1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <net/net_namespace.h> 38 #ifdef CONFIG_DCB 39 #include <net/dcbnl.h> 40 #endif 41 #include <net/netprio_cgroup.h> 42 #include <net/xdp.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <linux/hashtable.h> 50 51 struct netpoll_info; 52 struct device; 53 struct ethtool_ops; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void synchronize_net(void); 74 void netdev_set_default_ethtool_ops(struct net_device *dev, 75 const struct ethtool_ops *ops); 76 77 /* Backlog congestion levels */ 78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 79 #define NET_RX_DROP 1 /* packet dropped */ 80 81 #define MAX_NEST_DEV 8 82 83 /* 84 * Transmit return codes: transmit return codes originate from three different 85 * namespaces: 86 * 87 * - qdisc return codes 88 * - driver transmit return codes 89 * - errno values 90 * 91 * Drivers are allowed to return any one of those in their hard_start_xmit() 92 * function. Real network devices commonly used with qdiscs should only return 93 * the driver transmit return codes though - when qdiscs are used, the actual 94 * transmission happens asynchronously, so the value is not propagated to 95 * higher layers. Virtual network devices transmit synchronously; in this case 96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 97 * others are propagated to higher layers. 98 */ 99 100 /* qdisc ->enqueue() return codes. */ 101 #define NET_XMIT_SUCCESS 0x00 102 #define NET_XMIT_DROP 0x01 /* skb dropped */ 103 #define NET_XMIT_CN 0x02 /* congestion notification */ 104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 105 106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 107 * indicates that the device will soon be dropping packets, or already drops 108 * some packets of the same priority; prompting us to send less aggressively. */ 109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 111 112 /* Driver transmit return codes */ 113 #define NETDEV_TX_MASK 0xf0 114 115 enum netdev_tx { 116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 117 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 119 }; 120 typedef enum netdev_tx netdev_tx_t; 121 122 /* 123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 125 */ 126 static inline bool dev_xmit_complete(int rc) 127 { 128 /* 129 * Positive cases with an skb consumed by a driver: 130 * - successful transmission (rc == NETDEV_TX_OK) 131 * - error while transmitting (rc < 0) 132 * - error while queueing to a different device (rc & NET_XMIT_MASK) 133 */ 134 if (likely(rc < NET_XMIT_MASK)) 135 return true; 136 137 return false; 138 } 139 140 /* 141 * Compute the worst-case header length according to the protocols 142 * used. 143 */ 144 145 #if defined(CONFIG_HYPERV_NET) 146 # define LL_MAX_HEADER 128 147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 148 # if defined(CONFIG_MAC80211_MESH) 149 # define LL_MAX_HEADER 128 150 # else 151 # define LL_MAX_HEADER 96 152 # endif 153 #else 154 # define LL_MAX_HEADER 32 155 #endif 156 157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 159 #define MAX_HEADER LL_MAX_HEADER 160 #else 161 #define MAX_HEADER (LL_MAX_HEADER + 48) 162 #endif 163 164 /* 165 * Old network device statistics. Fields are native words 166 * (unsigned long) so they can be read and written atomically. 167 */ 168 169 struct net_device_stats { 170 unsigned long rx_packets; 171 unsigned long tx_packets; 172 unsigned long rx_bytes; 173 unsigned long tx_bytes; 174 unsigned long rx_errors; 175 unsigned long tx_errors; 176 unsigned long rx_dropped; 177 unsigned long tx_dropped; 178 unsigned long multicast; 179 unsigned long collisions; 180 unsigned long rx_length_errors; 181 unsigned long rx_over_errors; 182 unsigned long rx_crc_errors; 183 unsigned long rx_frame_errors; 184 unsigned long rx_fifo_errors; 185 unsigned long rx_missed_errors; 186 unsigned long tx_aborted_errors; 187 unsigned long tx_carrier_errors; 188 unsigned long tx_fifo_errors; 189 unsigned long tx_heartbeat_errors; 190 unsigned long tx_window_errors; 191 unsigned long rx_compressed; 192 unsigned long tx_compressed; 193 }; 194 195 196 #include <linux/cache.h> 197 #include <linux/skbuff.h> 198 199 #ifdef CONFIG_RPS 200 #include <linux/static_key.h> 201 extern struct static_key_false rps_needed; 202 extern struct static_key_false rfs_needed; 203 #endif 204 205 struct neighbour; 206 struct neigh_parms; 207 struct sk_buff; 208 209 struct netdev_hw_addr { 210 struct list_head list; 211 unsigned char addr[MAX_ADDR_LEN]; 212 unsigned char type; 213 #define NETDEV_HW_ADDR_T_LAN 1 214 #define NETDEV_HW_ADDR_T_SAN 2 215 #define NETDEV_HW_ADDR_T_UNICAST 3 216 #define NETDEV_HW_ADDR_T_MULTICAST 4 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 }; 351 352 enum { 353 NAPI_STATE_SCHED, /* Poll is scheduled */ 354 NAPI_STATE_MISSED, /* reschedule a napi */ 355 NAPI_STATE_DISABLE, /* Disable pending */ 356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 357 NAPI_STATE_LISTED, /* NAPI added to system lists */ 358 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 359 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 360 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 361 }; 362 363 enum { 364 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 365 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 366 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 367 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 368 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 369 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 370 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 371 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 372 }; 373 374 enum gro_result { 375 GRO_MERGED, 376 GRO_MERGED_FREE, 377 GRO_HELD, 378 GRO_NORMAL, 379 GRO_CONSUMED, 380 }; 381 typedef enum gro_result gro_result_t; 382 383 /* 384 * enum rx_handler_result - Possible return values for rx_handlers. 385 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 386 * further. 387 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 388 * case skb->dev was changed by rx_handler. 389 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 390 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 391 * 392 * rx_handlers are functions called from inside __netif_receive_skb(), to do 393 * special processing of the skb, prior to delivery to protocol handlers. 394 * 395 * Currently, a net_device can only have a single rx_handler registered. Trying 396 * to register a second rx_handler will return -EBUSY. 397 * 398 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 399 * To unregister a rx_handler on a net_device, use 400 * netdev_rx_handler_unregister(). 401 * 402 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 403 * do with the skb. 404 * 405 * If the rx_handler consumed the skb in some way, it should return 406 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 407 * the skb to be delivered in some other way. 408 * 409 * If the rx_handler changed skb->dev, to divert the skb to another 410 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 411 * new device will be called if it exists. 412 * 413 * If the rx_handler decides the skb should be ignored, it should return 414 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 415 * are registered on exact device (ptype->dev == skb->dev). 416 * 417 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 418 * delivered, it should return RX_HANDLER_PASS. 419 * 420 * A device without a registered rx_handler will behave as if rx_handler 421 * returned RX_HANDLER_PASS. 422 */ 423 424 enum rx_handler_result { 425 RX_HANDLER_CONSUMED, 426 RX_HANDLER_ANOTHER, 427 RX_HANDLER_EXACT, 428 RX_HANDLER_PASS, 429 }; 430 typedef enum rx_handler_result rx_handler_result_t; 431 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 432 433 void __napi_schedule(struct napi_struct *n); 434 void __napi_schedule_irqoff(struct napi_struct *n); 435 436 static inline bool napi_disable_pending(struct napi_struct *n) 437 { 438 return test_bit(NAPI_STATE_DISABLE, &n->state); 439 } 440 441 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 442 { 443 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 444 } 445 446 bool napi_schedule_prep(struct napi_struct *n); 447 448 /** 449 * napi_schedule - schedule NAPI poll 450 * @n: NAPI context 451 * 452 * Schedule NAPI poll routine to be called if it is not already 453 * running. 454 */ 455 static inline void napi_schedule(struct napi_struct *n) 456 { 457 if (napi_schedule_prep(n)) 458 __napi_schedule(n); 459 } 460 461 /** 462 * napi_schedule_irqoff - schedule NAPI poll 463 * @n: NAPI context 464 * 465 * Variant of napi_schedule(), assuming hard irqs are masked. 466 */ 467 static inline void napi_schedule_irqoff(struct napi_struct *n) 468 { 469 if (napi_schedule_prep(n)) 470 __napi_schedule_irqoff(n); 471 } 472 473 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 474 static inline bool napi_reschedule(struct napi_struct *napi) 475 { 476 if (napi_schedule_prep(napi)) { 477 __napi_schedule(napi); 478 return true; 479 } 480 return false; 481 } 482 483 bool napi_complete_done(struct napi_struct *n, int work_done); 484 /** 485 * napi_complete - NAPI processing complete 486 * @n: NAPI context 487 * 488 * Mark NAPI processing as complete. 489 * Consider using napi_complete_done() instead. 490 * Return false if device should avoid rearming interrupts. 491 */ 492 static inline bool napi_complete(struct napi_struct *n) 493 { 494 return napi_complete_done(n, 0); 495 } 496 497 /** 498 * napi_disable - prevent NAPI from scheduling 499 * @n: NAPI context 500 * 501 * Stop NAPI from being scheduled on this context. 502 * Waits till any outstanding processing completes. 503 */ 504 void napi_disable(struct napi_struct *n); 505 506 /** 507 * napi_enable - enable NAPI scheduling 508 * @n: NAPI context 509 * 510 * Resume NAPI from being scheduled on this context. 511 * Must be paired with napi_disable. 512 */ 513 static inline void napi_enable(struct napi_struct *n) 514 { 515 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 516 smp_mb__before_atomic(); 517 clear_bit(NAPI_STATE_SCHED, &n->state); 518 clear_bit(NAPI_STATE_NPSVC, &n->state); 519 } 520 521 /** 522 * napi_synchronize - wait until NAPI is not running 523 * @n: NAPI context 524 * 525 * Wait until NAPI is done being scheduled on this context. 526 * Waits till any outstanding processing completes but 527 * does not disable future activations. 528 */ 529 static inline void napi_synchronize(const struct napi_struct *n) 530 { 531 if (IS_ENABLED(CONFIG_SMP)) 532 while (test_bit(NAPI_STATE_SCHED, &n->state)) 533 msleep(1); 534 else 535 barrier(); 536 } 537 538 /** 539 * napi_if_scheduled_mark_missed - if napi is running, set the 540 * NAPIF_STATE_MISSED 541 * @n: NAPI context 542 * 543 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 544 * NAPI is scheduled. 545 **/ 546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 547 { 548 unsigned long val, new; 549 550 do { 551 val = READ_ONCE(n->state); 552 if (val & NAPIF_STATE_DISABLE) 553 return true; 554 555 if (!(val & NAPIF_STATE_SCHED)) 556 return false; 557 558 new = val | NAPIF_STATE_MISSED; 559 } while (cmpxchg(&n->state, val, new) != val); 560 561 return true; 562 } 563 564 enum netdev_queue_state_t { 565 __QUEUE_STATE_DRV_XOFF, 566 __QUEUE_STATE_STACK_XOFF, 567 __QUEUE_STATE_FROZEN, 568 }; 569 570 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 571 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 572 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 573 574 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 576 QUEUE_STATE_FROZEN) 577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 578 QUEUE_STATE_FROZEN) 579 580 /* 581 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 582 * netif_tx_* functions below are used to manipulate this flag. The 583 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 584 * queue independently. The netif_xmit_*stopped functions below are called 585 * to check if the queue has been stopped by the driver or stack (either 586 * of the XOFF bits are set in the state). Drivers should not need to call 587 * netif_xmit*stopped functions, they should only be using netif_tx_*. 588 */ 589 590 struct netdev_queue { 591 /* 592 * read-mostly part 593 */ 594 struct net_device *dev; 595 struct Qdisc __rcu *qdisc; 596 struct Qdisc *qdisc_sleeping; 597 #ifdef CONFIG_SYSFS 598 struct kobject kobj; 599 #endif 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 int numa_node; 602 #endif 603 unsigned long tx_maxrate; 604 /* 605 * Number of TX timeouts for this queue 606 * (/sys/class/net/DEV/Q/trans_timeout) 607 */ 608 unsigned long trans_timeout; 609 610 /* Subordinate device that the queue has been assigned to */ 611 struct net_device *sb_dev; 612 #ifdef CONFIG_XDP_SOCKETS 613 struct xsk_buff_pool *pool; 614 #endif 615 /* 616 * write-mostly part 617 */ 618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 619 int xmit_lock_owner; 620 /* 621 * Time (in jiffies) of last Tx 622 */ 623 unsigned long trans_start; 624 625 unsigned long state; 626 627 #ifdef CONFIG_BQL 628 struct dql dql; 629 #endif 630 } ____cacheline_aligned_in_smp; 631 632 extern int sysctl_fb_tunnels_only_for_init_net; 633 extern int sysctl_devconf_inherit_init_net; 634 635 /* 636 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 637 * == 1 : For initns only 638 * == 2 : For none. 639 */ 640 static inline bool net_has_fallback_tunnels(const struct net *net) 641 { 642 return !IS_ENABLED(CONFIG_SYSCTL) || 643 !sysctl_fb_tunnels_only_for_init_net || 644 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 645 } 646 647 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 648 { 649 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 650 return q->numa_node; 651 #else 652 return NUMA_NO_NODE; 653 #endif 654 } 655 656 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 657 { 658 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 659 q->numa_node = node; 660 #endif 661 } 662 663 #ifdef CONFIG_RPS 664 /* 665 * This structure holds an RPS map which can be of variable length. The 666 * map is an array of CPUs. 667 */ 668 struct rps_map { 669 unsigned int len; 670 struct rcu_head rcu; 671 u16 cpus[]; 672 }; 673 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 674 675 /* 676 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 677 * tail pointer for that CPU's input queue at the time of last enqueue, and 678 * a hardware filter index. 679 */ 680 struct rps_dev_flow { 681 u16 cpu; 682 u16 filter; 683 unsigned int last_qtail; 684 }; 685 #define RPS_NO_FILTER 0xffff 686 687 /* 688 * The rps_dev_flow_table structure contains a table of flow mappings. 689 */ 690 struct rps_dev_flow_table { 691 unsigned int mask; 692 struct rcu_head rcu; 693 struct rps_dev_flow flows[]; 694 }; 695 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 696 ((_num) * sizeof(struct rps_dev_flow))) 697 698 /* 699 * The rps_sock_flow_table contains mappings of flows to the last CPU 700 * on which they were processed by the application (set in recvmsg). 701 * Each entry is a 32bit value. Upper part is the high-order bits 702 * of flow hash, lower part is CPU number. 703 * rps_cpu_mask is used to partition the space, depending on number of 704 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 705 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 706 * meaning we use 32-6=26 bits for the hash. 707 */ 708 struct rps_sock_flow_table { 709 u32 mask; 710 711 u32 ents[] ____cacheline_aligned_in_smp; 712 }; 713 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 714 715 #define RPS_NO_CPU 0xffff 716 717 extern u32 rps_cpu_mask; 718 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 719 720 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 721 u32 hash) 722 { 723 if (table && hash) { 724 unsigned int index = hash & table->mask; 725 u32 val = hash & ~rps_cpu_mask; 726 727 /* We only give a hint, preemption can change CPU under us */ 728 val |= raw_smp_processor_id(); 729 730 if (table->ents[index] != val) 731 table->ents[index] = val; 732 } 733 } 734 735 #ifdef CONFIG_RFS_ACCEL 736 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 737 u16 filter_id); 738 #endif 739 #endif /* CONFIG_RPS */ 740 741 /* This structure contains an instance of an RX queue. */ 742 struct netdev_rx_queue { 743 #ifdef CONFIG_RPS 744 struct rps_map __rcu *rps_map; 745 struct rps_dev_flow_table __rcu *rps_flow_table; 746 #endif 747 struct kobject kobj; 748 struct net_device *dev; 749 struct xdp_rxq_info xdp_rxq; 750 #ifdef CONFIG_XDP_SOCKETS 751 struct xsk_buff_pool *pool; 752 #endif 753 } ____cacheline_aligned_in_smp; 754 755 /* 756 * RX queue sysfs structures and functions. 757 */ 758 struct rx_queue_attribute { 759 struct attribute attr; 760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 761 ssize_t (*store)(struct netdev_rx_queue *queue, 762 const char *buf, size_t len); 763 }; 764 765 #ifdef CONFIG_XPS 766 /* 767 * This structure holds an XPS map which can be of variable length. The 768 * map is an array of queues. 769 */ 770 struct xps_map { 771 unsigned int len; 772 unsigned int alloc_len; 773 struct rcu_head rcu; 774 u16 queues[]; 775 }; 776 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 777 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 778 - sizeof(struct xps_map)) / sizeof(u16)) 779 780 /* 781 * This structure holds all XPS maps for device. Maps are indexed by CPU. 782 */ 783 struct xps_dev_maps { 784 struct rcu_head rcu; 785 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 786 }; 787 788 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 789 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 790 791 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 792 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 793 794 #endif /* CONFIG_XPS */ 795 796 #define TC_MAX_QUEUE 16 797 #define TC_BITMASK 15 798 /* HW offloaded queuing disciplines txq count and offset maps */ 799 struct netdev_tc_txq { 800 u16 count; 801 u16 offset; 802 }; 803 804 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 805 /* 806 * This structure is to hold information about the device 807 * configured to run FCoE protocol stack. 808 */ 809 struct netdev_fcoe_hbainfo { 810 char manufacturer[64]; 811 char serial_number[64]; 812 char hardware_version[64]; 813 char driver_version[64]; 814 char optionrom_version[64]; 815 char firmware_version[64]; 816 char model[256]; 817 char model_description[256]; 818 }; 819 #endif 820 821 #define MAX_PHYS_ITEM_ID_LEN 32 822 823 /* This structure holds a unique identifier to identify some 824 * physical item (port for example) used by a netdevice. 825 */ 826 struct netdev_phys_item_id { 827 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 828 unsigned char id_len; 829 }; 830 831 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 832 struct netdev_phys_item_id *b) 833 { 834 return a->id_len == b->id_len && 835 memcmp(a->id, b->id, a->id_len) == 0; 836 } 837 838 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 839 struct sk_buff *skb, 840 struct net_device *sb_dev); 841 842 enum tc_setup_type { 843 TC_SETUP_QDISC_MQPRIO, 844 TC_SETUP_CLSU32, 845 TC_SETUP_CLSFLOWER, 846 TC_SETUP_CLSMATCHALL, 847 TC_SETUP_CLSBPF, 848 TC_SETUP_BLOCK, 849 TC_SETUP_QDISC_CBS, 850 TC_SETUP_QDISC_RED, 851 TC_SETUP_QDISC_PRIO, 852 TC_SETUP_QDISC_MQ, 853 TC_SETUP_QDISC_ETF, 854 TC_SETUP_ROOT_QDISC, 855 TC_SETUP_QDISC_GRED, 856 TC_SETUP_QDISC_TAPRIO, 857 TC_SETUP_FT, 858 TC_SETUP_QDISC_ETS, 859 TC_SETUP_QDISC_TBF, 860 TC_SETUP_QDISC_FIFO, 861 TC_SETUP_QDISC_HTB, 862 }; 863 864 /* These structures hold the attributes of bpf state that are being passed 865 * to the netdevice through the bpf op. 866 */ 867 enum bpf_netdev_command { 868 /* Set or clear a bpf program used in the earliest stages of packet 869 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 870 * is responsible for calling bpf_prog_put on any old progs that are 871 * stored. In case of error, the callee need not release the new prog 872 * reference, but on success it takes ownership and must bpf_prog_put 873 * when it is no longer used. 874 */ 875 XDP_SETUP_PROG, 876 XDP_SETUP_PROG_HW, 877 /* BPF program for offload callbacks, invoked at program load time. */ 878 BPF_OFFLOAD_MAP_ALLOC, 879 BPF_OFFLOAD_MAP_FREE, 880 XDP_SETUP_XSK_POOL, 881 }; 882 883 struct bpf_prog_offload_ops; 884 struct netlink_ext_ack; 885 struct xdp_umem; 886 struct xdp_dev_bulk_queue; 887 struct bpf_xdp_link; 888 889 enum bpf_xdp_mode { 890 XDP_MODE_SKB = 0, 891 XDP_MODE_DRV = 1, 892 XDP_MODE_HW = 2, 893 __MAX_XDP_MODE 894 }; 895 896 struct bpf_xdp_entity { 897 struct bpf_prog *prog; 898 struct bpf_xdp_link *link; 899 }; 900 901 struct netdev_bpf { 902 enum bpf_netdev_command command; 903 union { 904 /* XDP_SETUP_PROG */ 905 struct { 906 u32 flags; 907 struct bpf_prog *prog; 908 struct netlink_ext_ack *extack; 909 }; 910 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 911 struct { 912 struct bpf_offloaded_map *offmap; 913 }; 914 /* XDP_SETUP_XSK_POOL */ 915 struct { 916 struct xsk_buff_pool *pool; 917 u16 queue_id; 918 } xsk; 919 }; 920 }; 921 922 /* Flags for ndo_xsk_wakeup. */ 923 #define XDP_WAKEUP_RX (1 << 0) 924 #define XDP_WAKEUP_TX (1 << 1) 925 926 #ifdef CONFIG_XFRM_OFFLOAD 927 struct xfrmdev_ops { 928 int (*xdo_dev_state_add) (struct xfrm_state *x); 929 void (*xdo_dev_state_delete) (struct xfrm_state *x); 930 void (*xdo_dev_state_free) (struct xfrm_state *x); 931 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 932 struct xfrm_state *x); 933 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 934 }; 935 #endif 936 937 struct dev_ifalias { 938 struct rcu_head rcuhead; 939 char ifalias[]; 940 }; 941 942 struct devlink; 943 struct tlsdev_ops; 944 945 struct netdev_name_node { 946 struct hlist_node hlist; 947 struct list_head list; 948 struct net_device *dev; 949 const char *name; 950 }; 951 952 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 953 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 954 955 struct netdev_net_notifier { 956 struct list_head list; 957 struct notifier_block *nb; 958 }; 959 960 /* 961 * This structure defines the management hooks for network devices. 962 * The following hooks can be defined; unless noted otherwise, they are 963 * optional and can be filled with a null pointer. 964 * 965 * int (*ndo_init)(struct net_device *dev); 966 * This function is called once when a network device is registered. 967 * The network device can use this for any late stage initialization 968 * or semantic validation. It can fail with an error code which will 969 * be propagated back to register_netdev. 970 * 971 * void (*ndo_uninit)(struct net_device *dev); 972 * This function is called when device is unregistered or when registration 973 * fails. It is not called if init fails. 974 * 975 * int (*ndo_open)(struct net_device *dev); 976 * This function is called when a network device transitions to the up 977 * state. 978 * 979 * int (*ndo_stop)(struct net_device *dev); 980 * This function is called when a network device transitions to the down 981 * state. 982 * 983 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 984 * struct net_device *dev); 985 * Called when a packet needs to be transmitted. 986 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 987 * the queue before that can happen; it's for obsolete devices and weird 988 * corner cases, but the stack really does a non-trivial amount 989 * of useless work if you return NETDEV_TX_BUSY. 990 * Required; cannot be NULL. 991 * 992 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 993 * struct net_device *dev 994 * netdev_features_t features); 995 * Called by core transmit path to determine if device is capable of 996 * performing offload operations on a given packet. This is to give 997 * the device an opportunity to implement any restrictions that cannot 998 * be otherwise expressed by feature flags. The check is called with 999 * the set of features that the stack has calculated and it returns 1000 * those the driver believes to be appropriate. 1001 * 1002 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1003 * struct net_device *sb_dev); 1004 * Called to decide which queue to use when device supports multiple 1005 * transmit queues. 1006 * 1007 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1008 * This function is called to allow device receiver to make 1009 * changes to configuration when multicast or promiscuous is enabled. 1010 * 1011 * void (*ndo_set_rx_mode)(struct net_device *dev); 1012 * This function is called device changes address list filtering. 1013 * If driver handles unicast address filtering, it should set 1014 * IFF_UNICAST_FLT in its priv_flags. 1015 * 1016 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1017 * This function is called when the Media Access Control address 1018 * needs to be changed. If this interface is not defined, the 1019 * MAC address can not be changed. 1020 * 1021 * int (*ndo_validate_addr)(struct net_device *dev); 1022 * Test if Media Access Control address is valid for the device. 1023 * 1024 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1025 * Called when a user requests an ioctl which can't be handled by 1026 * the generic interface code. If not defined ioctls return 1027 * not supported error code. 1028 * 1029 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1030 * Used to set network devices bus interface parameters. This interface 1031 * is retained for legacy reasons; new devices should use the bus 1032 * interface (PCI) for low level management. 1033 * 1034 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1035 * Called when a user wants to change the Maximum Transfer Unit 1036 * of a device. 1037 * 1038 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1039 * Callback used when the transmitter has not made any progress 1040 * for dev->watchdog ticks. 1041 * 1042 * void (*ndo_get_stats64)(struct net_device *dev, 1043 * struct rtnl_link_stats64 *storage); 1044 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1045 * Called when a user wants to get the network device usage 1046 * statistics. Drivers must do one of the following: 1047 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1048 * rtnl_link_stats64 structure passed by the caller. 1049 * 2. Define @ndo_get_stats to update a net_device_stats structure 1050 * (which should normally be dev->stats) and return a pointer to 1051 * it. The structure may be changed asynchronously only if each 1052 * field is written atomically. 1053 * 3. Update dev->stats asynchronously and atomically, and define 1054 * neither operation. 1055 * 1056 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1057 * Return true if this device supports offload stats of this attr_id. 1058 * 1059 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1060 * void *attr_data) 1061 * Get statistics for offload operations by attr_id. Write it into the 1062 * attr_data pointer. 1063 * 1064 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1065 * If device supports VLAN filtering this function is called when a 1066 * VLAN id is registered. 1067 * 1068 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1069 * If device supports VLAN filtering this function is called when a 1070 * VLAN id is unregistered. 1071 * 1072 * void (*ndo_poll_controller)(struct net_device *dev); 1073 * 1074 * SR-IOV management functions. 1075 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1076 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1077 * u8 qos, __be16 proto); 1078 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1079 * int max_tx_rate); 1080 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1081 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1082 * int (*ndo_get_vf_config)(struct net_device *dev, 1083 * int vf, struct ifla_vf_info *ivf); 1084 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1085 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1086 * struct nlattr *port[]); 1087 * 1088 * Enable or disable the VF ability to query its RSS Redirection Table and 1089 * Hash Key. This is needed since on some devices VF share this information 1090 * with PF and querying it may introduce a theoretical security risk. 1091 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1092 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1093 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1094 * void *type_data); 1095 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1096 * This is always called from the stack with the rtnl lock held and netif 1097 * tx queues stopped. This allows the netdevice to perform queue 1098 * management safely. 1099 * 1100 * Fiber Channel over Ethernet (FCoE) offload functions. 1101 * int (*ndo_fcoe_enable)(struct net_device *dev); 1102 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1103 * so the underlying device can perform whatever needed configuration or 1104 * initialization to support acceleration of FCoE traffic. 1105 * 1106 * int (*ndo_fcoe_disable)(struct net_device *dev); 1107 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1108 * so the underlying device can perform whatever needed clean-ups to 1109 * stop supporting acceleration of FCoE traffic. 1110 * 1111 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1112 * struct scatterlist *sgl, unsigned int sgc); 1113 * Called when the FCoE Initiator wants to initialize an I/O that 1114 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1115 * perform necessary setup and returns 1 to indicate the device is set up 1116 * successfully to perform DDP on this I/O, otherwise this returns 0. 1117 * 1118 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1119 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1120 * indicated by the FC exchange id 'xid', so the underlying device can 1121 * clean up and reuse resources for later DDP requests. 1122 * 1123 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1124 * struct scatterlist *sgl, unsigned int sgc); 1125 * Called when the FCoE Target wants to initialize an I/O that 1126 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1127 * perform necessary setup and returns 1 to indicate the device is set up 1128 * successfully to perform DDP on this I/O, otherwise this returns 0. 1129 * 1130 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1131 * struct netdev_fcoe_hbainfo *hbainfo); 1132 * Called when the FCoE Protocol stack wants information on the underlying 1133 * device. This information is utilized by the FCoE protocol stack to 1134 * register attributes with Fiber Channel management service as per the 1135 * FC-GS Fabric Device Management Information(FDMI) specification. 1136 * 1137 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1138 * Called when the underlying device wants to override default World Wide 1139 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1140 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1141 * protocol stack to use. 1142 * 1143 * RFS acceleration. 1144 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1145 * u16 rxq_index, u32 flow_id); 1146 * Set hardware filter for RFS. rxq_index is the target queue index; 1147 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1148 * Return the filter ID on success, or a negative error code. 1149 * 1150 * Slave management functions (for bridge, bonding, etc). 1151 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1152 * Called to make another netdev an underling. 1153 * 1154 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1155 * Called to release previously enslaved netdev. 1156 * 1157 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1158 * struct sk_buff *skb, 1159 * bool all_slaves); 1160 * Get the xmit slave of master device. If all_slaves is true, function 1161 * assume all the slaves can transmit. 1162 * 1163 * Feature/offload setting functions. 1164 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1165 * netdev_features_t features); 1166 * Adjusts the requested feature flags according to device-specific 1167 * constraints, and returns the resulting flags. Must not modify 1168 * the device state. 1169 * 1170 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1171 * Called to update device configuration to new features. Passed 1172 * feature set might be less than what was returned by ndo_fix_features()). 1173 * Must return >0 or -errno if it changed dev->features itself. 1174 * 1175 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1176 * struct net_device *dev, 1177 * const unsigned char *addr, u16 vid, u16 flags, 1178 * struct netlink_ext_ack *extack); 1179 * Adds an FDB entry to dev for addr. 1180 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1181 * struct net_device *dev, 1182 * const unsigned char *addr, u16 vid) 1183 * Deletes the FDB entry from dev coresponding to addr. 1184 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1185 * struct net_device *dev, struct net_device *filter_dev, 1186 * int *idx) 1187 * Used to add FDB entries to dump requests. Implementers should add 1188 * entries to skb and update idx with the number of entries. 1189 * 1190 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1191 * u16 flags, struct netlink_ext_ack *extack) 1192 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1193 * struct net_device *dev, u32 filter_mask, 1194 * int nlflags) 1195 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1196 * u16 flags); 1197 * 1198 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1199 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1200 * which do not represent real hardware may define this to allow their 1201 * userspace components to manage their virtual carrier state. Devices 1202 * that determine carrier state from physical hardware properties (eg 1203 * network cables) or protocol-dependent mechanisms (eg 1204 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1205 * 1206 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1207 * struct netdev_phys_item_id *ppid); 1208 * Called to get ID of physical port of this device. If driver does 1209 * not implement this, it is assumed that the hw is not able to have 1210 * multiple net devices on single physical port. 1211 * 1212 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1213 * struct netdev_phys_item_id *ppid) 1214 * Called to get the parent ID of the physical port of this device. 1215 * 1216 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1217 * struct net_device *dev) 1218 * Called by upper layer devices to accelerate switching or other 1219 * station functionality into hardware. 'pdev is the lowerdev 1220 * to use for the offload and 'dev' is the net device that will 1221 * back the offload. Returns a pointer to the private structure 1222 * the upper layer will maintain. 1223 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1224 * Called by upper layer device to delete the station created 1225 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1226 * the station and priv is the structure returned by the add 1227 * operation. 1228 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1229 * int queue_index, u32 maxrate); 1230 * Called when a user wants to set a max-rate limitation of specific 1231 * TX queue. 1232 * int (*ndo_get_iflink)(const struct net_device *dev); 1233 * Called to get the iflink value of this device. 1234 * void (*ndo_change_proto_down)(struct net_device *dev, 1235 * bool proto_down); 1236 * This function is used to pass protocol port error state information 1237 * to the switch driver. The switch driver can react to the proto_down 1238 * by doing a phys down on the associated switch port. 1239 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1240 * This function is used to get egress tunnel information for given skb. 1241 * This is useful for retrieving outer tunnel header parameters while 1242 * sampling packet. 1243 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1244 * This function is used to specify the headroom that the skb must 1245 * consider when allocation skb during packet reception. Setting 1246 * appropriate rx headroom value allows avoiding skb head copy on 1247 * forward. Setting a negative value resets the rx headroom to the 1248 * default value. 1249 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1250 * This function is used to set or query state related to XDP on the 1251 * netdevice and manage BPF offload. See definition of 1252 * enum bpf_netdev_command for details. 1253 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1254 * u32 flags); 1255 * This function is used to submit @n XDP packets for transmit on a 1256 * netdevice. Returns number of frames successfully transmitted, frames 1257 * that got dropped are freed/returned via xdp_return_frame(). 1258 * Returns negative number, means general error invoking ndo, meaning 1259 * no frames were xmit'ed and core-caller will free all frames. 1260 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1261 * This function is used to wake up the softirq, ksoftirqd or kthread 1262 * responsible for sending and/or receiving packets on a specific 1263 * queue id bound to an AF_XDP socket. The flags field specifies if 1264 * only RX, only Tx, or both should be woken up using the flags 1265 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1266 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1267 * Get devlink port instance associated with a given netdev. 1268 * Called with a reference on the netdevice and devlink locks only, 1269 * rtnl_lock is not held. 1270 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1271 * int cmd); 1272 * Add, change, delete or get information on an IPv4 tunnel. 1273 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1274 * If a device is paired with a peer device, return the peer instance. 1275 * The caller must be under RCU read context. 1276 */ 1277 struct net_device_ops { 1278 int (*ndo_init)(struct net_device *dev); 1279 void (*ndo_uninit)(struct net_device *dev); 1280 int (*ndo_open)(struct net_device *dev); 1281 int (*ndo_stop)(struct net_device *dev); 1282 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1283 struct net_device *dev); 1284 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1285 struct net_device *dev, 1286 netdev_features_t features); 1287 u16 (*ndo_select_queue)(struct net_device *dev, 1288 struct sk_buff *skb, 1289 struct net_device *sb_dev); 1290 void (*ndo_change_rx_flags)(struct net_device *dev, 1291 int flags); 1292 void (*ndo_set_rx_mode)(struct net_device *dev); 1293 int (*ndo_set_mac_address)(struct net_device *dev, 1294 void *addr); 1295 int (*ndo_validate_addr)(struct net_device *dev); 1296 int (*ndo_do_ioctl)(struct net_device *dev, 1297 struct ifreq *ifr, int cmd); 1298 int (*ndo_set_config)(struct net_device *dev, 1299 struct ifmap *map); 1300 int (*ndo_change_mtu)(struct net_device *dev, 1301 int new_mtu); 1302 int (*ndo_neigh_setup)(struct net_device *dev, 1303 struct neigh_parms *); 1304 void (*ndo_tx_timeout) (struct net_device *dev, 1305 unsigned int txqueue); 1306 1307 void (*ndo_get_stats64)(struct net_device *dev, 1308 struct rtnl_link_stats64 *storage); 1309 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1310 int (*ndo_get_offload_stats)(int attr_id, 1311 const struct net_device *dev, 1312 void *attr_data); 1313 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1314 1315 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1316 __be16 proto, u16 vid); 1317 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1318 __be16 proto, u16 vid); 1319 #ifdef CONFIG_NET_POLL_CONTROLLER 1320 void (*ndo_poll_controller)(struct net_device *dev); 1321 int (*ndo_netpoll_setup)(struct net_device *dev, 1322 struct netpoll_info *info); 1323 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1324 #endif 1325 int (*ndo_set_vf_mac)(struct net_device *dev, 1326 int queue, u8 *mac); 1327 int (*ndo_set_vf_vlan)(struct net_device *dev, 1328 int queue, u16 vlan, 1329 u8 qos, __be16 proto); 1330 int (*ndo_set_vf_rate)(struct net_device *dev, 1331 int vf, int min_tx_rate, 1332 int max_tx_rate); 1333 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1334 int vf, bool setting); 1335 int (*ndo_set_vf_trust)(struct net_device *dev, 1336 int vf, bool setting); 1337 int (*ndo_get_vf_config)(struct net_device *dev, 1338 int vf, 1339 struct ifla_vf_info *ivf); 1340 int (*ndo_set_vf_link_state)(struct net_device *dev, 1341 int vf, int link_state); 1342 int (*ndo_get_vf_stats)(struct net_device *dev, 1343 int vf, 1344 struct ifla_vf_stats 1345 *vf_stats); 1346 int (*ndo_set_vf_port)(struct net_device *dev, 1347 int vf, 1348 struct nlattr *port[]); 1349 int (*ndo_get_vf_port)(struct net_device *dev, 1350 int vf, struct sk_buff *skb); 1351 int (*ndo_get_vf_guid)(struct net_device *dev, 1352 int vf, 1353 struct ifla_vf_guid *node_guid, 1354 struct ifla_vf_guid *port_guid); 1355 int (*ndo_set_vf_guid)(struct net_device *dev, 1356 int vf, u64 guid, 1357 int guid_type); 1358 int (*ndo_set_vf_rss_query_en)( 1359 struct net_device *dev, 1360 int vf, bool setting); 1361 int (*ndo_setup_tc)(struct net_device *dev, 1362 enum tc_setup_type type, 1363 void *type_data); 1364 #if IS_ENABLED(CONFIG_FCOE) 1365 int (*ndo_fcoe_enable)(struct net_device *dev); 1366 int (*ndo_fcoe_disable)(struct net_device *dev); 1367 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1368 u16 xid, 1369 struct scatterlist *sgl, 1370 unsigned int sgc); 1371 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1372 u16 xid); 1373 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1374 u16 xid, 1375 struct scatterlist *sgl, 1376 unsigned int sgc); 1377 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1378 struct netdev_fcoe_hbainfo *hbainfo); 1379 #endif 1380 1381 #if IS_ENABLED(CONFIG_LIBFCOE) 1382 #define NETDEV_FCOE_WWNN 0 1383 #define NETDEV_FCOE_WWPN 1 1384 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1385 u64 *wwn, int type); 1386 #endif 1387 1388 #ifdef CONFIG_RFS_ACCEL 1389 int (*ndo_rx_flow_steer)(struct net_device *dev, 1390 const struct sk_buff *skb, 1391 u16 rxq_index, 1392 u32 flow_id); 1393 #endif 1394 int (*ndo_add_slave)(struct net_device *dev, 1395 struct net_device *slave_dev, 1396 struct netlink_ext_ack *extack); 1397 int (*ndo_del_slave)(struct net_device *dev, 1398 struct net_device *slave_dev); 1399 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1400 struct sk_buff *skb, 1401 bool all_slaves); 1402 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1403 struct sock *sk); 1404 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1405 netdev_features_t features); 1406 int (*ndo_set_features)(struct net_device *dev, 1407 netdev_features_t features); 1408 int (*ndo_neigh_construct)(struct net_device *dev, 1409 struct neighbour *n); 1410 void (*ndo_neigh_destroy)(struct net_device *dev, 1411 struct neighbour *n); 1412 1413 int (*ndo_fdb_add)(struct ndmsg *ndm, 1414 struct nlattr *tb[], 1415 struct net_device *dev, 1416 const unsigned char *addr, 1417 u16 vid, 1418 u16 flags, 1419 struct netlink_ext_ack *extack); 1420 int (*ndo_fdb_del)(struct ndmsg *ndm, 1421 struct nlattr *tb[], 1422 struct net_device *dev, 1423 const unsigned char *addr, 1424 u16 vid); 1425 int (*ndo_fdb_dump)(struct sk_buff *skb, 1426 struct netlink_callback *cb, 1427 struct net_device *dev, 1428 struct net_device *filter_dev, 1429 int *idx); 1430 int (*ndo_fdb_get)(struct sk_buff *skb, 1431 struct nlattr *tb[], 1432 struct net_device *dev, 1433 const unsigned char *addr, 1434 u16 vid, u32 portid, u32 seq, 1435 struct netlink_ext_ack *extack); 1436 int (*ndo_bridge_setlink)(struct net_device *dev, 1437 struct nlmsghdr *nlh, 1438 u16 flags, 1439 struct netlink_ext_ack *extack); 1440 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1441 u32 pid, u32 seq, 1442 struct net_device *dev, 1443 u32 filter_mask, 1444 int nlflags); 1445 int (*ndo_bridge_dellink)(struct net_device *dev, 1446 struct nlmsghdr *nlh, 1447 u16 flags); 1448 int (*ndo_change_carrier)(struct net_device *dev, 1449 bool new_carrier); 1450 int (*ndo_get_phys_port_id)(struct net_device *dev, 1451 struct netdev_phys_item_id *ppid); 1452 int (*ndo_get_port_parent_id)(struct net_device *dev, 1453 struct netdev_phys_item_id *ppid); 1454 int (*ndo_get_phys_port_name)(struct net_device *dev, 1455 char *name, size_t len); 1456 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1457 struct net_device *dev); 1458 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1459 void *priv); 1460 1461 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1462 int queue_index, 1463 u32 maxrate); 1464 int (*ndo_get_iflink)(const struct net_device *dev); 1465 int (*ndo_change_proto_down)(struct net_device *dev, 1466 bool proto_down); 1467 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1468 struct sk_buff *skb); 1469 void (*ndo_set_rx_headroom)(struct net_device *dev, 1470 int needed_headroom); 1471 int (*ndo_bpf)(struct net_device *dev, 1472 struct netdev_bpf *bpf); 1473 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1474 struct xdp_frame **xdp, 1475 u32 flags); 1476 int (*ndo_xsk_wakeup)(struct net_device *dev, 1477 u32 queue_id, u32 flags); 1478 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1479 int (*ndo_tunnel_ctl)(struct net_device *dev, 1480 struct ip_tunnel_parm *p, int cmd); 1481 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1482 }; 1483 1484 /** 1485 * enum netdev_priv_flags - &struct net_device priv_flags 1486 * 1487 * These are the &struct net_device, they are only set internally 1488 * by drivers and used in the kernel. These flags are invisible to 1489 * userspace; this means that the order of these flags can change 1490 * during any kernel release. 1491 * 1492 * You should have a pretty good reason to be extending these flags. 1493 * 1494 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1495 * @IFF_EBRIDGE: Ethernet bridging device 1496 * @IFF_BONDING: bonding master or slave 1497 * @IFF_ISATAP: ISATAP interface (RFC4214) 1498 * @IFF_WAN_HDLC: WAN HDLC device 1499 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1500 * release skb->dst 1501 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1502 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1503 * @IFF_MACVLAN_PORT: device used as macvlan port 1504 * @IFF_BRIDGE_PORT: device used as bridge port 1505 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1506 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1507 * @IFF_UNICAST_FLT: Supports unicast filtering 1508 * @IFF_TEAM_PORT: device used as team port 1509 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1510 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1511 * change when it's running 1512 * @IFF_MACVLAN: Macvlan device 1513 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1514 * underlying stacked devices 1515 * @IFF_L3MDEV_MASTER: device is an L3 master device 1516 * @IFF_NO_QUEUE: device can run without qdisc attached 1517 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1518 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1519 * @IFF_TEAM: device is a team device 1520 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1521 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1522 * entity (i.e. the master device for bridged veth) 1523 * @IFF_MACSEC: device is a MACsec device 1524 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1525 * @IFF_FAILOVER: device is a failover master device 1526 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1527 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1528 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1529 */ 1530 enum netdev_priv_flags { 1531 IFF_802_1Q_VLAN = 1<<0, 1532 IFF_EBRIDGE = 1<<1, 1533 IFF_BONDING = 1<<2, 1534 IFF_ISATAP = 1<<3, 1535 IFF_WAN_HDLC = 1<<4, 1536 IFF_XMIT_DST_RELEASE = 1<<5, 1537 IFF_DONT_BRIDGE = 1<<6, 1538 IFF_DISABLE_NETPOLL = 1<<7, 1539 IFF_MACVLAN_PORT = 1<<8, 1540 IFF_BRIDGE_PORT = 1<<9, 1541 IFF_OVS_DATAPATH = 1<<10, 1542 IFF_TX_SKB_SHARING = 1<<11, 1543 IFF_UNICAST_FLT = 1<<12, 1544 IFF_TEAM_PORT = 1<<13, 1545 IFF_SUPP_NOFCS = 1<<14, 1546 IFF_LIVE_ADDR_CHANGE = 1<<15, 1547 IFF_MACVLAN = 1<<16, 1548 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1549 IFF_L3MDEV_MASTER = 1<<18, 1550 IFF_NO_QUEUE = 1<<19, 1551 IFF_OPENVSWITCH = 1<<20, 1552 IFF_L3MDEV_SLAVE = 1<<21, 1553 IFF_TEAM = 1<<22, 1554 IFF_RXFH_CONFIGURED = 1<<23, 1555 IFF_PHONY_HEADROOM = 1<<24, 1556 IFF_MACSEC = 1<<25, 1557 IFF_NO_RX_HANDLER = 1<<26, 1558 IFF_FAILOVER = 1<<27, 1559 IFF_FAILOVER_SLAVE = 1<<28, 1560 IFF_L3MDEV_RX_HANDLER = 1<<29, 1561 IFF_LIVE_RENAME_OK = 1<<30, 1562 }; 1563 1564 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1565 #define IFF_EBRIDGE IFF_EBRIDGE 1566 #define IFF_BONDING IFF_BONDING 1567 #define IFF_ISATAP IFF_ISATAP 1568 #define IFF_WAN_HDLC IFF_WAN_HDLC 1569 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1570 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1571 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1572 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1573 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1574 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1575 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1576 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1577 #define IFF_TEAM_PORT IFF_TEAM_PORT 1578 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1579 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1580 #define IFF_MACVLAN IFF_MACVLAN 1581 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1582 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1583 #define IFF_NO_QUEUE IFF_NO_QUEUE 1584 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1585 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1586 #define IFF_TEAM IFF_TEAM 1587 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1588 #define IFF_MACSEC IFF_MACSEC 1589 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1590 #define IFF_FAILOVER IFF_FAILOVER 1591 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1592 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1593 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1594 1595 /** 1596 * struct net_device - The DEVICE structure. 1597 * 1598 * Actually, this whole structure is a big mistake. It mixes I/O 1599 * data with strictly "high-level" data, and it has to know about 1600 * almost every data structure used in the INET module. 1601 * 1602 * @name: This is the first field of the "visible" part of this structure 1603 * (i.e. as seen by users in the "Space.c" file). It is the name 1604 * of the interface. 1605 * 1606 * @name_node: Name hashlist node 1607 * @ifalias: SNMP alias 1608 * @mem_end: Shared memory end 1609 * @mem_start: Shared memory start 1610 * @base_addr: Device I/O address 1611 * @irq: Device IRQ number 1612 * 1613 * @state: Generic network queuing layer state, see netdev_state_t 1614 * @dev_list: The global list of network devices 1615 * @napi_list: List entry used for polling NAPI devices 1616 * @unreg_list: List entry when we are unregistering the 1617 * device; see the function unregister_netdev 1618 * @close_list: List entry used when we are closing the device 1619 * @ptype_all: Device-specific packet handlers for all protocols 1620 * @ptype_specific: Device-specific, protocol-specific packet handlers 1621 * 1622 * @adj_list: Directly linked devices, like slaves for bonding 1623 * @features: Currently active device features 1624 * @hw_features: User-changeable features 1625 * 1626 * @wanted_features: User-requested features 1627 * @vlan_features: Mask of features inheritable by VLAN devices 1628 * 1629 * @hw_enc_features: Mask of features inherited by encapsulating devices 1630 * This field indicates what encapsulation 1631 * offloads the hardware is capable of doing, 1632 * and drivers will need to set them appropriately. 1633 * 1634 * @mpls_features: Mask of features inheritable by MPLS 1635 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1636 * 1637 * @ifindex: interface index 1638 * @group: The group the device belongs to 1639 * 1640 * @stats: Statistics struct, which was left as a legacy, use 1641 * rtnl_link_stats64 instead 1642 * 1643 * @rx_dropped: Dropped packets by core network, 1644 * do not use this in drivers 1645 * @tx_dropped: Dropped packets by core network, 1646 * do not use this in drivers 1647 * @rx_nohandler: nohandler dropped packets by core network on 1648 * inactive devices, do not use this in drivers 1649 * @carrier_up_count: Number of times the carrier has been up 1650 * @carrier_down_count: Number of times the carrier has been down 1651 * 1652 * @wireless_handlers: List of functions to handle Wireless Extensions, 1653 * instead of ioctl, 1654 * see <net/iw_handler.h> for details. 1655 * @wireless_data: Instance data managed by the core of wireless extensions 1656 * 1657 * @netdev_ops: Includes several pointers to callbacks, 1658 * if one wants to override the ndo_*() functions 1659 * @ethtool_ops: Management operations 1660 * @l3mdev_ops: Layer 3 master device operations 1661 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1662 * discovery handling. Necessary for e.g. 6LoWPAN. 1663 * @xfrmdev_ops: Transformation offload operations 1664 * @tlsdev_ops: Transport Layer Security offload operations 1665 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1666 * of Layer 2 headers. 1667 * 1668 * @flags: Interface flags (a la BSD) 1669 * @priv_flags: Like 'flags' but invisible to userspace, 1670 * see if.h for the definitions 1671 * @gflags: Global flags ( kept as legacy ) 1672 * @padded: How much padding added by alloc_netdev() 1673 * @operstate: RFC2863 operstate 1674 * @link_mode: Mapping policy to operstate 1675 * @if_port: Selectable AUI, TP, ... 1676 * @dma: DMA channel 1677 * @mtu: Interface MTU value 1678 * @min_mtu: Interface Minimum MTU value 1679 * @max_mtu: Interface Maximum MTU value 1680 * @type: Interface hardware type 1681 * @hard_header_len: Maximum hardware header length. 1682 * @min_header_len: Minimum hardware header length 1683 * 1684 * @needed_headroom: Extra headroom the hardware may need, but not in all 1685 * cases can this be guaranteed 1686 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1687 * cases can this be guaranteed. Some cases also use 1688 * LL_MAX_HEADER instead to allocate the skb 1689 * 1690 * interface address info: 1691 * 1692 * @perm_addr: Permanent hw address 1693 * @addr_assign_type: Hw address assignment type 1694 * @addr_len: Hardware address length 1695 * @upper_level: Maximum depth level of upper devices. 1696 * @lower_level: Maximum depth level of lower devices. 1697 * @neigh_priv_len: Used in neigh_alloc() 1698 * @dev_id: Used to differentiate devices that share 1699 * the same link layer address 1700 * @dev_port: Used to differentiate devices that share 1701 * the same function 1702 * @addr_list_lock: XXX: need comments on this one 1703 * @name_assign_type: network interface name assignment type 1704 * @uc_promisc: Counter that indicates promiscuous mode 1705 * has been enabled due to the need to listen to 1706 * additional unicast addresses in a device that 1707 * does not implement ndo_set_rx_mode() 1708 * @uc: unicast mac addresses 1709 * @mc: multicast mac addresses 1710 * @dev_addrs: list of device hw addresses 1711 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1712 * @promiscuity: Number of times the NIC is told to work in 1713 * promiscuous mode; if it becomes 0 the NIC will 1714 * exit promiscuous mode 1715 * @allmulti: Counter, enables or disables allmulticast mode 1716 * 1717 * @vlan_info: VLAN info 1718 * @dsa_ptr: dsa specific data 1719 * @tipc_ptr: TIPC specific data 1720 * @atalk_ptr: AppleTalk link 1721 * @ip_ptr: IPv4 specific data 1722 * @dn_ptr: DECnet specific data 1723 * @ip6_ptr: IPv6 specific data 1724 * @ax25_ptr: AX.25 specific data 1725 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1726 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1727 * device struct 1728 * @mpls_ptr: mpls_dev struct pointer 1729 * 1730 * @dev_addr: Hw address (before bcast, 1731 * because most packets are unicast) 1732 * 1733 * @_rx: Array of RX queues 1734 * @num_rx_queues: Number of RX queues 1735 * allocated at register_netdev() time 1736 * @real_num_rx_queues: Number of RX queues currently active in device 1737 * @xdp_prog: XDP sockets filter program pointer 1738 * @gro_flush_timeout: timeout for GRO layer in NAPI 1739 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1740 * allow to avoid NIC hard IRQ, on busy queues. 1741 * 1742 * @rx_handler: handler for received packets 1743 * @rx_handler_data: XXX: need comments on this one 1744 * @miniq_ingress: ingress/clsact qdisc specific data for 1745 * ingress processing 1746 * @ingress_queue: XXX: need comments on this one 1747 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1748 * @broadcast: hw bcast address 1749 * 1750 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1751 * indexed by RX queue number. Assigned by driver. 1752 * This must only be set if the ndo_rx_flow_steer 1753 * operation is defined 1754 * @index_hlist: Device index hash chain 1755 * 1756 * @_tx: Array of TX queues 1757 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1758 * @real_num_tx_queues: Number of TX queues currently active in device 1759 * @qdisc: Root qdisc from userspace point of view 1760 * @tx_queue_len: Max frames per queue allowed 1761 * @tx_global_lock: XXX: need comments on this one 1762 * @xdp_bulkq: XDP device bulk queue 1763 * @xps_cpus_map: all CPUs map for XPS device 1764 * @xps_rxqs_map: all RXQs map for XPS device 1765 * 1766 * @xps_maps: XXX: need comments on this one 1767 * @miniq_egress: clsact qdisc specific data for 1768 * egress processing 1769 * @qdisc_hash: qdisc hash table 1770 * @watchdog_timeo: Represents the timeout that is used by 1771 * the watchdog (see dev_watchdog()) 1772 * @watchdog_timer: List of timers 1773 * 1774 * @proto_down_reason: reason a netdev interface is held down 1775 * @pcpu_refcnt: Number of references to this device 1776 * @todo_list: Delayed register/unregister 1777 * @link_watch_list: XXX: need comments on this one 1778 * 1779 * @reg_state: Register/unregister state machine 1780 * @dismantle: Device is going to be freed 1781 * @rtnl_link_state: This enum represents the phases of creating 1782 * a new link 1783 * 1784 * @needs_free_netdev: Should unregister perform free_netdev? 1785 * @priv_destructor: Called from unregister 1786 * @npinfo: XXX: need comments on this one 1787 * @nd_net: Network namespace this network device is inside 1788 * 1789 * @ml_priv: Mid-layer private 1790 * @lstats: Loopback statistics 1791 * @tstats: Tunnel statistics 1792 * @dstats: Dummy statistics 1793 * @vstats: Virtual ethernet statistics 1794 * 1795 * @garp_port: GARP 1796 * @mrp_port: MRP 1797 * 1798 * @dev: Class/net/name entry 1799 * @sysfs_groups: Space for optional device, statistics and wireless 1800 * sysfs groups 1801 * 1802 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1803 * @rtnl_link_ops: Rtnl_link_ops 1804 * 1805 * @gso_max_size: Maximum size of generic segmentation offload 1806 * @gso_max_segs: Maximum number of segments that can be passed to the 1807 * NIC for GSO 1808 * 1809 * @dcbnl_ops: Data Center Bridging netlink ops 1810 * @num_tc: Number of traffic classes in the net device 1811 * @tc_to_txq: XXX: need comments on this one 1812 * @prio_tc_map: XXX: need comments on this one 1813 * 1814 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1815 * 1816 * @priomap: XXX: need comments on this one 1817 * @phydev: Physical device may attach itself 1818 * for hardware timestamping 1819 * @sfp_bus: attached &struct sfp_bus structure. 1820 * 1821 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1822 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1823 * 1824 * @proto_down: protocol port state information can be sent to the 1825 * switch driver and used to set the phys state of the 1826 * switch port. 1827 * 1828 * @wol_enabled: Wake-on-LAN is enabled 1829 * 1830 * @net_notifier_list: List of per-net netdev notifier block 1831 * that follow this device when it is moved 1832 * to another network namespace. 1833 * 1834 * @macsec_ops: MACsec offloading ops 1835 * 1836 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1837 * offload capabilities of the device 1838 * @udp_tunnel_nic: UDP tunnel offload state 1839 * @xdp_state: stores info on attached XDP BPF programs 1840 * 1841 * @nested_level: Used as as a parameter of spin_lock_nested() of 1842 * dev->addr_list_lock. 1843 * @unlink_list: As netif_addr_lock() can be called recursively, 1844 * keep a list of interfaces to be deleted. 1845 * 1846 * FIXME: cleanup struct net_device such that network protocol info 1847 * moves out. 1848 */ 1849 1850 struct net_device { 1851 char name[IFNAMSIZ]; 1852 struct netdev_name_node *name_node; 1853 struct dev_ifalias __rcu *ifalias; 1854 /* 1855 * I/O specific fields 1856 * FIXME: Merge these and struct ifmap into one 1857 */ 1858 unsigned long mem_end; 1859 unsigned long mem_start; 1860 unsigned long base_addr; 1861 int irq; 1862 1863 /* 1864 * Some hardware also needs these fields (state,dev_list, 1865 * napi_list,unreg_list,close_list) but they are not 1866 * part of the usual set specified in Space.c. 1867 */ 1868 1869 unsigned long state; 1870 1871 struct list_head dev_list; 1872 struct list_head napi_list; 1873 struct list_head unreg_list; 1874 struct list_head close_list; 1875 struct list_head ptype_all; 1876 struct list_head ptype_specific; 1877 1878 struct { 1879 struct list_head upper; 1880 struct list_head lower; 1881 } adj_list; 1882 1883 netdev_features_t features; 1884 netdev_features_t hw_features; 1885 netdev_features_t wanted_features; 1886 netdev_features_t vlan_features; 1887 netdev_features_t hw_enc_features; 1888 netdev_features_t mpls_features; 1889 netdev_features_t gso_partial_features; 1890 1891 int ifindex; 1892 int group; 1893 1894 struct net_device_stats stats; 1895 1896 atomic_long_t rx_dropped; 1897 atomic_long_t tx_dropped; 1898 atomic_long_t rx_nohandler; 1899 1900 /* Stats to monitor link on/off, flapping */ 1901 atomic_t carrier_up_count; 1902 atomic_t carrier_down_count; 1903 1904 #ifdef CONFIG_WIRELESS_EXT 1905 const struct iw_handler_def *wireless_handlers; 1906 struct iw_public_data *wireless_data; 1907 #endif 1908 const struct net_device_ops *netdev_ops; 1909 const struct ethtool_ops *ethtool_ops; 1910 #ifdef CONFIG_NET_L3_MASTER_DEV 1911 const struct l3mdev_ops *l3mdev_ops; 1912 #endif 1913 #if IS_ENABLED(CONFIG_IPV6) 1914 const struct ndisc_ops *ndisc_ops; 1915 #endif 1916 1917 #ifdef CONFIG_XFRM_OFFLOAD 1918 const struct xfrmdev_ops *xfrmdev_ops; 1919 #endif 1920 1921 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1922 const struct tlsdev_ops *tlsdev_ops; 1923 #endif 1924 1925 const struct header_ops *header_ops; 1926 1927 unsigned int flags; 1928 unsigned int priv_flags; 1929 1930 unsigned short gflags; 1931 unsigned short padded; 1932 1933 unsigned char operstate; 1934 unsigned char link_mode; 1935 1936 unsigned char if_port; 1937 unsigned char dma; 1938 1939 /* Note : dev->mtu is often read without holding a lock. 1940 * Writers usually hold RTNL. 1941 * It is recommended to use READ_ONCE() to annotate the reads, 1942 * and to use WRITE_ONCE() to annotate the writes. 1943 */ 1944 unsigned int mtu; 1945 unsigned int min_mtu; 1946 unsigned int max_mtu; 1947 unsigned short type; 1948 unsigned short hard_header_len; 1949 unsigned char min_header_len; 1950 unsigned char name_assign_type; 1951 1952 unsigned short needed_headroom; 1953 unsigned short needed_tailroom; 1954 1955 /* Interface address info. */ 1956 unsigned char perm_addr[MAX_ADDR_LEN]; 1957 unsigned char addr_assign_type; 1958 unsigned char addr_len; 1959 unsigned char upper_level; 1960 unsigned char lower_level; 1961 1962 unsigned short neigh_priv_len; 1963 unsigned short dev_id; 1964 unsigned short dev_port; 1965 spinlock_t addr_list_lock; 1966 1967 struct netdev_hw_addr_list uc; 1968 struct netdev_hw_addr_list mc; 1969 struct netdev_hw_addr_list dev_addrs; 1970 1971 #ifdef CONFIG_SYSFS 1972 struct kset *queues_kset; 1973 #endif 1974 #ifdef CONFIG_LOCKDEP 1975 struct list_head unlink_list; 1976 #endif 1977 unsigned int promiscuity; 1978 unsigned int allmulti; 1979 bool uc_promisc; 1980 #ifdef CONFIG_LOCKDEP 1981 unsigned char nested_level; 1982 #endif 1983 1984 1985 /* Protocol-specific pointers */ 1986 1987 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1988 struct vlan_info __rcu *vlan_info; 1989 #endif 1990 #if IS_ENABLED(CONFIG_NET_DSA) 1991 struct dsa_port *dsa_ptr; 1992 #endif 1993 #if IS_ENABLED(CONFIG_TIPC) 1994 struct tipc_bearer __rcu *tipc_ptr; 1995 #endif 1996 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1997 void *atalk_ptr; 1998 #endif 1999 struct in_device __rcu *ip_ptr; 2000 #if IS_ENABLED(CONFIG_DECNET) 2001 struct dn_dev __rcu *dn_ptr; 2002 #endif 2003 struct inet6_dev __rcu *ip6_ptr; 2004 #if IS_ENABLED(CONFIG_AX25) 2005 void *ax25_ptr; 2006 #endif 2007 struct wireless_dev *ieee80211_ptr; 2008 struct wpan_dev *ieee802154_ptr; 2009 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2010 struct mpls_dev __rcu *mpls_ptr; 2011 #endif 2012 2013 /* 2014 * Cache lines mostly used on receive path (including eth_type_trans()) 2015 */ 2016 /* Interface address info used in eth_type_trans() */ 2017 unsigned char *dev_addr; 2018 2019 struct netdev_rx_queue *_rx; 2020 unsigned int num_rx_queues; 2021 unsigned int real_num_rx_queues; 2022 2023 struct bpf_prog __rcu *xdp_prog; 2024 unsigned long gro_flush_timeout; 2025 int napi_defer_hard_irqs; 2026 rx_handler_func_t __rcu *rx_handler; 2027 void __rcu *rx_handler_data; 2028 2029 #ifdef CONFIG_NET_CLS_ACT 2030 struct mini_Qdisc __rcu *miniq_ingress; 2031 #endif 2032 struct netdev_queue __rcu *ingress_queue; 2033 #ifdef CONFIG_NETFILTER_INGRESS 2034 struct nf_hook_entries __rcu *nf_hooks_ingress; 2035 #endif 2036 2037 unsigned char broadcast[MAX_ADDR_LEN]; 2038 #ifdef CONFIG_RFS_ACCEL 2039 struct cpu_rmap *rx_cpu_rmap; 2040 #endif 2041 struct hlist_node index_hlist; 2042 2043 /* 2044 * Cache lines mostly used on transmit path 2045 */ 2046 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2047 unsigned int num_tx_queues; 2048 unsigned int real_num_tx_queues; 2049 struct Qdisc *qdisc; 2050 unsigned int tx_queue_len; 2051 spinlock_t tx_global_lock; 2052 2053 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2054 2055 #ifdef CONFIG_XPS 2056 struct xps_dev_maps __rcu *xps_cpus_map; 2057 struct xps_dev_maps __rcu *xps_rxqs_map; 2058 #endif 2059 #ifdef CONFIG_NET_CLS_ACT 2060 struct mini_Qdisc __rcu *miniq_egress; 2061 #endif 2062 2063 #ifdef CONFIG_NET_SCHED 2064 DECLARE_HASHTABLE (qdisc_hash, 4); 2065 #endif 2066 /* These may be needed for future network-power-down code. */ 2067 struct timer_list watchdog_timer; 2068 int watchdog_timeo; 2069 2070 u32 proto_down_reason; 2071 2072 struct list_head todo_list; 2073 int __percpu *pcpu_refcnt; 2074 2075 struct list_head link_watch_list; 2076 2077 enum { NETREG_UNINITIALIZED=0, 2078 NETREG_REGISTERED, /* completed register_netdevice */ 2079 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2080 NETREG_UNREGISTERED, /* completed unregister todo */ 2081 NETREG_RELEASED, /* called free_netdev */ 2082 NETREG_DUMMY, /* dummy device for NAPI poll */ 2083 } reg_state:8; 2084 2085 bool dismantle; 2086 2087 enum { 2088 RTNL_LINK_INITIALIZED, 2089 RTNL_LINK_INITIALIZING, 2090 } rtnl_link_state:16; 2091 2092 bool needs_free_netdev; 2093 void (*priv_destructor)(struct net_device *dev); 2094 2095 #ifdef CONFIG_NETPOLL 2096 struct netpoll_info __rcu *npinfo; 2097 #endif 2098 2099 possible_net_t nd_net; 2100 2101 /* mid-layer private */ 2102 union { 2103 void *ml_priv; 2104 struct pcpu_lstats __percpu *lstats; 2105 struct pcpu_sw_netstats __percpu *tstats; 2106 struct pcpu_dstats __percpu *dstats; 2107 }; 2108 2109 #if IS_ENABLED(CONFIG_GARP) 2110 struct garp_port __rcu *garp_port; 2111 #endif 2112 #if IS_ENABLED(CONFIG_MRP) 2113 struct mrp_port __rcu *mrp_port; 2114 #endif 2115 2116 struct device dev; 2117 const struct attribute_group *sysfs_groups[4]; 2118 const struct attribute_group *sysfs_rx_queue_group; 2119 2120 const struct rtnl_link_ops *rtnl_link_ops; 2121 2122 /* for setting kernel sock attribute on TCP connection setup */ 2123 #define GSO_MAX_SIZE 65536 2124 unsigned int gso_max_size; 2125 #define GSO_MAX_SEGS 65535 2126 u16 gso_max_segs; 2127 2128 #ifdef CONFIG_DCB 2129 const struct dcbnl_rtnl_ops *dcbnl_ops; 2130 #endif 2131 s16 num_tc; 2132 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2133 u8 prio_tc_map[TC_BITMASK + 1]; 2134 2135 #if IS_ENABLED(CONFIG_FCOE) 2136 unsigned int fcoe_ddp_xid; 2137 #endif 2138 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2139 struct netprio_map __rcu *priomap; 2140 #endif 2141 struct phy_device *phydev; 2142 struct sfp_bus *sfp_bus; 2143 struct lock_class_key *qdisc_tx_busylock; 2144 struct lock_class_key *qdisc_running_key; 2145 bool proto_down; 2146 unsigned wol_enabled:1; 2147 2148 struct list_head net_notifier_list; 2149 2150 #if IS_ENABLED(CONFIG_MACSEC) 2151 /* MACsec management functions */ 2152 const struct macsec_ops *macsec_ops; 2153 #endif 2154 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2155 struct udp_tunnel_nic *udp_tunnel_nic; 2156 2157 /* protected by rtnl_lock */ 2158 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2159 }; 2160 #define to_net_dev(d) container_of(d, struct net_device, dev) 2161 2162 static inline bool netif_elide_gro(const struct net_device *dev) 2163 { 2164 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2165 return true; 2166 return false; 2167 } 2168 2169 #define NETDEV_ALIGN 32 2170 2171 static inline 2172 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2173 { 2174 return dev->prio_tc_map[prio & TC_BITMASK]; 2175 } 2176 2177 static inline 2178 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2179 { 2180 if (tc >= dev->num_tc) 2181 return -EINVAL; 2182 2183 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2184 return 0; 2185 } 2186 2187 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2188 void netdev_reset_tc(struct net_device *dev); 2189 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2190 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2191 2192 static inline 2193 int netdev_get_num_tc(struct net_device *dev) 2194 { 2195 return dev->num_tc; 2196 } 2197 2198 static inline void net_prefetch(void *p) 2199 { 2200 prefetch(p); 2201 #if L1_CACHE_BYTES < 128 2202 prefetch((u8 *)p + L1_CACHE_BYTES); 2203 #endif 2204 } 2205 2206 static inline void net_prefetchw(void *p) 2207 { 2208 prefetchw(p); 2209 #if L1_CACHE_BYTES < 128 2210 prefetchw((u8 *)p + L1_CACHE_BYTES); 2211 #endif 2212 } 2213 2214 void netdev_unbind_sb_channel(struct net_device *dev, 2215 struct net_device *sb_dev); 2216 int netdev_bind_sb_channel_queue(struct net_device *dev, 2217 struct net_device *sb_dev, 2218 u8 tc, u16 count, u16 offset); 2219 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2220 static inline int netdev_get_sb_channel(struct net_device *dev) 2221 { 2222 return max_t(int, -dev->num_tc, 0); 2223 } 2224 2225 static inline 2226 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2227 unsigned int index) 2228 { 2229 return &dev->_tx[index]; 2230 } 2231 2232 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2233 const struct sk_buff *skb) 2234 { 2235 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2236 } 2237 2238 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2239 void (*f)(struct net_device *, 2240 struct netdev_queue *, 2241 void *), 2242 void *arg) 2243 { 2244 unsigned int i; 2245 2246 for (i = 0; i < dev->num_tx_queues; i++) 2247 f(dev, &dev->_tx[i], arg); 2248 } 2249 2250 #define netdev_lockdep_set_classes(dev) \ 2251 { \ 2252 static struct lock_class_key qdisc_tx_busylock_key; \ 2253 static struct lock_class_key qdisc_running_key; \ 2254 static struct lock_class_key qdisc_xmit_lock_key; \ 2255 static struct lock_class_key dev_addr_list_lock_key; \ 2256 unsigned int i; \ 2257 \ 2258 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2259 (dev)->qdisc_running_key = &qdisc_running_key; \ 2260 lockdep_set_class(&(dev)->addr_list_lock, \ 2261 &dev_addr_list_lock_key); \ 2262 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2263 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2264 &qdisc_xmit_lock_key); \ 2265 } 2266 2267 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2268 struct net_device *sb_dev); 2269 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2270 struct sk_buff *skb, 2271 struct net_device *sb_dev); 2272 2273 /* returns the headroom that the master device needs to take in account 2274 * when forwarding to this dev 2275 */ 2276 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2277 { 2278 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2279 } 2280 2281 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2282 { 2283 if (dev->netdev_ops->ndo_set_rx_headroom) 2284 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2285 } 2286 2287 /* set the device rx headroom to the dev's default */ 2288 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2289 { 2290 netdev_set_rx_headroom(dev, -1); 2291 } 2292 2293 /* 2294 * Net namespace inlines 2295 */ 2296 static inline 2297 struct net *dev_net(const struct net_device *dev) 2298 { 2299 return read_pnet(&dev->nd_net); 2300 } 2301 2302 static inline 2303 void dev_net_set(struct net_device *dev, struct net *net) 2304 { 2305 write_pnet(&dev->nd_net, net); 2306 } 2307 2308 /** 2309 * netdev_priv - access network device private data 2310 * @dev: network device 2311 * 2312 * Get network device private data 2313 */ 2314 static inline void *netdev_priv(const struct net_device *dev) 2315 { 2316 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2317 } 2318 2319 /* Set the sysfs physical device reference for the network logical device 2320 * if set prior to registration will cause a symlink during initialization. 2321 */ 2322 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2323 2324 /* Set the sysfs device type for the network logical device to allow 2325 * fine-grained identification of different network device types. For 2326 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2327 */ 2328 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2329 2330 /* Default NAPI poll() weight 2331 * Device drivers are strongly advised to not use bigger value 2332 */ 2333 #define NAPI_POLL_WEIGHT 64 2334 2335 /** 2336 * netif_napi_add - initialize a NAPI context 2337 * @dev: network device 2338 * @napi: NAPI context 2339 * @poll: polling function 2340 * @weight: default weight 2341 * 2342 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2343 * *any* of the other NAPI-related functions. 2344 */ 2345 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2346 int (*poll)(struct napi_struct *, int), int weight); 2347 2348 /** 2349 * netif_tx_napi_add - initialize a NAPI context 2350 * @dev: network device 2351 * @napi: NAPI context 2352 * @poll: polling function 2353 * @weight: default weight 2354 * 2355 * This variant of netif_napi_add() should be used from drivers using NAPI 2356 * to exclusively poll a TX queue. 2357 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2358 */ 2359 static inline void netif_tx_napi_add(struct net_device *dev, 2360 struct napi_struct *napi, 2361 int (*poll)(struct napi_struct *, int), 2362 int weight) 2363 { 2364 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2365 netif_napi_add(dev, napi, poll, weight); 2366 } 2367 2368 /** 2369 * __netif_napi_del - remove a NAPI context 2370 * @napi: NAPI context 2371 * 2372 * Warning: caller must observe RCU grace period before freeing memory 2373 * containing @napi. Drivers might want to call this helper to combine 2374 * all the needed RCU grace periods into a single one. 2375 */ 2376 void __netif_napi_del(struct napi_struct *napi); 2377 2378 /** 2379 * netif_napi_del - remove a NAPI context 2380 * @napi: NAPI context 2381 * 2382 * netif_napi_del() removes a NAPI context from the network device NAPI list 2383 */ 2384 static inline void netif_napi_del(struct napi_struct *napi) 2385 { 2386 __netif_napi_del(napi); 2387 synchronize_net(); 2388 } 2389 2390 struct napi_gro_cb { 2391 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2392 void *frag0; 2393 2394 /* Length of frag0. */ 2395 unsigned int frag0_len; 2396 2397 /* This indicates where we are processing relative to skb->data. */ 2398 int data_offset; 2399 2400 /* This is non-zero if the packet cannot be merged with the new skb. */ 2401 u16 flush; 2402 2403 /* Save the IP ID here and check when we get to the transport layer */ 2404 u16 flush_id; 2405 2406 /* Number of segments aggregated. */ 2407 u16 count; 2408 2409 /* Start offset for remote checksum offload */ 2410 u16 gro_remcsum_start; 2411 2412 /* jiffies when first packet was created/queued */ 2413 unsigned long age; 2414 2415 /* Used in ipv6_gro_receive() and foo-over-udp */ 2416 u16 proto; 2417 2418 /* This is non-zero if the packet may be of the same flow. */ 2419 u8 same_flow:1; 2420 2421 /* Used in tunnel GRO receive */ 2422 u8 encap_mark:1; 2423 2424 /* GRO checksum is valid */ 2425 u8 csum_valid:1; 2426 2427 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2428 u8 csum_cnt:3; 2429 2430 /* Free the skb? */ 2431 u8 free:2; 2432 #define NAPI_GRO_FREE 1 2433 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2434 2435 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2436 u8 is_ipv6:1; 2437 2438 /* Used in GRE, set in fou/gue_gro_receive */ 2439 u8 is_fou:1; 2440 2441 /* Used to determine if flush_id can be ignored */ 2442 u8 is_atomic:1; 2443 2444 /* Number of gro_receive callbacks this packet already went through */ 2445 u8 recursion_counter:4; 2446 2447 /* GRO is done by frag_list pointer chaining. */ 2448 u8 is_flist:1; 2449 2450 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2451 __wsum csum; 2452 2453 /* used in skb_gro_receive() slow path */ 2454 struct sk_buff *last; 2455 }; 2456 2457 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2458 2459 #define GRO_RECURSION_LIMIT 15 2460 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2461 { 2462 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2463 } 2464 2465 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2466 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2467 struct list_head *head, 2468 struct sk_buff *skb) 2469 { 2470 if (unlikely(gro_recursion_inc_test(skb))) { 2471 NAPI_GRO_CB(skb)->flush |= 1; 2472 return NULL; 2473 } 2474 2475 return cb(head, skb); 2476 } 2477 2478 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2479 struct sk_buff *); 2480 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2481 struct sock *sk, 2482 struct list_head *head, 2483 struct sk_buff *skb) 2484 { 2485 if (unlikely(gro_recursion_inc_test(skb))) { 2486 NAPI_GRO_CB(skb)->flush |= 1; 2487 return NULL; 2488 } 2489 2490 return cb(sk, head, skb); 2491 } 2492 2493 struct packet_type { 2494 __be16 type; /* This is really htons(ether_type). */ 2495 bool ignore_outgoing; 2496 struct net_device *dev; /* NULL is wildcarded here */ 2497 int (*func) (struct sk_buff *, 2498 struct net_device *, 2499 struct packet_type *, 2500 struct net_device *); 2501 void (*list_func) (struct list_head *, 2502 struct packet_type *, 2503 struct net_device *); 2504 bool (*id_match)(struct packet_type *ptype, 2505 struct sock *sk); 2506 void *af_packet_priv; 2507 struct list_head list; 2508 }; 2509 2510 struct offload_callbacks { 2511 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2512 netdev_features_t features); 2513 struct sk_buff *(*gro_receive)(struct list_head *head, 2514 struct sk_buff *skb); 2515 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2516 }; 2517 2518 struct packet_offload { 2519 __be16 type; /* This is really htons(ether_type). */ 2520 u16 priority; 2521 struct offload_callbacks callbacks; 2522 struct list_head list; 2523 }; 2524 2525 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2526 struct pcpu_sw_netstats { 2527 u64 rx_packets; 2528 u64 rx_bytes; 2529 u64 tx_packets; 2530 u64 tx_bytes; 2531 struct u64_stats_sync syncp; 2532 } __aligned(4 * sizeof(u64)); 2533 2534 struct pcpu_lstats { 2535 u64_stats_t packets; 2536 u64_stats_t bytes; 2537 struct u64_stats_sync syncp; 2538 } __aligned(2 * sizeof(u64)); 2539 2540 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2541 2542 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2543 { 2544 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2545 2546 u64_stats_update_begin(&tstats->syncp); 2547 tstats->rx_bytes += len; 2548 tstats->rx_packets++; 2549 u64_stats_update_end(&tstats->syncp); 2550 } 2551 2552 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2553 unsigned int packets, 2554 unsigned int len) 2555 { 2556 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2557 2558 u64_stats_update_begin(&tstats->syncp); 2559 tstats->tx_bytes += len; 2560 tstats->tx_packets += packets; 2561 u64_stats_update_end(&tstats->syncp); 2562 } 2563 2564 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2565 { 2566 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2567 2568 u64_stats_update_begin(&lstats->syncp); 2569 u64_stats_add(&lstats->bytes, len); 2570 u64_stats_inc(&lstats->packets); 2571 u64_stats_update_end(&lstats->syncp); 2572 } 2573 2574 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2575 ({ \ 2576 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2577 if (pcpu_stats) { \ 2578 int __cpu; \ 2579 for_each_possible_cpu(__cpu) { \ 2580 typeof(type) *stat; \ 2581 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2582 u64_stats_init(&stat->syncp); \ 2583 } \ 2584 } \ 2585 pcpu_stats; \ 2586 }) 2587 2588 #define netdev_alloc_pcpu_stats(type) \ 2589 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2590 2591 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2592 ({ \ 2593 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2594 if (pcpu_stats) { \ 2595 int __cpu; \ 2596 for_each_possible_cpu(__cpu) { \ 2597 typeof(type) *stat; \ 2598 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2599 u64_stats_init(&stat->syncp); \ 2600 } \ 2601 } \ 2602 pcpu_stats; \ 2603 }) 2604 2605 enum netdev_lag_tx_type { 2606 NETDEV_LAG_TX_TYPE_UNKNOWN, 2607 NETDEV_LAG_TX_TYPE_RANDOM, 2608 NETDEV_LAG_TX_TYPE_BROADCAST, 2609 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2610 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2611 NETDEV_LAG_TX_TYPE_HASH, 2612 }; 2613 2614 enum netdev_lag_hash { 2615 NETDEV_LAG_HASH_NONE, 2616 NETDEV_LAG_HASH_L2, 2617 NETDEV_LAG_HASH_L34, 2618 NETDEV_LAG_HASH_L23, 2619 NETDEV_LAG_HASH_E23, 2620 NETDEV_LAG_HASH_E34, 2621 NETDEV_LAG_HASH_VLAN_SRCMAC, 2622 NETDEV_LAG_HASH_UNKNOWN, 2623 }; 2624 2625 struct netdev_lag_upper_info { 2626 enum netdev_lag_tx_type tx_type; 2627 enum netdev_lag_hash hash_type; 2628 }; 2629 2630 struct netdev_lag_lower_state_info { 2631 u8 link_up : 1, 2632 tx_enabled : 1; 2633 }; 2634 2635 #include <linux/notifier.h> 2636 2637 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2638 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2639 * adding new types. 2640 */ 2641 enum netdev_cmd { 2642 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2643 NETDEV_DOWN, 2644 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2645 detected a hardware crash and restarted 2646 - we can use this eg to kick tcp sessions 2647 once done */ 2648 NETDEV_CHANGE, /* Notify device state change */ 2649 NETDEV_REGISTER, 2650 NETDEV_UNREGISTER, 2651 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2652 NETDEV_CHANGEADDR, /* notify after the address change */ 2653 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2654 NETDEV_GOING_DOWN, 2655 NETDEV_CHANGENAME, 2656 NETDEV_FEAT_CHANGE, 2657 NETDEV_BONDING_FAILOVER, 2658 NETDEV_PRE_UP, 2659 NETDEV_PRE_TYPE_CHANGE, 2660 NETDEV_POST_TYPE_CHANGE, 2661 NETDEV_POST_INIT, 2662 NETDEV_RELEASE, 2663 NETDEV_NOTIFY_PEERS, 2664 NETDEV_JOIN, 2665 NETDEV_CHANGEUPPER, 2666 NETDEV_RESEND_IGMP, 2667 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2668 NETDEV_CHANGEINFODATA, 2669 NETDEV_BONDING_INFO, 2670 NETDEV_PRECHANGEUPPER, 2671 NETDEV_CHANGELOWERSTATE, 2672 NETDEV_UDP_TUNNEL_PUSH_INFO, 2673 NETDEV_UDP_TUNNEL_DROP_INFO, 2674 NETDEV_CHANGE_TX_QUEUE_LEN, 2675 NETDEV_CVLAN_FILTER_PUSH_INFO, 2676 NETDEV_CVLAN_FILTER_DROP_INFO, 2677 NETDEV_SVLAN_FILTER_PUSH_INFO, 2678 NETDEV_SVLAN_FILTER_DROP_INFO, 2679 }; 2680 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2681 2682 int register_netdevice_notifier(struct notifier_block *nb); 2683 int unregister_netdevice_notifier(struct notifier_block *nb); 2684 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2685 int unregister_netdevice_notifier_net(struct net *net, 2686 struct notifier_block *nb); 2687 int register_netdevice_notifier_dev_net(struct net_device *dev, 2688 struct notifier_block *nb, 2689 struct netdev_net_notifier *nn); 2690 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2691 struct notifier_block *nb, 2692 struct netdev_net_notifier *nn); 2693 2694 struct netdev_notifier_info { 2695 struct net_device *dev; 2696 struct netlink_ext_ack *extack; 2697 }; 2698 2699 struct netdev_notifier_info_ext { 2700 struct netdev_notifier_info info; /* must be first */ 2701 union { 2702 u32 mtu; 2703 } ext; 2704 }; 2705 2706 struct netdev_notifier_change_info { 2707 struct netdev_notifier_info info; /* must be first */ 2708 unsigned int flags_changed; 2709 }; 2710 2711 struct netdev_notifier_changeupper_info { 2712 struct netdev_notifier_info info; /* must be first */ 2713 struct net_device *upper_dev; /* new upper dev */ 2714 bool master; /* is upper dev master */ 2715 bool linking; /* is the notification for link or unlink */ 2716 void *upper_info; /* upper dev info */ 2717 }; 2718 2719 struct netdev_notifier_changelowerstate_info { 2720 struct netdev_notifier_info info; /* must be first */ 2721 void *lower_state_info; /* is lower dev state */ 2722 }; 2723 2724 struct netdev_notifier_pre_changeaddr_info { 2725 struct netdev_notifier_info info; /* must be first */ 2726 const unsigned char *dev_addr; 2727 }; 2728 2729 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2730 struct net_device *dev) 2731 { 2732 info->dev = dev; 2733 info->extack = NULL; 2734 } 2735 2736 static inline struct net_device * 2737 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2738 { 2739 return info->dev; 2740 } 2741 2742 static inline struct netlink_ext_ack * 2743 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2744 { 2745 return info->extack; 2746 } 2747 2748 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2749 2750 2751 extern rwlock_t dev_base_lock; /* Device list lock */ 2752 2753 #define for_each_netdev(net, d) \ 2754 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2755 #define for_each_netdev_reverse(net, d) \ 2756 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2757 #define for_each_netdev_rcu(net, d) \ 2758 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2759 #define for_each_netdev_safe(net, d, n) \ 2760 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2761 #define for_each_netdev_continue(net, d) \ 2762 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2763 #define for_each_netdev_continue_reverse(net, d) \ 2764 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2765 dev_list) 2766 #define for_each_netdev_continue_rcu(net, d) \ 2767 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2768 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2769 for_each_netdev_rcu(&init_net, slave) \ 2770 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2771 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2772 2773 static inline struct net_device *next_net_device(struct net_device *dev) 2774 { 2775 struct list_head *lh; 2776 struct net *net; 2777 2778 net = dev_net(dev); 2779 lh = dev->dev_list.next; 2780 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2781 } 2782 2783 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2784 { 2785 struct list_head *lh; 2786 struct net *net; 2787 2788 net = dev_net(dev); 2789 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2790 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2791 } 2792 2793 static inline struct net_device *first_net_device(struct net *net) 2794 { 2795 return list_empty(&net->dev_base_head) ? NULL : 2796 net_device_entry(net->dev_base_head.next); 2797 } 2798 2799 static inline struct net_device *first_net_device_rcu(struct net *net) 2800 { 2801 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2802 2803 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2804 } 2805 2806 int netdev_boot_setup_check(struct net_device *dev); 2807 unsigned long netdev_boot_base(const char *prefix, int unit); 2808 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2809 const char *hwaddr); 2810 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2811 void dev_add_pack(struct packet_type *pt); 2812 void dev_remove_pack(struct packet_type *pt); 2813 void __dev_remove_pack(struct packet_type *pt); 2814 void dev_add_offload(struct packet_offload *po); 2815 void dev_remove_offload(struct packet_offload *po); 2816 2817 int dev_get_iflink(const struct net_device *dev); 2818 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2819 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2820 unsigned short mask); 2821 struct net_device *dev_get_by_name(struct net *net, const char *name); 2822 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2823 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2824 int dev_alloc_name(struct net_device *dev, const char *name); 2825 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2826 void dev_close(struct net_device *dev); 2827 void dev_close_many(struct list_head *head, bool unlink); 2828 void dev_disable_lro(struct net_device *dev); 2829 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2830 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2831 struct net_device *sb_dev); 2832 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2833 struct net_device *sb_dev); 2834 2835 int dev_queue_xmit(struct sk_buff *skb); 2836 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2837 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2838 2839 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2840 { 2841 int ret; 2842 2843 ret = __dev_direct_xmit(skb, queue_id); 2844 if (!dev_xmit_complete(ret)) 2845 kfree_skb(skb); 2846 return ret; 2847 } 2848 2849 int register_netdevice(struct net_device *dev); 2850 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2851 void unregister_netdevice_many(struct list_head *head); 2852 static inline void unregister_netdevice(struct net_device *dev) 2853 { 2854 unregister_netdevice_queue(dev, NULL); 2855 } 2856 2857 int netdev_refcnt_read(const struct net_device *dev); 2858 void free_netdev(struct net_device *dev); 2859 void netdev_freemem(struct net_device *dev); 2860 int init_dummy_netdev(struct net_device *dev); 2861 2862 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2863 struct sk_buff *skb, 2864 bool all_slaves); 2865 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2866 struct sock *sk); 2867 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2868 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2869 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2870 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2871 int netdev_get_name(struct net *net, char *name, int ifindex); 2872 int dev_restart(struct net_device *dev); 2873 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2874 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2875 2876 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2877 { 2878 return NAPI_GRO_CB(skb)->data_offset; 2879 } 2880 2881 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2882 { 2883 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2884 } 2885 2886 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2887 { 2888 NAPI_GRO_CB(skb)->data_offset += len; 2889 } 2890 2891 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2892 unsigned int offset) 2893 { 2894 return NAPI_GRO_CB(skb)->frag0 + offset; 2895 } 2896 2897 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2898 { 2899 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2900 } 2901 2902 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2903 { 2904 NAPI_GRO_CB(skb)->frag0 = NULL; 2905 NAPI_GRO_CB(skb)->frag0_len = 0; 2906 } 2907 2908 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2909 unsigned int offset) 2910 { 2911 if (!pskb_may_pull(skb, hlen)) 2912 return NULL; 2913 2914 skb_gro_frag0_invalidate(skb); 2915 return skb->data + offset; 2916 } 2917 2918 static inline void *skb_gro_network_header(struct sk_buff *skb) 2919 { 2920 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2921 skb_network_offset(skb); 2922 } 2923 2924 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2925 const void *start, unsigned int len) 2926 { 2927 if (NAPI_GRO_CB(skb)->csum_valid) 2928 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2929 csum_partial(start, len, 0)); 2930 } 2931 2932 /* GRO checksum functions. These are logical equivalents of the normal 2933 * checksum functions (in skbuff.h) except that they operate on the GRO 2934 * offsets and fields in sk_buff. 2935 */ 2936 2937 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2938 2939 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2940 { 2941 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2942 } 2943 2944 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2945 bool zero_okay, 2946 __sum16 check) 2947 { 2948 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2949 skb_checksum_start_offset(skb) < 2950 skb_gro_offset(skb)) && 2951 !skb_at_gro_remcsum_start(skb) && 2952 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2953 (!zero_okay || check)); 2954 } 2955 2956 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2957 __wsum psum) 2958 { 2959 if (NAPI_GRO_CB(skb)->csum_valid && 2960 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2961 return 0; 2962 2963 NAPI_GRO_CB(skb)->csum = psum; 2964 2965 return __skb_gro_checksum_complete(skb); 2966 } 2967 2968 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2969 { 2970 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2971 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2972 NAPI_GRO_CB(skb)->csum_cnt--; 2973 } else { 2974 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2975 * verified a new top level checksum or an encapsulated one 2976 * during GRO. This saves work if we fallback to normal path. 2977 */ 2978 __skb_incr_checksum_unnecessary(skb); 2979 } 2980 } 2981 2982 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2983 compute_pseudo) \ 2984 ({ \ 2985 __sum16 __ret = 0; \ 2986 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2987 __ret = __skb_gro_checksum_validate_complete(skb, \ 2988 compute_pseudo(skb, proto)); \ 2989 if (!__ret) \ 2990 skb_gro_incr_csum_unnecessary(skb); \ 2991 __ret; \ 2992 }) 2993 2994 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2995 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2996 2997 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2998 compute_pseudo) \ 2999 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 3000 3001 #define skb_gro_checksum_simple_validate(skb) \ 3002 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 3003 3004 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3005 { 3006 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3007 !NAPI_GRO_CB(skb)->csum_valid); 3008 } 3009 3010 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3011 __wsum pseudo) 3012 { 3013 NAPI_GRO_CB(skb)->csum = ~pseudo; 3014 NAPI_GRO_CB(skb)->csum_valid = 1; 3015 } 3016 3017 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3018 do { \ 3019 if (__skb_gro_checksum_convert_check(skb)) \ 3020 __skb_gro_checksum_convert(skb, \ 3021 compute_pseudo(skb, proto)); \ 3022 } while (0) 3023 3024 struct gro_remcsum { 3025 int offset; 3026 __wsum delta; 3027 }; 3028 3029 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3030 { 3031 grc->offset = 0; 3032 grc->delta = 0; 3033 } 3034 3035 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3036 unsigned int off, size_t hdrlen, 3037 int start, int offset, 3038 struct gro_remcsum *grc, 3039 bool nopartial) 3040 { 3041 __wsum delta; 3042 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3043 3044 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3045 3046 if (!nopartial) { 3047 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3048 return ptr; 3049 } 3050 3051 ptr = skb_gro_header_fast(skb, off); 3052 if (skb_gro_header_hard(skb, off + plen)) { 3053 ptr = skb_gro_header_slow(skb, off + plen, off); 3054 if (!ptr) 3055 return NULL; 3056 } 3057 3058 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3059 start, offset); 3060 3061 /* Adjust skb->csum since we changed the packet */ 3062 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3063 3064 grc->offset = off + hdrlen + offset; 3065 grc->delta = delta; 3066 3067 return ptr; 3068 } 3069 3070 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3071 struct gro_remcsum *grc) 3072 { 3073 void *ptr; 3074 size_t plen = grc->offset + sizeof(u16); 3075 3076 if (!grc->delta) 3077 return; 3078 3079 ptr = skb_gro_header_fast(skb, grc->offset); 3080 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3081 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3082 if (!ptr) 3083 return; 3084 } 3085 3086 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3087 } 3088 3089 #ifdef CONFIG_XFRM_OFFLOAD 3090 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3091 { 3092 if (PTR_ERR(pp) != -EINPROGRESS) 3093 NAPI_GRO_CB(skb)->flush |= flush; 3094 } 3095 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3096 struct sk_buff *pp, 3097 int flush, 3098 struct gro_remcsum *grc) 3099 { 3100 if (PTR_ERR(pp) != -EINPROGRESS) { 3101 NAPI_GRO_CB(skb)->flush |= flush; 3102 skb_gro_remcsum_cleanup(skb, grc); 3103 skb->remcsum_offload = 0; 3104 } 3105 } 3106 #else 3107 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3108 { 3109 NAPI_GRO_CB(skb)->flush |= flush; 3110 } 3111 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3112 struct sk_buff *pp, 3113 int flush, 3114 struct gro_remcsum *grc) 3115 { 3116 NAPI_GRO_CB(skb)->flush |= flush; 3117 skb_gro_remcsum_cleanup(skb, grc); 3118 skb->remcsum_offload = 0; 3119 } 3120 #endif 3121 3122 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3123 unsigned short type, 3124 const void *daddr, const void *saddr, 3125 unsigned int len) 3126 { 3127 if (!dev->header_ops || !dev->header_ops->create) 3128 return 0; 3129 3130 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3131 } 3132 3133 static inline int dev_parse_header(const struct sk_buff *skb, 3134 unsigned char *haddr) 3135 { 3136 const struct net_device *dev = skb->dev; 3137 3138 if (!dev->header_ops || !dev->header_ops->parse) 3139 return 0; 3140 return dev->header_ops->parse(skb, haddr); 3141 } 3142 3143 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3144 { 3145 const struct net_device *dev = skb->dev; 3146 3147 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3148 return 0; 3149 return dev->header_ops->parse_protocol(skb); 3150 } 3151 3152 /* ll_header must have at least hard_header_len allocated */ 3153 static inline bool dev_validate_header(const struct net_device *dev, 3154 char *ll_header, int len) 3155 { 3156 if (likely(len >= dev->hard_header_len)) 3157 return true; 3158 if (len < dev->min_header_len) 3159 return false; 3160 3161 if (capable(CAP_SYS_RAWIO)) { 3162 memset(ll_header + len, 0, dev->hard_header_len - len); 3163 return true; 3164 } 3165 3166 if (dev->header_ops && dev->header_ops->validate) 3167 return dev->header_ops->validate(ll_header, len); 3168 3169 return false; 3170 } 3171 3172 static inline bool dev_has_header(const struct net_device *dev) 3173 { 3174 return dev->header_ops && dev->header_ops->create; 3175 } 3176 3177 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3178 int len, int size); 3179 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3180 static inline int unregister_gifconf(unsigned int family) 3181 { 3182 return register_gifconf(family, NULL); 3183 } 3184 3185 #ifdef CONFIG_NET_FLOW_LIMIT 3186 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3187 struct sd_flow_limit { 3188 u64 count; 3189 unsigned int num_buckets; 3190 unsigned int history_head; 3191 u16 history[FLOW_LIMIT_HISTORY]; 3192 u8 buckets[]; 3193 }; 3194 3195 extern int netdev_flow_limit_table_len; 3196 #endif /* CONFIG_NET_FLOW_LIMIT */ 3197 3198 /* 3199 * Incoming packets are placed on per-CPU queues 3200 */ 3201 struct softnet_data { 3202 struct list_head poll_list; 3203 struct sk_buff_head process_queue; 3204 3205 /* stats */ 3206 unsigned int processed; 3207 unsigned int time_squeeze; 3208 unsigned int received_rps; 3209 #ifdef CONFIG_RPS 3210 struct softnet_data *rps_ipi_list; 3211 #endif 3212 #ifdef CONFIG_NET_FLOW_LIMIT 3213 struct sd_flow_limit __rcu *flow_limit; 3214 #endif 3215 struct Qdisc *output_queue; 3216 struct Qdisc **output_queue_tailp; 3217 struct sk_buff *completion_queue; 3218 #ifdef CONFIG_XFRM_OFFLOAD 3219 struct sk_buff_head xfrm_backlog; 3220 #endif 3221 /* written and read only by owning cpu: */ 3222 struct { 3223 u16 recursion; 3224 u8 more; 3225 } xmit; 3226 #ifdef CONFIG_RPS 3227 /* input_queue_head should be written by cpu owning this struct, 3228 * and only read by other cpus. Worth using a cache line. 3229 */ 3230 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3231 3232 /* Elements below can be accessed between CPUs for RPS/RFS */ 3233 call_single_data_t csd ____cacheline_aligned_in_smp; 3234 struct softnet_data *rps_ipi_next; 3235 unsigned int cpu; 3236 unsigned int input_queue_tail; 3237 #endif 3238 unsigned int dropped; 3239 struct sk_buff_head input_pkt_queue; 3240 struct napi_struct backlog; 3241 3242 }; 3243 3244 static inline void input_queue_head_incr(struct softnet_data *sd) 3245 { 3246 #ifdef CONFIG_RPS 3247 sd->input_queue_head++; 3248 #endif 3249 } 3250 3251 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3252 unsigned int *qtail) 3253 { 3254 #ifdef CONFIG_RPS 3255 *qtail = ++sd->input_queue_tail; 3256 #endif 3257 } 3258 3259 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3260 3261 static inline int dev_recursion_level(void) 3262 { 3263 return this_cpu_read(softnet_data.xmit.recursion); 3264 } 3265 3266 #define XMIT_RECURSION_LIMIT 8 3267 static inline bool dev_xmit_recursion(void) 3268 { 3269 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3270 XMIT_RECURSION_LIMIT); 3271 } 3272 3273 static inline void dev_xmit_recursion_inc(void) 3274 { 3275 __this_cpu_inc(softnet_data.xmit.recursion); 3276 } 3277 3278 static inline void dev_xmit_recursion_dec(void) 3279 { 3280 __this_cpu_dec(softnet_data.xmit.recursion); 3281 } 3282 3283 void __netif_schedule(struct Qdisc *q); 3284 void netif_schedule_queue(struct netdev_queue *txq); 3285 3286 static inline void netif_tx_schedule_all(struct net_device *dev) 3287 { 3288 unsigned int i; 3289 3290 for (i = 0; i < dev->num_tx_queues; i++) 3291 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3292 } 3293 3294 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3295 { 3296 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3297 } 3298 3299 /** 3300 * netif_start_queue - allow transmit 3301 * @dev: network device 3302 * 3303 * Allow upper layers to call the device hard_start_xmit routine. 3304 */ 3305 static inline void netif_start_queue(struct net_device *dev) 3306 { 3307 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3308 } 3309 3310 static inline void netif_tx_start_all_queues(struct net_device *dev) 3311 { 3312 unsigned int i; 3313 3314 for (i = 0; i < dev->num_tx_queues; i++) { 3315 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3316 netif_tx_start_queue(txq); 3317 } 3318 } 3319 3320 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3321 3322 /** 3323 * netif_wake_queue - restart transmit 3324 * @dev: network device 3325 * 3326 * Allow upper layers to call the device hard_start_xmit routine. 3327 * Used for flow control when transmit resources are available. 3328 */ 3329 static inline void netif_wake_queue(struct net_device *dev) 3330 { 3331 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3332 } 3333 3334 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3335 { 3336 unsigned int i; 3337 3338 for (i = 0; i < dev->num_tx_queues; i++) { 3339 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3340 netif_tx_wake_queue(txq); 3341 } 3342 } 3343 3344 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3345 { 3346 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3347 } 3348 3349 /** 3350 * netif_stop_queue - stop transmitted packets 3351 * @dev: network device 3352 * 3353 * Stop upper layers calling the device hard_start_xmit routine. 3354 * Used for flow control when transmit resources are unavailable. 3355 */ 3356 static inline void netif_stop_queue(struct net_device *dev) 3357 { 3358 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3359 } 3360 3361 void netif_tx_stop_all_queues(struct net_device *dev); 3362 3363 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3364 { 3365 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3366 } 3367 3368 /** 3369 * netif_queue_stopped - test if transmit queue is flowblocked 3370 * @dev: network device 3371 * 3372 * Test if transmit queue on device is currently unable to send. 3373 */ 3374 static inline bool netif_queue_stopped(const struct net_device *dev) 3375 { 3376 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3377 } 3378 3379 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3380 { 3381 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3382 } 3383 3384 static inline bool 3385 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3386 { 3387 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3388 } 3389 3390 static inline bool 3391 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3392 { 3393 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3394 } 3395 3396 /** 3397 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3398 * @dev_queue: pointer to transmit queue 3399 * 3400 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3401 * to give appropriate hint to the CPU. 3402 */ 3403 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3404 { 3405 #ifdef CONFIG_BQL 3406 prefetchw(&dev_queue->dql.num_queued); 3407 #endif 3408 } 3409 3410 /** 3411 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3412 * @dev_queue: pointer to transmit queue 3413 * 3414 * BQL enabled drivers might use this helper in their TX completion path, 3415 * to give appropriate hint to the CPU. 3416 */ 3417 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3418 { 3419 #ifdef CONFIG_BQL 3420 prefetchw(&dev_queue->dql.limit); 3421 #endif 3422 } 3423 3424 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3425 unsigned int bytes) 3426 { 3427 #ifdef CONFIG_BQL 3428 dql_queued(&dev_queue->dql, bytes); 3429 3430 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3431 return; 3432 3433 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3434 3435 /* 3436 * The XOFF flag must be set before checking the dql_avail below, 3437 * because in netdev_tx_completed_queue we update the dql_completed 3438 * before checking the XOFF flag. 3439 */ 3440 smp_mb(); 3441 3442 /* check again in case another CPU has just made room avail */ 3443 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3444 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3445 #endif 3446 } 3447 3448 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3449 * that they should not test BQL status themselves. 3450 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3451 * skb of a batch. 3452 * Returns true if the doorbell must be used to kick the NIC. 3453 */ 3454 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3455 unsigned int bytes, 3456 bool xmit_more) 3457 { 3458 if (xmit_more) { 3459 #ifdef CONFIG_BQL 3460 dql_queued(&dev_queue->dql, bytes); 3461 #endif 3462 return netif_tx_queue_stopped(dev_queue); 3463 } 3464 netdev_tx_sent_queue(dev_queue, bytes); 3465 return true; 3466 } 3467 3468 /** 3469 * netdev_sent_queue - report the number of bytes queued to hardware 3470 * @dev: network device 3471 * @bytes: number of bytes queued to the hardware device queue 3472 * 3473 * Report the number of bytes queued for sending/completion to the network 3474 * device hardware queue. @bytes should be a good approximation and should 3475 * exactly match netdev_completed_queue() @bytes 3476 */ 3477 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3478 { 3479 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3480 } 3481 3482 static inline bool __netdev_sent_queue(struct net_device *dev, 3483 unsigned int bytes, 3484 bool xmit_more) 3485 { 3486 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3487 xmit_more); 3488 } 3489 3490 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3491 unsigned int pkts, unsigned int bytes) 3492 { 3493 #ifdef CONFIG_BQL 3494 if (unlikely(!bytes)) 3495 return; 3496 3497 dql_completed(&dev_queue->dql, bytes); 3498 3499 /* 3500 * Without the memory barrier there is a small possiblity that 3501 * netdev_tx_sent_queue will miss the update and cause the queue to 3502 * be stopped forever 3503 */ 3504 smp_mb(); 3505 3506 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3507 return; 3508 3509 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3510 netif_schedule_queue(dev_queue); 3511 #endif 3512 } 3513 3514 /** 3515 * netdev_completed_queue - report bytes and packets completed by device 3516 * @dev: network device 3517 * @pkts: actual number of packets sent over the medium 3518 * @bytes: actual number of bytes sent over the medium 3519 * 3520 * Report the number of bytes and packets transmitted by the network device 3521 * hardware queue over the physical medium, @bytes must exactly match the 3522 * @bytes amount passed to netdev_sent_queue() 3523 */ 3524 static inline void netdev_completed_queue(struct net_device *dev, 3525 unsigned int pkts, unsigned int bytes) 3526 { 3527 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3528 } 3529 3530 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3531 { 3532 #ifdef CONFIG_BQL 3533 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3534 dql_reset(&q->dql); 3535 #endif 3536 } 3537 3538 /** 3539 * netdev_reset_queue - reset the packets and bytes count of a network device 3540 * @dev_queue: network device 3541 * 3542 * Reset the bytes and packet count of a network device and clear the 3543 * software flow control OFF bit for this network device 3544 */ 3545 static inline void netdev_reset_queue(struct net_device *dev_queue) 3546 { 3547 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3548 } 3549 3550 /** 3551 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3552 * @dev: network device 3553 * @queue_index: given tx queue index 3554 * 3555 * Returns 0 if given tx queue index >= number of device tx queues, 3556 * otherwise returns the originally passed tx queue index. 3557 */ 3558 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3559 { 3560 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3561 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3562 dev->name, queue_index, 3563 dev->real_num_tx_queues); 3564 return 0; 3565 } 3566 3567 return queue_index; 3568 } 3569 3570 /** 3571 * netif_running - test if up 3572 * @dev: network device 3573 * 3574 * Test if the device has been brought up. 3575 */ 3576 static inline bool netif_running(const struct net_device *dev) 3577 { 3578 return test_bit(__LINK_STATE_START, &dev->state); 3579 } 3580 3581 /* 3582 * Routines to manage the subqueues on a device. We only need start, 3583 * stop, and a check if it's stopped. All other device management is 3584 * done at the overall netdevice level. 3585 * Also test the device if we're multiqueue. 3586 */ 3587 3588 /** 3589 * netif_start_subqueue - allow sending packets on subqueue 3590 * @dev: network device 3591 * @queue_index: sub queue index 3592 * 3593 * Start individual transmit queue of a device with multiple transmit queues. 3594 */ 3595 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3596 { 3597 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3598 3599 netif_tx_start_queue(txq); 3600 } 3601 3602 /** 3603 * netif_stop_subqueue - stop sending packets on subqueue 3604 * @dev: network device 3605 * @queue_index: sub queue index 3606 * 3607 * Stop individual transmit queue of a device with multiple transmit queues. 3608 */ 3609 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3610 { 3611 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3612 netif_tx_stop_queue(txq); 3613 } 3614 3615 /** 3616 * __netif_subqueue_stopped - test status of subqueue 3617 * @dev: network device 3618 * @queue_index: sub queue index 3619 * 3620 * Check individual transmit queue of a device with multiple transmit queues. 3621 */ 3622 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3623 u16 queue_index) 3624 { 3625 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3626 3627 return netif_tx_queue_stopped(txq); 3628 } 3629 3630 /** 3631 * netif_subqueue_stopped - test status of subqueue 3632 * @dev: network device 3633 * @skb: sub queue buffer pointer 3634 * 3635 * Check individual transmit queue of a device with multiple transmit queues. 3636 */ 3637 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3638 struct sk_buff *skb) 3639 { 3640 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3641 } 3642 3643 /** 3644 * netif_wake_subqueue - allow sending packets on subqueue 3645 * @dev: network device 3646 * @queue_index: sub queue index 3647 * 3648 * Resume individual transmit queue of a device with multiple transmit queues. 3649 */ 3650 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3651 { 3652 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3653 3654 netif_tx_wake_queue(txq); 3655 } 3656 3657 #ifdef CONFIG_XPS 3658 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3659 u16 index); 3660 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3661 u16 index, bool is_rxqs_map); 3662 3663 /** 3664 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3665 * @j: CPU/Rx queue index 3666 * @mask: bitmask of all cpus/rx queues 3667 * @nr_bits: number of bits in the bitmask 3668 * 3669 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3670 */ 3671 static inline bool netif_attr_test_mask(unsigned long j, 3672 const unsigned long *mask, 3673 unsigned int nr_bits) 3674 { 3675 cpu_max_bits_warn(j, nr_bits); 3676 return test_bit(j, mask); 3677 } 3678 3679 /** 3680 * netif_attr_test_online - Test for online CPU/Rx queue 3681 * @j: CPU/Rx queue index 3682 * @online_mask: bitmask for CPUs/Rx queues that are online 3683 * @nr_bits: number of bits in the bitmask 3684 * 3685 * Returns true if a CPU/Rx queue is online. 3686 */ 3687 static inline bool netif_attr_test_online(unsigned long j, 3688 const unsigned long *online_mask, 3689 unsigned int nr_bits) 3690 { 3691 cpu_max_bits_warn(j, nr_bits); 3692 3693 if (online_mask) 3694 return test_bit(j, online_mask); 3695 3696 return (j < nr_bits); 3697 } 3698 3699 /** 3700 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3701 * @n: CPU/Rx queue index 3702 * @srcp: the cpumask/Rx queue mask pointer 3703 * @nr_bits: number of bits in the bitmask 3704 * 3705 * Returns >= nr_bits if no further CPUs/Rx queues set. 3706 */ 3707 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3708 unsigned int nr_bits) 3709 { 3710 /* -1 is a legal arg here. */ 3711 if (n != -1) 3712 cpu_max_bits_warn(n, nr_bits); 3713 3714 if (srcp) 3715 return find_next_bit(srcp, nr_bits, n + 1); 3716 3717 return n + 1; 3718 } 3719 3720 /** 3721 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3722 * @n: CPU/Rx queue index 3723 * @src1p: the first CPUs/Rx queues mask pointer 3724 * @src2p: the second CPUs/Rx queues mask pointer 3725 * @nr_bits: number of bits in the bitmask 3726 * 3727 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3728 */ 3729 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3730 const unsigned long *src2p, 3731 unsigned int nr_bits) 3732 { 3733 /* -1 is a legal arg here. */ 3734 if (n != -1) 3735 cpu_max_bits_warn(n, nr_bits); 3736 3737 if (src1p && src2p) 3738 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3739 else if (src1p) 3740 return find_next_bit(src1p, nr_bits, n + 1); 3741 else if (src2p) 3742 return find_next_bit(src2p, nr_bits, n + 1); 3743 3744 return n + 1; 3745 } 3746 #else 3747 static inline int netif_set_xps_queue(struct net_device *dev, 3748 const struct cpumask *mask, 3749 u16 index) 3750 { 3751 return 0; 3752 } 3753 3754 static inline int __netif_set_xps_queue(struct net_device *dev, 3755 const unsigned long *mask, 3756 u16 index, bool is_rxqs_map) 3757 { 3758 return 0; 3759 } 3760 #endif 3761 3762 /** 3763 * netif_is_multiqueue - test if device has multiple transmit queues 3764 * @dev: network device 3765 * 3766 * Check if device has multiple transmit queues 3767 */ 3768 static inline bool netif_is_multiqueue(const struct net_device *dev) 3769 { 3770 return dev->num_tx_queues > 1; 3771 } 3772 3773 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3774 3775 #ifdef CONFIG_SYSFS 3776 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3777 #else 3778 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3779 unsigned int rxqs) 3780 { 3781 dev->real_num_rx_queues = rxqs; 3782 return 0; 3783 } 3784 #endif 3785 3786 static inline struct netdev_rx_queue * 3787 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3788 { 3789 return dev->_rx + rxq; 3790 } 3791 3792 #ifdef CONFIG_SYSFS 3793 static inline unsigned int get_netdev_rx_queue_index( 3794 struct netdev_rx_queue *queue) 3795 { 3796 struct net_device *dev = queue->dev; 3797 int index = queue - dev->_rx; 3798 3799 BUG_ON(index >= dev->num_rx_queues); 3800 return index; 3801 } 3802 #endif 3803 3804 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3805 int netif_get_num_default_rss_queues(void); 3806 3807 enum skb_free_reason { 3808 SKB_REASON_CONSUMED, 3809 SKB_REASON_DROPPED, 3810 }; 3811 3812 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3813 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3814 3815 /* 3816 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3817 * interrupt context or with hardware interrupts being disabled. 3818 * (in_irq() || irqs_disabled()) 3819 * 3820 * We provide four helpers that can be used in following contexts : 3821 * 3822 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3823 * replacing kfree_skb(skb) 3824 * 3825 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3826 * Typically used in place of consume_skb(skb) in TX completion path 3827 * 3828 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3829 * replacing kfree_skb(skb) 3830 * 3831 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3832 * and consumed a packet. Used in place of consume_skb(skb) 3833 */ 3834 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3835 { 3836 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3837 } 3838 3839 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3840 { 3841 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3842 } 3843 3844 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3845 { 3846 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3847 } 3848 3849 static inline void dev_consume_skb_any(struct sk_buff *skb) 3850 { 3851 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3852 } 3853 3854 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3855 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3856 int netif_rx(struct sk_buff *skb); 3857 int netif_rx_ni(struct sk_buff *skb); 3858 int netif_rx_any_context(struct sk_buff *skb); 3859 int netif_receive_skb(struct sk_buff *skb); 3860 int netif_receive_skb_core(struct sk_buff *skb); 3861 void netif_receive_skb_list(struct list_head *head); 3862 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3863 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3864 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3865 gro_result_t napi_gro_frags(struct napi_struct *napi); 3866 struct packet_offload *gro_find_receive_by_type(__be16 type); 3867 struct packet_offload *gro_find_complete_by_type(__be16 type); 3868 3869 static inline void napi_free_frags(struct napi_struct *napi) 3870 { 3871 kfree_skb(napi->skb); 3872 napi->skb = NULL; 3873 } 3874 3875 bool netdev_is_rx_handler_busy(struct net_device *dev); 3876 int netdev_rx_handler_register(struct net_device *dev, 3877 rx_handler_func_t *rx_handler, 3878 void *rx_handler_data); 3879 void netdev_rx_handler_unregister(struct net_device *dev); 3880 3881 bool dev_valid_name(const char *name); 3882 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3883 bool *need_copyout); 3884 int dev_ifconf(struct net *net, struct ifconf *, int); 3885 int dev_ethtool(struct net *net, struct ifreq *); 3886 unsigned int dev_get_flags(const struct net_device *); 3887 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3888 struct netlink_ext_ack *extack); 3889 int dev_change_flags(struct net_device *dev, unsigned int flags, 3890 struct netlink_ext_ack *extack); 3891 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3892 unsigned int gchanges); 3893 int dev_change_name(struct net_device *, const char *); 3894 int dev_set_alias(struct net_device *, const char *, size_t); 3895 int dev_get_alias(const struct net_device *, char *, size_t); 3896 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3897 int __dev_set_mtu(struct net_device *, int); 3898 int dev_validate_mtu(struct net_device *dev, int mtu, 3899 struct netlink_ext_ack *extack); 3900 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3901 struct netlink_ext_ack *extack); 3902 int dev_set_mtu(struct net_device *, int); 3903 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3904 void dev_set_group(struct net_device *, int); 3905 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3906 struct netlink_ext_ack *extack); 3907 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3908 struct netlink_ext_ack *extack); 3909 int dev_change_carrier(struct net_device *, bool new_carrier); 3910 int dev_get_phys_port_id(struct net_device *dev, 3911 struct netdev_phys_item_id *ppid); 3912 int dev_get_phys_port_name(struct net_device *dev, 3913 char *name, size_t len); 3914 int dev_get_port_parent_id(struct net_device *dev, 3915 struct netdev_phys_item_id *ppid, bool recurse); 3916 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3917 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3918 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3919 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3920 u32 value); 3921 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3922 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3923 struct netdev_queue *txq, int *ret); 3924 3925 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3926 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3927 int fd, int expected_fd, u32 flags); 3928 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3929 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3930 3931 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3932 3933 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3934 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3935 bool is_skb_forwardable(const struct net_device *dev, 3936 const struct sk_buff *skb); 3937 3938 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3939 struct sk_buff *skb) 3940 { 3941 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3942 unlikely(!is_skb_forwardable(dev, skb))) { 3943 atomic_long_inc(&dev->rx_dropped); 3944 kfree_skb(skb); 3945 return NET_RX_DROP; 3946 } 3947 3948 skb_scrub_packet(skb, true); 3949 skb->priority = 0; 3950 return 0; 3951 } 3952 3953 bool dev_nit_active(struct net_device *dev); 3954 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3955 3956 extern int netdev_budget; 3957 extern unsigned int netdev_budget_usecs; 3958 3959 /* Called by rtnetlink.c:rtnl_unlock() */ 3960 void netdev_run_todo(void); 3961 3962 /** 3963 * dev_put - release reference to device 3964 * @dev: network device 3965 * 3966 * Release reference to device to allow it to be freed. 3967 */ 3968 static inline void dev_put(struct net_device *dev) 3969 { 3970 this_cpu_dec(*dev->pcpu_refcnt); 3971 } 3972 3973 /** 3974 * dev_hold - get reference to device 3975 * @dev: network device 3976 * 3977 * Hold reference to device to keep it from being freed. 3978 */ 3979 static inline void dev_hold(struct net_device *dev) 3980 { 3981 this_cpu_inc(*dev->pcpu_refcnt); 3982 } 3983 3984 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3985 * and _off may be called from IRQ context, but it is caller 3986 * who is responsible for serialization of these calls. 3987 * 3988 * The name carrier is inappropriate, these functions should really be 3989 * called netif_lowerlayer_*() because they represent the state of any 3990 * kind of lower layer not just hardware media. 3991 */ 3992 3993 void linkwatch_init_dev(struct net_device *dev); 3994 void linkwatch_fire_event(struct net_device *dev); 3995 void linkwatch_forget_dev(struct net_device *dev); 3996 3997 /** 3998 * netif_carrier_ok - test if carrier present 3999 * @dev: network device 4000 * 4001 * Check if carrier is present on device 4002 */ 4003 static inline bool netif_carrier_ok(const struct net_device *dev) 4004 { 4005 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4006 } 4007 4008 unsigned long dev_trans_start(struct net_device *dev); 4009 4010 void __netdev_watchdog_up(struct net_device *dev); 4011 4012 void netif_carrier_on(struct net_device *dev); 4013 4014 void netif_carrier_off(struct net_device *dev); 4015 4016 /** 4017 * netif_dormant_on - mark device as dormant. 4018 * @dev: network device 4019 * 4020 * Mark device as dormant (as per RFC2863). 4021 * 4022 * The dormant state indicates that the relevant interface is not 4023 * actually in a condition to pass packets (i.e., it is not 'up') but is 4024 * in a "pending" state, waiting for some external event. For "on- 4025 * demand" interfaces, this new state identifies the situation where the 4026 * interface is waiting for events to place it in the up state. 4027 */ 4028 static inline void netif_dormant_on(struct net_device *dev) 4029 { 4030 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4031 linkwatch_fire_event(dev); 4032 } 4033 4034 /** 4035 * netif_dormant_off - set device as not dormant. 4036 * @dev: network device 4037 * 4038 * Device is not in dormant state. 4039 */ 4040 static inline void netif_dormant_off(struct net_device *dev) 4041 { 4042 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4043 linkwatch_fire_event(dev); 4044 } 4045 4046 /** 4047 * netif_dormant - test if device is dormant 4048 * @dev: network device 4049 * 4050 * Check if device is dormant. 4051 */ 4052 static inline bool netif_dormant(const struct net_device *dev) 4053 { 4054 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4055 } 4056 4057 4058 /** 4059 * netif_testing_on - mark device as under test. 4060 * @dev: network device 4061 * 4062 * Mark device as under test (as per RFC2863). 4063 * 4064 * The testing state indicates that some test(s) must be performed on 4065 * the interface. After completion, of the test, the interface state 4066 * will change to up, dormant, or down, as appropriate. 4067 */ 4068 static inline void netif_testing_on(struct net_device *dev) 4069 { 4070 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4071 linkwatch_fire_event(dev); 4072 } 4073 4074 /** 4075 * netif_testing_off - set device as not under test. 4076 * @dev: network device 4077 * 4078 * Device is not in testing state. 4079 */ 4080 static inline void netif_testing_off(struct net_device *dev) 4081 { 4082 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4083 linkwatch_fire_event(dev); 4084 } 4085 4086 /** 4087 * netif_testing - test if device is under test 4088 * @dev: network device 4089 * 4090 * Check if device is under test 4091 */ 4092 static inline bool netif_testing(const struct net_device *dev) 4093 { 4094 return test_bit(__LINK_STATE_TESTING, &dev->state); 4095 } 4096 4097 4098 /** 4099 * netif_oper_up - test if device is operational 4100 * @dev: network device 4101 * 4102 * Check if carrier is operational 4103 */ 4104 static inline bool netif_oper_up(const struct net_device *dev) 4105 { 4106 return (dev->operstate == IF_OPER_UP || 4107 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4108 } 4109 4110 /** 4111 * netif_device_present - is device available or removed 4112 * @dev: network device 4113 * 4114 * Check if device has not been removed from system. 4115 */ 4116 static inline bool netif_device_present(struct net_device *dev) 4117 { 4118 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4119 } 4120 4121 void netif_device_detach(struct net_device *dev); 4122 4123 void netif_device_attach(struct net_device *dev); 4124 4125 /* 4126 * Network interface message level settings 4127 */ 4128 4129 enum { 4130 NETIF_MSG_DRV_BIT, 4131 NETIF_MSG_PROBE_BIT, 4132 NETIF_MSG_LINK_BIT, 4133 NETIF_MSG_TIMER_BIT, 4134 NETIF_MSG_IFDOWN_BIT, 4135 NETIF_MSG_IFUP_BIT, 4136 NETIF_MSG_RX_ERR_BIT, 4137 NETIF_MSG_TX_ERR_BIT, 4138 NETIF_MSG_TX_QUEUED_BIT, 4139 NETIF_MSG_INTR_BIT, 4140 NETIF_MSG_TX_DONE_BIT, 4141 NETIF_MSG_RX_STATUS_BIT, 4142 NETIF_MSG_PKTDATA_BIT, 4143 NETIF_MSG_HW_BIT, 4144 NETIF_MSG_WOL_BIT, 4145 4146 /* When you add a new bit above, update netif_msg_class_names array 4147 * in net/ethtool/common.c 4148 */ 4149 NETIF_MSG_CLASS_COUNT, 4150 }; 4151 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4152 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4153 4154 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4155 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4156 4157 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4158 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4159 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4160 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4161 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4162 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4163 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4164 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4165 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4166 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4167 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4168 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4169 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4170 #define NETIF_MSG_HW __NETIF_MSG(HW) 4171 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4172 4173 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4174 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4175 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4176 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4177 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4178 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4179 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4180 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4181 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4182 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4183 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4184 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4185 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4186 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4187 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4188 4189 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4190 { 4191 /* use default */ 4192 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4193 return default_msg_enable_bits; 4194 if (debug_value == 0) /* no output */ 4195 return 0; 4196 /* set low N bits */ 4197 return (1U << debug_value) - 1; 4198 } 4199 4200 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4201 { 4202 spin_lock(&txq->_xmit_lock); 4203 txq->xmit_lock_owner = cpu; 4204 } 4205 4206 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4207 { 4208 __acquire(&txq->_xmit_lock); 4209 return true; 4210 } 4211 4212 static inline void __netif_tx_release(struct netdev_queue *txq) 4213 { 4214 __release(&txq->_xmit_lock); 4215 } 4216 4217 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4218 { 4219 spin_lock_bh(&txq->_xmit_lock); 4220 txq->xmit_lock_owner = smp_processor_id(); 4221 } 4222 4223 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4224 { 4225 bool ok = spin_trylock(&txq->_xmit_lock); 4226 if (likely(ok)) 4227 txq->xmit_lock_owner = smp_processor_id(); 4228 return ok; 4229 } 4230 4231 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4232 { 4233 txq->xmit_lock_owner = -1; 4234 spin_unlock(&txq->_xmit_lock); 4235 } 4236 4237 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4238 { 4239 txq->xmit_lock_owner = -1; 4240 spin_unlock_bh(&txq->_xmit_lock); 4241 } 4242 4243 static inline void txq_trans_update(struct netdev_queue *txq) 4244 { 4245 if (txq->xmit_lock_owner != -1) 4246 txq->trans_start = jiffies; 4247 } 4248 4249 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4250 static inline void netif_trans_update(struct net_device *dev) 4251 { 4252 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4253 4254 if (txq->trans_start != jiffies) 4255 txq->trans_start = jiffies; 4256 } 4257 4258 /** 4259 * netif_tx_lock - grab network device transmit lock 4260 * @dev: network device 4261 * 4262 * Get network device transmit lock 4263 */ 4264 static inline void netif_tx_lock(struct net_device *dev) 4265 { 4266 unsigned int i; 4267 int cpu; 4268 4269 spin_lock(&dev->tx_global_lock); 4270 cpu = smp_processor_id(); 4271 for (i = 0; i < dev->num_tx_queues; i++) { 4272 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4273 4274 /* We are the only thread of execution doing a 4275 * freeze, but we have to grab the _xmit_lock in 4276 * order to synchronize with threads which are in 4277 * the ->hard_start_xmit() handler and already 4278 * checked the frozen bit. 4279 */ 4280 __netif_tx_lock(txq, cpu); 4281 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4282 __netif_tx_unlock(txq); 4283 } 4284 } 4285 4286 static inline void netif_tx_lock_bh(struct net_device *dev) 4287 { 4288 local_bh_disable(); 4289 netif_tx_lock(dev); 4290 } 4291 4292 static inline void netif_tx_unlock(struct net_device *dev) 4293 { 4294 unsigned int i; 4295 4296 for (i = 0; i < dev->num_tx_queues; i++) { 4297 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4298 4299 /* No need to grab the _xmit_lock here. If the 4300 * queue is not stopped for another reason, we 4301 * force a schedule. 4302 */ 4303 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4304 netif_schedule_queue(txq); 4305 } 4306 spin_unlock(&dev->tx_global_lock); 4307 } 4308 4309 static inline void netif_tx_unlock_bh(struct net_device *dev) 4310 { 4311 netif_tx_unlock(dev); 4312 local_bh_enable(); 4313 } 4314 4315 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4316 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4317 __netif_tx_lock(txq, cpu); \ 4318 } else { \ 4319 __netif_tx_acquire(txq); \ 4320 } \ 4321 } 4322 4323 #define HARD_TX_TRYLOCK(dev, txq) \ 4324 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4325 __netif_tx_trylock(txq) : \ 4326 __netif_tx_acquire(txq)) 4327 4328 #define HARD_TX_UNLOCK(dev, txq) { \ 4329 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4330 __netif_tx_unlock(txq); \ 4331 } else { \ 4332 __netif_tx_release(txq); \ 4333 } \ 4334 } 4335 4336 static inline void netif_tx_disable(struct net_device *dev) 4337 { 4338 unsigned int i; 4339 int cpu; 4340 4341 local_bh_disable(); 4342 cpu = smp_processor_id(); 4343 for (i = 0; i < dev->num_tx_queues; i++) { 4344 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4345 4346 __netif_tx_lock(txq, cpu); 4347 netif_tx_stop_queue(txq); 4348 __netif_tx_unlock(txq); 4349 } 4350 local_bh_enable(); 4351 } 4352 4353 static inline void netif_addr_lock(struct net_device *dev) 4354 { 4355 unsigned char nest_level = 0; 4356 4357 #ifdef CONFIG_LOCKDEP 4358 nest_level = dev->nested_level; 4359 #endif 4360 spin_lock_nested(&dev->addr_list_lock, nest_level); 4361 } 4362 4363 static inline void netif_addr_lock_bh(struct net_device *dev) 4364 { 4365 unsigned char nest_level = 0; 4366 4367 #ifdef CONFIG_LOCKDEP 4368 nest_level = dev->nested_level; 4369 #endif 4370 local_bh_disable(); 4371 spin_lock_nested(&dev->addr_list_lock, nest_level); 4372 } 4373 4374 static inline void netif_addr_unlock(struct net_device *dev) 4375 { 4376 spin_unlock(&dev->addr_list_lock); 4377 } 4378 4379 static inline void netif_addr_unlock_bh(struct net_device *dev) 4380 { 4381 spin_unlock_bh(&dev->addr_list_lock); 4382 } 4383 4384 /* 4385 * dev_addrs walker. Should be used only for read access. Call with 4386 * rcu_read_lock held. 4387 */ 4388 #define for_each_dev_addr(dev, ha) \ 4389 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4390 4391 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4392 4393 void ether_setup(struct net_device *dev); 4394 4395 /* Support for loadable net-drivers */ 4396 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4397 unsigned char name_assign_type, 4398 void (*setup)(struct net_device *), 4399 unsigned int txqs, unsigned int rxqs); 4400 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4401 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4402 4403 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4404 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4405 count) 4406 4407 int register_netdev(struct net_device *dev); 4408 void unregister_netdev(struct net_device *dev); 4409 4410 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4411 4412 /* General hardware address lists handling functions */ 4413 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4414 struct netdev_hw_addr_list *from_list, int addr_len); 4415 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4416 struct netdev_hw_addr_list *from_list, int addr_len); 4417 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4418 struct net_device *dev, 4419 int (*sync)(struct net_device *, const unsigned char *), 4420 int (*unsync)(struct net_device *, 4421 const unsigned char *)); 4422 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4423 struct net_device *dev, 4424 int (*sync)(struct net_device *, 4425 const unsigned char *, int), 4426 int (*unsync)(struct net_device *, 4427 const unsigned char *, int)); 4428 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4429 struct net_device *dev, 4430 int (*unsync)(struct net_device *, 4431 const unsigned char *, int)); 4432 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4433 struct net_device *dev, 4434 int (*unsync)(struct net_device *, 4435 const unsigned char *)); 4436 void __hw_addr_init(struct netdev_hw_addr_list *list); 4437 4438 /* Functions used for device addresses handling */ 4439 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4440 unsigned char addr_type); 4441 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4442 unsigned char addr_type); 4443 void dev_addr_flush(struct net_device *dev); 4444 int dev_addr_init(struct net_device *dev); 4445 4446 /* Functions used for unicast addresses handling */ 4447 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4448 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4449 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4450 int dev_uc_sync(struct net_device *to, struct net_device *from); 4451 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4452 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4453 void dev_uc_flush(struct net_device *dev); 4454 void dev_uc_init(struct net_device *dev); 4455 4456 /** 4457 * __dev_uc_sync - Synchonize device's unicast list 4458 * @dev: device to sync 4459 * @sync: function to call if address should be added 4460 * @unsync: function to call if address should be removed 4461 * 4462 * Add newly added addresses to the interface, and release 4463 * addresses that have been deleted. 4464 */ 4465 static inline int __dev_uc_sync(struct net_device *dev, 4466 int (*sync)(struct net_device *, 4467 const unsigned char *), 4468 int (*unsync)(struct net_device *, 4469 const unsigned char *)) 4470 { 4471 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4472 } 4473 4474 /** 4475 * __dev_uc_unsync - Remove synchronized addresses from device 4476 * @dev: device to sync 4477 * @unsync: function to call if address should be removed 4478 * 4479 * Remove all addresses that were added to the device by dev_uc_sync(). 4480 */ 4481 static inline void __dev_uc_unsync(struct net_device *dev, 4482 int (*unsync)(struct net_device *, 4483 const unsigned char *)) 4484 { 4485 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4486 } 4487 4488 /* Functions used for multicast addresses handling */ 4489 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4490 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4491 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4492 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4493 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4494 int dev_mc_sync(struct net_device *to, struct net_device *from); 4495 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4496 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4497 void dev_mc_flush(struct net_device *dev); 4498 void dev_mc_init(struct net_device *dev); 4499 4500 /** 4501 * __dev_mc_sync - Synchonize device's multicast list 4502 * @dev: device to sync 4503 * @sync: function to call if address should be added 4504 * @unsync: function to call if address should be removed 4505 * 4506 * Add newly added addresses to the interface, and release 4507 * addresses that have been deleted. 4508 */ 4509 static inline int __dev_mc_sync(struct net_device *dev, 4510 int (*sync)(struct net_device *, 4511 const unsigned char *), 4512 int (*unsync)(struct net_device *, 4513 const unsigned char *)) 4514 { 4515 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4516 } 4517 4518 /** 4519 * __dev_mc_unsync - Remove synchronized addresses from device 4520 * @dev: device to sync 4521 * @unsync: function to call if address should be removed 4522 * 4523 * Remove all addresses that were added to the device by dev_mc_sync(). 4524 */ 4525 static inline void __dev_mc_unsync(struct net_device *dev, 4526 int (*unsync)(struct net_device *, 4527 const unsigned char *)) 4528 { 4529 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4530 } 4531 4532 /* Functions used for secondary unicast and multicast support */ 4533 void dev_set_rx_mode(struct net_device *dev); 4534 void __dev_set_rx_mode(struct net_device *dev); 4535 int dev_set_promiscuity(struct net_device *dev, int inc); 4536 int dev_set_allmulti(struct net_device *dev, int inc); 4537 void netdev_state_change(struct net_device *dev); 4538 void __netdev_notify_peers(struct net_device *dev); 4539 void netdev_notify_peers(struct net_device *dev); 4540 void netdev_features_change(struct net_device *dev); 4541 /* Load a device via the kmod */ 4542 void dev_load(struct net *net, const char *name); 4543 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4544 struct rtnl_link_stats64 *storage); 4545 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4546 const struct net_device_stats *netdev_stats); 4547 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4548 const struct pcpu_sw_netstats __percpu *netstats); 4549 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4550 4551 extern int netdev_max_backlog; 4552 extern int netdev_tstamp_prequeue; 4553 extern int weight_p; 4554 extern int dev_weight_rx_bias; 4555 extern int dev_weight_tx_bias; 4556 extern int dev_rx_weight; 4557 extern int dev_tx_weight; 4558 extern int gro_normal_batch; 4559 4560 enum { 4561 NESTED_SYNC_IMM_BIT, 4562 NESTED_SYNC_TODO_BIT, 4563 }; 4564 4565 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4566 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4567 4568 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4569 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4570 4571 struct netdev_nested_priv { 4572 unsigned char flags; 4573 void *data; 4574 }; 4575 4576 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4577 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4578 struct list_head **iter); 4579 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4580 struct list_head **iter); 4581 4582 #ifdef CONFIG_LOCKDEP 4583 static LIST_HEAD(net_unlink_list); 4584 4585 static inline void net_unlink_todo(struct net_device *dev) 4586 { 4587 if (list_empty(&dev->unlink_list)) 4588 list_add_tail(&dev->unlink_list, &net_unlink_list); 4589 } 4590 #endif 4591 4592 /* iterate through upper list, must be called under RCU read lock */ 4593 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4594 for (iter = &(dev)->adj_list.upper, \ 4595 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4596 updev; \ 4597 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4598 4599 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4600 int (*fn)(struct net_device *upper_dev, 4601 struct netdev_nested_priv *priv), 4602 struct netdev_nested_priv *priv); 4603 4604 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4605 struct net_device *upper_dev); 4606 4607 bool netdev_has_any_upper_dev(struct net_device *dev); 4608 4609 void *netdev_lower_get_next_private(struct net_device *dev, 4610 struct list_head **iter); 4611 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4612 struct list_head **iter); 4613 4614 #define netdev_for_each_lower_private(dev, priv, iter) \ 4615 for (iter = (dev)->adj_list.lower.next, \ 4616 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4617 priv; \ 4618 priv = netdev_lower_get_next_private(dev, &(iter))) 4619 4620 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4621 for (iter = &(dev)->adj_list.lower, \ 4622 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4623 priv; \ 4624 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4625 4626 void *netdev_lower_get_next(struct net_device *dev, 4627 struct list_head **iter); 4628 4629 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4630 for (iter = (dev)->adj_list.lower.next, \ 4631 ldev = netdev_lower_get_next(dev, &(iter)); \ 4632 ldev; \ 4633 ldev = netdev_lower_get_next(dev, &(iter))) 4634 4635 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4636 struct list_head **iter); 4637 int netdev_walk_all_lower_dev(struct net_device *dev, 4638 int (*fn)(struct net_device *lower_dev, 4639 struct netdev_nested_priv *priv), 4640 struct netdev_nested_priv *priv); 4641 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4642 int (*fn)(struct net_device *lower_dev, 4643 struct netdev_nested_priv *priv), 4644 struct netdev_nested_priv *priv); 4645 4646 void *netdev_adjacent_get_private(struct list_head *adj_list); 4647 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4648 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4649 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4650 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4651 struct netlink_ext_ack *extack); 4652 int netdev_master_upper_dev_link(struct net_device *dev, 4653 struct net_device *upper_dev, 4654 void *upper_priv, void *upper_info, 4655 struct netlink_ext_ack *extack); 4656 void netdev_upper_dev_unlink(struct net_device *dev, 4657 struct net_device *upper_dev); 4658 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4659 struct net_device *new_dev, 4660 struct net_device *dev, 4661 struct netlink_ext_ack *extack); 4662 void netdev_adjacent_change_commit(struct net_device *old_dev, 4663 struct net_device *new_dev, 4664 struct net_device *dev); 4665 void netdev_adjacent_change_abort(struct net_device *old_dev, 4666 struct net_device *new_dev, 4667 struct net_device *dev); 4668 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4669 void *netdev_lower_dev_get_private(struct net_device *dev, 4670 struct net_device *lower_dev); 4671 void netdev_lower_state_changed(struct net_device *lower_dev, 4672 void *lower_state_info); 4673 4674 /* RSS keys are 40 or 52 bytes long */ 4675 #define NETDEV_RSS_KEY_LEN 52 4676 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4677 void netdev_rss_key_fill(void *buffer, size_t len); 4678 4679 int skb_checksum_help(struct sk_buff *skb); 4680 int skb_crc32c_csum_help(struct sk_buff *skb); 4681 int skb_csum_hwoffload_help(struct sk_buff *skb, 4682 const netdev_features_t features); 4683 4684 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4685 netdev_features_t features, bool tx_path); 4686 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4687 netdev_features_t features); 4688 4689 struct netdev_bonding_info { 4690 ifslave slave; 4691 ifbond master; 4692 }; 4693 4694 struct netdev_notifier_bonding_info { 4695 struct netdev_notifier_info info; /* must be first */ 4696 struct netdev_bonding_info bonding_info; 4697 }; 4698 4699 void netdev_bonding_info_change(struct net_device *dev, 4700 struct netdev_bonding_info *bonding_info); 4701 4702 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4703 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4704 #else 4705 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4706 const void *data) 4707 { 4708 } 4709 #endif 4710 4711 static inline 4712 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4713 { 4714 return __skb_gso_segment(skb, features, true); 4715 } 4716 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4717 4718 static inline bool can_checksum_protocol(netdev_features_t features, 4719 __be16 protocol) 4720 { 4721 if (protocol == htons(ETH_P_FCOE)) 4722 return !!(features & NETIF_F_FCOE_CRC); 4723 4724 /* Assume this is an IP checksum (not SCTP CRC) */ 4725 4726 if (features & NETIF_F_HW_CSUM) { 4727 /* Can checksum everything */ 4728 return true; 4729 } 4730 4731 switch (protocol) { 4732 case htons(ETH_P_IP): 4733 return !!(features & NETIF_F_IP_CSUM); 4734 case htons(ETH_P_IPV6): 4735 return !!(features & NETIF_F_IPV6_CSUM); 4736 default: 4737 return false; 4738 } 4739 } 4740 4741 #ifdef CONFIG_BUG 4742 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4743 #else 4744 static inline void netdev_rx_csum_fault(struct net_device *dev, 4745 struct sk_buff *skb) 4746 { 4747 } 4748 #endif 4749 /* rx skb timestamps */ 4750 void net_enable_timestamp(void); 4751 void net_disable_timestamp(void); 4752 4753 #ifdef CONFIG_PROC_FS 4754 int __init dev_proc_init(void); 4755 #else 4756 #define dev_proc_init() 0 4757 #endif 4758 4759 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4760 struct sk_buff *skb, struct net_device *dev, 4761 bool more) 4762 { 4763 __this_cpu_write(softnet_data.xmit.more, more); 4764 return ops->ndo_start_xmit(skb, dev); 4765 } 4766 4767 static inline bool netdev_xmit_more(void) 4768 { 4769 return __this_cpu_read(softnet_data.xmit.more); 4770 } 4771 4772 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4773 struct netdev_queue *txq, bool more) 4774 { 4775 const struct net_device_ops *ops = dev->netdev_ops; 4776 netdev_tx_t rc; 4777 4778 rc = __netdev_start_xmit(ops, skb, dev, more); 4779 if (rc == NETDEV_TX_OK) 4780 txq_trans_update(txq); 4781 4782 return rc; 4783 } 4784 4785 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4786 const void *ns); 4787 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4788 const void *ns); 4789 4790 extern const struct kobj_ns_type_operations net_ns_type_operations; 4791 4792 const char *netdev_drivername(const struct net_device *dev); 4793 4794 void linkwatch_run_queue(void); 4795 4796 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4797 netdev_features_t f2) 4798 { 4799 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4800 if (f1 & NETIF_F_HW_CSUM) 4801 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4802 else 4803 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4804 } 4805 4806 return f1 & f2; 4807 } 4808 4809 static inline netdev_features_t netdev_get_wanted_features( 4810 struct net_device *dev) 4811 { 4812 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4813 } 4814 netdev_features_t netdev_increment_features(netdev_features_t all, 4815 netdev_features_t one, netdev_features_t mask); 4816 4817 /* Allow TSO being used on stacked device : 4818 * Performing the GSO segmentation before last device 4819 * is a performance improvement. 4820 */ 4821 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4822 netdev_features_t mask) 4823 { 4824 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4825 } 4826 4827 int __netdev_update_features(struct net_device *dev); 4828 void netdev_update_features(struct net_device *dev); 4829 void netdev_change_features(struct net_device *dev); 4830 4831 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4832 struct net_device *dev); 4833 4834 netdev_features_t passthru_features_check(struct sk_buff *skb, 4835 struct net_device *dev, 4836 netdev_features_t features); 4837 netdev_features_t netif_skb_features(struct sk_buff *skb); 4838 4839 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4840 { 4841 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4842 4843 /* check flags correspondence */ 4844 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4845 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4846 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4847 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4848 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4849 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4850 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4851 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4852 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4853 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4854 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4855 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4856 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4857 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4858 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4859 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4860 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4861 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4862 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4863 4864 return (features & feature) == feature; 4865 } 4866 4867 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4868 { 4869 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4870 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4871 } 4872 4873 static inline bool netif_needs_gso(struct sk_buff *skb, 4874 netdev_features_t features) 4875 { 4876 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4877 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4878 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4879 } 4880 4881 static inline void netif_set_gso_max_size(struct net_device *dev, 4882 unsigned int size) 4883 { 4884 dev->gso_max_size = size; 4885 } 4886 4887 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4888 int pulled_hlen, u16 mac_offset, 4889 int mac_len) 4890 { 4891 skb->protocol = protocol; 4892 skb->encapsulation = 1; 4893 skb_push(skb, pulled_hlen); 4894 skb_reset_transport_header(skb); 4895 skb->mac_header = mac_offset; 4896 skb->network_header = skb->mac_header + mac_len; 4897 skb->mac_len = mac_len; 4898 } 4899 4900 static inline bool netif_is_macsec(const struct net_device *dev) 4901 { 4902 return dev->priv_flags & IFF_MACSEC; 4903 } 4904 4905 static inline bool netif_is_macvlan(const struct net_device *dev) 4906 { 4907 return dev->priv_flags & IFF_MACVLAN; 4908 } 4909 4910 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4911 { 4912 return dev->priv_flags & IFF_MACVLAN_PORT; 4913 } 4914 4915 static inline bool netif_is_bond_master(const struct net_device *dev) 4916 { 4917 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4918 } 4919 4920 static inline bool netif_is_bond_slave(const struct net_device *dev) 4921 { 4922 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4923 } 4924 4925 static inline bool netif_supports_nofcs(struct net_device *dev) 4926 { 4927 return dev->priv_flags & IFF_SUPP_NOFCS; 4928 } 4929 4930 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4931 { 4932 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4933 } 4934 4935 static inline bool netif_is_l3_master(const struct net_device *dev) 4936 { 4937 return dev->priv_flags & IFF_L3MDEV_MASTER; 4938 } 4939 4940 static inline bool netif_is_l3_slave(const struct net_device *dev) 4941 { 4942 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4943 } 4944 4945 static inline bool netif_is_bridge_master(const struct net_device *dev) 4946 { 4947 return dev->priv_flags & IFF_EBRIDGE; 4948 } 4949 4950 static inline bool netif_is_bridge_port(const struct net_device *dev) 4951 { 4952 return dev->priv_flags & IFF_BRIDGE_PORT; 4953 } 4954 4955 static inline bool netif_is_ovs_master(const struct net_device *dev) 4956 { 4957 return dev->priv_flags & IFF_OPENVSWITCH; 4958 } 4959 4960 static inline bool netif_is_ovs_port(const struct net_device *dev) 4961 { 4962 return dev->priv_flags & IFF_OVS_DATAPATH; 4963 } 4964 4965 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4966 { 4967 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4968 } 4969 4970 static inline bool netif_is_team_master(const struct net_device *dev) 4971 { 4972 return dev->priv_flags & IFF_TEAM; 4973 } 4974 4975 static inline bool netif_is_team_port(const struct net_device *dev) 4976 { 4977 return dev->priv_flags & IFF_TEAM_PORT; 4978 } 4979 4980 static inline bool netif_is_lag_master(const struct net_device *dev) 4981 { 4982 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4983 } 4984 4985 static inline bool netif_is_lag_port(const struct net_device *dev) 4986 { 4987 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4988 } 4989 4990 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4991 { 4992 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4993 } 4994 4995 static inline bool netif_is_failover(const struct net_device *dev) 4996 { 4997 return dev->priv_flags & IFF_FAILOVER; 4998 } 4999 5000 static inline bool netif_is_failover_slave(const struct net_device *dev) 5001 { 5002 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5003 } 5004 5005 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5006 static inline void netif_keep_dst(struct net_device *dev) 5007 { 5008 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5009 } 5010 5011 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5012 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5013 { 5014 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5015 return dev->priv_flags & IFF_MACSEC; 5016 } 5017 5018 extern struct pernet_operations __net_initdata loopback_net_ops; 5019 5020 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5021 5022 /* netdev_printk helpers, similar to dev_printk */ 5023 5024 static inline const char *netdev_name(const struct net_device *dev) 5025 { 5026 if (!dev->name[0] || strchr(dev->name, '%')) 5027 return "(unnamed net_device)"; 5028 return dev->name; 5029 } 5030 5031 static inline bool netdev_unregistering(const struct net_device *dev) 5032 { 5033 return dev->reg_state == NETREG_UNREGISTERING; 5034 } 5035 5036 static inline const char *netdev_reg_state(const struct net_device *dev) 5037 { 5038 switch (dev->reg_state) { 5039 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5040 case NETREG_REGISTERED: return ""; 5041 case NETREG_UNREGISTERING: return " (unregistering)"; 5042 case NETREG_UNREGISTERED: return " (unregistered)"; 5043 case NETREG_RELEASED: return " (released)"; 5044 case NETREG_DUMMY: return " (dummy)"; 5045 } 5046 5047 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5048 return " (unknown)"; 5049 } 5050 5051 __printf(3, 4) __cold 5052 void netdev_printk(const char *level, const struct net_device *dev, 5053 const char *format, ...); 5054 __printf(2, 3) __cold 5055 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5056 __printf(2, 3) __cold 5057 void netdev_alert(const struct net_device *dev, const char *format, ...); 5058 __printf(2, 3) __cold 5059 void netdev_crit(const struct net_device *dev, const char *format, ...); 5060 __printf(2, 3) __cold 5061 void netdev_err(const struct net_device *dev, const char *format, ...); 5062 __printf(2, 3) __cold 5063 void netdev_warn(const struct net_device *dev, const char *format, ...); 5064 __printf(2, 3) __cold 5065 void netdev_notice(const struct net_device *dev, const char *format, ...); 5066 __printf(2, 3) __cold 5067 void netdev_info(const struct net_device *dev, const char *format, ...); 5068 5069 #define netdev_level_once(level, dev, fmt, ...) \ 5070 do { \ 5071 static bool __print_once __read_mostly; \ 5072 \ 5073 if (!__print_once) { \ 5074 __print_once = true; \ 5075 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5076 } \ 5077 } while (0) 5078 5079 #define netdev_emerg_once(dev, fmt, ...) \ 5080 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5081 #define netdev_alert_once(dev, fmt, ...) \ 5082 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5083 #define netdev_crit_once(dev, fmt, ...) \ 5084 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5085 #define netdev_err_once(dev, fmt, ...) \ 5086 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5087 #define netdev_warn_once(dev, fmt, ...) \ 5088 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5089 #define netdev_notice_once(dev, fmt, ...) \ 5090 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5091 #define netdev_info_once(dev, fmt, ...) \ 5092 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5093 5094 #define MODULE_ALIAS_NETDEV(device) \ 5095 MODULE_ALIAS("netdev-" device) 5096 5097 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5098 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5099 #define netdev_dbg(__dev, format, args...) \ 5100 do { \ 5101 dynamic_netdev_dbg(__dev, format, ##args); \ 5102 } while (0) 5103 #elif defined(DEBUG) 5104 #define netdev_dbg(__dev, format, args...) \ 5105 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5106 #else 5107 #define netdev_dbg(__dev, format, args...) \ 5108 ({ \ 5109 if (0) \ 5110 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5111 }) 5112 #endif 5113 5114 #if defined(VERBOSE_DEBUG) 5115 #define netdev_vdbg netdev_dbg 5116 #else 5117 5118 #define netdev_vdbg(dev, format, args...) \ 5119 ({ \ 5120 if (0) \ 5121 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5122 0; \ 5123 }) 5124 #endif 5125 5126 /* 5127 * netdev_WARN() acts like dev_printk(), but with the key difference 5128 * of using a WARN/WARN_ON to get the message out, including the 5129 * file/line information and a backtrace. 5130 */ 5131 #define netdev_WARN(dev, format, args...) \ 5132 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5133 netdev_reg_state(dev), ##args) 5134 5135 #define netdev_WARN_ONCE(dev, format, args...) \ 5136 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5137 netdev_reg_state(dev), ##args) 5138 5139 /* netif printk helpers, similar to netdev_printk */ 5140 5141 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5142 do { \ 5143 if (netif_msg_##type(priv)) \ 5144 netdev_printk(level, (dev), fmt, ##args); \ 5145 } while (0) 5146 5147 #define netif_level(level, priv, type, dev, fmt, args...) \ 5148 do { \ 5149 if (netif_msg_##type(priv)) \ 5150 netdev_##level(dev, fmt, ##args); \ 5151 } while (0) 5152 5153 #define netif_emerg(priv, type, dev, fmt, args...) \ 5154 netif_level(emerg, priv, type, dev, fmt, ##args) 5155 #define netif_alert(priv, type, dev, fmt, args...) \ 5156 netif_level(alert, priv, type, dev, fmt, ##args) 5157 #define netif_crit(priv, type, dev, fmt, args...) \ 5158 netif_level(crit, priv, type, dev, fmt, ##args) 5159 #define netif_err(priv, type, dev, fmt, args...) \ 5160 netif_level(err, priv, type, dev, fmt, ##args) 5161 #define netif_warn(priv, type, dev, fmt, args...) \ 5162 netif_level(warn, priv, type, dev, fmt, ##args) 5163 #define netif_notice(priv, type, dev, fmt, args...) \ 5164 netif_level(notice, priv, type, dev, fmt, ##args) 5165 #define netif_info(priv, type, dev, fmt, args...) \ 5166 netif_level(info, priv, type, dev, fmt, ##args) 5167 5168 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5169 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5170 #define netif_dbg(priv, type, netdev, format, args...) \ 5171 do { \ 5172 if (netif_msg_##type(priv)) \ 5173 dynamic_netdev_dbg(netdev, format, ##args); \ 5174 } while (0) 5175 #elif defined(DEBUG) 5176 #define netif_dbg(priv, type, dev, format, args...) \ 5177 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5178 #else 5179 #define netif_dbg(priv, type, dev, format, args...) \ 5180 ({ \ 5181 if (0) \ 5182 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5183 0; \ 5184 }) 5185 #endif 5186 5187 /* if @cond then downgrade to debug, else print at @level */ 5188 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5189 do { \ 5190 if (cond) \ 5191 netif_dbg(priv, type, netdev, fmt, ##args); \ 5192 else \ 5193 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5194 } while (0) 5195 5196 #if defined(VERBOSE_DEBUG) 5197 #define netif_vdbg netif_dbg 5198 #else 5199 #define netif_vdbg(priv, type, dev, format, args...) \ 5200 ({ \ 5201 if (0) \ 5202 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5203 0; \ 5204 }) 5205 #endif 5206 5207 /* 5208 * The list of packet types we will receive (as opposed to discard) 5209 * and the routines to invoke. 5210 * 5211 * Why 16. Because with 16 the only overlap we get on a hash of the 5212 * low nibble of the protocol value is RARP/SNAP/X.25. 5213 * 5214 * 0800 IP 5215 * 0001 802.3 5216 * 0002 AX.25 5217 * 0004 802.2 5218 * 8035 RARP 5219 * 0005 SNAP 5220 * 0805 X.25 5221 * 0806 ARP 5222 * 8137 IPX 5223 * 0009 Localtalk 5224 * 86DD IPv6 5225 */ 5226 #define PTYPE_HASH_SIZE (16) 5227 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5228 5229 extern struct net_device *blackhole_netdev; 5230 5231 #endif /* _LINUX_NETDEVICE_H */ 5232