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