1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void netdev_set_default_ethtool_ops(struct net_device *dev, 74 const struct ethtool_ops *ops); 75 76 /* Backlog congestion levels */ 77 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 78 #define NET_RX_DROP 1 /* packet dropped */ 79 80 #define MAX_NEST_DEV 8 81 82 /* 83 * Transmit return codes: transmit return codes originate from three different 84 * namespaces: 85 * 86 * - qdisc return codes 87 * - driver transmit return codes 88 * - errno values 89 * 90 * Drivers are allowed to return any one of those in their hard_start_xmit() 91 * function. Real network devices commonly used with qdiscs should only return 92 * the driver transmit return codes though - when qdiscs are used, the actual 93 * transmission happens asynchronously, so the value is not propagated to 94 * higher layers. Virtual network devices transmit synchronously; in this case 95 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 96 * others are propagated to higher layers. 97 */ 98 99 /* qdisc ->enqueue() return codes. */ 100 #define NET_XMIT_SUCCESS 0x00 101 #define NET_XMIT_DROP 0x01 /* skb dropped */ 102 #define NET_XMIT_CN 0x02 /* congestion notification */ 103 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 104 105 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 106 * indicates that the device will soon be dropping packets, or already drops 107 * some packets of the same priority; prompting us to send less aggressively. */ 108 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 109 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 110 111 /* Driver transmit return codes */ 112 #define NETDEV_TX_MASK 0xf0 113 114 enum netdev_tx { 115 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 116 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 117 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 118 }; 119 typedef enum netdev_tx netdev_tx_t; 120 121 /* 122 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 123 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 124 */ 125 static inline bool dev_xmit_complete(int rc) 126 { 127 /* 128 * Positive cases with an skb consumed by a driver: 129 * - successful transmission (rc == NETDEV_TX_OK) 130 * - error while transmitting (rc < 0) 131 * - error while queueing to a different device (rc & NET_XMIT_MASK) 132 */ 133 if (likely(rc < NET_XMIT_MASK)) 134 return true; 135 136 return false; 137 } 138 139 /* 140 * Compute the worst-case header length according to the protocols 141 * used. 142 */ 143 144 #if defined(CONFIG_HYPERV_NET) 145 # define LL_MAX_HEADER 128 146 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 147 # if defined(CONFIG_MAC80211_MESH) 148 # define LL_MAX_HEADER 128 149 # else 150 # define LL_MAX_HEADER 96 151 # endif 152 #else 153 # define LL_MAX_HEADER 32 154 #endif 155 156 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 157 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 158 #define MAX_HEADER LL_MAX_HEADER 159 #else 160 #define MAX_HEADER (LL_MAX_HEADER + 48) 161 #endif 162 163 /* 164 * Old network device statistics. Fields are native words 165 * (unsigned long) so they can be read and written atomically. 166 */ 167 168 struct net_device_stats { 169 unsigned long rx_packets; 170 unsigned long tx_packets; 171 unsigned long rx_bytes; 172 unsigned long tx_bytes; 173 unsigned long rx_errors; 174 unsigned long tx_errors; 175 unsigned long rx_dropped; 176 unsigned long tx_dropped; 177 unsigned long multicast; 178 unsigned long collisions; 179 unsigned long rx_length_errors; 180 unsigned long rx_over_errors; 181 unsigned long rx_crc_errors; 182 unsigned long rx_frame_errors; 183 unsigned long rx_fifo_errors; 184 unsigned long rx_missed_errors; 185 unsigned long tx_aborted_errors; 186 unsigned long tx_carrier_errors; 187 unsigned long tx_fifo_errors; 188 unsigned long tx_heartbeat_errors; 189 unsigned long tx_window_errors; 190 unsigned long rx_compressed; 191 unsigned long tx_compressed; 192 }; 193 194 195 #include <linux/cache.h> 196 #include <linux/skbuff.h> 197 198 #ifdef CONFIG_RPS 199 #include <linux/static_key.h> 200 extern struct static_key_false rps_needed; 201 extern struct static_key_false rfs_needed; 202 #endif 203 204 struct neighbour; 205 struct neigh_parms; 206 struct sk_buff; 207 208 struct netdev_hw_addr { 209 struct list_head list; 210 unsigned char addr[MAX_ADDR_LEN]; 211 unsigned char type; 212 #define NETDEV_HW_ADDR_T_LAN 1 213 #define NETDEV_HW_ADDR_T_SAN 2 214 #define NETDEV_HW_ADDR_T_SLAVE 3 215 #define NETDEV_HW_ADDR_T_UNICAST 4 216 #define NETDEV_HW_ADDR_T_MULTICAST 5 217 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_HASHED, /* In NAPI hash (busy polling possible) */ 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 }; 361 362 enum { 363 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 364 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 365 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 366 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 367 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 368 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 369 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 370 }; 371 372 enum gro_result { 373 GRO_MERGED, 374 GRO_MERGED_FREE, 375 GRO_HELD, 376 GRO_NORMAL, 377 GRO_DROP, 378 GRO_CONSUMED, 379 }; 380 typedef enum gro_result gro_result_t; 381 382 /* 383 * enum rx_handler_result - Possible return values for rx_handlers. 384 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 385 * further. 386 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 387 * case skb->dev was changed by rx_handler. 388 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 389 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 390 * 391 * rx_handlers are functions called from inside __netif_receive_skb(), to do 392 * special processing of the skb, prior to delivery to protocol handlers. 393 * 394 * Currently, a net_device can only have a single rx_handler registered. Trying 395 * to register a second rx_handler will return -EBUSY. 396 * 397 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 398 * To unregister a rx_handler on a net_device, use 399 * netdev_rx_handler_unregister(). 400 * 401 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 402 * do with the skb. 403 * 404 * If the rx_handler consumed the skb in some way, it should return 405 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 406 * the skb to be delivered in some other way. 407 * 408 * If the rx_handler changed skb->dev, to divert the skb to another 409 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 410 * new device will be called if it exists. 411 * 412 * If the rx_handler decides the skb should be ignored, it should return 413 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 414 * are registered on exact device (ptype->dev == skb->dev). 415 * 416 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 417 * delivered, it should return RX_HANDLER_PASS. 418 * 419 * A device without a registered rx_handler will behave as if rx_handler 420 * returned RX_HANDLER_PASS. 421 */ 422 423 enum rx_handler_result { 424 RX_HANDLER_CONSUMED, 425 RX_HANDLER_ANOTHER, 426 RX_HANDLER_EXACT, 427 RX_HANDLER_PASS, 428 }; 429 typedef enum rx_handler_result rx_handler_result_t; 430 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 431 432 void __napi_schedule(struct napi_struct *n); 433 void __napi_schedule_irqoff(struct napi_struct *n); 434 435 static inline bool napi_disable_pending(struct napi_struct *n) 436 { 437 return test_bit(NAPI_STATE_DISABLE, &n->state); 438 } 439 440 bool napi_schedule_prep(struct napi_struct *n); 441 442 /** 443 * napi_schedule - schedule NAPI poll 444 * @n: NAPI context 445 * 446 * Schedule NAPI poll routine to be called if it is not already 447 * running. 448 */ 449 static inline void napi_schedule(struct napi_struct *n) 450 { 451 if (napi_schedule_prep(n)) 452 __napi_schedule(n); 453 } 454 455 /** 456 * napi_schedule_irqoff - schedule NAPI poll 457 * @n: NAPI context 458 * 459 * Variant of napi_schedule(), assuming hard irqs are masked. 460 */ 461 static inline void napi_schedule_irqoff(struct napi_struct *n) 462 { 463 if (napi_schedule_prep(n)) 464 __napi_schedule_irqoff(n); 465 } 466 467 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 468 static inline bool napi_reschedule(struct napi_struct *napi) 469 { 470 if (napi_schedule_prep(napi)) { 471 __napi_schedule(napi); 472 return true; 473 } 474 return false; 475 } 476 477 bool napi_complete_done(struct napi_struct *n, int work_done); 478 /** 479 * napi_complete - NAPI processing complete 480 * @n: NAPI context 481 * 482 * Mark NAPI processing as complete. 483 * Consider using napi_complete_done() instead. 484 * Return false if device should avoid rearming interrupts. 485 */ 486 static inline bool napi_complete(struct napi_struct *n) 487 { 488 return napi_complete_done(n, 0); 489 } 490 491 /** 492 * napi_hash_del - remove a NAPI from global table 493 * @napi: NAPI context 494 * 495 * Warning: caller must observe RCU grace period 496 * before freeing memory containing @napi, if 497 * this function returns true. 498 * Note: core networking stack automatically calls it 499 * from netif_napi_del(). 500 * Drivers might want to call this helper to combine all 501 * the needed RCU grace periods into a single one. 502 */ 503 bool napi_hash_del(struct napi_struct *napi); 504 505 /** 506 * napi_disable - prevent NAPI from scheduling 507 * @n: NAPI context 508 * 509 * Stop NAPI from being scheduled on this context. 510 * Waits till any outstanding processing completes. 511 */ 512 void napi_disable(struct napi_struct *n); 513 514 /** 515 * napi_enable - enable NAPI scheduling 516 * @n: NAPI context 517 * 518 * Resume NAPI from being scheduled on this context. 519 * Must be paired with napi_disable. 520 */ 521 static inline void napi_enable(struct napi_struct *n) 522 { 523 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 524 smp_mb__before_atomic(); 525 clear_bit(NAPI_STATE_SCHED, &n->state); 526 clear_bit(NAPI_STATE_NPSVC, &n->state); 527 } 528 529 /** 530 * napi_synchronize - wait until NAPI is not running 531 * @n: NAPI context 532 * 533 * Wait until NAPI is done being scheduled on this context. 534 * Waits till any outstanding processing completes but 535 * does not disable future activations. 536 */ 537 static inline void napi_synchronize(const struct napi_struct *n) 538 { 539 if (IS_ENABLED(CONFIG_SMP)) 540 while (test_bit(NAPI_STATE_SCHED, &n->state)) 541 msleep(1); 542 else 543 barrier(); 544 } 545 546 /** 547 * napi_if_scheduled_mark_missed - if napi is running, set the 548 * NAPIF_STATE_MISSED 549 * @n: NAPI context 550 * 551 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 552 * NAPI is scheduled. 553 **/ 554 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 555 { 556 unsigned long val, new; 557 558 do { 559 val = READ_ONCE(n->state); 560 if (val & NAPIF_STATE_DISABLE) 561 return true; 562 563 if (!(val & NAPIF_STATE_SCHED)) 564 return false; 565 566 new = val | NAPIF_STATE_MISSED; 567 } while (cmpxchg(&n->state, val, new) != val); 568 569 return true; 570 } 571 572 enum netdev_queue_state_t { 573 __QUEUE_STATE_DRV_XOFF, 574 __QUEUE_STATE_STACK_XOFF, 575 __QUEUE_STATE_FROZEN, 576 }; 577 578 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 579 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 580 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 581 582 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 583 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 584 QUEUE_STATE_FROZEN) 585 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 586 QUEUE_STATE_FROZEN) 587 588 /* 589 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 590 * netif_tx_* functions below are used to manipulate this flag. The 591 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 592 * queue independently. The netif_xmit_*stopped functions below are called 593 * to check if the queue has been stopped by the driver or stack (either 594 * of the XOFF bits are set in the state). Drivers should not need to call 595 * netif_xmit*stopped functions, they should only be using netif_tx_*. 596 */ 597 598 struct netdev_queue { 599 /* 600 * read-mostly part 601 */ 602 struct net_device *dev; 603 struct Qdisc __rcu *qdisc; 604 struct Qdisc *qdisc_sleeping; 605 #ifdef CONFIG_SYSFS 606 struct kobject kobj; 607 #endif 608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 609 int numa_node; 610 #endif 611 unsigned long tx_maxrate; 612 /* 613 * Number of TX timeouts for this queue 614 * (/sys/class/net/DEV/Q/trans_timeout) 615 */ 616 unsigned long trans_timeout; 617 618 /* Subordinate device that the queue has been assigned to */ 619 struct net_device *sb_dev; 620 #ifdef CONFIG_XDP_SOCKETS 621 struct xdp_umem *umem; 622 #endif 623 /* 624 * write-mostly part 625 */ 626 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 627 int xmit_lock_owner; 628 /* 629 * Time (in jiffies) of last Tx 630 */ 631 unsigned long trans_start; 632 633 unsigned long state; 634 635 #ifdef CONFIG_BQL 636 struct dql dql; 637 #endif 638 } ____cacheline_aligned_in_smp; 639 640 extern int sysctl_fb_tunnels_only_for_init_net; 641 extern int sysctl_devconf_inherit_init_net; 642 643 static inline bool net_has_fallback_tunnels(const struct net *net) 644 { 645 return net == &init_net || 646 !IS_ENABLED(CONFIG_SYSCTL) || 647 !sysctl_fb_tunnels_only_for_init_net; 648 } 649 650 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 return q->numa_node; 654 #else 655 return NUMA_NO_NODE; 656 #endif 657 } 658 659 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 660 { 661 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 662 q->numa_node = node; 663 #endif 664 } 665 666 #ifdef CONFIG_RPS 667 /* 668 * This structure holds an RPS map which can be of variable length. The 669 * map is an array of CPUs. 670 */ 671 struct rps_map { 672 unsigned int len; 673 struct rcu_head rcu; 674 u16 cpus[]; 675 }; 676 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 677 678 /* 679 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 680 * tail pointer for that CPU's input queue at the time of last enqueue, and 681 * a hardware filter index. 682 */ 683 struct rps_dev_flow { 684 u16 cpu; 685 u16 filter; 686 unsigned int last_qtail; 687 }; 688 #define RPS_NO_FILTER 0xffff 689 690 /* 691 * The rps_dev_flow_table structure contains a table of flow mappings. 692 */ 693 struct rps_dev_flow_table { 694 unsigned int mask; 695 struct rcu_head rcu; 696 struct rps_dev_flow flows[]; 697 }; 698 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 699 ((_num) * sizeof(struct rps_dev_flow))) 700 701 /* 702 * The rps_sock_flow_table contains mappings of flows to the last CPU 703 * on which they were processed by the application (set in recvmsg). 704 * Each entry is a 32bit value. Upper part is the high-order bits 705 * of flow hash, lower part is CPU number. 706 * rps_cpu_mask is used to partition the space, depending on number of 707 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 708 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 709 * meaning we use 32-6=26 bits for the hash. 710 */ 711 struct rps_sock_flow_table { 712 u32 mask; 713 714 u32 ents[] ____cacheline_aligned_in_smp; 715 }; 716 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 717 718 #define RPS_NO_CPU 0xffff 719 720 extern u32 rps_cpu_mask; 721 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 722 723 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 724 u32 hash) 725 { 726 if (table && hash) { 727 unsigned int index = hash & table->mask; 728 u32 val = hash & ~rps_cpu_mask; 729 730 /* We only give a hint, preemption can change CPU under us */ 731 val |= raw_smp_processor_id(); 732 733 if (table->ents[index] != val) 734 table->ents[index] = val; 735 } 736 } 737 738 #ifdef CONFIG_RFS_ACCEL 739 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 740 u16 filter_id); 741 #endif 742 #endif /* CONFIG_RPS */ 743 744 /* This structure contains an instance of an RX queue. */ 745 struct netdev_rx_queue { 746 #ifdef CONFIG_RPS 747 struct rps_map __rcu *rps_map; 748 struct rps_dev_flow_table __rcu *rps_flow_table; 749 #endif 750 struct kobject kobj; 751 struct net_device *dev; 752 struct xdp_rxq_info xdp_rxq; 753 #ifdef CONFIG_XDP_SOCKETS 754 struct xdp_umem *umem; 755 #endif 756 } ____cacheline_aligned_in_smp; 757 758 /* 759 * RX queue sysfs structures and functions. 760 */ 761 struct rx_queue_attribute { 762 struct attribute attr; 763 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 764 ssize_t (*store)(struct netdev_rx_queue *queue, 765 const char *buf, size_t len); 766 }; 767 768 #ifdef CONFIG_XPS 769 /* 770 * This structure holds an XPS map which can be of variable length. The 771 * map is an array of queues. 772 */ 773 struct xps_map { 774 unsigned int len; 775 unsigned int alloc_len; 776 struct rcu_head rcu; 777 u16 queues[]; 778 }; 779 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 780 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 781 - sizeof(struct xps_map)) / sizeof(u16)) 782 783 /* 784 * This structure holds all XPS maps for device. Maps are indexed by CPU. 785 */ 786 struct xps_dev_maps { 787 struct rcu_head rcu; 788 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 789 }; 790 791 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 792 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 793 794 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 795 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 796 797 #endif /* CONFIG_XPS */ 798 799 #define TC_MAX_QUEUE 16 800 #define TC_BITMASK 15 801 /* HW offloaded queuing disciplines txq count and offset maps */ 802 struct netdev_tc_txq { 803 u16 count; 804 u16 offset; 805 }; 806 807 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 808 /* 809 * This structure is to hold information about the device 810 * configured to run FCoE protocol stack. 811 */ 812 struct netdev_fcoe_hbainfo { 813 char manufacturer[64]; 814 char serial_number[64]; 815 char hardware_version[64]; 816 char driver_version[64]; 817 char optionrom_version[64]; 818 char firmware_version[64]; 819 char model[256]; 820 char model_description[256]; 821 }; 822 #endif 823 824 #define MAX_PHYS_ITEM_ID_LEN 32 825 826 /* This structure holds a unique identifier to identify some 827 * physical item (port for example) used by a netdevice. 828 */ 829 struct netdev_phys_item_id { 830 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 831 unsigned char id_len; 832 }; 833 834 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 835 struct netdev_phys_item_id *b) 836 { 837 return a->id_len == b->id_len && 838 memcmp(a->id, b->id, a->id_len) == 0; 839 } 840 841 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 842 struct sk_buff *skb, 843 struct net_device *sb_dev); 844 845 enum tc_setup_type { 846 TC_SETUP_QDISC_MQPRIO, 847 TC_SETUP_CLSU32, 848 TC_SETUP_CLSFLOWER, 849 TC_SETUP_CLSMATCHALL, 850 TC_SETUP_CLSBPF, 851 TC_SETUP_BLOCK, 852 TC_SETUP_QDISC_CBS, 853 TC_SETUP_QDISC_RED, 854 TC_SETUP_QDISC_PRIO, 855 TC_SETUP_QDISC_MQ, 856 TC_SETUP_QDISC_ETF, 857 TC_SETUP_ROOT_QDISC, 858 TC_SETUP_QDISC_GRED, 859 TC_SETUP_QDISC_TAPRIO, 860 TC_SETUP_FT, 861 TC_SETUP_QDISC_ETS, 862 TC_SETUP_QDISC_TBF, 863 TC_SETUP_QDISC_FIFO, 864 }; 865 866 /* These structures hold the attributes of bpf state that are being passed 867 * to the netdevice through the bpf op. 868 */ 869 enum bpf_netdev_command { 870 /* Set or clear a bpf program used in the earliest stages of packet 871 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 872 * is responsible for calling bpf_prog_put on any old progs that are 873 * stored. In case of error, the callee need not release the new prog 874 * reference, but on success it takes ownership and must bpf_prog_put 875 * when it is no longer used. 876 */ 877 XDP_SETUP_PROG, 878 XDP_SETUP_PROG_HW, 879 XDP_QUERY_PROG, 880 XDP_QUERY_PROG_HW, 881 /* BPF program for offload callbacks, invoked at program load time. */ 882 BPF_OFFLOAD_MAP_ALLOC, 883 BPF_OFFLOAD_MAP_FREE, 884 XDP_SETUP_XSK_UMEM, 885 }; 886 887 struct bpf_prog_offload_ops; 888 struct netlink_ext_ack; 889 struct xdp_umem; 890 struct xdp_dev_bulk_queue; 891 892 struct netdev_bpf { 893 enum bpf_netdev_command command; 894 union { 895 /* XDP_SETUP_PROG */ 896 struct { 897 u32 flags; 898 struct bpf_prog *prog; 899 struct netlink_ext_ack *extack; 900 }; 901 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 902 struct { 903 u32 prog_id; 904 /* flags with which program was installed */ 905 u32 prog_flags; 906 }; 907 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 908 struct { 909 struct bpf_offloaded_map *offmap; 910 }; 911 /* XDP_SETUP_XSK_UMEM */ 912 struct { 913 struct xdp_umem *umem; 914 u16 queue_id; 915 } xsk; 916 }; 917 }; 918 919 /* Flags for ndo_xsk_wakeup. */ 920 #define XDP_WAKEUP_RX (1 << 0) 921 #define XDP_WAKEUP_TX (1 << 1) 922 923 #ifdef CONFIG_XFRM_OFFLOAD 924 struct xfrmdev_ops { 925 int (*xdo_dev_state_add) (struct xfrm_state *x); 926 void (*xdo_dev_state_delete) (struct xfrm_state *x); 927 void (*xdo_dev_state_free) (struct xfrm_state *x); 928 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 929 struct xfrm_state *x); 930 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 931 }; 932 #endif 933 934 struct dev_ifalias { 935 struct rcu_head rcuhead; 936 char ifalias[]; 937 }; 938 939 struct devlink; 940 struct tlsdev_ops; 941 942 struct netdev_name_node { 943 struct hlist_node hlist; 944 struct list_head list; 945 struct net_device *dev; 946 const char *name; 947 }; 948 949 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 950 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 951 952 struct netdev_net_notifier { 953 struct list_head list; 954 struct notifier_block *nb; 955 }; 956 957 /* 958 * This structure defines the management hooks for network devices. 959 * The following hooks can be defined; unless noted otherwise, they are 960 * optional and can be filled with a null pointer. 961 * 962 * int (*ndo_init)(struct net_device *dev); 963 * This function is called once when a network device is registered. 964 * The network device can use this for any late stage initialization 965 * or semantic validation. It can fail with an error code which will 966 * be propagated back to register_netdev. 967 * 968 * void (*ndo_uninit)(struct net_device *dev); 969 * This function is called when device is unregistered or when registration 970 * fails. It is not called if init fails. 971 * 972 * int (*ndo_open)(struct net_device *dev); 973 * This function is called when a network device transitions to the up 974 * state. 975 * 976 * int (*ndo_stop)(struct net_device *dev); 977 * This function is called when a network device transitions to the down 978 * state. 979 * 980 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 981 * struct net_device *dev); 982 * Called when a packet needs to be transmitted. 983 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 984 * the queue before that can happen; it's for obsolete devices and weird 985 * corner cases, but the stack really does a non-trivial amount 986 * of useless work if you return NETDEV_TX_BUSY. 987 * Required; cannot be NULL. 988 * 989 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 990 * struct net_device *dev 991 * netdev_features_t features); 992 * Called by core transmit path to determine if device is capable of 993 * performing offload operations on a given packet. This is to give 994 * the device an opportunity to implement any restrictions that cannot 995 * be otherwise expressed by feature flags. The check is called with 996 * the set of features that the stack has calculated and it returns 997 * those the driver believes to be appropriate. 998 * 999 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1000 * struct net_device *sb_dev); 1001 * Called to decide which queue to use when device supports multiple 1002 * transmit queues. 1003 * 1004 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1005 * This function is called to allow device receiver to make 1006 * changes to configuration when multicast or promiscuous is enabled. 1007 * 1008 * void (*ndo_set_rx_mode)(struct net_device *dev); 1009 * This function is called device changes address list filtering. 1010 * If driver handles unicast address filtering, it should set 1011 * IFF_UNICAST_FLT in its priv_flags. 1012 * 1013 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1014 * This function is called when the Media Access Control address 1015 * needs to be changed. If this interface is not defined, the 1016 * MAC address can not be changed. 1017 * 1018 * int (*ndo_validate_addr)(struct net_device *dev); 1019 * Test if Media Access Control address is valid for the device. 1020 * 1021 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1022 * Called when a user requests an ioctl which can't be handled by 1023 * the generic interface code. If not defined ioctls return 1024 * not supported error code. 1025 * 1026 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1027 * Used to set network devices bus interface parameters. This interface 1028 * is retained for legacy reasons; new devices should use the bus 1029 * interface (PCI) for low level management. 1030 * 1031 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1032 * Called when a user wants to change the Maximum Transfer Unit 1033 * of a device. 1034 * 1035 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1036 * Callback used when the transmitter has not made any progress 1037 * for dev->watchdog ticks. 1038 * 1039 * void (*ndo_get_stats64)(struct net_device *dev, 1040 * struct rtnl_link_stats64 *storage); 1041 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1042 * Called when a user wants to get the network device usage 1043 * statistics. Drivers must do one of the following: 1044 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1045 * rtnl_link_stats64 structure passed by the caller. 1046 * 2. Define @ndo_get_stats to update a net_device_stats structure 1047 * (which should normally be dev->stats) and return a pointer to 1048 * it. The structure may be changed asynchronously only if each 1049 * field is written atomically. 1050 * 3. Update dev->stats asynchronously and atomically, and define 1051 * neither operation. 1052 * 1053 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1054 * Return true if this device supports offload stats of this attr_id. 1055 * 1056 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1057 * void *attr_data) 1058 * Get statistics for offload operations by attr_id. Write it into the 1059 * attr_data pointer. 1060 * 1061 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1062 * If device supports VLAN filtering this function is called when a 1063 * VLAN id is registered. 1064 * 1065 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1066 * If device supports VLAN filtering this function is called when a 1067 * VLAN id is unregistered. 1068 * 1069 * void (*ndo_poll_controller)(struct net_device *dev); 1070 * 1071 * SR-IOV management functions. 1072 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1073 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1074 * u8 qos, __be16 proto); 1075 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1076 * int max_tx_rate); 1077 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1078 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1079 * int (*ndo_get_vf_config)(struct net_device *dev, 1080 * int vf, struct ifla_vf_info *ivf); 1081 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1082 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1083 * struct nlattr *port[]); 1084 * 1085 * Enable or disable the VF ability to query its RSS Redirection Table and 1086 * Hash Key. This is needed since on some devices VF share this information 1087 * with PF and querying it may introduce a theoretical security risk. 1088 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1089 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1090 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1091 * void *type_data); 1092 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1093 * This is always called from the stack with the rtnl lock held and netif 1094 * tx queues stopped. This allows the netdevice to perform queue 1095 * management safely. 1096 * 1097 * Fiber Channel over Ethernet (FCoE) offload functions. 1098 * int (*ndo_fcoe_enable)(struct net_device *dev); 1099 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1100 * so the underlying device can perform whatever needed configuration or 1101 * initialization to support acceleration of FCoE traffic. 1102 * 1103 * int (*ndo_fcoe_disable)(struct net_device *dev); 1104 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1105 * so the underlying device can perform whatever needed clean-ups to 1106 * stop supporting acceleration of FCoE traffic. 1107 * 1108 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1109 * struct scatterlist *sgl, unsigned int sgc); 1110 * Called when the FCoE Initiator wants to initialize an I/O that 1111 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1112 * perform necessary setup and returns 1 to indicate the device is set up 1113 * successfully to perform DDP on this I/O, otherwise this returns 0. 1114 * 1115 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1116 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1117 * indicated by the FC exchange id 'xid', so the underlying device can 1118 * clean up and reuse resources for later DDP requests. 1119 * 1120 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1121 * struct scatterlist *sgl, unsigned int sgc); 1122 * Called when the FCoE Target wants to initialize an I/O that 1123 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1124 * perform necessary setup and returns 1 to indicate the device is set up 1125 * successfully to perform DDP on this I/O, otherwise this returns 0. 1126 * 1127 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1128 * struct netdev_fcoe_hbainfo *hbainfo); 1129 * Called when the FCoE Protocol stack wants information on the underlying 1130 * device. This information is utilized by the FCoE protocol stack to 1131 * register attributes with Fiber Channel management service as per the 1132 * FC-GS Fabric Device Management Information(FDMI) specification. 1133 * 1134 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1135 * Called when the underlying device wants to override default World Wide 1136 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1137 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1138 * protocol stack to use. 1139 * 1140 * RFS acceleration. 1141 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1142 * u16 rxq_index, u32 flow_id); 1143 * Set hardware filter for RFS. rxq_index is the target queue index; 1144 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1145 * Return the filter ID on success, or a negative error code. 1146 * 1147 * Slave management functions (for bridge, bonding, etc). 1148 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1149 * Called to make another netdev an underling. 1150 * 1151 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1152 * Called to release previously enslaved netdev. 1153 * 1154 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1155 * struct sk_buff *skb, 1156 * bool all_slaves); 1157 * Get the xmit slave of master device. If all_slaves is true, function 1158 * assume all the slaves can transmit. 1159 * 1160 * Feature/offload setting functions. 1161 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1162 * netdev_features_t features); 1163 * Adjusts the requested feature flags according to device-specific 1164 * constraints, and returns the resulting flags. Must not modify 1165 * the device state. 1166 * 1167 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1168 * Called to update device configuration to new features. Passed 1169 * feature set might be less than what was returned by ndo_fix_features()). 1170 * Must return >0 or -errno if it changed dev->features itself. 1171 * 1172 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1173 * struct net_device *dev, 1174 * const unsigned char *addr, u16 vid, u16 flags, 1175 * struct netlink_ext_ack *extack); 1176 * Adds an FDB entry to dev for addr. 1177 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1178 * struct net_device *dev, 1179 * const unsigned char *addr, u16 vid) 1180 * Deletes the FDB entry from dev coresponding to addr. 1181 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1182 * struct net_device *dev, struct net_device *filter_dev, 1183 * int *idx) 1184 * Used to add FDB entries to dump requests. Implementers should add 1185 * entries to skb and update idx with the number of entries. 1186 * 1187 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1188 * u16 flags, struct netlink_ext_ack *extack) 1189 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1190 * struct net_device *dev, u32 filter_mask, 1191 * int nlflags) 1192 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1193 * u16 flags); 1194 * 1195 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1196 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1197 * which do not represent real hardware may define this to allow their 1198 * userspace components to manage their virtual carrier state. Devices 1199 * that determine carrier state from physical hardware properties (eg 1200 * network cables) or protocol-dependent mechanisms (eg 1201 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1202 * 1203 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1204 * struct netdev_phys_item_id *ppid); 1205 * Called to get ID of physical port of this device. If driver does 1206 * not implement this, it is assumed that the hw is not able to have 1207 * multiple net devices on single physical port. 1208 * 1209 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1210 * struct netdev_phys_item_id *ppid) 1211 * Called to get the parent ID of the physical port of this device. 1212 * 1213 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1214 * struct udp_tunnel_info *ti); 1215 * Called by UDP tunnel to notify a driver about the UDP port and socket 1216 * address family that a UDP tunnel is listnening to. It is called only 1217 * when a new port starts listening. The operation is protected by the 1218 * RTNL. 1219 * 1220 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1221 * struct udp_tunnel_info *ti); 1222 * Called by UDP tunnel to notify the driver about a UDP port and socket 1223 * address family that the UDP tunnel is not listening to anymore. The 1224 * operation is protected by the RTNL. 1225 * 1226 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1227 * struct net_device *dev) 1228 * Called by upper layer devices to accelerate switching or other 1229 * station functionality into hardware. 'pdev is the lowerdev 1230 * to use for the offload and 'dev' is the net device that will 1231 * back the offload. Returns a pointer to the private structure 1232 * the upper layer will maintain. 1233 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1234 * Called by upper layer device to delete the station created 1235 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1236 * the station and priv is the structure returned by the add 1237 * operation. 1238 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1239 * int queue_index, u32 maxrate); 1240 * Called when a user wants to set a max-rate limitation of specific 1241 * TX queue. 1242 * int (*ndo_get_iflink)(const struct net_device *dev); 1243 * Called to get the iflink value of this device. 1244 * void (*ndo_change_proto_down)(struct net_device *dev, 1245 * bool proto_down); 1246 * This function is used to pass protocol port error state information 1247 * to the switch driver. The switch driver can react to the proto_down 1248 * by doing a phys down on the associated switch port. 1249 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1250 * This function is used to get egress tunnel information for given skb. 1251 * This is useful for retrieving outer tunnel header parameters while 1252 * sampling packet. 1253 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1254 * This function is used to specify the headroom that the skb must 1255 * consider when allocation skb during packet reception. Setting 1256 * appropriate rx headroom value allows avoiding skb head copy on 1257 * forward. Setting a negative value resets the rx headroom to the 1258 * default value. 1259 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1260 * This function is used to set or query state related to XDP on the 1261 * netdevice and manage BPF offload. See definition of 1262 * enum bpf_netdev_command for details. 1263 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1264 * u32 flags); 1265 * This function is used to submit @n XDP packets for transmit on a 1266 * netdevice. Returns number of frames successfully transmitted, frames 1267 * that got dropped are freed/returned via xdp_return_frame(). 1268 * Returns negative number, means general error invoking ndo, meaning 1269 * no frames were xmit'ed and core-caller will free all frames. 1270 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1271 * This function is used to wake up the softirq, ksoftirqd or kthread 1272 * responsible for sending and/or receiving packets on a specific 1273 * queue id bound to an AF_XDP socket. The flags field specifies if 1274 * only RX, only Tx, or both should be woken up using the flags 1275 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1276 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1277 * Get devlink port instance associated with a given netdev. 1278 * Called with a reference on the netdevice and devlink locks only, 1279 * rtnl_lock is not held. 1280 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1281 * int cmd); 1282 * Add, change, delete or get information on an IPv4 tunnel. 1283 */ 1284 struct net_device_ops { 1285 int (*ndo_init)(struct net_device *dev); 1286 void (*ndo_uninit)(struct net_device *dev); 1287 int (*ndo_open)(struct net_device *dev); 1288 int (*ndo_stop)(struct net_device *dev); 1289 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1290 struct net_device *dev); 1291 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1292 struct net_device *dev, 1293 netdev_features_t features); 1294 u16 (*ndo_select_queue)(struct net_device *dev, 1295 struct sk_buff *skb, 1296 struct net_device *sb_dev); 1297 void (*ndo_change_rx_flags)(struct net_device *dev, 1298 int flags); 1299 void (*ndo_set_rx_mode)(struct net_device *dev); 1300 int (*ndo_set_mac_address)(struct net_device *dev, 1301 void *addr); 1302 int (*ndo_validate_addr)(struct net_device *dev); 1303 int (*ndo_do_ioctl)(struct net_device *dev, 1304 struct ifreq *ifr, int cmd); 1305 int (*ndo_set_config)(struct net_device *dev, 1306 struct ifmap *map); 1307 int (*ndo_change_mtu)(struct net_device *dev, 1308 int new_mtu); 1309 int (*ndo_neigh_setup)(struct net_device *dev, 1310 struct neigh_parms *); 1311 void (*ndo_tx_timeout) (struct net_device *dev, 1312 unsigned int txqueue); 1313 1314 void (*ndo_get_stats64)(struct net_device *dev, 1315 struct rtnl_link_stats64 *storage); 1316 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1317 int (*ndo_get_offload_stats)(int attr_id, 1318 const struct net_device *dev, 1319 void *attr_data); 1320 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1321 1322 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1323 __be16 proto, u16 vid); 1324 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1325 __be16 proto, u16 vid); 1326 #ifdef CONFIG_NET_POLL_CONTROLLER 1327 void (*ndo_poll_controller)(struct net_device *dev); 1328 int (*ndo_netpoll_setup)(struct net_device *dev, 1329 struct netpoll_info *info); 1330 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1331 #endif 1332 int (*ndo_set_vf_mac)(struct net_device *dev, 1333 int queue, u8 *mac); 1334 int (*ndo_set_vf_vlan)(struct net_device *dev, 1335 int queue, u16 vlan, 1336 u8 qos, __be16 proto); 1337 int (*ndo_set_vf_rate)(struct net_device *dev, 1338 int vf, int min_tx_rate, 1339 int max_tx_rate); 1340 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1341 int vf, bool setting); 1342 int (*ndo_set_vf_trust)(struct net_device *dev, 1343 int vf, bool setting); 1344 int (*ndo_get_vf_config)(struct net_device *dev, 1345 int vf, 1346 struct ifla_vf_info *ivf); 1347 int (*ndo_set_vf_link_state)(struct net_device *dev, 1348 int vf, int link_state); 1349 int (*ndo_get_vf_stats)(struct net_device *dev, 1350 int vf, 1351 struct ifla_vf_stats 1352 *vf_stats); 1353 int (*ndo_set_vf_port)(struct net_device *dev, 1354 int vf, 1355 struct nlattr *port[]); 1356 int (*ndo_get_vf_port)(struct net_device *dev, 1357 int vf, struct sk_buff *skb); 1358 int (*ndo_get_vf_guid)(struct net_device *dev, 1359 int vf, 1360 struct ifla_vf_guid *node_guid, 1361 struct ifla_vf_guid *port_guid); 1362 int (*ndo_set_vf_guid)(struct net_device *dev, 1363 int vf, u64 guid, 1364 int guid_type); 1365 int (*ndo_set_vf_rss_query_en)( 1366 struct net_device *dev, 1367 int vf, bool setting); 1368 int (*ndo_setup_tc)(struct net_device *dev, 1369 enum tc_setup_type type, 1370 void *type_data); 1371 #if IS_ENABLED(CONFIG_FCOE) 1372 int (*ndo_fcoe_enable)(struct net_device *dev); 1373 int (*ndo_fcoe_disable)(struct net_device *dev); 1374 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1375 u16 xid, 1376 struct scatterlist *sgl, 1377 unsigned int sgc); 1378 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1379 u16 xid); 1380 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1381 u16 xid, 1382 struct scatterlist *sgl, 1383 unsigned int sgc); 1384 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1385 struct netdev_fcoe_hbainfo *hbainfo); 1386 #endif 1387 1388 #if IS_ENABLED(CONFIG_LIBFCOE) 1389 #define NETDEV_FCOE_WWNN 0 1390 #define NETDEV_FCOE_WWPN 1 1391 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1392 u64 *wwn, int type); 1393 #endif 1394 1395 #ifdef CONFIG_RFS_ACCEL 1396 int (*ndo_rx_flow_steer)(struct net_device *dev, 1397 const struct sk_buff *skb, 1398 u16 rxq_index, 1399 u32 flow_id); 1400 #endif 1401 int (*ndo_add_slave)(struct net_device *dev, 1402 struct net_device *slave_dev, 1403 struct netlink_ext_ack *extack); 1404 int (*ndo_del_slave)(struct net_device *dev, 1405 struct net_device *slave_dev); 1406 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1407 struct sk_buff *skb, 1408 bool all_slaves); 1409 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1410 netdev_features_t features); 1411 int (*ndo_set_features)(struct net_device *dev, 1412 netdev_features_t features); 1413 int (*ndo_neigh_construct)(struct net_device *dev, 1414 struct neighbour *n); 1415 void (*ndo_neigh_destroy)(struct net_device *dev, 1416 struct neighbour *n); 1417 1418 int (*ndo_fdb_add)(struct ndmsg *ndm, 1419 struct nlattr *tb[], 1420 struct net_device *dev, 1421 const unsigned char *addr, 1422 u16 vid, 1423 u16 flags, 1424 struct netlink_ext_ack *extack); 1425 int (*ndo_fdb_del)(struct ndmsg *ndm, 1426 struct nlattr *tb[], 1427 struct net_device *dev, 1428 const unsigned char *addr, 1429 u16 vid); 1430 int (*ndo_fdb_dump)(struct sk_buff *skb, 1431 struct netlink_callback *cb, 1432 struct net_device *dev, 1433 struct net_device *filter_dev, 1434 int *idx); 1435 int (*ndo_fdb_get)(struct sk_buff *skb, 1436 struct nlattr *tb[], 1437 struct net_device *dev, 1438 const unsigned char *addr, 1439 u16 vid, u32 portid, u32 seq, 1440 struct netlink_ext_ack *extack); 1441 int (*ndo_bridge_setlink)(struct net_device *dev, 1442 struct nlmsghdr *nlh, 1443 u16 flags, 1444 struct netlink_ext_ack *extack); 1445 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1446 u32 pid, u32 seq, 1447 struct net_device *dev, 1448 u32 filter_mask, 1449 int nlflags); 1450 int (*ndo_bridge_dellink)(struct net_device *dev, 1451 struct nlmsghdr *nlh, 1452 u16 flags); 1453 int (*ndo_change_carrier)(struct net_device *dev, 1454 bool new_carrier); 1455 int (*ndo_get_phys_port_id)(struct net_device *dev, 1456 struct netdev_phys_item_id *ppid); 1457 int (*ndo_get_port_parent_id)(struct net_device *dev, 1458 struct netdev_phys_item_id *ppid); 1459 int (*ndo_get_phys_port_name)(struct net_device *dev, 1460 char *name, size_t len); 1461 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1462 struct udp_tunnel_info *ti); 1463 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1464 struct udp_tunnel_info *ti); 1465 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1466 struct net_device *dev); 1467 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1468 void *priv); 1469 1470 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1471 int queue_index, 1472 u32 maxrate); 1473 int (*ndo_get_iflink)(const struct net_device *dev); 1474 int (*ndo_change_proto_down)(struct net_device *dev, 1475 bool proto_down); 1476 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1477 struct sk_buff *skb); 1478 void (*ndo_set_rx_headroom)(struct net_device *dev, 1479 int needed_headroom); 1480 int (*ndo_bpf)(struct net_device *dev, 1481 struct netdev_bpf *bpf); 1482 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1483 struct xdp_frame **xdp, 1484 u32 flags); 1485 int (*ndo_xsk_wakeup)(struct net_device *dev, 1486 u32 queue_id, u32 flags); 1487 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1488 int (*ndo_tunnel_ctl)(struct net_device *dev, 1489 struct ip_tunnel_parm *p, int cmd); 1490 }; 1491 1492 /** 1493 * enum net_device_priv_flags - &struct net_device priv_flags 1494 * 1495 * These are the &struct net_device, they are only set internally 1496 * by drivers and used in the kernel. These flags are invisible to 1497 * userspace; this means that the order of these flags can change 1498 * during any kernel release. 1499 * 1500 * You should have a pretty good reason to be extending these flags. 1501 * 1502 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1503 * @IFF_EBRIDGE: Ethernet bridging device 1504 * @IFF_BONDING: bonding master or slave 1505 * @IFF_ISATAP: ISATAP interface (RFC4214) 1506 * @IFF_WAN_HDLC: WAN HDLC device 1507 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1508 * release skb->dst 1509 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1510 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1511 * @IFF_MACVLAN_PORT: device used as macvlan port 1512 * @IFF_BRIDGE_PORT: device used as bridge port 1513 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1514 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1515 * @IFF_UNICAST_FLT: Supports unicast filtering 1516 * @IFF_TEAM_PORT: device used as team port 1517 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1518 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1519 * change when it's running 1520 * @IFF_MACVLAN: Macvlan device 1521 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1522 * underlying stacked devices 1523 * @IFF_L3MDEV_MASTER: device is an L3 master device 1524 * @IFF_NO_QUEUE: device can run without qdisc attached 1525 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1526 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1527 * @IFF_TEAM: device is a team device 1528 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1529 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1530 * entity (i.e. the master device for bridged veth) 1531 * @IFF_MACSEC: device is a MACsec device 1532 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1533 * @IFF_FAILOVER: device is a failover master device 1534 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1535 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1536 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1537 */ 1538 enum netdev_priv_flags { 1539 IFF_802_1Q_VLAN = 1<<0, 1540 IFF_EBRIDGE = 1<<1, 1541 IFF_BONDING = 1<<2, 1542 IFF_ISATAP = 1<<3, 1543 IFF_WAN_HDLC = 1<<4, 1544 IFF_XMIT_DST_RELEASE = 1<<5, 1545 IFF_DONT_BRIDGE = 1<<6, 1546 IFF_DISABLE_NETPOLL = 1<<7, 1547 IFF_MACVLAN_PORT = 1<<8, 1548 IFF_BRIDGE_PORT = 1<<9, 1549 IFF_OVS_DATAPATH = 1<<10, 1550 IFF_TX_SKB_SHARING = 1<<11, 1551 IFF_UNICAST_FLT = 1<<12, 1552 IFF_TEAM_PORT = 1<<13, 1553 IFF_SUPP_NOFCS = 1<<14, 1554 IFF_LIVE_ADDR_CHANGE = 1<<15, 1555 IFF_MACVLAN = 1<<16, 1556 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1557 IFF_L3MDEV_MASTER = 1<<18, 1558 IFF_NO_QUEUE = 1<<19, 1559 IFF_OPENVSWITCH = 1<<20, 1560 IFF_L3MDEV_SLAVE = 1<<21, 1561 IFF_TEAM = 1<<22, 1562 IFF_RXFH_CONFIGURED = 1<<23, 1563 IFF_PHONY_HEADROOM = 1<<24, 1564 IFF_MACSEC = 1<<25, 1565 IFF_NO_RX_HANDLER = 1<<26, 1566 IFF_FAILOVER = 1<<27, 1567 IFF_FAILOVER_SLAVE = 1<<28, 1568 IFF_L3MDEV_RX_HANDLER = 1<<29, 1569 IFF_LIVE_RENAME_OK = 1<<30, 1570 }; 1571 1572 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1573 #define IFF_EBRIDGE IFF_EBRIDGE 1574 #define IFF_BONDING IFF_BONDING 1575 #define IFF_ISATAP IFF_ISATAP 1576 #define IFF_WAN_HDLC IFF_WAN_HDLC 1577 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1578 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1579 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1580 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1581 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1582 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1583 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1584 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1585 #define IFF_TEAM_PORT IFF_TEAM_PORT 1586 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1587 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1588 #define IFF_MACVLAN IFF_MACVLAN 1589 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1590 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1591 #define IFF_NO_QUEUE IFF_NO_QUEUE 1592 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1593 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1594 #define IFF_TEAM IFF_TEAM 1595 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1596 #define IFF_MACSEC IFF_MACSEC 1597 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1598 #define IFF_FAILOVER IFF_FAILOVER 1599 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1600 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1601 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1602 1603 /** 1604 * struct net_device - The DEVICE structure. 1605 * 1606 * Actually, this whole structure is a big mistake. It mixes I/O 1607 * data with strictly "high-level" data, and it has to know about 1608 * almost every data structure used in the INET module. 1609 * 1610 * @name: This is the first field of the "visible" part of this structure 1611 * (i.e. as seen by users in the "Space.c" file). It is the name 1612 * of the interface. 1613 * 1614 * @name_node: Name hashlist node 1615 * @ifalias: SNMP alias 1616 * @mem_end: Shared memory end 1617 * @mem_start: Shared memory start 1618 * @base_addr: Device I/O address 1619 * @irq: Device IRQ number 1620 * 1621 * @state: Generic network queuing layer state, see netdev_state_t 1622 * @dev_list: The global list of network devices 1623 * @napi_list: List entry used for polling NAPI devices 1624 * @unreg_list: List entry when we are unregistering the 1625 * device; see the function unregister_netdev 1626 * @close_list: List entry used when we are closing the device 1627 * @ptype_all: Device-specific packet handlers for all protocols 1628 * @ptype_specific: Device-specific, protocol-specific packet handlers 1629 * 1630 * @adj_list: Directly linked devices, like slaves for bonding 1631 * @features: Currently active device features 1632 * @hw_features: User-changeable features 1633 * 1634 * @wanted_features: User-requested features 1635 * @vlan_features: Mask of features inheritable by VLAN devices 1636 * 1637 * @hw_enc_features: Mask of features inherited by encapsulating devices 1638 * This field indicates what encapsulation 1639 * offloads the hardware is capable of doing, 1640 * and drivers will need to set them appropriately. 1641 * 1642 * @mpls_features: Mask of features inheritable by MPLS 1643 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1644 * 1645 * @ifindex: interface index 1646 * @group: The group the device belongs to 1647 * 1648 * @stats: Statistics struct, which was left as a legacy, use 1649 * rtnl_link_stats64 instead 1650 * 1651 * @rx_dropped: Dropped packets by core network, 1652 * do not use this in drivers 1653 * @tx_dropped: Dropped packets by core network, 1654 * do not use this in drivers 1655 * @rx_nohandler: nohandler dropped packets by core network on 1656 * inactive devices, do not use this in drivers 1657 * @carrier_up_count: Number of times the carrier has been up 1658 * @carrier_down_count: Number of times the carrier has been down 1659 * 1660 * @wireless_handlers: List of functions to handle Wireless Extensions, 1661 * instead of ioctl, 1662 * see <net/iw_handler.h> for details. 1663 * @wireless_data: Instance data managed by the core of wireless extensions 1664 * 1665 * @netdev_ops: Includes several pointers to callbacks, 1666 * if one wants to override the ndo_*() functions 1667 * @ethtool_ops: Management operations 1668 * @l3mdev_ops: Layer 3 master device operations 1669 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1670 * discovery handling. Necessary for e.g. 6LoWPAN. 1671 * @xfrmdev_ops: Transformation offload operations 1672 * @tlsdev_ops: Transport Layer Security offload operations 1673 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1674 * of Layer 2 headers. 1675 * 1676 * @flags: Interface flags (a la BSD) 1677 * @priv_flags: Like 'flags' but invisible to userspace, 1678 * see if.h for the definitions 1679 * @gflags: Global flags ( kept as legacy ) 1680 * @padded: How much padding added by alloc_netdev() 1681 * @operstate: RFC2863 operstate 1682 * @link_mode: Mapping policy to operstate 1683 * @if_port: Selectable AUI, TP, ... 1684 * @dma: DMA channel 1685 * @mtu: Interface MTU value 1686 * @min_mtu: Interface Minimum MTU value 1687 * @max_mtu: Interface Maximum MTU value 1688 * @type: Interface hardware type 1689 * @hard_header_len: Maximum hardware header length. 1690 * @min_header_len: Minimum hardware header length 1691 * 1692 * @needed_headroom: Extra headroom the hardware may need, but not in all 1693 * cases can this be guaranteed 1694 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1695 * cases can this be guaranteed. Some cases also use 1696 * LL_MAX_HEADER instead to allocate the skb 1697 * 1698 * interface address info: 1699 * 1700 * @perm_addr: Permanent hw address 1701 * @addr_assign_type: Hw address assignment type 1702 * @addr_len: Hardware address length 1703 * @upper_level: Maximum depth level of upper devices. 1704 * @lower_level: Maximum depth level of lower devices. 1705 * @neigh_priv_len: Used in neigh_alloc() 1706 * @dev_id: Used to differentiate devices that share 1707 * the same link layer address 1708 * @dev_port: Used to differentiate devices that share 1709 * the same function 1710 * @addr_list_lock: XXX: need comments on this one 1711 * @name_assign_type: network interface name assignment type 1712 * @uc_promisc: Counter that indicates promiscuous mode 1713 * has been enabled due to the need to listen to 1714 * additional unicast addresses in a device that 1715 * does not implement ndo_set_rx_mode() 1716 * @uc: unicast mac addresses 1717 * @mc: multicast mac addresses 1718 * @dev_addrs: list of device hw addresses 1719 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1720 * @promiscuity: Number of times the NIC is told to work in 1721 * promiscuous mode; if it becomes 0 the NIC will 1722 * exit promiscuous mode 1723 * @allmulti: Counter, enables or disables allmulticast mode 1724 * 1725 * @vlan_info: VLAN info 1726 * @dsa_ptr: dsa specific data 1727 * @tipc_ptr: TIPC specific data 1728 * @atalk_ptr: AppleTalk link 1729 * @ip_ptr: IPv4 specific data 1730 * @dn_ptr: DECnet specific data 1731 * @ip6_ptr: IPv6 specific data 1732 * @ax25_ptr: AX.25 specific data 1733 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1734 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1735 * device struct 1736 * @mpls_ptr: mpls_dev struct pointer 1737 * 1738 * @dev_addr: Hw address (before bcast, 1739 * because most packets are unicast) 1740 * 1741 * @_rx: Array of RX queues 1742 * @num_rx_queues: Number of RX queues 1743 * allocated at register_netdev() time 1744 * @real_num_rx_queues: Number of RX queues currently active in device 1745 * @xdp_prog: XDP sockets filter program pointer 1746 * @gro_flush_timeout: timeout for GRO layer in NAPI 1747 * 1748 * @rx_handler: handler for received packets 1749 * @rx_handler_data: XXX: need comments on this one 1750 * @miniq_ingress: ingress/clsact qdisc specific data for 1751 * ingress processing 1752 * @ingress_queue: XXX: need comments on this one 1753 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1754 * @broadcast: hw bcast address 1755 * 1756 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1757 * indexed by RX queue number. Assigned by driver. 1758 * This must only be set if the ndo_rx_flow_steer 1759 * operation is defined 1760 * @index_hlist: Device index hash chain 1761 * 1762 * @_tx: Array of TX queues 1763 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1764 * @real_num_tx_queues: Number of TX queues currently active in device 1765 * @qdisc: Root qdisc from userspace point of view 1766 * @tx_queue_len: Max frames per queue allowed 1767 * @tx_global_lock: XXX: need comments on this one 1768 * @xdp_bulkq: XDP device bulk queue 1769 * @xps_cpus_map: all CPUs map for XPS device 1770 * @xps_rxqs_map: all RXQs map for XPS device 1771 * 1772 * @xps_maps: XXX: need comments on this one 1773 * @miniq_egress: clsact qdisc specific data for 1774 * egress processing 1775 * @qdisc_hash: qdisc hash table 1776 * @watchdog_timeo: Represents the timeout that is used by 1777 * the watchdog (see dev_watchdog()) 1778 * @watchdog_timer: List of timers 1779 * 1780 * @pcpu_refcnt: Number of references to this device 1781 * @todo_list: Delayed register/unregister 1782 * @link_watch_list: XXX: need comments on this one 1783 * 1784 * @reg_state: Register/unregister state machine 1785 * @dismantle: Device is going to be freed 1786 * @rtnl_link_state: This enum represents the phases of creating 1787 * a new link 1788 * 1789 * @needs_free_netdev: Should unregister perform free_netdev? 1790 * @priv_destructor: Called from unregister 1791 * @npinfo: XXX: need comments on this one 1792 * @nd_net: Network namespace this network device is inside 1793 * 1794 * @ml_priv: Mid-layer private 1795 * @lstats: Loopback statistics 1796 * @tstats: Tunnel statistics 1797 * @dstats: Dummy statistics 1798 * @vstats: Virtual ethernet statistics 1799 * 1800 * @garp_port: GARP 1801 * @mrp_port: MRP 1802 * 1803 * @dev: Class/net/name entry 1804 * @sysfs_groups: Space for optional device, statistics and wireless 1805 * sysfs groups 1806 * 1807 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1808 * @rtnl_link_ops: Rtnl_link_ops 1809 * 1810 * @gso_max_size: Maximum size of generic segmentation offload 1811 * @gso_max_segs: Maximum number of segments that can be passed to the 1812 * NIC for GSO 1813 * 1814 * @dcbnl_ops: Data Center Bridging netlink ops 1815 * @num_tc: Number of traffic classes in the net device 1816 * @tc_to_txq: XXX: need comments on this one 1817 * @prio_tc_map: XXX: need comments on this one 1818 * 1819 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1820 * 1821 * @priomap: XXX: need comments on this one 1822 * @phydev: Physical device may attach itself 1823 * for hardware timestamping 1824 * @sfp_bus: attached &struct sfp_bus structure. 1825 * 1826 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1827 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1828 * 1829 * @proto_down: protocol port state information can be sent to the 1830 * switch driver and used to set the phys state of the 1831 * switch port. 1832 * 1833 * @wol_enabled: Wake-on-LAN is enabled 1834 * 1835 * @net_notifier_list: List of per-net netdev notifier block 1836 * that follow this device when it is moved 1837 * to another network namespace. 1838 * 1839 * @macsec_ops: MACsec offloading ops 1840 * 1841 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1842 * offload capabilities of the device 1843 * @udp_tunnel_nic: UDP tunnel offload state 1844 * 1845 * FIXME: cleanup struct net_device such that network protocol info 1846 * moves out. 1847 */ 1848 1849 struct net_device { 1850 char name[IFNAMSIZ]; 1851 struct netdev_name_node *name_node; 1852 struct dev_ifalias __rcu *ifalias; 1853 /* 1854 * I/O specific fields 1855 * FIXME: Merge these and struct ifmap into one 1856 */ 1857 unsigned long mem_end; 1858 unsigned long mem_start; 1859 unsigned long base_addr; 1860 int irq; 1861 1862 /* 1863 * Some hardware also needs these fields (state,dev_list, 1864 * napi_list,unreg_list,close_list) but they are not 1865 * part of the usual set specified in Space.c. 1866 */ 1867 1868 unsigned long state; 1869 1870 struct list_head dev_list; 1871 struct list_head napi_list; 1872 struct list_head unreg_list; 1873 struct list_head close_list; 1874 struct list_head ptype_all; 1875 struct list_head ptype_specific; 1876 1877 struct { 1878 struct list_head upper; 1879 struct list_head lower; 1880 } adj_list; 1881 1882 netdev_features_t features; 1883 netdev_features_t hw_features; 1884 netdev_features_t wanted_features; 1885 netdev_features_t vlan_features; 1886 netdev_features_t hw_enc_features; 1887 netdev_features_t mpls_features; 1888 netdev_features_t gso_partial_features; 1889 1890 int ifindex; 1891 int group; 1892 1893 struct net_device_stats stats; 1894 1895 atomic_long_t rx_dropped; 1896 atomic_long_t tx_dropped; 1897 atomic_long_t rx_nohandler; 1898 1899 /* Stats to monitor link on/off, flapping */ 1900 atomic_t carrier_up_count; 1901 atomic_t carrier_down_count; 1902 1903 #ifdef CONFIG_WIRELESS_EXT 1904 const struct iw_handler_def *wireless_handlers; 1905 struct iw_public_data *wireless_data; 1906 #endif 1907 const struct net_device_ops *netdev_ops; 1908 const struct ethtool_ops *ethtool_ops; 1909 #ifdef CONFIG_NET_L3_MASTER_DEV 1910 const struct l3mdev_ops *l3mdev_ops; 1911 #endif 1912 #if IS_ENABLED(CONFIG_IPV6) 1913 const struct ndisc_ops *ndisc_ops; 1914 #endif 1915 1916 #ifdef CONFIG_XFRM_OFFLOAD 1917 const struct xfrmdev_ops *xfrmdev_ops; 1918 #endif 1919 1920 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1921 const struct tlsdev_ops *tlsdev_ops; 1922 #endif 1923 1924 const struct header_ops *header_ops; 1925 1926 unsigned int flags; 1927 unsigned int priv_flags; 1928 1929 unsigned short gflags; 1930 unsigned short padded; 1931 1932 unsigned char operstate; 1933 unsigned char link_mode; 1934 1935 unsigned char if_port; 1936 unsigned char dma; 1937 1938 /* Note : dev->mtu is often read without holding a lock. 1939 * Writers usually hold RTNL. 1940 * It is recommended to use READ_ONCE() to annotate the reads, 1941 * and to use WRITE_ONCE() to annotate the writes. 1942 */ 1943 unsigned int mtu; 1944 unsigned int min_mtu; 1945 unsigned int max_mtu; 1946 unsigned short type; 1947 unsigned short hard_header_len; 1948 unsigned char min_header_len; 1949 1950 unsigned short needed_headroom; 1951 unsigned short needed_tailroom; 1952 1953 /* Interface address info. */ 1954 unsigned char perm_addr[MAX_ADDR_LEN]; 1955 unsigned char addr_assign_type; 1956 unsigned char addr_len; 1957 unsigned char upper_level; 1958 unsigned char lower_level; 1959 unsigned short neigh_priv_len; 1960 unsigned short dev_id; 1961 unsigned short dev_port; 1962 spinlock_t addr_list_lock; 1963 unsigned char name_assign_type; 1964 bool uc_promisc; 1965 struct netdev_hw_addr_list uc; 1966 struct netdev_hw_addr_list mc; 1967 struct netdev_hw_addr_list dev_addrs; 1968 1969 #ifdef CONFIG_SYSFS 1970 struct kset *queues_kset; 1971 #endif 1972 unsigned int promiscuity; 1973 unsigned int allmulti; 1974 1975 1976 /* Protocol-specific pointers */ 1977 1978 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1979 struct vlan_info __rcu *vlan_info; 1980 #endif 1981 #if IS_ENABLED(CONFIG_NET_DSA) 1982 struct dsa_port *dsa_ptr; 1983 #endif 1984 #if IS_ENABLED(CONFIG_TIPC) 1985 struct tipc_bearer __rcu *tipc_ptr; 1986 #endif 1987 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1988 void *atalk_ptr; 1989 #endif 1990 struct in_device __rcu *ip_ptr; 1991 #if IS_ENABLED(CONFIG_DECNET) 1992 struct dn_dev __rcu *dn_ptr; 1993 #endif 1994 struct inet6_dev __rcu *ip6_ptr; 1995 #if IS_ENABLED(CONFIG_AX25) 1996 void *ax25_ptr; 1997 #endif 1998 struct wireless_dev *ieee80211_ptr; 1999 struct wpan_dev *ieee802154_ptr; 2000 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2001 struct mpls_dev __rcu *mpls_ptr; 2002 #endif 2003 2004 /* 2005 * Cache lines mostly used on receive path (including eth_type_trans()) 2006 */ 2007 /* Interface address info used in eth_type_trans() */ 2008 unsigned char *dev_addr; 2009 2010 struct netdev_rx_queue *_rx; 2011 unsigned int num_rx_queues; 2012 unsigned int real_num_rx_queues; 2013 2014 struct bpf_prog __rcu *xdp_prog; 2015 unsigned long gro_flush_timeout; 2016 int napi_defer_hard_irqs; 2017 rx_handler_func_t __rcu *rx_handler; 2018 void __rcu *rx_handler_data; 2019 2020 #ifdef CONFIG_NET_CLS_ACT 2021 struct mini_Qdisc __rcu *miniq_ingress; 2022 #endif 2023 struct netdev_queue __rcu *ingress_queue; 2024 #ifdef CONFIG_NETFILTER_INGRESS 2025 struct nf_hook_entries __rcu *nf_hooks_ingress; 2026 #endif 2027 2028 unsigned char broadcast[MAX_ADDR_LEN]; 2029 #ifdef CONFIG_RFS_ACCEL 2030 struct cpu_rmap *rx_cpu_rmap; 2031 #endif 2032 struct hlist_node index_hlist; 2033 2034 /* 2035 * Cache lines mostly used on transmit path 2036 */ 2037 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2038 unsigned int num_tx_queues; 2039 unsigned int real_num_tx_queues; 2040 struct Qdisc *qdisc; 2041 unsigned int tx_queue_len; 2042 spinlock_t tx_global_lock; 2043 2044 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2045 2046 #ifdef CONFIG_XPS 2047 struct xps_dev_maps __rcu *xps_cpus_map; 2048 struct xps_dev_maps __rcu *xps_rxqs_map; 2049 #endif 2050 #ifdef CONFIG_NET_CLS_ACT 2051 struct mini_Qdisc __rcu *miniq_egress; 2052 #endif 2053 2054 #ifdef CONFIG_NET_SCHED 2055 DECLARE_HASHTABLE (qdisc_hash, 4); 2056 #endif 2057 /* These may be needed for future network-power-down code. */ 2058 struct timer_list watchdog_timer; 2059 int watchdog_timeo; 2060 2061 struct list_head todo_list; 2062 int __percpu *pcpu_refcnt; 2063 2064 struct list_head link_watch_list; 2065 2066 enum { NETREG_UNINITIALIZED=0, 2067 NETREG_REGISTERED, /* completed register_netdevice */ 2068 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2069 NETREG_UNREGISTERED, /* completed unregister todo */ 2070 NETREG_RELEASED, /* called free_netdev */ 2071 NETREG_DUMMY, /* dummy device for NAPI poll */ 2072 } reg_state:8; 2073 2074 bool dismantle; 2075 2076 enum { 2077 RTNL_LINK_INITIALIZED, 2078 RTNL_LINK_INITIALIZING, 2079 } rtnl_link_state:16; 2080 2081 bool needs_free_netdev; 2082 void (*priv_destructor)(struct net_device *dev); 2083 2084 #ifdef CONFIG_NETPOLL 2085 struct netpoll_info __rcu *npinfo; 2086 #endif 2087 2088 possible_net_t nd_net; 2089 2090 /* mid-layer private */ 2091 union { 2092 void *ml_priv; 2093 struct pcpu_lstats __percpu *lstats; 2094 struct pcpu_sw_netstats __percpu *tstats; 2095 struct pcpu_dstats __percpu *dstats; 2096 }; 2097 2098 #if IS_ENABLED(CONFIG_GARP) 2099 struct garp_port __rcu *garp_port; 2100 #endif 2101 #if IS_ENABLED(CONFIG_MRP) 2102 struct mrp_port __rcu *mrp_port; 2103 #endif 2104 2105 struct device dev; 2106 const struct attribute_group *sysfs_groups[4]; 2107 const struct attribute_group *sysfs_rx_queue_group; 2108 2109 const struct rtnl_link_ops *rtnl_link_ops; 2110 2111 /* for setting kernel sock attribute on TCP connection setup */ 2112 #define GSO_MAX_SIZE 65536 2113 unsigned int gso_max_size; 2114 #define GSO_MAX_SEGS 65535 2115 u16 gso_max_segs; 2116 2117 #ifdef CONFIG_DCB 2118 const struct dcbnl_rtnl_ops *dcbnl_ops; 2119 #endif 2120 s16 num_tc; 2121 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2122 u8 prio_tc_map[TC_BITMASK + 1]; 2123 2124 #if IS_ENABLED(CONFIG_FCOE) 2125 unsigned int fcoe_ddp_xid; 2126 #endif 2127 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2128 struct netprio_map __rcu *priomap; 2129 #endif 2130 struct phy_device *phydev; 2131 struct sfp_bus *sfp_bus; 2132 struct lock_class_key *qdisc_tx_busylock; 2133 struct lock_class_key *qdisc_running_key; 2134 bool proto_down; 2135 unsigned wol_enabled:1; 2136 2137 struct list_head net_notifier_list; 2138 2139 #if IS_ENABLED(CONFIG_MACSEC) 2140 /* MACsec management functions */ 2141 const struct macsec_ops *macsec_ops; 2142 #endif 2143 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2144 struct udp_tunnel_nic *udp_tunnel_nic; 2145 }; 2146 #define to_net_dev(d) container_of(d, struct net_device, dev) 2147 2148 static inline bool netif_elide_gro(const struct net_device *dev) 2149 { 2150 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2151 return true; 2152 return false; 2153 } 2154 2155 #define NETDEV_ALIGN 32 2156 2157 static inline 2158 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2159 { 2160 return dev->prio_tc_map[prio & TC_BITMASK]; 2161 } 2162 2163 static inline 2164 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2165 { 2166 if (tc >= dev->num_tc) 2167 return -EINVAL; 2168 2169 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2170 return 0; 2171 } 2172 2173 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2174 void netdev_reset_tc(struct net_device *dev); 2175 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2176 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2177 2178 static inline 2179 int netdev_get_num_tc(struct net_device *dev) 2180 { 2181 return dev->num_tc; 2182 } 2183 2184 void netdev_unbind_sb_channel(struct net_device *dev, 2185 struct net_device *sb_dev); 2186 int netdev_bind_sb_channel_queue(struct net_device *dev, 2187 struct net_device *sb_dev, 2188 u8 tc, u16 count, u16 offset); 2189 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2190 static inline int netdev_get_sb_channel(struct net_device *dev) 2191 { 2192 return max_t(int, -dev->num_tc, 0); 2193 } 2194 2195 static inline 2196 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2197 unsigned int index) 2198 { 2199 return &dev->_tx[index]; 2200 } 2201 2202 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2203 const struct sk_buff *skb) 2204 { 2205 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2206 } 2207 2208 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2209 void (*f)(struct net_device *, 2210 struct netdev_queue *, 2211 void *), 2212 void *arg) 2213 { 2214 unsigned int i; 2215 2216 for (i = 0; i < dev->num_tx_queues; i++) 2217 f(dev, &dev->_tx[i], arg); 2218 } 2219 2220 #define netdev_lockdep_set_classes(dev) \ 2221 { \ 2222 static struct lock_class_key qdisc_tx_busylock_key; \ 2223 static struct lock_class_key qdisc_running_key; \ 2224 static struct lock_class_key qdisc_xmit_lock_key; \ 2225 static struct lock_class_key dev_addr_list_lock_key; \ 2226 unsigned int i; \ 2227 \ 2228 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2229 (dev)->qdisc_running_key = &qdisc_running_key; \ 2230 lockdep_set_class(&(dev)->addr_list_lock, \ 2231 &dev_addr_list_lock_key); \ 2232 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2233 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2234 &qdisc_xmit_lock_key); \ 2235 } 2236 2237 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2238 struct net_device *sb_dev); 2239 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2240 struct sk_buff *skb, 2241 struct net_device *sb_dev); 2242 2243 /* returns the headroom that the master device needs to take in account 2244 * when forwarding to this dev 2245 */ 2246 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2247 { 2248 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2249 } 2250 2251 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2252 { 2253 if (dev->netdev_ops->ndo_set_rx_headroom) 2254 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2255 } 2256 2257 /* set the device rx headroom to the dev's default */ 2258 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2259 { 2260 netdev_set_rx_headroom(dev, -1); 2261 } 2262 2263 /* 2264 * Net namespace inlines 2265 */ 2266 static inline 2267 struct net *dev_net(const struct net_device *dev) 2268 { 2269 return read_pnet(&dev->nd_net); 2270 } 2271 2272 static inline 2273 void dev_net_set(struct net_device *dev, struct net *net) 2274 { 2275 write_pnet(&dev->nd_net, net); 2276 } 2277 2278 /** 2279 * netdev_priv - access network device private data 2280 * @dev: network device 2281 * 2282 * Get network device private data 2283 */ 2284 static inline void *netdev_priv(const struct net_device *dev) 2285 { 2286 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2287 } 2288 2289 /* Set the sysfs physical device reference for the network logical device 2290 * if set prior to registration will cause a symlink during initialization. 2291 */ 2292 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2293 2294 /* Set the sysfs device type for the network logical device to allow 2295 * fine-grained identification of different network device types. For 2296 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2297 */ 2298 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2299 2300 /* Default NAPI poll() weight 2301 * Device drivers are strongly advised to not use bigger value 2302 */ 2303 #define NAPI_POLL_WEIGHT 64 2304 2305 /** 2306 * netif_napi_add - initialize a NAPI context 2307 * @dev: network device 2308 * @napi: NAPI context 2309 * @poll: polling function 2310 * @weight: default weight 2311 * 2312 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2313 * *any* of the other NAPI-related functions. 2314 */ 2315 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2316 int (*poll)(struct napi_struct *, int), int weight); 2317 2318 /** 2319 * netif_tx_napi_add - initialize a NAPI context 2320 * @dev: network device 2321 * @napi: NAPI context 2322 * @poll: polling function 2323 * @weight: default weight 2324 * 2325 * This variant of netif_napi_add() should be used from drivers using NAPI 2326 * to exclusively poll a TX queue. 2327 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2328 */ 2329 static inline void netif_tx_napi_add(struct net_device *dev, 2330 struct napi_struct *napi, 2331 int (*poll)(struct napi_struct *, int), 2332 int weight) 2333 { 2334 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2335 netif_napi_add(dev, napi, poll, weight); 2336 } 2337 2338 /** 2339 * netif_napi_del - remove a NAPI context 2340 * @napi: NAPI context 2341 * 2342 * netif_napi_del() removes a NAPI context from the network device NAPI list 2343 */ 2344 void netif_napi_del(struct napi_struct *napi); 2345 2346 struct napi_gro_cb { 2347 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2348 void *frag0; 2349 2350 /* Length of frag0. */ 2351 unsigned int frag0_len; 2352 2353 /* This indicates where we are processing relative to skb->data. */ 2354 int data_offset; 2355 2356 /* This is non-zero if the packet cannot be merged with the new skb. */ 2357 u16 flush; 2358 2359 /* Save the IP ID here and check when we get to the transport layer */ 2360 u16 flush_id; 2361 2362 /* Number of segments aggregated. */ 2363 u16 count; 2364 2365 /* Start offset for remote checksum offload */ 2366 u16 gro_remcsum_start; 2367 2368 /* jiffies when first packet was created/queued */ 2369 unsigned long age; 2370 2371 /* Used in ipv6_gro_receive() and foo-over-udp */ 2372 u16 proto; 2373 2374 /* This is non-zero if the packet may be of the same flow. */ 2375 u8 same_flow:1; 2376 2377 /* Used in tunnel GRO receive */ 2378 u8 encap_mark:1; 2379 2380 /* GRO checksum is valid */ 2381 u8 csum_valid:1; 2382 2383 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2384 u8 csum_cnt:3; 2385 2386 /* Free the skb? */ 2387 u8 free:2; 2388 #define NAPI_GRO_FREE 1 2389 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2390 2391 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2392 u8 is_ipv6:1; 2393 2394 /* Used in GRE, set in fou/gue_gro_receive */ 2395 u8 is_fou:1; 2396 2397 /* Used to determine if flush_id can be ignored */ 2398 u8 is_atomic:1; 2399 2400 /* Number of gro_receive callbacks this packet already went through */ 2401 u8 recursion_counter:4; 2402 2403 /* GRO is done by frag_list pointer chaining. */ 2404 u8 is_flist:1; 2405 2406 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2407 __wsum csum; 2408 2409 /* used in skb_gro_receive() slow path */ 2410 struct sk_buff *last; 2411 }; 2412 2413 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2414 2415 #define GRO_RECURSION_LIMIT 15 2416 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2417 { 2418 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2419 } 2420 2421 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2422 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2423 struct list_head *head, 2424 struct sk_buff *skb) 2425 { 2426 if (unlikely(gro_recursion_inc_test(skb))) { 2427 NAPI_GRO_CB(skb)->flush |= 1; 2428 return NULL; 2429 } 2430 2431 return cb(head, skb); 2432 } 2433 2434 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2435 struct sk_buff *); 2436 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2437 struct sock *sk, 2438 struct list_head *head, 2439 struct sk_buff *skb) 2440 { 2441 if (unlikely(gro_recursion_inc_test(skb))) { 2442 NAPI_GRO_CB(skb)->flush |= 1; 2443 return NULL; 2444 } 2445 2446 return cb(sk, head, skb); 2447 } 2448 2449 struct packet_type { 2450 __be16 type; /* This is really htons(ether_type). */ 2451 bool ignore_outgoing; 2452 struct net_device *dev; /* NULL is wildcarded here */ 2453 int (*func) (struct sk_buff *, 2454 struct net_device *, 2455 struct packet_type *, 2456 struct net_device *); 2457 void (*list_func) (struct list_head *, 2458 struct packet_type *, 2459 struct net_device *); 2460 bool (*id_match)(struct packet_type *ptype, 2461 struct sock *sk); 2462 void *af_packet_priv; 2463 struct list_head list; 2464 }; 2465 2466 struct offload_callbacks { 2467 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2468 netdev_features_t features); 2469 struct sk_buff *(*gro_receive)(struct list_head *head, 2470 struct sk_buff *skb); 2471 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2472 }; 2473 2474 struct packet_offload { 2475 __be16 type; /* This is really htons(ether_type). */ 2476 u16 priority; 2477 struct offload_callbacks callbacks; 2478 struct list_head list; 2479 }; 2480 2481 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2482 struct pcpu_sw_netstats { 2483 u64 rx_packets; 2484 u64 rx_bytes; 2485 u64 tx_packets; 2486 u64 tx_bytes; 2487 struct u64_stats_sync syncp; 2488 } __aligned(4 * sizeof(u64)); 2489 2490 struct pcpu_lstats { 2491 u64_stats_t packets; 2492 u64_stats_t bytes; 2493 struct u64_stats_sync syncp; 2494 } __aligned(2 * sizeof(u64)); 2495 2496 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2497 2498 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2499 { 2500 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2501 2502 u64_stats_update_begin(&lstats->syncp); 2503 u64_stats_add(&lstats->bytes, len); 2504 u64_stats_inc(&lstats->packets); 2505 u64_stats_update_end(&lstats->syncp); 2506 } 2507 2508 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2509 ({ \ 2510 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2511 if (pcpu_stats) { \ 2512 int __cpu; \ 2513 for_each_possible_cpu(__cpu) { \ 2514 typeof(type) *stat; \ 2515 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2516 u64_stats_init(&stat->syncp); \ 2517 } \ 2518 } \ 2519 pcpu_stats; \ 2520 }) 2521 2522 #define netdev_alloc_pcpu_stats(type) \ 2523 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2524 2525 enum netdev_lag_tx_type { 2526 NETDEV_LAG_TX_TYPE_UNKNOWN, 2527 NETDEV_LAG_TX_TYPE_RANDOM, 2528 NETDEV_LAG_TX_TYPE_BROADCAST, 2529 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2530 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2531 NETDEV_LAG_TX_TYPE_HASH, 2532 }; 2533 2534 enum netdev_lag_hash { 2535 NETDEV_LAG_HASH_NONE, 2536 NETDEV_LAG_HASH_L2, 2537 NETDEV_LAG_HASH_L34, 2538 NETDEV_LAG_HASH_L23, 2539 NETDEV_LAG_HASH_E23, 2540 NETDEV_LAG_HASH_E34, 2541 NETDEV_LAG_HASH_UNKNOWN, 2542 }; 2543 2544 struct netdev_lag_upper_info { 2545 enum netdev_lag_tx_type tx_type; 2546 enum netdev_lag_hash hash_type; 2547 }; 2548 2549 struct netdev_lag_lower_state_info { 2550 u8 link_up : 1, 2551 tx_enabled : 1; 2552 }; 2553 2554 #include <linux/notifier.h> 2555 2556 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2557 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2558 * adding new types. 2559 */ 2560 enum netdev_cmd { 2561 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2562 NETDEV_DOWN, 2563 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2564 detected a hardware crash and restarted 2565 - we can use this eg to kick tcp sessions 2566 once done */ 2567 NETDEV_CHANGE, /* Notify device state change */ 2568 NETDEV_REGISTER, 2569 NETDEV_UNREGISTER, 2570 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2571 NETDEV_CHANGEADDR, /* notify after the address change */ 2572 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2573 NETDEV_GOING_DOWN, 2574 NETDEV_CHANGENAME, 2575 NETDEV_FEAT_CHANGE, 2576 NETDEV_BONDING_FAILOVER, 2577 NETDEV_PRE_UP, 2578 NETDEV_PRE_TYPE_CHANGE, 2579 NETDEV_POST_TYPE_CHANGE, 2580 NETDEV_POST_INIT, 2581 NETDEV_RELEASE, 2582 NETDEV_NOTIFY_PEERS, 2583 NETDEV_JOIN, 2584 NETDEV_CHANGEUPPER, 2585 NETDEV_RESEND_IGMP, 2586 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2587 NETDEV_CHANGEINFODATA, 2588 NETDEV_BONDING_INFO, 2589 NETDEV_PRECHANGEUPPER, 2590 NETDEV_CHANGELOWERSTATE, 2591 NETDEV_UDP_TUNNEL_PUSH_INFO, 2592 NETDEV_UDP_TUNNEL_DROP_INFO, 2593 NETDEV_CHANGE_TX_QUEUE_LEN, 2594 NETDEV_CVLAN_FILTER_PUSH_INFO, 2595 NETDEV_CVLAN_FILTER_DROP_INFO, 2596 NETDEV_SVLAN_FILTER_PUSH_INFO, 2597 NETDEV_SVLAN_FILTER_DROP_INFO, 2598 }; 2599 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2600 2601 int register_netdevice_notifier(struct notifier_block *nb); 2602 int unregister_netdevice_notifier(struct notifier_block *nb); 2603 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2604 int unregister_netdevice_notifier_net(struct net *net, 2605 struct notifier_block *nb); 2606 int register_netdevice_notifier_dev_net(struct net_device *dev, 2607 struct notifier_block *nb, 2608 struct netdev_net_notifier *nn); 2609 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2610 struct notifier_block *nb, 2611 struct netdev_net_notifier *nn); 2612 2613 struct netdev_notifier_info { 2614 struct net_device *dev; 2615 struct netlink_ext_ack *extack; 2616 }; 2617 2618 struct netdev_notifier_info_ext { 2619 struct netdev_notifier_info info; /* must be first */ 2620 union { 2621 u32 mtu; 2622 } ext; 2623 }; 2624 2625 struct netdev_notifier_change_info { 2626 struct netdev_notifier_info info; /* must be first */ 2627 unsigned int flags_changed; 2628 }; 2629 2630 struct netdev_notifier_changeupper_info { 2631 struct netdev_notifier_info info; /* must be first */ 2632 struct net_device *upper_dev; /* new upper dev */ 2633 bool master; /* is upper dev master */ 2634 bool linking; /* is the notification for link or unlink */ 2635 void *upper_info; /* upper dev info */ 2636 }; 2637 2638 struct netdev_notifier_changelowerstate_info { 2639 struct netdev_notifier_info info; /* must be first */ 2640 void *lower_state_info; /* is lower dev state */ 2641 }; 2642 2643 struct netdev_notifier_pre_changeaddr_info { 2644 struct netdev_notifier_info info; /* must be first */ 2645 const unsigned char *dev_addr; 2646 }; 2647 2648 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2649 struct net_device *dev) 2650 { 2651 info->dev = dev; 2652 info->extack = NULL; 2653 } 2654 2655 static inline struct net_device * 2656 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2657 { 2658 return info->dev; 2659 } 2660 2661 static inline struct netlink_ext_ack * 2662 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2663 { 2664 return info->extack; 2665 } 2666 2667 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2668 2669 2670 extern rwlock_t dev_base_lock; /* Device list lock */ 2671 2672 #define for_each_netdev(net, d) \ 2673 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2674 #define for_each_netdev_reverse(net, d) \ 2675 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2676 #define for_each_netdev_rcu(net, d) \ 2677 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2678 #define for_each_netdev_safe(net, d, n) \ 2679 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2680 #define for_each_netdev_continue(net, d) \ 2681 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2682 #define for_each_netdev_continue_reverse(net, d) \ 2683 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2684 dev_list) 2685 #define for_each_netdev_continue_rcu(net, d) \ 2686 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2687 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2688 for_each_netdev_rcu(&init_net, slave) \ 2689 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2690 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2691 2692 static inline struct net_device *next_net_device(struct net_device *dev) 2693 { 2694 struct list_head *lh; 2695 struct net *net; 2696 2697 net = dev_net(dev); 2698 lh = dev->dev_list.next; 2699 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2700 } 2701 2702 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2703 { 2704 struct list_head *lh; 2705 struct net *net; 2706 2707 net = dev_net(dev); 2708 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2709 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2710 } 2711 2712 static inline struct net_device *first_net_device(struct net *net) 2713 { 2714 return list_empty(&net->dev_base_head) ? NULL : 2715 net_device_entry(net->dev_base_head.next); 2716 } 2717 2718 static inline struct net_device *first_net_device_rcu(struct net *net) 2719 { 2720 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2721 2722 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2723 } 2724 2725 int netdev_boot_setup_check(struct net_device *dev); 2726 unsigned long netdev_boot_base(const char *prefix, int unit); 2727 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2728 const char *hwaddr); 2729 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2730 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2731 void dev_add_pack(struct packet_type *pt); 2732 void dev_remove_pack(struct packet_type *pt); 2733 void __dev_remove_pack(struct packet_type *pt); 2734 void dev_add_offload(struct packet_offload *po); 2735 void dev_remove_offload(struct packet_offload *po); 2736 2737 int dev_get_iflink(const struct net_device *dev); 2738 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2739 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2740 unsigned short mask); 2741 struct net_device *dev_get_by_name(struct net *net, const char *name); 2742 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2743 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2744 int dev_alloc_name(struct net_device *dev, const char *name); 2745 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2746 void dev_close(struct net_device *dev); 2747 void dev_close_many(struct list_head *head, bool unlink); 2748 void dev_disable_lro(struct net_device *dev); 2749 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2750 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2751 struct net_device *sb_dev); 2752 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2753 struct net_device *sb_dev); 2754 int dev_queue_xmit(struct sk_buff *skb); 2755 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2756 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2757 int register_netdevice(struct net_device *dev); 2758 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2759 void unregister_netdevice_many(struct list_head *head); 2760 static inline void unregister_netdevice(struct net_device *dev) 2761 { 2762 unregister_netdevice_queue(dev, NULL); 2763 } 2764 2765 int netdev_refcnt_read(const struct net_device *dev); 2766 void free_netdev(struct net_device *dev); 2767 void netdev_freemem(struct net_device *dev); 2768 void synchronize_net(void); 2769 int init_dummy_netdev(struct net_device *dev); 2770 2771 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2772 struct sk_buff *skb, 2773 bool all_slaves); 2774 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2775 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2776 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2777 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2778 int netdev_get_name(struct net *net, char *name, int ifindex); 2779 int dev_restart(struct net_device *dev); 2780 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2781 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2782 2783 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2784 { 2785 return NAPI_GRO_CB(skb)->data_offset; 2786 } 2787 2788 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2789 { 2790 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2791 } 2792 2793 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2794 { 2795 NAPI_GRO_CB(skb)->data_offset += len; 2796 } 2797 2798 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2799 unsigned int offset) 2800 { 2801 return NAPI_GRO_CB(skb)->frag0 + offset; 2802 } 2803 2804 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2805 { 2806 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2807 } 2808 2809 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2810 { 2811 NAPI_GRO_CB(skb)->frag0 = NULL; 2812 NAPI_GRO_CB(skb)->frag0_len = 0; 2813 } 2814 2815 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2816 unsigned int offset) 2817 { 2818 if (!pskb_may_pull(skb, hlen)) 2819 return NULL; 2820 2821 skb_gro_frag0_invalidate(skb); 2822 return skb->data + offset; 2823 } 2824 2825 static inline void *skb_gro_network_header(struct sk_buff *skb) 2826 { 2827 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2828 skb_network_offset(skb); 2829 } 2830 2831 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2832 const void *start, unsigned int len) 2833 { 2834 if (NAPI_GRO_CB(skb)->csum_valid) 2835 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2836 csum_partial(start, len, 0)); 2837 } 2838 2839 /* GRO checksum functions. These are logical equivalents of the normal 2840 * checksum functions (in skbuff.h) except that they operate on the GRO 2841 * offsets and fields in sk_buff. 2842 */ 2843 2844 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2845 2846 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2847 { 2848 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2849 } 2850 2851 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2852 bool zero_okay, 2853 __sum16 check) 2854 { 2855 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2856 skb_checksum_start_offset(skb) < 2857 skb_gro_offset(skb)) && 2858 !skb_at_gro_remcsum_start(skb) && 2859 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2860 (!zero_okay || check)); 2861 } 2862 2863 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2864 __wsum psum) 2865 { 2866 if (NAPI_GRO_CB(skb)->csum_valid && 2867 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2868 return 0; 2869 2870 NAPI_GRO_CB(skb)->csum = psum; 2871 2872 return __skb_gro_checksum_complete(skb); 2873 } 2874 2875 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2876 { 2877 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2878 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2879 NAPI_GRO_CB(skb)->csum_cnt--; 2880 } else { 2881 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2882 * verified a new top level checksum or an encapsulated one 2883 * during GRO. This saves work if we fallback to normal path. 2884 */ 2885 __skb_incr_checksum_unnecessary(skb); 2886 } 2887 } 2888 2889 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2890 compute_pseudo) \ 2891 ({ \ 2892 __sum16 __ret = 0; \ 2893 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2894 __ret = __skb_gro_checksum_validate_complete(skb, \ 2895 compute_pseudo(skb, proto)); \ 2896 if (!__ret) \ 2897 skb_gro_incr_csum_unnecessary(skb); \ 2898 __ret; \ 2899 }) 2900 2901 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2902 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2903 2904 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2905 compute_pseudo) \ 2906 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2907 2908 #define skb_gro_checksum_simple_validate(skb) \ 2909 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2910 2911 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2912 { 2913 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2914 !NAPI_GRO_CB(skb)->csum_valid); 2915 } 2916 2917 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2918 __wsum pseudo) 2919 { 2920 NAPI_GRO_CB(skb)->csum = ~pseudo; 2921 NAPI_GRO_CB(skb)->csum_valid = 1; 2922 } 2923 2924 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2925 do { \ 2926 if (__skb_gro_checksum_convert_check(skb)) \ 2927 __skb_gro_checksum_convert(skb, \ 2928 compute_pseudo(skb, proto)); \ 2929 } while (0) 2930 2931 struct gro_remcsum { 2932 int offset; 2933 __wsum delta; 2934 }; 2935 2936 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2937 { 2938 grc->offset = 0; 2939 grc->delta = 0; 2940 } 2941 2942 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2943 unsigned int off, size_t hdrlen, 2944 int start, int offset, 2945 struct gro_remcsum *grc, 2946 bool nopartial) 2947 { 2948 __wsum delta; 2949 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2950 2951 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2952 2953 if (!nopartial) { 2954 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2955 return ptr; 2956 } 2957 2958 ptr = skb_gro_header_fast(skb, off); 2959 if (skb_gro_header_hard(skb, off + plen)) { 2960 ptr = skb_gro_header_slow(skb, off + plen, off); 2961 if (!ptr) 2962 return NULL; 2963 } 2964 2965 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2966 start, offset); 2967 2968 /* Adjust skb->csum since we changed the packet */ 2969 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2970 2971 grc->offset = off + hdrlen + offset; 2972 grc->delta = delta; 2973 2974 return ptr; 2975 } 2976 2977 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2978 struct gro_remcsum *grc) 2979 { 2980 void *ptr; 2981 size_t plen = grc->offset + sizeof(u16); 2982 2983 if (!grc->delta) 2984 return; 2985 2986 ptr = skb_gro_header_fast(skb, grc->offset); 2987 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2988 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2989 if (!ptr) 2990 return; 2991 } 2992 2993 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2994 } 2995 2996 #ifdef CONFIG_XFRM_OFFLOAD 2997 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2998 { 2999 if (PTR_ERR(pp) != -EINPROGRESS) 3000 NAPI_GRO_CB(skb)->flush |= flush; 3001 } 3002 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3003 struct sk_buff *pp, 3004 int flush, 3005 struct gro_remcsum *grc) 3006 { 3007 if (PTR_ERR(pp) != -EINPROGRESS) { 3008 NAPI_GRO_CB(skb)->flush |= flush; 3009 skb_gro_remcsum_cleanup(skb, grc); 3010 skb->remcsum_offload = 0; 3011 } 3012 } 3013 #else 3014 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3015 { 3016 NAPI_GRO_CB(skb)->flush |= flush; 3017 } 3018 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3019 struct sk_buff *pp, 3020 int flush, 3021 struct gro_remcsum *grc) 3022 { 3023 NAPI_GRO_CB(skb)->flush |= flush; 3024 skb_gro_remcsum_cleanup(skb, grc); 3025 skb->remcsum_offload = 0; 3026 } 3027 #endif 3028 3029 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3030 unsigned short type, 3031 const void *daddr, const void *saddr, 3032 unsigned int len) 3033 { 3034 if (!dev->header_ops || !dev->header_ops->create) 3035 return 0; 3036 3037 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3038 } 3039 3040 static inline int dev_parse_header(const struct sk_buff *skb, 3041 unsigned char *haddr) 3042 { 3043 const struct net_device *dev = skb->dev; 3044 3045 if (!dev->header_ops || !dev->header_ops->parse) 3046 return 0; 3047 return dev->header_ops->parse(skb, haddr); 3048 } 3049 3050 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3051 { 3052 const struct net_device *dev = skb->dev; 3053 3054 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3055 return 0; 3056 return dev->header_ops->parse_protocol(skb); 3057 } 3058 3059 /* ll_header must have at least hard_header_len allocated */ 3060 static inline bool dev_validate_header(const struct net_device *dev, 3061 char *ll_header, int len) 3062 { 3063 if (likely(len >= dev->hard_header_len)) 3064 return true; 3065 if (len < dev->min_header_len) 3066 return false; 3067 3068 if (capable(CAP_SYS_RAWIO)) { 3069 memset(ll_header + len, 0, dev->hard_header_len - len); 3070 return true; 3071 } 3072 3073 if (dev->header_ops && dev->header_ops->validate) 3074 return dev->header_ops->validate(ll_header, len); 3075 3076 return false; 3077 } 3078 3079 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3080 int len, int size); 3081 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3082 static inline int unregister_gifconf(unsigned int family) 3083 { 3084 return register_gifconf(family, NULL); 3085 } 3086 3087 #ifdef CONFIG_NET_FLOW_LIMIT 3088 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3089 struct sd_flow_limit { 3090 u64 count; 3091 unsigned int num_buckets; 3092 unsigned int history_head; 3093 u16 history[FLOW_LIMIT_HISTORY]; 3094 u8 buckets[]; 3095 }; 3096 3097 extern int netdev_flow_limit_table_len; 3098 #endif /* CONFIG_NET_FLOW_LIMIT */ 3099 3100 /* 3101 * Incoming packets are placed on per-CPU queues 3102 */ 3103 struct softnet_data { 3104 struct list_head poll_list; 3105 struct sk_buff_head process_queue; 3106 3107 /* stats */ 3108 unsigned int processed; 3109 unsigned int time_squeeze; 3110 unsigned int received_rps; 3111 #ifdef CONFIG_RPS 3112 struct softnet_data *rps_ipi_list; 3113 #endif 3114 #ifdef CONFIG_NET_FLOW_LIMIT 3115 struct sd_flow_limit __rcu *flow_limit; 3116 #endif 3117 struct Qdisc *output_queue; 3118 struct Qdisc **output_queue_tailp; 3119 struct sk_buff *completion_queue; 3120 #ifdef CONFIG_XFRM_OFFLOAD 3121 struct sk_buff_head xfrm_backlog; 3122 #endif 3123 /* written and read only by owning cpu: */ 3124 struct { 3125 u16 recursion; 3126 u8 more; 3127 } xmit; 3128 #ifdef CONFIG_RPS 3129 /* input_queue_head should be written by cpu owning this struct, 3130 * and only read by other cpus. Worth using a cache line. 3131 */ 3132 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3133 3134 /* Elements below can be accessed between CPUs for RPS/RFS */ 3135 call_single_data_t csd ____cacheline_aligned_in_smp; 3136 struct softnet_data *rps_ipi_next; 3137 unsigned int cpu; 3138 unsigned int input_queue_tail; 3139 #endif 3140 unsigned int dropped; 3141 struct sk_buff_head input_pkt_queue; 3142 struct napi_struct backlog; 3143 3144 }; 3145 3146 static inline void input_queue_head_incr(struct softnet_data *sd) 3147 { 3148 #ifdef CONFIG_RPS 3149 sd->input_queue_head++; 3150 #endif 3151 } 3152 3153 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3154 unsigned int *qtail) 3155 { 3156 #ifdef CONFIG_RPS 3157 *qtail = ++sd->input_queue_tail; 3158 #endif 3159 } 3160 3161 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3162 3163 static inline int dev_recursion_level(void) 3164 { 3165 return this_cpu_read(softnet_data.xmit.recursion); 3166 } 3167 3168 #define XMIT_RECURSION_LIMIT 8 3169 static inline bool dev_xmit_recursion(void) 3170 { 3171 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3172 XMIT_RECURSION_LIMIT); 3173 } 3174 3175 static inline void dev_xmit_recursion_inc(void) 3176 { 3177 __this_cpu_inc(softnet_data.xmit.recursion); 3178 } 3179 3180 static inline void dev_xmit_recursion_dec(void) 3181 { 3182 __this_cpu_dec(softnet_data.xmit.recursion); 3183 } 3184 3185 void __netif_schedule(struct Qdisc *q); 3186 void netif_schedule_queue(struct netdev_queue *txq); 3187 3188 static inline void netif_tx_schedule_all(struct net_device *dev) 3189 { 3190 unsigned int i; 3191 3192 for (i = 0; i < dev->num_tx_queues; i++) 3193 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3194 } 3195 3196 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3197 { 3198 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3199 } 3200 3201 /** 3202 * netif_start_queue - allow transmit 3203 * @dev: network device 3204 * 3205 * Allow upper layers to call the device hard_start_xmit routine. 3206 */ 3207 static inline void netif_start_queue(struct net_device *dev) 3208 { 3209 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3210 } 3211 3212 static inline void netif_tx_start_all_queues(struct net_device *dev) 3213 { 3214 unsigned int i; 3215 3216 for (i = 0; i < dev->num_tx_queues; i++) { 3217 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3218 netif_tx_start_queue(txq); 3219 } 3220 } 3221 3222 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3223 3224 /** 3225 * netif_wake_queue - restart transmit 3226 * @dev: network device 3227 * 3228 * Allow upper layers to call the device hard_start_xmit routine. 3229 * Used for flow control when transmit resources are available. 3230 */ 3231 static inline void netif_wake_queue(struct net_device *dev) 3232 { 3233 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3234 } 3235 3236 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3237 { 3238 unsigned int i; 3239 3240 for (i = 0; i < dev->num_tx_queues; i++) { 3241 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3242 netif_tx_wake_queue(txq); 3243 } 3244 } 3245 3246 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3247 { 3248 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3249 } 3250 3251 /** 3252 * netif_stop_queue - stop transmitted packets 3253 * @dev: network device 3254 * 3255 * Stop upper layers calling the device hard_start_xmit routine. 3256 * Used for flow control when transmit resources are unavailable. 3257 */ 3258 static inline void netif_stop_queue(struct net_device *dev) 3259 { 3260 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3261 } 3262 3263 void netif_tx_stop_all_queues(struct net_device *dev); 3264 3265 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3266 { 3267 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3268 } 3269 3270 /** 3271 * netif_queue_stopped - test if transmit queue is flowblocked 3272 * @dev: network device 3273 * 3274 * Test if transmit queue on device is currently unable to send. 3275 */ 3276 static inline bool netif_queue_stopped(const struct net_device *dev) 3277 { 3278 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3279 } 3280 3281 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3282 { 3283 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3284 } 3285 3286 static inline bool 3287 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3288 { 3289 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3290 } 3291 3292 static inline bool 3293 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3294 { 3295 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3296 } 3297 3298 /** 3299 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3300 * @dev_queue: pointer to transmit queue 3301 * 3302 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3303 * to give appropriate hint to the CPU. 3304 */ 3305 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3306 { 3307 #ifdef CONFIG_BQL 3308 prefetchw(&dev_queue->dql.num_queued); 3309 #endif 3310 } 3311 3312 /** 3313 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3314 * @dev_queue: pointer to transmit queue 3315 * 3316 * BQL enabled drivers might use this helper in their TX completion path, 3317 * to give appropriate hint to the CPU. 3318 */ 3319 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3320 { 3321 #ifdef CONFIG_BQL 3322 prefetchw(&dev_queue->dql.limit); 3323 #endif 3324 } 3325 3326 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3327 unsigned int bytes) 3328 { 3329 #ifdef CONFIG_BQL 3330 dql_queued(&dev_queue->dql, bytes); 3331 3332 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3333 return; 3334 3335 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3336 3337 /* 3338 * The XOFF flag must be set before checking the dql_avail below, 3339 * because in netdev_tx_completed_queue we update the dql_completed 3340 * before checking the XOFF flag. 3341 */ 3342 smp_mb(); 3343 3344 /* check again in case another CPU has just made room avail */ 3345 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3346 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3347 #endif 3348 } 3349 3350 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3351 * that they should not test BQL status themselves. 3352 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3353 * skb of a batch. 3354 * Returns true if the doorbell must be used to kick the NIC. 3355 */ 3356 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3357 unsigned int bytes, 3358 bool xmit_more) 3359 { 3360 if (xmit_more) { 3361 #ifdef CONFIG_BQL 3362 dql_queued(&dev_queue->dql, bytes); 3363 #endif 3364 return netif_tx_queue_stopped(dev_queue); 3365 } 3366 netdev_tx_sent_queue(dev_queue, bytes); 3367 return true; 3368 } 3369 3370 /** 3371 * netdev_sent_queue - report the number of bytes queued to hardware 3372 * @dev: network device 3373 * @bytes: number of bytes queued to the hardware device queue 3374 * 3375 * Report the number of bytes queued for sending/completion to the network 3376 * device hardware queue. @bytes should be a good approximation and should 3377 * exactly match netdev_completed_queue() @bytes 3378 */ 3379 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3380 { 3381 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3382 } 3383 3384 static inline bool __netdev_sent_queue(struct net_device *dev, 3385 unsigned int bytes, 3386 bool xmit_more) 3387 { 3388 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3389 xmit_more); 3390 } 3391 3392 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3393 unsigned int pkts, unsigned int bytes) 3394 { 3395 #ifdef CONFIG_BQL 3396 if (unlikely(!bytes)) 3397 return; 3398 3399 dql_completed(&dev_queue->dql, bytes); 3400 3401 /* 3402 * Without the memory barrier there is a small possiblity that 3403 * netdev_tx_sent_queue will miss the update and cause the queue to 3404 * be stopped forever 3405 */ 3406 smp_mb(); 3407 3408 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3409 return; 3410 3411 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3412 netif_schedule_queue(dev_queue); 3413 #endif 3414 } 3415 3416 /** 3417 * netdev_completed_queue - report bytes and packets completed by device 3418 * @dev: network device 3419 * @pkts: actual number of packets sent over the medium 3420 * @bytes: actual number of bytes sent over the medium 3421 * 3422 * Report the number of bytes and packets transmitted by the network device 3423 * hardware queue over the physical medium, @bytes must exactly match the 3424 * @bytes amount passed to netdev_sent_queue() 3425 */ 3426 static inline void netdev_completed_queue(struct net_device *dev, 3427 unsigned int pkts, unsigned int bytes) 3428 { 3429 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3430 } 3431 3432 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3433 { 3434 #ifdef CONFIG_BQL 3435 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3436 dql_reset(&q->dql); 3437 #endif 3438 } 3439 3440 /** 3441 * netdev_reset_queue - reset the packets and bytes count of a network device 3442 * @dev_queue: network device 3443 * 3444 * Reset the bytes and packet count of a network device and clear the 3445 * software flow control OFF bit for this network device 3446 */ 3447 static inline void netdev_reset_queue(struct net_device *dev_queue) 3448 { 3449 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3450 } 3451 3452 /** 3453 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3454 * @dev: network device 3455 * @queue_index: given tx queue index 3456 * 3457 * Returns 0 if given tx queue index >= number of device tx queues, 3458 * otherwise returns the originally passed tx queue index. 3459 */ 3460 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3461 { 3462 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3463 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3464 dev->name, queue_index, 3465 dev->real_num_tx_queues); 3466 return 0; 3467 } 3468 3469 return queue_index; 3470 } 3471 3472 /** 3473 * netif_running - test if up 3474 * @dev: network device 3475 * 3476 * Test if the device has been brought up. 3477 */ 3478 static inline bool netif_running(const struct net_device *dev) 3479 { 3480 return test_bit(__LINK_STATE_START, &dev->state); 3481 } 3482 3483 /* 3484 * Routines to manage the subqueues on a device. We only need start, 3485 * stop, and a check if it's stopped. All other device management is 3486 * done at the overall netdevice level. 3487 * Also test the device if we're multiqueue. 3488 */ 3489 3490 /** 3491 * netif_start_subqueue - allow sending packets on subqueue 3492 * @dev: network device 3493 * @queue_index: sub queue index 3494 * 3495 * Start individual transmit queue of a device with multiple transmit queues. 3496 */ 3497 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3498 { 3499 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3500 3501 netif_tx_start_queue(txq); 3502 } 3503 3504 /** 3505 * netif_stop_subqueue - stop sending packets on subqueue 3506 * @dev: network device 3507 * @queue_index: sub queue index 3508 * 3509 * Stop individual transmit queue of a device with multiple transmit queues. 3510 */ 3511 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3512 { 3513 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3514 netif_tx_stop_queue(txq); 3515 } 3516 3517 /** 3518 * netif_subqueue_stopped - test status of subqueue 3519 * @dev: network device 3520 * @queue_index: sub queue index 3521 * 3522 * Check individual transmit queue of a device with multiple transmit queues. 3523 */ 3524 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3525 u16 queue_index) 3526 { 3527 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3528 3529 return netif_tx_queue_stopped(txq); 3530 } 3531 3532 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3533 struct sk_buff *skb) 3534 { 3535 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3536 } 3537 3538 /** 3539 * netif_wake_subqueue - allow sending packets on subqueue 3540 * @dev: network device 3541 * @queue_index: sub queue index 3542 * 3543 * Resume individual transmit queue of a device with multiple transmit queues. 3544 */ 3545 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3546 { 3547 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3548 3549 netif_tx_wake_queue(txq); 3550 } 3551 3552 #ifdef CONFIG_XPS 3553 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3554 u16 index); 3555 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3556 u16 index, bool is_rxqs_map); 3557 3558 /** 3559 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3560 * @j: CPU/Rx queue index 3561 * @mask: bitmask of all cpus/rx queues 3562 * @nr_bits: number of bits in the bitmask 3563 * 3564 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3565 */ 3566 static inline bool netif_attr_test_mask(unsigned long j, 3567 const unsigned long *mask, 3568 unsigned int nr_bits) 3569 { 3570 cpu_max_bits_warn(j, nr_bits); 3571 return test_bit(j, mask); 3572 } 3573 3574 /** 3575 * netif_attr_test_online - Test for online CPU/Rx queue 3576 * @j: CPU/Rx queue index 3577 * @online_mask: bitmask for CPUs/Rx queues that are online 3578 * @nr_bits: number of bits in the bitmask 3579 * 3580 * Returns true if a CPU/Rx queue is online. 3581 */ 3582 static inline bool netif_attr_test_online(unsigned long j, 3583 const unsigned long *online_mask, 3584 unsigned int nr_bits) 3585 { 3586 cpu_max_bits_warn(j, nr_bits); 3587 3588 if (online_mask) 3589 return test_bit(j, online_mask); 3590 3591 return (j < nr_bits); 3592 } 3593 3594 /** 3595 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3596 * @n: CPU/Rx queue index 3597 * @srcp: the cpumask/Rx queue mask pointer 3598 * @nr_bits: number of bits in the bitmask 3599 * 3600 * Returns >= nr_bits if no further CPUs/Rx queues set. 3601 */ 3602 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3603 unsigned int nr_bits) 3604 { 3605 /* -1 is a legal arg here. */ 3606 if (n != -1) 3607 cpu_max_bits_warn(n, nr_bits); 3608 3609 if (srcp) 3610 return find_next_bit(srcp, nr_bits, n + 1); 3611 3612 return n + 1; 3613 } 3614 3615 /** 3616 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3617 * @n: CPU/Rx queue index 3618 * @src1p: the first CPUs/Rx queues mask pointer 3619 * @src2p: the second CPUs/Rx queues mask pointer 3620 * @nr_bits: number of bits in the bitmask 3621 * 3622 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3623 */ 3624 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3625 const unsigned long *src2p, 3626 unsigned int nr_bits) 3627 { 3628 /* -1 is a legal arg here. */ 3629 if (n != -1) 3630 cpu_max_bits_warn(n, nr_bits); 3631 3632 if (src1p && src2p) 3633 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3634 else if (src1p) 3635 return find_next_bit(src1p, nr_bits, n + 1); 3636 else if (src2p) 3637 return find_next_bit(src2p, nr_bits, n + 1); 3638 3639 return n + 1; 3640 } 3641 #else 3642 static inline int netif_set_xps_queue(struct net_device *dev, 3643 const struct cpumask *mask, 3644 u16 index) 3645 { 3646 return 0; 3647 } 3648 3649 static inline int __netif_set_xps_queue(struct net_device *dev, 3650 const unsigned long *mask, 3651 u16 index, bool is_rxqs_map) 3652 { 3653 return 0; 3654 } 3655 #endif 3656 3657 /** 3658 * netif_is_multiqueue - test if device has multiple transmit queues 3659 * @dev: network device 3660 * 3661 * Check if device has multiple transmit queues 3662 */ 3663 static inline bool netif_is_multiqueue(const struct net_device *dev) 3664 { 3665 return dev->num_tx_queues > 1; 3666 } 3667 3668 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3669 3670 #ifdef CONFIG_SYSFS 3671 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3672 #else 3673 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3674 unsigned int rxqs) 3675 { 3676 dev->real_num_rx_queues = rxqs; 3677 return 0; 3678 } 3679 #endif 3680 3681 static inline struct netdev_rx_queue * 3682 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3683 { 3684 return dev->_rx + rxq; 3685 } 3686 3687 #ifdef CONFIG_SYSFS 3688 static inline unsigned int get_netdev_rx_queue_index( 3689 struct netdev_rx_queue *queue) 3690 { 3691 struct net_device *dev = queue->dev; 3692 int index = queue - dev->_rx; 3693 3694 BUG_ON(index >= dev->num_rx_queues); 3695 return index; 3696 } 3697 #endif 3698 3699 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3700 int netif_get_num_default_rss_queues(void); 3701 3702 enum skb_free_reason { 3703 SKB_REASON_CONSUMED, 3704 SKB_REASON_DROPPED, 3705 }; 3706 3707 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3708 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3709 3710 /* 3711 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3712 * interrupt context or with hardware interrupts being disabled. 3713 * (in_irq() || irqs_disabled()) 3714 * 3715 * We provide four helpers that can be used in following contexts : 3716 * 3717 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3718 * replacing kfree_skb(skb) 3719 * 3720 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3721 * Typically used in place of consume_skb(skb) in TX completion path 3722 * 3723 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3724 * replacing kfree_skb(skb) 3725 * 3726 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3727 * and consumed a packet. Used in place of consume_skb(skb) 3728 */ 3729 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3730 { 3731 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3732 } 3733 3734 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3735 { 3736 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3737 } 3738 3739 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3740 { 3741 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3742 } 3743 3744 static inline void dev_consume_skb_any(struct sk_buff *skb) 3745 { 3746 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3747 } 3748 3749 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3750 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3751 int netif_rx(struct sk_buff *skb); 3752 int netif_rx_ni(struct sk_buff *skb); 3753 int netif_receive_skb(struct sk_buff *skb); 3754 int netif_receive_skb_core(struct sk_buff *skb); 3755 void netif_receive_skb_list(struct list_head *head); 3756 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3757 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3758 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3759 gro_result_t napi_gro_frags(struct napi_struct *napi); 3760 struct packet_offload *gro_find_receive_by_type(__be16 type); 3761 struct packet_offload *gro_find_complete_by_type(__be16 type); 3762 3763 static inline void napi_free_frags(struct napi_struct *napi) 3764 { 3765 kfree_skb(napi->skb); 3766 napi->skb = NULL; 3767 } 3768 3769 bool netdev_is_rx_handler_busy(struct net_device *dev); 3770 int netdev_rx_handler_register(struct net_device *dev, 3771 rx_handler_func_t *rx_handler, 3772 void *rx_handler_data); 3773 void netdev_rx_handler_unregister(struct net_device *dev); 3774 3775 bool dev_valid_name(const char *name); 3776 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3777 bool *need_copyout); 3778 int dev_ifconf(struct net *net, struct ifconf *, int); 3779 int dev_ethtool(struct net *net, struct ifreq *); 3780 unsigned int dev_get_flags(const struct net_device *); 3781 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3782 struct netlink_ext_ack *extack); 3783 int dev_change_flags(struct net_device *dev, unsigned int flags, 3784 struct netlink_ext_ack *extack); 3785 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3786 unsigned int gchanges); 3787 int dev_change_name(struct net_device *, const char *); 3788 int dev_set_alias(struct net_device *, const char *, size_t); 3789 int dev_get_alias(const struct net_device *, char *, size_t); 3790 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3791 int __dev_set_mtu(struct net_device *, int); 3792 int dev_validate_mtu(struct net_device *dev, int mtu, 3793 struct netlink_ext_ack *extack); 3794 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3795 struct netlink_ext_ack *extack); 3796 int dev_set_mtu(struct net_device *, int); 3797 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3798 void dev_set_group(struct net_device *, int); 3799 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3800 struct netlink_ext_ack *extack); 3801 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3802 struct netlink_ext_ack *extack); 3803 int dev_change_carrier(struct net_device *, bool new_carrier); 3804 int dev_get_phys_port_id(struct net_device *dev, 3805 struct netdev_phys_item_id *ppid); 3806 int dev_get_phys_port_name(struct net_device *dev, 3807 char *name, size_t len); 3808 int dev_get_port_parent_id(struct net_device *dev, 3809 struct netdev_phys_item_id *ppid, bool recurse); 3810 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3811 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3812 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3813 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3814 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3815 struct netdev_queue *txq, int *ret); 3816 3817 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3818 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3819 int fd, int expected_fd, u32 flags); 3820 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3821 enum bpf_netdev_command cmd); 3822 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3823 3824 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3825 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3826 bool is_skb_forwardable(const struct net_device *dev, 3827 const struct sk_buff *skb); 3828 3829 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3830 struct sk_buff *skb) 3831 { 3832 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3833 unlikely(!is_skb_forwardable(dev, skb))) { 3834 atomic_long_inc(&dev->rx_dropped); 3835 kfree_skb(skb); 3836 return NET_RX_DROP; 3837 } 3838 3839 skb_scrub_packet(skb, true); 3840 skb->priority = 0; 3841 return 0; 3842 } 3843 3844 bool dev_nit_active(struct net_device *dev); 3845 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3846 3847 extern int netdev_budget; 3848 extern unsigned int netdev_budget_usecs; 3849 3850 /* Called by rtnetlink.c:rtnl_unlock() */ 3851 void netdev_run_todo(void); 3852 3853 /** 3854 * dev_put - release reference to device 3855 * @dev: network device 3856 * 3857 * Release reference to device to allow it to be freed. 3858 */ 3859 static inline void dev_put(struct net_device *dev) 3860 { 3861 this_cpu_dec(*dev->pcpu_refcnt); 3862 } 3863 3864 /** 3865 * dev_hold - get reference to device 3866 * @dev: network device 3867 * 3868 * Hold reference to device to keep it from being freed. 3869 */ 3870 static inline void dev_hold(struct net_device *dev) 3871 { 3872 this_cpu_inc(*dev->pcpu_refcnt); 3873 } 3874 3875 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3876 * and _off may be called from IRQ context, but it is caller 3877 * who is responsible for serialization of these calls. 3878 * 3879 * The name carrier is inappropriate, these functions should really be 3880 * called netif_lowerlayer_*() because they represent the state of any 3881 * kind of lower layer not just hardware media. 3882 */ 3883 3884 void linkwatch_init_dev(struct net_device *dev); 3885 void linkwatch_fire_event(struct net_device *dev); 3886 void linkwatch_forget_dev(struct net_device *dev); 3887 3888 /** 3889 * netif_carrier_ok - test if carrier present 3890 * @dev: network device 3891 * 3892 * Check if carrier is present on device 3893 */ 3894 static inline bool netif_carrier_ok(const struct net_device *dev) 3895 { 3896 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3897 } 3898 3899 unsigned long dev_trans_start(struct net_device *dev); 3900 3901 void __netdev_watchdog_up(struct net_device *dev); 3902 3903 void netif_carrier_on(struct net_device *dev); 3904 3905 void netif_carrier_off(struct net_device *dev); 3906 3907 /** 3908 * netif_dormant_on - mark device as dormant. 3909 * @dev: network device 3910 * 3911 * Mark device as dormant (as per RFC2863). 3912 * 3913 * The dormant state indicates that the relevant interface is not 3914 * actually in a condition to pass packets (i.e., it is not 'up') but is 3915 * in a "pending" state, waiting for some external event. For "on- 3916 * demand" interfaces, this new state identifies the situation where the 3917 * interface is waiting for events to place it in the up state. 3918 */ 3919 static inline void netif_dormant_on(struct net_device *dev) 3920 { 3921 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3922 linkwatch_fire_event(dev); 3923 } 3924 3925 /** 3926 * netif_dormant_off - set device as not dormant. 3927 * @dev: network device 3928 * 3929 * Device is not in dormant state. 3930 */ 3931 static inline void netif_dormant_off(struct net_device *dev) 3932 { 3933 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3934 linkwatch_fire_event(dev); 3935 } 3936 3937 /** 3938 * netif_dormant - test if device is dormant 3939 * @dev: network device 3940 * 3941 * Check if device is dormant. 3942 */ 3943 static inline bool netif_dormant(const struct net_device *dev) 3944 { 3945 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3946 } 3947 3948 3949 /** 3950 * netif_testing_on - mark device as under test. 3951 * @dev: network device 3952 * 3953 * Mark device as under test (as per RFC2863). 3954 * 3955 * The testing state indicates that some test(s) must be performed on 3956 * the interface. After completion, of the test, the interface state 3957 * will change to up, dormant, or down, as appropriate. 3958 */ 3959 static inline void netif_testing_on(struct net_device *dev) 3960 { 3961 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 3962 linkwatch_fire_event(dev); 3963 } 3964 3965 /** 3966 * netif_testing_off - set device as not under test. 3967 * @dev: network device 3968 * 3969 * Device is not in testing state. 3970 */ 3971 static inline void netif_testing_off(struct net_device *dev) 3972 { 3973 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 3974 linkwatch_fire_event(dev); 3975 } 3976 3977 /** 3978 * netif_testing - test if device is under test 3979 * @dev: network device 3980 * 3981 * Check if device is under test 3982 */ 3983 static inline bool netif_testing(const struct net_device *dev) 3984 { 3985 return test_bit(__LINK_STATE_TESTING, &dev->state); 3986 } 3987 3988 3989 /** 3990 * netif_oper_up - test if device is operational 3991 * @dev: network device 3992 * 3993 * Check if carrier is operational 3994 */ 3995 static inline bool netif_oper_up(const struct net_device *dev) 3996 { 3997 return (dev->operstate == IF_OPER_UP || 3998 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3999 } 4000 4001 /** 4002 * netif_device_present - is device available or removed 4003 * @dev: network device 4004 * 4005 * Check if device has not been removed from system. 4006 */ 4007 static inline bool netif_device_present(struct net_device *dev) 4008 { 4009 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4010 } 4011 4012 void netif_device_detach(struct net_device *dev); 4013 4014 void netif_device_attach(struct net_device *dev); 4015 4016 /* 4017 * Network interface message level settings 4018 */ 4019 4020 enum { 4021 NETIF_MSG_DRV_BIT, 4022 NETIF_MSG_PROBE_BIT, 4023 NETIF_MSG_LINK_BIT, 4024 NETIF_MSG_TIMER_BIT, 4025 NETIF_MSG_IFDOWN_BIT, 4026 NETIF_MSG_IFUP_BIT, 4027 NETIF_MSG_RX_ERR_BIT, 4028 NETIF_MSG_TX_ERR_BIT, 4029 NETIF_MSG_TX_QUEUED_BIT, 4030 NETIF_MSG_INTR_BIT, 4031 NETIF_MSG_TX_DONE_BIT, 4032 NETIF_MSG_RX_STATUS_BIT, 4033 NETIF_MSG_PKTDATA_BIT, 4034 NETIF_MSG_HW_BIT, 4035 NETIF_MSG_WOL_BIT, 4036 4037 /* When you add a new bit above, update netif_msg_class_names array 4038 * in net/ethtool/common.c 4039 */ 4040 NETIF_MSG_CLASS_COUNT, 4041 }; 4042 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4043 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4044 4045 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4046 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4047 4048 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4049 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4050 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4051 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4052 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4053 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4054 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4055 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4056 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4057 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4058 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4059 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4060 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4061 #define NETIF_MSG_HW __NETIF_MSG(HW) 4062 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4063 4064 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4065 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4066 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4067 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4068 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4069 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4070 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4071 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4072 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4073 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4074 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4075 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4076 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4077 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4078 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4079 4080 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4081 { 4082 /* use default */ 4083 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4084 return default_msg_enable_bits; 4085 if (debug_value == 0) /* no output */ 4086 return 0; 4087 /* set low N bits */ 4088 return (1U << debug_value) - 1; 4089 } 4090 4091 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4092 { 4093 spin_lock(&txq->_xmit_lock); 4094 txq->xmit_lock_owner = cpu; 4095 } 4096 4097 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4098 { 4099 __acquire(&txq->_xmit_lock); 4100 return true; 4101 } 4102 4103 static inline void __netif_tx_release(struct netdev_queue *txq) 4104 { 4105 __release(&txq->_xmit_lock); 4106 } 4107 4108 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4109 { 4110 spin_lock_bh(&txq->_xmit_lock); 4111 txq->xmit_lock_owner = smp_processor_id(); 4112 } 4113 4114 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4115 { 4116 bool ok = spin_trylock(&txq->_xmit_lock); 4117 if (likely(ok)) 4118 txq->xmit_lock_owner = smp_processor_id(); 4119 return ok; 4120 } 4121 4122 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4123 { 4124 txq->xmit_lock_owner = -1; 4125 spin_unlock(&txq->_xmit_lock); 4126 } 4127 4128 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4129 { 4130 txq->xmit_lock_owner = -1; 4131 spin_unlock_bh(&txq->_xmit_lock); 4132 } 4133 4134 static inline void txq_trans_update(struct netdev_queue *txq) 4135 { 4136 if (txq->xmit_lock_owner != -1) 4137 txq->trans_start = jiffies; 4138 } 4139 4140 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4141 static inline void netif_trans_update(struct net_device *dev) 4142 { 4143 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4144 4145 if (txq->trans_start != jiffies) 4146 txq->trans_start = jiffies; 4147 } 4148 4149 /** 4150 * netif_tx_lock - grab network device transmit lock 4151 * @dev: network device 4152 * 4153 * Get network device transmit lock 4154 */ 4155 static inline void netif_tx_lock(struct net_device *dev) 4156 { 4157 unsigned int i; 4158 int cpu; 4159 4160 spin_lock(&dev->tx_global_lock); 4161 cpu = smp_processor_id(); 4162 for (i = 0; i < dev->num_tx_queues; i++) { 4163 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4164 4165 /* We are the only thread of execution doing a 4166 * freeze, but we have to grab the _xmit_lock in 4167 * order to synchronize with threads which are in 4168 * the ->hard_start_xmit() handler and already 4169 * checked the frozen bit. 4170 */ 4171 __netif_tx_lock(txq, cpu); 4172 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4173 __netif_tx_unlock(txq); 4174 } 4175 } 4176 4177 static inline void netif_tx_lock_bh(struct net_device *dev) 4178 { 4179 local_bh_disable(); 4180 netif_tx_lock(dev); 4181 } 4182 4183 static inline void netif_tx_unlock(struct net_device *dev) 4184 { 4185 unsigned int i; 4186 4187 for (i = 0; i < dev->num_tx_queues; i++) { 4188 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4189 4190 /* No need to grab the _xmit_lock here. If the 4191 * queue is not stopped for another reason, we 4192 * force a schedule. 4193 */ 4194 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4195 netif_schedule_queue(txq); 4196 } 4197 spin_unlock(&dev->tx_global_lock); 4198 } 4199 4200 static inline void netif_tx_unlock_bh(struct net_device *dev) 4201 { 4202 netif_tx_unlock(dev); 4203 local_bh_enable(); 4204 } 4205 4206 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4207 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4208 __netif_tx_lock(txq, cpu); \ 4209 } else { \ 4210 __netif_tx_acquire(txq); \ 4211 } \ 4212 } 4213 4214 #define HARD_TX_TRYLOCK(dev, txq) \ 4215 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4216 __netif_tx_trylock(txq) : \ 4217 __netif_tx_acquire(txq)) 4218 4219 #define HARD_TX_UNLOCK(dev, txq) { \ 4220 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4221 __netif_tx_unlock(txq); \ 4222 } else { \ 4223 __netif_tx_release(txq); \ 4224 } \ 4225 } 4226 4227 static inline void netif_tx_disable(struct net_device *dev) 4228 { 4229 unsigned int i; 4230 int cpu; 4231 4232 local_bh_disable(); 4233 cpu = smp_processor_id(); 4234 for (i = 0; i < dev->num_tx_queues; i++) { 4235 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4236 4237 __netif_tx_lock(txq, cpu); 4238 netif_tx_stop_queue(txq); 4239 __netif_tx_unlock(txq); 4240 } 4241 local_bh_enable(); 4242 } 4243 4244 static inline void netif_addr_lock(struct net_device *dev) 4245 { 4246 spin_lock(&dev->addr_list_lock); 4247 } 4248 4249 static inline void netif_addr_lock_nested(struct net_device *dev) 4250 { 4251 spin_lock_nested(&dev->addr_list_lock, dev->lower_level); 4252 } 4253 4254 static inline void netif_addr_lock_bh(struct net_device *dev) 4255 { 4256 spin_lock_bh(&dev->addr_list_lock); 4257 } 4258 4259 static inline void netif_addr_unlock(struct net_device *dev) 4260 { 4261 spin_unlock(&dev->addr_list_lock); 4262 } 4263 4264 static inline void netif_addr_unlock_bh(struct net_device *dev) 4265 { 4266 spin_unlock_bh(&dev->addr_list_lock); 4267 } 4268 4269 /* 4270 * dev_addrs walker. Should be used only for read access. Call with 4271 * rcu_read_lock held. 4272 */ 4273 #define for_each_dev_addr(dev, ha) \ 4274 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4275 4276 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4277 4278 void ether_setup(struct net_device *dev); 4279 4280 /* Support for loadable net-drivers */ 4281 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4282 unsigned char name_assign_type, 4283 void (*setup)(struct net_device *), 4284 unsigned int txqs, unsigned int rxqs); 4285 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4286 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4287 4288 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4289 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4290 count) 4291 4292 int register_netdev(struct net_device *dev); 4293 void unregister_netdev(struct net_device *dev); 4294 4295 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4296 4297 /* General hardware address lists handling functions */ 4298 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4299 struct netdev_hw_addr_list *from_list, int addr_len); 4300 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4301 struct netdev_hw_addr_list *from_list, int addr_len); 4302 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4303 struct net_device *dev, 4304 int (*sync)(struct net_device *, const unsigned char *), 4305 int (*unsync)(struct net_device *, 4306 const unsigned char *)); 4307 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4308 struct net_device *dev, 4309 int (*sync)(struct net_device *, 4310 const unsigned char *, int), 4311 int (*unsync)(struct net_device *, 4312 const unsigned char *, int)); 4313 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4314 struct net_device *dev, 4315 int (*unsync)(struct net_device *, 4316 const unsigned char *, int)); 4317 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4318 struct net_device *dev, 4319 int (*unsync)(struct net_device *, 4320 const unsigned char *)); 4321 void __hw_addr_init(struct netdev_hw_addr_list *list); 4322 4323 /* Functions used for device addresses handling */ 4324 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4325 unsigned char addr_type); 4326 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4327 unsigned char addr_type); 4328 void dev_addr_flush(struct net_device *dev); 4329 int dev_addr_init(struct net_device *dev); 4330 4331 /* Functions used for unicast addresses handling */ 4332 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4333 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4334 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4335 int dev_uc_sync(struct net_device *to, struct net_device *from); 4336 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4337 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4338 void dev_uc_flush(struct net_device *dev); 4339 void dev_uc_init(struct net_device *dev); 4340 4341 /** 4342 * __dev_uc_sync - Synchonize device's unicast list 4343 * @dev: device to sync 4344 * @sync: function to call if address should be added 4345 * @unsync: function to call if address should be removed 4346 * 4347 * Add newly added addresses to the interface, and release 4348 * addresses that have been deleted. 4349 */ 4350 static inline int __dev_uc_sync(struct net_device *dev, 4351 int (*sync)(struct net_device *, 4352 const unsigned char *), 4353 int (*unsync)(struct net_device *, 4354 const unsigned char *)) 4355 { 4356 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4357 } 4358 4359 /** 4360 * __dev_uc_unsync - Remove synchronized addresses from device 4361 * @dev: device to sync 4362 * @unsync: function to call if address should be removed 4363 * 4364 * Remove all addresses that were added to the device by dev_uc_sync(). 4365 */ 4366 static inline void __dev_uc_unsync(struct net_device *dev, 4367 int (*unsync)(struct net_device *, 4368 const unsigned char *)) 4369 { 4370 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4371 } 4372 4373 /* Functions used for multicast addresses handling */ 4374 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4375 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4376 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4377 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4378 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4379 int dev_mc_sync(struct net_device *to, struct net_device *from); 4380 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4381 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4382 void dev_mc_flush(struct net_device *dev); 4383 void dev_mc_init(struct net_device *dev); 4384 4385 /** 4386 * __dev_mc_sync - Synchonize device's multicast list 4387 * @dev: device to sync 4388 * @sync: function to call if address should be added 4389 * @unsync: function to call if address should be removed 4390 * 4391 * Add newly added addresses to the interface, and release 4392 * addresses that have been deleted. 4393 */ 4394 static inline int __dev_mc_sync(struct net_device *dev, 4395 int (*sync)(struct net_device *, 4396 const unsigned char *), 4397 int (*unsync)(struct net_device *, 4398 const unsigned char *)) 4399 { 4400 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4401 } 4402 4403 /** 4404 * __dev_mc_unsync - Remove synchronized addresses from device 4405 * @dev: device to sync 4406 * @unsync: function to call if address should be removed 4407 * 4408 * Remove all addresses that were added to the device by dev_mc_sync(). 4409 */ 4410 static inline void __dev_mc_unsync(struct net_device *dev, 4411 int (*unsync)(struct net_device *, 4412 const unsigned char *)) 4413 { 4414 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4415 } 4416 4417 /* Functions used for secondary unicast and multicast support */ 4418 void dev_set_rx_mode(struct net_device *dev); 4419 void __dev_set_rx_mode(struct net_device *dev); 4420 int dev_set_promiscuity(struct net_device *dev, int inc); 4421 int dev_set_allmulti(struct net_device *dev, int inc); 4422 void netdev_state_change(struct net_device *dev); 4423 void netdev_notify_peers(struct net_device *dev); 4424 void netdev_features_change(struct net_device *dev); 4425 /* Load a device via the kmod */ 4426 void dev_load(struct net *net, const char *name); 4427 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4428 struct rtnl_link_stats64 *storage); 4429 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4430 const struct net_device_stats *netdev_stats); 4431 4432 extern int netdev_max_backlog; 4433 extern int netdev_tstamp_prequeue; 4434 extern int weight_p; 4435 extern int dev_weight_rx_bias; 4436 extern int dev_weight_tx_bias; 4437 extern int dev_rx_weight; 4438 extern int dev_tx_weight; 4439 extern int gro_normal_batch; 4440 4441 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4442 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4443 struct list_head **iter); 4444 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4445 struct list_head **iter); 4446 4447 /* iterate through upper list, must be called under RCU read lock */ 4448 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4449 for (iter = &(dev)->adj_list.upper, \ 4450 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4451 updev; \ 4452 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4453 4454 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4455 int (*fn)(struct net_device *upper_dev, 4456 void *data), 4457 void *data); 4458 4459 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4460 struct net_device *upper_dev); 4461 4462 bool netdev_has_any_upper_dev(struct net_device *dev); 4463 4464 void *netdev_lower_get_next_private(struct net_device *dev, 4465 struct list_head **iter); 4466 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4467 struct list_head **iter); 4468 4469 #define netdev_for_each_lower_private(dev, priv, iter) \ 4470 for (iter = (dev)->adj_list.lower.next, \ 4471 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4472 priv; \ 4473 priv = netdev_lower_get_next_private(dev, &(iter))) 4474 4475 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4476 for (iter = &(dev)->adj_list.lower, \ 4477 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4478 priv; \ 4479 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4480 4481 void *netdev_lower_get_next(struct net_device *dev, 4482 struct list_head **iter); 4483 4484 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4485 for (iter = (dev)->adj_list.lower.next, \ 4486 ldev = netdev_lower_get_next(dev, &(iter)); \ 4487 ldev; \ 4488 ldev = netdev_lower_get_next(dev, &(iter))) 4489 4490 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4491 struct list_head **iter); 4492 int netdev_walk_all_lower_dev(struct net_device *dev, 4493 int (*fn)(struct net_device *lower_dev, 4494 void *data), 4495 void *data); 4496 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4497 int (*fn)(struct net_device *lower_dev, 4498 void *data), 4499 void *data); 4500 4501 void *netdev_adjacent_get_private(struct list_head *adj_list); 4502 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4503 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4504 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4505 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4506 struct netlink_ext_ack *extack); 4507 int netdev_master_upper_dev_link(struct net_device *dev, 4508 struct net_device *upper_dev, 4509 void *upper_priv, void *upper_info, 4510 struct netlink_ext_ack *extack); 4511 void netdev_upper_dev_unlink(struct net_device *dev, 4512 struct net_device *upper_dev); 4513 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4514 struct net_device *new_dev, 4515 struct net_device *dev, 4516 struct netlink_ext_ack *extack); 4517 void netdev_adjacent_change_commit(struct net_device *old_dev, 4518 struct net_device *new_dev, 4519 struct net_device *dev); 4520 void netdev_adjacent_change_abort(struct net_device *old_dev, 4521 struct net_device *new_dev, 4522 struct net_device *dev); 4523 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4524 void *netdev_lower_dev_get_private(struct net_device *dev, 4525 struct net_device *lower_dev); 4526 void netdev_lower_state_changed(struct net_device *lower_dev, 4527 void *lower_state_info); 4528 4529 /* RSS keys are 40 or 52 bytes long */ 4530 #define NETDEV_RSS_KEY_LEN 52 4531 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4532 void netdev_rss_key_fill(void *buffer, size_t len); 4533 4534 int skb_checksum_help(struct sk_buff *skb); 4535 int skb_crc32c_csum_help(struct sk_buff *skb); 4536 int skb_csum_hwoffload_help(struct sk_buff *skb, 4537 const netdev_features_t features); 4538 4539 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4540 netdev_features_t features, bool tx_path); 4541 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4542 netdev_features_t features); 4543 4544 struct netdev_bonding_info { 4545 ifslave slave; 4546 ifbond master; 4547 }; 4548 4549 struct netdev_notifier_bonding_info { 4550 struct netdev_notifier_info info; /* must be first */ 4551 struct netdev_bonding_info bonding_info; 4552 }; 4553 4554 void netdev_bonding_info_change(struct net_device *dev, 4555 struct netdev_bonding_info *bonding_info); 4556 4557 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4558 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4559 #else 4560 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4561 const void *data) 4562 { 4563 } 4564 #endif 4565 4566 static inline 4567 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4568 { 4569 return __skb_gso_segment(skb, features, true); 4570 } 4571 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4572 4573 static inline bool can_checksum_protocol(netdev_features_t features, 4574 __be16 protocol) 4575 { 4576 if (protocol == htons(ETH_P_FCOE)) 4577 return !!(features & NETIF_F_FCOE_CRC); 4578 4579 /* Assume this is an IP checksum (not SCTP CRC) */ 4580 4581 if (features & NETIF_F_HW_CSUM) { 4582 /* Can checksum everything */ 4583 return true; 4584 } 4585 4586 switch (protocol) { 4587 case htons(ETH_P_IP): 4588 return !!(features & NETIF_F_IP_CSUM); 4589 case htons(ETH_P_IPV6): 4590 return !!(features & NETIF_F_IPV6_CSUM); 4591 default: 4592 return false; 4593 } 4594 } 4595 4596 #ifdef CONFIG_BUG 4597 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4598 #else 4599 static inline void netdev_rx_csum_fault(struct net_device *dev, 4600 struct sk_buff *skb) 4601 { 4602 } 4603 #endif 4604 /* rx skb timestamps */ 4605 void net_enable_timestamp(void); 4606 void net_disable_timestamp(void); 4607 4608 #ifdef CONFIG_PROC_FS 4609 int __init dev_proc_init(void); 4610 #else 4611 #define dev_proc_init() 0 4612 #endif 4613 4614 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4615 struct sk_buff *skb, struct net_device *dev, 4616 bool more) 4617 { 4618 __this_cpu_write(softnet_data.xmit.more, more); 4619 return ops->ndo_start_xmit(skb, dev); 4620 } 4621 4622 static inline bool netdev_xmit_more(void) 4623 { 4624 return __this_cpu_read(softnet_data.xmit.more); 4625 } 4626 4627 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4628 struct netdev_queue *txq, bool more) 4629 { 4630 const struct net_device_ops *ops = dev->netdev_ops; 4631 netdev_tx_t rc; 4632 4633 rc = __netdev_start_xmit(ops, skb, dev, more); 4634 if (rc == NETDEV_TX_OK) 4635 txq_trans_update(txq); 4636 4637 return rc; 4638 } 4639 4640 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4641 const void *ns); 4642 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4643 const void *ns); 4644 4645 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4646 { 4647 return netdev_class_create_file_ns(class_attr, NULL); 4648 } 4649 4650 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4651 { 4652 netdev_class_remove_file_ns(class_attr, NULL); 4653 } 4654 4655 extern const struct kobj_ns_type_operations net_ns_type_operations; 4656 4657 const char *netdev_drivername(const struct net_device *dev); 4658 4659 void linkwatch_run_queue(void); 4660 4661 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4662 netdev_features_t f2) 4663 { 4664 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4665 if (f1 & NETIF_F_HW_CSUM) 4666 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4667 else 4668 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4669 } 4670 4671 return f1 & f2; 4672 } 4673 4674 static inline netdev_features_t netdev_get_wanted_features( 4675 struct net_device *dev) 4676 { 4677 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4678 } 4679 netdev_features_t netdev_increment_features(netdev_features_t all, 4680 netdev_features_t one, netdev_features_t mask); 4681 4682 /* Allow TSO being used on stacked device : 4683 * Performing the GSO segmentation before last device 4684 * is a performance improvement. 4685 */ 4686 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4687 netdev_features_t mask) 4688 { 4689 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4690 } 4691 4692 int __netdev_update_features(struct net_device *dev); 4693 void netdev_update_features(struct net_device *dev); 4694 void netdev_change_features(struct net_device *dev); 4695 4696 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4697 struct net_device *dev); 4698 4699 netdev_features_t passthru_features_check(struct sk_buff *skb, 4700 struct net_device *dev, 4701 netdev_features_t features); 4702 netdev_features_t netif_skb_features(struct sk_buff *skb); 4703 4704 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4705 { 4706 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4707 4708 /* check flags correspondence */ 4709 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4710 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4711 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4712 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4713 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4714 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4715 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4716 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4717 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4718 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4719 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4720 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4721 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4722 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4723 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4724 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4725 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4726 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4727 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4728 4729 return (features & feature) == feature; 4730 } 4731 4732 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4733 { 4734 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4735 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4736 } 4737 4738 static inline bool netif_needs_gso(struct sk_buff *skb, 4739 netdev_features_t features) 4740 { 4741 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4742 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4743 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4744 } 4745 4746 static inline void netif_set_gso_max_size(struct net_device *dev, 4747 unsigned int size) 4748 { 4749 dev->gso_max_size = size; 4750 } 4751 4752 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4753 int pulled_hlen, u16 mac_offset, 4754 int mac_len) 4755 { 4756 skb->protocol = protocol; 4757 skb->encapsulation = 1; 4758 skb_push(skb, pulled_hlen); 4759 skb_reset_transport_header(skb); 4760 skb->mac_header = mac_offset; 4761 skb->network_header = skb->mac_header + mac_len; 4762 skb->mac_len = mac_len; 4763 } 4764 4765 static inline bool netif_is_macsec(const struct net_device *dev) 4766 { 4767 return dev->priv_flags & IFF_MACSEC; 4768 } 4769 4770 static inline bool netif_is_macvlan(const struct net_device *dev) 4771 { 4772 return dev->priv_flags & IFF_MACVLAN; 4773 } 4774 4775 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4776 { 4777 return dev->priv_flags & IFF_MACVLAN_PORT; 4778 } 4779 4780 static inline bool netif_is_bond_master(const struct net_device *dev) 4781 { 4782 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4783 } 4784 4785 static inline bool netif_is_bond_slave(const struct net_device *dev) 4786 { 4787 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4788 } 4789 4790 static inline bool netif_supports_nofcs(struct net_device *dev) 4791 { 4792 return dev->priv_flags & IFF_SUPP_NOFCS; 4793 } 4794 4795 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4796 { 4797 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4798 } 4799 4800 static inline bool netif_is_l3_master(const struct net_device *dev) 4801 { 4802 return dev->priv_flags & IFF_L3MDEV_MASTER; 4803 } 4804 4805 static inline bool netif_is_l3_slave(const struct net_device *dev) 4806 { 4807 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4808 } 4809 4810 static inline bool netif_is_bridge_master(const struct net_device *dev) 4811 { 4812 return dev->priv_flags & IFF_EBRIDGE; 4813 } 4814 4815 static inline bool netif_is_bridge_port(const struct net_device *dev) 4816 { 4817 return dev->priv_flags & IFF_BRIDGE_PORT; 4818 } 4819 4820 static inline bool netif_is_ovs_master(const struct net_device *dev) 4821 { 4822 return dev->priv_flags & IFF_OPENVSWITCH; 4823 } 4824 4825 static inline bool netif_is_ovs_port(const struct net_device *dev) 4826 { 4827 return dev->priv_flags & IFF_OVS_DATAPATH; 4828 } 4829 4830 static inline bool netif_is_team_master(const struct net_device *dev) 4831 { 4832 return dev->priv_flags & IFF_TEAM; 4833 } 4834 4835 static inline bool netif_is_team_port(const struct net_device *dev) 4836 { 4837 return dev->priv_flags & IFF_TEAM_PORT; 4838 } 4839 4840 static inline bool netif_is_lag_master(const struct net_device *dev) 4841 { 4842 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4843 } 4844 4845 static inline bool netif_is_lag_port(const struct net_device *dev) 4846 { 4847 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4848 } 4849 4850 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4851 { 4852 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4853 } 4854 4855 static inline bool netif_is_failover(const struct net_device *dev) 4856 { 4857 return dev->priv_flags & IFF_FAILOVER; 4858 } 4859 4860 static inline bool netif_is_failover_slave(const struct net_device *dev) 4861 { 4862 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4863 } 4864 4865 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4866 static inline void netif_keep_dst(struct net_device *dev) 4867 { 4868 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4869 } 4870 4871 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4872 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4873 { 4874 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4875 return dev->priv_flags & IFF_MACSEC; 4876 } 4877 4878 extern struct pernet_operations __net_initdata loopback_net_ops; 4879 4880 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4881 4882 /* netdev_printk helpers, similar to dev_printk */ 4883 4884 static inline const char *netdev_name(const struct net_device *dev) 4885 { 4886 if (!dev->name[0] || strchr(dev->name, '%')) 4887 return "(unnamed net_device)"; 4888 return dev->name; 4889 } 4890 4891 static inline bool netdev_unregistering(const struct net_device *dev) 4892 { 4893 return dev->reg_state == NETREG_UNREGISTERING; 4894 } 4895 4896 static inline const char *netdev_reg_state(const struct net_device *dev) 4897 { 4898 switch (dev->reg_state) { 4899 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4900 case NETREG_REGISTERED: return ""; 4901 case NETREG_UNREGISTERING: return " (unregistering)"; 4902 case NETREG_UNREGISTERED: return " (unregistered)"; 4903 case NETREG_RELEASED: return " (released)"; 4904 case NETREG_DUMMY: return " (dummy)"; 4905 } 4906 4907 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4908 return " (unknown)"; 4909 } 4910 4911 __printf(3, 4) __cold 4912 void netdev_printk(const char *level, const struct net_device *dev, 4913 const char *format, ...); 4914 __printf(2, 3) __cold 4915 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4916 __printf(2, 3) __cold 4917 void netdev_alert(const struct net_device *dev, const char *format, ...); 4918 __printf(2, 3) __cold 4919 void netdev_crit(const struct net_device *dev, const char *format, ...); 4920 __printf(2, 3) __cold 4921 void netdev_err(const struct net_device *dev, const char *format, ...); 4922 __printf(2, 3) __cold 4923 void netdev_warn(const struct net_device *dev, const char *format, ...); 4924 __printf(2, 3) __cold 4925 void netdev_notice(const struct net_device *dev, const char *format, ...); 4926 __printf(2, 3) __cold 4927 void netdev_info(const struct net_device *dev, const char *format, ...); 4928 4929 #define netdev_level_once(level, dev, fmt, ...) \ 4930 do { \ 4931 static bool __print_once __read_mostly; \ 4932 \ 4933 if (!__print_once) { \ 4934 __print_once = true; \ 4935 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4936 } \ 4937 } while (0) 4938 4939 #define netdev_emerg_once(dev, fmt, ...) \ 4940 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4941 #define netdev_alert_once(dev, fmt, ...) \ 4942 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4943 #define netdev_crit_once(dev, fmt, ...) \ 4944 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4945 #define netdev_err_once(dev, fmt, ...) \ 4946 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4947 #define netdev_warn_once(dev, fmt, ...) \ 4948 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4949 #define netdev_notice_once(dev, fmt, ...) \ 4950 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4951 #define netdev_info_once(dev, fmt, ...) \ 4952 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4953 4954 #define MODULE_ALIAS_NETDEV(device) \ 4955 MODULE_ALIAS("netdev-" device) 4956 4957 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 4958 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 4959 #define netdev_dbg(__dev, format, args...) \ 4960 do { \ 4961 dynamic_netdev_dbg(__dev, format, ##args); \ 4962 } while (0) 4963 #elif defined(DEBUG) 4964 #define netdev_dbg(__dev, format, args...) \ 4965 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4966 #else 4967 #define netdev_dbg(__dev, format, args...) \ 4968 ({ \ 4969 if (0) \ 4970 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4971 }) 4972 #endif 4973 4974 #if defined(VERBOSE_DEBUG) 4975 #define netdev_vdbg netdev_dbg 4976 #else 4977 4978 #define netdev_vdbg(dev, format, args...) \ 4979 ({ \ 4980 if (0) \ 4981 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4982 0; \ 4983 }) 4984 #endif 4985 4986 /* 4987 * netdev_WARN() acts like dev_printk(), but with the key difference 4988 * of using a WARN/WARN_ON to get the message out, including the 4989 * file/line information and a backtrace. 4990 */ 4991 #define netdev_WARN(dev, format, args...) \ 4992 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4993 netdev_reg_state(dev), ##args) 4994 4995 #define netdev_WARN_ONCE(dev, format, args...) \ 4996 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4997 netdev_reg_state(dev), ##args) 4998 4999 /* netif printk helpers, similar to netdev_printk */ 5000 5001 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5002 do { \ 5003 if (netif_msg_##type(priv)) \ 5004 netdev_printk(level, (dev), fmt, ##args); \ 5005 } while (0) 5006 5007 #define netif_level(level, priv, type, dev, fmt, args...) \ 5008 do { \ 5009 if (netif_msg_##type(priv)) \ 5010 netdev_##level(dev, fmt, ##args); \ 5011 } while (0) 5012 5013 #define netif_emerg(priv, type, dev, fmt, args...) \ 5014 netif_level(emerg, priv, type, dev, fmt, ##args) 5015 #define netif_alert(priv, type, dev, fmt, args...) \ 5016 netif_level(alert, priv, type, dev, fmt, ##args) 5017 #define netif_crit(priv, type, dev, fmt, args...) \ 5018 netif_level(crit, priv, type, dev, fmt, ##args) 5019 #define netif_err(priv, type, dev, fmt, args...) \ 5020 netif_level(err, priv, type, dev, fmt, ##args) 5021 #define netif_warn(priv, type, dev, fmt, args...) \ 5022 netif_level(warn, priv, type, dev, fmt, ##args) 5023 #define netif_notice(priv, type, dev, fmt, args...) \ 5024 netif_level(notice, priv, type, dev, fmt, ##args) 5025 #define netif_info(priv, type, dev, fmt, args...) \ 5026 netif_level(info, priv, type, dev, fmt, ##args) 5027 5028 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5029 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5030 #define netif_dbg(priv, type, netdev, format, args...) \ 5031 do { \ 5032 if (netif_msg_##type(priv)) \ 5033 dynamic_netdev_dbg(netdev, format, ##args); \ 5034 } while (0) 5035 #elif defined(DEBUG) 5036 #define netif_dbg(priv, type, dev, format, args...) \ 5037 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5038 #else 5039 #define netif_dbg(priv, type, dev, format, args...) \ 5040 ({ \ 5041 if (0) \ 5042 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5043 0; \ 5044 }) 5045 #endif 5046 5047 /* if @cond then downgrade to debug, else print at @level */ 5048 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5049 do { \ 5050 if (cond) \ 5051 netif_dbg(priv, type, netdev, fmt, ##args); \ 5052 else \ 5053 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5054 } while (0) 5055 5056 #if defined(VERBOSE_DEBUG) 5057 #define netif_vdbg netif_dbg 5058 #else 5059 #define netif_vdbg(priv, type, dev, format, args...) \ 5060 ({ \ 5061 if (0) \ 5062 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5063 0; \ 5064 }) 5065 #endif 5066 5067 /* 5068 * The list of packet types we will receive (as opposed to discard) 5069 * and the routines to invoke. 5070 * 5071 * Why 16. Because with 16 the only overlap we get on a hash of the 5072 * low nibble of the protocol value is RARP/SNAP/X.25. 5073 * 5074 * 0800 IP 5075 * 0001 802.3 5076 * 0002 AX.25 5077 * 0004 802.2 5078 * 8035 RARP 5079 * 0005 SNAP 5080 * 0805 X.25 5081 * 0806 ARP 5082 * 8137 IPX 5083 * 0009 Localtalk 5084 * 86DD IPv6 5085 */ 5086 #define PTYPE_HASH_SIZE (16) 5087 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5088 5089 extern struct net_device *blackhole_netdev; 5090 5091 #endif /* _LINUX_NETDEVICE_H */ 5092