dsa_ptr; 2198 #endif 2199 #if IS_ENABLED(CONFIG_TIPC) 2200 struct tipc_bearer __rcu *tipc_ptr; 2201 #endif 2202 #if IS_ENABLED(CONFIG_ATALK) 2203 void *atalk_ptr; 2204 #endif 2205 #if IS_ENABLED(CONFIG_AX25) 2206 void *ax25_ptr; 2207 #endif 2208 /* Android KMI hack to allow vendors to have their own wifi changes in modules */ 2209 #ifdef __GENKSYMS__ 2210 void *ieee80211_ptr; 2211 #else 2212 struct wireless_dev *ieee80211_ptr; 2213 #endif 2214 2215 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2216 struct wpan_dev *ieee802154_ptr; 2217 #endif 2218 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2219 struct mpls_dev __rcu *mpls_ptr; 2220 #endif 2221 #if IS_ENABLED(CONFIG_MCTP) 2222 struct mctp_dev __rcu *mctp_ptr; 2223 #endif 2224 2225 /* 2226 * Cache lines mostly used on receive path (including eth_type_trans()) 2227 */ 2228 /* Interface address info used in eth_type_trans() */ 2229 const unsigned char *dev_addr; 2230 2231 struct netdev_rx_queue *_rx; 2232 unsigned int num_rx_queues; 2233 unsigned int real_num_rx_queues; 2234 2235 struct bpf_prog __rcu *xdp_prog; 2236 unsigned long gro_flush_timeout; 2237 int napi_defer_hard_irqs; 2238 #define GRO_LEGACY_MAX_SIZE 65536u 2239 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2240 * and shinfo->gso_segs is a 16bit field. 2241 */ 2242 #define GRO_MAX_SIZE (8 * 65535u) 2243 unsigned int gro_max_size; 2244 rx_handler_func_t __rcu *rx_handler; 2245 void __rcu *rx_handler_data; 2246 2247 #ifdef CONFIG_NET_CLS_ACT 2248 struct mini_Qdisc __rcu *miniq_ingress; 2249 #endif 2250 struct netdev_queue __rcu *ingress_queue; 2251 #ifdef CONFIG_NETFILTER_INGRESS 2252 struct nf_hook_entries __rcu *nf_hooks_ingress; 2253 #endif 2254 2255 unsigned char broadcast[MAX_ADDR_LEN]; 2256 #ifdef CONFIG_RFS_ACCEL 2257 struct cpu_rmap *rx_cpu_rmap; 2258 #endif 2259 struct hlist_node index_hlist; 2260 2261 /* 2262 * Cache lines mostly used on transmit path 2263 */ 2264 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2265 unsigned int num_tx_queues; 2266 unsigned int real_num_tx_queues; 2267 struct Qdisc __rcu *qdisc; 2268 unsigned int tx_queue_len; 2269 spinlock_t tx_global_lock; 2270 2271 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2272 2273 #ifdef CONFIG_XPS 2274 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2275 #endif 2276 #ifdef CONFIG_NET_CLS_ACT 2277 struct mini_Qdisc __rcu *miniq_egress; 2278 #endif 2279 #ifdef CONFIG_NETFILTER_EGRESS 2280 struct nf_hook_entries __rcu *nf_hooks_egress; 2281 #endif 2282 2283 #ifdef CONFIG_NET_SCHED 2284 DECLARE_HASHTABLE (qdisc_hash, 4); 2285 #endif 2286 /* These may be needed for future network-power-down code. */ 2287 struct timer_list watchdog_timer; 2288 int watchdog_timeo; 2289 2290 u32 proto_down_reason; 2291 2292 struct list_head todo_list; 2293 2294 #ifdef CONFIG_PCPU_DEV_REFCNT 2295 int __percpu *pcpu_refcnt; 2296 #else 2297 refcount_t dev_refcnt; 2298 #endif 2299 struct ref_tracker_dir refcnt_tracker; 2300 2301 struct list_head link_watch_list; 2302 2303 enum { NETREG_UNINITIALIZED=0, 2304 NETREG_REGISTERED, /* completed register_netdevice */ 2305 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2306 NETREG_UNREGISTERED, /* completed unregister todo */ 2307 NETREG_RELEASED, /* called free_netdev */ 2308 NETREG_DUMMY, /* dummy device for NAPI poll */ 2309 } reg_state:8; 2310 2311 bool dismantle; 2312 2313 enum { 2314 RTNL_LINK_INITIALIZED, 2315 RTNL_LINK_INITIALIZING, 2316 } rtnl_link_state:16; 2317 2318 bool needs_free_netdev; 2319 void (*priv_destructor)(struct net_device *dev); 2320 2321 #ifdef CONFIG_NETPOLL 2322 struct netpoll_info __rcu *npinfo; 2323 #endif 2324 2325 possible_net_t nd_net; 2326 2327 /* mid-layer private */ 2328 void *ml_priv; 2329 enum netdev_ml_priv_type ml_priv_type; 2330 2331 union { 2332 struct pcpu_lstats __percpu *lstats; 2333 struct pcpu_sw_netstats __percpu *tstats; 2334 struct pcpu_dstats __percpu *dstats; 2335 }; 2336 2337 #if IS_ENABLED(CONFIG_GARP) 2338 struct garp_port __rcu *garp_port; 2339 #endif 2340 #if IS_ENABLED(CONFIG_MRP) 2341 struct mrp_port __rcu *mrp_port; 2342 #endif 2343 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2344 struct dm_hw_stat_delta __rcu *dm_private; 2345 #endif 2346 struct device dev; 2347 const struct attribute_group *sysfs_groups[4]; 2348 const struct attribute_group *sysfs_rx_queue_group; 2349 2350 const struct rtnl_link_ops *rtnl_link_ops; 2351 2352 /* for setting kernel sock attribute on TCP connection setup */ 2353 #define GSO_MAX_SEGS 65535u 2354 #define GSO_LEGACY_MAX_SIZE 65536u 2355 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2356 * and shinfo->gso_segs is a 16bit field. 2357 */ 2358 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2359 2360 unsigned int gso_max_size; 2361 #define TSO_LEGACY_MAX_SIZE 65536 2362 #define TSO_MAX_SIZE UINT_MAX 2363 unsigned int tso_max_size; 2364 u16 gso_max_segs; 2365 #define TSO_MAX_SEGS U16_MAX 2366 u16 tso_max_segs; 2367 2368 #ifdef CONFIG_DCB 2369 const struct dcbnl_rtnl_ops *dcbnl_ops; 2370 #endif 2371 s16 num_tc; 2372 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2373 u8 prio_tc_map[TC_BITMASK + 1]; 2374 2375 #if IS_ENABLED(CONFIG_FCOE) 2376 unsigned int fcoe_ddp_xid; 2377 #endif 2378 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2379 struct netprio_map __rcu *priomap; 2380 #endif 2381 struct phy_device *phydev; 2382 struct sfp_bus *sfp_bus; 2383 struct lock_class_key *qdisc_tx_busylock; 2384 bool proto_down; 2385 unsigned wol_enabled:1; 2386 unsigned threaded:1; 2387 2388 struct list_head net_notifier_list; 2389 2390 #if IS_ENABLED(CONFIG_MACSEC) 2391 /* MACsec management functions */ 2392 const struct macsec_ops *macsec_ops; 2393 #endif 2394 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2395 struct udp_tunnel_nic *udp_tunnel_nic; 2396 2397 /* protected by rtnl_lock */ 2398 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2399 2400 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2401 netdevice_tracker linkwatch_dev_tracker; 2402 netdevice_tracker watchdog_dev_tracker; 2403 netdevice_tracker dev_registered_tracker; 2404 struct rtnl_hw_stats64 *offload_xstats_l3; 2405 2406 ANDROID_KABI_RESERVE(1); 2407 ANDROID_KABI_RESERVE(2); 2408 ANDROID_KABI_RESERVE(3); 2409 ANDROID_KABI_RESERVE(4); 2410 ANDROID_KABI_RESERVE(5); 2411 ANDROID_KABI_RESERVE(6); 2412 ANDROID_KABI_RESERVE(7); 2413 ANDROID_KABI_RESERVE(8); 2414 }; 2415 #define to_net_dev(d) container_of(d, struct net_device, dev) 2416 netif_elide_gro(const struct net_device * dev)2417 static inline bool netif_elide_gro(const struct net_device *dev) 2418 { 2419 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2420 return true; 2421 return false; 2422 } 2423 2424 #define NETDEV_ALIGN 32 2425 2426 static inline netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2427 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2428 { 2429 return dev->prio_tc_map[prio & TC_BITMASK]; 2430 } 2431 2432 static inline netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2433 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2434 { 2435 if (tc >= dev->num_tc) 2436 return -EINVAL; 2437 2438 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2439 return 0; 2440 } 2441 2442 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2443 void netdev_reset_tc(struct net_device *dev); 2444 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2445 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2446 2447 static inline netdev_get_num_tc(struct net_device * dev)2448 int netdev_get_num_tc(struct net_device *dev) 2449 { 2450 return dev->num_tc; 2451 } 2452 net_prefetch(void * p)2453 static inline void net_prefetch(void *p) 2454 { 2455 prefetch(p); 2456 #if L1_CACHE_BYTES < 128 2457 prefetch((u8 *)p + L1_CACHE_BYTES); 2458 #endif 2459 } 2460 net_prefetchw(void * p)2461 static inline void net_prefetchw(void *p) 2462 { 2463 prefetchw(p); 2464 #if L1_CACHE_BYTES < 128 2465 prefetchw((u8 *)p + L1_CACHE_BYTES); 2466 #endif 2467 } 2468 2469 void netdev_unbind_sb_channel(struct net_device *dev, 2470 struct net_device *sb_dev); 2471 int netdev_bind_sb_channel_queue(struct net_device *dev, 2472 struct net_device *sb_dev, 2473 u8 tc, u16 count, u16 offset); 2474 int netdev_set_sb_channel(struct net_device *dev, u16 channel); netdev_get_sb_channel(struct net_device * dev)2475 static inline int netdev_get_sb_channel(struct net_device *dev) 2476 { 2477 return max_t(int, -dev->num_tc, 0); 2478 } 2479 2480 static inline netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2481 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2482 unsigned int index) 2483 { 2484 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2485 return &dev->_tx[index]; 2486 } 2487 skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2488 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2489 const struct sk_buff *skb) 2490 { 2491 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2492 } 2493 netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2494 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2495 void (*f)(struct net_device *, 2496 struct netdev_queue *, 2497 void *), 2498 void *arg) 2499 { 2500 unsigned int i; 2501 2502 for (i = 0; i < dev->num_tx_queues; i++) 2503 f(dev, &dev->_tx[i], arg); 2504 } 2505 2506 #define netdev_lockdep_set_classes(dev) \ 2507 { \ 2508 static struct lock_class_key qdisc_tx_busylock_key; \ 2509 static struct lock_class_key qdisc_xmit_lock_key; \ 2510 static struct lock_class_key dev_addr_list_lock_key; \ 2511 unsigned int i; \ 2512 \ 2513 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2514 lockdep_set_class(&(dev)->addr_list_lock, \ 2515 &dev_addr_list_lock_key); \ 2516 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2517 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2518 &qdisc_xmit_lock_key); \ 2519 } 2520 2521 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2522 struct net_device *sb_dev); 2523 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2524 struct sk_buff *skb, 2525 struct net_device *sb_dev); 2526 2527 /* returns the headroom that the master device needs to take in account 2528 * when forwarding to this dev 2529 */ netdev_get_fwd_headroom(struct net_device * dev)2530 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2531 { 2532 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2533 } 2534 netdev_set_rx_headroom(struct net_device * dev,int new_hr)2535 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2536 { 2537 if (dev->netdev_ops->ndo_set_rx_headroom) 2538 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2539 } 2540 2541 /* set the device rx headroom to the dev's default */ netdev_reset_rx_headroom(struct net_device * dev)2542 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2543 { 2544 netdev_set_rx_headroom(dev, -1); 2545 } 2546 netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2547 static inline void *netdev_get_ml_priv(struct net_device *dev, 2548 enum netdev_ml_priv_type type) 2549 { 2550 if (dev->ml_priv_type != type) 2551 return NULL; 2552 2553 return dev->ml_priv; 2554 } 2555 netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2556 static inline void netdev_set_ml_priv(struct net_device *dev, 2557 void *ml_priv, 2558 enum netdev_ml_priv_type type) 2559 { 2560 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2561 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2562 dev->ml_priv_type, type); 2563 WARN(!dev->ml_priv_type && dev->ml_priv, 2564 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2565 2566 dev->ml_priv = ml_priv; 2567 dev->ml_priv_type = type; 2568 } 2569 2570 /* 2571 * Net namespace inlines 2572 */ 2573 static inline dev_net(const struct net_device * dev)2574 struct net *dev_net(const struct net_device *dev) 2575 { 2576 return read_pnet(&dev->nd_net); 2577 } 2578 2579 static inline dev_net_set(struct net_device * dev,struct net * net)2580 void dev_net_set(struct net_device *dev, struct net *net) 2581 { 2582 write_pnet(&dev->nd_net, net); 2583 } 2584 2585 /** 2586 * netdev_priv - access network device private data 2587 * @dev: network device 2588 * 2589 * Get network device private data 2590 */ netdev_priv(const struct net_device * dev)2591 static inline void *netdev_priv(const struct net_device *dev) 2592 { 2593 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2594 } 2595 2596 /* Set the sysfs physical device reference for the network logical device 2597 * if set prior to registration will cause a symlink during initialization. 2598 */ 2599 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2600 2601 /* Set the sysfs device type for the network logical device to allow 2602 * fine-grained identification of different network device types. For 2603 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2604 */ 2605 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2606 2607 /* Default NAPI poll() weight 2608 * Device drivers are strongly advised to not use bigger value 2609 */ 2610 #define NAPI_POLL_WEIGHT 64 2611 2612 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2613 int (*poll)(struct napi_struct *, int), int weight); 2614 2615 /** 2616 * netif_napi_add() - initialize a NAPI context 2617 * @dev: network device 2618 * @napi: NAPI context 2619 * @poll: polling function 2620 * 2621 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2622 * *any* of the other NAPI-related functions. 2623 */ 2624 static inline void netif_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2625 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2626 int (*poll)(struct napi_struct *, int)) 2627 { 2628 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2629 } 2630 2631 static inline void netif_napi_add_tx_weight(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2632 netif_napi_add_tx_weight(struct net_device *dev, 2633 struct napi_struct *napi, 2634 int (*poll)(struct napi_struct *, int), 2635 int weight) 2636 { 2637 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2638 netif_napi_add_weight(dev, napi, poll, weight); 2639 } 2640 2641 /** 2642 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2643 * @dev: network device 2644 * @napi: NAPI context 2645 * @poll: polling function 2646 * 2647 * This variant of netif_napi_add() should be used from drivers using NAPI 2648 * to exclusively poll a TX queue. 2649 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2650 */ netif_napi_add_tx(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2651 static inline void netif_napi_add_tx(struct net_device *dev, 2652 struct napi_struct *napi, 2653 int (*poll)(struct napi_struct *, int)) 2654 { 2655 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2656 } 2657 2658 /** 2659 * __netif_napi_del - remove a NAPI context 2660 * @napi: NAPI context 2661 * 2662 * Warning: caller must observe RCU grace period before freeing memory 2663 * containing @napi. Drivers might want to call this helper to combine 2664 * all the needed RCU grace periods into a single one. 2665 */ 2666 void __netif_napi_del(struct napi_struct *napi); 2667 2668 /** 2669 * netif_napi_del - remove a NAPI context 2670 * @napi: NAPI context 2671 * 2672 * netif_napi_del() removes a NAPI context from the network device NAPI list 2673 */ netif_napi_del(struct napi_struct * napi)2674 static inline void netif_napi_del(struct napi_struct *napi) 2675 { 2676 __netif_napi_del(napi); 2677 synchronize_net(); 2678 } 2679 2680 struct packet_type { 2681 __be16 type; /* This is really htons(ether_type). */ 2682 bool ignore_outgoing; 2683 struct net_device *dev; /* NULL is wildcarded here */ 2684 netdevice_tracker dev_tracker; 2685 int (*func) (struct sk_buff *, 2686 struct net_device *, 2687 struct packet_type *, 2688 struct net_device *); 2689 void (*list_func) (struct list_head *, 2690 struct packet_type *, 2691 struct net_device *); 2692 bool (*id_match)(struct packet_type *ptype, 2693 struct sock *sk); 2694 struct net *af_packet_net; 2695 void *af_packet_priv; 2696 struct list_head list; 2697 2698 ANDROID_KABI_RESERVE(1); 2699 ANDROID_KABI_RESERVE(2); 2700 ANDROID_KABI_RESERVE(3); 2701 ANDROID_KABI_RESERVE(4); 2702 }; 2703 2704 struct offload_callbacks { 2705 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2706 netdev_features_t features); 2707 struct sk_buff *(*gro_receive)(struct list_head *head, 2708 struct sk_buff *skb); 2709 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2710 }; 2711 2712 struct packet_offload { 2713 __be16 type; /* This is really htons(ether_type). */ 2714 u16 priority; 2715 struct offload_callbacks callbacks; 2716 struct list_head list; 2717 }; 2718 2719 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2720 struct pcpu_sw_netstats { 2721 u64_stats_t rx_packets; 2722 u64_stats_t rx_bytes; 2723 u64_stats_t tx_packets; 2724 u64_stats_t tx_bytes; 2725 struct u64_stats_sync syncp; 2726 } __aligned(4 * sizeof(u64)); 2727 2728 struct pcpu_lstats { 2729 u64_stats_t packets; 2730 u64_stats_t bytes; 2731 struct u64_stats_sync syncp; 2732 } __aligned(2 * sizeof(u64)); 2733 2734 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2735 dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2736 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2737 { 2738 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2739 2740 u64_stats_update_begin(&tstats->syncp); 2741 u64_stats_add(&tstats->rx_bytes, len); 2742 u64_stats_inc(&tstats->rx_packets); 2743 u64_stats_update_end(&tstats->syncp); 2744 } 2745 dev_sw_netstats_tx_add(struct net_device * dev,unsigned int packets,unsigned int len)2746 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2747 unsigned int packets, 2748 unsigned int len) 2749 { 2750 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2751 2752 u64_stats_update_begin(&tstats->syncp); 2753 u64_stats_add(&tstats->tx_bytes, len); 2754 u64_stats_add(&tstats->tx_packets, packets); 2755 u64_stats_update_end(&tstats->syncp); 2756 } 2757 dev_lstats_add(struct net_device * dev,unsigned int len)2758 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2759 { 2760 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2761 2762 u64_stats_update_begin(&lstats->syncp); 2763 u64_stats_add(&lstats->bytes, len); 2764 u64_stats_inc(&lstats->packets); 2765 u64_stats_update_end(&lstats->syncp); 2766 } 2767 2768 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2769 ({ \ 2770 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2771 if (pcpu_stats) { \ 2772 int __cpu; \ 2773 for_each_possible_cpu(__cpu) { \ 2774 typeof(type) *stat; \ 2775 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2776 u64_stats_init(&stat->syncp); \ 2777 } \ 2778 } \ 2779 pcpu_stats; \ 2780 }) 2781 2782 #define netdev_alloc_pcpu_stats(type) \ 2783 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2784 2785 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2786 ({ \ 2787 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2788 if (pcpu_stats) { \ 2789 int __cpu; \ 2790 for_each_possible_cpu(__cpu) { \ 2791 typeof(type) *stat; \ 2792 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2793 u64_stats_init(&stat->syncp); \ 2794 } \ 2795 } \ 2796 pcpu_stats; \ 2797 }) 2798 2799 enum netdev_lag_tx_type { 2800 NETDEV_LAG_TX_TYPE_UNKNOWN, 2801 NETDEV_LAG_TX_TYPE_RANDOM, 2802 NETDEV_LAG_TX_TYPE_BROADCAST, 2803 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2804 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2805 NETDEV_LAG_TX_TYPE_HASH, 2806 }; 2807 2808 enum netdev_lag_hash { 2809 NETDEV_LAG_HASH_NONE, 2810 NETDEV_LAG_HASH_L2, 2811 NETDEV_LAG_HASH_L34, 2812 NETDEV_LAG_HASH_L23, 2813 NETDEV_LAG_HASH_E23, 2814 NETDEV_LAG_HASH_E34, 2815 NETDEV_LAG_HASH_VLAN_SRCMAC, 2816 NETDEV_LAG_HASH_UNKNOWN, 2817 }; 2818 2819 struct netdev_lag_upper_info { 2820 enum netdev_lag_tx_type tx_type; 2821 enum netdev_lag_hash hash_type; 2822 }; 2823 2824 struct netdev_lag_lower_state_info { 2825 u8 link_up : 1, 2826 tx_enabled : 1; 2827 }; 2828 2829 #include <linux/notifier.h> 2830 2831 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2832 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2833 * adding new types. 2834 */ 2835 enum netdev_cmd { 2836 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2837 NETDEV_DOWN, 2838 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2839 detected a hardware crash and restarted 2840 - we can use this eg to kick tcp sessions 2841 once done */ 2842 NETDEV_CHANGE, /* Notify device state change */ 2843 NETDEV_REGISTER, 2844 NETDEV_UNREGISTER, 2845 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2846 NETDEV_CHANGEADDR, /* notify after the address change */ 2847 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2848 NETDEV_GOING_DOWN, 2849 NETDEV_CHANGENAME, 2850 NETDEV_FEAT_CHANGE, 2851 NETDEV_BONDING_FAILOVER, 2852 NETDEV_PRE_UP, 2853 NETDEV_PRE_TYPE_CHANGE, 2854 NETDEV_POST_TYPE_CHANGE, 2855 NETDEV_POST_INIT, 2856 NETDEV_RELEASE, 2857 NETDEV_NOTIFY_PEERS, 2858 NETDEV_JOIN, 2859 NETDEV_CHANGEUPPER, 2860 NETDEV_RESEND_IGMP, 2861 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2862 NETDEV_CHANGEINFODATA, 2863 NETDEV_BONDING_INFO, 2864 NETDEV_PRECHANGEUPPER, 2865 NETDEV_CHANGELOWERSTATE, 2866 NETDEV_UDP_TUNNEL_PUSH_INFO, 2867 NETDEV_UDP_TUNNEL_DROP_INFO, 2868 NETDEV_CHANGE_TX_QUEUE_LEN, 2869 NETDEV_CVLAN_FILTER_PUSH_INFO, 2870 NETDEV_CVLAN_FILTER_DROP_INFO, 2871 NETDEV_SVLAN_FILTER_PUSH_INFO, 2872 NETDEV_SVLAN_FILTER_DROP_INFO, 2873 NETDEV_OFFLOAD_XSTATS_ENABLE, 2874 NETDEV_OFFLOAD_XSTATS_DISABLE, 2875 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 2876 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 2877 }; 2878 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2879 2880 int register_netdevice_notifier(struct notifier_block *nb); 2881 int unregister_netdevice_notifier(struct notifier_block *nb); 2882 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2883 int unregister_netdevice_notifier_net(struct net *net, 2884 struct notifier_block *nb); 2885 int register_netdevice_notifier_dev_net(struct net_device *dev, 2886 struct notifier_block *nb, 2887 struct netdev_net_notifier *nn); 2888 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2889 struct notifier_block *nb, 2890 struct netdev_net_notifier *nn); 2891 2892 struct netdev_notifier_info { 2893 struct net_device *dev; 2894 struct netlink_ext_ack *extack; 2895 }; 2896 2897 struct netdev_notifier_info_ext { 2898 struct netdev_notifier_info info; /* must be first */ 2899 union { 2900 u32 mtu; 2901 } ext; 2902 }; 2903 2904 struct netdev_notifier_change_info { 2905 struct netdev_notifier_info info; /* must be first */ 2906 unsigned int flags_changed; 2907 }; 2908 2909 struct netdev_notifier_changeupper_info { 2910 struct netdev_notifier_info info; /* must be first */ 2911 struct net_device *upper_dev; /* new upper dev */ 2912 bool master; /* is upper dev master */ 2913 bool linking; /* is the notification for link or unlink */ 2914 void *upper_info; /* upper dev info */ 2915 }; 2916 2917 struct netdev_notifier_changelowerstate_info { 2918 struct netdev_notifier_info info; /* must be first */ 2919 void *lower_state_info; /* is lower dev state */ 2920 }; 2921 2922 struct netdev_notifier_pre_changeaddr_info { 2923 struct netdev_notifier_info info; /* must be first */ 2924 const unsigned char *dev_addr; 2925 }; 2926 2927 enum netdev_offload_xstats_type { 2928 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 2929 }; 2930 2931 struct netdev_notifier_offload_xstats_info { 2932 struct netdev_notifier_info info; /* must be first */ 2933 enum netdev_offload_xstats_type type; 2934 2935 union { 2936 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 2937 struct netdev_notifier_offload_xstats_rd *report_delta; 2938 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 2939 struct netdev_notifier_offload_xstats_ru *report_used; 2940 }; 2941 }; 2942 2943 int netdev_offload_xstats_enable(struct net_device *dev, 2944 enum netdev_offload_xstats_type type, 2945 struct netlink_ext_ack *extack); 2946 int netdev_offload_xstats_disable(struct net_device *dev, 2947 enum netdev_offload_xstats_type type); 2948 bool netdev_offload_xstats_enabled(const struct net_device *dev, 2949 enum netdev_offload_xstats_type type); 2950 int netdev_offload_xstats_get(struct net_device *dev, 2951 enum netdev_offload_xstats_type type, 2952 struct rtnl_hw_stats64 *stats, bool *used, 2953 struct netlink_ext_ack *extack); 2954 void 2955 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 2956 const struct rtnl_hw_stats64 *stats); 2957 void 2958 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 2959 void netdev_offload_xstats_push_delta(struct net_device *dev, 2960 enum netdev_offload_xstats_type type, 2961 const struct rtnl_hw_stats64 *stats); 2962 netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2963 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2964 struct net_device *dev) 2965 { 2966 info->dev = dev; 2967 info->extack = NULL; 2968 } 2969 2970 static inline struct net_device * netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2971 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2972 { 2973 return info->dev; 2974 } 2975 2976 static inline struct netlink_ext_ack * netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2977 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2978 { 2979 return info->extack; 2980 } 2981 2982 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2983 2984 2985 extern rwlock_t dev_base_lock; /* Device list lock */ 2986 2987 #define for_each_netdev(net, d) \ 2988 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2989 #define for_each_netdev_reverse(net, d) \ 2990 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2991 #define for_each_netdev_rcu(net, d) \ 2992 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2993 #define for_each_netdev_safe(net, d, n) \ 2994 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2995 #define for_each_netdev_continue(net, d) \ 2996 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2997 #define for_each_netdev_continue_reverse(net, d) \ 2998 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2999 dev_list) 3000 #define for_each_netdev_continue_rcu(net, d) \ 3001 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3002 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3003 for_each_netdev_rcu(&init_net, slave) \ 3004 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3005 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3006 next_net_device(struct net_device * dev)3007 static inline struct net_device *next_net_device(struct net_device *dev) 3008 { 3009 struct list_head *lh; 3010 struct net *net; 3011 3012 net = dev_net(dev); 3013 lh = dev->dev_list.next; 3014 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3015 } 3016 next_net_device_rcu(struct net_device * dev)3017 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3018 { 3019 struct list_head *lh; 3020 struct net *net; 3021 3022 net = dev_net(dev); 3023 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3024 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3025 } 3026 first_net_device(struct net * net)3027 static inline struct net_device *first_net_device(struct net *net) 3028 { 3029 return list_empty(&net->dev_base_head) ? NULL : 3030 net_device_entry(net->dev_base_head.next); 3031 } 3032 first_net_device_rcu(struct net * net)3033 static inline struct net_device *first_net_device_rcu(struct net *net) 3034 { 3035 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 3036 3037 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3038 } 3039 3040 int netdev_boot_setup_check(struct net_device *dev); 3041 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3042 const char *hwaddr); 3043 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3044 void dev_add_pack(struct packet_type *pt); 3045 void dev_remove_pack(struct packet_type *pt); 3046 void __dev_remove_pack(struct packet_type *pt); 3047 void dev_add_offload(struct packet_offload *po); 3048 void dev_remove_offload(struct packet_offload *po); 3049 3050 int dev_get_iflink(const struct net_device *dev); 3051 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3052 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3053 struct net_device_path_stack *stack); 3054 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 3055 unsigned short mask); 3056 struct net_device *dev_get_by_name(struct net *net, const char *name); 3057 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3058 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3059 bool netdev_name_in_use(struct net *net, const char *name); 3060 int dev_alloc_name(struct net_device *dev, const char *name); 3061 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3062 void dev_close(struct net_device *dev); 3063 void dev_close_many(struct list_head *head, bool unlink); 3064 void dev_disable_lro(struct net_device *dev); 3065 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3066 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3067 struct net_device *sb_dev); 3068 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 3069 struct net_device *sb_dev); 3070 3071 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3072 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3073 dev_queue_xmit(struct sk_buff * skb)3074 static inline int dev_queue_xmit(struct sk_buff *skb) 3075 { 3076 return __dev_queue_xmit(skb, NULL); 3077 } 3078 dev_queue_xmit_accel(struct sk_buff * skb,struct net_device * sb_dev)3079 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3080 struct net_device *sb_dev) 3081 { 3082 return __dev_queue_xmit(skb, sb_dev); 3083 } 3084 dev_direct_xmit(struct sk_buff * skb,u16 queue_id)3085 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3086 { 3087 int ret; 3088 3089 ret = __dev_direct_xmit(skb, queue_id); 3090 if (!dev_xmit_complete(ret)) 3091 kfree_skb(skb); 3092 return ret; 3093 } 3094 3095 int register_netdevice(struct net_device *dev); 3096 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3097 void unregister_netdevice_many(struct list_head *head); unregister_netdevice(struct net_device * dev)3098 static inline void unregister_netdevice(struct net_device *dev) 3099 { 3100 unregister_netdevice_queue(dev, NULL); 3101 } 3102 3103 int netdev_refcnt_read(const struct net_device *dev); 3104 void free_netdev(struct net_device *dev); 3105 void netdev_freemem(struct net_device *dev); 3106 int init_dummy_netdev(struct net_device *dev); 3107 3108 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3109 struct sk_buff *skb, 3110 bool all_slaves); 3111 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3112 struct sock *sk); 3113 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3114 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3115 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3116 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3117 int dev_restart(struct net_device *dev); 3118 3119 dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3120 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3121 unsigned short type, 3122 const void *daddr, const void *saddr, 3123 unsigned int len) 3124 { 3125 if (!dev->header_ops || !dev->header_ops->create) 3126 return 0; 3127 3128 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3129 } 3130 dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3131 static inline int dev_parse_header(const struct sk_buff *skb, 3132 unsigned char *haddr) 3133 { 3134 const struct net_device *dev = skb->dev; 3135 3136 if (!dev->header_ops || !dev->header_ops->parse) 3137 return 0; 3138 return dev->header_ops->parse(skb, haddr); 3139 } 3140 dev_parse_header_protocol(const struct sk_buff * skb)3141 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3142 { 3143 const struct net_device *dev = skb->dev; 3144 3145 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3146 return 0; 3147 return dev->header_ops->parse_protocol(skb); 3148 } 3149 3150 /* ll_header must have at least hard_header_len allocated */ dev_validate_header(const struct net_device * dev,char * ll_header,int len)3151 static inline bool dev_validate_header(const struct net_device *dev, 3152 char *ll_header, int len) 3153 { 3154 if (likely(len >= dev->hard_header_len)) 3155 return true; 3156 if (len < dev->min_header_len) 3157 return false; 3158 3159 if (capable(CAP_SYS_RAWIO)) { 3160 memset(ll_header + len, 0, dev->hard_header_len - len); 3161 return true; 3162 } 3163 3164 if (dev->header_ops && dev->header_ops->validate) 3165 return dev->header_ops->validate(ll_header, len); 3166 3167 return false; 3168 } 3169 dev_has_header(const struct net_device * dev)3170 static inline bool dev_has_header(const struct net_device *dev) 3171 { 3172 return dev->header_ops && dev->header_ops->create; 3173 } 3174 3175 /* 3176 * Incoming packets are placed on per-CPU queues 3177 */ 3178 struct softnet_data { 3179 struct list_head poll_list; 3180 struct sk_buff_head process_queue; 3181 3182 /* stats */ 3183 unsigned int processed; 3184 unsigned int time_squeeze; 3185 unsigned int received_rps; 3186 #ifdef CONFIG_RPS 3187 struct softnet_data *rps_ipi_list; 3188 #endif 3189 #ifdef CONFIG_NET_FLOW_LIMIT 3190 struct sd_flow_limit __rcu *flow_limit; 3191 #endif 3192 struct Qdisc *output_queue; 3193 struct Qdisc **output_queue_tailp; 3194 struct sk_buff *completion_queue; 3195 #ifdef CONFIG_XFRM_OFFLOAD 3196 struct sk_buff_head xfrm_backlog; 3197 #endif 3198 /* written and read only by owning cpu: */ 3199 struct { 3200 u16 recursion; 3201 u8 more; 3202 #ifdef CONFIG_NET_EGRESS 3203 u8 skip_txqueue; 3204 #endif 3205 } xmit; 3206 #ifdef CONFIG_RPS 3207 /* input_queue_head should be written by cpu owning this struct, 3208 * and only read by other cpus. Worth using a cache line. 3209 */ 3210 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3211 3212 /* Elements below can be accessed between CPUs for RPS/RFS */ 3213 call_single_data_t csd ____cacheline_aligned_in_smp; 3214 struct softnet_data *rps_ipi_next; 3215 unsigned int cpu; 3216 unsigned int input_queue_tail; 3217 #endif 3218 unsigned int dropped; 3219 struct sk_buff_head input_pkt_queue; 3220 struct napi_struct backlog; 3221 3222 /* Another possibly contended cache line */ 3223 spinlock_t defer_lock ____cacheline_aligned_in_smp; 3224 int defer_count; 3225 int defer_ipi_scheduled; 3226 struct sk_buff *defer_list; 3227 call_single_data_t defer_csd; 3228 }; 3229 input_queue_head_incr(struct softnet_data * sd)3230 static inline void input_queue_head_incr(struct softnet_data *sd) 3231 { 3232 #ifdef CONFIG_RPS 3233 sd->input_queue_head++; 3234 #endif 3235 } 3236 input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3237 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3238 unsigned int *qtail) 3239 { 3240 #ifdef CONFIG_RPS 3241 *qtail = ++sd->input_queue_tail; 3242 #endif 3243 } 3244 3245 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3246 dev_recursion_level(void)3247 static inline int dev_recursion_level(void) 3248 { 3249 return this_cpu_read(softnet_data.xmit.recursion); 3250 } 3251 3252 #define XMIT_RECURSION_LIMIT 8 dev_xmit_recursion(void)3253 static inline bool dev_xmit_recursion(void) 3254 { 3255 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3256 XMIT_RECURSION_LIMIT); 3257 } 3258 dev_xmit_recursion_inc(void)3259 static inline void dev_xmit_recursion_inc(void) 3260 { 3261 __this_cpu_inc(softnet_data.xmit.recursion); 3262 } 3263 dev_xmit_recursion_dec(void)3264 static inline void dev_xmit_recursion_dec(void) 3265 { 3266 __this_cpu_dec(softnet_data.xmit.recursion); 3267 } 3268 3269 void __netif_schedule(struct Qdisc *q); 3270 void netif_schedule_queue(struct netdev_queue *txq); 3271 netif_tx_schedule_all(struct net_device * dev)3272 static inline void netif_tx_schedule_all(struct net_device *dev) 3273 { 3274 unsigned int i; 3275 3276 for (i = 0; i < dev->num_tx_queues; i++) 3277 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3278 } 3279 netif_tx_start_queue(struct netdev_queue * dev_queue)3280 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3281 { 3282 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3283 } 3284 3285 /** 3286 * netif_start_queue - allow transmit 3287 * @dev: network device 3288 * 3289 * Allow upper layers to call the device hard_start_xmit routine. 3290 */ netif_start_queue(struct net_device * dev)3291 static inline void netif_start_queue(struct net_device *dev) 3292 { 3293 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3294 } 3295 netif_tx_start_all_queues(struct net_device * dev)3296 static inline void netif_tx_start_all_queues(struct net_device *dev) 3297 { 3298 unsigned int i; 3299 3300 for (i = 0; i < dev->num_tx_queues; i++) { 3301 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3302 netif_tx_start_queue(txq); 3303 } 3304 } 3305 3306 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3307 3308 /** 3309 * netif_wake_queue - restart transmit 3310 * @dev: network device 3311 * 3312 * Allow upper layers to call the device hard_start_xmit routine. 3313 * Used for flow control when transmit resources are available. 3314 */ netif_wake_queue(struct net_device * dev)3315 static inline void netif_wake_queue(struct net_device *dev) 3316 { 3317 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3318 } 3319 netif_tx_wake_all_queues(struct net_device * dev)3320 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3321 { 3322 unsigned int i; 3323 3324 for (i = 0; i < dev->num_tx_queues; i++) { 3325 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3326 netif_tx_wake_queue(txq); 3327 } 3328 } 3329 netif_tx_stop_queue(struct netdev_queue * dev_queue)3330 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3331 { 3332 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3333 } 3334 3335 /** 3336 * netif_stop_queue - stop transmitted packets 3337 * @dev: network device 3338 * 3339 * Stop upper layers calling the device hard_start_xmit routine. 3340 * Used for flow control when transmit resources are unavailable. 3341 */ netif_stop_queue(struct net_device * dev)3342 static inline void netif_stop_queue(struct net_device *dev) 3343 { 3344 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3345 } 3346 3347 void netif_tx_stop_all_queues(struct net_device *dev); 3348 netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3349 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3350 { 3351 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3352 } 3353 3354 /** 3355 * netif_queue_stopped - test if transmit queue is flowblocked 3356 * @dev: network device 3357 * 3358 * Test if transmit queue on device is currently unable to send. 3359 */ netif_queue_stopped(const struct net_device * dev)3360 static inline bool netif_queue_stopped(const struct net_device *dev) 3361 { 3362 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3363 } 3364 netif_xmit_stopped(const struct netdev_queue * dev_queue)3365 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3366 { 3367 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3368 } 3369 3370 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3371 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3372 { 3373 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3374 } 3375 3376 static inline bool netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3377 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3378 { 3379 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3380 } 3381 3382 /** 3383 * netdev_queue_set_dql_min_limit - set dql minimum limit 3384 * @dev_queue: pointer to transmit queue 3385 * @min_limit: dql minimum limit 3386 * 3387 * Forces xmit_more() to return true until the minimum threshold 3388 * defined by @min_limit is reached (or until the tx queue is 3389 * empty). Warning: to be use with care, misuse will impact the 3390 * latency. 3391 */ netdev_queue_set_dql_min_limit(struct netdev_queue * dev_queue,unsigned int min_limit)3392 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3393 unsigned int min_limit) 3394 { 3395 #ifdef CONFIG_BQL 3396 dev_queue->dql.min_limit = min_limit; 3397 #endif 3398 } 3399 3400 /** 3401 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3402 * @dev_queue: pointer to transmit queue 3403 * 3404 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3405 * to give appropriate hint to the CPU. 3406 */ netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3407 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3408 { 3409 #ifdef CONFIG_BQL 3410 prefetchw(&dev_queue->dql.num_queued); 3411 #endif 3412 } 3413 3414 /** 3415 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3416 * @dev_queue: pointer to transmit queue 3417 * 3418 * BQL enabled drivers might use this helper in their TX completion path, 3419 * to give appropriate hint to the CPU. 3420 */ netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3421 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3422 { 3423 #ifdef CONFIG_BQL 3424 prefetchw(&dev_queue->dql.limit); 3425 #endif 3426 } 3427 3428 /** 3429 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3430 * @dev_queue: network device queue 3431 * @bytes: number of bytes queued to the device queue 3432 * 3433 * Report the number of bytes queued for sending/completion to the network 3434 * device hardware queue. @bytes should be a good approximation and should 3435 * exactly match netdev_completed_queue() @bytes. 3436 * This is typically called once per packet, from ndo_start_xmit(). 3437 */ netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3438 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3439 unsigned int bytes) 3440 { 3441 #ifdef CONFIG_BQL 3442 dql_queued(&dev_queue->dql, bytes); 3443 3444 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3445 return; 3446 3447 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3448 3449 /* 3450 * The XOFF flag must be set before checking the dql_avail below, 3451 * because in netdev_tx_completed_queue we update the dql_completed 3452 * before checking the XOFF flag. 3453 */ 3454 smp_mb(); 3455 3456 /* check again in case another CPU has just made room avail */ 3457 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3458 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3459 #endif 3460 } 3461 3462 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3463 * that they should not test BQL status themselves. 3464 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3465 * skb of a batch. 3466 * Returns true if the doorbell must be used to kick the NIC. 3467 */ __netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3468 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3469 unsigned int bytes, 3470 bool xmit_more) 3471 { 3472 if (xmit_more) { 3473 #ifdef CONFIG_BQL 3474 dql_queued(&dev_queue->dql, bytes); 3475 #endif 3476 return netif_tx_queue_stopped(dev_queue); 3477 } 3478 netdev_tx_sent_queue(dev_queue, bytes); 3479 return true; 3480 } 3481 3482 /** 3483 * netdev_sent_queue - report the number of bytes queued to hardware 3484 * @dev: network device 3485 * @bytes: number of bytes queued to the hardware device queue 3486 * 3487 * Report the number of bytes queued for sending/completion to the network 3488 * device hardware queue#0. @bytes should be a good approximation and should 3489 * exactly match netdev_completed_queue() @bytes. 3490 * This is typically called once per packet, from ndo_start_xmit(). 3491 */ netdev_sent_queue(struct net_device * dev,unsigned int bytes)3492 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3493 { 3494 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3495 } 3496 __netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3497 static inline bool __netdev_sent_queue(struct net_device *dev, 3498 unsigned int bytes, 3499 bool xmit_more) 3500 { 3501 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3502 xmit_more); 3503 } 3504 3505 /** 3506 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3507 * @dev_queue: network device queue 3508 * @pkts: number of packets (currently ignored) 3509 * @bytes: number of bytes dequeued from the device queue 3510 * 3511 * Must be called at most once per TX completion round (and not per 3512 * individual packet), so that BQL can adjust its limits appropriately. 3513 */ netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3514 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3515 unsigned int pkts, unsigned int bytes) 3516 { 3517 #ifdef CONFIG_BQL 3518 if (unlikely(!bytes)) 3519 return; 3520 3521 dql_completed(&dev_queue->dql, bytes); 3522 3523 /* 3524 * Without the memory barrier there is a small possiblity that 3525 * netdev_tx_sent_queue will miss the update and cause the queue to 3526 * be stopped forever 3527 */ 3528 smp_mb(); 3529 3530 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3531 return; 3532 3533 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3534 netif_schedule_queue(dev_queue); 3535 #endif 3536 } 3537 3538 /** 3539 * netdev_completed_queue - report bytes and packets completed by device 3540 * @dev: network device 3541 * @pkts: actual number of packets sent over the medium 3542 * @bytes: actual number of bytes sent over the medium 3543 * 3544 * Report the number of bytes and packets transmitted by the network device 3545 * hardware queue over the physical medium, @bytes must exactly match the 3546 * @bytes amount passed to netdev_sent_queue() 3547 */ netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3548 static inline void netdev_completed_queue(struct net_device *dev, 3549 unsigned int pkts, unsigned int bytes) 3550 { 3551 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3552 } 3553 netdev_tx_reset_queue(struct netdev_queue * q)3554 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3555 { 3556 #ifdef CONFIG_BQL 3557 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3558 dql_reset(&q->dql); 3559 #endif 3560 } 3561 3562 /** 3563 * netdev_reset_queue - reset the packets and bytes count of a network device 3564 * @dev_queue: network device 3565 * 3566 * Reset the bytes and packet count of a network device and clear the 3567 * software flow control OFF bit for this network device 3568 */ netdev_reset_queue(struct net_device * dev_queue)3569 static inline void netdev_reset_queue(struct net_device *dev_queue) 3570 { 3571 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3572 } 3573 3574 /** 3575 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3576 * @dev: network device 3577 * @queue_index: given tx queue index 3578 * 3579 * Returns 0 if given tx queue index >= number of device tx queues, 3580 * otherwise returns the originally passed tx queue index. 3581 */ netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3582 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3583 { 3584 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3585 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3586 dev->name, queue_index, 3587 dev->real_num_tx_queues); 3588 return 0; 3589 } 3590 3591 return queue_index; 3592 } 3593 3594 /** 3595 * netif_running - test if up 3596 * @dev: network device 3597 * 3598 * Test if the device has been brought up. 3599 */ netif_running(const struct net_device * dev)3600 static inline bool netif_running(const struct net_device *dev) 3601 { 3602 return test_bit(__LINK_STATE_START, &dev->state); 3603 } 3604 3605 /* 3606 * Routines to manage the subqueues on a device. We only need start, 3607 * stop, and a check if it's stopped. All other device management is 3608 * done at the overall netdevice level. 3609 * Also test the device if we're multiqueue. 3610 */ 3611 3612 /** 3613 * netif_start_subqueue - allow sending packets on subqueue 3614 * @dev: network device 3615 * @queue_index: sub queue index 3616 * 3617 * Start individual transmit queue of a device with multiple transmit queues. 3618 */ netif_start_subqueue(struct net_device * dev,u16 queue_index)3619 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3620 { 3621 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3622 3623 netif_tx_start_queue(txq); 3624 } 3625 3626 /** 3627 * netif_stop_subqueue - stop sending packets on subqueue 3628 * @dev: network device 3629 * @queue_index: sub queue index 3630 * 3631 * Stop individual transmit queue of a device with multiple transmit queues. 3632 */ netif_stop_subqueue(struct net_device * dev,u16 queue_index)3633 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3634 { 3635 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3636 netif_tx_stop_queue(txq); 3637 } 3638 3639 /** 3640 * __netif_subqueue_stopped - test status of subqueue 3641 * @dev: network device 3642 * @queue_index: sub queue index 3643 * 3644 * Check individual transmit queue of a device with multiple transmit queues. 3645 */ __netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3646 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3647 u16 queue_index) 3648 { 3649 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3650 3651 return netif_tx_queue_stopped(txq); 3652 } 3653 3654 /** 3655 * netif_subqueue_stopped - test status of subqueue 3656 * @dev: network device 3657 * @skb: sub queue buffer pointer 3658 * 3659 * Check individual transmit queue of a device with multiple transmit queues. 3660 */ netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3661 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3662 struct sk_buff *skb) 3663 { 3664 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3665 } 3666 3667 /** 3668 * netif_wake_subqueue - allow sending packets on subqueue 3669 * @dev: network device 3670 * @queue_index: sub queue index 3671 * 3672 * Resume individual transmit queue of a device with multiple transmit queues. 3673 */ netif_wake_subqueue(struct net_device * dev,u16 queue_index)3674 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3675 { 3676 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3677 3678 netif_tx_wake_queue(txq); 3679 } 3680 3681 #ifdef CONFIG_XPS 3682 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3683 u16 index); 3684 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3685 u16 index, enum xps_map_type type); 3686 3687 /** 3688 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3689 * @j: CPU/Rx queue index 3690 * @mask: bitmask of all cpus/rx queues 3691 * @nr_bits: number of bits in the bitmask 3692 * 3693 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3694 */ netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3695 static inline bool netif_attr_test_mask(unsigned long j, 3696 const unsigned long *mask, 3697 unsigned int nr_bits) 3698 { 3699 cpu_max_bits_warn(j, nr_bits); 3700 return test_bit(j, mask); 3701 } 3702 3703 /** 3704 * netif_attr_test_online - Test for online CPU/Rx queue 3705 * @j: CPU/Rx queue index 3706 * @online_mask: bitmask for CPUs/Rx queues that are online 3707 * @nr_bits: number of bits in the bitmask 3708 * 3709 * Returns true if a CPU/Rx queue is online. 3710 */ netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3711 static inline bool netif_attr_test_online(unsigned long j, 3712 const unsigned long *online_mask, 3713 unsigned int nr_bits) 3714 { 3715 cpu_max_bits_warn(j, nr_bits); 3716 3717 if (online_mask) 3718 return test_bit(j, online_mask); 3719 3720 return (j < nr_bits); 3721 } 3722 3723 /** 3724 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3725 * @n: CPU/Rx queue index 3726 * @srcp: the cpumask/Rx queue mask pointer 3727 * @nr_bits: number of bits in the bitmask 3728 * 3729 * Returns >= nr_bits if no further CPUs/Rx queues set. 3730 */ netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3731 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3732 unsigned int nr_bits) 3733 { 3734 /* -1 is a legal arg here. */ 3735 if (n != -1) 3736 cpu_max_bits_warn(n, nr_bits); 3737 3738 if (srcp) 3739 return find_next_bit(srcp, nr_bits, n + 1); 3740 3741 return n + 1; 3742 } 3743 3744 /** 3745 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3746 * @n: CPU/Rx queue index 3747 * @src1p: the first CPUs/Rx queues mask pointer 3748 * @src2p: the second CPUs/Rx queues mask pointer 3749 * @nr_bits: number of bits in the bitmask 3750 * 3751 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3752 */ netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3753 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3754 const unsigned long *src2p, 3755 unsigned int nr_bits) 3756 { 3757 /* -1 is a legal arg here. */ 3758 if (n != -1) 3759 cpu_max_bits_warn(n, nr_bits); 3760 3761 if (src1p && src2p) 3762 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3763 else if (src1p) 3764 return find_next_bit(src1p, nr_bits, n + 1); 3765 else if (src2p) 3766 return find_next_bit(src2p, nr_bits, n + 1); 3767 3768 return n + 1; 3769 } 3770 #else netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3771 static inline int netif_set_xps_queue(struct net_device *dev, 3772 const struct cpumask *mask, 3773 u16 index) 3774 { 3775 return 0; 3776 } 3777 __netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,enum xps_map_type type)3778 static inline int __netif_set_xps_queue(struct net_device *dev, 3779 const unsigned long *mask, 3780 u16 index, enum xps_map_type type) 3781 { 3782 return 0; 3783 } 3784 #endif 3785 3786 /** 3787 * netif_is_multiqueue - test if device has multiple transmit queues 3788 * @dev: network device 3789 * 3790 * Check if device has multiple transmit queues 3791 */ netif_is_multiqueue(const struct net_device * dev)3792 static inline bool netif_is_multiqueue(const struct net_device *dev) 3793 { 3794 return dev->num_tx_queues > 1; 3795 } 3796 3797 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3798 3799 #ifdef CONFIG_SYSFS 3800 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3801 #else netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3802 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3803 unsigned int rxqs) 3804 { 3805 dev->real_num_rx_queues = rxqs; 3806 return 0; 3807 } 3808 #endif 3809 int netif_set_real_num_queues(struct net_device *dev, 3810 unsigned int txq, unsigned int rxq); 3811 3812 static inline struct netdev_rx_queue * __netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3813 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3814 { 3815 return dev->_rx + rxq; 3816 } 3817 3818 #ifdef CONFIG_SYSFS get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3819 static inline unsigned int get_netdev_rx_queue_index( 3820 struct netdev_rx_queue *queue) 3821 { 3822 struct net_device *dev = queue->dev; 3823 int index = queue - dev->_rx; 3824 3825 BUG_ON(index >= dev->num_rx_queues); 3826 return index; 3827 } 3828 #endif 3829 3830 int netif_get_num_default_rss_queues(void); 3831 3832 enum skb_free_reason { 3833 SKB_REASON_CONSUMED, 3834 SKB_REASON_DROPPED, 3835 }; 3836 3837 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3838 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3839 3840 /* 3841 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3842 * interrupt context or with hardware interrupts being disabled. 3843 * (in_hardirq() || irqs_disabled()) 3844 * 3845 * We provide four helpers that can be used in following contexts : 3846 * 3847 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3848 * replacing kfree_skb(skb) 3849 * 3850 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3851 * Typically used in place of consume_skb(skb) in TX completion path 3852 * 3853 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3854 * replacing kfree_skb(skb) 3855 * 3856 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3857 * and consumed a packet. Used in place of consume_skb(skb) 3858 */ dev_kfree_skb_irq(struct sk_buff * skb)3859 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3860 { 3861 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3862 } 3863 dev_consume_skb_irq(struct sk_buff * skb)3864 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3865 { 3866 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3867 } 3868 dev_kfree_skb_any(struct sk_buff * skb)3869 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3870 { 3871 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3872 } 3873 dev_consume_skb_any(struct sk_buff * skb)3874 static inline void dev_consume_skb_any(struct sk_buff *skb) 3875 { 3876 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3877 } 3878 3879 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3880 struct bpf_prog *xdp_prog); 3881 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3882 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3883 int netif_rx(struct sk_buff *skb); 3884 int __netif_rx(struct sk_buff *skb); 3885 3886 int netif_receive_skb(struct sk_buff *skb); 3887 int netif_receive_skb_core(struct sk_buff *skb); 3888 void netif_receive_skb_list_internal(struct list_head *head); 3889 void netif_receive_skb_list(struct list_head *head); 3890 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3891 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3892 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3893 void napi_get_frags_check(struct napi_struct *napi); 3894 gro_result_t napi_gro_frags(struct napi_struct *napi); 3895 struct packet_offload *gro_find_receive_by_type(__be16 type); 3896 struct packet_offload *gro_find_complete_by_type(__be16 type); 3897 napi_free_frags(struct napi_struct * napi)3898 static inline void napi_free_frags(struct napi_struct *napi) 3899 { 3900 kfree_skb(napi->skb); 3901 napi->skb = NULL; 3902 } 3903 3904 bool netdev_is_rx_handler_busy(struct net_device *dev); 3905 int netdev_rx_handler_register(struct net_device *dev, 3906 rx_handler_func_t *rx_handler, 3907 void *rx_handler_data); 3908 void netdev_rx_handler_unregister(struct net_device *dev); 3909 3910 bool dev_valid_name(const char *name); is_socket_ioctl_cmd(unsigned int cmd)3911 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3912 { 3913 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3914 } 3915 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3916 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3917 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3918 void __user *data, bool *need_copyout); 3919 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3920 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3921 unsigned int dev_get_flags(const struct net_device *); 3922 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3923 struct netlink_ext_ack *extack); 3924 int dev_change_flags(struct net_device *dev, unsigned int flags, 3925 struct netlink_ext_ack *extack); 3926 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3927 unsigned int gchanges); 3928 int dev_set_alias(struct net_device *, const char *, size_t); 3929 int dev_get_alias(const struct net_device *, char *, size_t); 3930 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3931 const char *pat, int new_ifindex); 3932 static inline dev_change_net_namespace(struct net_device * dev,struct net * net,const char * pat)3933 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3934 const char *pat) 3935 { 3936 return __dev_change_net_namespace(dev, net, pat, 0); 3937 } 3938 int __dev_set_mtu(struct net_device *, int); 3939 int dev_set_mtu(struct net_device *, int); 3940 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3941 struct netlink_ext_ack *extack); 3942 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3943 struct netlink_ext_ack *extack); 3944 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3945 struct netlink_ext_ack *extack); 3946 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3947 int dev_get_port_parent_id(struct net_device *dev, 3948 struct netdev_phys_item_id *ppid, bool recurse); 3949 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3950 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3951 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3952 struct netdev_queue *txq, int *ret); 3953 3954 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3955 u8 dev_xdp_prog_count(struct net_device *dev); 3956 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3957 3958 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3959 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3960 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3961 bool is_skb_forwardable(const struct net_device *dev, 3962 const struct sk_buff *skb); 3963 __is_skb_forwardable(const struct net_device * dev,const struct sk_buff * skb,const bool check_mtu)3964 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3965 const struct sk_buff *skb, 3966 const bool check_mtu) 3967 { 3968 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3969 unsigned int len; 3970 3971 if (!(dev->flags & IFF_UP)) 3972 return false; 3973 3974 if (!check_mtu) 3975 return true; 3976 3977 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3978 if (skb->len <= len) 3979 return true; 3980 3981 /* if TSO is enabled, we don't care about the length as the packet 3982 * could be forwarded without being segmented before 3983 */ 3984 if (skb_is_gso(skb)) 3985 return true; 3986 3987 return false; 3988 } 3989 3990 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev); 3991 dev_core_stats(struct net_device * dev)3992 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev) 3993 { 3994 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */ 3995 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats); 3996 3997 if (likely(p)) 3998 return p; 3999 4000 return netdev_core_stats_alloc(dev); 4001 } 4002 4003 #define DEV_CORE_STATS_INC(FIELD) \ 4004 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4005 { \ 4006 struct net_device_core_stats __percpu *p; \ 4007 \ 4008 p = dev_core_stats(dev); \ 4009 if (p) \ 4010 this_cpu_inc(p->FIELD); \ 4011 } 4012 DEV_CORE_STATS_INC(rx_dropped) DEV_CORE_STATS_INC(tx_dropped)4013 DEV_CORE_STATS_INC(tx_dropped) 4014 DEV_CORE_STATS_INC(rx_nohandler) 4015 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4016 4017 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4018 struct sk_buff *skb, 4019 const bool check_mtu) 4020 { 4021 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4022 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4023 dev_core_stats_rx_dropped_inc(dev); 4024 kfree_skb(skb); 4025 return NET_RX_DROP; 4026 } 4027 4028 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4029 skb->priority = 0; 4030 return 0; 4031 } 4032 4033 bool dev_nit_active(struct net_device *dev); 4034 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4035 __dev_put(struct net_device * dev)4036 static inline void __dev_put(struct net_device *dev) 4037 { 4038 if (dev) { 4039 #ifdef CONFIG_PCPU_DEV_REFCNT 4040 this_cpu_dec(*dev->pcpu_refcnt); 4041 #else 4042 refcount_dec(&dev->dev_refcnt); 4043 #endif 4044 } 4045 } 4046 __dev_hold(struct net_device * dev)4047 static inline void __dev_hold(struct net_device *dev) 4048 { 4049 if (dev) { 4050 #ifdef CONFIG_PCPU_DEV_REFCNT 4051 this_cpu_inc(*dev->pcpu_refcnt); 4052 #else 4053 refcount_inc(&dev->dev_refcnt); 4054 #endif 4055 } 4056 } 4057 __netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4058 static inline void __netdev_tracker_alloc(struct net_device *dev, 4059 netdevice_tracker *tracker, 4060 gfp_t gfp) 4061 { 4062 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4063 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4064 #endif 4065 } 4066 4067 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4068 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4069 */ netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4070 static inline void netdev_tracker_alloc(struct net_device *dev, 4071 netdevice_tracker *tracker, gfp_t gfp) 4072 { 4073 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4074 refcount_dec(&dev->refcnt_tracker.no_tracker); 4075 __netdev_tracker_alloc(dev, tracker, gfp); 4076 #endif 4077 } 4078 netdev_tracker_free(struct net_device * dev,netdevice_tracker * tracker)4079 static inline void netdev_tracker_free(struct net_device *dev, 4080 netdevice_tracker *tracker) 4081 { 4082 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4083 ref_tracker_free(&dev->refcnt_tracker, tracker); 4084 #endif 4085 } 4086 netdev_hold(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4087 static inline void netdev_hold(struct net_device *dev, 4088 netdevice_tracker *tracker, gfp_t gfp) 4089 { 4090 if (dev) { 4091 __dev_hold(dev); 4092 __netdev_tracker_alloc(dev, tracker, gfp); 4093 } 4094 } 4095 netdev_put(struct net_device * dev,netdevice_tracker * tracker)4096 static inline void netdev_put(struct net_device *dev, 4097 netdevice_tracker *tracker) 4098 { 4099 if (dev) { 4100 netdev_tracker_free(dev, tracker); 4101 __dev_put(dev); 4102 } 4103 } 4104 4105 /** 4106 * dev_hold - get reference to device 4107 * @dev: network device 4108 * 4109 * Hold reference to device to keep it from being freed. 4110 * Try using netdev_hold() instead. 4111 */ dev_hold(struct net_device * dev)4112 static inline void dev_hold(struct net_device *dev) 4113 { 4114 netdev_hold(dev, NULL, GFP_ATOMIC); 4115 } 4116 4117 /** 4118 * dev_put - release reference to device 4119 * @dev: network device 4120 * 4121 * Release reference to device to allow it to be freed. 4122 * Try using netdev_put() instead. 4123 */ dev_put(struct net_device * dev)4124 static inline void dev_put(struct net_device *dev) 4125 { 4126 netdev_put(dev, NULL); 4127 } 4128 netdev_ref_replace(struct net_device * odev,struct net_device * ndev,netdevice_tracker * tracker,gfp_t gfp)4129 static inline void netdev_ref_replace(struct net_device *odev, 4130 struct net_device *ndev, 4131 netdevice_tracker *tracker, 4132 gfp_t gfp) 4133 { 4134 if (odev) 4135 netdev_tracker_free(odev, tracker); 4136 4137 __dev_hold(ndev); 4138 __dev_put(odev); 4139 4140 if (ndev) 4141 __netdev_tracker_alloc(ndev, tracker, gfp); 4142 } 4143 4144 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4145 * and _off may be called from IRQ context, but it is caller 4146 * who is responsible for serialization of these calls. 4147 * 4148 * The name carrier is inappropriate, these functions should really be 4149 * called netif_lowerlayer_*() because they represent the state of any 4150 * kind of lower layer not just hardware media. 4151 */ 4152 void linkwatch_fire_event(struct net_device *dev); 4153 4154 /** 4155 * netif_carrier_ok - test if carrier present 4156 * @dev: network device 4157 * 4158 * Check if carrier is present on device 4159 */ netif_carrier_ok(const struct net_device * dev)4160 static inline bool netif_carrier_ok(const struct net_device *dev) 4161 { 4162 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4163 } 4164 4165 unsigned long dev_trans_start(struct net_device *dev); 4166 4167 void __netdev_watchdog_up(struct net_device *dev); 4168 4169 void netif_carrier_on(struct net_device *dev); 4170 void netif_carrier_off(struct net_device *dev); 4171 void netif_carrier_event(struct net_device *dev); 4172 4173 /** 4174 * netif_dormant_on - mark device as dormant. 4175 * @dev: network device 4176 * 4177 * Mark device as dormant (as per RFC2863). 4178 * 4179 * The dormant state indicates that the relevant interface is not 4180 * actually in a condition to pass packets (i.e., it is not 'up') but is 4181 * in a "pending" state, waiting for some external event. For "on- 4182 * demand" interfaces, this new state identifies the situation where the 4183 * interface is waiting for events to place it in the up state. 4184 */ netif_dormant_on(struct net_device * dev)4185 static inline void netif_dormant_on(struct net_device *dev) 4186 { 4187 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4188 linkwatch_fire_event(dev); 4189 } 4190 4191 /** 4192 * netif_dormant_off - set device as not dormant. 4193 * @dev: network device 4194 * 4195 * Device is not in dormant state. 4196 */ netif_dormant_off(struct net_device * dev)4197 static inline void netif_dormant_off(struct net_device *dev) 4198 { 4199 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4200 linkwatch_fire_event(dev); 4201 } 4202 4203 /** 4204 * netif_dormant - test if device is dormant 4205 * @dev: network device 4206 * 4207 * Check if device is dormant. 4208 */ netif_dormant(const struct net_device * dev)4209 static inline bool netif_dormant(const struct net_device *dev) 4210 { 4211 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4212 } 4213 4214 4215 /** 4216 * netif_testing_on - mark device as under test. 4217 * @dev: network device 4218 * 4219 * Mark device as under test (as per RFC2863). 4220 * 4221 * The testing state indicates that some test(s) must be performed on 4222 * the interface. After completion, of the test, the interface state 4223 * will change to up, dormant, or down, as appropriate. 4224 */ netif_testing_on(struct net_device * dev)4225 static inline void netif_testing_on(struct net_device *dev) 4226 { 4227 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4228 linkwatch_fire_event(dev); 4229 } 4230 4231 /** 4232 * netif_testing_off - set device as not under test. 4233 * @dev: network device 4234 * 4235 * Device is not in testing state. 4236 */ netif_testing_off(struct net_device * dev)4237 static inline void netif_testing_off(struct net_device *dev) 4238 { 4239 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4240 linkwatch_fire_event(dev); 4241 } 4242 4243 /** 4244 * netif_testing - test if device is under test 4245 * @dev: network device 4246 * 4247 * Check if device is under test 4248 */ netif_testing(const struct net_device * dev)4249 static inline bool netif_testing(const struct net_device *dev) 4250 { 4251 return test_bit(__LINK_STATE_TESTING, &dev->state); 4252 } 4253 4254 4255 /** 4256 * netif_oper_up - test if device is operational 4257 * @dev: network device 4258 * 4259 * Check if carrier is operational 4260 */ netif_oper_up(const struct net_device * dev)4261 static inline bool netif_oper_up(const struct net_device *dev) 4262 { 4263 return (dev->operstate == IF_OPER_UP || 4264 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4265 } 4266 4267 /** 4268 * netif_device_present - is device available or removed 4269 * @dev: network device 4270 * 4271 * Check if device has not been removed from system. 4272 */ netif_device_present(const struct net_device * dev)4273 static inline bool netif_device_present(const struct net_device *dev) 4274 { 4275 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4276 } 4277 4278 void netif_device_detach(struct net_device *dev); 4279 4280 void netif_device_attach(struct net_device *dev); 4281 4282 /* 4283 * Network interface message level settings 4284 */ 4285 4286 enum { 4287 NETIF_MSG_DRV_BIT, 4288 NETIF_MSG_PROBE_BIT, 4289 NETIF_MSG_LINK_BIT, 4290 NETIF_MSG_TIMER_BIT, 4291 NETIF_MSG_IFDOWN_BIT, 4292 NETIF_MSG_IFUP_BIT, 4293 NETIF_MSG_RX_ERR_BIT, 4294 NETIF_MSG_TX_ERR_BIT, 4295 NETIF_MSG_TX_QUEUED_BIT, 4296 NETIF_MSG_INTR_BIT, 4297 NETIF_MSG_TX_DONE_BIT, 4298 NETIF_MSG_RX_STATUS_BIT, 4299 NETIF_MSG_PKTDATA_BIT, 4300 NETIF_MSG_HW_BIT, 4301 NETIF_MSG_WOL_BIT, 4302 4303 /* When you add a new bit above, update netif_msg_class_names array 4304 * in net/ethtool/common.c 4305 */ 4306 NETIF_MSG_CLASS_COUNT, 4307 }; 4308 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4309 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4310 4311 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4312 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4313 4314 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4315 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4316 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4317 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4318 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4319 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4320 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4321 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4322 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4323 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4324 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4325 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4326 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4327 #define NETIF_MSG_HW __NETIF_MSG(HW) 4328 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4329 4330 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4331 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4332 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4333 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4334 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4335 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4336 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4337 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4338 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4339 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4340 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4341 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4342 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4343 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4344 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4345 netif_msg_init(int debug_value,int default_msg_enable_bits)4346 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4347 { 4348 /* use default */ 4349 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4350 return default_msg_enable_bits; 4351 if (debug_value == 0) /* no output */ 4352 return 0; 4353 /* set low N bits */ 4354 return (1U << debug_value) - 1; 4355 } 4356 __netif_tx_lock(struct netdev_queue * txq,int cpu)4357 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4358 { 4359 spin_lock(&txq->_xmit_lock); 4360 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4361 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4362 } 4363 __netif_tx_acquire(struct netdev_queue * txq)4364 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4365 { 4366 __acquire(&txq->_xmit_lock); 4367 return true; 4368 } 4369 __netif_tx_release(struct netdev_queue * txq)4370 static inline void __netif_tx_release(struct netdev_queue *txq) 4371 { 4372 __release(&txq->_xmit_lock); 4373 } 4374 __netif_tx_lock_bh(struct netdev_queue * txq)4375 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4376 { 4377 spin_lock_bh(&txq->_xmit_lock); 4378 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4379 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4380 } 4381 __netif_tx_trylock(struct netdev_queue * txq)4382 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4383 { 4384 bool ok = spin_trylock(&txq->_xmit_lock); 4385 4386 if (likely(ok)) { 4387 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4388 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4389 } 4390 return ok; 4391 } 4392 __netif_tx_unlock(struct netdev_queue * txq)4393 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4394 { 4395 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4396 WRITE_ONCE(txq->xmit_lock_owner, -1); 4397 spin_unlock(&txq->_xmit_lock); 4398 } 4399 __netif_tx_unlock_bh(struct netdev_queue * txq)4400 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4401 { 4402 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4403 WRITE_ONCE(txq->xmit_lock_owner, -1); 4404 spin_unlock_bh(&txq->_xmit_lock); 4405 } 4406 4407 /* 4408 * txq->trans_start can be read locklessly from dev_watchdog() 4409 */ txq_trans_update(struct netdev_queue * txq)4410 static inline void txq_trans_update(struct netdev_queue *txq) 4411 { 4412 if (txq->xmit_lock_owner != -1) 4413 WRITE_ONCE(txq->trans_start, jiffies); 4414 } 4415 txq_trans_cond_update(struct netdev_queue * txq)4416 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4417 { 4418 unsigned long now = jiffies; 4419 4420 if (READ_ONCE(txq->trans_start) != now) 4421 WRITE_ONCE(txq->trans_start, now); 4422 } 4423 4424 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ netif_trans_update(struct net_device * dev)4425 static inline void netif_trans_update(struct net_device *dev) 4426 { 4427 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4428 4429 txq_trans_cond_update(txq); 4430 } 4431 4432 /** 4433 * netif_tx_lock - grab network device transmit lock 4434 * @dev: network device 4435 * 4436 * Get network device transmit lock 4437 */ 4438 void netif_tx_lock(struct net_device *dev); 4439 netif_tx_lock_bh(struct net_device * dev)4440 static inline void netif_tx_lock_bh(struct net_device *dev) 4441 { 4442 local_bh_disable(); 4443 netif_tx_lock(dev); 4444 } 4445 4446 void netif_tx_unlock(struct net_device *dev); 4447 netif_tx_unlock_bh(struct net_device * dev)4448 static inline void netif_tx_unlock_bh(struct net_device *dev) 4449 { 4450 netif_tx_unlock(dev); 4451 local_bh_enable(); 4452 } 4453 4454 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4455 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4456 __netif_tx_lock(txq, cpu); \ 4457 } else { \ 4458 __netif_tx_acquire(txq); \ 4459 } \ 4460 } 4461 4462 #define HARD_TX_TRYLOCK(dev, txq) \ 4463 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4464 __netif_tx_trylock(txq) : \ 4465 __netif_tx_acquire(txq)) 4466 4467 #define HARD_TX_UNLOCK(dev, txq) { \ 4468 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4469 __netif_tx_unlock(txq); \ 4470 } else { \ 4471 __netif_tx_release(txq); \ 4472 } \ 4473 } 4474 netif_tx_disable(struct net_device * dev)4475 static inline void netif_tx_disable(struct net_device *dev) 4476 { 4477 unsigned int i; 4478 int cpu; 4479 4480 local_bh_disable(); 4481 cpu = smp_processor_id(); 4482 spin_lock(&dev->tx_global_lock); 4483 for (i = 0; i < dev->num_tx_queues; i++) { 4484 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4485 4486 __netif_tx_lock(txq, cpu); 4487 netif_tx_stop_queue(txq); 4488 __netif_tx_unlock(txq); 4489 } 4490 spin_unlock(&dev->tx_global_lock); 4491 local_bh_enable(); 4492 } 4493 netif_addr_lock(struct net_device * dev)4494 static inline void netif_addr_lock(struct net_device *dev) 4495 { 4496 unsigned char nest_level = 0; 4497 4498 #ifdef CONFIG_LOCKDEP 4499 nest_level = dev->nested_level; 4500 #endif 4501 spin_lock_nested(&dev->addr_list_lock, nest_level); 4502 } 4503 netif_addr_lock_bh(struct net_device * dev)4504 static inline void netif_addr_lock_bh(struct net_device *dev) 4505 { 4506 unsigned char nest_level = 0; 4507 4508 #ifdef CONFIG_LOCKDEP 4509 nest_level = dev->nested_level; 4510 #endif 4511 local_bh_disable(); 4512 spin_lock_nested(&dev->addr_list_lock, nest_level); 4513 } 4514 netif_addr_unlock(struct net_device * dev)4515 static inline void netif_addr_unlock(struct net_device *dev) 4516 { 4517 spin_unlock(&dev->addr_list_lock); 4518 } 4519 netif_addr_unlock_bh(struct net_device * dev)4520 static inline void netif_addr_unlock_bh(struct net_device *dev) 4521 { 4522 spin_unlock_bh(&dev->addr_list_lock); 4523 } 4524 4525 /* 4526 * dev_addrs walker. Should be used only for read access. Call with 4527 * rcu_read_lock held. 4528 */ 4529 #define for_each_dev_addr(dev, ha) \ 4530 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4531 4532 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4533 4534 void ether_setup(struct net_device *dev); 4535 4536 /* Support for loadable net-drivers */ 4537 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4538 unsigned char name_assign_type, 4539 void (*setup)(struct net_device *), 4540 unsigned int txqs, unsigned int rxqs); 4541 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4542 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4543 4544 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4545 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4546 count) 4547 4548 int register_netdev(struct net_device *dev); 4549 void unregister_netdev(struct net_device *dev); 4550 4551 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4552 4553 /* General hardware address lists handling functions */ 4554 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4555 struct netdev_hw_addr_list *from_list, int addr_len); 4556 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4557 struct netdev_hw_addr_list *from_list, int addr_len); 4558 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4559 struct net_device *dev, 4560 int (*sync)(struct net_device *, const unsigned char *), 4561 int (*unsync)(struct net_device *, 4562 const unsigned char *)); 4563 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4564 struct net_device *dev, 4565 int (*sync)(struct net_device *, 4566 const unsigned char *, int), 4567 int (*unsync)(struct net_device *, 4568 const unsigned char *, int)); 4569 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4570 struct net_device *dev, 4571 int (*unsync)(struct net_device *, 4572 const unsigned char *, int)); 4573 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4574 struct net_device *dev, 4575 int (*unsync)(struct net_device *, 4576 const unsigned char *)); 4577 void __hw_addr_init(struct netdev_hw_addr_list *list); 4578 4579 /* Functions used for device addresses handling */ 4580 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4581 const void *addr, size_t len); 4582 4583 static inline void __dev_addr_set(struct net_device * dev,const void * addr,size_t len)4584 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4585 { 4586 dev_addr_mod(dev, 0, addr, len); 4587 } 4588 dev_addr_set(struct net_device * dev,const u8 * addr)4589 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4590 { 4591 __dev_addr_set(dev, addr, dev->addr_len); 4592 } 4593 4594 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4595 unsigned char addr_type); 4596 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4597 unsigned char addr_type); 4598 4599 /* Functions used for unicast addresses handling */ 4600 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4601 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4602 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4603 int dev_uc_sync(struct net_device *to, struct net_device *from); 4604 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4605 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4606 void dev_uc_flush(struct net_device *dev); 4607 void dev_uc_init(struct net_device *dev); 4608 4609 /** 4610 * __dev_uc_sync - Synchonize device's unicast list 4611 * @dev: device to sync 4612 * @sync: function to call if address should be added 4613 * @unsync: function to call if address should be removed 4614 * 4615 * Add newly added addresses to the interface, and release 4616 * addresses that have been deleted. 4617 */ __dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4618 static inline int __dev_uc_sync(struct net_device *dev, 4619 int (*sync)(struct net_device *, 4620 const unsigned char *), 4621 int (*unsync)(struct net_device *, 4622 const unsigned char *)) 4623 { 4624 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4625 } 4626 4627 /** 4628 * __dev_uc_unsync - Remove synchronized addresses from device 4629 * @dev: device to sync 4630 * @unsync: function to call if address should be removed 4631 * 4632 * Remove all addresses that were added to the device by dev_uc_sync(). 4633 */ __dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4634 static inline void __dev_uc_unsync(struct net_device *dev, 4635 int (*unsync)(struct net_device *, 4636 const unsigned char *)) 4637 { 4638 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4639 } 4640 4641 /* Functions used for multicast addresses handling */ 4642 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4643 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4644 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4645 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4646 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4647 int dev_mc_sync(struct net_device *to, struct net_device *from); 4648 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4649 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4650 void dev_mc_flush(struct net_device *dev); 4651 void dev_mc_init(struct net_device *dev); 4652 4653 /** 4654 * __dev_mc_sync - Synchonize device's multicast list 4655 * @dev: device to sync 4656 * @sync: function to call if address should be added 4657 * @unsync: function to call if address should be removed 4658 * 4659 * Add newly added addresses to the interface, and release 4660 * addresses that have been deleted. 4661 */ __dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4662 static inline int __dev_mc_sync(struct net_device *dev, 4663 int (*sync)(struct net_device *, 4664 const unsigned char *), 4665 int (*unsync)(struct net_device *, 4666 const unsigned char *)) 4667 { 4668 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4669 } 4670 4671 /** 4672 * __dev_mc_unsync - Remove synchronized addresses from device 4673 * @dev: device to sync 4674 * @unsync: function to call if address should be removed 4675 * 4676 * Remove all addresses that were added to the device by dev_mc_sync(). 4677 */ __dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4678 static inline void __dev_mc_unsync(struct net_device *dev, 4679 int (*unsync)(struct net_device *, 4680 const unsigned char *)) 4681 { 4682 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4683 } 4684 4685 /* Functions used for secondary unicast and multicast support */ 4686 void dev_set_rx_mode(struct net_device *dev); 4687 int dev_set_promiscuity(struct net_device *dev, int inc); 4688 int dev_set_allmulti(struct net_device *dev, int inc); 4689 void netdev_state_change(struct net_device *dev); 4690 void __netdev_notify_peers(struct net_device *dev); 4691 void netdev_notify_peers(struct net_device *dev); 4692 void netdev_features_change(struct net_device *dev); 4693 /* Load a device via the kmod */ 4694 void dev_load(struct net *net, const char *name); 4695 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4696 struct rtnl_link_stats64 *storage); 4697 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4698 const struct net_device_stats *netdev_stats); 4699 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4700 const struct pcpu_sw_netstats __percpu *netstats); 4701 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4702 4703 extern int netdev_max_backlog; 4704 extern int dev_rx_weight; 4705 extern int dev_tx_weight; 4706 extern int gro_normal_batch; 4707 4708 enum { 4709 NESTED_SYNC_IMM_BIT, 4710 NESTED_SYNC_TODO_BIT, 4711 }; 4712 4713 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4714 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4715 4716 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4717 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4718 4719 struct netdev_nested_priv { 4720 unsigned char flags; 4721 void *data; 4722 }; 4723 4724 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4725 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4726 struct list_head **iter); 4727 4728 /* iterate through upper list, must be called under RCU read lock */ 4729 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4730 for (iter = &(dev)->adj_list.upper, \ 4731 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4732 updev; \ 4733 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4734 4735 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4736 int (*fn)(struct net_device *upper_dev, 4737 struct netdev_nested_priv *priv), 4738 struct netdev_nested_priv *priv); 4739 4740 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4741 struct net_device *upper_dev); 4742 4743 bool netdev_has_any_upper_dev(struct net_device *dev); 4744 4745 void *netdev_lower_get_next_private(struct net_device *dev, 4746 struct list_head **iter); 4747 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4748 struct list_head **iter); 4749 4750 #define netdev_for_each_lower_private(dev, priv, iter) \ 4751 for (iter = (dev)->adj_list.lower.next, \ 4752 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4753 priv; \ 4754 priv = netdev_lower_get_next_private(dev, &(iter))) 4755 4756 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4757 for (iter = &(dev)->adj_list.lower, \ 4758 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4759 priv; \ 4760 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4761 4762 void *netdev_lower_get_next(struct net_device *dev, 4763 struct list_head **iter); 4764 4765 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4766 for (iter = (dev)->adj_list.lower.next, \ 4767 ldev = netdev_lower_get_next(dev, &(iter)); \ 4768 ldev; \ 4769 ldev = netdev_lower_get_next(dev, &(iter))) 4770 4771 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4772 struct list_head **iter); 4773 int netdev_walk_all_lower_dev(struct net_device *dev, 4774 int (*fn)(struct net_device *lower_dev, 4775 struct netdev_nested_priv *priv), 4776 struct netdev_nested_priv *priv); 4777 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4778 int (*fn)(struct net_device *lower_dev, 4779 struct netdev_nested_priv *priv), 4780 struct netdev_nested_priv *priv); 4781 4782 void *netdev_adjacent_get_private(struct list_head *adj_list); 4783 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4784 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4785 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4786 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4787 struct netlink_ext_ack *extack); 4788 int netdev_master_upper_dev_link(struct net_device *dev, 4789 struct net_device *upper_dev, 4790 void *upper_priv, void *upper_info, 4791 struct netlink_ext_ack *extack); 4792 void netdev_upper_dev_unlink(struct net_device *dev, 4793 struct net_device *upper_dev); 4794 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4795 struct net_device *new_dev, 4796 struct net_device *dev, 4797 struct netlink_ext_ack *extack); 4798 void netdev_adjacent_change_commit(struct net_device *old_dev, 4799 struct net_device *new_dev, 4800 struct net_device *dev); 4801 void netdev_adjacent_change_abort(struct net_device *old_dev, 4802 struct net_device *new_dev, 4803 struct net_device *dev); 4804 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4805 void *netdev_lower_dev_get_private(struct net_device *dev, 4806 struct net_device *lower_dev); 4807 void netdev_lower_state_changed(struct net_device *lower_dev, 4808 void *lower_state_info); 4809 4810 /* RSS keys are 40 or 52 bytes long */ 4811 #define NETDEV_RSS_KEY_LEN 52 4812 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4813 void netdev_rss_key_fill(void *buffer, size_t len); 4814 4815 int skb_checksum_help(struct sk_buff *skb); 4816 int skb_crc32c_csum_help(struct sk_buff *skb); 4817 int skb_csum_hwoffload_help(struct sk_buff *skb, 4818 const netdev_features_t features); 4819 4820 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4821 netdev_features_t features, bool tx_path); 4822 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb, 4823 netdev_features_t features, __be16 type); 4824 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4825 netdev_features_t features); 4826 4827 struct netdev_bonding_info { 4828 ifslave slave; 4829 ifbond master; 4830 }; 4831 4832 struct netdev_notifier_bonding_info { 4833 struct netdev_notifier_info info; /* must be first */ 4834 struct netdev_bonding_info bonding_info; 4835 }; 4836 4837 void netdev_bonding_info_change(struct net_device *dev, 4838 struct netdev_bonding_info *bonding_info); 4839 4840 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4841 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4842 #else ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4843 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4844 const void *data) 4845 { 4846 } 4847 #endif 4848 4849 static inline skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4850 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4851 { 4852 return __skb_gso_segment(skb, features, true); 4853 } 4854 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4855 can_checksum_protocol(netdev_features_t features,__be16 protocol)4856 static inline bool can_checksum_protocol(netdev_features_t features, 4857 __be16 protocol) 4858 { 4859 if (protocol == htons(ETH_P_FCOE)) 4860 return !!(features & NETIF_F_FCOE_CRC); 4861 4862 /* Assume this is an IP checksum (not SCTP CRC) */ 4863 4864 if (features & NETIF_F_HW_CSUM) { 4865 /* Can checksum everything */ 4866 return true; 4867 } 4868 4869 switch (protocol) { 4870 case htons(ETH_P_IP): 4871 return !!(features & NETIF_F_IP_CSUM); 4872 case htons(ETH_P_IPV6): 4873 return !!(features & NETIF_F_IPV6_CSUM); 4874 default: 4875 return false; 4876 } 4877 } 4878 4879 #ifdef CONFIG_BUG 4880 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4881 #else netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4882 static inline void netdev_rx_csum_fault(struct net_device *dev, 4883 struct sk_buff *skb) 4884 { 4885 } 4886 #endif 4887 /* rx skb timestamps */ 4888 void net_enable_timestamp(void); 4889 void net_disable_timestamp(void); 4890 netdev_get_tstamp(struct net_device * dev,const struct skb_shared_hwtstamps * hwtstamps,bool cycles)4891 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 4892 const struct skb_shared_hwtstamps *hwtstamps, 4893 bool cycles) 4894 { 4895 const struct net_device_ops *ops = dev->netdev_ops; 4896 4897 if (ops->ndo_get_tstamp) 4898 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 4899 4900 return hwtstamps->hwtstamp; 4901 } 4902 __netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4903 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4904 struct sk_buff *skb, struct net_device *dev, 4905 bool more) 4906 { 4907 __this_cpu_write(softnet_data.xmit.more, more); 4908 return ops->ndo_start_xmit(skb, dev); 4909 } 4910 netdev_xmit_more(void)4911 static inline bool netdev_xmit_more(void) 4912 { 4913 return __this_cpu_read(softnet_data.xmit.more); 4914 } 4915 netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4916 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4917 struct netdev_queue *txq, bool more) 4918 { 4919 const struct net_device_ops *ops = dev->netdev_ops; 4920 netdev_tx_t rc; 4921 4922 rc = __netdev_start_xmit(ops, skb, dev, more); 4923 if (rc == NETDEV_TX_OK) 4924 txq_trans_update(txq); 4925 4926 return rc; 4927 } 4928 4929 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4930 const void *ns); 4931 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4932 const void *ns); 4933 4934 extern const struct kobj_ns_type_operations net_ns_type_operations; 4935 4936 const char *netdev_drivername(const struct net_device *dev); 4937 netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4938 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4939 netdev_features_t f2) 4940 { 4941 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4942 if (f1 & NETIF_F_HW_CSUM) 4943 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4944 else 4945 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4946 } 4947 4948 return f1 & f2; 4949 } 4950 netdev_get_wanted_features(struct net_device * dev)4951 static inline netdev_features_t netdev_get_wanted_features( 4952 struct net_device *dev) 4953 { 4954 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4955 } 4956 netdev_features_t netdev_increment_features(netdev_features_t all, 4957 netdev_features_t one, netdev_features_t mask); 4958 4959 /* Allow TSO being used on stacked device : 4960 * Performing the GSO segmentation before last device 4961 * is a performance improvement. 4962 */ netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4963 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4964 netdev_features_t mask) 4965 { 4966 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4967 } 4968 4969 int __netdev_update_features(struct net_device *dev); 4970 void netdev_update_features(struct net_device *dev); 4971 void netdev_change_features(struct net_device *dev); 4972 4973 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4974 struct net_device *dev); 4975 4976 netdev_features_t passthru_features_check(struct sk_buff *skb, 4977 struct net_device *dev, 4978 netdev_features_t features); 4979 netdev_features_t netif_skb_features(struct sk_buff *skb); 4980 net_gso_ok(netdev_features_t features,int gso_type)4981 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4982 { 4983 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4984 4985 /* check flags correspondence */ 4986 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4987 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4988 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4989 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4990 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4991 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4992 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4993 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4994 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4995 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4996 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4997 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4998 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4999 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5000 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5001 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5002 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5003 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5004 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5005 5006 return (features & feature) == feature; 5007 } 5008 skb_gso_ok(struct sk_buff * skb,netdev_features_t features)5009 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5010 { 5011 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5012 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5013 } 5014 netif_needs_gso(struct sk_buff * skb,netdev_features_t features)5015 static inline bool netif_needs_gso(struct sk_buff *skb, 5016 netdev_features_t features) 5017 { 5018 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5019 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5020 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5021 } 5022 5023 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5024 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5025 void netif_inherit_tso_max(struct net_device *to, 5026 const struct net_device *from); 5027 skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)5028 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 5029 int pulled_hlen, u16 mac_offset, 5030 int mac_len) 5031 { 5032 skb->protocol = protocol; 5033 skb->encapsulation = 1; 5034 skb_push(skb, pulled_hlen); 5035 skb_reset_transport_header(skb); 5036 skb->mac_header = mac_offset; 5037 skb->network_header = skb->mac_header + mac_len; 5038 skb->mac_len = mac_len; 5039 } 5040 netif_is_macsec(const struct net_device * dev)5041 static inline bool netif_is_macsec(const struct net_device *dev) 5042 { 5043 return dev->priv_flags & IFF_MACSEC; 5044 } 5045 netif_is_macvlan(const struct net_device * dev)5046 static inline bool netif_is_macvlan(const struct net_device *dev) 5047 { 5048 return dev->priv_flags & IFF_MACVLAN; 5049 } 5050 netif_is_macvlan_port(const struct net_device * dev)5051 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5052 { 5053 return dev->priv_flags & IFF_MACVLAN_PORT; 5054 } 5055 netif_is_bond_master(const struct net_device * dev)5056 static inline bool netif_is_bond_master(const struct net_device *dev) 5057 { 5058 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5059 } 5060 netif_is_bond_slave(const struct net_device * dev)5061 static inline bool netif_is_bond_slave(const struct net_device *dev) 5062 { 5063 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5064 } 5065 netif_supports_nofcs(struct net_device * dev)5066 static inline bool netif_supports_nofcs(struct net_device *dev) 5067 { 5068 return dev->priv_flags & IFF_SUPP_NOFCS; 5069 } 5070 netif_has_l3_rx_handler(const struct net_device * dev)5071 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5072 { 5073 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5074 } 5075 netif_is_l3_master(const struct net_device * dev)5076 static inline bool netif_is_l3_master(const struct net_device *dev) 5077 { 5078 return dev->priv_flags & IFF_L3MDEV_MASTER; 5079 } 5080 netif_is_l3_slave(const struct net_device * dev)5081 static inline bool netif_is_l3_slave(const struct net_device *dev) 5082 { 5083 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5084 } 5085 dev_sdif(const struct net_device * dev)5086 static inline int dev_sdif(const struct net_device *dev) 5087 { 5088 #ifdef CONFIG_NET_L3_MASTER_DEV 5089 if (netif_is_l3_slave(dev)) 5090 return dev->ifindex; 5091 #endif 5092 return 0; 5093 } 5094 netif_is_bridge_master(const struct net_device * dev)5095 static inline bool netif_is_bridge_master(const struct net_device *dev) 5096 { 5097 return dev->priv_flags & IFF_EBRIDGE; 5098 } 5099 netif_is_bridge_port(const struct net_device * dev)5100 static inline bool netif_is_bridge_port(const struct net_device *dev) 5101 { 5102 return dev->priv_flags & IFF_BRIDGE_PORT; 5103 } 5104 netif_is_ovs_master(const struct net_device * dev)5105 static inline bool netif_is_ovs_master(const struct net_device *dev) 5106 { 5107 return dev->priv_flags & IFF_OPENVSWITCH; 5108 } 5109 netif_is_ovs_port(const struct net_device * dev)5110 static inline bool netif_is_ovs_port(const struct net_device *dev) 5111 { 5112 return dev->priv_flags & IFF_OVS_DATAPATH; 5113 } 5114 netif_is_any_bridge_port(const struct net_device * dev)5115 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5116 { 5117 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5118 } 5119 netif_is_team_master(const struct net_device * dev)5120 static inline bool netif_is_team_master(const struct net_device *dev) 5121 { 5122 return dev->priv_flags & IFF_TEAM; 5123 } 5124 netif_is_team_port(const struct net_device * dev)5125 static inline bool netif_is_team_port(const struct net_device *dev) 5126 { 5127 return dev->priv_flags & IFF_TEAM_PORT; 5128 } 5129 netif_is_lag_master(const struct net_device * dev)5130 static inline bool netif_is_lag_master(const struct net_device *dev) 5131 { 5132 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5133 } 5134 netif_is_lag_port(const struct net_device * dev)5135 static inline bool netif_is_lag_port(const struct net_device *dev) 5136 { 5137 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5138 } 5139 netif_is_rxfh_configured(const struct net_device * dev)5140 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5141 { 5142 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5143 } 5144 netif_is_failover(const struct net_device * dev)5145 static inline bool netif_is_failover(const struct net_device *dev) 5146 { 5147 return dev->priv_flags & IFF_FAILOVER; 5148 } 5149 netif_is_failover_slave(const struct net_device * dev)5150 static inline bool netif_is_failover_slave(const struct net_device *dev) 5151 { 5152 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5153 } 5154 5155 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ netif_keep_dst(struct net_device * dev)5156 static inline void netif_keep_dst(struct net_device *dev) 5157 { 5158 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5159 } 5160 5161 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ netif_reduces_vlan_mtu(struct net_device * dev)5162 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5163 { 5164 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5165 return netif_is_macsec(dev); 5166 } 5167 5168 extern struct pernet_operations __net_initdata loopback_net_ops; 5169 5170 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5171 5172 /* netdev_printk helpers, similar to dev_printk */ 5173 netdev_name(const struct net_device * dev)5174 static inline const char *netdev_name(const struct net_device *dev) 5175 { 5176 if (!dev->name[0] || strchr(dev->name, '%')) 5177 return "(unnamed net_device)"; 5178 return dev->name; 5179 } 5180 netdev_unregistering(const struct net_device * dev)5181 static inline bool netdev_unregistering(const struct net_device *dev) 5182 { 5183 return dev->reg_state == NETREG_UNREGISTERING; 5184 } 5185 netdev_reg_state(const struct net_device * dev)5186 static inline const char *netdev_reg_state(const struct net_device *dev) 5187 { 5188 switch (dev->reg_state) { 5189 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5190 case NETREG_REGISTERED: return ""; 5191 case NETREG_UNREGISTERING: return " (unregistering)"; 5192 case NETREG_UNREGISTERED: return " (unregistered)"; 5193 case NETREG_RELEASED: return " (released)"; 5194 case NETREG_DUMMY: return " (dummy)"; 5195 } 5196 5197 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5198 return " (unknown)"; 5199 } 5200 5201 #define MODULE_ALIAS_NETDEV(device) \ 5202 MODULE_ALIAS("netdev-" device) 5203 5204 /* 5205 * netdev_WARN() acts like dev_printk(), but with the key difference 5206 * of using a WARN/WARN_ON to get the message out, including the 5207 * file/line information and a backtrace. 5208 */ 5209 #define netdev_WARN(dev, format, args...) \ 5210 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5211 netdev_reg_state(dev), ##args) 5212 5213 #define netdev_WARN_ONCE(dev, format, args...) \ 5214 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5215 netdev_reg_state(dev), ##args) 5216 5217 /* 5218 * The list of packet types we will receive (as opposed to discard) 5219 * and the routines to invoke. 5220 * 5221 * Why 16. Because with 16 the only overlap we get on a hash of the 5222 * low nibble of the protocol value is RARP/SNAP/X.25. 5223 * 5224 * 0800 IP 5225 * 0001 802.3 5226 * 0002 AX.25 5227 * 0004 802.2 5228 * 8035 RARP 5229 * 0005 SNAP 5230 * 0805 X.25 5231 * 0806 ARP 5232 * 8137 IPX 5233 * 0009 Localtalk 5234 * 86DD IPv6 5235 */ 5236 #define PTYPE_HASH_SIZE (16) 5237 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5238 5239 extern struct list_head ptype_all __read_mostly; 5240 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5241 5242 extern struct net_device *blackhole_netdev; 5243 5244 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5245 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5246 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5247 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5248 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5249 5250 #endif /* _LINUX_NETDEVICE_H */ 5251