1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 15 * Bjorn Ekwall. <bj0rn@blox.se> 16 * Pekka Riikonen <priikone@poseidon.pspt.fi> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/workqueue.h> 39 #include <linux/dynamic_queue_limits.h> 40 41 #include <linux/ethtool.h> 42 #include <net/net_namespace.h> 43 #ifdef CONFIG_DCB 44 #include <net/dcbnl.h> 45 #endif 46 #include <net/netprio_cgroup.h> 47 #include <net/xdp.h> 48 49 #include <linux/netdev_features.h> 50 #include <linux/neighbour.h> 51 #include <uapi/linux/netdevice.h> 52 #include <uapi/linux/if_bonding.h> 53 #include <uapi/linux/pkt_cls.h> 54 #include <linux/hashtable.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 struct dsa_port; 60 61 struct sfp_bus; 62 /* 802.11 specific */ 63 struct wireless_dev; 64 /* 802.15.4 specific */ 65 struct wpan_dev; 66 struct mpls_dev; 67 /* UDP Tunnel offloads */ 68 struct udp_tunnel_info; 69 struct bpf_prog; 70 struct xdp_buff; 71 72 void netdev_set_default_ethtool_ops(struct net_device *dev, 73 const struct ethtool_ops *ops); 74 75 /* Backlog congestion levels */ 76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 77 #define NET_RX_DROP 1 /* packet dropped */ 78 79 /* 80 * Transmit return codes: transmit return codes originate from three different 81 * namespaces: 82 * 83 * - qdisc return codes 84 * - driver transmit return codes 85 * - errno values 86 * 87 * Drivers are allowed to return any one of those in their hard_start_xmit() 88 * function. Real network devices commonly used with qdiscs should only return 89 * the driver transmit return codes though - when qdiscs are used, the actual 90 * transmission happens asynchronously, so the value is not propagated to 91 * higher layers. Virtual network devices transmit synchronously; in this case 92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 93 * others are propagated to higher layers. 94 */ 95 96 /* qdisc ->enqueue() return codes. */ 97 #define NET_XMIT_SUCCESS 0x00 98 #define NET_XMIT_DROP 0x01 /* skb dropped */ 99 #define NET_XMIT_CN 0x02 /* congestion notification */ 100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 101 102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 103 * indicates that the device will soon be dropping packets, or already drops 104 * some packets of the same priority; prompting us to send less aggressively. */ 105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 107 108 /* Driver transmit return codes */ 109 #define NETDEV_TX_MASK 0xf0 110 111 enum netdev_tx { 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 113 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 115 }; 116 typedef enum netdev_tx netdev_tx_t; 117 118 /* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122 static inline bool dev_xmit_complete(int rc) 123 { 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134 } 135 136 /* 137 * Compute the worst-case header length according to the protocols 138 * used. 139 */ 140 141 #if defined(CONFIG_HYPERV_NET) 142 # define LL_MAX_HEADER 128 143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 144 # if defined(CONFIG_MAC80211_MESH) 145 # define LL_MAX_HEADER 128 146 # else 147 # define LL_MAX_HEADER 96 148 # endif 149 #else 150 # define LL_MAX_HEADER 32 151 #endif 152 153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 155 #define MAX_HEADER LL_MAX_HEADER 156 #else 157 #define MAX_HEADER (LL_MAX_HEADER + 48) 158 #endif 159 160 /* 161 * Old network device statistics. Fields are native words 162 * (unsigned long) so they can be read and written atomically. 163 */ 164 165 struct net_device_stats { 166 unsigned long rx_packets; 167 unsigned long tx_packets; 168 unsigned long rx_bytes; 169 unsigned long tx_bytes; 170 unsigned long rx_errors; 171 unsigned long tx_errors; 172 unsigned long rx_dropped; 173 unsigned long tx_dropped; 174 unsigned long multicast; 175 unsigned long collisions; 176 unsigned long rx_length_errors; 177 unsigned long rx_over_errors; 178 unsigned long rx_crc_errors; 179 unsigned long rx_frame_errors; 180 unsigned long rx_fifo_errors; 181 unsigned long rx_missed_errors; 182 unsigned long tx_aborted_errors; 183 unsigned long tx_carrier_errors; 184 unsigned long tx_fifo_errors; 185 unsigned long tx_heartbeat_errors; 186 unsigned long tx_window_errors; 187 unsigned long rx_compressed; 188 unsigned long tx_compressed; 189 }; 190 191 192 #include <linux/cache.h> 193 #include <linux/skbuff.h> 194 195 #ifdef CONFIG_RPS 196 #include <linux/static_key.h> 197 extern struct static_key_false rps_needed; 198 extern struct static_key_false rfs_needed; 199 #endif 200 201 struct neighbour; 202 struct neigh_parms; 203 struct sk_buff; 204 205 struct netdev_hw_addr { 206 struct list_head list; 207 unsigned char addr[MAX_ADDR_LEN]; 208 unsigned char type; 209 #define NETDEV_HW_ADDR_T_LAN 1 210 #define NETDEV_HW_ADDR_T_SAN 2 211 #define NETDEV_HW_ADDR_T_SLAVE 3 212 #define NETDEV_HW_ADDR_T_UNICAST 4 213 #define NETDEV_HW_ADDR_T_MULTICAST 5 214 bool global_use; 215 int sync_cnt; 216 int refcount; 217 int synced; 218 struct rcu_head rcu_head; 219 }; 220 221 struct netdev_hw_addr_list { 222 struct list_head list; 223 int count; 224 }; 225 226 #define netdev_hw_addr_list_count(l) ((l)->count) 227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 228 #define netdev_hw_addr_list_for_each(ha, l) \ 229 list_for_each_entry(ha, &(l)->list, list) 230 231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 233 #define netdev_for_each_uc_addr(ha, dev) \ 234 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 235 236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 238 #define netdev_for_each_mc_addr(ha, dev) \ 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 240 241 struct hh_cache { 242 unsigned int hh_len; 243 seqlock_t hh_lock; 244 245 /* cached hardware header; allow for machine alignment needs. */ 246 #define HH_DATA_MOD 16 247 #define HH_DATA_OFF(__len) \ 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 249 #define HH_DATA_ALIGN(__len) \ 250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 252 }; 253 254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 255 * Alternative is: 256 * dev->hard_header_len ? (dev->hard_header_len + 257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 258 * 259 * We could use other alignment values, but we must maintain the 260 * relationship HH alignment <= LL alignment. 261 */ 262 #define LL_RESERVED_SPACE(dev) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 266 267 struct header_ops { 268 int (*create) (struct sk_buff *skb, struct net_device *dev, 269 unsigned short type, const void *daddr, 270 const void *saddr, unsigned int len); 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 273 void (*cache_update)(struct hh_cache *hh, 274 const struct net_device *dev, 275 const unsigned char *haddr); 276 bool (*validate)(const char *ll_header, unsigned int len); 277 __be16 (*parse_protocol)(const struct sk_buff *skb); 278 }; 279 280 /* These flag bits are private to the generic network queueing 281 * layer; they may not be explicitly referenced by any other 282 * code. 283 */ 284 285 enum netdev_state_t { 286 __LINK_STATE_START, 287 __LINK_STATE_PRESENT, 288 __LINK_STATE_NOCARRIER, 289 __LINK_STATE_LINKWATCH_PENDING, 290 __LINK_STATE_DORMANT, 291 }; 292 293 294 /* 295 * This structure holds boot-time configured netdevice settings. They 296 * are then used in the device probing. 297 */ 298 struct netdev_boot_setup { 299 char name[IFNAMSIZ]; 300 struct ifmap map; 301 }; 302 #define NETDEV_BOOT_SETUP_MAX 8 303 304 int __init netdev_boot_setup(char *str); 305 306 struct gro_list { 307 struct list_head list; 308 int count; 309 }; 310 311 /* 312 * size of gro hash buckets, must less than bit number of 313 * napi_struct::gro_bitmask 314 */ 315 #define GRO_HASH_BUCKETS 8 316 317 /* 318 * Structure for NAPI scheduling similar to tasklet but with weighting 319 */ 320 struct napi_struct { 321 /* The poll_list must only be managed by the entity which 322 * changes the state of the NAPI_STATE_SCHED bit. This means 323 * whoever atomically sets that bit can add this napi_struct 324 * to the per-CPU poll_list, and whoever clears that bit 325 * can remove from the list right before clearing the bit. 326 */ 327 struct list_head poll_list; 328 329 unsigned long state; 330 int weight; 331 unsigned long gro_bitmask; 332 int (*poll)(struct napi_struct *, int); 333 #ifdef CONFIG_NETPOLL 334 int poll_owner; 335 #endif 336 struct net_device *dev; 337 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 338 struct sk_buff *skb; 339 struct hrtimer timer; 340 struct list_head dev_list; 341 struct hlist_node napi_hash_node; 342 unsigned int napi_id; 343 }; 344 345 enum { 346 NAPI_STATE_SCHED, /* Poll is scheduled */ 347 NAPI_STATE_MISSED, /* reschedule a napi */ 348 NAPI_STATE_DISABLE, /* Disable pending */ 349 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 350 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 351 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 352 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 353 }; 354 355 enum { 356 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 357 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 358 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 359 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 360 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 361 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 362 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 363 }; 364 365 enum gro_result { 366 GRO_MERGED, 367 GRO_MERGED_FREE, 368 GRO_HELD, 369 GRO_NORMAL, 370 GRO_DROP, 371 GRO_CONSUMED, 372 }; 373 typedef enum gro_result gro_result_t; 374 375 /* 376 * enum rx_handler_result - Possible return values for rx_handlers. 377 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 378 * further. 379 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 380 * case skb->dev was changed by rx_handler. 381 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 382 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 383 * 384 * rx_handlers are functions called from inside __netif_receive_skb(), to do 385 * special processing of the skb, prior to delivery to protocol handlers. 386 * 387 * Currently, a net_device can only have a single rx_handler registered. Trying 388 * to register a second rx_handler will return -EBUSY. 389 * 390 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 391 * To unregister a rx_handler on a net_device, use 392 * netdev_rx_handler_unregister(). 393 * 394 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 395 * do with the skb. 396 * 397 * If the rx_handler consumed the skb in some way, it should return 398 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 399 * the skb to be delivered in some other way. 400 * 401 * If the rx_handler changed skb->dev, to divert the skb to another 402 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 403 * new device will be called if it exists. 404 * 405 * If the rx_handler decides the skb should be ignored, it should return 406 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 407 * are registered on exact device (ptype->dev == skb->dev). 408 * 409 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 410 * delivered, it should return RX_HANDLER_PASS. 411 * 412 * A device without a registered rx_handler will behave as if rx_handler 413 * returned RX_HANDLER_PASS. 414 */ 415 416 enum rx_handler_result { 417 RX_HANDLER_CONSUMED, 418 RX_HANDLER_ANOTHER, 419 RX_HANDLER_EXACT, 420 RX_HANDLER_PASS, 421 }; 422 typedef enum rx_handler_result rx_handler_result_t; 423 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 424 425 void __napi_schedule(struct napi_struct *n); 426 void __napi_schedule_irqoff(struct napi_struct *n); 427 428 static inline bool napi_disable_pending(struct napi_struct *n) 429 { 430 return test_bit(NAPI_STATE_DISABLE, &n->state); 431 } 432 433 bool napi_schedule_prep(struct napi_struct *n); 434 435 /** 436 * napi_schedule - schedule NAPI poll 437 * @n: NAPI context 438 * 439 * Schedule NAPI poll routine to be called if it is not already 440 * running. 441 */ 442 static inline void napi_schedule(struct napi_struct *n) 443 { 444 if (napi_schedule_prep(n)) 445 __napi_schedule(n); 446 } 447 448 /** 449 * napi_schedule_irqoff - schedule NAPI poll 450 * @n: NAPI context 451 * 452 * Variant of napi_schedule(), assuming hard irqs are masked. 453 */ 454 static inline void napi_schedule_irqoff(struct napi_struct *n) 455 { 456 if (napi_schedule_prep(n)) 457 __napi_schedule_irqoff(n); 458 } 459 460 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 461 static inline bool napi_reschedule(struct napi_struct *napi) 462 { 463 if (napi_schedule_prep(napi)) { 464 __napi_schedule(napi); 465 return true; 466 } 467 return false; 468 } 469 470 bool napi_complete_done(struct napi_struct *n, int work_done); 471 /** 472 * napi_complete - NAPI processing complete 473 * @n: NAPI context 474 * 475 * Mark NAPI processing as complete. 476 * Consider using napi_complete_done() instead. 477 * Return false if device should avoid rearming interrupts. 478 */ 479 static inline bool napi_complete(struct napi_struct *n) 480 { 481 return napi_complete_done(n, 0); 482 } 483 484 /** 485 * napi_hash_del - remove a NAPI from global table 486 * @napi: NAPI context 487 * 488 * Warning: caller must observe RCU grace period 489 * before freeing memory containing @napi, if 490 * this function returns true. 491 * Note: core networking stack automatically calls it 492 * from netif_napi_del(). 493 * Drivers might want to call this helper to combine all 494 * the needed RCU grace periods into a single one. 495 */ 496 bool napi_hash_del(struct napi_struct *napi); 497 498 /** 499 * napi_disable - prevent NAPI from scheduling 500 * @n: NAPI context 501 * 502 * Stop NAPI from being scheduled on this context. 503 * Waits till any outstanding processing completes. 504 */ 505 void napi_disable(struct napi_struct *n); 506 507 /** 508 * napi_enable - enable NAPI scheduling 509 * @n: NAPI context 510 * 511 * Resume NAPI from being scheduled on this context. 512 * Must be paired with napi_disable. 513 */ 514 static inline void napi_enable(struct napi_struct *n) 515 { 516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 517 smp_mb__before_atomic(); 518 clear_bit(NAPI_STATE_SCHED, &n->state); 519 clear_bit(NAPI_STATE_NPSVC, &n->state); 520 } 521 522 /** 523 * napi_synchronize - wait until NAPI is not running 524 * @n: NAPI context 525 * 526 * Wait until NAPI is done being scheduled on this context. 527 * Waits till any outstanding processing completes but 528 * does not disable future activations. 529 */ 530 static inline void napi_synchronize(const struct napi_struct *n) 531 { 532 if (IS_ENABLED(CONFIG_SMP)) 533 while (test_bit(NAPI_STATE_SCHED, &n->state)) 534 msleep(1); 535 else 536 barrier(); 537 } 538 539 /** 540 * napi_if_scheduled_mark_missed - if napi is running, set the 541 * NAPIF_STATE_MISSED 542 * @n: NAPI context 543 * 544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 545 * NAPI is scheduled. 546 **/ 547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 548 { 549 unsigned long val, new; 550 551 do { 552 val = READ_ONCE(n->state); 553 if (val & NAPIF_STATE_DISABLE) 554 return true; 555 556 if (!(val & NAPIF_STATE_SCHED)) 557 return false; 558 559 new = val | NAPIF_STATE_MISSED; 560 } while (cmpxchg(&n->state, val, new) != val); 561 562 return true; 563 } 564 565 enum netdev_queue_state_t { 566 __QUEUE_STATE_DRV_XOFF, 567 __QUEUE_STATE_STACK_XOFF, 568 __QUEUE_STATE_FROZEN, 569 }; 570 571 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 572 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 573 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 574 575 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 579 QUEUE_STATE_FROZEN) 580 581 /* 582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 583 * netif_tx_* functions below are used to manipulate this flag. The 584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 585 * queue independently. The netif_xmit_*stopped functions below are called 586 * to check if the queue has been stopped by the driver or stack (either 587 * of the XOFF bits are set in the state). Drivers should not need to call 588 * netif_xmit*stopped functions, they should only be using netif_tx_*. 589 */ 590 591 struct netdev_queue { 592 /* 593 * read-mostly part 594 */ 595 struct net_device *dev; 596 struct Qdisc __rcu *qdisc; 597 struct Qdisc *qdisc_sleeping; 598 #ifdef CONFIG_SYSFS 599 struct kobject kobj; 600 #endif 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 602 int numa_node; 603 #endif 604 unsigned long tx_maxrate; 605 /* 606 * Number of TX timeouts for this queue 607 * (/sys/class/net/DEV/Q/trans_timeout) 608 */ 609 unsigned long trans_timeout; 610 611 /* Subordinate device that the queue has been assigned to */ 612 struct net_device *sb_dev; 613 #ifdef CONFIG_XDP_SOCKETS 614 struct xdp_umem *umem; 615 #endif 616 /* 617 * write-mostly part 618 */ 619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 620 int xmit_lock_owner; 621 /* 622 * Time (in jiffies) of last Tx 623 */ 624 unsigned long trans_start; 625 626 unsigned long state; 627 628 #ifdef CONFIG_BQL 629 struct dql dql; 630 #endif 631 } ____cacheline_aligned_in_smp; 632 633 extern int sysctl_fb_tunnels_only_for_init_net; 634 extern int sysctl_devconf_inherit_init_net; 635 636 static inline bool net_has_fallback_tunnels(const struct net *net) 637 { 638 return net == &init_net || 639 !IS_ENABLED(CONFIG_SYSCTL) || 640 !sysctl_fb_tunnels_only_for_init_net; 641 } 642 643 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 644 { 645 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 646 return q->numa_node; 647 #else 648 return NUMA_NO_NODE; 649 #endif 650 } 651 652 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 653 { 654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 655 q->numa_node = node; 656 #endif 657 } 658 659 #ifdef CONFIG_RPS 660 /* 661 * This structure holds an RPS map which can be of variable length. The 662 * map is an array of CPUs. 663 */ 664 struct rps_map { 665 unsigned int len; 666 struct rcu_head rcu; 667 u16 cpus[0]; 668 }; 669 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 670 671 /* 672 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 673 * tail pointer for that CPU's input queue at the time of last enqueue, and 674 * a hardware filter index. 675 */ 676 struct rps_dev_flow { 677 u16 cpu; 678 u16 filter; 679 unsigned int last_qtail; 680 }; 681 #define RPS_NO_FILTER 0xffff 682 683 /* 684 * The rps_dev_flow_table structure contains a table of flow mappings. 685 */ 686 struct rps_dev_flow_table { 687 unsigned int mask; 688 struct rcu_head rcu; 689 struct rps_dev_flow flows[0]; 690 }; 691 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 692 ((_num) * sizeof(struct rps_dev_flow))) 693 694 /* 695 * The rps_sock_flow_table contains mappings of flows to the last CPU 696 * on which they were processed by the application (set in recvmsg). 697 * Each entry is a 32bit value. Upper part is the high-order bits 698 * of flow hash, lower part is CPU number. 699 * rps_cpu_mask is used to partition the space, depending on number of 700 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 701 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 702 * meaning we use 32-6=26 bits for the hash. 703 */ 704 struct rps_sock_flow_table { 705 u32 mask; 706 707 u32 ents[0] ____cacheline_aligned_in_smp; 708 }; 709 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 710 711 #define RPS_NO_CPU 0xffff 712 713 extern u32 rps_cpu_mask; 714 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 715 716 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 717 u32 hash) 718 { 719 if (table && hash) { 720 unsigned int index = hash & table->mask; 721 u32 val = hash & ~rps_cpu_mask; 722 723 /* We only give a hint, preemption can change CPU under us */ 724 val |= raw_smp_processor_id(); 725 726 if (table->ents[index] != val) 727 table->ents[index] = val; 728 } 729 } 730 731 #ifdef CONFIG_RFS_ACCEL 732 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 733 u16 filter_id); 734 #endif 735 #endif /* CONFIG_RPS */ 736 737 /* This structure contains an instance of an RX queue. */ 738 struct netdev_rx_queue { 739 #ifdef CONFIG_RPS 740 struct rps_map __rcu *rps_map; 741 struct rps_dev_flow_table __rcu *rps_flow_table; 742 #endif 743 struct kobject kobj; 744 struct net_device *dev; 745 struct xdp_rxq_info xdp_rxq; 746 #ifdef CONFIG_XDP_SOCKETS 747 struct xdp_umem *umem; 748 #endif 749 } ____cacheline_aligned_in_smp; 750 751 /* 752 * RX queue sysfs structures and functions. 753 */ 754 struct rx_queue_attribute { 755 struct attribute attr; 756 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 757 ssize_t (*store)(struct netdev_rx_queue *queue, 758 const char *buf, size_t len); 759 }; 760 761 #ifdef CONFIG_XPS 762 /* 763 * This structure holds an XPS map which can be of variable length. The 764 * map is an array of queues. 765 */ 766 struct xps_map { 767 unsigned int len; 768 unsigned int alloc_len; 769 struct rcu_head rcu; 770 u16 queues[0]; 771 }; 772 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 773 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 774 - sizeof(struct xps_map)) / sizeof(u16)) 775 776 /* 777 * This structure holds all XPS maps for device. Maps are indexed by CPU. 778 */ 779 struct xps_dev_maps { 780 struct rcu_head rcu; 781 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 782 }; 783 784 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 785 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 786 787 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 788 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 789 790 #endif /* CONFIG_XPS */ 791 792 #define TC_MAX_QUEUE 16 793 #define TC_BITMASK 15 794 /* HW offloaded queuing disciplines txq count and offset maps */ 795 struct netdev_tc_txq { 796 u16 count; 797 u16 offset; 798 }; 799 800 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 801 /* 802 * This structure is to hold information about the device 803 * configured to run FCoE protocol stack. 804 */ 805 struct netdev_fcoe_hbainfo { 806 char manufacturer[64]; 807 char serial_number[64]; 808 char hardware_version[64]; 809 char driver_version[64]; 810 char optionrom_version[64]; 811 char firmware_version[64]; 812 char model[256]; 813 char model_description[256]; 814 }; 815 #endif 816 817 #define MAX_PHYS_ITEM_ID_LEN 32 818 819 /* This structure holds a unique identifier to identify some 820 * physical item (port for example) used by a netdevice. 821 */ 822 struct netdev_phys_item_id { 823 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 824 unsigned char id_len; 825 }; 826 827 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 828 struct netdev_phys_item_id *b) 829 { 830 return a->id_len == b->id_len && 831 memcmp(a->id, b->id, a->id_len) == 0; 832 } 833 834 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 835 struct sk_buff *skb, 836 struct net_device *sb_dev); 837 838 enum tc_setup_type { 839 TC_SETUP_QDISC_MQPRIO, 840 TC_SETUP_CLSU32, 841 TC_SETUP_CLSFLOWER, 842 TC_SETUP_CLSMATCHALL, 843 TC_SETUP_CLSBPF, 844 TC_SETUP_BLOCK, 845 TC_SETUP_QDISC_CBS, 846 TC_SETUP_QDISC_RED, 847 TC_SETUP_QDISC_PRIO, 848 TC_SETUP_QDISC_MQ, 849 TC_SETUP_QDISC_ETF, 850 TC_SETUP_ROOT_QDISC, 851 TC_SETUP_QDISC_GRED, 852 }; 853 854 /* These structures hold the attributes of bpf state that are being passed 855 * to the netdevice through the bpf op. 856 */ 857 enum bpf_netdev_command { 858 /* Set or clear a bpf program used in the earliest stages of packet 859 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 860 * is responsible for calling bpf_prog_put on any old progs that are 861 * stored. In case of error, the callee need not release the new prog 862 * reference, but on success it takes ownership and must bpf_prog_put 863 * when it is no longer used. 864 */ 865 XDP_SETUP_PROG, 866 XDP_SETUP_PROG_HW, 867 XDP_QUERY_PROG, 868 XDP_QUERY_PROG_HW, 869 /* BPF program for offload callbacks, invoked at program load time. */ 870 BPF_OFFLOAD_MAP_ALLOC, 871 BPF_OFFLOAD_MAP_FREE, 872 XDP_SETUP_XSK_UMEM, 873 }; 874 875 struct bpf_prog_offload_ops; 876 struct netlink_ext_ack; 877 struct xdp_umem; 878 879 struct netdev_bpf { 880 enum bpf_netdev_command command; 881 union { 882 /* XDP_SETUP_PROG */ 883 struct { 884 u32 flags; 885 struct bpf_prog *prog; 886 struct netlink_ext_ack *extack; 887 }; 888 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 889 struct { 890 u32 prog_id; 891 /* flags with which program was installed */ 892 u32 prog_flags; 893 }; 894 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 895 struct { 896 struct bpf_offloaded_map *offmap; 897 }; 898 /* XDP_SETUP_XSK_UMEM */ 899 struct { 900 struct xdp_umem *umem; 901 u16 queue_id; 902 } xsk; 903 }; 904 }; 905 906 #ifdef CONFIG_XFRM_OFFLOAD 907 struct xfrmdev_ops { 908 int (*xdo_dev_state_add) (struct xfrm_state *x); 909 void (*xdo_dev_state_delete) (struct xfrm_state *x); 910 void (*xdo_dev_state_free) (struct xfrm_state *x); 911 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 912 struct xfrm_state *x); 913 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 914 }; 915 #endif 916 917 #if IS_ENABLED(CONFIG_TLS_DEVICE) 918 enum tls_offload_ctx_dir { 919 TLS_OFFLOAD_CTX_DIR_RX, 920 TLS_OFFLOAD_CTX_DIR_TX, 921 }; 922 923 struct tls_crypto_info; 924 struct tls_context; 925 926 struct tlsdev_ops { 927 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk, 928 enum tls_offload_ctx_dir direction, 929 struct tls_crypto_info *crypto_info, 930 u32 start_offload_tcp_sn); 931 void (*tls_dev_del)(struct net_device *netdev, 932 struct tls_context *ctx, 933 enum tls_offload_ctx_dir direction); 934 void (*tls_dev_resync_rx)(struct net_device *netdev, 935 struct sock *sk, u32 seq, u64 rcd_sn); 936 }; 937 #endif 938 939 struct dev_ifalias { 940 struct rcu_head rcuhead; 941 char ifalias[]; 942 }; 943 944 struct devlink; 945 946 /* 947 * This structure defines the management hooks for network devices. 948 * The following hooks can be defined; unless noted otherwise, they are 949 * optional and can be filled with a null pointer. 950 * 951 * int (*ndo_init)(struct net_device *dev); 952 * This function is called once when a network device is registered. 953 * The network device can use this for any late stage initialization 954 * or semantic validation. It can fail with an error code which will 955 * be propagated back to register_netdev. 956 * 957 * void (*ndo_uninit)(struct net_device *dev); 958 * This function is called when device is unregistered or when registration 959 * fails. It is not called if init fails. 960 * 961 * int (*ndo_open)(struct net_device *dev); 962 * This function is called when a network device transitions to the up 963 * state. 964 * 965 * int (*ndo_stop)(struct net_device *dev); 966 * This function is called when a network device transitions to the down 967 * state. 968 * 969 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 970 * struct net_device *dev); 971 * Called when a packet needs to be transmitted. 972 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 973 * the queue before that can happen; it's for obsolete devices and weird 974 * corner cases, but the stack really does a non-trivial amount 975 * of useless work if you return NETDEV_TX_BUSY. 976 * Required; cannot be NULL. 977 * 978 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 979 * struct net_device *dev 980 * netdev_features_t features); 981 * Called by core transmit path to determine if device is capable of 982 * performing offload operations on a given packet. This is to give 983 * the device an opportunity to implement any restrictions that cannot 984 * be otherwise expressed by feature flags. The check is called with 985 * the set of features that the stack has calculated and it returns 986 * those the driver believes to be appropriate. 987 * 988 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 989 * struct net_device *sb_dev); 990 * Called to decide which queue to use when device supports multiple 991 * transmit queues. 992 * 993 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 994 * This function is called to allow device receiver to make 995 * changes to configuration when multicast or promiscuous is enabled. 996 * 997 * void (*ndo_set_rx_mode)(struct net_device *dev); 998 * This function is called device changes address list filtering. 999 * If driver handles unicast address filtering, it should set 1000 * IFF_UNICAST_FLT in its priv_flags. 1001 * 1002 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1003 * This function is called when the Media Access Control address 1004 * needs to be changed. If this interface is not defined, the 1005 * MAC address can not be changed. 1006 * 1007 * int (*ndo_validate_addr)(struct net_device *dev); 1008 * Test if Media Access Control address is valid for the device. 1009 * 1010 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1011 * Called when a user requests an ioctl which can't be handled by 1012 * the generic interface code. If not defined ioctls return 1013 * not supported error code. 1014 * 1015 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1016 * Used to set network devices bus interface parameters. This interface 1017 * is retained for legacy reasons; new devices should use the bus 1018 * interface (PCI) for low level management. 1019 * 1020 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1021 * Called when a user wants to change the Maximum Transfer Unit 1022 * of a device. 1023 * 1024 * void (*ndo_tx_timeout)(struct net_device *dev); 1025 * Callback used when the transmitter has not made any progress 1026 * for dev->watchdog ticks. 1027 * 1028 * void (*ndo_get_stats64)(struct net_device *dev, 1029 * struct rtnl_link_stats64 *storage); 1030 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1031 * Called when a user wants to get the network device usage 1032 * statistics. Drivers must do one of the following: 1033 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1034 * rtnl_link_stats64 structure passed by the caller. 1035 * 2. Define @ndo_get_stats to update a net_device_stats structure 1036 * (which should normally be dev->stats) and return a pointer to 1037 * it. The structure may be changed asynchronously only if each 1038 * field is written atomically. 1039 * 3. Update dev->stats asynchronously and atomically, and define 1040 * neither operation. 1041 * 1042 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1043 * Return true if this device supports offload stats of this attr_id. 1044 * 1045 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1046 * void *attr_data) 1047 * Get statistics for offload operations by attr_id. Write it into the 1048 * attr_data pointer. 1049 * 1050 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1051 * If device supports VLAN filtering this function is called when a 1052 * VLAN id is registered. 1053 * 1054 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1055 * If device supports VLAN filtering this function is called when a 1056 * VLAN id is unregistered. 1057 * 1058 * void (*ndo_poll_controller)(struct net_device *dev); 1059 * 1060 * SR-IOV management functions. 1061 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1062 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1063 * u8 qos, __be16 proto); 1064 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1065 * int max_tx_rate); 1066 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1067 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1068 * int (*ndo_get_vf_config)(struct net_device *dev, 1069 * int vf, struct ifla_vf_info *ivf); 1070 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1071 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1072 * struct nlattr *port[]); 1073 * 1074 * Enable or disable the VF ability to query its RSS Redirection Table and 1075 * Hash Key. This is needed since on some devices VF share this information 1076 * with PF and querying it may introduce a theoretical security risk. 1077 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1078 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1079 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1080 * void *type_data); 1081 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1082 * This is always called from the stack with the rtnl lock held and netif 1083 * tx queues stopped. This allows the netdevice to perform queue 1084 * management safely. 1085 * 1086 * Fiber Channel over Ethernet (FCoE) offload functions. 1087 * int (*ndo_fcoe_enable)(struct net_device *dev); 1088 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1089 * so the underlying device can perform whatever needed configuration or 1090 * initialization to support acceleration of FCoE traffic. 1091 * 1092 * int (*ndo_fcoe_disable)(struct net_device *dev); 1093 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1094 * so the underlying device can perform whatever needed clean-ups to 1095 * stop supporting acceleration of FCoE traffic. 1096 * 1097 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1098 * struct scatterlist *sgl, unsigned int sgc); 1099 * Called when the FCoE Initiator wants to initialize an I/O that 1100 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1101 * perform necessary setup and returns 1 to indicate the device is set up 1102 * successfully to perform DDP on this I/O, otherwise this returns 0. 1103 * 1104 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1105 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1106 * indicated by the FC exchange id 'xid', so the underlying device can 1107 * clean up and reuse resources for later DDP requests. 1108 * 1109 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1110 * struct scatterlist *sgl, unsigned int sgc); 1111 * Called when the FCoE Target wants to initialize an I/O that 1112 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1113 * perform necessary setup and returns 1 to indicate the device is set up 1114 * successfully to perform DDP on this I/O, otherwise this returns 0. 1115 * 1116 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1117 * struct netdev_fcoe_hbainfo *hbainfo); 1118 * Called when the FCoE Protocol stack wants information on the underlying 1119 * device. This information is utilized by the FCoE protocol stack to 1120 * register attributes with Fiber Channel management service as per the 1121 * FC-GS Fabric Device Management Information(FDMI) specification. 1122 * 1123 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1124 * Called when the underlying device wants to override default World Wide 1125 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1126 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1127 * protocol stack to use. 1128 * 1129 * RFS acceleration. 1130 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1131 * u16 rxq_index, u32 flow_id); 1132 * Set hardware filter for RFS. rxq_index is the target queue index; 1133 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1134 * Return the filter ID on success, or a negative error code. 1135 * 1136 * Slave management functions (for bridge, bonding, etc). 1137 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1138 * Called to make another netdev an underling. 1139 * 1140 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1141 * Called to release previously enslaved netdev. 1142 * 1143 * Feature/offload setting functions. 1144 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1145 * netdev_features_t features); 1146 * Adjusts the requested feature flags according to device-specific 1147 * constraints, and returns the resulting flags. Must not modify 1148 * the device state. 1149 * 1150 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1151 * Called to update device configuration to new features. Passed 1152 * feature set might be less than what was returned by ndo_fix_features()). 1153 * Must return >0 or -errno if it changed dev->features itself. 1154 * 1155 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1156 * struct net_device *dev, 1157 * const unsigned char *addr, u16 vid, u16 flags, 1158 * struct netlink_ext_ack *extack); 1159 * Adds an FDB entry to dev for addr. 1160 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1161 * struct net_device *dev, 1162 * const unsigned char *addr, u16 vid) 1163 * Deletes the FDB entry from dev coresponding to addr. 1164 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1165 * struct net_device *dev, struct net_device *filter_dev, 1166 * int *idx) 1167 * Used to add FDB entries to dump requests. Implementers should add 1168 * entries to skb and update idx with the number of entries. 1169 * 1170 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1171 * u16 flags, struct netlink_ext_ack *extack) 1172 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1173 * struct net_device *dev, u32 filter_mask, 1174 * int nlflags) 1175 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1176 * u16 flags); 1177 * 1178 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1179 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1180 * which do not represent real hardware may define this to allow their 1181 * userspace components to manage their virtual carrier state. Devices 1182 * that determine carrier state from physical hardware properties (eg 1183 * network cables) or protocol-dependent mechanisms (eg 1184 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1185 * 1186 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1187 * struct netdev_phys_item_id *ppid); 1188 * Called to get ID of physical port of this device. If driver does 1189 * not implement this, it is assumed that the hw is not able to have 1190 * multiple net devices on single physical port. 1191 * 1192 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1193 * struct netdev_phys_item_id *ppid) 1194 * Called to get the parent ID of the physical port of this device. 1195 * 1196 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1197 * struct udp_tunnel_info *ti); 1198 * Called by UDP tunnel to notify a driver about the UDP port and socket 1199 * address family that a UDP tunnel is listnening to. It is called only 1200 * when a new port starts listening. The operation is protected by the 1201 * RTNL. 1202 * 1203 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1204 * struct udp_tunnel_info *ti); 1205 * Called by UDP tunnel to notify the driver about a UDP port and socket 1206 * address family that the UDP tunnel is not listening to anymore. The 1207 * operation is protected by the RTNL. 1208 * 1209 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1210 * struct net_device *dev) 1211 * Called by upper layer devices to accelerate switching or other 1212 * station functionality into hardware. 'pdev is the lowerdev 1213 * to use for the offload and 'dev' is the net device that will 1214 * back the offload. Returns a pointer to the private structure 1215 * the upper layer will maintain. 1216 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1217 * Called by upper layer device to delete the station created 1218 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1219 * the station and priv is the structure returned by the add 1220 * operation. 1221 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1222 * int queue_index, u32 maxrate); 1223 * Called when a user wants to set a max-rate limitation of specific 1224 * TX queue. 1225 * int (*ndo_get_iflink)(const struct net_device *dev); 1226 * Called to get the iflink value of this device. 1227 * void (*ndo_change_proto_down)(struct net_device *dev, 1228 * bool proto_down); 1229 * This function is used to pass protocol port error state information 1230 * to the switch driver. The switch driver can react to the proto_down 1231 * by doing a phys down on the associated switch port. 1232 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1233 * This function is used to get egress tunnel information for given skb. 1234 * This is useful for retrieving outer tunnel header parameters while 1235 * sampling packet. 1236 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1237 * This function is used to specify the headroom that the skb must 1238 * consider when allocation skb during packet reception. Setting 1239 * appropriate rx headroom value allows avoiding skb head copy on 1240 * forward. Setting a negative value resets the rx headroom to the 1241 * default value. 1242 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1243 * This function is used to set or query state related to XDP on the 1244 * netdevice and manage BPF offload. See definition of 1245 * enum bpf_netdev_command for details. 1246 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1247 * u32 flags); 1248 * This function is used to submit @n XDP packets for transmit on a 1249 * netdevice. Returns number of frames successfully transmitted, frames 1250 * that got dropped are freed/returned via xdp_return_frame(). 1251 * Returns negative number, means general error invoking ndo, meaning 1252 * no frames were xmit'ed and core-caller will free all frames. 1253 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1254 * Get devlink port instance associated with a given netdev. 1255 * Called with a reference on the netdevice and devlink locks only, 1256 * rtnl_lock is not held. 1257 */ 1258 struct net_device_ops { 1259 int (*ndo_init)(struct net_device *dev); 1260 void (*ndo_uninit)(struct net_device *dev); 1261 int (*ndo_open)(struct net_device *dev); 1262 int (*ndo_stop)(struct net_device *dev); 1263 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1264 struct net_device *dev); 1265 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1266 struct net_device *dev, 1267 netdev_features_t features); 1268 u16 (*ndo_select_queue)(struct net_device *dev, 1269 struct sk_buff *skb, 1270 struct net_device *sb_dev); 1271 void (*ndo_change_rx_flags)(struct net_device *dev, 1272 int flags); 1273 void (*ndo_set_rx_mode)(struct net_device *dev); 1274 int (*ndo_set_mac_address)(struct net_device *dev, 1275 void *addr); 1276 int (*ndo_validate_addr)(struct net_device *dev); 1277 int (*ndo_do_ioctl)(struct net_device *dev, 1278 struct ifreq *ifr, int cmd); 1279 int (*ndo_set_config)(struct net_device *dev, 1280 struct ifmap *map); 1281 int (*ndo_change_mtu)(struct net_device *dev, 1282 int new_mtu); 1283 int (*ndo_neigh_setup)(struct net_device *dev, 1284 struct neigh_parms *); 1285 void (*ndo_tx_timeout) (struct net_device *dev); 1286 1287 void (*ndo_get_stats64)(struct net_device *dev, 1288 struct rtnl_link_stats64 *storage); 1289 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1290 int (*ndo_get_offload_stats)(int attr_id, 1291 const struct net_device *dev, 1292 void *attr_data); 1293 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1294 1295 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1296 __be16 proto, u16 vid); 1297 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1298 __be16 proto, u16 vid); 1299 #ifdef CONFIG_NET_POLL_CONTROLLER 1300 void (*ndo_poll_controller)(struct net_device *dev); 1301 int (*ndo_netpoll_setup)(struct net_device *dev, 1302 struct netpoll_info *info); 1303 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1304 #endif 1305 int (*ndo_set_vf_mac)(struct net_device *dev, 1306 int queue, u8 *mac); 1307 int (*ndo_set_vf_vlan)(struct net_device *dev, 1308 int queue, u16 vlan, 1309 u8 qos, __be16 proto); 1310 int (*ndo_set_vf_rate)(struct net_device *dev, 1311 int vf, int min_tx_rate, 1312 int max_tx_rate); 1313 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1314 int vf, bool setting); 1315 int (*ndo_set_vf_trust)(struct net_device *dev, 1316 int vf, bool setting); 1317 int (*ndo_get_vf_config)(struct net_device *dev, 1318 int vf, 1319 struct ifla_vf_info *ivf); 1320 int (*ndo_set_vf_link_state)(struct net_device *dev, 1321 int vf, int link_state); 1322 int (*ndo_get_vf_stats)(struct net_device *dev, 1323 int vf, 1324 struct ifla_vf_stats 1325 *vf_stats); 1326 int (*ndo_set_vf_port)(struct net_device *dev, 1327 int vf, 1328 struct nlattr *port[]); 1329 int (*ndo_get_vf_port)(struct net_device *dev, 1330 int vf, struct sk_buff *skb); 1331 int (*ndo_set_vf_guid)(struct net_device *dev, 1332 int vf, u64 guid, 1333 int guid_type); 1334 int (*ndo_set_vf_rss_query_en)( 1335 struct net_device *dev, 1336 int vf, bool setting); 1337 int (*ndo_setup_tc)(struct net_device *dev, 1338 enum tc_setup_type type, 1339 void *type_data); 1340 #if IS_ENABLED(CONFIG_FCOE) 1341 int (*ndo_fcoe_enable)(struct net_device *dev); 1342 int (*ndo_fcoe_disable)(struct net_device *dev); 1343 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1344 u16 xid, 1345 struct scatterlist *sgl, 1346 unsigned int sgc); 1347 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1348 u16 xid); 1349 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1350 u16 xid, 1351 struct scatterlist *sgl, 1352 unsigned int sgc); 1353 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1354 struct netdev_fcoe_hbainfo *hbainfo); 1355 #endif 1356 1357 #if IS_ENABLED(CONFIG_LIBFCOE) 1358 #define NETDEV_FCOE_WWNN 0 1359 #define NETDEV_FCOE_WWPN 1 1360 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1361 u64 *wwn, int type); 1362 #endif 1363 1364 #ifdef CONFIG_RFS_ACCEL 1365 int (*ndo_rx_flow_steer)(struct net_device *dev, 1366 const struct sk_buff *skb, 1367 u16 rxq_index, 1368 u32 flow_id); 1369 #endif 1370 int (*ndo_add_slave)(struct net_device *dev, 1371 struct net_device *slave_dev, 1372 struct netlink_ext_ack *extack); 1373 int (*ndo_del_slave)(struct net_device *dev, 1374 struct net_device *slave_dev); 1375 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1376 netdev_features_t features); 1377 int (*ndo_set_features)(struct net_device *dev, 1378 netdev_features_t features); 1379 int (*ndo_neigh_construct)(struct net_device *dev, 1380 struct neighbour *n); 1381 void (*ndo_neigh_destroy)(struct net_device *dev, 1382 struct neighbour *n); 1383 1384 int (*ndo_fdb_add)(struct ndmsg *ndm, 1385 struct nlattr *tb[], 1386 struct net_device *dev, 1387 const unsigned char *addr, 1388 u16 vid, 1389 u16 flags, 1390 struct netlink_ext_ack *extack); 1391 int (*ndo_fdb_del)(struct ndmsg *ndm, 1392 struct nlattr *tb[], 1393 struct net_device *dev, 1394 const unsigned char *addr, 1395 u16 vid); 1396 int (*ndo_fdb_dump)(struct sk_buff *skb, 1397 struct netlink_callback *cb, 1398 struct net_device *dev, 1399 struct net_device *filter_dev, 1400 int *idx); 1401 int (*ndo_fdb_get)(struct sk_buff *skb, 1402 struct nlattr *tb[], 1403 struct net_device *dev, 1404 const unsigned char *addr, 1405 u16 vid, u32 portid, u32 seq, 1406 struct netlink_ext_ack *extack); 1407 int (*ndo_bridge_setlink)(struct net_device *dev, 1408 struct nlmsghdr *nlh, 1409 u16 flags, 1410 struct netlink_ext_ack *extack); 1411 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1412 u32 pid, u32 seq, 1413 struct net_device *dev, 1414 u32 filter_mask, 1415 int nlflags); 1416 int (*ndo_bridge_dellink)(struct net_device *dev, 1417 struct nlmsghdr *nlh, 1418 u16 flags); 1419 int (*ndo_change_carrier)(struct net_device *dev, 1420 bool new_carrier); 1421 int (*ndo_get_phys_port_id)(struct net_device *dev, 1422 struct netdev_phys_item_id *ppid); 1423 int (*ndo_get_port_parent_id)(struct net_device *dev, 1424 struct netdev_phys_item_id *ppid); 1425 int (*ndo_get_phys_port_name)(struct net_device *dev, 1426 char *name, size_t len); 1427 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1428 struct udp_tunnel_info *ti); 1429 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1430 struct udp_tunnel_info *ti); 1431 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1432 struct net_device *dev); 1433 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1434 void *priv); 1435 1436 int (*ndo_get_lock_subclass)(struct net_device *dev); 1437 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1438 int queue_index, 1439 u32 maxrate); 1440 int (*ndo_get_iflink)(const struct net_device *dev); 1441 int (*ndo_change_proto_down)(struct net_device *dev, 1442 bool proto_down); 1443 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1444 struct sk_buff *skb); 1445 void (*ndo_set_rx_headroom)(struct net_device *dev, 1446 int needed_headroom); 1447 int (*ndo_bpf)(struct net_device *dev, 1448 struct netdev_bpf *bpf); 1449 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1450 struct xdp_frame **xdp, 1451 u32 flags); 1452 int (*ndo_xsk_async_xmit)(struct net_device *dev, 1453 u32 queue_id); 1454 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1455 }; 1456 1457 /** 1458 * enum net_device_priv_flags - &struct net_device priv_flags 1459 * 1460 * These are the &struct net_device, they are only set internally 1461 * by drivers and used in the kernel. These flags are invisible to 1462 * userspace; this means that the order of these flags can change 1463 * during any kernel release. 1464 * 1465 * You should have a pretty good reason to be extending these flags. 1466 * 1467 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1468 * @IFF_EBRIDGE: Ethernet bridging device 1469 * @IFF_BONDING: bonding master or slave 1470 * @IFF_ISATAP: ISATAP interface (RFC4214) 1471 * @IFF_WAN_HDLC: WAN HDLC device 1472 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1473 * release skb->dst 1474 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1475 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1476 * @IFF_MACVLAN_PORT: device used as macvlan port 1477 * @IFF_BRIDGE_PORT: device used as bridge port 1478 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1479 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1480 * @IFF_UNICAST_FLT: Supports unicast filtering 1481 * @IFF_TEAM_PORT: device used as team port 1482 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1483 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1484 * change when it's running 1485 * @IFF_MACVLAN: Macvlan device 1486 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1487 * underlying stacked devices 1488 * @IFF_L3MDEV_MASTER: device is an L3 master device 1489 * @IFF_NO_QUEUE: device can run without qdisc attached 1490 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1491 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1492 * @IFF_TEAM: device is a team device 1493 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1494 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1495 * entity (i.e. the master device for bridged veth) 1496 * @IFF_MACSEC: device is a MACsec device 1497 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1498 * @IFF_FAILOVER: device is a failover master device 1499 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1500 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1501 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1502 */ 1503 enum netdev_priv_flags { 1504 IFF_802_1Q_VLAN = 1<<0, 1505 IFF_EBRIDGE = 1<<1, 1506 IFF_BONDING = 1<<2, 1507 IFF_ISATAP = 1<<3, 1508 IFF_WAN_HDLC = 1<<4, 1509 IFF_XMIT_DST_RELEASE = 1<<5, 1510 IFF_DONT_BRIDGE = 1<<6, 1511 IFF_DISABLE_NETPOLL = 1<<7, 1512 IFF_MACVLAN_PORT = 1<<8, 1513 IFF_BRIDGE_PORT = 1<<9, 1514 IFF_OVS_DATAPATH = 1<<10, 1515 IFF_TX_SKB_SHARING = 1<<11, 1516 IFF_UNICAST_FLT = 1<<12, 1517 IFF_TEAM_PORT = 1<<13, 1518 IFF_SUPP_NOFCS = 1<<14, 1519 IFF_LIVE_ADDR_CHANGE = 1<<15, 1520 IFF_MACVLAN = 1<<16, 1521 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1522 IFF_L3MDEV_MASTER = 1<<18, 1523 IFF_NO_QUEUE = 1<<19, 1524 IFF_OPENVSWITCH = 1<<20, 1525 IFF_L3MDEV_SLAVE = 1<<21, 1526 IFF_TEAM = 1<<22, 1527 IFF_RXFH_CONFIGURED = 1<<23, 1528 IFF_PHONY_HEADROOM = 1<<24, 1529 IFF_MACSEC = 1<<25, 1530 IFF_NO_RX_HANDLER = 1<<26, 1531 IFF_FAILOVER = 1<<27, 1532 IFF_FAILOVER_SLAVE = 1<<28, 1533 IFF_L3MDEV_RX_HANDLER = 1<<29, 1534 IFF_LIVE_RENAME_OK = 1<<30, 1535 }; 1536 1537 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1538 #define IFF_EBRIDGE IFF_EBRIDGE 1539 #define IFF_BONDING IFF_BONDING 1540 #define IFF_ISATAP IFF_ISATAP 1541 #define IFF_WAN_HDLC IFF_WAN_HDLC 1542 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1543 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1544 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1545 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1546 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1547 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1548 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1549 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1550 #define IFF_TEAM_PORT IFF_TEAM_PORT 1551 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1552 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1553 #define IFF_MACVLAN IFF_MACVLAN 1554 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1555 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1556 #define IFF_NO_QUEUE IFF_NO_QUEUE 1557 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1558 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1559 #define IFF_TEAM IFF_TEAM 1560 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1561 #define IFF_MACSEC IFF_MACSEC 1562 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1563 #define IFF_FAILOVER IFF_FAILOVER 1564 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1565 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1566 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1567 1568 /** 1569 * struct net_device - The DEVICE structure. 1570 * 1571 * Actually, this whole structure is a big mistake. It mixes I/O 1572 * data with strictly "high-level" data, and it has to know about 1573 * almost every data structure used in the INET module. 1574 * 1575 * @name: This is the first field of the "visible" part of this structure 1576 * (i.e. as seen by users in the "Space.c" file). It is the name 1577 * of the interface. 1578 * 1579 * @name_hlist: Device name hash chain, please keep it close to name[] 1580 * @ifalias: SNMP alias 1581 * @mem_end: Shared memory end 1582 * @mem_start: Shared memory start 1583 * @base_addr: Device I/O address 1584 * @irq: Device IRQ number 1585 * 1586 * @state: Generic network queuing layer state, see netdev_state_t 1587 * @dev_list: The global list of network devices 1588 * @napi_list: List entry used for polling NAPI devices 1589 * @unreg_list: List entry when we are unregistering the 1590 * device; see the function unregister_netdev 1591 * @close_list: List entry used when we are closing the device 1592 * @ptype_all: Device-specific packet handlers for all protocols 1593 * @ptype_specific: Device-specific, protocol-specific packet handlers 1594 * 1595 * @adj_list: Directly linked devices, like slaves for bonding 1596 * @features: Currently active device features 1597 * @hw_features: User-changeable features 1598 * 1599 * @wanted_features: User-requested features 1600 * @vlan_features: Mask of features inheritable by VLAN devices 1601 * 1602 * @hw_enc_features: Mask of features inherited by encapsulating devices 1603 * This field indicates what encapsulation 1604 * offloads the hardware is capable of doing, 1605 * and drivers will need to set them appropriately. 1606 * 1607 * @mpls_features: Mask of features inheritable by MPLS 1608 * 1609 * @ifindex: interface index 1610 * @group: The group the device belongs to 1611 * 1612 * @stats: Statistics struct, which was left as a legacy, use 1613 * rtnl_link_stats64 instead 1614 * 1615 * @rx_dropped: Dropped packets by core network, 1616 * do not use this in drivers 1617 * @tx_dropped: Dropped packets by core network, 1618 * do not use this in drivers 1619 * @rx_nohandler: nohandler dropped packets by core network on 1620 * inactive devices, do not use this in drivers 1621 * @carrier_up_count: Number of times the carrier has been up 1622 * @carrier_down_count: Number of times the carrier has been down 1623 * 1624 * @wireless_handlers: List of functions to handle Wireless Extensions, 1625 * instead of ioctl, 1626 * see <net/iw_handler.h> for details. 1627 * @wireless_data: Instance data managed by the core of wireless extensions 1628 * 1629 * @netdev_ops: Includes several pointers to callbacks, 1630 * if one wants to override the ndo_*() functions 1631 * @ethtool_ops: Management operations 1632 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1633 * discovery handling. Necessary for e.g. 6LoWPAN. 1634 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1635 * of Layer 2 headers. 1636 * 1637 * @flags: Interface flags (a la BSD) 1638 * @priv_flags: Like 'flags' but invisible to userspace, 1639 * see if.h for the definitions 1640 * @gflags: Global flags ( kept as legacy ) 1641 * @padded: How much padding added by alloc_netdev() 1642 * @operstate: RFC2863 operstate 1643 * @link_mode: Mapping policy to operstate 1644 * @if_port: Selectable AUI, TP, ... 1645 * @dma: DMA channel 1646 * @mtu: Interface MTU value 1647 * @min_mtu: Interface Minimum MTU value 1648 * @max_mtu: Interface Maximum MTU value 1649 * @type: Interface hardware type 1650 * @hard_header_len: Maximum hardware header length. 1651 * @min_header_len: Minimum hardware header length 1652 * 1653 * @needed_headroom: Extra headroom the hardware may need, but not in all 1654 * cases can this be guaranteed 1655 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1656 * cases can this be guaranteed. Some cases also use 1657 * LL_MAX_HEADER instead to allocate the skb 1658 * 1659 * interface address info: 1660 * 1661 * @perm_addr: Permanent hw address 1662 * @addr_assign_type: Hw address assignment type 1663 * @addr_len: Hardware address length 1664 * @neigh_priv_len: Used in neigh_alloc() 1665 * @dev_id: Used to differentiate devices that share 1666 * the same link layer address 1667 * @dev_port: Used to differentiate devices that share 1668 * the same function 1669 * @addr_list_lock: XXX: need comments on this one 1670 * @uc_promisc: Counter that indicates promiscuous mode 1671 * has been enabled due to the need to listen to 1672 * additional unicast addresses in a device that 1673 * does not implement ndo_set_rx_mode() 1674 * @uc: unicast mac addresses 1675 * @mc: multicast mac addresses 1676 * @dev_addrs: list of device hw addresses 1677 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1678 * @promiscuity: Number of times the NIC is told to work in 1679 * promiscuous mode; if it becomes 0 the NIC will 1680 * exit promiscuous mode 1681 * @allmulti: Counter, enables or disables allmulticast mode 1682 * 1683 * @vlan_info: VLAN info 1684 * @dsa_ptr: dsa specific data 1685 * @tipc_ptr: TIPC specific data 1686 * @atalk_ptr: AppleTalk link 1687 * @ip_ptr: IPv4 specific data 1688 * @dn_ptr: DECnet specific data 1689 * @ip6_ptr: IPv6 specific data 1690 * @ax25_ptr: AX.25 specific data 1691 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1692 * 1693 * @dev_addr: Hw address (before bcast, 1694 * because most packets are unicast) 1695 * 1696 * @_rx: Array of RX queues 1697 * @num_rx_queues: Number of RX queues 1698 * allocated at register_netdev() time 1699 * @real_num_rx_queues: Number of RX queues currently active in device 1700 * 1701 * @rx_handler: handler for received packets 1702 * @rx_handler_data: XXX: need comments on this one 1703 * @miniq_ingress: ingress/clsact qdisc specific data for 1704 * ingress processing 1705 * @ingress_queue: XXX: need comments on this one 1706 * @broadcast: hw bcast address 1707 * 1708 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1709 * indexed by RX queue number. Assigned by driver. 1710 * This must only be set if the ndo_rx_flow_steer 1711 * operation is defined 1712 * @index_hlist: Device index hash chain 1713 * 1714 * @_tx: Array of TX queues 1715 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1716 * @real_num_tx_queues: Number of TX queues currently active in device 1717 * @qdisc: Root qdisc from userspace point of view 1718 * @tx_queue_len: Max frames per queue allowed 1719 * @tx_global_lock: XXX: need comments on this one 1720 * 1721 * @xps_maps: XXX: need comments on this one 1722 * @miniq_egress: clsact qdisc specific data for 1723 * egress processing 1724 * @watchdog_timeo: Represents the timeout that is used by 1725 * the watchdog (see dev_watchdog()) 1726 * @watchdog_timer: List of timers 1727 * 1728 * @pcpu_refcnt: Number of references to this device 1729 * @todo_list: Delayed register/unregister 1730 * @link_watch_list: XXX: need comments on this one 1731 * 1732 * @reg_state: Register/unregister state machine 1733 * @dismantle: Device is going to be freed 1734 * @rtnl_link_state: This enum represents the phases of creating 1735 * a new link 1736 * 1737 * @needs_free_netdev: Should unregister perform free_netdev? 1738 * @priv_destructor: Called from unregister 1739 * @npinfo: XXX: need comments on this one 1740 * @nd_net: Network namespace this network device is inside 1741 * 1742 * @ml_priv: Mid-layer private 1743 * @lstats: Loopback statistics 1744 * @tstats: Tunnel statistics 1745 * @dstats: Dummy statistics 1746 * @vstats: Virtual ethernet statistics 1747 * 1748 * @garp_port: GARP 1749 * @mrp_port: MRP 1750 * 1751 * @dev: Class/net/name entry 1752 * @sysfs_groups: Space for optional device, statistics and wireless 1753 * sysfs groups 1754 * 1755 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1756 * @rtnl_link_ops: Rtnl_link_ops 1757 * 1758 * @gso_max_size: Maximum size of generic segmentation offload 1759 * @gso_max_segs: Maximum number of segments that can be passed to the 1760 * NIC for GSO 1761 * 1762 * @dcbnl_ops: Data Center Bridging netlink ops 1763 * @num_tc: Number of traffic classes in the net device 1764 * @tc_to_txq: XXX: need comments on this one 1765 * @prio_tc_map: XXX: need comments on this one 1766 * 1767 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1768 * 1769 * @priomap: XXX: need comments on this one 1770 * @phydev: Physical device may attach itself 1771 * for hardware timestamping 1772 * @sfp_bus: attached &struct sfp_bus structure. 1773 * 1774 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1775 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1776 * 1777 * @proto_down: protocol port state information can be sent to the 1778 * switch driver and used to set the phys state of the 1779 * switch port. 1780 * 1781 * @wol_enabled: Wake-on-LAN is enabled 1782 * 1783 * FIXME: cleanup struct net_device such that network protocol info 1784 * moves out. 1785 */ 1786 1787 struct net_device { 1788 char name[IFNAMSIZ]; 1789 struct hlist_node name_hlist; 1790 struct dev_ifalias __rcu *ifalias; 1791 /* 1792 * I/O specific fields 1793 * FIXME: Merge these and struct ifmap into one 1794 */ 1795 unsigned long mem_end; 1796 unsigned long mem_start; 1797 unsigned long base_addr; 1798 int irq; 1799 1800 /* 1801 * Some hardware also needs these fields (state,dev_list, 1802 * napi_list,unreg_list,close_list) but they are not 1803 * part of the usual set specified in Space.c. 1804 */ 1805 1806 unsigned long state; 1807 1808 struct list_head dev_list; 1809 struct list_head napi_list; 1810 struct list_head unreg_list; 1811 struct list_head close_list; 1812 struct list_head ptype_all; 1813 struct list_head ptype_specific; 1814 1815 struct { 1816 struct list_head upper; 1817 struct list_head lower; 1818 } adj_list; 1819 1820 netdev_features_t features; 1821 netdev_features_t hw_features; 1822 netdev_features_t wanted_features; 1823 netdev_features_t vlan_features; 1824 netdev_features_t hw_enc_features; 1825 netdev_features_t mpls_features; 1826 netdev_features_t gso_partial_features; 1827 1828 int ifindex; 1829 int group; 1830 1831 struct net_device_stats stats; 1832 1833 atomic_long_t rx_dropped; 1834 atomic_long_t tx_dropped; 1835 atomic_long_t rx_nohandler; 1836 1837 /* Stats to monitor link on/off, flapping */ 1838 atomic_t carrier_up_count; 1839 atomic_t carrier_down_count; 1840 1841 #ifdef CONFIG_WIRELESS_EXT 1842 const struct iw_handler_def *wireless_handlers; 1843 struct iw_public_data *wireless_data; 1844 #endif 1845 const struct net_device_ops *netdev_ops; 1846 const struct ethtool_ops *ethtool_ops; 1847 #ifdef CONFIG_NET_L3_MASTER_DEV 1848 const struct l3mdev_ops *l3mdev_ops; 1849 #endif 1850 #if IS_ENABLED(CONFIG_IPV6) 1851 const struct ndisc_ops *ndisc_ops; 1852 #endif 1853 1854 #ifdef CONFIG_XFRM_OFFLOAD 1855 const struct xfrmdev_ops *xfrmdev_ops; 1856 #endif 1857 1858 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1859 const struct tlsdev_ops *tlsdev_ops; 1860 #endif 1861 1862 const struct header_ops *header_ops; 1863 1864 unsigned int flags; 1865 unsigned int priv_flags; 1866 1867 unsigned short gflags; 1868 unsigned short padded; 1869 1870 unsigned char operstate; 1871 unsigned char link_mode; 1872 1873 unsigned char if_port; 1874 unsigned char dma; 1875 1876 unsigned int mtu; 1877 unsigned int min_mtu; 1878 unsigned int max_mtu; 1879 unsigned short type; 1880 unsigned short hard_header_len; 1881 unsigned char min_header_len; 1882 1883 unsigned short needed_headroom; 1884 unsigned short needed_tailroom; 1885 1886 /* Interface address info. */ 1887 unsigned char perm_addr[MAX_ADDR_LEN]; 1888 unsigned char addr_assign_type; 1889 unsigned char addr_len; 1890 unsigned short neigh_priv_len; 1891 unsigned short dev_id; 1892 unsigned short dev_port; 1893 spinlock_t addr_list_lock; 1894 unsigned char name_assign_type; 1895 bool uc_promisc; 1896 struct netdev_hw_addr_list uc; 1897 struct netdev_hw_addr_list mc; 1898 struct netdev_hw_addr_list dev_addrs; 1899 1900 #ifdef CONFIG_SYSFS 1901 struct kset *queues_kset; 1902 #endif 1903 unsigned int promiscuity; 1904 unsigned int allmulti; 1905 1906 1907 /* Protocol-specific pointers */ 1908 1909 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1910 struct vlan_info __rcu *vlan_info; 1911 #endif 1912 #if IS_ENABLED(CONFIG_NET_DSA) 1913 struct dsa_port *dsa_ptr; 1914 #endif 1915 #if IS_ENABLED(CONFIG_TIPC) 1916 struct tipc_bearer __rcu *tipc_ptr; 1917 #endif 1918 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1919 void *atalk_ptr; 1920 #endif 1921 struct in_device __rcu *ip_ptr; 1922 #if IS_ENABLED(CONFIG_DECNET) 1923 struct dn_dev __rcu *dn_ptr; 1924 #endif 1925 struct inet6_dev __rcu *ip6_ptr; 1926 #if IS_ENABLED(CONFIG_AX25) 1927 void *ax25_ptr; 1928 #endif 1929 struct wireless_dev *ieee80211_ptr; 1930 struct wpan_dev *ieee802154_ptr; 1931 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1932 struct mpls_dev __rcu *mpls_ptr; 1933 #endif 1934 1935 /* 1936 * Cache lines mostly used on receive path (including eth_type_trans()) 1937 */ 1938 /* Interface address info used in eth_type_trans() */ 1939 unsigned char *dev_addr; 1940 1941 struct netdev_rx_queue *_rx; 1942 unsigned int num_rx_queues; 1943 unsigned int real_num_rx_queues; 1944 1945 struct bpf_prog __rcu *xdp_prog; 1946 unsigned long gro_flush_timeout; 1947 rx_handler_func_t __rcu *rx_handler; 1948 void __rcu *rx_handler_data; 1949 1950 #ifdef CONFIG_NET_CLS_ACT 1951 struct mini_Qdisc __rcu *miniq_ingress; 1952 #endif 1953 struct netdev_queue __rcu *ingress_queue; 1954 #ifdef CONFIG_NETFILTER_INGRESS 1955 struct nf_hook_entries __rcu *nf_hooks_ingress; 1956 #endif 1957 1958 unsigned char broadcast[MAX_ADDR_LEN]; 1959 #ifdef CONFIG_RFS_ACCEL 1960 struct cpu_rmap *rx_cpu_rmap; 1961 #endif 1962 struct hlist_node index_hlist; 1963 1964 /* 1965 * Cache lines mostly used on transmit path 1966 */ 1967 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1968 unsigned int num_tx_queues; 1969 unsigned int real_num_tx_queues; 1970 struct Qdisc *qdisc; 1971 #ifdef CONFIG_NET_SCHED 1972 DECLARE_HASHTABLE (qdisc_hash, 4); 1973 #endif 1974 unsigned int tx_queue_len; 1975 spinlock_t tx_global_lock; 1976 int watchdog_timeo; 1977 1978 #ifdef CONFIG_XPS 1979 struct xps_dev_maps __rcu *xps_cpus_map; 1980 struct xps_dev_maps __rcu *xps_rxqs_map; 1981 #endif 1982 #ifdef CONFIG_NET_CLS_ACT 1983 struct mini_Qdisc __rcu *miniq_egress; 1984 #endif 1985 1986 /* These may be needed for future network-power-down code. */ 1987 struct timer_list watchdog_timer; 1988 1989 int __percpu *pcpu_refcnt; 1990 struct list_head todo_list; 1991 1992 struct list_head link_watch_list; 1993 1994 enum { NETREG_UNINITIALIZED=0, 1995 NETREG_REGISTERED, /* completed register_netdevice */ 1996 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1997 NETREG_UNREGISTERED, /* completed unregister todo */ 1998 NETREG_RELEASED, /* called free_netdev */ 1999 NETREG_DUMMY, /* dummy device for NAPI poll */ 2000 } reg_state:8; 2001 2002 bool dismantle; 2003 2004 enum { 2005 RTNL_LINK_INITIALIZED, 2006 RTNL_LINK_INITIALIZING, 2007 } rtnl_link_state:16; 2008 2009 bool needs_free_netdev; 2010 void (*priv_destructor)(struct net_device *dev); 2011 2012 #ifdef CONFIG_NETPOLL 2013 struct netpoll_info __rcu *npinfo; 2014 #endif 2015 2016 possible_net_t nd_net; 2017 2018 /* mid-layer private */ 2019 union { 2020 void *ml_priv; 2021 struct pcpu_lstats __percpu *lstats; 2022 struct pcpu_sw_netstats __percpu *tstats; 2023 struct pcpu_dstats __percpu *dstats; 2024 }; 2025 2026 #if IS_ENABLED(CONFIG_GARP) 2027 struct garp_port __rcu *garp_port; 2028 #endif 2029 #if IS_ENABLED(CONFIG_MRP) 2030 struct mrp_port __rcu *mrp_port; 2031 #endif 2032 2033 struct device dev; 2034 const struct attribute_group *sysfs_groups[4]; 2035 const struct attribute_group *sysfs_rx_queue_group; 2036 2037 const struct rtnl_link_ops *rtnl_link_ops; 2038 2039 /* for setting kernel sock attribute on TCP connection setup */ 2040 #define GSO_MAX_SIZE 65536 2041 unsigned int gso_max_size; 2042 #define GSO_MAX_SEGS 65535 2043 u16 gso_max_segs; 2044 2045 #ifdef CONFIG_DCB 2046 const struct dcbnl_rtnl_ops *dcbnl_ops; 2047 #endif 2048 s16 num_tc; 2049 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2050 u8 prio_tc_map[TC_BITMASK + 1]; 2051 2052 #if IS_ENABLED(CONFIG_FCOE) 2053 unsigned int fcoe_ddp_xid; 2054 #endif 2055 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2056 struct netprio_map __rcu *priomap; 2057 #endif 2058 struct phy_device *phydev; 2059 struct sfp_bus *sfp_bus; 2060 struct lock_class_key *qdisc_tx_busylock; 2061 struct lock_class_key *qdisc_running_key; 2062 bool proto_down; 2063 unsigned wol_enabled:1; 2064 }; 2065 #define to_net_dev(d) container_of(d, struct net_device, dev) 2066 2067 static inline bool netif_elide_gro(const struct net_device *dev) 2068 { 2069 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2070 return true; 2071 return false; 2072 } 2073 2074 #define NETDEV_ALIGN 32 2075 2076 static inline 2077 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2078 { 2079 return dev->prio_tc_map[prio & TC_BITMASK]; 2080 } 2081 2082 static inline 2083 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2084 { 2085 if (tc >= dev->num_tc) 2086 return -EINVAL; 2087 2088 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2089 return 0; 2090 } 2091 2092 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2093 void netdev_reset_tc(struct net_device *dev); 2094 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2095 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2096 2097 static inline 2098 int netdev_get_num_tc(struct net_device *dev) 2099 { 2100 return dev->num_tc; 2101 } 2102 2103 void netdev_unbind_sb_channel(struct net_device *dev, 2104 struct net_device *sb_dev); 2105 int netdev_bind_sb_channel_queue(struct net_device *dev, 2106 struct net_device *sb_dev, 2107 u8 tc, u16 count, u16 offset); 2108 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2109 static inline int netdev_get_sb_channel(struct net_device *dev) 2110 { 2111 return max_t(int, -dev->num_tc, 0); 2112 } 2113 2114 static inline 2115 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2116 unsigned int index) 2117 { 2118 return &dev->_tx[index]; 2119 } 2120 2121 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2122 const struct sk_buff *skb) 2123 { 2124 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2125 } 2126 2127 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2128 void (*f)(struct net_device *, 2129 struct netdev_queue *, 2130 void *), 2131 void *arg) 2132 { 2133 unsigned int i; 2134 2135 for (i = 0; i < dev->num_tx_queues; i++) 2136 f(dev, &dev->_tx[i], arg); 2137 } 2138 2139 #define netdev_lockdep_set_classes(dev) \ 2140 { \ 2141 static struct lock_class_key qdisc_tx_busylock_key; \ 2142 static struct lock_class_key qdisc_running_key; \ 2143 static struct lock_class_key qdisc_xmit_lock_key; \ 2144 static struct lock_class_key dev_addr_list_lock_key; \ 2145 unsigned int i; \ 2146 \ 2147 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2148 (dev)->qdisc_running_key = &qdisc_running_key; \ 2149 lockdep_set_class(&(dev)->addr_list_lock, \ 2150 &dev_addr_list_lock_key); \ 2151 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2152 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2153 &qdisc_xmit_lock_key); \ 2154 } 2155 2156 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2157 struct net_device *sb_dev); 2158 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2159 struct sk_buff *skb, 2160 struct net_device *sb_dev); 2161 2162 /* returns the headroom that the master device needs to take in account 2163 * when forwarding to this dev 2164 */ 2165 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2166 { 2167 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2168 } 2169 2170 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2171 { 2172 if (dev->netdev_ops->ndo_set_rx_headroom) 2173 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2174 } 2175 2176 /* set the device rx headroom to the dev's default */ 2177 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2178 { 2179 netdev_set_rx_headroom(dev, -1); 2180 } 2181 2182 /* 2183 * Net namespace inlines 2184 */ 2185 static inline 2186 struct net *dev_net(const struct net_device *dev) 2187 { 2188 return read_pnet(&dev->nd_net); 2189 } 2190 2191 static inline 2192 void dev_net_set(struct net_device *dev, struct net *net) 2193 { 2194 write_pnet(&dev->nd_net, net); 2195 } 2196 2197 /** 2198 * netdev_priv - access network device private data 2199 * @dev: network device 2200 * 2201 * Get network device private data 2202 */ 2203 static inline void *netdev_priv(const struct net_device *dev) 2204 { 2205 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2206 } 2207 2208 /* Set the sysfs physical device reference for the network logical device 2209 * if set prior to registration will cause a symlink during initialization. 2210 */ 2211 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2212 2213 /* Set the sysfs device type for the network logical device to allow 2214 * fine-grained identification of different network device types. For 2215 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2216 */ 2217 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2218 2219 /* Default NAPI poll() weight 2220 * Device drivers are strongly advised to not use bigger value 2221 */ 2222 #define NAPI_POLL_WEIGHT 64 2223 2224 /** 2225 * netif_napi_add - initialize a NAPI context 2226 * @dev: network device 2227 * @napi: NAPI context 2228 * @poll: polling function 2229 * @weight: default weight 2230 * 2231 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2232 * *any* of the other NAPI-related functions. 2233 */ 2234 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2235 int (*poll)(struct napi_struct *, int), int weight); 2236 2237 /** 2238 * netif_tx_napi_add - initialize a NAPI context 2239 * @dev: network device 2240 * @napi: NAPI context 2241 * @poll: polling function 2242 * @weight: default weight 2243 * 2244 * This variant of netif_napi_add() should be used from drivers using NAPI 2245 * to exclusively poll a TX queue. 2246 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2247 */ 2248 static inline void netif_tx_napi_add(struct net_device *dev, 2249 struct napi_struct *napi, 2250 int (*poll)(struct napi_struct *, int), 2251 int weight) 2252 { 2253 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2254 netif_napi_add(dev, napi, poll, weight); 2255 } 2256 2257 /** 2258 * netif_napi_del - remove a NAPI context 2259 * @napi: NAPI context 2260 * 2261 * netif_napi_del() removes a NAPI context from the network device NAPI list 2262 */ 2263 void netif_napi_del(struct napi_struct *napi); 2264 2265 struct napi_gro_cb { 2266 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2267 void *frag0; 2268 2269 /* Length of frag0. */ 2270 unsigned int frag0_len; 2271 2272 /* This indicates where we are processing relative to skb->data. */ 2273 int data_offset; 2274 2275 /* This is non-zero if the packet cannot be merged with the new skb. */ 2276 u16 flush; 2277 2278 /* Save the IP ID here and check when we get to the transport layer */ 2279 u16 flush_id; 2280 2281 /* Number of segments aggregated. */ 2282 u16 count; 2283 2284 /* Start offset for remote checksum offload */ 2285 u16 gro_remcsum_start; 2286 2287 /* jiffies when first packet was created/queued */ 2288 unsigned long age; 2289 2290 /* Used in ipv6_gro_receive() and foo-over-udp */ 2291 u16 proto; 2292 2293 /* This is non-zero if the packet may be of the same flow. */ 2294 u8 same_flow:1; 2295 2296 /* Used in tunnel GRO receive */ 2297 u8 encap_mark:1; 2298 2299 /* GRO checksum is valid */ 2300 u8 csum_valid:1; 2301 2302 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2303 u8 csum_cnt:3; 2304 2305 /* Free the skb? */ 2306 u8 free:2; 2307 #define NAPI_GRO_FREE 1 2308 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2309 2310 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2311 u8 is_ipv6:1; 2312 2313 /* Used in GRE, set in fou/gue_gro_receive */ 2314 u8 is_fou:1; 2315 2316 /* Used to determine if flush_id can be ignored */ 2317 u8 is_atomic:1; 2318 2319 /* Number of gro_receive callbacks this packet already went through */ 2320 u8 recursion_counter:4; 2321 2322 /* 1 bit hole */ 2323 2324 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2325 __wsum csum; 2326 2327 /* used in skb_gro_receive() slow path */ 2328 struct sk_buff *last; 2329 }; 2330 2331 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2332 2333 #define GRO_RECURSION_LIMIT 15 2334 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2335 { 2336 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2337 } 2338 2339 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2340 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2341 struct list_head *head, 2342 struct sk_buff *skb) 2343 { 2344 if (unlikely(gro_recursion_inc_test(skb))) { 2345 NAPI_GRO_CB(skb)->flush |= 1; 2346 return NULL; 2347 } 2348 2349 return cb(head, skb); 2350 } 2351 2352 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2353 struct sk_buff *); 2354 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2355 struct sock *sk, 2356 struct list_head *head, 2357 struct sk_buff *skb) 2358 { 2359 if (unlikely(gro_recursion_inc_test(skb))) { 2360 NAPI_GRO_CB(skb)->flush |= 1; 2361 return NULL; 2362 } 2363 2364 return cb(sk, head, skb); 2365 } 2366 2367 struct packet_type { 2368 __be16 type; /* This is really htons(ether_type). */ 2369 bool ignore_outgoing; 2370 struct net_device *dev; /* NULL is wildcarded here */ 2371 int (*func) (struct sk_buff *, 2372 struct net_device *, 2373 struct packet_type *, 2374 struct net_device *); 2375 void (*list_func) (struct list_head *, 2376 struct packet_type *, 2377 struct net_device *); 2378 bool (*id_match)(struct packet_type *ptype, 2379 struct sock *sk); 2380 void *af_packet_priv; 2381 struct list_head list; 2382 }; 2383 2384 struct offload_callbacks { 2385 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2386 netdev_features_t features); 2387 struct sk_buff *(*gro_receive)(struct list_head *head, 2388 struct sk_buff *skb); 2389 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2390 }; 2391 2392 struct packet_offload { 2393 __be16 type; /* This is really htons(ether_type). */ 2394 u16 priority; 2395 struct offload_callbacks callbacks; 2396 struct list_head list; 2397 }; 2398 2399 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2400 struct pcpu_sw_netstats { 2401 u64 rx_packets; 2402 u64 rx_bytes; 2403 u64 tx_packets; 2404 u64 tx_bytes; 2405 struct u64_stats_sync syncp; 2406 } __aligned(4 * sizeof(u64)); 2407 2408 struct pcpu_lstats { 2409 u64 packets; 2410 u64 bytes; 2411 struct u64_stats_sync syncp; 2412 } __aligned(2 * sizeof(u64)); 2413 2414 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2415 ({ \ 2416 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2417 if (pcpu_stats) { \ 2418 int __cpu; \ 2419 for_each_possible_cpu(__cpu) { \ 2420 typeof(type) *stat; \ 2421 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2422 u64_stats_init(&stat->syncp); \ 2423 } \ 2424 } \ 2425 pcpu_stats; \ 2426 }) 2427 2428 #define netdev_alloc_pcpu_stats(type) \ 2429 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2430 2431 enum netdev_lag_tx_type { 2432 NETDEV_LAG_TX_TYPE_UNKNOWN, 2433 NETDEV_LAG_TX_TYPE_RANDOM, 2434 NETDEV_LAG_TX_TYPE_BROADCAST, 2435 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2436 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2437 NETDEV_LAG_TX_TYPE_HASH, 2438 }; 2439 2440 enum netdev_lag_hash { 2441 NETDEV_LAG_HASH_NONE, 2442 NETDEV_LAG_HASH_L2, 2443 NETDEV_LAG_HASH_L34, 2444 NETDEV_LAG_HASH_L23, 2445 NETDEV_LAG_HASH_E23, 2446 NETDEV_LAG_HASH_E34, 2447 NETDEV_LAG_HASH_UNKNOWN, 2448 }; 2449 2450 struct netdev_lag_upper_info { 2451 enum netdev_lag_tx_type tx_type; 2452 enum netdev_lag_hash hash_type; 2453 }; 2454 2455 struct netdev_lag_lower_state_info { 2456 u8 link_up : 1, 2457 tx_enabled : 1; 2458 }; 2459 2460 #include <linux/notifier.h> 2461 2462 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2463 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2464 * adding new types. 2465 */ 2466 enum netdev_cmd { 2467 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2468 NETDEV_DOWN, 2469 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2470 detected a hardware crash and restarted 2471 - we can use this eg to kick tcp sessions 2472 once done */ 2473 NETDEV_CHANGE, /* Notify device state change */ 2474 NETDEV_REGISTER, 2475 NETDEV_UNREGISTER, 2476 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2477 NETDEV_CHANGEADDR, /* notify after the address change */ 2478 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2479 NETDEV_GOING_DOWN, 2480 NETDEV_CHANGENAME, 2481 NETDEV_FEAT_CHANGE, 2482 NETDEV_BONDING_FAILOVER, 2483 NETDEV_PRE_UP, 2484 NETDEV_PRE_TYPE_CHANGE, 2485 NETDEV_POST_TYPE_CHANGE, 2486 NETDEV_POST_INIT, 2487 NETDEV_RELEASE, 2488 NETDEV_NOTIFY_PEERS, 2489 NETDEV_JOIN, 2490 NETDEV_CHANGEUPPER, 2491 NETDEV_RESEND_IGMP, 2492 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2493 NETDEV_CHANGEINFODATA, 2494 NETDEV_BONDING_INFO, 2495 NETDEV_PRECHANGEUPPER, 2496 NETDEV_CHANGELOWERSTATE, 2497 NETDEV_UDP_TUNNEL_PUSH_INFO, 2498 NETDEV_UDP_TUNNEL_DROP_INFO, 2499 NETDEV_CHANGE_TX_QUEUE_LEN, 2500 NETDEV_CVLAN_FILTER_PUSH_INFO, 2501 NETDEV_CVLAN_FILTER_DROP_INFO, 2502 NETDEV_SVLAN_FILTER_PUSH_INFO, 2503 NETDEV_SVLAN_FILTER_DROP_INFO, 2504 }; 2505 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2506 2507 int register_netdevice_notifier(struct notifier_block *nb); 2508 int unregister_netdevice_notifier(struct notifier_block *nb); 2509 2510 struct netdev_notifier_info { 2511 struct net_device *dev; 2512 struct netlink_ext_ack *extack; 2513 }; 2514 2515 struct netdev_notifier_info_ext { 2516 struct netdev_notifier_info info; /* must be first */ 2517 union { 2518 u32 mtu; 2519 } ext; 2520 }; 2521 2522 struct netdev_notifier_change_info { 2523 struct netdev_notifier_info info; /* must be first */ 2524 unsigned int flags_changed; 2525 }; 2526 2527 struct netdev_notifier_changeupper_info { 2528 struct netdev_notifier_info info; /* must be first */ 2529 struct net_device *upper_dev; /* new upper dev */ 2530 bool master; /* is upper dev master */ 2531 bool linking; /* is the notification for link or unlink */ 2532 void *upper_info; /* upper dev info */ 2533 }; 2534 2535 struct netdev_notifier_changelowerstate_info { 2536 struct netdev_notifier_info info; /* must be first */ 2537 void *lower_state_info; /* is lower dev state */ 2538 }; 2539 2540 struct netdev_notifier_pre_changeaddr_info { 2541 struct netdev_notifier_info info; /* must be first */ 2542 const unsigned char *dev_addr; 2543 }; 2544 2545 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2546 struct net_device *dev) 2547 { 2548 info->dev = dev; 2549 info->extack = NULL; 2550 } 2551 2552 static inline struct net_device * 2553 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2554 { 2555 return info->dev; 2556 } 2557 2558 static inline struct netlink_ext_ack * 2559 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2560 { 2561 return info->extack; 2562 } 2563 2564 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2565 2566 2567 extern rwlock_t dev_base_lock; /* Device list lock */ 2568 2569 #define for_each_netdev(net, d) \ 2570 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2571 #define for_each_netdev_reverse(net, d) \ 2572 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2573 #define for_each_netdev_rcu(net, d) \ 2574 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2575 #define for_each_netdev_safe(net, d, n) \ 2576 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2577 #define for_each_netdev_continue(net, d) \ 2578 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2579 #define for_each_netdev_continue_rcu(net, d) \ 2580 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2581 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2582 for_each_netdev_rcu(&init_net, slave) \ 2583 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2584 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2585 2586 static inline struct net_device *next_net_device(struct net_device *dev) 2587 { 2588 struct list_head *lh; 2589 struct net *net; 2590 2591 net = dev_net(dev); 2592 lh = dev->dev_list.next; 2593 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2594 } 2595 2596 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2597 { 2598 struct list_head *lh; 2599 struct net *net; 2600 2601 net = dev_net(dev); 2602 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2603 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2604 } 2605 2606 static inline struct net_device *first_net_device(struct net *net) 2607 { 2608 return list_empty(&net->dev_base_head) ? NULL : 2609 net_device_entry(net->dev_base_head.next); 2610 } 2611 2612 static inline struct net_device *first_net_device_rcu(struct net *net) 2613 { 2614 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2615 2616 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2617 } 2618 2619 int netdev_boot_setup_check(struct net_device *dev); 2620 unsigned long netdev_boot_base(const char *prefix, int unit); 2621 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2622 const char *hwaddr); 2623 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2624 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2625 void dev_add_pack(struct packet_type *pt); 2626 void dev_remove_pack(struct packet_type *pt); 2627 void __dev_remove_pack(struct packet_type *pt); 2628 void dev_add_offload(struct packet_offload *po); 2629 void dev_remove_offload(struct packet_offload *po); 2630 2631 int dev_get_iflink(const struct net_device *dev); 2632 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2633 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2634 unsigned short mask); 2635 struct net_device *dev_get_by_name(struct net *net, const char *name); 2636 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2637 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2638 int dev_alloc_name(struct net_device *dev, const char *name); 2639 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2640 void dev_close(struct net_device *dev); 2641 void dev_close_many(struct list_head *head, bool unlink); 2642 void dev_disable_lro(struct net_device *dev); 2643 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2644 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2645 struct net_device *sb_dev); 2646 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2647 struct net_device *sb_dev); 2648 int dev_queue_xmit(struct sk_buff *skb); 2649 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2650 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2651 int register_netdevice(struct net_device *dev); 2652 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2653 void unregister_netdevice_many(struct list_head *head); 2654 static inline void unregister_netdevice(struct net_device *dev) 2655 { 2656 unregister_netdevice_queue(dev, NULL); 2657 } 2658 2659 int netdev_refcnt_read(const struct net_device *dev); 2660 void free_netdev(struct net_device *dev); 2661 void netdev_freemem(struct net_device *dev); 2662 void synchronize_net(void); 2663 int init_dummy_netdev(struct net_device *dev); 2664 2665 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2666 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2667 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2668 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2669 int netdev_get_name(struct net *net, char *name, int ifindex); 2670 int dev_restart(struct net_device *dev); 2671 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2672 2673 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2674 { 2675 return NAPI_GRO_CB(skb)->data_offset; 2676 } 2677 2678 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2679 { 2680 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2681 } 2682 2683 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2684 { 2685 NAPI_GRO_CB(skb)->data_offset += len; 2686 } 2687 2688 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2689 unsigned int offset) 2690 { 2691 return NAPI_GRO_CB(skb)->frag0 + offset; 2692 } 2693 2694 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2695 { 2696 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2697 } 2698 2699 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2700 { 2701 NAPI_GRO_CB(skb)->frag0 = NULL; 2702 NAPI_GRO_CB(skb)->frag0_len = 0; 2703 } 2704 2705 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2706 unsigned int offset) 2707 { 2708 if (!pskb_may_pull(skb, hlen)) 2709 return NULL; 2710 2711 skb_gro_frag0_invalidate(skb); 2712 return skb->data + offset; 2713 } 2714 2715 static inline void *skb_gro_network_header(struct sk_buff *skb) 2716 { 2717 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2718 skb_network_offset(skb); 2719 } 2720 2721 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2722 const void *start, unsigned int len) 2723 { 2724 if (NAPI_GRO_CB(skb)->csum_valid) 2725 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2726 csum_partial(start, len, 0)); 2727 } 2728 2729 /* GRO checksum functions. These are logical equivalents of the normal 2730 * checksum functions (in skbuff.h) except that they operate on the GRO 2731 * offsets and fields in sk_buff. 2732 */ 2733 2734 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2735 2736 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2737 { 2738 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2739 } 2740 2741 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2742 bool zero_okay, 2743 __sum16 check) 2744 { 2745 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2746 skb_checksum_start_offset(skb) < 2747 skb_gro_offset(skb)) && 2748 !skb_at_gro_remcsum_start(skb) && 2749 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2750 (!zero_okay || check)); 2751 } 2752 2753 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2754 __wsum psum) 2755 { 2756 if (NAPI_GRO_CB(skb)->csum_valid && 2757 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2758 return 0; 2759 2760 NAPI_GRO_CB(skb)->csum = psum; 2761 2762 return __skb_gro_checksum_complete(skb); 2763 } 2764 2765 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2766 { 2767 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2768 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2769 NAPI_GRO_CB(skb)->csum_cnt--; 2770 } else { 2771 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2772 * verified a new top level checksum or an encapsulated one 2773 * during GRO. This saves work if we fallback to normal path. 2774 */ 2775 __skb_incr_checksum_unnecessary(skb); 2776 } 2777 } 2778 2779 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2780 compute_pseudo) \ 2781 ({ \ 2782 __sum16 __ret = 0; \ 2783 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2784 __ret = __skb_gro_checksum_validate_complete(skb, \ 2785 compute_pseudo(skb, proto)); \ 2786 if (!__ret) \ 2787 skb_gro_incr_csum_unnecessary(skb); \ 2788 __ret; \ 2789 }) 2790 2791 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2792 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2793 2794 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2795 compute_pseudo) \ 2796 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2797 2798 #define skb_gro_checksum_simple_validate(skb) \ 2799 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2800 2801 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2802 { 2803 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2804 !NAPI_GRO_CB(skb)->csum_valid); 2805 } 2806 2807 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2808 __sum16 check, __wsum pseudo) 2809 { 2810 NAPI_GRO_CB(skb)->csum = ~pseudo; 2811 NAPI_GRO_CB(skb)->csum_valid = 1; 2812 } 2813 2814 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2815 do { \ 2816 if (__skb_gro_checksum_convert_check(skb)) \ 2817 __skb_gro_checksum_convert(skb, check, \ 2818 compute_pseudo(skb, proto)); \ 2819 } while (0) 2820 2821 struct gro_remcsum { 2822 int offset; 2823 __wsum delta; 2824 }; 2825 2826 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2827 { 2828 grc->offset = 0; 2829 grc->delta = 0; 2830 } 2831 2832 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2833 unsigned int off, size_t hdrlen, 2834 int start, int offset, 2835 struct gro_remcsum *grc, 2836 bool nopartial) 2837 { 2838 __wsum delta; 2839 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2840 2841 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2842 2843 if (!nopartial) { 2844 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2845 return ptr; 2846 } 2847 2848 ptr = skb_gro_header_fast(skb, off); 2849 if (skb_gro_header_hard(skb, off + plen)) { 2850 ptr = skb_gro_header_slow(skb, off + plen, off); 2851 if (!ptr) 2852 return NULL; 2853 } 2854 2855 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2856 start, offset); 2857 2858 /* Adjust skb->csum since we changed the packet */ 2859 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2860 2861 grc->offset = off + hdrlen + offset; 2862 grc->delta = delta; 2863 2864 return ptr; 2865 } 2866 2867 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2868 struct gro_remcsum *grc) 2869 { 2870 void *ptr; 2871 size_t plen = grc->offset + sizeof(u16); 2872 2873 if (!grc->delta) 2874 return; 2875 2876 ptr = skb_gro_header_fast(skb, grc->offset); 2877 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2878 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2879 if (!ptr) 2880 return; 2881 } 2882 2883 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2884 } 2885 2886 #ifdef CONFIG_XFRM_OFFLOAD 2887 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2888 { 2889 if (PTR_ERR(pp) != -EINPROGRESS) 2890 NAPI_GRO_CB(skb)->flush |= flush; 2891 } 2892 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2893 struct sk_buff *pp, 2894 int flush, 2895 struct gro_remcsum *grc) 2896 { 2897 if (PTR_ERR(pp) != -EINPROGRESS) { 2898 NAPI_GRO_CB(skb)->flush |= flush; 2899 skb_gro_remcsum_cleanup(skb, grc); 2900 skb->remcsum_offload = 0; 2901 } 2902 } 2903 #else 2904 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2905 { 2906 NAPI_GRO_CB(skb)->flush |= flush; 2907 } 2908 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2909 struct sk_buff *pp, 2910 int flush, 2911 struct gro_remcsum *grc) 2912 { 2913 NAPI_GRO_CB(skb)->flush |= flush; 2914 skb_gro_remcsum_cleanup(skb, grc); 2915 skb->remcsum_offload = 0; 2916 } 2917 #endif 2918 2919 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2920 unsigned short type, 2921 const void *daddr, const void *saddr, 2922 unsigned int len) 2923 { 2924 if (!dev->header_ops || !dev->header_ops->create) 2925 return 0; 2926 2927 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2928 } 2929 2930 static inline int dev_parse_header(const struct sk_buff *skb, 2931 unsigned char *haddr) 2932 { 2933 const struct net_device *dev = skb->dev; 2934 2935 if (!dev->header_ops || !dev->header_ops->parse) 2936 return 0; 2937 return dev->header_ops->parse(skb, haddr); 2938 } 2939 2940 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2941 { 2942 const struct net_device *dev = skb->dev; 2943 2944 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2945 return 0; 2946 return dev->header_ops->parse_protocol(skb); 2947 } 2948 2949 /* ll_header must have at least hard_header_len allocated */ 2950 static inline bool dev_validate_header(const struct net_device *dev, 2951 char *ll_header, int len) 2952 { 2953 if (likely(len >= dev->hard_header_len)) 2954 return true; 2955 if (len < dev->min_header_len) 2956 return false; 2957 2958 if (capable(CAP_SYS_RAWIO)) { 2959 memset(ll_header + len, 0, dev->hard_header_len - len); 2960 return true; 2961 } 2962 2963 if (dev->header_ops && dev->header_ops->validate) 2964 return dev->header_ops->validate(ll_header, len); 2965 2966 return false; 2967 } 2968 2969 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2970 int len, int size); 2971 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2972 static inline int unregister_gifconf(unsigned int family) 2973 { 2974 return register_gifconf(family, NULL); 2975 } 2976 2977 #ifdef CONFIG_NET_FLOW_LIMIT 2978 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2979 struct sd_flow_limit { 2980 u64 count; 2981 unsigned int num_buckets; 2982 unsigned int history_head; 2983 u16 history[FLOW_LIMIT_HISTORY]; 2984 u8 buckets[]; 2985 }; 2986 2987 extern int netdev_flow_limit_table_len; 2988 #endif /* CONFIG_NET_FLOW_LIMIT */ 2989 2990 /* 2991 * Incoming packets are placed on per-CPU queues 2992 */ 2993 struct softnet_data { 2994 struct list_head poll_list; 2995 struct sk_buff_head process_queue; 2996 2997 /* stats */ 2998 unsigned int processed; 2999 unsigned int time_squeeze; 3000 unsigned int received_rps; 3001 #ifdef CONFIG_RPS 3002 struct softnet_data *rps_ipi_list; 3003 #endif 3004 #ifdef CONFIG_NET_FLOW_LIMIT 3005 struct sd_flow_limit __rcu *flow_limit; 3006 #endif 3007 struct Qdisc *output_queue; 3008 struct Qdisc **output_queue_tailp; 3009 struct sk_buff *completion_queue; 3010 #ifdef CONFIG_XFRM_OFFLOAD 3011 struct sk_buff_head xfrm_backlog; 3012 #endif 3013 /* written and read only by owning cpu: */ 3014 struct { 3015 u16 recursion; 3016 u8 more; 3017 } xmit; 3018 #ifdef CONFIG_RPS 3019 /* input_queue_head should be written by cpu owning this struct, 3020 * and only read by other cpus. Worth using a cache line. 3021 */ 3022 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3023 3024 /* Elements below can be accessed between CPUs for RPS/RFS */ 3025 call_single_data_t csd ____cacheline_aligned_in_smp; 3026 struct softnet_data *rps_ipi_next; 3027 unsigned int cpu; 3028 unsigned int input_queue_tail; 3029 #endif 3030 unsigned int dropped; 3031 struct sk_buff_head input_pkt_queue; 3032 struct napi_struct backlog; 3033 3034 }; 3035 3036 static inline void input_queue_head_incr(struct softnet_data *sd) 3037 { 3038 #ifdef CONFIG_RPS 3039 sd->input_queue_head++; 3040 #endif 3041 } 3042 3043 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3044 unsigned int *qtail) 3045 { 3046 #ifdef CONFIG_RPS 3047 *qtail = ++sd->input_queue_tail; 3048 #endif 3049 } 3050 3051 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3052 3053 static inline int dev_recursion_level(void) 3054 { 3055 return this_cpu_read(softnet_data.xmit.recursion); 3056 } 3057 3058 #define XMIT_RECURSION_LIMIT 10 3059 static inline bool dev_xmit_recursion(void) 3060 { 3061 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3062 XMIT_RECURSION_LIMIT); 3063 } 3064 3065 static inline void dev_xmit_recursion_inc(void) 3066 { 3067 __this_cpu_inc(softnet_data.xmit.recursion); 3068 } 3069 3070 static inline void dev_xmit_recursion_dec(void) 3071 { 3072 __this_cpu_dec(softnet_data.xmit.recursion); 3073 } 3074 3075 void __netif_schedule(struct Qdisc *q); 3076 void netif_schedule_queue(struct netdev_queue *txq); 3077 3078 static inline void netif_tx_schedule_all(struct net_device *dev) 3079 { 3080 unsigned int i; 3081 3082 for (i = 0; i < dev->num_tx_queues; i++) 3083 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3084 } 3085 3086 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3087 { 3088 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3089 } 3090 3091 /** 3092 * netif_start_queue - allow transmit 3093 * @dev: network device 3094 * 3095 * Allow upper layers to call the device hard_start_xmit routine. 3096 */ 3097 static inline void netif_start_queue(struct net_device *dev) 3098 { 3099 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3100 } 3101 3102 static inline void netif_tx_start_all_queues(struct net_device *dev) 3103 { 3104 unsigned int i; 3105 3106 for (i = 0; i < dev->num_tx_queues; i++) { 3107 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3108 netif_tx_start_queue(txq); 3109 } 3110 } 3111 3112 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3113 3114 /** 3115 * netif_wake_queue - restart transmit 3116 * @dev: network device 3117 * 3118 * Allow upper layers to call the device hard_start_xmit routine. 3119 * Used for flow control when transmit resources are available. 3120 */ 3121 static inline void netif_wake_queue(struct net_device *dev) 3122 { 3123 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3124 } 3125 3126 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3127 { 3128 unsigned int i; 3129 3130 for (i = 0; i < dev->num_tx_queues; i++) { 3131 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3132 netif_tx_wake_queue(txq); 3133 } 3134 } 3135 3136 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3137 { 3138 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3139 } 3140 3141 /** 3142 * netif_stop_queue - stop transmitted packets 3143 * @dev: network device 3144 * 3145 * Stop upper layers calling the device hard_start_xmit routine. 3146 * Used for flow control when transmit resources are unavailable. 3147 */ 3148 static inline void netif_stop_queue(struct net_device *dev) 3149 { 3150 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3151 } 3152 3153 void netif_tx_stop_all_queues(struct net_device *dev); 3154 3155 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3156 { 3157 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3158 } 3159 3160 /** 3161 * netif_queue_stopped - test if transmit queue is flowblocked 3162 * @dev: network device 3163 * 3164 * Test if transmit queue on device is currently unable to send. 3165 */ 3166 static inline bool netif_queue_stopped(const struct net_device *dev) 3167 { 3168 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3169 } 3170 3171 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3172 { 3173 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3174 } 3175 3176 static inline bool 3177 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3178 { 3179 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3180 } 3181 3182 static inline bool 3183 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3184 { 3185 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3186 } 3187 3188 /** 3189 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3190 * @dev_queue: pointer to transmit queue 3191 * 3192 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3193 * to give appropriate hint to the CPU. 3194 */ 3195 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3196 { 3197 #ifdef CONFIG_BQL 3198 prefetchw(&dev_queue->dql.num_queued); 3199 #endif 3200 } 3201 3202 /** 3203 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3204 * @dev_queue: pointer to transmit queue 3205 * 3206 * BQL enabled drivers might use this helper in their TX completion path, 3207 * to give appropriate hint to the CPU. 3208 */ 3209 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3210 { 3211 #ifdef CONFIG_BQL 3212 prefetchw(&dev_queue->dql.limit); 3213 #endif 3214 } 3215 3216 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3217 unsigned int bytes) 3218 { 3219 #ifdef CONFIG_BQL 3220 dql_queued(&dev_queue->dql, bytes); 3221 3222 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3223 return; 3224 3225 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3226 3227 /* 3228 * The XOFF flag must be set before checking the dql_avail below, 3229 * because in netdev_tx_completed_queue we update the dql_completed 3230 * before checking the XOFF flag. 3231 */ 3232 smp_mb(); 3233 3234 /* check again in case another CPU has just made room avail */ 3235 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3236 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3237 #endif 3238 } 3239 3240 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3241 * that they should not test BQL status themselves. 3242 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3243 * skb of a batch. 3244 * Returns true if the doorbell must be used to kick the NIC. 3245 */ 3246 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3247 unsigned int bytes, 3248 bool xmit_more) 3249 { 3250 if (xmit_more) { 3251 #ifdef CONFIG_BQL 3252 dql_queued(&dev_queue->dql, bytes); 3253 #endif 3254 return netif_tx_queue_stopped(dev_queue); 3255 } 3256 netdev_tx_sent_queue(dev_queue, bytes); 3257 return true; 3258 } 3259 3260 /** 3261 * netdev_sent_queue - report the number of bytes queued to hardware 3262 * @dev: network device 3263 * @bytes: number of bytes queued to the hardware device queue 3264 * 3265 * Report the number of bytes queued for sending/completion to the network 3266 * device hardware queue. @bytes should be a good approximation and should 3267 * exactly match netdev_completed_queue() @bytes 3268 */ 3269 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3270 { 3271 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3272 } 3273 3274 static inline bool __netdev_sent_queue(struct net_device *dev, 3275 unsigned int bytes, 3276 bool xmit_more) 3277 { 3278 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3279 xmit_more); 3280 } 3281 3282 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3283 unsigned int pkts, unsigned int bytes) 3284 { 3285 #ifdef CONFIG_BQL 3286 if (unlikely(!bytes)) 3287 return; 3288 3289 dql_completed(&dev_queue->dql, bytes); 3290 3291 /* 3292 * Without the memory barrier there is a small possiblity that 3293 * netdev_tx_sent_queue will miss the update and cause the queue to 3294 * be stopped forever 3295 */ 3296 smp_mb(); 3297 3298 if (dql_avail(&dev_queue->dql) < 0) 3299 return; 3300 3301 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3302 netif_schedule_queue(dev_queue); 3303 #endif 3304 } 3305 3306 /** 3307 * netdev_completed_queue - report bytes and packets completed by device 3308 * @dev: network device 3309 * @pkts: actual number of packets sent over the medium 3310 * @bytes: actual number of bytes sent over the medium 3311 * 3312 * Report the number of bytes and packets transmitted by the network device 3313 * hardware queue over the physical medium, @bytes must exactly match the 3314 * @bytes amount passed to netdev_sent_queue() 3315 */ 3316 static inline void netdev_completed_queue(struct net_device *dev, 3317 unsigned int pkts, unsigned int bytes) 3318 { 3319 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3320 } 3321 3322 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3323 { 3324 #ifdef CONFIG_BQL 3325 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3326 dql_reset(&q->dql); 3327 #endif 3328 } 3329 3330 /** 3331 * netdev_reset_queue - reset the packets and bytes count of a network device 3332 * @dev_queue: network device 3333 * 3334 * Reset the bytes and packet count of a network device and clear the 3335 * software flow control OFF bit for this network device 3336 */ 3337 static inline void netdev_reset_queue(struct net_device *dev_queue) 3338 { 3339 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3340 } 3341 3342 /** 3343 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3344 * @dev: network device 3345 * @queue_index: given tx queue index 3346 * 3347 * Returns 0 if given tx queue index >= number of device tx queues, 3348 * otherwise returns the originally passed tx queue index. 3349 */ 3350 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3351 { 3352 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3353 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3354 dev->name, queue_index, 3355 dev->real_num_tx_queues); 3356 return 0; 3357 } 3358 3359 return queue_index; 3360 } 3361 3362 /** 3363 * netif_running - test if up 3364 * @dev: network device 3365 * 3366 * Test if the device has been brought up. 3367 */ 3368 static inline bool netif_running(const struct net_device *dev) 3369 { 3370 return test_bit(__LINK_STATE_START, &dev->state); 3371 } 3372 3373 /* 3374 * Routines to manage the subqueues on a device. We only need start, 3375 * stop, and a check if it's stopped. All other device management is 3376 * done at the overall netdevice level. 3377 * Also test the device if we're multiqueue. 3378 */ 3379 3380 /** 3381 * netif_start_subqueue - allow sending packets on subqueue 3382 * @dev: network device 3383 * @queue_index: sub queue index 3384 * 3385 * Start individual transmit queue of a device with multiple transmit queues. 3386 */ 3387 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3388 { 3389 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3390 3391 netif_tx_start_queue(txq); 3392 } 3393 3394 /** 3395 * netif_stop_subqueue - stop sending packets on subqueue 3396 * @dev: network device 3397 * @queue_index: sub queue index 3398 * 3399 * Stop individual transmit queue of a device with multiple transmit queues. 3400 */ 3401 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3402 { 3403 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3404 netif_tx_stop_queue(txq); 3405 } 3406 3407 /** 3408 * netif_subqueue_stopped - test status of subqueue 3409 * @dev: network device 3410 * @queue_index: sub queue index 3411 * 3412 * Check individual transmit queue of a device with multiple transmit queues. 3413 */ 3414 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3415 u16 queue_index) 3416 { 3417 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3418 3419 return netif_tx_queue_stopped(txq); 3420 } 3421 3422 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3423 struct sk_buff *skb) 3424 { 3425 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3426 } 3427 3428 /** 3429 * netif_wake_subqueue - allow sending packets on subqueue 3430 * @dev: network device 3431 * @queue_index: sub queue index 3432 * 3433 * Resume individual transmit queue of a device with multiple transmit queues. 3434 */ 3435 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3436 { 3437 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3438 3439 netif_tx_wake_queue(txq); 3440 } 3441 3442 #ifdef CONFIG_XPS 3443 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3444 u16 index); 3445 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3446 u16 index, bool is_rxqs_map); 3447 3448 /** 3449 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3450 * @j: CPU/Rx queue index 3451 * @mask: bitmask of all cpus/rx queues 3452 * @nr_bits: number of bits in the bitmask 3453 * 3454 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3455 */ 3456 static inline bool netif_attr_test_mask(unsigned long j, 3457 const unsigned long *mask, 3458 unsigned int nr_bits) 3459 { 3460 cpu_max_bits_warn(j, nr_bits); 3461 return test_bit(j, mask); 3462 } 3463 3464 /** 3465 * netif_attr_test_online - Test for online CPU/Rx queue 3466 * @j: CPU/Rx queue index 3467 * @online_mask: bitmask for CPUs/Rx queues that are online 3468 * @nr_bits: number of bits in the bitmask 3469 * 3470 * Returns true if a CPU/Rx queue is online. 3471 */ 3472 static inline bool netif_attr_test_online(unsigned long j, 3473 const unsigned long *online_mask, 3474 unsigned int nr_bits) 3475 { 3476 cpu_max_bits_warn(j, nr_bits); 3477 3478 if (online_mask) 3479 return test_bit(j, online_mask); 3480 3481 return (j < nr_bits); 3482 } 3483 3484 /** 3485 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3486 * @n: CPU/Rx queue index 3487 * @srcp: the cpumask/Rx queue mask pointer 3488 * @nr_bits: number of bits in the bitmask 3489 * 3490 * Returns >= nr_bits if no further CPUs/Rx queues set. 3491 */ 3492 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3493 unsigned int nr_bits) 3494 { 3495 /* -1 is a legal arg here. */ 3496 if (n != -1) 3497 cpu_max_bits_warn(n, nr_bits); 3498 3499 if (srcp) 3500 return find_next_bit(srcp, nr_bits, n + 1); 3501 3502 return n + 1; 3503 } 3504 3505 /** 3506 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3507 * @n: CPU/Rx queue index 3508 * @src1p: the first CPUs/Rx queues mask pointer 3509 * @src2p: the second CPUs/Rx queues mask pointer 3510 * @nr_bits: number of bits in the bitmask 3511 * 3512 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3513 */ 3514 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3515 const unsigned long *src2p, 3516 unsigned int nr_bits) 3517 { 3518 /* -1 is a legal arg here. */ 3519 if (n != -1) 3520 cpu_max_bits_warn(n, nr_bits); 3521 3522 if (src1p && src2p) 3523 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3524 else if (src1p) 3525 return find_next_bit(src1p, nr_bits, n + 1); 3526 else if (src2p) 3527 return find_next_bit(src2p, nr_bits, n + 1); 3528 3529 return n + 1; 3530 } 3531 #else 3532 static inline int netif_set_xps_queue(struct net_device *dev, 3533 const struct cpumask *mask, 3534 u16 index) 3535 { 3536 return 0; 3537 } 3538 3539 static inline int __netif_set_xps_queue(struct net_device *dev, 3540 const unsigned long *mask, 3541 u16 index, bool is_rxqs_map) 3542 { 3543 return 0; 3544 } 3545 #endif 3546 3547 /** 3548 * netif_is_multiqueue - test if device has multiple transmit queues 3549 * @dev: network device 3550 * 3551 * Check if device has multiple transmit queues 3552 */ 3553 static inline bool netif_is_multiqueue(const struct net_device *dev) 3554 { 3555 return dev->num_tx_queues > 1; 3556 } 3557 3558 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3559 3560 #ifdef CONFIG_SYSFS 3561 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3562 #else 3563 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3564 unsigned int rxqs) 3565 { 3566 dev->real_num_rx_queues = rxqs; 3567 return 0; 3568 } 3569 #endif 3570 3571 static inline struct netdev_rx_queue * 3572 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3573 { 3574 return dev->_rx + rxq; 3575 } 3576 3577 #ifdef CONFIG_SYSFS 3578 static inline unsigned int get_netdev_rx_queue_index( 3579 struct netdev_rx_queue *queue) 3580 { 3581 struct net_device *dev = queue->dev; 3582 int index = queue - dev->_rx; 3583 3584 BUG_ON(index >= dev->num_rx_queues); 3585 return index; 3586 } 3587 #endif 3588 3589 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3590 int netif_get_num_default_rss_queues(void); 3591 3592 enum skb_free_reason { 3593 SKB_REASON_CONSUMED, 3594 SKB_REASON_DROPPED, 3595 }; 3596 3597 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3598 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3599 3600 /* 3601 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3602 * interrupt context or with hardware interrupts being disabled. 3603 * (in_irq() || irqs_disabled()) 3604 * 3605 * We provide four helpers that can be used in following contexts : 3606 * 3607 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3608 * replacing kfree_skb(skb) 3609 * 3610 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3611 * Typically used in place of consume_skb(skb) in TX completion path 3612 * 3613 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3614 * replacing kfree_skb(skb) 3615 * 3616 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3617 * and consumed a packet. Used in place of consume_skb(skb) 3618 */ 3619 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3620 { 3621 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3622 } 3623 3624 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3625 { 3626 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3627 } 3628 3629 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3630 { 3631 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3632 } 3633 3634 static inline void dev_consume_skb_any(struct sk_buff *skb) 3635 { 3636 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3637 } 3638 3639 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3640 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3641 int netif_rx(struct sk_buff *skb); 3642 int netif_rx_ni(struct sk_buff *skb); 3643 int netif_receive_skb(struct sk_buff *skb); 3644 int netif_receive_skb_core(struct sk_buff *skb); 3645 void netif_receive_skb_list(struct list_head *head); 3646 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3647 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3648 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3649 gro_result_t napi_gro_frags(struct napi_struct *napi); 3650 struct packet_offload *gro_find_receive_by_type(__be16 type); 3651 struct packet_offload *gro_find_complete_by_type(__be16 type); 3652 3653 static inline void napi_free_frags(struct napi_struct *napi) 3654 { 3655 kfree_skb(napi->skb); 3656 napi->skb = NULL; 3657 } 3658 3659 bool netdev_is_rx_handler_busy(struct net_device *dev); 3660 int netdev_rx_handler_register(struct net_device *dev, 3661 rx_handler_func_t *rx_handler, 3662 void *rx_handler_data); 3663 void netdev_rx_handler_unregister(struct net_device *dev); 3664 3665 bool dev_valid_name(const char *name); 3666 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3667 bool *need_copyout); 3668 int dev_ifconf(struct net *net, struct ifconf *, int); 3669 int dev_ethtool(struct net *net, struct ifreq *); 3670 unsigned int dev_get_flags(const struct net_device *); 3671 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3672 struct netlink_ext_ack *extack); 3673 int dev_change_flags(struct net_device *dev, unsigned int flags, 3674 struct netlink_ext_ack *extack); 3675 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3676 unsigned int gchanges); 3677 int dev_change_name(struct net_device *, const char *); 3678 int dev_set_alias(struct net_device *, const char *, size_t); 3679 int dev_get_alias(const struct net_device *, char *, size_t); 3680 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3681 int __dev_set_mtu(struct net_device *, int); 3682 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3683 struct netlink_ext_ack *extack); 3684 int dev_set_mtu(struct net_device *, int); 3685 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3686 void dev_set_group(struct net_device *, int); 3687 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3688 struct netlink_ext_ack *extack); 3689 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3690 struct netlink_ext_ack *extack); 3691 int dev_change_carrier(struct net_device *, bool new_carrier); 3692 int dev_get_phys_port_id(struct net_device *dev, 3693 struct netdev_phys_item_id *ppid); 3694 int dev_get_phys_port_name(struct net_device *dev, 3695 char *name, size_t len); 3696 int dev_get_port_parent_id(struct net_device *dev, 3697 struct netdev_phys_item_id *ppid, bool recurse); 3698 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3699 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3700 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3701 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3702 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3703 struct netdev_queue *txq, int *ret); 3704 3705 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3706 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3707 int fd, u32 flags); 3708 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3709 enum bpf_netdev_command cmd); 3710 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3711 3712 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3713 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3714 bool is_skb_forwardable(const struct net_device *dev, 3715 const struct sk_buff *skb); 3716 3717 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3718 struct sk_buff *skb) 3719 { 3720 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3721 unlikely(!is_skb_forwardable(dev, skb))) { 3722 atomic_long_inc(&dev->rx_dropped); 3723 kfree_skb(skb); 3724 return NET_RX_DROP; 3725 } 3726 3727 skb_scrub_packet(skb, true); 3728 skb->priority = 0; 3729 return 0; 3730 } 3731 3732 bool dev_nit_active(struct net_device *dev); 3733 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3734 3735 extern int netdev_budget; 3736 extern unsigned int netdev_budget_usecs; 3737 3738 /* Called by rtnetlink.c:rtnl_unlock() */ 3739 void netdev_run_todo(void); 3740 3741 /** 3742 * dev_put - release reference to device 3743 * @dev: network device 3744 * 3745 * Release reference to device to allow it to be freed. 3746 */ 3747 static inline void dev_put(struct net_device *dev) 3748 { 3749 this_cpu_dec(*dev->pcpu_refcnt); 3750 } 3751 3752 /** 3753 * dev_hold - get reference to device 3754 * @dev: network device 3755 * 3756 * Hold reference to device to keep it from being freed. 3757 */ 3758 static inline void dev_hold(struct net_device *dev) 3759 { 3760 this_cpu_inc(*dev->pcpu_refcnt); 3761 } 3762 3763 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3764 * and _off may be called from IRQ context, but it is caller 3765 * who is responsible for serialization of these calls. 3766 * 3767 * The name carrier is inappropriate, these functions should really be 3768 * called netif_lowerlayer_*() because they represent the state of any 3769 * kind of lower layer not just hardware media. 3770 */ 3771 3772 void linkwatch_init_dev(struct net_device *dev); 3773 void linkwatch_fire_event(struct net_device *dev); 3774 void linkwatch_forget_dev(struct net_device *dev); 3775 3776 /** 3777 * netif_carrier_ok - test if carrier present 3778 * @dev: network device 3779 * 3780 * Check if carrier is present on device 3781 */ 3782 static inline bool netif_carrier_ok(const struct net_device *dev) 3783 { 3784 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3785 } 3786 3787 unsigned long dev_trans_start(struct net_device *dev); 3788 3789 void __netdev_watchdog_up(struct net_device *dev); 3790 3791 void netif_carrier_on(struct net_device *dev); 3792 3793 void netif_carrier_off(struct net_device *dev); 3794 3795 /** 3796 * netif_dormant_on - mark device as dormant. 3797 * @dev: network device 3798 * 3799 * Mark device as dormant (as per RFC2863). 3800 * 3801 * The dormant state indicates that the relevant interface is not 3802 * actually in a condition to pass packets (i.e., it is not 'up') but is 3803 * in a "pending" state, waiting for some external event. For "on- 3804 * demand" interfaces, this new state identifies the situation where the 3805 * interface is waiting for events to place it in the up state. 3806 */ 3807 static inline void netif_dormant_on(struct net_device *dev) 3808 { 3809 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3810 linkwatch_fire_event(dev); 3811 } 3812 3813 /** 3814 * netif_dormant_off - set device as not dormant. 3815 * @dev: network device 3816 * 3817 * Device is not in dormant state. 3818 */ 3819 static inline void netif_dormant_off(struct net_device *dev) 3820 { 3821 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3822 linkwatch_fire_event(dev); 3823 } 3824 3825 /** 3826 * netif_dormant - test if device is dormant 3827 * @dev: network device 3828 * 3829 * Check if device is dormant. 3830 */ 3831 static inline bool netif_dormant(const struct net_device *dev) 3832 { 3833 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3834 } 3835 3836 3837 /** 3838 * netif_oper_up - test if device is operational 3839 * @dev: network device 3840 * 3841 * Check if carrier is operational 3842 */ 3843 static inline bool netif_oper_up(const struct net_device *dev) 3844 { 3845 return (dev->operstate == IF_OPER_UP || 3846 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3847 } 3848 3849 /** 3850 * netif_device_present - is device available or removed 3851 * @dev: network device 3852 * 3853 * Check if device has not been removed from system. 3854 */ 3855 static inline bool netif_device_present(struct net_device *dev) 3856 { 3857 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3858 } 3859 3860 void netif_device_detach(struct net_device *dev); 3861 3862 void netif_device_attach(struct net_device *dev); 3863 3864 /* 3865 * Network interface message level settings 3866 */ 3867 3868 enum { 3869 NETIF_MSG_DRV = 0x0001, 3870 NETIF_MSG_PROBE = 0x0002, 3871 NETIF_MSG_LINK = 0x0004, 3872 NETIF_MSG_TIMER = 0x0008, 3873 NETIF_MSG_IFDOWN = 0x0010, 3874 NETIF_MSG_IFUP = 0x0020, 3875 NETIF_MSG_RX_ERR = 0x0040, 3876 NETIF_MSG_TX_ERR = 0x0080, 3877 NETIF_MSG_TX_QUEUED = 0x0100, 3878 NETIF_MSG_INTR = 0x0200, 3879 NETIF_MSG_TX_DONE = 0x0400, 3880 NETIF_MSG_RX_STATUS = 0x0800, 3881 NETIF_MSG_PKTDATA = 0x1000, 3882 NETIF_MSG_HW = 0x2000, 3883 NETIF_MSG_WOL = 0x4000, 3884 }; 3885 3886 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3887 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3888 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3889 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3890 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3891 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3892 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3893 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3894 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3895 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3896 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3897 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3898 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3899 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3900 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3901 3902 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3903 { 3904 /* use default */ 3905 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3906 return default_msg_enable_bits; 3907 if (debug_value == 0) /* no output */ 3908 return 0; 3909 /* set low N bits */ 3910 return (1U << debug_value) - 1; 3911 } 3912 3913 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3914 { 3915 spin_lock(&txq->_xmit_lock); 3916 txq->xmit_lock_owner = cpu; 3917 } 3918 3919 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3920 { 3921 __acquire(&txq->_xmit_lock); 3922 return true; 3923 } 3924 3925 static inline void __netif_tx_release(struct netdev_queue *txq) 3926 { 3927 __release(&txq->_xmit_lock); 3928 } 3929 3930 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3931 { 3932 spin_lock_bh(&txq->_xmit_lock); 3933 txq->xmit_lock_owner = smp_processor_id(); 3934 } 3935 3936 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3937 { 3938 bool ok = spin_trylock(&txq->_xmit_lock); 3939 if (likely(ok)) 3940 txq->xmit_lock_owner = smp_processor_id(); 3941 return ok; 3942 } 3943 3944 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3945 { 3946 txq->xmit_lock_owner = -1; 3947 spin_unlock(&txq->_xmit_lock); 3948 } 3949 3950 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3951 { 3952 txq->xmit_lock_owner = -1; 3953 spin_unlock_bh(&txq->_xmit_lock); 3954 } 3955 3956 static inline void txq_trans_update(struct netdev_queue *txq) 3957 { 3958 if (txq->xmit_lock_owner != -1) 3959 txq->trans_start = jiffies; 3960 } 3961 3962 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3963 static inline void netif_trans_update(struct net_device *dev) 3964 { 3965 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3966 3967 if (txq->trans_start != jiffies) 3968 txq->trans_start = jiffies; 3969 } 3970 3971 /** 3972 * netif_tx_lock - grab network device transmit lock 3973 * @dev: network device 3974 * 3975 * Get network device transmit lock 3976 */ 3977 static inline void netif_tx_lock(struct net_device *dev) 3978 { 3979 unsigned int i; 3980 int cpu; 3981 3982 spin_lock(&dev->tx_global_lock); 3983 cpu = smp_processor_id(); 3984 for (i = 0; i < dev->num_tx_queues; i++) { 3985 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3986 3987 /* We are the only thread of execution doing a 3988 * freeze, but we have to grab the _xmit_lock in 3989 * order to synchronize with threads which are in 3990 * the ->hard_start_xmit() handler and already 3991 * checked the frozen bit. 3992 */ 3993 __netif_tx_lock(txq, cpu); 3994 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3995 __netif_tx_unlock(txq); 3996 } 3997 } 3998 3999 static inline void netif_tx_lock_bh(struct net_device *dev) 4000 { 4001 local_bh_disable(); 4002 netif_tx_lock(dev); 4003 } 4004 4005 static inline void netif_tx_unlock(struct net_device *dev) 4006 { 4007 unsigned int i; 4008 4009 for (i = 0; i < dev->num_tx_queues; i++) { 4010 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4011 4012 /* No need to grab the _xmit_lock here. If the 4013 * queue is not stopped for another reason, we 4014 * force a schedule. 4015 */ 4016 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4017 netif_schedule_queue(txq); 4018 } 4019 spin_unlock(&dev->tx_global_lock); 4020 } 4021 4022 static inline void netif_tx_unlock_bh(struct net_device *dev) 4023 { 4024 netif_tx_unlock(dev); 4025 local_bh_enable(); 4026 } 4027 4028 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4029 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4030 __netif_tx_lock(txq, cpu); \ 4031 } else { \ 4032 __netif_tx_acquire(txq); \ 4033 } \ 4034 } 4035 4036 #define HARD_TX_TRYLOCK(dev, txq) \ 4037 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4038 __netif_tx_trylock(txq) : \ 4039 __netif_tx_acquire(txq)) 4040 4041 #define HARD_TX_UNLOCK(dev, txq) { \ 4042 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4043 __netif_tx_unlock(txq); \ 4044 } else { \ 4045 __netif_tx_release(txq); \ 4046 } \ 4047 } 4048 4049 static inline void netif_tx_disable(struct net_device *dev) 4050 { 4051 unsigned int i; 4052 int cpu; 4053 4054 local_bh_disable(); 4055 cpu = smp_processor_id(); 4056 for (i = 0; i < dev->num_tx_queues; i++) { 4057 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4058 4059 __netif_tx_lock(txq, cpu); 4060 netif_tx_stop_queue(txq); 4061 __netif_tx_unlock(txq); 4062 } 4063 local_bh_enable(); 4064 } 4065 4066 static inline void netif_addr_lock(struct net_device *dev) 4067 { 4068 spin_lock(&dev->addr_list_lock); 4069 } 4070 4071 static inline void netif_addr_lock_nested(struct net_device *dev) 4072 { 4073 int subclass = SINGLE_DEPTH_NESTING; 4074 4075 if (dev->netdev_ops->ndo_get_lock_subclass) 4076 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 4077 4078 spin_lock_nested(&dev->addr_list_lock, subclass); 4079 } 4080 4081 static inline void netif_addr_lock_bh(struct net_device *dev) 4082 { 4083 spin_lock_bh(&dev->addr_list_lock); 4084 } 4085 4086 static inline void netif_addr_unlock(struct net_device *dev) 4087 { 4088 spin_unlock(&dev->addr_list_lock); 4089 } 4090 4091 static inline void netif_addr_unlock_bh(struct net_device *dev) 4092 { 4093 spin_unlock_bh(&dev->addr_list_lock); 4094 } 4095 4096 /* 4097 * dev_addrs walker. Should be used only for read access. Call with 4098 * rcu_read_lock held. 4099 */ 4100 #define for_each_dev_addr(dev, ha) \ 4101 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4102 4103 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4104 4105 void ether_setup(struct net_device *dev); 4106 4107 /* Support for loadable net-drivers */ 4108 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4109 unsigned char name_assign_type, 4110 void (*setup)(struct net_device *), 4111 unsigned int txqs, unsigned int rxqs); 4112 int dev_get_valid_name(struct net *net, struct net_device *dev, 4113 const char *name); 4114 4115 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4116 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4117 4118 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4119 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4120 count) 4121 4122 int register_netdev(struct net_device *dev); 4123 void unregister_netdev(struct net_device *dev); 4124 4125 /* General hardware address lists handling functions */ 4126 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4127 struct netdev_hw_addr_list *from_list, int addr_len); 4128 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4129 struct netdev_hw_addr_list *from_list, int addr_len); 4130 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4131 struct net_device *dev, 4132 int (*sync)(struct net_device *, const unsigned char *), 4133 int (*unsync)(struct net_device *, 4134 const unsigned char *)); 4135 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4136 struct net_device *dev, 4137 int (*sync)(struct net_device *, 4138 const unsigned char *, int), 4139 int (*unsync)(struct net_device *, 4140 const unsigned char *, int)); 4141 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4142 struct net_device *dev, 4143 int (*unsync)(struct net_device *, 4144 const unsigned char *, int)); 4145 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4146 struct net_device *dev, 4147 int (*unsync)(struct net_device *, 4148 const unsigned char *)); 4149 void __hw_addr_init(struct netdev_hw_addr_list *list); 4150 4151 /* Functions used for device addresses handling */ 4152 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4153 unsigned char addr_type); 4154 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4155 unsigned char addr_type); 4156 void dev_addr_flush(struct net_device *dev); 4157 int dev_addr_init(struct net_device *dev); 4158 4159 /* Functions used for unicast addresses handling */ 4160 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4161 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4162 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4163 int dev_uc_sync(struct net_device *to, struct net_device *from); 4164 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4165 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4166 void dev_uc_flush(struct net_device *dev); 4167 void dev_uc_init(struct net_device *dev); 4168 4169 /** 4170 * __dev_uc_sync - Synchonize device's unicast list 4171 * @dev: device to sync 4172 * @sync: function to call if address should be added 4173 * @unsync: function to call if address should be removed 4174 * 4175 * Add newly added addresses to the interface, and release 4176 * addresses that have been deleted. 4177 */ 4178 static inline int __dev_uc_sync(struct net_device *dev, 4179 int (*sync)(struct net_device *, 4180 const unsigned char *), 4181 int (*unsync)(struct net_device *, 4182 const unsigned char *)) 4183 { 4184 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4185 } 4186 4187 /** 4188 * __dev_uc_unsync - Remove synchronized addresses from device 4189 * @dev: device to sync 4190 * @unsync: function to call if address should be removed 4191 * 4192 * Remove all addresses that were added to the device by dev_uc_sync(). 4193 */ 4194 static inline void __dev_uc_unsync(struct net_device *dev, 4195 int (*unsync)(struct net_device *, 4196 const unsigned char *)) 4197 { 4198 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4199 } 4200 4201 /* Functions used for multicast addresses handling */ 4202 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4203 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4204 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4205 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4206 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4207 int dev_mc_sync(struct net_device *to, struct net_device *from); 4208 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4209 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4210 void dev_mc_flush(struct net_device *dev); 4211 void dev_mc_init(struct net_device *dev); 4212 4213 /** 4214 * __dev_mc_sync - Synchonize device's multicast list 4215 * @dev: device to sync 4216 * @sync: function to call if address should be added 4217 * @unsync: function to call if address should be removed 4218 * 4219 * Add newly added addresses to the interface, and release 4220 * addresses that have been deleted. 4221 */ 4222 static inline int __dev_mc_sync(struct net_device *dev, 4223 int (*sync)(struct net_device *, 4224 const unsigned char *), 4225 int (*unsync)(struct net_device *, 4226 const unsigned char *)) 4227 { 4228 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4229 } 4230 4231 /** 4232 * __dev_mc_unsync - Remove synchronized addresses from device 4233 * @dev: device to sync 4234 * @unsync: function to call if address should be removed 4235 * 4236 * Remove all addresses that were added to the device by dev_mc_sync(). 4237 */ 4238 static inline void __dev_mc_unsync(struct net_device *dev, 4239 int (*unsync)(struct net_device *, 4240 const unsigned char *)) 4241 { 4242 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4243 } 4244 4245 /* Functions used for secondary unicast and multicast support */ 4246 void dev_set_rx_mode(struct net_device *dev); 4247 void __dev_set_rx_mode(struct net_device *dev); 4248 int dev_set_promiscuity(struct net_device *dev, int inc); 4249 int dev_set_allmulti(struct net_device *dev, int inc); 4250 void netdev_state_change(struct net_device *dev); 4251 void netdev_notify_peers(struct net_device *dev); 4252 void netdev_features_change(struct net_device *dev); 4253 /* Load a device via the kmod */ 4254 void dev_load(struct net *net, const char *name); 4255 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4256 struct rtnl_link_stats64 *storage); 4257 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4258 const struct net_device_stats *netdev_stats); 4259 4260 extern int netdev_max_backlog; 4261 extern int netdev_tstamp_prequeue; 4262 extern int weight_p; 4263 extern int dev_weight_rx_bias; 4264 extern int dev_weight_tx_bias; 4265 extern int dev_rx_weight; 4266 extern int dev_tx_weight; 4267 4268 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4269 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4270 struct list_head **iter); 4271 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4272 struct list_head **iter); 4273 4274 /* iterate through upper list, must be called under RCU read lock */ 4275 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4276 for (iter = &(dev)->adj_list.upper, \ 4277 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4278 updev; \ 4279 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4280 4281 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4282 int (*fn)(struct net_device *upper_dev, 4283 void *data), 4284 void *data); 4285 4286 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4287 struct net_device *upper_dev); 4288 4289 bool netdev_has_any_upper_dev(struct net_device *dev); 4290 4291 void *netdev_lower_get_next_private(struct net_device *dev, 4292 struct list_head **iter); 4293 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4294 struct list_head **iter); 4295 4296 #define netdev_for_each_lower_private(dev, priv, iter) \ 4297 for (iter = (dev)->adj_list.lower.next, \ 4298 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4299 priv; \ 4300 priv = netdev_lower_get_next_private(dev, &(iter))) 4301 4302 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4303 for (iter = &(dev)->adj_list.lower, \ 4304 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4305 priv; \ 4306 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4307 4308 void *netdev_lower_get_next(struct net_device *dev, 4309 struct list_head **iter); 4310 4311 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4312 for (iter = (dev)->adj_list.lower.next, \ 4313 ldev = netdev_lower_get_next(dev, &(iter)); \ 4314 ldev; \ 4315 ldev = netdev_lower_get_next(dev, &(iter))) 4316 4317 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4318 struct list_head **iter); 4319 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4320 struct list_head **iter); 4321 4322 int netdev_walk_all_lower_dev(struct net_device *dev, 4323 int (*fn)(struct net_device *lower_dev, 4324 void *data), 4325 void *data); 4326 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4327 int (*fn)(struct net_device *lower_dev, 4328 void *data), 4329 void *data); 4330 4331 void *netdev_adjacent_get_private(struct list_head *adj_list); 4332 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4333 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4334 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4335 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4336 struct netlink_ext_ack *extack); 4337 int netdev_master_upper_dev_link(struct net_device *dev, 4338 struct net_device *upper_dev, 4339 void *upper_priv, void *upper_info, 4340 struct netlink_ext_ack *extack); 4341 void netdev_upper_dev_unlink(struct net_device *dev, 4342 struct net_device *upper_dev); 4343 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4344 void *netdev_lower_dev_get_private(struct net_device *dev, 4345 struct net_device *lower_dev); 4346 void netdev_lower_state_changed(struct net_device *lower_dev, 4347 void *lower_state_info); 4348 4349 /* RSS keys are 40 or 52 bytes long */ 4350 #define NETDEV_RSS_KEY_LEN 52 4351 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4352 void netdev_rss_key_fill(void *buffer, size_t len); 4353 4354 int dev_get_nest_level(struct net_device *dev); 4355 int skb_checksum_help(struct sk_buff *skb); 4356 int skb_crc32c_csum_help(struct sk_buff *skb); 4357 int skb_csum_hwoffload_help(struct sk_buff *skb, 4358 const netdev_features_t features); 4359 4360 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4361 netdev_features_t features, bool tx_path); 4362 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4363 netdev_features_t features); 4364 4365 struct netdev_bonding_info { 4366 ifslave slave; 4367 ifbond master; 4368 }; 4369 4370 struct netdev_notifier_bonding_info { 4371 struct netdev_notifier_info info; /* must be first */ 4372 struct netdev_bonding_info bonding_info; 4373 }; 4374 4375 void netdev_bonding_info_change(struct net_device *dev, 4376 struct netdev_bonding_info *bonding_info); 4377 4378 static inline 4379 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4380 { 4381 return __skb_gso_segment(skb, features, true); 4382 } 4383 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4384 4385 static inline bool can_checksum_protocol(netdev_features_t features, 4386 __be16 protocol) 4387 { 4388 if (protocol == htons(ETH_P_FCOE)) 4389 return !!(features & NETIF_F_FCOE_CRC); 4390 4391 /* Assume this is an IP checksum (not SCTP CRC) */ 4392 4393 if (features & NETIF_F_HW_CSUM) { 4394 /* Can checksum everything */ 4395 return true; 4396 } 4397 4398 switch (protocol) { 4399 case htons(ETH_P_IP): 4400 return !!(features & NETIF_F_IP_CSUM); 4401 case htons(ETH_P_IPV6): 4402 return !!(features & NETIF_F_IPV6_CSUM); 4403 default: 4404 return false; 4405 } 4406 } 4407 4408 #ifdef CONFIG_BUG 4409 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4410 #else 4411 static inline void netdev_rx_csum_fault(struct net_device *dev, 4412 struct sk_buff *skb) 4413 { 4414 } 4415 #endif 4416 /* rx skb timestamps */ 4417 void net_enable_timestamp(void); 4418 void net_disable_timestamp(void); 4419 4420 #ifdef CONFIG_PROC_FS 4421 int __init dev_proc_init(void); 4422 #else 4423 #define dev_proc_init() 0 4424 #endif 4425 4426 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4427 struct sk_buff *skb, struct net_device *dev, 4428 bool more) 4429 { 4430 __this_cpu_write(softnet_data.xmit.more, more); 4431 return ops->ndo_start_xmit(skb, dev); 4432 } 4433 4434 static inline bool netdev_xmit_more(void) 4435 { 4436 return __this_cpu_read(softnet_data.xmit.more); 4437 } 4438 4439 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4440 struct netdev_queue *txq, bool more) 4441 { 4442 const struct net_device_ops *ops = dev->netdev_ops; 4443 netdev_tx_t rc; 4444 4445 rc = __netdev_start_xmit(ops, skb, dev, more); 4446 if (rc == NETDEV_TX_OK) 4447 txq_trans_update(txq); 4448 4449 return rc; 4450 } 4451 4452 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4453 const void *ns); 4454 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4455 const void *ns); 4456 4457 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4458 { 4459 return netdev_class_create_file_ns(class_attr, NULL); 4460 } 4461 4462 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4463 { 4464 netdev_class_remove_file_ns(class_attr, NULL); 4465 } 4466 4467 extern const struct kobj_ns_type_operations net_ns_type_operations; 4468 4469 const char *netdev_drivername(const struct net_device *dev); 4470 4471 void linkwatch_run_queue(void); 4472 4473 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4474 netdev_features_t f2) 4475 { 4476 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4477 if (f1 & NETIF_F_HW_CSUM) 4478 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4479 else 4480 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4481 } 4482 4483 return f1 & f2; 4484 } 4485 4486 static inline netdev_features_t netdev_get_wanted_features( 4487 struct net_device *dev) 4488 { 4489 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4490 } 4491 netdev_features_t netdev_increment_features(netdev_features_t all, 4492 netdev_features_t one, netdev_features_t mask); 4493 4494 /* Allow TSO being used on stacked device : 4495 * Performing the GSO segmentation before last device 4496 * is a performance improvement. 4497 */ 4498 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4499 netdev_features_t mask) 4500 { 4501 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4502 } 4503 4504 int __netdev_update_features(struct net_device *dev); 4505 void netdev_update_features(struct net_device *dev); 4506 void netdev_change_features(struct net_device *dev); 4507 4508 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4509 struct net_device *dev); 4510 4511 netdev_features_t passthru_features_check(struct sk_buff *skb, 4512 struct net_device *dev, 4513 netdev_features_t features); 4514 netdev_features_t netif_skb_features(struct sk_buff *skb); 4515 4516 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4517 { 4518 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4519 4520 /* check flags correspondence */ 4521 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4522 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4523 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4524 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4525 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4526 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4527 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4528 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4529 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4530 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4531 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4532 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4533 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4534 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4535 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4536 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4537 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4538 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4539 4540 return (features & feature) == feature; 4541 } 4542 4543 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4544 { 4545 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4546 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4547 } 4548 4549 static inline bool netif_needs_gso(struct sk_buff *skb, 4550 netdev_features_t features) 4551 { 4552 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4553 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4554 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4555 } 4556 4557 static inline void netif_set_gso_max_size(struct net_device *dev, 4558 unsigned int size) 4559 { 4560 dev->gso_max_size = size; 4561 } 4562 4563 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4564 int pulled_hlen, u16 mac_offset, 4565 int mac_len) 4566 { 4567 skb->protocol = protocol; 4568 skb->encapsulation = 1; 4569 skb_push(skb, pulled_hlen); 4570 skb_reset_transport_header(skb); 4571 skb->mac_header = mac_offset; 4572 skb->network_header = skb->mac_header + mac_len; 4573 skb->mac_len = mac_len; 4574 } 4575 4576 static inline bool netif_is_macsec(const struct net_device *dev) 4577 { 4578 return dev->priv_flags & IFF_MACSEC; 4579 } 4580 4581 static inline bool netif_is_macvlan(const struct net_device *dev) 4582 { 4583 return dev->priv_flags & IFF_MACVLAN; 4584 } 4585 4586 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4587 { 4588 return dev->priv_flags & IFF_MACVLAN_PORT; 4589 } 4590 4591 static inline bool netif_is_bond_master(const struct net_device *dev) 4592 { 4593 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4594 } 4595 4596 static inline bool netif_is_bond_slave(const struct net_device *dev) 4597 { 4598 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4599 } 4600 4601 static inline bool netif_supports_nofcs(struct net_device *dev) 4602 { 4603 return dev->priv_flags & IFF_SUPP_NOFCS; 4604 } 4605 4606 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4607 { 4608 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4609 } 4610 4611 static inline bool netif_is_l3_master(const struct net_device *dev) 4612 { 4613 return dev->priv_flags & IFF_L3MDEV_MASTER; 4614 } 4615 4616 static inline bool netif_is_l3_slave(const struct net_device *dev) 4617 { 4618 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4619 } 4620 4621 static inline bool netif_is_bridge_master(const struct net_device *dev) 4622 { 4623 return dev->priv_flags & IFF_EBRIDGE; 4624 } 4625 4626 static inline bool netif_is_bridge_port(const struct net_device *dev) 4627 { 4628 return dev->priv_flags & IFF_BRIDGE_PORT; 4629 } 4630 4631 static inline bool netif_is_ovs_master(const struct net_device *dev) 4632 { 4633 return dev->priv_flags & IFF_OPENVSWITCH; 4634 } 4635 4636 static inline bool netif_is_ovs_port(const struct net_device *dev) 4637 { 4638 return dev->priv_flags & IFF_OVS_DATAPATH; 4639 } 4640 4641 static inline bool netif_is_team_master(const struct net_device *dev) 4642 { 4643 return dev->priv_flags & IFF_TEAM; 4644 } 4645 4646 static inline bool netif_is_team_port(const struct net_device *dev) 4647 { 4648 return dev->priv_flags & IFF_TEAM_PORT; 4649 } 4650 4651 static inline bool netif_is_lag_master(const struct net_device *dev) 4652 { 4653 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4654 } 4655 4656 static inline bool netif_is_lag_port(const struct net_device *dev) 4657 { 4658 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4659 } 4660 4661 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4662 { 4663 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4664 } 4665 4666 static inline bool netif_is_failover(const struct net_device *dev) 4667 { 4668 return dev->priv_flags & IFF_FAILOVER; 4669 } 4670 4671 static inline bool netif_is_failover_slave(const struct net_device *dev) 4672 { 4673 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4674 } 4675 4676 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4677 static inline void netif_keep_dst(struct net_device *dev) 4678 { 4679 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4680 } 4681 4682 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4683 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4684 { 4685 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4686 return dev->priv_flags & IFF_MACSEC; 4687 } 4688 4689 extern struct pernet_operations __net_initdata loopback_net_ops; 4690 4691 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4692 4693 /* netdev_printk helpers, similar to dev_printk */ 4694 4695 static inline const char *netdev_name(const struct net_device *dev) 4696 { 4697 if (!dev->name[0] || strchr(dev->name, '%')) 4698 return "(unnamed net_device)"; 4699 return dev->name; 4700 } 4701 4702 static inline bool netdev_unregistering(const struct net_device *dev) 4703 { 4704 return dev->reg_state == NETREG_UNREGISTERING; 4705 } 4706 4707 static inline const char *netdev_reg_state(const struct net_device *dev) 4708 { 4709 switch (dev->reg_state) { 4710 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4711 case NETREG_REGISTERED: return ""; 4712 case NETREG_UNREGISTERING: return " (unregistering)"; 4713 case NETREG_UNREGISTERED: return " (unregistered)"; 4714 case NETREG_RELEASED: return " (released)"; 4715 case NETREG_DUMMY: return " (dummy)"; 4716 } 4717 4718 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4719 return " (unknown)"; 4720 } 4721 4722 __printf(3, 4) __cold 4723 void netdev_printk(const char *level, const struct net_device *dev, 4724 const char *format, ...); 4725 __printf(2, 3) __cold 4726 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4727 __printf(2, 3) __cold 4728 void netdev_alert(const struct net_device *dev, const char *format, ...); 4729 __printf(2, 3) __cold 4730 void netdev_crit(const struct net_device *dev, const char *format, ...); 4731 __printf(2, 3) __cold 4732 void netdev_err(const struct net_device *dev, const char *format, ...); 4733 __printf(2, 3) __cold 4734 void netdev_warn(const struct net_device *dev, const char *format, ...); 4735 __printf(2, 3) __cold 4736 void netdev_notice(const struct net_device *dev, const char *format, ...); 4737 __printf(2, 3) __cold 4738 void netdev_info(const struct net_device *dev, const char *format, ...); 4739 4740 #define netdev_level_once(level, dev, fmt, ...) \ 4741 do { \ 4742 static bool __print_once __read_mostly; \ 4743 \ 4744 if (!__print_once) { \ 4745 __print_once = true; \ 4746 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4747 } \ 4748 } while (0) 4749 4750 #define netdev_emerg_once(dev, fmt, ...) \ 4751 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4752 #define netdev_alert_once(dev, fmt, ...) \ 4753 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4754 #define netdev_crit_once(dev, fmt, ...) \ 4755 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4756 #define netdev_err_once(dev, fmt, ...) \ 4757 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4758 #define netdev_warn_once(dev, fmt, ...) \ 4759 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4760 #define netdev_notice_once(dev, fmt, ...) \ 4761 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4762 #define netdev_info_once(dev, fmt, ...) \ 4763 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4764 4765 #define MODULE_ALIAS_NETDEV(device) \ 4766 MODULE_ALIAS("netdev-" device) 4767 4768 #if defined(CONFIG_DYNAMIC_DEBUG) 4769 #define netdev_dbg(__dev, format, args...) \ 4770 do { \ 4771 dynamic_netdev_dbg(__dev, format, ##args); \ 4772 } while (0) 4773 #elif defined(DEBUG) 4774 #define netdev_dbg(__dev, format, args...) \ 4775 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4776 #else 4777 #define netdev_dbg(__dev, format, args...) \ 4778 ({ \ 4779 if (0) \ 4780 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4781 }) 4782 #endif 4783 4784 #if defined(VERBOSE_DEBUG) 4785 #define netdev_vdbg netdev_dbg 4786 #else 4787 4788 #define netdev_vdbg(dev, format, args...) \ 4789 ({ \ 4790 if (0) \ 4791 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4792 0; \ 4793 }) 4794 #endif 4795 4796 /* 4797 * netdev_WARN() acts like dev_printk(), but with the key difference 4798 * of using a WARN/WARN_ON to get the message out, including the 4799 * file/line information and a backtrace. 4800 */ 4801 #define netdev_WARN(dev, format, args...) \ 4802 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4803 netdev_reg_state(dev), ##args) 4804 4805 #define netdev_WARN_ONCE(dev, format, args...) \ 4806 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4807 netdev_reg_state(dev), ##args) 4808 4809 /* netif printk helpers, similar to netdev_printk */ 4810 4811 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4812 do { \ 4813 if (netif_msg_##type(priv)) \ 4814 netdev_printk(level, (dev), fmt, ##args); \ 4815 } while (0) 4816 4817 #define netif_level(level, priv, type, dev, fmt, args...) \ 4818 do { \ 4819 if (netif_msg_##type(priv)) \ 4820 netdev_##level(dev, fmt, ##args); \ 4821 } while (0) 4822 4823 #define netif_emerg(priv, type, dev, fmt, args...) \ 4824 netif_level(emerg, priv, type, dev, fmt, ##args) 4825 #define netif_alert(priv, type, dev, fmt, args...) \ 4826 netif_level(alert, priv, type, dev, fmt, ##args) 4827 #define netif_crit(priv, type, dev, fmt, args...) \ 4828 netif_level(crit, priv, type, dev, fmt, ##args) 4829 #define netif_err(priv, type, dev, fmt, args...) \ 4830 netif_level(err, priv, type, dev, fmt, ##args) 4831 #define netif_warn(priv, type, dev, fmt, args...) \ 4832 netif_level(warn, priv, type, dev, fmt, ##args) 4833 #define netif_notice(priv, type, dev, fmt, args...) \ 4834 netif_level(notice, priv, type, dev, fmt, ##args) 4835 #define netif_info(priv, type, dev, fmt, args...) \ 4836 netif_level(info, priv, type, dev, fmt, ##args) 4837 4838 #if defined(CONFIG_DYNAMIC_DEBUG) 4839 #define netif_dbg(priv, type, netdev, format, args...) \ 4840 do { \ 4841 if (netif_msg_##type(priv)) \ 4842 dynamic_netdev_dbg(netdev, format, ##args); \ 4843 } while (0) 4844 #elif defined(DEBUG) 4845 #define netif_dbg(priv, type, dev, format, args...) \ 4846 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4847 #else 4848 #define netif_dbg(priv, type, dev, format, args...) \ 4849 ({ \ 4850 if (0) \ 4851 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4852 0; \ 4853 }) 4854 #endif 4855 4856 /* if @cond then downgrade to debug, else print at @level */ 4857 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4858 do { \ 4859 if (cond) \ 4860 netif_dbg(priv, type, netdev, fmt, ##args); \ 4861 else \ 4862 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4863 } while (0) 4864 4865 #if defined(VERBOSE_DEBUG) 4866 #define netif_vdbg netif_dbg 4867 #else 4868 #define netif_vdbg(priv, type, dev, format, args...) \ 4869 ({ \ 4870 if (0) \ 4871 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4872 0; \ 4873 }) 4874 #endif 4875 4876 /* 4877 * The list of packet types we will receive (as opposed to discard) 4878 * and the routines to invoke. 4879 * 4880 * Why 16. Because with 16 the only overlap we get on a hash of the 4881 * low nibble of the protocol value is RARP/SNAP/X.25. 4882 * 4883 * 0800 IP 4884 * 0001 802.3 4885 * 0002 AX.25 4886 * 0004 802.2 4887 * 8035 RARP 4888 * 0005 SNAP 4889 * 0805 X.25 4890 * 0806 ARP 4891 * 8137 IPX 4892 * 0009 Localtalk 4893 * 86DD IPv6 4894 */ 4895 #define PTYPE_HASH_SIZE (16) 4896 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4897 4898 #endif /* _LINUX_NETDEVICE_H */ 4899