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