1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 57 struct sfp_bus; 58 /* 802.11 specific */ 59 struct wireless_dev; 60 /* 802.15.4 specific */ 61 struct wpan_dev; 62 struct mpls_dev; 63 /* UDP Tunnel offloads */ 64 struct udp_tunnel_info; 65 struct bpf_prog; 66 struct xdp_buff; 67 68 void netdev_set_default_ethtool_ops(struct net_device *dev, 69 const struct ethtool_ops *ops); 70 71 /* Backlog congestion levels */ 72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 73 #define NET_RX_DROP 1 /* packet dropped */ 74 75 #define MAX_NEST_DEV 8 76 77 /* 78 * Transmit return codes: transmit return codes originate from three different 79 * namespaces: 80 * 81 * - qdisc return codes 82 * - driver transmit return codes 83 * - errno values 84 * 85 * Drivers are allowed to return any one of those in their hard_start_xmit() 86 * function. Real network devices commonly used with qdiscs should only return 87 * the driver transmit return codes though - when qdiscs are used, the actual 88 * transmission happens asynchronously, so the value is not propagated to 89 * higher layers. Virtual network devices transmit synchronously; in this case 90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 91 * others are propagated to higher layers. 92 */ 93 94 /* qdisc ->enqueue() return codes. */ 95 #define NET_XMIT_SUCCESS 0x00 96 #define NET_XMIT_DROP 0x01 /* skb dropped */ 97 #define NET_XMIT_CN 0x02 /* congestion notification */ 98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 99 100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 101 * indicates that the device will soon be dropping packets, or already drops 102 * some packets of the same priority; prompting us to send less aggressively. */ 103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 105 106 /* Driver transmit return codes */ 107 #define NETDEV_TX_MASK 0xf0 108 109 enum netdev_tx { 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 111 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 113 }; 114 typedef enum netdev_tx netdev_tx_t; 115 116 /* 117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 119 */ 120 static inline bool dev_xmit_complete(int rc) 121 { 122 /* 123 * Positive cases with an skb consumed by a driver: 124 * - successful transmission (rc == NETDEV_TX_OK) 125 * - error while transmitting (rc < 0) 126 * - error while queueing to a different device (rc & NET_XMIT_MASK) 127 */ 128 if (likely(rc < NET_XMIT_MASK)) 129 return true; 130 131 return false; 132 } 133 134 /* 135 * Compute the worst-case header length according to the protocols 136 * used. 137 */ 138 139 #if defined(CONFIG_HYPERV_NET) 140 # define LL_MAX_HEADER 128 141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 142 # if defined(CONFIG_MAC80211_MESH) 143 # define LL_MAX_HEADER 128 144 # else 145 # define LL_MAX_HEADER 96 146 # endif 147 #else 148 # define LL_MAX_HEADER 32 149 #endif 150 151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 153 #define MAX_HEADER LL_MAX_HEADER 154 #else 155 #define MAX_HEADER (LL_MAX_HEADER + 48) 156 #endif 157 158 /* 159 * Old network device statistics. Fields are native words 160 * (unsigned long) so they can be read and written atomically. 161 */ 162 163 struct net_device_stats { 164 unsigned long rx_packets; 165 unsigned long tx_packets; 166 unsigned long rx_bytes; 167 unsigned long tx_bytes; 168 unsigned long rx_errors; 169 unsigned long tx_errors; 170 unsigned long rx_dropped; 171 unsigned long tx_dropped; 172 unsigned long multicast; 173 unsigned long collisions; 174 unsigned long rx_length_errors; 175 unsigned long rx_over_errors; 176 unsigned long rx_crc_errors; 177 unsigned long rx_frame_errors; 178 unsigned long rx_fifo_errors; 179 unsigned long rx_missed_errors; 180 unsigned long tx_aborted_errors; 181 unsigned long tx_carrier_errors; 182 unsigned long tx_fifo_errors; 183 unsigned long tx_heartbeat_errors; 184 unsigned long tx_window_errors; 185 unsigned long rx_compressed; 186 unsigned long tx_compressed; 187 }; 188 189 190 #include <linux/cache.h> 191 #include <linux/skbuff.h> 192 193 #ifdef CONFIG_RPS 194 #include <linux/static_key.h> 195 extern struct static_key_false rps_needed; 196 extern struct static_key_false rfs_needed; 197 #endif 198 199 struct neighbour; 200 struct neigh_parms; 201 struct sk_buff; 202 203 struct netdev_hw_addr { 204 struct list_head list; 205 unsigned char addr[MAX_ADDR_LEN]; 206 unsigned char type; 207 #define NETDEV_HW_ADDR_T_LAN 1 208 #define NETDEV_HW_ADDR_T_SAN 2 209 #define NETDEV_HW_ADDR_T_SLAVE 3 210 #define NETDEV_HW_ADDR_T_UNICAST 4 211 #define NETDEV_HW_ADDR_T_MULTICAST 5 212 bool global_use; 213 int sync_cnt; 214 int refcount; 215 int synced; 216 struct rcu_head rcu_head; 217 }; 218 219 struct netdev_hw_addr_list { 220 struct list_head list; 221 int count; 222 }; 223 224 #define netdev_hw_addr_list_count(l) ((l)->count) 225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 226 #define netdev_hw_addr_list_for_each(ha, l) \ 227 list_for_each_entry(ha, &(l)->list, list) 228 229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 231 #define netdev_for_each_uc_addr(ha, dev) \ 232 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 233 234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 236 #define netdev_for_each_mc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 238 239 struct hh_cache { 240 unsigned int hh_len; 241 seqlock_t hh_lock; 242 243 /* cached hardware header; allow for machine alignment needs. */ 244 #define HH_DATA_MOD 16 245 #define HH_DATA_OFF(__len) \ 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 247 #define HH_DATA_ALIGN(__len) \ 248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 250 }; 251 252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 253 * Alternative is: 254 * dev->hard_header_len ? (dev->hard_header_len + 255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 256 * 257 * We could use other alignment values, but we must maintain the 258 * relationship HH alignment <= LL alignment. 259 */ 260 #define LL_RESERVED_SPACE(dev) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 265 struct header_ops { 266 int (*create) (struct sk_buff *skb, struct net_device *dev, 267 unsigned short type, const void *daddr, 268 const void *saddr, unsigned int len); 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 271 void (*cache_update)(struct hh_cache *hh, 272 const struct net_device *dev, 273 const unsigned char *haddr); 274 bool (*validate)(const char *ll_header, unsigned int len); 275 __be16 (*parse_protocol)(const struct sk_buff *skb); 276 }; 277 278 /* These flag bits are private to the generic network queueing 279 * layer; they may not be explicitly referenced by any other 280 * code. 281 */ 282 283 enum netdev_state_t { 284 __LINK_STATE_START, 285 __LINK_STATE_PRESENT, 286 __LINK_STATE_NOCARRIER, 287 __LINK_STATE_LINKWATCH_PENDING, 288 __LINK_STATE_DORMANT, 289 }; 290 291 292 /* 293 * This structure holds boot-time configured netdevice settings. They 294 * are then used in the device probing. 295 */ 296 struct netdev_boot_setup { 297 char name[IFNAMSIZ]; 298 struct ifmap map; 299 }; 300 #define NETDEV_BOOT_SETUP_MAX 8 301 302 int __init netdev_boot_setup(char *str); 303 304 struct gro_list { 305 struct list_head list; 306 int count; 307 }; 308 309 /* 310 * size of gro hash buckets, must less than bit number of 311 * napi_struct::gro_bitmask 312 */ 313 #define GRO_HASH_BUCKETS 8 314 315 /* 316 * Structure for NAPI scheduling similar to tasklet but with weighting 317 */ 318 struct napi_struct { 319 /* The poll_list must only be managed by the entity which 320 * changes the state of the NAPI_STATE_SCHED bit. This means 321 * whoever atomically sets that bit can add this napi_struct 322 * to the per-CPU poll_list, and whoever clears that bit 323 * can remove from the list right before clearing the bit. 324 */ 325 struct list_head poll_list; 326 327 unsigned long state; 328 int weight; 329 unsigned long gro_bitmask; 330 int (*poll)(struct napi_struct *, int); 331 #ifdef CONFIG_NETPOLL 332 int poll_owner; 333 #endif 334 struct net_device *dev; 335 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 336 struct sk_buff *skb; 337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 338 int rx_count; /* length of rx_list */ 339 struct hrtimer timer; 340 struct list_head dev_list; 341 struct hlist_node napi_hash_node; 342 unsigned int napi_id; 343 }; 344 345 enum { 346 NAPI_STATE_SCHED, /* Poll is scheduled */ 347 NAPI_STATE_MISSED, /* reschedule a napi */ 348 NAPI_STATE_DISABLE, /* Disable pending */ 349 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 350 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 351 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 352 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 353 }; 354 355 enum { 356 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 357 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 358 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 359 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 360 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 361 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 362 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 363 }; 364 365 enum gro_result { 366 GRO_MERGED, 367 GRO_MERGED_FREE, 368 GRO_HELD, 369 GRO_NORMAL, 370 GRO_DROP, 371 GRO_CONSUMED, 372 }; 373 typedef enum gro_result gro_result_t; 374 375 /* 376 * enum rx_handler_result - Possible return values for rx_handlers. 377 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 378 * further. 379 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 380 * case skb->dev was changed by rx_handler. 381 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 382 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 383 * 384 * rx_handlers are functions called from inside __netif_receive_skb(), to do 385 * special processing of the skb, prior to delivery to protocol handlers. 386 * 387 * Currently, a net_device can only have a single rx_handler registered. Trying 388 * to register a second rx_handler will return -EBUSY. 389 * 390 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 391 * To unregister a rx_handler on a net_device, use 392 * netdev_rx_handler_unregister(). 393 * 394 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 395 * do with the skb. 396 * 397 * If the rx_handler consumed the skb in some way, it should return 398 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 399 * the skb to be delivered in some other way. 400 * 401 * If the rx_handler changed skb->dev, to divert the skb to another 402 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 403 * new device will be called if it exists. 404 * 405 * If the rx_handler decides the skb should be ignored, it should return 406 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 407 * are registered on exact device (ptype->dev == skb->dev). 408 * 409 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 410 * delivered, it should return RX_HANDLER_PASS. 411 * 412 * A device without a registered rx_handler will behave as if rx_handler 413 * returned RX_HANDLER_PASS. 414 */ 415 416 enum rx_handler_result { 417 RX_HANDLER_CONSUMED, 418 RX_HANDLER_ANOTHER, 419 RX_HANDLER_EXACT, 420 RX_HANDLER_PASS, 421 }; 422 typedef enum rx_handler_result rx_handler_result_t; 423 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 424 425 void __napi_schedule(struct napi_struct *n); 426 void __napi_schedule_irqoff(struct napi_struct *n); 427 428 static inline bool napi_disable_pending(struct napi_struct *n) 429 { 430 return test_bit(NAPI_STATE_DISABLE, &n->state); 431 } 432 433 bool napi_schedule_prep(struct napi_struct *n); 434 435 /** 436 * napi_schedule - schedule NAPI poll 437 * @n: NAPI context 438 * 439 * Schedule NAPI poll routine to be called if it is not already 440 * running. 441 */ 442 static inline void napi_schedule(struct napi_struct *n) 443 { 444 if (napi_schedule_prep(n)) 445 __napi_schedule(n); 446 } 447 448 /** 449 * napi_schedule_irqoff - schedule NAPI poll 450 * @n: NAPI context 451 * 452 * Variant of napi_schedule(), assuming hard irqs are masked. 453 */ 454 static inline void napi_schedule_irqoff(struct napi_struct *n) 455 { 456 if (napi_schedule_prep(n)) 457 __napi_schedule_irqoff(n); 458 } 459 460 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 461 static inline bool napi_reschedule(struct napi_struct *napi) 462 { 463 if (napi_schedule_prep(napi)) { 464 __napi_schedule(napi); 465 return true; 466 } 467 return false; 468 } 469 470 bool napi_complete_done(struct napi_struct *n, int work_done); 471 /** 472 * napi_complete - NAPI processing complete 473 * @n: NAPI context 474 * 475 * Mark NAPI processing as complete. 476 * Consider using napi_complete_done() instead. 477 * Return false if device should avoid rearming interrupts. 478 */ 479 static inline bool napi_complete(struct napi_struct *n) 480 { 481 return napi_complete_done(n, 0); 482 } 483 484 /** 485 * napi_hash_del - remove a NAPI from global table 486 * @napi: NAPI context 487 * 488 * Warning: caller must observe RCU grace period 489 * before freeing memory containing @napi, if 490 * this function returns true. 491 * Note: core networking stack automatically calls it 492 * from netif_napi_del(). 493 * Drivers might want to call this helper to combine all 494 * the needed RCU grace periods into a single one. 495 */ 496 bool napi_hash_del(struct napi_struct *napi); 497 498 /** 499 * napi_disable - prevent NAPI from scheduling 500 * @n: NAPI context 501 * 502 * Stop NAPI from being scheduled on this context. 503 * Waits till any outstanding processing completes. 504 */ 505 void napi_disable(struct napi_struct *n); 506 507 /** 508 * napi_enable - enable NAPI scheduling 509 * @n: NAPI context 510 * 511 * Resume NAPI from being scheduled on this context. 512 * Must be paired with napi_disable. 513 */ 514 static inline void napi_enable(struct napi_struct *n) 515 { 516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 517 smp_mb__before_atomic(); 518 clear_bit(NAPI_STATE_SCHED, &n->state); 519 clear_bit(NAPI_STATE_NPSVC, &n->state); 520 } 521 522 /** 523 * napi_synchronize - wait until NAPI is not running 524 * @n: NAPI context 525 * 526 * Wait until NAPI is done being scheduled on this context. 527 * Waits till any outstanding processing completes but 528 * does not disable future activations. 529 */ 530 static inline void napi_synchronize(const struct napi_struct *n) 531 { 532 if (IS_ENABLED(CONFIG_SMP)) 533 while (test_bit(NAPI_STATE_SCHED, &n->state)) 534 msleep(1); 535 else 536 barrier(); 537 } 538 539 /** 540 * napi_if_scheduled_mark_missed - if napi is running, set the 541 * NAPIF_STATE_MISSED 542 * @n: NAPI context 543 * 544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 545 * NAPI is scheduled. 546 **/ 547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 548 { 549 unsigned long val, new; 550 551 do { 552 val = READ_ONCE(n->state); 553 if (val & NAPIF_STATE_DISABLE) 554 return true; 555 556 if (!(val & NAPIF_STATE_SCHED)) 557 return false; 558 559 new = val | NAPIF_STATE_MISSED; 560 } while (cmpxchg(&n->state, val, new) != val); 561 562 return true; 563 } 564 565 enum netdev_queue_state_t { 566 __QUEUE_STATE_DRV_XOFF, 567 __QUEUE_STATE_STACK_XOFF, 568 __QUEUE_STATE_FROZEN, 569 }; 570 571 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 572 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 573 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 574 575 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 579 QUEUE_STATE_FROZEN) 580 581 /* 582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 583 * netif_tx_* functions below are used to manipulate this flag. The 584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 585 * queue independently. The netif_xmit_*stopped functions below are called 586 * to check if the queue has been stopped by the driver or stack (either 587 * of the XOFF bits are set in the state). Drivers should not need to call 588 * netif_xmit*stopped functions, they should only be using netif_tx_*. 589 */ 590 591 struct netdev_queue { 592 /* 593 * read-mostly part 594 */ 595 struct net_device *dev; 596 struct Qdisc __rcu *qdisc; 597 struct Qdisc *qdisc_sleeping; 598 #ifdef CONFIG_SYSFS 599 struct kobject kobj; 600 #endif 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 602 int numa_node; 603 #endif 604 unsigned long tx_maxrate; 605 /* 606 * Number of TX timeouts for this queue 607 * (/sys/class/net/DEV/Q/trans_timeout) 608 */ 609 unsigned long trans_timeout; 610 611 /* Subordinate device that the queue has been assigned to */ 612 struct net_device *sb_dev; 613 #ifdef CONFIG_XDP_SOCKETS 614 struct xdp_umem *umem; 615 #endif 616 /* 617 * write-mostly part 618 */ 619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 620 int xmit_lock_owner; 621 /* 622 * Time (in jiffies) of last Tx 623 */ 624 unsigned long trans_start; 625 626 unsigned long state; 627 628 #ifdef CONFIG_BQL 629 struct dql dql; 630 #endif 631 } ____cacheline_aligned_in_smp; 632 633 extern int sysctl_fb_tunnels_only_for_init_net; 634 extern int sysctl_devconf_inherit_init_net; 635 636 static inline bool net_has_fallback_tunnels(const struct net *net) 637 { 638 return net == &init_net || 639 !IS_ENABLED(CONFIG_SYSCTL) || 640 !sysctl_fb_tunnels_only_for_init_net; 641 } 642 643 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 644 { 645 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 646 return q->numa_node; 647 #else 648 return NUMA_NO_NODE; 649 #endif 650 } 651 652 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 653 { 654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 655 q->numa_node = node; 656 #endif 657 } 658 659 #ifdef CONFIG_RPS 660 /* 661 * This structure holds an RPS map which can be of variable length. The 662 * map is an array of CPUs. 663 */ 664 struct rps_map { 665 unsigned int len; 666 struct rcu_head rcu; 667 u16 cpus[0]; 668 }; 669 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 670 671 /* 672 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 673 * tail pointer for that CPU's input queue at the time of last enqueue, and 674 * a hardware filter index. 675 */ 676 struct rps_dev_flow { 677 u16 cpu; 678 u16 filter; 679 unsigned int last_qtail; 680 }; 681 #define RPS_NO_FILTER 0xffff 682 683 /* 684 * The rps_dev_flow_table structure contains a table of flow mappings. 685 */ 686 struct rps_dev_flow_table { 687 unsigned int mask; 688 struct rcu_head rcu; 689 struct rps_dev_flow flows[0]; 690 }; 691 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 692 ((_num) * sizeof(struct rps_dev_flow))) 693 694 /* 695 * The rps_sock_flow_table contains mappings of flows to the last CPU 696 * on which they were processed by the application (set in recvmsg). 697 * Each entry is a 32bit value. Upper part is the high-order bits 698 * of flow hash, lower part is CPU number. 699 * rps_cpu_mask is used to partition the space, depending on number of 700 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 701 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 702 * meaning we use 32-6=26 bits for the hash. 703 */ 704 struct rps_sock_flow_table { 705 u32 mask; 706 707 u32 ents[0] ____cacheline_aligned_in_smp; 708 }; 709 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 710 711 #define RPS_NO_CPU 0xffff 712 713 extern u32 rps_cpu_mask; 714 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 715 716 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 717 u32 hash) 718 { 719 if (table && hash) { 720 unsigned int index = hash & table->mask; 721 u32 val = hash & ~rps_cpu_mask; 722 723 /* We only give a hint, preemption can change CPU under us */ 724 val |= raw_smp_processor_id(); 725 726 if (table->ents[index] != val) 727 table->ents[index] = val; 728 } 729 } 730 731 #ifdef CONFIG_RFS_ACCEL 732 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 733 u16 filter_id); 734 #endif 735 #endif /* CONFIG_RPS */ 736 737 /* This structure contains an instance of an RX queue. */ 738 struct netdev_rx_queue { 739 #ifdef CONFIG_RPS 740 struct rps_map __rcu *rps_map; 741 struct rps_dev_flow_table __rcu *rps_flow_table; 742 #endif 743 struct kobject kobj; 744 struct net_device *dev; 745 struct xdp_rxq_info xdp_rxq; 746 #ifdef CONFIG_XDP_SOCKETS 747 struct xdp_umem *umem; 748 #endif 749 } ____cacheline_aligned_in_smp; 750 751 /* 752 * RX queue sysfs structures and functions. 753 */ 754 struct rx_queue_attribute { 755 struct attribute attr; 756 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 757 ssize_t (*store)(struct netdev_rx_queue *queue, 758 const char *buf, size_t len); 759 }; 760 761 #ifdef CONFIG_XPS 762 /* 763 * This structure holds an XPS map which can be of variable length. The 764 * map is an array of queues. 765 */ 766 struct xps_map { 767 unsigned int len; 768 unsigned int alloc_len; 769 struct rcu_head rcu; 770 u16 queues[0]; 771 }; 772 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 773 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 774 - sizeof(struct xps_map)) / sizeof(u16)) 775 776 /* 777 * This structure holds all XPS maps for device. Maps are indexed by CPU. 778 */ 779 struct xps_dev_maps { 780 struct rcu_head rcu; 781 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 782 }; 783 784 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 785 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 786 787 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 788 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 789 790 #endif /* CONFIG_XPS */ 791 792 #define TC_MAX_QUEUE 16 793 #define TC_BITMASK 15 794 /* HW offloaded queuing disciplines txq count and offset maps */ 795 struct netdev_tc_txq { 796 u16 count; 797 u16 offset; 798 }; 799 800 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 801 /* 802 * This structure is to hold information about the device 803 * configured to run FCoE protocol stack. 804 */ 805 struct netdev_fcoe_hbainfo { 806 char manufacturer[64]; 807 char serial_number[64]; 808 char hardware_version[64]; 809 char driver_version[64]; 810 char optionrom_version[64]; 811 char firmware_version[64]; 812 char model[256]; 813 char model_description[256]; 814 }; 815 #endif 816 817 #define MAX_PHYS_ITEM_ID_LEN 32 818 819 /* This structure holds a unique identifier to identify some 820 * physical item (port for example) used by a netdevice. 821 */ 822 struct netdev_phys_item_id { 823 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 824 unsigned char id_len; 825 }; 826 827 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 828 struct netdev_phys_item_id *b) 829 { 830 return a->id_len == b->id_len && 831 memcmp(a->id, b->id, a->id_len) == 0; 832 } 833 834 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 835 struct sk_buff *skb, 836 struct net_device *sb_dev); 837 838 enum tc_setup_type { 839 TC_SETUP_QDISC_MQPRIO, 840 TC_SETUP_CLSU32, 841 TC_SETUP_CLSFLOWER, 842 TC_SETUP_CLSMATCHALL, 843 TC_SETUP_CLSBPF, 844 TC_SETUP_BLOCK, 845 TC_SETUP_QDISC_CBS, 846 TC_SETUP_QDISC_RED, 847 TC_SETUP_QDISC_PRIO, 848 TC_SETUP_QDISC_MQ, 849 TC_SETUP_QDISC_ETF, 850 TC_SETUP_ROOT_QDISC, 851 TC_SETUP_QDISC_GRED, 852 TC_SETUP_QDISC_TAPRIO, 853 TC_SETUP_FT, 854 TC_SETUP_QDISC_ETS, 855 TC_SETUP_QDISC_TBF, 856 }; 857 858 /* These structures hold the attributes of bpf state that are being passed 859 * to the netdevice through the bpf op. 860 */ 861 enum bpf_netdev_command { 862 /* Set or clear a bpf program used in the earliest stages of packet 863 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 864 * is responsible for calling bpf_prog_put on any old progs that are 865 * stored. In case of error, the callee need not release the new prog 866 * reference, but on success it takes ownership and must bpf_prog_put 867 * when it is no longer used. 868 */ 869 XDP_SETUP_PROG, 870 XDP_SETUP_PROG_HW, 871 XDP_QUERY_PROG, 872 XDP_QUERY_PROG_HW, 873 /* BPF program for offload callbacks, invoked at program load time. */ 874 BPF_OFFLOAD_MAP_ALLOC, 875 BPF_OFFLOAD_MAP_FREE, 876 XDP_SETUP_XSK_UMEM, 877 }; 878 879 struct bpf_prog_offload_ops; 880 struct netlink_ext_ack; 881 struct xdp_umem; 882 struct xdp_dev_bulk_queue; 883 884 struct netdev_bpf { 885 enum bpf_netdev_command command; 886 union { 887 /* XDP_SETUP_PROG */ 888 struct { 889 u32 flags; 890 struct bpf_prog *prog; 891 struct netlink_ext_ack *extack; 892 }; 893 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 894 struct { 895 u32 prog_id; 896 /* flags with which program was installed */ 897 u32 prog_flags; 898 }; 899 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 900 struct { 901 struct bpf_offloaded_map *offmap; 902 }; 903 /* XDP_SETUP_XSK_UMEM */ 904 struct { 905 struct xdp_umem *umem; 906 u16 queue_id; 907 } xsk; 908 }; 909 }; 910 911 /* Flags for ndo_xsk_wakeup. */ 912 #define XDP_WAKEUP_RX (1 << 0) 913 #define XDP_WAKEUP_TX (1 << 1) 914 915 #ifdef CONFIG_XFRM_OFFLOAD 916 struct xfrmdev_ops { 917 int (*xdo_dev_state_add) (struct xfrm_state *x); 918 void (*xdo_dev_state_delete) (struct xfrm_state *x); 919 void (*xdo_dev_state_free) (struct xfrm_state *x); 920 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 921 struct xfrm_state *x); 922 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 923 }; 924 #endif 925 926 struct dev_ifalias { 927 struct rcu_head rcuhead; 928 char ifalias[]; 929 }; 930 931 struct devlink; 932 struct tlsdev_ops; 933 934 struct netdev_name_node { 935 struct hlist_node hlist; 936 struct list_head list; 937 struct net_device *dev; 938 const char *name; 939 }; 940 941 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 942 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 943 944 struct netdev_net_notifier { 945 struct list_head list; 946 struct notifier_block *nb; 947 }; 948 949 /* 950 * This structure defines the management hooks for network devices. 951 * The following hooks can be defined; unless noted otherwise, they are 952 * optional and can be filled with a null pointer. 953 * 954 * int (*ndo_init)(struct net_device *dev); 955 * This function is called once when a network device is registered. 956 * The network device can use this for any late stage initialization 957 * or semantic validation. It can fail with an error code which will 958 * be propagated back to register_netdev. 959 * 960 * void (*ndo_uninit)(struct net_device *dev); 961 * This function is called when device is unregistered or when registration 962 * fails. It is not called if init fails. 963 * 964 * int (*ndo_open)(struct net_device *dev); 965 * This function is called when a network device transitions to the up 966 * state. 967 * 968 * int (*ndo_stop)(struct net_device *dev); 969 * This function is called when a network device transitions to the down 970 * state. 971 * 972 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 973 * struct net_device *dev); 974 * Called when a packet needs to be transmitted. 975 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 976 * the queue before that can happen; it's for obsolete devices and weird 977 * corner cases, but the stack really does a non-trivial amount 978 * of useless work if you return NETDEV_TX_BUSY. 979 * Required; cannot be NULL. 980 * 981 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 982 * struct net_device *dev 983 * netdev_features_t features); 984 * Called by core transmit path to determine if device is capable of 985 * performing offload operations on a given packet. This is to give 986 * the device an opportunity to implement any restrictions that cannot 987 * be otherwise expressed by feature flags. The check is called with 988 * the set of features that the stack has calculated and it returns 989 * those the driver believes to be appropriate. 990 * 991 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 992 * struct net_device *sb_dev); 993 * Called to decide which queue to use when device supports multiple 994 * transmit queues. 995 * 996 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 997 * This function is called to allow device receiver to make 998 * changes to configuration when multicast or promiscuous is enabled. 999 * 1000 * void (*ndo_set_rx_mode)(struct net_device *dev); 1001 * This function is called device changes address list filtering. 1002 * If driver handles unicast address filtering, it should set 1003 * IFF_UNICAST_FLT in its priv_flags. 1004 * 1005 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1006 * This function is called when the Media Access Control address 1007 * needs to be changed. If this interface is not defined, the 1008 * MAC address can not be changed. 1009 * 1010 * int (*ndo_validate_addr)(struct net_device *dev); 1011 * Test if Media Access Control address is valid for the device. 1012 * 1013 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1014 * Called when a user requests an ioctl which can't be handled by 1015 * the generic interface code. If not defined ioctls return 1016 * not supported error code. 1017 * 1018 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1019 * Used to set network devices bus interface parameters. This interface 1020 * is retained for legacy reasons; new devices should use the bus 1021 * interface (PCI) for low level management. 1022 * 1023 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1024 * Called when a user wants to change the Maximum Transfer Unit 1025 * of a device. 1026 * 1027 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1028 * Callback used when the transmitter has not made any progress 1029 * for dev->watchdog ticks. 1030 * 1031 * void (*ndo_get_stats64)(struct net_device *dev, 1032 * struct rtnl_link_stats64 *storage); 1033 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1034 * Called when a user wants to get the network device usage 1035 * statistics. Drivers must do one of the following: 1036 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1037 * rtnl_link_stats64 structure passed by the caller. 1038 * 2. Define @ndo_get_stats to update a net_device_stats structure 1039 * (which should normally be dev->stats) and return a pointer to 1040 * it. The structure may be changed asynchronously only if each 1041 * field is written atomically. 1042 * 3. Update dev->stats asynchronously and atomically, and define 1043 * neither operation. 1044 * 1045 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1046 * Return true if this device supports offload stats of this attr_id. 1047 * 1048 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1049 * void *attr_data) 1050 * Get statistics for offload operations by attr_id. Write it into the 1051 * attr_data pointer. 1052 * 1053 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1054 * If device supports VLAN filtering this function is called when a 1055 * VLAN id is registered. 1056 * 1057 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1058 * If device supports VLAN filtering this function is called when a 1059 * VLAN id is unregistered. 1060 * 1061 * void (*ndo_poll_controller)(struct net_device *dev); 1062 * 1063 * SR-IOV management functions. 1064 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1065 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1066 * u8 qos, __be16 proto); 1067 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1068 * int max_tx_rate); 1069 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1070 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1071 * int (*ndo_get_vf_config)(struct net_device *dev, 1072 * int vf, struct ifla_vf_info *ivf); 1073 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1074 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1075 * struct nlattr *port[]); 1076 * 1077 * Enable or disable the VF ability to query its RSS Redirection Table and 1078 * Hash Key. This is needed since on some devices VF share this information 1079 * with PF and querying it may introduce a theoretical security risk. 1080 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1081 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1082 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1083 * void *type_data); 1084 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1085 * This is always called from the stack with the rtnl lock held and netif 1086 * tx queues stopped. This allows the netdevice to perform queue 1087 * management safely. 1088 * 1089 * Fiber Channel over Ethernet (FCoE) offload functions. 1090 * int (*ndo_fcoe_enable)(struct net_device *dev); 1091 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1092 * so the underlying device can perform whatever needed configuration or 1093 * initialization to support acceleration of FCoE traffic. 1094 * 1095 * int (*ndo_fcoe_disable)(struct net_device *dev); 1096 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1097 * so the underlying device can perform whatever needed clean-ups to 1098 * stop supporting acceleration of FCoE traffic. 1099 * 1100 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1101 * struct scatterlist *sgl, unsigned int sgc); 1102 * Called when the FCoE Initiator wants to initialize an I/O that 1103 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1104 * perform necessary setup and returns 1 to indicate the device is set up 1105 * successfully to perform DDP on this I/O, otherwise this returns 0. 1106 * 1107 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1108 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1109 * indicated by the FC exchange id 'xid', so the underlying device can 1110 * clean up and reuse resources for later DDP requests. 1111 * 1112 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1113 * struct scatterlist *sgl, unsigned int sgc); 1114 * Called when the FCoE Target wants to initialize an I/O that 1115 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1116 * perform necessary setup and returns 1 to indicate the device is set up 1117 * successfully to perform DDP on this I/O, otherwise this returns 0. 1118 * 1119 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1120 * struct netdev_fcoe_hbainfo *hbainfo); 1121 * Called when the FCoE Protocol stack wants information on the underlying 1122 * device. This information is utilized by the FCoE protocol stack to 1123 * register attributes with Fiber Channel management service as per the 1124 * FC-GS Fabric Device Management Information(FDMI) specification. 1125 * 1126 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1127 * Called when the underlying device wants to override default World Wide 1128 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1129 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1130 * protocol stack to use. 1131 * 1132 * RFS acceleration. 1133 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1134 * u16 rxq_index, u32 flow_id); 1135 * Set hardware filter for RFS. rxq_index is the target queue index; 1136 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1137 * Return the filter ID on success, or a negative error code. 1138 * 1139 * Slave management functions (for bridge, bonding, etc). 1140 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1141 * Called to make another netdev an underling. 1142 * 1143 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1144 * Called to release previously enslaved netdev. 1145 * 1146 * Feature/offload setting functions. 1147 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1148 * netdev_features_t features); 1149 * Adjusts the requested feature flags according to device-specific 1150 * constraints, and returns the resulting flags. Must not modify 1151 * the device state. 1152 * 1153 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1154 * Called to update device configuration to new features. Passed 1155 * feature set might be less than what was returned by ndo_fix_features()). 1156 * Must return >0 or -errno if it changed dev->features itself. 1157 * 1158 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1159 * struct net_device *dev, 1160 * const unsigned char *addr, u16 vid, u16 flags, 1161 * struct netlink_ext_ack *extack); 1162 * Adds an FDB entry to dev for addr. 1163 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1164 * struct net_device *dev, 1165 * const unsigned char *addr, u16 vid) 1166 * Deletes the FDB entry from dev coresponding to addr. 1167 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1168 * struct net_device *dev, struct net_device *filter_dev, 1169 * int *idx) 1170 * Used to add FDB entries to dump requests. Implementers should add 1171 * entries to skb and update idx with the number of entries. 1172 * 1173 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1174 * u16 flags, struct netlink_ext_ack *extack) 1175 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1176 * struct net_device *dev, u32 filter_mask, 1177 * int nlflags) 1178 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1179 * u16 flags); 1180 * 1181 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1182 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1183 * which do not represent real hardware may define this to allow their 1184 * userspace components to manage their virtual carrier state. Devices 1185 * that determine carrier state from physical hardware properties (eg 1186 * network cables) or protocol-dependent mechanisms (eg 1187 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1188 * 1189 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1190 * struct netdev_phys_item_id *ppid); 1191 * Called to get ID of physical port of this device. If driver does 1192 * not implement this, it is assumed that the hw is not able to have 1193 * multiple net devices on single physical port. 1194 * 1195 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1196 * struct netdev_phys_item_id *ppid) 1197 * Called to get the parent ID of the physical port of this device. 1198 * 1199 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1200 * struct udp_tunnel_info *ti); 1201 * Called by UDP tunnel to notify a driver about the UDP port and socket 1202 * address family that a UDP tunnel is listnening to. It is called only 1203 * when a new port starts listening. The operation is protected by the 1204 * RTNL. 1205 * 1206 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1207 * struct udp_tunnel_info *ti); 1208 * Called by UDP tunnel to notify the driver about a UDP port and socket 1209 * address family that the UDP tunnel is not listening to anymore. The 1210 * operation is protected by the RTNL. 1211 * 1212 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1213 * struct net_device *dev) 1214 * Called by upper layer devices to accelerate switching or other 1215 * station functionality into hardware. 'pdev is the lowerdev 1216 * to use for the offload and 'dev' is the net device that will 1217 * back the offload. Returns a pointer to the private structure 1218 * the upper layer will maintain. 1219 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1220 * Called by upper layer device to delete the station created 1221 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1222 * the station and priv is the structure returned by the add 1223 * operation. 1224 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1225 * int queue_index, u32 maxrate); 1226 * Called when a user wants to set a max-rate limitation of specific 1227 * TX queue. 1228 * int (*ndo_get_iflink)(const struct net_device *dev); 1229 * Called to get the iflink value of this device. 1230 * void (*ndo_change_proto_down)(struct net_device *dev, 1231 * bool proto_down); 1232 * This function is used to pass protocol port error state information 1233 * to the switch driver. The switch driver can react to the proto_down 1234 * by doing a phys down on the associated switch port. 1235 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1236 * This function is used to get egress tunnel information for given skb. 1237 * This is useful for retrieving outer tunnel header parameters while 1238 * sampling packet. 1239 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1240 * This function is used to specify the headroom that the skb must 1241 * consider when allocation skb during packet reception. Setting 1242 * appropriate rx headroom value allows avoiding skb head copy on 1243 * forward. Setting a negative value resets the rx headroom to the 1244 * default value. 1245 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1246 * This function is used to set or query state related to XDP on the 1247 * netdevice and manage BPF offload. See definition of 1248 * enum bpf_netdev_command for details. 1249 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1250 * u32 flags); 1251 * This function is used to submit @n XDP packets for transmit on a 1252 * netdevice. Returns number of frames successfully transmitted, frames 1253 * that got dropped are freed/returned via xdp_return_frame(). 1254 * Returns negative number, means general error invoking ndo, meaning 1255 * no frames were xmit'ed and core-caller will free all frames. 1256 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1257 * This function is used to wake up the softirq, ksoftirqd or kthread 1258 * responsible for sending and/or receiving packets on a specific 1259 * queue id bound to an AF_XDP socket. The flags field specifies if 1260 * only RX, only Tx, or both should be woken up using the flags 1261 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1262 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1263 * Get devlink port instance associated with a given netdev. 1264 * Called with a reference on the netdevice and devlink locks only, 1265 * rtnl_lock is not held. 1266 */ 1267 struct net_device_ops { 1268 int (*ndo_init)(struct net_device *dev); 1269 void (*ndo_uninit)(struct net_device *dev); 1270 int (*ndo_open)(struct net_device *dev); 1271 int (*ndo_stop)(struct net_device *dev); 1272 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1273 struct net_device *dev); 1274 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1275 struct net_device *dev, 1276 netdev_features_t features); 1277 u16 (*ndo_select_queue)(struct net_device *dev, 1278 struct sk_buff *skb, 1279 struct net_device *sb_dev); 1280 void (*ndo_change_rx_flags)(struct net_device *dev, 1281 int flags); 1282 void (*ndo_set_rx_mode)(struct net_device *dev); 1283 int (*ndo_set_mac_address)(struct net_device *dev, 1284 void *addr); 1285 int (*ndo_validate_addr)(struct net_device *dev); 1286 int (*ndo_do_ioctl)(struct net_device *dev, 1287 struct ifreq *ifr, int cmd); 1288 int (*ndo_set_config)(struct net_device *dev, 1289 struct ifmap *map); 1290 int (*ndo_change_mtu)(struct net_device *dev, 1291 int new_mtu); 1292 int (*ndo_neigh_setup)(struct net_device *dev, 1293 struct neigh_parms *); 1294 void (*ndo_tx_timeout) (struct net_device *dev, 1295 unsigned int txqueue); 1296 1297 void (*ndo_get_stats64)(struct net_device *dev, 1298 struct rtnl_link_stats64 *storage); 1299 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1300 int (*ndo_get_offload_stats)(int attr_id, 1301 const struct net_device *dev, 1302 void *attr_data); 1303 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1304 1305 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1306 __be16 proto, u16 vid); 1307 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1308 __be16 proto, u16 vid); 1309 #ifdef CONFIG_NET_POLL_CONTROLLER 1310 void (*ndo_poll_controller)(struct net_device *dev); 1311 int (*ndo_netpoll_setup)(struct net_device *dev, 1312 struct netpoll_info *info); 1313 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1314 #endif 1315 int (*ndo_set_vf_mac)(struct net_device *dev, 1316 int queue, u8 *mac); 1317 int (*ndo_set_vf_vlan)(struct net_device *dev, 1318 int queue, u16 vlan, 1319 u8 qos, __be16 proto); 1320 int (*ndo_set_vf_rate)(struct net_device *dev, 1321 int vf, int min_tx_rate, 1322 int max_tx_rate); 1323 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1324 int vf, bool setting); 1325 int (*ndo_set_vf_trust)(struct net_device *dev, 1326 int vf, bool setting); 1327 int (*ndo_get_vf_config)(struct net_device *dev, 1328 int vf, 1329 struct ifla_vf_info *ivf); 1330 int (*ndo_set_vf_link_state)(struct net_device *dev, 1331 int vf, int link_state); 1332 int (*ndo_get_vf_stats)(struct net_device *dev, 1333 int vf, 1334 struct ifla_vf_stats 1335 *vf_stats); 1336 int (*ndo_set_vf_port)(struct net_device *dev, 1337 int vf, 1338 struct nlattr *port[]); 1339 int (*ndo_get_vf_port)(struct net_device *dev, 1340 int vf, struct sk_buff *skb); 1341 int (*ndo_get_vf_guid)(struct net_device *dev, 1342 int vf, 1343 struct ifla_vf_guid *node_guid, 1344 struct ifla_vf_guid *port_guid); 1345 int (*ndo_set_vf_guid)(struct net_device *dev, 1346 int vf, u64 guid, 1347 int guid_type); 1348 int (*ndo_set_vf_rss_query_en)( 1349 struct net_device *dev, 1350 int vf, bool setting); 1351 int (*ndo_setup_tc)(struct net_device *dev, 1352 enum tc_setup_type type, 1353 void *type_data); 1354 #if IS_ENABLED(CONFIG_FCOE) 1355 int (*ndo_fcoe_enable)(struct net_device *dev); 1356 int (*ndo_fcoe_disable)(struct net_device *dev); 1357 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1358 u16 xid, 1359 struct scatterlist *sgl, 1360 unsigned int sgc); 1361 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1362 u16 xid); 1363 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1364 u16 xid, 1365 struct scatterlist *sgl, 1366 unsigned int sgc); 1367 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1368 struct netdev_fcoe_hbainfo *hbainfo); 1369 #endif 1370 1371 #if IS_ENABLED(CONFIG_LIBFCOE) 1372 #define NETDEV_FCOE_WWNN 0 1373 #define NETDEV_FCOE_WWPN 1 1374 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1375 u64 *wwn, int type); 1376 #endif 1377 1378 #ifdef CONFIG_RFS_ACCEL 1379 int (*ndo_rx_flow_steer)(struct net_device *dev, 1380 const struct sk_buff *skb, 1381 u16 rxq_index, 1382 u32 flow_id); 1383 #endif 1384 int (*ndo_add_slave)(struct net_device *dev, 1385 struct net_device *slave_dev, 1386 struct netlink_ext_ack *extack); 1387 int (*ndo_del_slave)(struct net_device *dev, 1388 struct net_device *slave_dev); 1389 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1390 netdev_features_t features); 1391 int (*ndo_set_features)(struct net_device *dev, 1392 netdev_features_t features); 1393 int (*ndo_neigh_construct)(struct net_device *dev, 1394 struct neighbour *n); 1395 void (*ndo_neigh_destroy)(struct net_device *dev, 1396 struct neighbour *n); 1397 1398 int (*ndo_fdb_add)(struct ndmsg *ndm, 1399 struct nlattr *tb[], 1400 struct net_device *dev, 1401 const unsigned char *addr, 1402 u16 vid, 1403 u16 flags, 1404 struct netlink_ext_ack *extack); 1405 int (*ndo_fdb_del)(struct ndmsg *ndm, 1406 struct nlattr *tb[], 1407 struct net_device *dev, 1408 const unsigned char *addr, 1409 u16 vid); 1410 int (*ndo_fdb_dump)(struct sk_buff *skb, 1411 struct netlink_callback *cb, 1412 struct net_device *dev, 1413 struct net_device *filter_dev, 1414 int *idx); 1415 int (*ndo_fdb_get)(struct sk_buff *skb, 1416 struct nlattr *tb[], 1417 struct net_device *dev, 1418 const unsigned char *addr, 1419 u16 vid, u32 portid, u32 seq, 1420 struct netlink_ext_ack *extack); 1421 int (*ndo_bridge_setlink)(struct net_device *dev, 1422 struct nlmsghdr *nlh, 1423 u16 flags, 1424 struct netlink_ext_ack *extack); 1425 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1426 u32 pid, u32 seq, 1427 struct net_device *dev, 1428 u32 filter_mask, 1429 int nlflags); 1430 int (*ndo_bridge_dellink)(struct net_device *dev, 1431 struct nlmsghdr *nlh, 1432 u16 flags); 1433 int (*ndo_change_carrier)(struct net_device *dev, 1434 bool new_carrier); 1435 int (*ndo_get_phys_port_id)(struct net_device *dev, 1436 struct netdev_phys_item_id *ppid); 1437 int (*ndo_get_port_parent_id)(struct net_device *dev, 1438 struct netdev_phys_item_id *ppid); 1439 int (*ndo_get_phys_port_name)(struct net_device *dev, 1440 char *name, size_t len); 1441 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1442 struct udp_tunnel_info *ti); 1443 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1444 struct udp_tunnel_info *ti); 1445 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1446 struct net_device *dev); 1447 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1448 void *priv); 1449 1450 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1451 int queue_index, 1452 u32 maxrate); 1453 int (*ndo_get_iflink)(const struct net_device *dev); 1454 int (*ndo_change_proto_down)(struct net_device *dev, 1455 bool proto_down); 1456 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1457 struct sk_buff *skb); 1458 void (*ndo_set_rx_headroom)(struct net_device *dev, 1459 int needed_headroom); 1460 int (*ndo_bpf)(struct net_device *dev, 1461 struct netdev_bpf *bpf); 1462 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1463 struct xdp_frame **xdp, 1464 u32 flags); 1465 int (*ndo_xsk_wakeup)(struct net_device *dev, 1466 u32 queue_id, u32 flags); 1467 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1468 }; 1469 1470 /** 1471 * enum net_device_priv_flags - &struct net_device priv_flags 1472 * 1473 * These are the &struct net_device, they are only set internally 1474 * by drivers and used in the kernel. These flags are invisible to 1475 * userspace; this means that the order of these flags can change 1476 * during any kernel release. 1477 * 1478 * You should have a pretty good reason to be extending these flags. 1479 * 1480 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1481 * @IFF_EBRIDGE: Ethernet bridging device 1482 * @IFF_BONDING: bonding master or slave 1483 * @IFF_ISATAP: ISATAP interface (RFC4214) 1484 * @IFF_WAN_HDLC: WAN HDLC device 1485 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1486 * release skb->dst 1487 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1488 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1489 * @IFF_MACVLAN_PORT: device used as macvlan port 1490 * @IFF_BRIDGE_PORT: device used as bridge port 1491 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1492 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1493 * @IFF_UNICAST_FLT: Supports unicast filtering 1494 * @IFF_TEAM_PORT: device used as team port 1495 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1496 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1497 * change when it's running 1498 * @IFF_MACVLAN: Macvlan device 1499 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1500 * underlying stacked devices 1501 * @IFF_L3MDEV_MASTER: device is an L3 master device 1502 * @IFF_NO_QUEUE: device can run without qdisc attached 1503 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1504 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1505 * @IFF_TEAM: device is a team device 1506 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1507 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1508 * entity (i.e. the master device for bridged veth) 1509 * @IFF_MACSEC: device is a MACsec device 1510 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1511 * @IFF_FAILOVER: device is a failover master device 1512 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1513 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1514 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1515 */ 1516 enum netdev_priv_flags { 1517 IFF_802_1Q_VLAN = 1<<0, 1518 IFF_EBRIDGE = 1<<1, 1519 IFF_BONDING = 1<<2, 1520 IFF_ISATAP = 1<<3, 1521 IFF_WAN_HDLC = 1<<4, 1522 IFF_XMIT_DST_RELEASE = 1<<5, 1523 IFF_DONT_BRIDGE = 1<<6, 1524 IFF_DISABLE_NETPOLL = 1<<7, 1525 IFF_MACVLAN_PORT = 1<<8, 1526 IFF_BRIDGE_PORT = 1<<9, 1527 IFF_OVS_DATAPATH = 1<<10, 1528 IFF_TX_SKB_SHARING = 1<<11, 1529 IFF_UNICAST_FLT = 1<<12, 1530 IFF_TEAM_PORT = 1<<13, 1531 IFF_SUPP_NOFCS = 1<<14, 1532 IFF_LIVE_ADDR_CHANGE = 1<<15, 1533 IFF_MACVLAN = 1<<16, 1534 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1535 IFF_L3MDEV_MASTER = 1<<18, 1536 IFF_NO_QUEUE = 1<<19, 1537 IFF_OPENVSWITCH = 1<<20, 1538 IFF_L3MDEV_SLAVE = 1<<21, 1539 IFF_TEAM = 1<<22, 1540 IFF_RXFH_CONFIGURED = 1<<23, 1541 IFF_PHONY_HEADROOM = 1<<24, 1542 IFF_MACSEC = 1<<25, 1543 IFF_NO_RX_HANDLER = 1<<26, 1544 IFF_FAILOVER = 1<<27, 1545 IFF_FAILOVER_SLAVE = 1<<28, 1546 IFF_L3MDEV_RX_HANDLER = 1<<29, 1547 IFF_LIVE_RENAME_OK = 1<<30, 1548 }; 1549 1550 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1551 #define IFF_EBRIDGE IFF_EBRIDGE 1552 #define IFF_BONDING IFF_BONDING 1553 #define IFF_ISATAP IFF_ISATAP 1554 #define IFF_WAN_HDLC IFF_WAN_HDLC 1555 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1556 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1557 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1558 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1559 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1560 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1561 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1562 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1563 #define IFF_TEAM_PORT IFF_TEAM_PORT 1564 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1565 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1566 #define IFF_MACVLAN IFF_MACVLAN 1567 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1568 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1569 #define IFF_NO_QUEUE IFF_NO_QUEUE 1570 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1571 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1572 #define IFF_TEAM IFF_TEAM 1573 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1574 #define IFF_MACSEC IFF_MACSEC 1575 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1576 #define IFF_FAILOVER IFF_FAILOVER 1577 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1578 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1579 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1580 1581 /** 1582 * struct net_device - The DEVICE structure. 1583 * 1584 * Actually, this whole structure is a big mistake. It mixes I/O 1585 * data with strictly "high-level" data, and it has to know about 1586 * almost every data structure used in the INET module. 1587 * 1588 * @name: This is the first field of the "visible" part of this structure 1589 * (i.e. as seen by users in the "Space.c" file). It is the name 1590 * of the interface. 1591 * 1592 * @name_node: Name hashlist node 1593 * @ifalias: SNMP alias 1594 * @mem_end: Shared memory end 1595 * @mem_start: Shared memory start 1596 * @base_addr: Device I/O address 1597 * @irq: Device IRQ number 1598 * 1599 * @state: Generic network queuing layer state, see netdev_state_t 1600 * @dev_list: The global list of network devices 1601 * @napi_list: List entry used for polling NAPI devices 1602 * @unreg_list: List entry when we are unregistering the 1603 * device; see the function unregister_netdev 1604 * @close_list: List entry used when we are closing the device 1605 * @ptype_all: Device-specific packet handlers for all protocols 1606 * @ptype_specific: Device-specific, protocol-specific packet handlers 1607 * 1608 * @adj_list: Directly linked devices, like slaves for bonding 1609 * @features: Currently active device features 1610 * @hw_features: User-changeable features 1611 * 1612 * @wanted_features: User-requested features 1613 * @vlan_features: Mask of features inheritable by VLAN devices 1614 * 1615 * @hw_enc_features: Mask of features inherited by encapsulating devices 1616 * This field indicates what encapsulation 1617 * offloads the hardware is capable of doing, 1618 * and drivers will need to set them appropriately. 1619 * 1620 * @mpls_features: Mask of features inheritable by MPLS 1621 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1622 * 1623 * @ifindex: interface index 1624 * @group: The group the device belongs to 1625 * 1626 * @stats: Statistics struct, which was left as a legacy, use 1627 * rtnl_link_stats64 instead 1628 * 1629 * @rx_dropped: Dropped packets by core network, 1630 * do not use this in drivers 1631 * @tx_dropped: Dropped packets by core network, 1632 * do not use this in drivers 1633 * @rx_nohandler: nohandler dropped packets by core network on 1634 * inactive devices, do not use this in drivers 1635 * @carrier_up_count: Number of times the carrier has been up 1636 * @carrier_down_count: Number of times the carrier has been down 1637 * 1638 * @wireless_handlers: List of functions to handle Wireless Extensions, 1639 * instead of ioctl, 1640 * see <net/iw_handler.h> for details. 1641 * @wireless_data: Instance data managed by the core of wireless extensions 1642 * 1643 * @netdev_ops: Includes several pointers to callbacks, 1644 * if one wants to override the ndo_*() functions 1645 * @ethtool_ops: Management operations 1646 * @l3mdev_ops: Layer 3 master device operations 1647 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1648 * discovery handling. Necessary for e.g. 6LoWPAN. 1649 * @xfrmdev_ops: Transformation offload operations 1650 * @tlsdev_ops: Transport Layer Security offload operations 1651 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1652 * of Layer 2 headers. 1653 * 1654 * @flags: Interface flags (a la BSD) 1655 * @priv_flags: Like 'flags' but invisible to userspace, 1656 * see if.h for the definitions 1657 * @gflags: Global flags ( kept as legacy ) 1658 * @padded: How much padding added by alloc_netdev() 1659 * @operstate: RFC2863 operstate 1660 * @link_mode: Mapping policy to operstate 1661 * @if_port: Selectable AUI, TP, ... 1662 * @dma: DMA channel 1663 * @mtu: Interface MTU value 1664 * @min_mtu: Interface Minimum MTU value 1665 * @max_mtu: Interface Maximum MTU value 1666 * @type: Interface hardware type 1667 * @hard_header_len: Maximum hardware header length. 1668 * @min_header_len: Minimum hardware header length 1669 * 1670 * @needed_headroom: Extra headroom the hardware may need, but not in all 1671 * cases can this be guaranteed 1672 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1673 * cases can this be guaranteed. Some cases also use 1674 * LL_MAX_HEADER instead to allocate the skb 1675 * 1676 * interface address info: 1677 * 1678 * @perm_addr: Permanent hw address 1679 * @addr_assign_type: Hw address assignment type 1680 * @addr_len: Hardware address length 1681 * @upper_level: Maximum depth level of upper devices. 1682 * @lower_level: Maximum depth level of lower devices. 1683 * @neigh_priv_len: Used in neigh_alloc() 1684 * @dev_id: Used to differentiate devices that share 1685 * the same link layer address 1686 * @dev_port: Used to differentiate devices that share 1687 * the same function 1688 * @addr_list_lock: XXX: need comments on this one 1689 * @name_assign_type: network interface name assignment type 1690 * @uc_promisc: Counter that indicates promiscuous mode 1691 * has been enabled due to the need to listen to 1692 * additional unicast addresses in a device that 1693 * does not implement ndo_set_rx_mode() 1694 * @uc: unicast mac addresses 1695 * @mc: multicast mac addresses 1696 * @dev_addrs: list of device hw addresses 1697 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1698 * @promiscuity: Number of times the NIC is told to work in 1699 * promiscuous mode; if it becomes 0 the NIC will 1700 * exit promiscuous mode 1701 * @allmulti: Counter, enables or disables allmulticast mode 1702 * 1703 * @vlan_info: VLAN info 1704 * @dsa_ptr: dsa specific data 1705 * @tipc_ptr: TIPC specific data 1706 * @atalk_ptr: AppleTalk link 1707 * @ip_ptr: IPv4 specific data 1708 * @dn_ptr: DECnet specific data 1709 * @ip6_ptr: IPv6 specific data 1710 * @ax25_ptr: AX.25 specific data 1711 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1712 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1713 * device struct 1714 * @mpls_ptr: mpls_dev struct pointer 1715 * 1716 * @dev_addr: Hw address (before bcast, 1717 * because most packets are unicast) 1718 * 1719 * @_rx: Array of RX queues 1720 * @num_rx_queues: Number of RX queues 1721 * allocated at register_netdev() time 1722 * @real_num_rx_queues: Number of RX queues currently active in device 1723 * @xdp_prog: XDP sockets filter program pointer 1724 * @gro_flush_timeout: timeout for GRO layer in NAPI 1725 * 1726 * @rx_handler: handler for received packets 1727 * @rx_handler_data: XXX: need comments on this one 1728 * @miniq_ingress: ingress/clsact qdisc specific data for 1729 * ingress processing 1730 * @ingress_queue: XXX: need comments on this one 1731 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1732 * @broadcast: hw bcast address 1733 * 1734 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1735 * indexed by RX queue number. Assigned by driver. 1736 * This must only be set if the ndo_rx_flow_steer 1737 * operation is defined 1738 * @index_hlist: Device index hash chain 1739 * 1740 * @_tx: Array of TX queues 1741 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1742 * @real_num_tx_queues: Number of TX queues currently active in device 1743 * @qdisc: Root qdisc from userspace point of view 1744 * @tx_queue_len: Max frames per queue allowed 1745 * @tx_global_lock: XXX: need comments on this one 1746 * @xdp_bulkq: XDP device bulk queue 1747 * @xps_cpus_map: all CPUs map for XPS device 1748 * @xps_rxqs_map: all RXQs map for XPS device 1749 * 1750 * @xps_maps: XXX: need comments on this one 1751 * @miniq_egress: clsact qdisc specific data for 1752 * egress processing 1753 * @qdisc_hash: qdisc hash table 1754 * @watchdog_timeo: Represents the timeout that is used by 1755 * the watchdog (see dev_watchdog()) 1756 * @watchdog_timer: List of timers 1757 * 1758 * @pcpu_refcnt: Number of references to this device 1759 * @todo_list: Delayed register/unregister 1760 * @link_watch_list: XXX: need comments on this one 1761 * 1762 * @reg_state: Register/unregister state machine 1763 * @dismantle: Device is going to be freed 1764 * @rtnl_link_state: This enum represents the phases of creating 1765 * a new link 1766 * 1767 * @needs_free_netdev: Should unregister perform free_netdev? 1768 * @priv_destructor: Called from unregister 1769 * @npinfo: XXX: need comments on this one 1770 * @nd_net: Network namespace this network device is inside 1771 * 1772 * @ml_priv: Mid-layer private 1773 * @lstats: Loopback statistics 1774 * @tstats: Tunnel statistics 1775 * @dstats: Dummy statistics 1776 * @vstats: Virtual ethernet statistics 1777 * 1778 * @garp_port: GARP 1779 * @mrp_port: MRP 1780 * 1781 * @dev: Class/net/name entry 1782 * @sysfs_groups: Space for optional device, statistics and wireless 1783 * sysfs groups 1784 * 1785 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1786 * @rtnl_link_ops: Rtnl_link_ops 1787 * 1788 * @gso_max_size: Maximum size of generic segmentation offload 1789 * @gso_max_segs: Maximum number of segments that can be passed to the 1790 * NIC for GSO 1791 * 1792 * @dcbnl_ops: Data Center Bridging netlink ops 1793 * @num_tc: Number of traffic classes in the net device 1794 * @tc_to_txq: XXX: need comments on this one 1795 * @prio_tc_map: XXX: need comments on this one 1796 * 1797 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1798 * 1799 * @priomap: XXX: need comments on this one 1800 * @phydev: Physical device may attach itself 1801 * for hardware timestamping 1802 * @sfp_bus: attached &struct sfp_bus structure. 1803 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock 1804 * spinlock 1805 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1806 * @qdisc_xmit_lock_key: lockdep class annotating 1807 * netdev_queue->_xmit_lock spinlock 1808 * @addr_list_lock_key: lockdep class annotating 1809 * net_device->addr_list_lock spinlock 1810 * 1811 * @proto_down: protocol port state information can be sent to the 1812 * switch driver and used to set the phys state of the 1813 * switch port. 1814 * 1815 * @wol_enabled: Wake-on-LAN is enabled 1816 * 1817 * @net_notifier_list: List of per-net netdev notifier block 1818 * that follow this device when it is moved 1819 * to another network namespace. 1820 * 1821 * FIXME: cleanup struct net_device such that network protocol info 1822 * moves out. 1823 */ 1824 1825 struct net_device { 1826 char name[IFNAMSIZ]; 1827 struct netdev_name_node *name_node; 1828 struct dev_ifalias __rcu *ifalias; 1829 /* 1830 * I/O specific fields 1831 * FIXME: Merge these and struct ifmap into one 1832 */ 1833 unsigned long mem_end; 1834 unsigned long mem_start; 1835 unsigned long base_addr; 1836 int irq; 1837 1838 /* 1839 * Some hardware also needs these fields (state,dev_list, 1840 * napi_list,unreg_list,close_list) but they are not 1841 * part of the usual set specified in Space.c. 1842 */ 1843 1844 unsigned long state; 1845 1846 struct list_head dev_list; 1847 struct list_head napi_list; 1848 struct list_head unreg_list; 1849 struct list_head close_list; 1850 struct list_head ptype_all; 1851 struct list_head ptype_specific; 1852 1853 struct { 1854 struct list_head upper; 1855 struct list_head lower; 1856 } adj_list; 1857 1858 netdev_features_t features; 1859 netdev_features_t hw_features; 1860 netdev_features_t wanted_features; 1861 netdev_features_t vlan_features; 1862 netdev_features_t hw_enc_features; 1863 netdev_features_t mpls_features; 1864 netdev_features_t gso_partial_features; 1865 1866 int ifindex; 1867 int group; 1868 1869 struct net_device_stats stats; 1870 1871 atomic_long_t rx_dropped; 1872 atomic_long_t tx_dropped; 1873 atomic_long_t rx_nohandler; 1874 1875 /* Stats to monitor link on/off, flapping */ 1876 atomic_t carrier_up_count; 1877 atomic_t carrier_down_count; 1878 1879 #ifdef CONFIG_WIRELESS_EXT 1880 const struct iw_handler_def *wireless_handlers; 1881 struct iw_public_data *wireless_data; 1882 #endif 1883 const struct net_device_ops *netdev_ops; 1884 const struct ethtool_ops *ethtool_ops; 1885 #ifdef CONFIG_NET_L3_MASTER_DEV 1886 const struct l3mdev_ops *l3mdev_ops; 1887 #endif 1888 #if IS_ENABLED(CONFIG_IPV6) 1889 const struct ndisc_ops *ndisc_ops; 1890 #endif 1891 1892 #ifdef CONFIG_XFRM_OFFLOAD 1893 const struct xfrmdev_ops *xfrmdev_ops; 1894 #endif 1895 1896 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1897 const struct tlsdev_ops *tlsdev_ops; 1898 #endif 1899 1900 const struct header_ops *header_ops; 1901 1902 unsigned int flags; 1903 unsigned int priv_flags; 1904 1905 unsigned short gflags; 1906 unsigned short padded; 1907 1908 unsigned char operstate; 1909 unsigned char link_mode; 1910 1911 unsigned char if_port; 1912 unsigned char dma; 1913 1914 /* Note : dev->mtu is often read without holding a lock. 1915 * Writers usually hold RTNL. 1916 * It is recommended to use READ_ONCE() to annotate the reads, 1917 * and to use WRITE_ONCE() to annotate the writes. 1918 */ 1919 unsigned int mtu; 1920 unsigned int min_mtu; 1921 unsigned int max_mtu; 1922 unsigned short type; 1923 unsigned short hard_header_len; 1924 unsigned char min_header_len; 1925 1926 unsigned short needed_headroom; 1927 unsigned short needed_tailroom; 1928 1929 /* Interface address info. */ 1930 unsigned char perm_addr[MAX_ADDR_LEN]; 1931 unsigned char addr_assign_type; 1932 unsigned char addr_len; 1933 unsigned char upper_level; 1934 unsigned char lower_level; 1935 unsigned short neigh_priv_len; 1936 unsigned short dev_id; 1937 unsigned short dev_port; 1938 spinlock_t addr_list_lock; 1939 unsigned char name_assign_type; 1940 bool uc_promisc; 1941 struct netdev_hw_addr_list uc; 1942 struct netdev_hw_addr_list mc; 1943 struct netdev_hw_addr_list dev_addrs; 1944 1945 #ifdef CONFIG_SYSFS 1946 struct kset *queues_kset; 1947 #endif 1948 unsigned int promiscuity; 1949 unsigned int allmulti; 1950 1951 1952 /* Protocol-specific pointers */ 1953 1954 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1955 struct vlan_info __rcu *vlan_info; 1956 #endif 1957 #if IS_ENABLED(CONFIG_NET_DSA) 1958 struct dsa_port *dsa_ptr; 1959 #endif 1960 #if IS_ENABLED(CONFIG_TIPC) 1961 struct tipc_bearer __rcu *tipc_ptr; 1962 #endif 1963 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1964 void *atalk_ptr; 1965 #endif 1966 struct in_device __rcu *ip_ptr; 1967 #if IS_ENABLED(CONFIG_DECNET) 1968 struct dn_dev __rcu *dn_ptr; 1969 #endif 1970 struct inet6_dev __rcu *ip6_ptr; 1971 #if IS_ENABLED(CONFIG_AX25) 1972 void *ax25_ptr; 1973 #endif 1974 struct wireless_dev *ieee80211_ptr; 1975 struct wpan_dev *ieee802154_ptr; 1976 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1977 struct mpls_dev __rcu *mpls_ptr; 1978 #endif 1979 1980 /* 1981 * Cache lines mostly used on receive path (including eth_type_trans()) 1982 */ 1983 /* Interface address info used in eth_type_trans() */ 1984 unsigned char *dev_addr; 1985 1986 struct netdev_rx_queue *_rx; 1987 unsigned int num_rx_queues; 1988 unsigned int real_num_rx_queues; 1989 1990 struct bpf_prog __rcu *xdp_prog; 1991 unsigned long gro_flush_timeout; 1992 rx_handler_func_t __rcu *rx_handler; 1993 void __rcu *rx_handler_data; 1994 1995 #ifdef CONFIG_NET_CLS_ACT 1996 struct mini_Qdisc __rcu *miniq_ingress; 1997 #endif 1998 struct netdev_queue __rcu *ingress_queue; 1999 #ifdef CONFIG_NETFILTER_INGRESS 2000 struct nf_hook_entries __rcu *nf_hooks_ingress; 2001 #endif 2002 2003 unsigned char broadcast[MAX_ADDR_LEN]; 2004 #ifdef CONFIG_RFS_ACCEL 2005 struct cpu_rmap *rx_cpu_rmap; 2006 #endif 2007 struct hlist_node index_hlist; 2008 2009 /* 2010 * Cache lines mostly used on transmit path 2011 */ 2012 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2013 unsigned int num_tx_queues; 2014 unsigned int real_num_tx_queues; 2015 struct Qdisc *qdisc; 2016 unsigned int tx_queue_len; 2017 spinlock_t tx_global_lock; 2018 2019 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2020 2021 #ifdef CONFIG_XPS 2022 struct xps_dev_maps __rcu *xps_cpus_map; 2023 struct xps_dev_maps __rcu *xps_rxqs_map; 2024 #endif 2025 #ifdef CONFIG_NET_CLS_ACT 2026 struct mini_Qdisc __rcu *miniq_egress; 2027 #endif 2028 2029 #ifdef CONFIG_NET_SCHED 2030 DECLARE_HASHTABLE (qdisc_hash, 4); 2031 #endif 2032 /* These may be needed for future network-power-down code. */ 2033 struct timer_list watchdog_timer; 2034 int watchdog_timeo; 2035 2036 struct list_head todo_list; 2037 int __percpu *pcpu_refcnt; 2038 2039 struct list_head link_watch_list; 2040 2041 enum { NETREG_UNINITIALIZED=0, 2042 NETREG_REGISTERED, /* completed register_netdevice */ 2043 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2044 NETREG_UNREGISTERED, /* completed unregister todo */ 2045 NETREG_RELEASED, /* called free_netdev */ 2046 NETREG_DUMMY, /* dummy device for NAPI poll */ 2047 } reg_state:8; 2048 2049 bool dismantle; 2050 2051 enum { 2052 RTNL_LINK_INITIALIZED, 2053 RTNL_LINK_INITIALIZING, 2054 } rtnl_link_state:16; 2055 2056 bool needs_free_netdev; 2057 void (*priv_destructor)(struct net_device *dev); 2058 2059 #ifdef CONFIG_NETPOLL 2060 struct netpoll_info __rcu *npinfo; 2061 #endif 2062 2063 possible_net_t nd_net; 2064 2065 /* mid-layer private */ 2066 union { 2067 void *ml_priv; 2068 struct pcpu_lstats __percpu *lstats; 2069 struct pcpu_sw_netstats __percpu *tstats; 2070 struct pcpu_dstats __percpu *dstats; 2071 }; 2072 2073 #if IS_ENABLED(CONFIG_GARP) 2074 struct garp_port __rcu *garp_port; 2075 #endif 2076 #if IS_ENABLED(CONFIG_MRP) 2077 struct mrp_port __rcu *mrp_port; 2078 #endif 2079 2080 struct device dev; 2081 const struct attribute_group *sysfs_groups[4]; 2082 const struct attribute_group *sysfs_rx_queue_group; 2083 2084 const struct rtnl_link_ops *rtnl_link_ops; 2085 2086 /* for setting kernel sock attribute on TCP connection setup */ 2087 #define GSO_MAX_SIZE 65536 2088 unsigned int gso_max_size; 2089 #define GSO_MAX_SEGS 65535 2090 u16 gso_max_segs; 2091 2092 #ifdef CONFIG_DCB 2093 const struct dcbnl_rtnl_ops *dcbnl_ops; 2094 #endif 2095 s16 num_tc; 2096 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2097 u8 prio_tc_map[TC_BITMASK + 1]; 2098 2099 #if IS_ENABLED(CONFIG_FCOE) 2100 unsigned int fcoe_ddp_xid; 2101 #endif 2102 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2103 struct netprio_map __rcu *priomap; 2104 #endif 2105 struct phy_device *phydev; 2106 struct sfp_bus *sfp_bus; 2107 struct lock_class_key qdisc_tx_busylock_key; 2108 struct lock_class_key qdisc_running_key; 2109 struct lock_class_key qdisc_xmit_lock_key; 2110 struct lock_class_key addr_list_lock_key; 2111 bool proto_down; 2112 unsigned wol_enabled:1; 2113 2114 struct list_head net_notifier_list; 2115 }; 2116 #define to_net_dev(d) container_of(d, struct net_device, dev) 2117 2118 static inline bool netif_elide_gro(const struct net_device *dev) 2119 { 2120 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2121 return true; 2122 return false; 2123 } 2124 2125 #define NETDEV_ALIGN 32 2126 2127 static inline 2128 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2129 { 2130 return dev->prio_tc_map[prio & TC_BITMASK]; 2131 } 2132 2133 static inline 2134 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2135 { 2136 if (tc >= dev->num_tc) 2137 return -EINVAL; 2138 2139 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2140 return 0; 2141 } 2142 2143 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2144 void netdev_reset_tc(struct net_device *dev); 2145 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2146 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2147 2148 static inline 2149 int netdev_get_num_tc(struct net_device *dev) 2150 { 2151 return dev->num_tc; 2152 } 2153 2154 void netdev_unbind_sb_channel(struct net_device *dev, 2155 struct net_device *sb_dev); 2156 int netdev_bind_sb_channel_queue(struct net_device *dev, 2157 struct net_device *sb_dev, 2158 u8 tc, u16 count, u16 offset); 2159 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2160 static inline int netdev_get_sb_channel(struct net_device *dev) 2161 { 2162 return max_t(int, -dev->num_tc, 0); 2163 } 2164 2165 static inline 2166 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2167 unsigned int index) 2168 { 2169 return &dev->_tx[index]; 2170 } 2171 2172 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2173 const struct sk_buff *skb) 2174 { 2175 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2176 } 2177 2178 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2179 void (*f)(struct net_device *, 2180 struct netdev_queue *, 2181 void *), 2182 void *arg) 2183 { 2184 unsigned int i; 2185 2186 for (i = 0; i < dev->num_tx_queues; i++) 2187 f(dev, &dev->_tx[i], arg); 2188 } 2189 2190 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2191 struct net_device *sb_dev); 2192 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2193 struct sk_buff *skb, 2194 struct net_device *sb_dev); 2195 2196 /* returns the headroom that the master device needs to take in account 2197 * when forwarding to this dev 2198 */ 2199 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2200 { 2201 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2202 } 2203 2204 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2205 { 2206 if (dev->netdev_ops->ndo_set_rx_headroom) 2207 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2208 } 2209 2210 /* set the device rx headroom to the dev's default */ 2211 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2212 { 2213 netdev_set_rx_headroom(dev, -1); 2214 } 2215 2216 /* 2217 * Net namespace inlines 2218 */ 2219 static inline 2220 struct net *dev_net(const struct net_device *dev) 2221 { 2222 return read_pnet(&dev->nd_net); 2223 } 2224 2225 static inline 2226 void dev_net_set(struct net_device *dev, struct net *net) 2227 { 2228 write_pnet(&dev->nd_net, net); 2229 } 2230 2231 /** 2232 * netdev_priv - access network device private data 2233 * @dev: network device 2234 * 2235 * Get network device private data 2236 */ 2237 static inline void *netdev_priv(const struct net_device *dev) 2238 { 2239 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2240 } 2241 2242 /* Set the sysfs physical device reference for the network logical device 2243 * if set prior to registration will cause a symlink during initialization. 2244 */ 2245 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2246 2247 /* Set the sysfs device type for the network logical device to allow 2248 * fine-grained identification of different network device types. For 2249 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2250 */ 2251 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2252 2253 /* Default NAPI poll() weight 2254 * Device drivers are strongly advised to not use bigger value 2255 */ 2256 #define NAPI_POLL_WEIGHT 64 2257 2258 /** 2259 * netif_napi_add - initialize a NAPI context 2260 * @dev: network device 2261 * @napi: NAPI context 2262 * @poll: polling function 2263 * @weight: default weight 2264 * 2265 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2266 * *any* of the other NAPI-related functions. 2267 */ 2268 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2269 int (*poll)(struct napi_struct *, int), int weight); 2270 2271 /** 2272 * netif_tx_napi_add - initialize a NAPI context 2273 * @dev: network device 2274 * @napi: NAPI context 2275 * @poll: polling function 2276 * @weight: default weight 2277 * 2278 * This variant of netif_napi_add() should be used from drivers using NAPI 2279 * to exclusively poll a TX queue. 2280 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2281 */ 2282 static inline void netif_tx_napi_add(struct net_device *dev, 2283 struct napi_struct *napi, 2284 int (*poll)(struct napi_struct *, int), 2285 int weight) 2286 { 2287 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2288 netif_napi_add(dev, napi, poll, weight); 2289 } 2290 2291 /** 2292 * netif_napi_del - remove a NAPI context 2293 * @napi: NAPI context 2294 * 2295 * netif_napi_del() removes a NAPI context from the network device NAPI list 2296 */ 2297 void netif_napi_del(struct napi_struct *napi); 2298 2299 struct napi_gro_cb { 2300 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2301 void *frag0; 2302 2303 /* Length of frag0. */ 2304 unsigned int frag0_len; 2305 2306 /* This indicates where we are processing relative to skb->data. */ 2307 int data_offset; 2308 2309 /* This is non-zero if the packet cannot be merged with the new skb. */ 2310 u16 flush; 2311 2312 /* Save the IP ID here and check when we get to the transport layer */ 2313 u16 flush_id; 2314 2315 /* Number of segments aggregated. */ 2316 u16 count; 2317 2318 /* Start offset for remote checksum offload */ 2319 u16 gro_remcsum_start; 2320 2321 /* jiffies when first packet was created/queued */ 2322 unsigned long age; 2323 2324 /* Used in ipv6_gro_receive() and foo-over-udp */ 2325 u16 proto; 2326 2327 /* This is non-zero if the packet may be of the same flow. */ 2328 u8 same_flow:1; 2329 2330 /* Used in tunnel GRO receive */ 2331 u8 encap_mark:1; 2332 2333 /* GRO checksum is valid */ 2334 u8 csum_valid:1; 2335 2336 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2337 u8 csum_cnt:3; 2338 2339 /* Free the skb? */ 2340 u8 free:2; 2341 #define NAPI_GRO_FREE 1 2342 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2343 2344 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2345 u8 is_ipv6:1; 2346 2347 /* Used in GRE, set in fou/gue_gro_receive */ 2348 u8 is_fou:1; 2349 2350 /* Used to determine if flush_id can be ignored */ 2351 u8 is_atomic:1; 2352 2353 /* Number of gro_receive callbacks this packet already went through */ 2354 u8 recursion_counter:4; 2355 2356 /* GRO is done by frag_list pointer chaining. */ 2357 u8 is_flist:1; 2358 2359 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2360 __wsum csum; 2361 2362 /* used in skb_gro_receive() slow path */ 2363 struct sk_buff *last; 2364 }; 2365 2366 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2367 2368 #define GRO_RECURSION_LIMIT 15 2369 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2370 { 2371 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2372 } 2373 2374 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2375 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2376 struct list_head *head, 2377 struct sk_buff *skb) 2378 { 2379 if (unlikely(gro_recursion_inc_test(skb))) { 2380 NAPI_GRO_CB(skb)->flush |= 1; 2381 return NULL; 2382 } 2383 2384 return cb(head, skb); 2385 } 2386 2387 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2388 struct sk_buff *); 2389 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2390 struct sock *sk, 2391 struct list_head *head, 2392 struct sk_buff *skb) 2393 { 2394 if (unlikely(gro_recursion_inc_test(skb))) { 2395 NAPI_GRO_CB(skb)->flush |= 1; 2396 return NULL; 2397 } 2398 2399 return cb(sk, head, skb); 2400 } 2401 2402 struct packet_type { 2403 __be16 type; /* This is really htons(ether_type). */ 2404 bool ignore_outgoing; 2405 struct net_device *dev; /* NULL is wildcarded here */ 2406 int (*func) (struct sk_buff *, 2407 struct net_device *, 2408 struct packet_type *, 2409 struct net_device *); 2410 void (*list_func) (struct list_head *, 2411 struct packet_type *, 2412 struct net_device *); 2413 bool (*id_match)(struct packet_type *ptype, 2414 struct sock *sk); 2415 void *af_packet_priv; 2416 struct list_head list; 2417 }; 2418 2419 struct offload_callbacks { 2420 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2421 netdev_features_t features); 2422 struct sk_buff *(*gro_receive)(struct list_head *head, 2423 struct sk_buff *skb); 2424 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2425 }; 2426 2427 struct packet_offload { 2428 __be16 type; /* This is really htons(ether_type). */ 2429 u16 priority; 2430 struct offload_callbacks callbacks; 2431 struct list_head list; 2432 }; 2433 2434 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2435 struct pcpu_sw_netstats { 2436 u64 rx_packets; 2437 u64 rx_bytes; 2438 u64 tx_packets; 2439 u64 tx_bytes; 2440 struct u64_stats_sync syncp; 2441 } __aligned(4 * sizeof(u64)); 2442 2443 struct pcpu_lstats { 2444 u64_stats_t packets; 2445 u64_stats_t bytes; 2446 struct u64_stats_sync syncp; 2447 } __aligned(2 * sizeof(u64)); 2448 2449 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2450 2451 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2452 { 2453 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2454 2455 u64_stats_update_begin(&lstats->syncp); 2456 u64_stats_add(&lstats->bytes, len); 2457 u64_stats_inc(&lstats->packets); 2458 u64_stats_update_end(&lstats->syncp); 2459 } 2460 2461 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2462 ({ \ 2463 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2464 if (pcpu_stats) { \ 2465 int __cpu; \ 2466 for_each_possible_cpu(__cpu) { \ 2467 typeof(type) *stat; \ 2468 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2469 u64_stats_init(&stat->syncp); \ 2470 } \ 2471 } \ 2472 pcpu_stats; \ 2473 }) 2474 2475 #define netdev_alloc_pcpu_stats(type) \ 2476 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2477 2478 enum netdev_lag_tx_type { 2479 NETDEV_LAG_TX_TYPE_UNKNOWN, 2480 NETDEV_LAG_TX_TYPE_RANDOM, 2481 NETDEV_LAG_TX_TYPE_BROADCAST, 2482 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2483 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2484 NETDEV_LAG_TX_TYPE_HASH, 2485 }; 2486 2487 enum netdev_lag_hash { 2488 NETDEV_LAG_HASH_NONE, 2489 NETDEV_LAG_HASH_L2, 2490 NETDEV_LAG_HASH_L34, 2491 NETDEV_LAG_HASH_L23, 2492 NETDEV_LAG_HASH_E23, 2493 NETDEV_LAG_HASH_E34, 2494 NETDEV_LAG_HASH_UNKNOWN, 2495 }; 2496 2497 struct netdev_lag_upper_info { 2498 enum netdev_lag_tx_type tx_type; 2499 enum netdev_lag_hash hash_type; 2500 }; 2501 2502 struct netdev_lag_lower_state_info { 2503 u8 link_up : 1, 2504 tx_enabled : 1; 2505 }; 2506 2507 #include <linux/notifier.h> 2508 2509 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2510 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2511 * adding new types. 2512 */ 2513 enum netdev_cmd { 2514 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2515 NETDEV_DOWN, 2516 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2517 detected a hardware crash and restarted 2518 - we can use this eg to kick tcp sessions 2519 once done */ 2520 NETDEV_CHANGE, /* Notify device state change */ 2521 NETDEV_REGISTER, 2522 NETDEV_UNREGISTER, 2523 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2524 NETDEV_CHANGEADDR, /* notify after the address change */ 2525 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2526 NETDEV_GOING_DOWN, 2527 NETDEV_CHANGENAME, 2528 NETDEV_FEAT_CHANGE, 2529 NETDEV_BONDING_FAILOVER, 2530 NETDEV_PRE_UP, 2531 NETDEV_PRE_TYPE_CHANGE, 2532 NETDEV_POST_TYPE_CHANGE, 2533 NETDEV_POST_INIT, 2534 NETDEV_RELEASE, 2535 NETDEV_NOTIFY_PEERS, 2536 NETDEV_JOIN, 2537 NETDEV_CHANGEUPPER, 2538 NETDEV_RESEND_IGMP, 2539 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2540 NETDEV_CHANGEINFODATA, 2541 NETDEV_BONDING_INFO, 2542 NETDEV_PRECHANGEUPPER, 2543 NETDEV_CHANGELOWERSTATE, 2544 NETDEV_UDP_TUNNEL_PUSH_INFO, 2545 NETDEV_UDP_TUNNEL_DROP_INFO, 2546 NETDEV_CHANGE_TX_QUEUE_LEN, 2547 NETDEV_CVLAN_FILTER_PUSH_INFO, 2548 NETDEV_CVLAN_FILTER_DROP_INFO, 2549 NETDEV_SVLAN_FILTER_PUSH_INFO, 2550 NETDEV_SVLAN_FILTER_DROP_INFO, 2551 }; 2552 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2553 2554 int register_netdevice_notifier(struct notifier_block *nb); 2555 int unregister_netdevice_notifier(struct notifier_block *nb); 2556 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2557 int unregister_netdevice_notifier_net(struct net *net, 2558 struct notifier_block *nb); 2559 int register_netdevice_notifier_dev_net(struct net_device *dev, 2560 struct notifier_block *nb, 2561 struct netdev_net_notifier *nn); 2562 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2563 struct notifier_block *nb, 2564 struct netdev_net_notifier *nn); 2565 2566 struct netdev_notifier_info { 2567 struct net_device *dev; 2568 struct netlink_ext_ack *extack; 2569 }; 2570 2571 struct netdev_notifier_info_ext { 2572 struct netdev_notifier_info info; /* must be first */ 2573 union { 2574 u32 mtu; 2575 } ext; 2576 }; 2577 2578 struct netdev_notifier_change_info { 2579 struct netdev_notifier_info info; /* must be first */ 2580 unsigned int flags_changed; 2581 }; 2582 2583 struct netdev_notifier_changeupper_info { 2584 struct netdev_notifier_info info; /* must be first */ 2585 struct net_device *upper_dev; /* new upper dev */ 2586 bool master; /* is upper dev master */ 2587 bool linking; /* is the notification for link or unlink */ 2588 void *upper_info; /* upper dev info */ 2589 }; 2590 2591 struct netdev_notifier_changelowerstate_info { 2592 struct netdev_notifier_info info; /* must be first */ 2593 void *lower_state_info; /* is lower dev state */ 2594 }; 2595 2596 struct netdev_notifier_pre_changeaddr_info { 2597 struct netdev_notifier_info info; /* must be first */ 2598 const unsigned char *dev_addr; 2599 }; 2600 2601 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2602 struct net_device *dev) 2603 { 2604 info->dev = dev; 2605 info->extack = NULL; 2606 } 2607 2608 static inline struct net_device * 2609 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2610 { 2611 return info->dev; 2612 } 2613 2614 static inline struct netlink_ext_ack * 2615 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2616 { 2617 return info->extack; 2618 } 2619 2620 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2621 2622 2623 extern rwlock_t dev_base_lock; /* Device list lock */ 2624 2625 #define for_each_netdev(net, d) \ 2626 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2627 #define for_each_netdev_reverse(net, d) \ 2628 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2629 #define for_each_netdev_rcu(net, d) \ 2630 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2631 #define for_each_netdev_safe(net, d, n) \ 2632 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2633 #define for_each_netdev_continue(net, d) \ 2634 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2635 #define for_each_netdev_continue_reverse(net, d) \ 2636 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2637 dev_list) 2638 #define for_each_netdev_continue_rcu(net, d) \ 2639 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2640 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2641 for_each_netdev_rcu(&init_net, slave) \ 2642 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2643 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2644 2645 static inline struct net_device *next_net_device(struct net_device *dev) 2646 { 2647 struct list_head *lh; 2648 struct net *net; 2649 2650 net = dev_net(dev); 2651 lh = dev->dev_list.next; 2652 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2653 } 2654 2655 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2656 { 2657 struct list_head *lh; 2658 struct net *net; 2659 2660 net = dev_net(dev); 2661 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2662 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2663 } 2664 2665 static inline struct net_device *first_net_device(struct net *net) 2666 { 2667 return list_empty(&net->dev_base_head) ? NULL : 2668 net_device_entry(net->dev_base_head.next); 2669 } 2670 2671 static inline struct net_device *first_net_device_rcu(struct net *net) 2672 { 2673 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2674 2675 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2676 } 2677 2678 int netdev_boot_setup_check(struct net_device *dev); 2679 unsigned long netdev_boot_base(const char *prefix, int unit); 2680 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2681 const char *hwaddr); 2682 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2683 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2684 void dev_add_pack(struct packet_type *pt); 2685 void dev_remove_pack(struct packet_type *pt); 2686 void __dev_remove_pack(struct packet_type *pt); 2687 void dev_add_offload(struct packet_offload *po); 2688 void dev_remove_offload(struct packet_offload *po); 2689 2690 int dev_get_iflink(const struct net_device *dev); 2691 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2692 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2693 unsigned short mask); 2694 struct net_device *dev_get_by_name(struct net *net, const char *name); 2695 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2696 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2697 int dev_alloc_name(struct net_device *dev, const char *name); 2698 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2699 void dev_close(struct net_device *dev); 2700 void dev_close_many(struct list_head *head, bool unlink); 2701 void dev_disable_lro(struct net_device *dev); 2702 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2703 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2704 struct net_device *sb_dev); 2705 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2706 struct net_device *sb_dev); 2707 int dev_queue_xmit(struct sk_buff *skb); 2708 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2709 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2710 int register_netdevice(struct net_device *dev); 2711 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2712 void unregister_netdevice_many(struct list_head *head); 2713 static inline void unregister_netdevice(struct net_device *dev) 2714 { 2715 unregister_netdevice_queue(dev, NULL); 2716 } 2717 2718 int netdev_refcnt_read(const struct net_device *dev); 2719 void free_netdev(struct net_device *dev); 2720 void netdev_freemem(struct net_device *dev); 2721 void synchronize_net(void); 2722 int init_dummy_netdev(struct net_device *dev); 2723 2724 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2725 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2726 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2727 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2728 int netdev_get_name(struct net *net, char *name, int ifindex); 2729 int dev_restart(struct net_device *dev); 2730 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2731 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2732 2733 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2734 { 2735 return NAPI_GRO_CB(skb)->data_offset; 2736 } 2737 2738 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2739 { 2740 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2741 } 2742 2743 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2744 { 2745 NAPI_GRO_CB(skb)->data_offset += len; 2746 } 2747 2748 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2749 unsigned int offset) 2750 { 2751 return NAPI_GRO_CB(skb)->frag0 + offset; 2752 } 2753 2754 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2755 { 2756 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2757 } 2758 2759 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2760 { 2761 NAPI_GRO_CB(skb)->frag0 = NULL; 2762 NAPI_GRO_CB(skb)->frag0_len = 0; 2763 } 2764 2765 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2766 unsigned int offset) 2767 { 2768 if (!pskb_may_pull(skb, hlen)) 2769 return NULL; 2770 2771 skb_gro_frag0_invalidate(skb); 2772 return skb->data + offset; 2773 } 2774 2775 static inline void *skb_gro_network_header(struct sk_buff *skb) 2776 { 2777 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2778 skb_network_offset(skb); 2779 } 2780 2781 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2782 const void *start, unsigned int len) 2783 { 2784 if (NAPI_GRO_CB(skb)->csum_valid) 2785 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2786 csum_partial(start, len, 0)); 2787 } 2788 2789 /* GRO checksum functions. These are logical equivalents of the normal 2790 * checksum functions (in skbuff.h) except that they operate on the GRO 2791 * offsets and fields in sk_buff. 2792 */ 2793 2794 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2795 2796 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2797 { 2798 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2799 } 2800 2801 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2802 bool zero_okay, 2803 __sum16 check) 2804 { 2805 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2806 skb_checksum_start_offset(skb) < 2807 skb_gro_offset(skb)) && 2808 !skb_at_gro_remcsum_start(skb) && 2809 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2810 (!zero_okay || check)); 2811 } 2812 2813 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2814 __wsum psum) 2815 { 2816 if (NAPI_GRO_CB(skb)->csum_valid && 2817 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2818 return 0; 2819 2820 NAPI_GRO_CB(skb)->csum = psum; 2821 2822 return __skb_gro_checksum_complete(skb); 2823 } 2824 2825 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2826 { 2827 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2828 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2829 NAPI_GRO_CB(skb)->csum_cnt--; 2830 } else { 2831 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2832 * verified a new top level checksum or an encapsulated one 2833 * during GRO. This saves work if we fallback to normal path. 2834 */ 2835 __skb_incr_checksum_unnecessary(skb); 2836 } 2837 } 2838 2839 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2840 compute_pseudo) \ 2841 ({ \ 2842 __sum16 __ret = 0; \ 2843 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2844 __ret = __skb_gro_checksum_validate_complete(skb, \ 2845 compute_pseudo(skb, proto)); \ 2846 if (!__ret) \ 2847 skb_gro_incr_csum_unnecessary(skb); \ 2848 __ret; \ 2849 }) 2850 2851 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2852 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2853 2854 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2855 compute_pseudo) \ 2856 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2857 2858 #define skb_gro_checksum_simple_validate(skb) \ 2859 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2860 2861 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2862 { 2863 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2864 !NAPI_GRO_CB(skb)->csum_valid); 2865 } 2866 2867 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2868 __wsum pseudo) 2869 { 2870 NAPI_GRO_CB(skb)->csum = ~pseudo; 2871 NAPI_GRO_CB(skb)->csum_valid = 1; 2872 } 2873 2874 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2875 do { \ 2876 if (__skb_gro_checksum_convert_check(skb)) \ 2877 __skb_gro_checksum_convert(skb, \ 2878 compute_pseudo(skb, proto)); \ 2879 } while (0) 2880 2881 struct gro_remcsum { 2882 int offset; 2883 __wsum delta; 2884 }; 2885 2886 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2887 { 2888 grc->offset = 0; 2889 grc->delta = 0; 2890 } 2891 2892 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2893 unsigned int off, size_t hdrlen, 2894 int start, int offset, 2895 struct gro_remcsum *grc, 2896 bool nopartial) 2897 { 2898 __wsum delta; 2899 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2900 2901 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2902 2903 if (!nopartial) { 2904 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2905 return ptr; 2906 } 2907 2908 ptr = skb_gro_header_fast(skb, off); 2909 if (skb_gro_header_hard(skb, off + plen)) { 2910 ptr = skb_gro_header_slow(skb, off + plen, off); 2911 if (!ptr) 2912 return NULL; 2913 } 2914 2915 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2916 start, offset); 2917 2918 /* Adjust skb->csum since we changed the packet */ 2919 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2920 2921 grc->offset = off + hdrlen + offset; 2922 grc->delta = delta; 2923 2924 return ptr; 2925 } 2926 2927 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2928 struct gro_remcsum *grc) 2929 { 2930 void *ptr; 2931 size_t plen = grc->offset + sizeof(u16); 2932 2933 if (!grc->delta) 2934 return; 2935 2936 ptr = skb_gro_header_fast(skb, grc->offset); 2937 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2938 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2939 if (!ptr) 2940 return; 2941 } 2942 2943 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2944 } 2945 2946 #ifdef CONFIG_XFRM_OFFLOAD 2947 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2948 { 2949 if (PTR_ERR(pp) != -EINPROGRESS) 2950 NAPI_GRO_CB(skb)->flush |= flush; 2951 } 2952 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2953 struct sk_buff *pp, 2954 int flush, 2955 struct gro_remcsum *grc) 2956 { 2957 if (PTR_ERR(pp) != -EINPROGRESS) { 2958 NAPI_GRO_CB(skb)->flush |= flush; 2959 skb_gro_remcsum_cleanup(skb, grc); 2960 skb->remcsum_offload = 0; 2961 } 2962 } 2963 #else 2964 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2965 { 2966 NAPI_GRO_CB(skb)->flush |= flush; 2967 } 2968 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2969 struct sk_buff *pp, 2970 int flush, 2971 struct gro_remcsum *grc) 2972 { 2973 NAPI_GRO_CB(skb)->flush |= flush; 2974 skb_gro_remcsum_cleanup(skb, grc); 2975 skb->remcsum_offload = 0; 2976 } 2977 #endif 2978 2979 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2980 unsigned short type, 2981 const void *daddr, const void *saddr, 2982 unsigned int len) 2983 { 2984 if (!dev->header_ops || !dev->header_ops->create) 2985 return 0; 2986 2987 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2988 } 2989 2990 static inline int dev_parse_header(const struct sk_buff *skb, 2991 unsigned char *haddr) 2992 { 2993 const struct net_device *dev = skb->dev; 2994 2995 if (!dev->header_ops || !dev->header_ops->parse) 2996 return 0; 2997 return dev->header_ops->parse(skb, haddr); 2998 } 2999 3000 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3001 { 3002 const struct net_device *dev = skb->dev; 3003 3004 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3005 return 0; 3006 return dev->header_ops->parse_protocol(skb); 3007 } 3008 3009 /* ll_header must have at least hard_header_len allocated */ 3010 static inline bool dev_validate_header(const struct net_device *dev, 3011 char *ll_header, int len) 3012 { 3013 if (likely(len >= dev->hard_header_len)) 3014 return true; 3015 if (len < dev->min_header_len) 3016 return false; 3017 3018 if (capable(CAP_SYS_RAWIO)) { 3019 memset(ll_header + len, 0, dev->hard_header_len - len); 3020 return true; 3021 } 3022 3023 if (dev->header_ops && dev->header_ops->validate) 3024 return dev->header_ops->validate(ll_header, len); 3025 3026 return false; 3027 } 3028 3029 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3030 int len, int size); 3031 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3032 static inline int unregister_gifconf(unsigned int family) 3033 { 3034 return register_gifconf(family, NULL); 3035 } 3036 3037 #ifdef CONFIG_NET_FLOW_LIMIT 3038 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3039 struct sd_flow_limit { 3040 u64 count; 3041 unsigned int num_buckets; 3042 unsigned int history_head; 3043 u16 history[FLOW_LIMIT_HISTORY]; 3044 u8 buckets[]; 3045 }; 3046 3047 extern int netdev_flow_limit_table_len; 3048 #endif /* CONFIG_NET_FLOW_LIMIT */ 3049 3050 /* 3051 * Incoming packets are placed on per-CPU queues 3052 */ 3053 struct softnet_data { 3054 struct list_head poll_list; 3055 struct sk_buff_head process_queue; 3056 3057 /* stats */ 3058 unsigned int processed; 3059 unsigned int time_squeeze; 3060 unsigned int received_rps; 3061 #ifdef CONFIG_RPS 3062 struct softnet_data *rps_ipi_list; 3063 #endif 3064 #ifdef CONFIG_NET_FLOW_LIMIT 3065 struct sd_flow_limit __rcu *flow_limit; 3066 #endif 3067 struct Qdisc *output_queue; 3068 struct Qdisc **output_queue_tailp; 3069 struct sk_buff *completion_queue; 3070 #ifdef CONFIG_XFRM_OFFLOAD 3071 struct sk_buff_head xfrm_backlog; 3072 #endif 3073 /* written and read only by owning cpu: */ 3074 struct { 3075 u16 recursion; 3076 u8 more; 3077 } xmit; 3078 #ifdef CONFIG_RPS 3079 /* input_queue_head should be written by cpu owning this struct, 3080 * and only read by other cpus. Worth using a cache line. 3081 */ 3082 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3083 3084 /* Elements below can be accessed between CPUs for RPS/RFS */ 3085 call_single_data_t csd ____cacheline_aligned_in_smp; 3086 struct softnet_data *rps_ipi_next; 3087 unsigned int cpu; 3088 unsigned int input_queue_tail; 3089 #endif 3090 unsigned int dropped; 3091 struct sk_buff_head input_pkt_queue; 3092 struct napi_struct backlog; 3093 3094 }; 3095 3096 static inline void input_queue_head_incr(struct softnet_data *sd) 3097 { 3098 #ifdef CONFIG_RPS 3099 sd->input_queue_head++; 3100 #endif 3101 } 3102 3103 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3104 unsigned int *qtail) 3105 { 3106 #ifdef CONFIG_RPS 3107 *qtail = ++sd->input_queue_tail; 3108 #endif 3109 } 3110 3111 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3112 3113 static inline int dev_recursion_level(void) 3114 { 3115 return this_cpu_read(softnet_data.xmit.recursion); 3116 } 3117 3118 #define XMIT_RECURSION_LIMIT 10 3119 static inline bool dev_xmit_recursion(void) 3120 { 3121 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3122 XMIT_RECURSION_LIMIT); 3123 } 3124 3125 static inline void dev_xmit_recursion_inc(void) 3126 { 3127 __this_cpu_inc(softnet_data.xmit.recursion); 3128 } 3129 3130 static inline void dev_xmit_recursion_dec(void) 3131 { 3132 __this_cpu_dec(softnet_data.xmit.recursion); 3133 } 3134 3135 void __netif_schedule(struct Qdisc *q); 3136 void netif_schedule_queue(struct netdev_queue *txq); 3137 3138 static inline void netif_tx_schedule_all(struct net_device *dev) 3139 { 3140 unsigned int i; 3141 3142 for (i = 0; i < dev->num_tx_queues; i++) 3143 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3144 } 3145 3146 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3147 { 3148 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3149 } 3150 3151 /** 3152 * netif_start_queue - allow transmit 3153 * @dev: network device 3154 * 3155 * Allow upper layers to call the device hard_start_xmit routine. 3156 */ 3157 static inline void netif_start_queue(struct net_device *dev) 3158 { 3159 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3160 } 3161 3162 static inline void netif_tx_start_all_queues(struct net_device *dev) 3163 { 3164 unsigned int i; 3165 3166 for (i = 0; i < dev->num_tx_queues; i++) { 3167 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3168 netif_tx_start_queue(txq); 3169 } 3170 } 3171 3172 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3173 3174 /** 3175 * netif_wake_queue - restart transmit 3176 * @dev: network device 3177 * 3178 * Allow upper layers to call the device hard_start_xmit routine. 3179 * Used for flow control when transmit resources are available. 3180 */ 3181 static inline void netif_wake_queue(struct net_device *dev) 3182 { 3183 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3184 } 3185 3186 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3187 { 3188 unsigned int i; 3189 3190 for (i = 0; i < dev->num_tx_queues; i++) { 3191 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3192 netif_tx_wake_queue(txq); 3193 } 3194 } 3195 3196 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3197 { 3198 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3199 } 3200 3201 /** 3202 * netif_stop_queue - stop transmitted packets 3203 * @dev: network device 3204 * 3205 * Stop upper layers calling the device hard_start_xmit routine. 3206 * Used for flow control when transmit resources are unavailable. 3207 */ 3208 static inline void netif_stop_queue(struct net_device *dev) 3209 { 3210 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3211 } 3212 3213 void netif_tx_stop_all_queues(struct net_device *dev); 3214 void netdev_update_lockdep_key(struct net_device *dev); 3215 3216 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3217 { 3218 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3219 } 3220 3221 /** 3222 * netif_queue_stopped - test if transmit queue is flowblocked 3223 * @dev: network device 3224 * 3225 * Test if transmit queue on device is currently unable to send. 3226 */ 3227 static inline bool netif_queue_stopped(const struct net_device *dev) 3228 { 3229 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3230 } 3231 3232 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3233 { 3234 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3235 } 3236 3237 static inline bool 3238 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3239 { 3240 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3241 } 3242 3243 static inline bool 3244 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3245 { 3246 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3247 } 3248 3249 /** 3250 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3251 * @dev_queue: pointer to transmit queue 3252 * 3253 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3254 * to give appropriate hint to the CPU. 3255 */ 3256 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3257 { 3258 #ifdef CONFIG_BQL 3259 prefetchw(&dev_queue->dql.num_queued); 3260 #endif 3261 } 3262 3263 /** 3264 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3265 * @dev_queue: pointer to transmit queue 3266 * 3267 * BQL enabled drivers might use this helper in their TX completion path, 3268 * to give appropriate hint to the CPU. 3269 */ 3270 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3271 { 3272 #ifdef CONFIG_BQL 3273 prefetchw(&dev_queue->dql.limit); 3274 #endif 3275 } 3276 3277 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3278 unsigned int bytes) 3279 { 3280 #ifdef CONFIG_BQL 3281 dql_queued(&dev_queue->dql, bytes); 3282 3283 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3284 return; 3285 3286 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3287 3288 /* 3289 * The XOFF flag must be set before checking the dql_avail below, 3290 * because in netdev_tx_completed_queue we update the dql_completed 3291 * before checking the XOFF flag. 3292 */ 3293 smp_mb(); 3294 3295 /* check again in case another CPU has just made room avail */ 3296 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3297 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3298 #endif 3299 } 3300 3301 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3302 * that they should not test BQL status themselves. 3303 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3304 * skb of a batch. 3305 * Returns true if the doorbell must be used to kick the NIC. 3306 */ 3307 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3308 unsigned int bytes, 3309 bool xmit_more) 3310 { 3311 if (xmit_more) { 3312 #ifdef CONFIG_BQL 3313 dql_queued(&dev_queue->dql, bytes); 3314 #endif 3315 return netif_tx_queue_stopped(dev_queue); 3316 } 3317 netdev_tx_sent_queue(dev_queue, bytes); 3318 return true; 3319 } 3320 3321 /** 3322 * netdev_sent_queue - report the number of bytes queued to hardware 3323 * @dev: network device 3324 * @bytes: number of bytes queued to the hardware device queue 3325 * 3326 * Report the number of bytes queued for sending/completion to the network 3327 * device hardware queue. @bytes should be a good approximation and should 3328 * exactly match netdev_completed_queue() @bytes 3329 */ 3330 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3331 { 3332 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3333 } 3334 3335 static inline bool __netdev_sent_queue(struct net_device *dev, 3336 unsigned int bytes, 3337 bool xmit_more) 3338 { 3339 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3340 xmit_more); 3341 } 3342 3343 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3344 unsigned int pkts, unsigned int bytes) 3345 { 3346 #ifdef CONFIG_BQL 3347 if (unlikely(!bytes)) 3348 return; 3349 3350 dql_completed(&dev_queue->dql, bytes); 3351 3352 /* 3353 * Without the memory barrier there is a small possiblity that 3354 * netdev_tx_sent_queue will miss the update and cause the queue to 3355 * be stopped forever 3356 */ 3357 smp_mb(); 3358 3359 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3360 return; 3361 3362 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3363 netif_schedule_queue(dev_queue); 3364 #endif 3365 } 3366 3367 /** 3368 * netdev_completed_queue - report bytes and packets completed by device 3369 * @dev: network device 3370 * @pkts: actual number of packets sent over the medium 3371 * @bytes: actual number of bytes sent over the medium 3372 * 3373 * Report the number of bytes and packets transmitted by the network device 3374 * hardware queue over the physical medium, @bytes must exactly match the 3375 * @bytes amount passed to netdev_sent_queue() 3376 */ 3377 static inline void netdev_completed_queue(struct net_device *dev, 3378 unsigned int pkts, unsigned int bytes) 3379 { 3380 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3381 } 3382 3383 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3384 { 3385 #ifdef CONFIG_BQL 3386 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3387 dql_reset(&q->dql); 3388 #endif 3389 } 3390 3391 /** 3392 * netdev_reset_queue - reset the packets and bytes count of a network device 3393 * @dev_queue: network device 3394 * 3395 * Reset the bytes and packet count of a network device and clear the 3396 * software flow control OFF bit for this network device 3397 */ 3398 static inline void netdev_reset_queue(struct net_device *dev_queue) 3399 { 3400 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3401 } 3402 3403 /** 3404 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3405 * @dev: network device 3406 * @queue_index: given tx queue index 3407 * 3408 * Returns 0 if given tx queue index >= number of device tx queues, 3409 * otherwise returns the originally passed tx queue index. 3410 */ 3411 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3412 { 3413 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3414 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3415 dev->name, queue_index, 3416 dev->real_num_tx_queues); 3417 return 0; 3418 } 3419 3420 return queue_index; 3421 } 3422 3423 /** 3424 * netif_running - test if up 3425 * @dev: network device 3426 * 3427 * Test if the device has been brought up. 3428 */ 3429 static inline bool netif_running(const struct net_device *dev) 3430 { 3431 return test_bit(__LINK_STATE_START, &dev->state); 3432 } 3433 3434 /* 3435 * Routines to manage the subqueues on a device. We only need start, 3436 * stop, and a check if it's stopped. All other device management is 3437 * done at the overall netdevice level. 3438 * Also test the device if we're multiqueue. 3439 */ 3440 3441 /** 3442 * netif_start_subqueue - allow sending packets on subqueue 3443 * @dev: network device 3444 * @queue_index: sub queue index 3445 * 3446 * Start individual transmit queue of a device with multiple transmit queues. 3447 */ 3448 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3449 { 3450 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3451 3452 netif_tx_start_queue(txq); 3453 } 3454 3455 /** 3456 * netif_stop_subqueue - stop sending packets on subqueue 3457 * @dev: network device 3458 * @queue_index: sub queue index 3459 * 3460 * Stop individual transmit queue of a device with multiple transmit queues. 3461 */ 3462 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3463 { 3464 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3465 netif_tx_stop_queue(txq); 3466 } 3467 3468 /** 3469 * netif_subqueue_stopped - test status of subqueue 3470 * @dev: network device 3471 * @queue_index: sub queue index 3472 * 3473 * Check individual transmit queue of a device with multiple transmit queues. 3474 */ 3475 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3476 u16 queue_index) 3477 { 3478 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3479 3480 return netif_tx_queue_stopped(txq); 3481 } 3482 3483 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3484 struct sk_buff *skb) 3485 { 3486 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3487 } 3488 3489 /** 3490 * netif_wake_subqueue - allow sending packets on subqueue 3491 * @dev: network device 3492 * @queue_index: sub queue index 3493 * 3494 * Resume individual transmit queue of a device with multiple transmit queues. 3495 */ 3496 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3497 { 3498 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3499 3500 netif_tx_wake_queue(txq); 3501 } 3502 3503 #ifdef CONFIG_XPS 3504 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3505 u16 index); 3506 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3507 u16 index, bool is_rxqs_map); 3508 3509 /** 3510 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3511 * @j: CPU/Rx queue index 3512 * @mask: bitmask of all cpus/rx queues 3513 * @nr_bits: number of bits in the bitmask 3514 * 3515 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3516 */ 3517 static inline bool netif_attr_test_mask(unsigned long j, 3518 const unsigned long *mask, 3519 unsigned int nr_bits) 3520 { 3521 cpu_max_bits_warn(j, nr_bits); 3522 return test_bit(j, mask); 3523 } 3524 3525 /** 3526 * netif_attr_test_online - Test for online CPU/Rx queue 3527 * @j: CPU/Rx queue index 3528 * @online_mask: bitmask for CPUs/Rx queues that are online 3529 * @nr_bits: number of bits in the bitmask 3530 * 3531 * Returns true if a CPU/Rx queue is online. 3532 */ 3533 static inline bool netif_attr_test_online(unsigned long j, 3534 const unsigned long *online_mask, 3535 unsigned int nr_bits) 3536 { 3537 cpu_max_bits_warn(j, nr_bits); 3538 3539 if (online_mask) 3540 return test_bit(j, online_mask); 3541 3542 return (j < nr_bits); 3543 } 3544 3545 /** 3546 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3547 * @n: CPU/Rx queue index 3548 * @srcp: the cpumask/Rx queue mask pointer 3549 * @nr_bits: number of bits in the bitmask 3550 * 3551 * Returns >= nr_bits if no further CPUs/Rx queues set. 3552 */ 3553 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3554 unsigned int nr_bits) 3555 { 3556 /* -1 is a legal arg here. */ 3557 if (n != -1) 3558 cpu_max_bits_warn(n, nr_bits); 3559 3560 if (srcp) 3561 return find_next_bit(srcp, nr_bits, n + 1); 3562 3563 return n + 1; 3564 } 3565 3566 /** 3567 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3568 * @n: CPU/Rx queue index 3569 * @src1p: the first CPUs/Rx queues mask pointer 3570 * @src2p: the second CPUs/Rx queues mask pointer 3571 * @nr_bits: number of bits in the bitmask 3572 * 3573 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3574 */ 3575 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3576 const unsigned long *src2p, 3577 unsigned int nr_bits) 3578 { 3579 /* -1 is a legal arg here. */ 3580 if (n != -1) 3581 cpu_max_bits_warn(n, nr_bits); 3582 3583 if (src1p && src2p) 3584 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3585 else if (src1p) 3586 return find_next_bit(src1p, nr_bits, n + 1); 3587 else if (src2p) 3588 return find_next_bit(src2p, nr_bits, n + 1); 3589 3590 return n + 1; 3591 } 3592 #else 3593 static inline int netif_set_xps_queue(struct net_device *dev, 3594 const struct cpumask *mask, 3595 u16 index) 3596 { 3597 return 0; 3598 } 3599 3600 static inline int __netif_set_xps_queue(struct net_device *dev, 3601 const unsigned long *mask, 3602 u16 index, bool is_rxqs_map) 3603 { 3604 return 0; 3605 } 3606 #endif 3607 3608 /** 3609 * netif_is_multiqueue - test if device has multiple transmit queues 3610 * @dev: network device 3611 * 3612 * Check if device has multiple transmit queues 3613 */ 3614 static inline bool netif_is_multiqueue(const struct net_device *dev) 3615 { 3616 return dev->num_tx_queues > 1; 3617 } 3618 3619 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3620 3621 #ifdef CONFIG_SYSFS 3622 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3623 #else 3624 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3625 unsigned int rxqs) 3626 { 3627 dev->real_num_rx_queues = rxqs; 3628 return 0; 3629 } 3630 #endif 3631 3632 static inline struct netdev_rx_queue * 3633 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3634 { 3635 return dev->_rx + rxq; 3636 } 3637 3638 #ifdef CONFIG_SYSFS 3639 static inline unsigned int get_netdev_rx_queue_index( 3640 struct netdev_rx_queue *queue) 3641 { 3642 struct net_device *dev = queue->dev; 3643 int index = queue - dev->_rx; 3644 3645 BUG_ON(index >= dev->num_rx_queues); 3646 return index; 3647 } 3648 #endif 3649 3650 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3651 int netif_get_num_default_rss_queues(void); 3652 3653 enum skb_free_reason { 3654 SKB_REASON_CONSUMED, 3655 SKB_REASON_DROPPED, 3656 }; 3657 3658 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3659 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3660 3661 /* 3662 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3663 * interrupt context or with hardware interrupts being disabled. 3664 * (in_irq() || irqs_disabled()) 3665 * 3666 * We provide four helpers that can be used in following contexts : 3667 * 3668 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3669 * replacing kfree_skb(skb) 3670 * 3671 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3672 * Typically used in place of consume_skb(skb) in TX completion path 3673 * 3674 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3675 * replacing kfree_skb(skb) 3676 * 3677 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3678 * and consumed a packet. Used in place of consume_skb(skb) 3679 */ 3680 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3681 { 3682 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3683 } 3684 3685 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3686 { 3687 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3688 } 3689 3690 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3691 { 3692 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3693 } 3694 3695 static inline void dev_consume_skb_any(struct sk_buff *skb) 3696 { 3697 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3698 } 3699 3700 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3701 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3702 int netif_rx(struct sk_buff *skb); 3703 int netif_rx_ni(struct sk_buff *skb); 3704 int netif_receive_skb(struct sk_buff *skb); 3705 int netif_receive_skb_core(struct sk_buff *skb); 3706 void netif_receive_skb_list(struct list_head *head); 3707 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3708 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3709 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3710 gro_result_t napi_gro_frags(struct napi_struct *napi); 3711 struct packet_offload *gro_find_receive_by_type(__be16 type); 3712 struct packet_offload *gro_find_complete_by_type(__be16 type); 3713 3714 static inline void napi_free_frags(struct napi_struct *napi) 3715 { 3716 kfree_skb(napi->skb); 3717 napi->skb = NULL; 3718 } 3719 3720 bool netdev_is_rx_handler_busy(struct net_device *dev); 3721 int netdev_rx_handler_register(struct net_device *dev, 3722 rx_handler_func_t *rx_handler, 3723 void *rx_handler_data); 3724 void netdev_rx_handler_unregister(struct net_device *dev); 3725 3726 bool dev_valid_name(const char *name); 3727 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3728 bool *need_copyout); 3729 int dev_ifconf(struct net *net, struct ifconf *, int); 3730 int dev_ethtool(struct net *net, struct ifreq *); 3731 unsigned int dev_get_flags(const struct net_device *); 3732 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3733 struct netlink_ext_ack *extack); 3734 int dev_change_flags(struct net_device *dev, unsigned int flags, 3735 struct netlink_ext_ack *extack); 3736 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3737 unsigned int gchanges); 3738 int dev_change_name(struct net_device *, const char *); 3739 int dev_set_alias(struct net_device *, const char *, size_t); 3740 int dev_get_alias(const struct net_device *, char *, size_t); 3741 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3742 int __dev_set_mtu(struct net_device *, int); 3743 int dev_validate_mtu(struct net_device *dev, int mtu, 3744 struct netlink_ext_ack *extack); 3745 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3746 struct netlink_ext_ack *extack); 3747 int dev_set_mtu(struct net_device *, int); 3748 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3749 void dev_set_group(struct net_device *, int); 3750 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3751 struct netlink_ext_ack *extack); 3752 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3753 struct netlink_ext_ack *extack); 3754 int dev_change_carrier(struct net_device *, bool new_carrier); 3755 int dev_get_phys_port_id(struct net_device *dev, 3756 struct netdev_phys_item_id *ppid); 3757 int dev_get_phys_port_name(struct net_device *dev, 3758 char *name, size_t len); 3759 int dev_get_port_parent_id(struct net_device *dev, 3760 struct netdev_phys_item_id *ppid, bool recurse); 3761 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3762 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3763 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3764 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3765 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3766 struct netdev_queue *txq, int *ret); 3767 3768 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3769 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3770 int fd, u32 flags); 3771 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3772 enum bpf_netdev_command cmd); 3773 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3774 3775 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3776 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3777 bool is_skb_forwardable(const struct net_device *dev, 3778 const struct sk_buff *skb); 3779 3780 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3781 struct sk_buff *skb) 3782 { 3783 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3784 unlikely(!is_skb_forwardable(dev, skb))) { 3785 atomic_long_inc(&dev->rx_dropped); 3786 kfree_skb(skb); 3787 return NET_RX_DROP; 3788 } 3789 3790 skb_scrub_packet(skb, true); 3791 skb->priority = 0; 3792 return 0; 3793 } 3794 3795 bool dev_nit_active(struct net_device *dev); 3796 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3797 3798 extern int netdev_budget; 3799 extern unsigned int netdev_budget_usecs; 3800 3801 /* Called by rtnetlink.c:rtnl_unlock() */ 3802 void netdev_run_todo(void); 3803 3804 /** 3805 * dev_put - release reference to device 3806 * @dev: network device 3807 * 3808 * Release reference to device to allow it to be freed. 3809 */ 3810 static inline void dev_put(struct net_device *dev) 3811 { 3812 this_cpu_dec(*dev->pcpu_refcnt); 3813 } 3814 3815 /** 3816 * dev_hold - get reference to device 3817 * @dev: network device 3818 * 3819 * Hold reference to device to keep it from being freed. 3820 */ 3821 static inline void dev_hold(struct net_device *dev) 3822 { 3823 this_cpu_inc(*dev->pcpu_refcnt); 3824 } 3825 3826 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3827 * and _off may be called from IRQ context, but it is caller 3828 * who is responsible for serialization of these calls. 3829 * 3830 * The name carrier is inappropriate, these functions should really be 3831 * called netif_lowerlayer_*() because they represent the state of any 3832 * kind of lower layer not just hardware media. 3833 */ 3834 3835 void linkwatch_init_dev(struct net_device *dev); 3836 void linkwatch_fire_event(struct net_device *dev); 3837 void linkwatch_forget_dev(struct net_device *dev); 3838 3839 /** 3840 * netif_carrier_ok - test if carrier present 3841 * @dev: network device 3842 * 3843 * Check if carrier is present on device 3844 */ 3845 static inline bool netif_carrier_ok(const struct net_device *dev) 3846 { 3847 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3848 } 3849 3850 unsigned long dev_trans_start(struct net_device *dev); 3851 3852 void __netdev_watchdog_up(struct net_device *dev); 3853 3854 void netif_carrier_on(struct net_device *dev); 3855 3856 void netif_carrier_off(struct net_device *dev); 3857 3858 /** 3859 * netif_dormant_on - mark device as dormant. 3860 * @dev: network device 3861 * 3862 * Mark device as dormant (as per RFC2863). 3863 * 3864 * The dormant state indicates that the relevant interface is not 3865 * actually in a condition to pass packets (i.e., it is not 'up') but is 3866 * in a "pending" state, waiting for some external event. For "on- 3867 * demand" interfaces, this new state identifies the situation where the 3868 * interface is waiting for events to place it in the up state. 3869 */ 3870 static inline void netif_dormant_on(struct net_device *dev) 3871 { 3872 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3873 linkwatch_fire_event(dev); 3874 } 3875 3876 /** 3877 * netif_dormant_off - set device as not dormant. 3878 * @dev: network device 3879 * 3880 * Device is not in dormant state. 3881 */ 3882 static inline void netif_dormant_off(struct net_device *dev) 3883 { 3884 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3885 linkwatch_fire_event(dev); 3886 } 3887 3888 /** 3889 * netif_dormant - test if device is dormant 3890 * @dev: network device 3891 * 3892 * Check if device is dormant. 3893 */ 3894 static inline bool netif_dormant(const struct net_device *dev) 3895 { 3896 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3897 } 3898 3899 3900 /** 3901 * netif_oper_up - test if device is operational 3902 * @dev: network device 3903 * 3904 * Check if carrier is operational 3905 */ 3906 static inline bool netif_oper_up(const struct net_device *dev) 3907 { 3908 return (dev->operstate == IF_OPER_UP || 3909 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3910 } 3911 3912 /** 3913 * netif_device_present - is device available or removed 3914 * @dev: network device 3915 * 3916 * Check if device has not been removed from system. 3917 */ 3918 static inline bool netif_device_present(struct net_device *dev) 3919 { 3920 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3921 } 3922 3923 void netif_device_detach(struct net_device *dev); 3924 3925 void netif_device_attach(struct net_device *dev); 3926 3927 /* 3928 * Network interface message level settings 3929 */ 3930 3931 enum { 3932 NETIF_MSG_DRV_BIT, 3933 NETIF_MSG_PROBE_BIT, 3934 NETIF_MSG_LINK_BIT, 3935 NETIF_MSG_TIMER_BIT, 3936 NETIF_MSG_IFDOWN_BIT, 3937 NETIF_MSG_IFUP_BIT, 3938 NETIF_MSG_RX_ERR_BIT, 3939 NETIF_MSG_TX_ERR_BIT, 3940 NETIF_MSG_TX_QUEUED_BIT, 3941 NETIF_MSG_INTR_BIT, 3942 NETIF_MSG_TX_DONE_BIT, 3943 NETIF_MSG_RX_STATUS_BIT, 3944 NETIF_MSG_PKTDATA_BIT, 3945 NETIF_MSG_HW_BIT, 3946 NETIF_MSG_WOL_BIT, 3947 3948 /* When you add a new bit above, update netif_msg_class_names array 3949 * in net/ethtool/common.c 3950 */ 3951 NETIF_MSG_CLASS_COUNT, 3952 }; 3953 /* Both ethtool_ops interface and internal driver implementation use u32 */ 3954 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 3955 3956 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 3957 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 3958 3959 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 3960 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 3961 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 3962 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 3963 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 3964 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 3965 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 3966 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 3967 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 3968 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 3969 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 3970 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 3971 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 3972 #define NETIF_MSG_HW __NETIF_MSG(HW) 3973 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 3974 3975 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3976 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3977 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3978 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3979 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3980 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3981 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3982 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3983 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3984 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3985 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3986 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3987 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3988 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3989 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3990 3991 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3992 { 3993 /* use default */ 3994 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3995 return default_msg_enable_bits; 3996 if (debug_value == 0) /* no output */ 3997 return 0; 3998 /* set low N bits */ 3999 return (1U << debug_value) - 1; 4000 } 4001 4002 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4003 { 4004 spin_lock(&txq->_xmit_lock); 4005 txq->xmit_lock_owner = cpu; 4006 } 4007 4008 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4009 { 4010 __acquire(&txq->_xmit_lock); 4011 return true; 4012 } 4013 4014 static inline void __netif_tx_release(struct netdev_queue *txq) 4015 { 4016 __release(&txq->_xmit_lock); 4017 } 4018 4019 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4020 { 4021 spin_lock_bh(&txq->_xmit_lock); 4022 txq->xmit_lock_owner = smp_processor_id(); 4023 } 4024 4025 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4026 { 4027 bool ok = spin_trylock(&txq->_xmit_lock); 4028 if (likely(ok)) 4029 txq->xmit_lock_owner = smp_processor_id(); 4030 return ok; 4031 } 4032 4033 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4034 { 4035 txq->xmit_lock_owner = -1; 4036 spin_unlock(&txq->_xmit_lock); 4037 } 4038 4039 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4040 { 4041 txq->xmit_lock_owner = -1; 4042 spin_unlock_bh(&txq->_xmit_lock); 4043 } 4044 4045 static inline void txq_trans_update(struct netdev_queue *txq) 4046 { 4047 if (txq->xmit_lock_owner != -1) 4048 txq->trans_start = jiffies; 4049 } 4050 4051 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4052 static inline void netif_trans_update(struct net_device *dev) 4053 { 4054 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4055 4056 if (txq->trans_start != jiffies) 4057 txq->trans_start = jiffies; 4058 } 4059 4060 /** 4061 * netif_tx_lock - grab network device transmit lock 4062 * @dev: network device 4063 * 4064 * Get network device transmit lock 4065 */ 4066 static inline void netif_tx_lock(struct net_device *dev) 4067 { 4068 unsigned int i; 4069 int cpu; 4070 4071 spin_lock(&dev->tx_global_lock); 4072 cpu = smp_processor_id(); 4073 for (i = 0; i < dev->num_tx_queues; i++) { 4074 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4075 4076 /* We are the only thread of execution doing a 4077 * freeze, but we have to grab the _xmit_lock in 4078 * order to synchronize with threads which are in 4079 * the ->hard_start_xmit() handler and already 4080 * checked the frozen bit. 4081 */ 4082 __netif_tx_lock(txq, cpu); 4083 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4084 __netif_tx_unlock(txq); 4085 } 4086 } 4087 4088 static inline void netif_tx_lock_bh(struct net_device *dev) 4089 { 4090 local_bh_disable(); 4091 netif_tx_lock(dev); 4092 } 4093 4094 static inline void netif_tx_unlock(struct net_device *dev) 4095 { 4096 unsigned int i; 4097 4098 for (i = 0; i < dev->num_tx_queues; i++) { 4099 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4100 4101 /* No need to grab the _xmit_lock here. If the 4102 * queue is not stopped for another reason, we 4103 * force a schedule. 4104 */ 4105 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4106 netif_schedule_queue(txq); 4107 } 4108 spin_unlock(&dev->tx_global_lock); 4109 } 4110 4111 static inline void netif_tx_unlock_bh(struct net_device *dev) 4112 { 4113 netif_tx_unlock(dev); 4114 local_bh_enable(); 4115 } 4116 4117 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4118 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4119 __netif_tx_lock(txq, cpu); \ 4120 } else { \ 4121 __netif_tx_acquire(txq); \ 4122 } \ 4123 } 4124 4125 #define HARD_TX_TRYLOCK(dev, txq) \ 4126 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4127 __netif_tx_trylock(txq) : \ 4128 __netif_tx_acquire(txq)) 4129 4130 #define HARD_TX_UNLOCK(dev, txq) { \ 4131 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4132 __netif_tx_unlock(txq); \ 4133 } else { \ 4134 __netif_tx_release(txq); \ 4135 } \ 4136 } 4137 4138 static inline void netif_tx_disable(struct net_device *dev) 4139 { 4140 unsigned int i; 4141 int cpu; 4142 4143 local_bh_disable(); 4144 cpu = smp_processor_id(); 4145 for (i = 0; i < dev->num_tx_queues; i++) { 4146 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4147 4148 __netif_tx_lock(txq, cpu); 4149 netif_tx_stop_queue(txq); 4150 __netif_tx_unlock(txq); 4151 } 4152 local_bh_enable(); 4153 } 4154 4155 static inline void netif_addr_lock(struct net_device *dev) 4156 { 4157 spin_lock(&dev->addr_list_lock); 4158 } 4159 4160 static inline void netif_addr_lock_bh(struct net_device *dev) 4161 { 4162 spin_lock_bh(&dev->addr_list_lock); 4163 } 4164 4165 static inline void netif_addr_unlock(struct net_device *dev) 4166 { 4167 spin_unlock(&dev->addr_list_lock); 4168 } 4169 4170 static inline void netif_addr_unlock_bh(struct net_device *dev) 4171 { 4172 spin_unlock_bh(&dev->addr_list_lock); 4173 } 4174 4175 /* 4176 * dev_addrs walker. Should be used only for read access. Call with 4177 * rcu_read_lock held. 4178 */ 4179 #define for_each_dev_addr(dev, ha) \ 4180 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4181 4182 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4183 4184 void ether_setup(struct net_device *dev); 4185 4186 /* Support for loadable net-drivers */ 4187 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4188 unsigned char name_assign_type, 4189 void (*setup)(struct net_device *), 4190 unsigned int txqs, unsigned int rxqs); 4191 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4192 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4193 4194 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4195 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4196 count) 4197 4198 int register_netdev(struct net_device *dev); 4199 void unregister_netdev(struct net_device *dev); 4200 4201 /* General hardware address lists handling functions */ 4202 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4203 struct netdev_hw_addr_list *from_list, int addr_len); 4204 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4205 struct netdev_hw_addr_list *from_list, int addr_len); 4206 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4207 struct net_device *dev, 4208 int (*sync)(struct net_device *, const unsigned char *), 4209 int (*unsync)(struct net_device *, 4210 const unsigned char *)); 4211 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4212 struct net_device *dev, 4213 int (*sync)(struct net_device *, 4214 const unsigned char *, int), 4215 int (*unsync)(struct net_device *, 4216 const unsigned char *, int)); 4217 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4218 struct net_device *dev, 4219 int (*unsync)(struct net_device *, 4220 const unsigned char *, int)); 4221 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4222 struct net_device *dev, 4223 int (*unsync)(struct net_device *, 4224 const unsigned char *)); 4225 void __hw_addr_init(struct netdev_hw_addr_list *list); 4226 4227 /* Functions used for device addresses handling */ 4228 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4229 unsigned char addr_type); 4230 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4231 unsigned char addr_type); 4232 void dev_addr_flush(struct net_device *dev); 4233 int dev_addr_init(struct net_device *dev); 4234 4235 /* Functions used for unicast addresses handling */ 4236 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4237 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4238 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4239 int dev_uc_sync(struct net_device *to, struct net_device *from); 4240 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4241 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4242 void dev_uc_flush(struct net_device *dev); 4243 void dev_uc_init(struct net_device *dev); 4244 4245 /** 4246 * __dev_uc_sync - Synchonize device's unicast list 4247 * @dev: device to sync 4248 * @sync: function to call if address should be added 4249 * @unsync: function to call if address should be removed 4250 * 4251 * Add newly added addresses to the interface, and release 4252 * addresses that have been deleted. 4253 */ 4254 static inline int __dev_uc_sync(struct net_device *dev, 4255 int (*sync)(struct net_device *, 4256 const unsigned char *), 4257 int (*unsync)(struct net_device *, 4258 const unsigned char *)) 4259 { 4260 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4261 } 4262 4263 /** 4264 * __dev_uc_unsync - Remove synchronized addresses from device 4265 * @dev: device to sync 4266 * @unsync: function to call if address should be removed 4267 * 4268 * Remove all addresses that were added to the device by dev_uc_sync(). 4269 */ 4270 static inline void __dev_uc_unsync(struct net_device *dev, 4271 int (*unsync)(struct net_device *, 4272 const unsigned char *)) 4273 { 4274 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4275 } 4276 4277 /* Functions used for multicast addresses handling */ 4278 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4279 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4280 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4281 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4282 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4283 int dev_mc_sync(struct net_device *to, struct net_device *from); 4284 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4285 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4286 void dev_mc_flush(struct net_device *dev); 4287 void dev_mc_init(struct net_device *dev); 4288 4289 /** 4290 * __dev_mc_sync - Synchonize device's multicast list 4291 * @dev: device to sync 4292 * @sync: function to call if address should be added 4293 * @unsync: function to call if address should be removed 4294 * 4295 * Add newly added addresses to the interface, and release 4296 * addresses that have been deleted. 4297 */ 4298 static inline int __dev_mc_sync(struct net_device *dev, 4299 int (*sync)(struct net_device *, 4300 const unsigned char *), 4301 int (*unsync)(struct net_device *, 4302 const unsigned char *)) 4303 { 4304 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4305 } 4306 4307 /** 4308 * __dev_mc_unsync - Remove synchronized addresses from device 4309 * @dev: device to sync 4310 * @unsync: function to call if address should be removed 4311 * 4312 * Remove all addresses that were added to the device by dev_mc_sync(). 4313 */ 4314 static inline void __dev_mc_unsync(struct net_device *dev, 4315 int (*unsync)(struct net_device *, 4316 const unsigned char *)) 4317 { 4318 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4319 } 4320 4321 /* Functions used for secondary unicast and multicast support */ 4322 void dev_set_rx_mode(struct net_device *dev); 4323 void __dev_set_rx_mode(struct net_device *dev); 4324 int dev_set_promiscuity(struct net_device *dev, int inc); 4325 int dev_set_allmulti(struct net_device *dev, int inc); 4326 void netdev_state_change(struct net_device *dev); 4327 void netdev_notify_peers(struct net_device *dev); 4328 void netdev_features_change(struct net_device *dev); 4329 /* Load a device via the kmod */ 4330 void dev_load(struct net *net, const char *name); 4331 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4332 struct rtnl_link_stats64 *storage); 4333 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4334 const struct net_device_stats *netdev_stats); 4335 4336 extern int netdev_max_backlog; 4337 extern int netdev_tstamp_prequeue; 4338 extern int weight_p; 4339 extern int dev_weight_rx_bias; 4340 extern int dev_weight_tx_bias; 4341 extern int dev_rx_weight; 4342 extern int dev_tx_weight; 4343 extern int gro_normal_batch; 4344 4345 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4346 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4347 struct list_head **iter); 4348 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4349 struct list_head **iter); 4350 4351 /* iterate through upper list, must be called under RCU read lock */ 4352 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4353 for (iter = &(dev)->adj_list.upper, \ 4354 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4355 updev; \ 4356 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4357 4358 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4359 int (*fn)(struct net_device *upper_dev, 4360 void *data), 4361 void *data); 4362 4363 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4364 struct net_device *upper_dev); 4365 4366 bool netdev_has_any_upper_dev(struct net_device *dev); 4367 4368 void *netdev_lower_get_next_private(struct net_device *dev, 4369 struct list_head **iter); 4370 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4371 struct list_head **iter); 4372 4373 #define netdev_for_each_lower_private(dev, priv, iter) \ 4374 for (iter = (dev)->adj_list.lower.next, \ 4375 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4376 priv; \ 4377 priv = netdev_lower_get_next_private(dev, &(iter))) 4378 4379 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4380 for (iter = &(dev)->adj_list.lower, \ 4381 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4382 priv; \ 4383 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4384 4385 void *netdev_lower_get_next(struct net_device *dev, 4386 struct list_head **iter); 4387 4388 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4389 for (iter = (dev)->adj_list.lower.next, \ 4390 ldev = netdev_lower_get_next(dev, &(iter)); \ 4391 ldev; \ 4392 ldev = netdev_lower_get_next(dev, &(iter))) 4393 4394 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4395 struct list_head **iter); 4396 int netdev_walk_all_lower_dev(struct net_device *dev, 4397 int (*fn)(struct net_device *lower_dev, 4398 void *data), 4399 void *data); 4400 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4401 int (*fn)(struct net_device *lower_dev, 4402 void *data), 4403 void *data); 4404 4405 void *netdev_adjacent_get_private(struct list_head *adj_list); 4406 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4407 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4408 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4409 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4410 struct netlink_ext_ack *extack); 4411 int netdev_master_upper_dev_link(struct net_device *dev, 4412 struct net_device *upper_dev, 4413 void *upper_priv, void *upper_info, 4414 struct netlink_ext_ack *extack); 4415 void netdev_upper_dev_unlink(struct net_device *dev, 4416 struct net_device *upper_dev); 4417 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4418 struct net_device *new_dev, 4419 struct net_device *dev, 4420 struct netlink_ext_ack *extack); 4421 void netdev_adjacent_change_commit(struct net_device *old_dev, 4422 struct net_device *new_dev, 4423 struct net_device *dev); 4424 void netdev_adjacent_change_abort(struct net_device *old_dev, 4425 struct net_device *new_dev, 4426 struct net_device *dev); 4427 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4428 void *netdev_lower_dev_get_private(struct net_device *dev, 4429 struct net_device *lower_dev); 4430 void netdev_lower_state_changed(struct net_device *lower_dev, 4431 void *lower_state_info); 4432 4433 /* RSS keys are 40 or 52 bytes long */ 4434 #define NETDEV_RSS_KEY_LEN 52 4435 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4436 void netdev_rss_key_fill(void *buffer, size_t len); 4437 4438 int skb_checksum_help(struct sk_buff *skb); 4439 int skb_crc32c_csum_help(struct sk_buff *skb); 4440 int skb_csum_hwoffload_help(struct sk_buff *skb, 4441 const netdev_features_t features); 4442 4443 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4444 netdev_features_t features, bool tx_path); 4445 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4446 netdev_features_t features); 4447 4448 struct netdev_bonding_info { 4449 ifslave slave; 4450 ifbond master; 4451 }; 4452 4453 struct netdev_notifier_bonding_info { 4454 struct netdev_notifier_info info; /* must be first */ 4455 struct netdev_bonding_info bonding_info; 4456 }; 4457 4458 void netdev_bonding_info_change(struct net_device *dev, 4459 struct netdev_bonding_info *bonding_info); 4460 4461 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4462 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4463 #else 4464 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4465 const void *data) 4466 { 4467 } 4468 #endif 4469 4470 static inline 4471 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4472 { 4473 return __skb_gso_segment(skb, features, true); 4474 } 4475 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4476 4477 static inline bool can_checksum_protocol(netdev_features_t features, 4478 __be16 protocol) 4479 { 4480 if (protocol == htons(ETH_P_FCOE)) 4481 return !!(features & NETIF_F_FCOE_CRC); 4482 4483 /* Assume this is an IP checksum (not SCTP CRC) */ 4484 4485 if (features & NETIF_F_HW_CSUM) { 4486 /* Can checksum everything */ 4487 return true; 4488 } 4489 4490 switch (protocol) { 4491 case htons(ETH_P_IP): 4492 return !!(features & NETIF_F_IP_CSUM); 4493 case htons(ETH_P_IPV6): 4494 return !!(features & NETIF_F_IPV6_CSUM); 4495 default: 4496 return false; 4497 } 4498 } 4499 4500 #ifdef CONFIG_BUG 4501 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4502 #else 4503 static inline void netdev_rx_csum_fault(struct net_device *dev, 4504 struct sk_buff *skb) 4505 { 4506 } 4507 #endif 4508 /* rx skb timestamps */ 4509 void net_enable_timestamp(void); 4510 void net_disable_timestamp(void); 4511 4512 #ifdef CONFIG_PROC_FS 4513 int __init dev_proc_init(void); 4514 #else 4515 #define dev_proc_init() 0 4516 #endif 4517 4518 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4519 struct sk_buff *skb, struct net_device *dev, 4520 bool more) 4521 { 4522 __this_cpu_write(softnet_data.xmit.more, more); 4523 return ops->ndo_start_xmit(skb, dev); 4524 } 4525 4526 static inline bool netdev_xmit_more(void) 4527 { 4528 return __this_cpu_read(softnet_data.xmit.more); 4529 } 4530 4531 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4532 struct netdev_queue *txq, bool more) 4533 { 4534 const struct net_device_ops *ops = dev->netdev_ops; 4535 netdev_tx_t rc; 4536 4537 rc = __netdev_start_xmit(ops, skb, dev, more); 4538 if (rc == NETDEV_TX_OK) 4539 txq_trans_update(txq); 4540 4541 return rc; 4542 } 4543 4544 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4545 const void *ns); 4546 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4547 const void *ns); 4548 4549 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4550 { 4551 return netdev_class_create_file_ns(class_attr, NULL); 4552 } 4553 4554 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4555 { 4556 netdev_class_remove_file_ns(class_attr, NULL); 4557 } 4558 4559 extern const struct kobj_ns_type_operations net_ns_type_operations; 4560 4561 const char *netdev_drivername(const struct net_device *dev); 4562 4563 void linkwatch_run_queue(void); 4564 4565 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4566 netdev_features_t f2) 4567 { 4568 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4569 if (f1 & NETIF_F_HW_CSUM) 4570 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4571 else 4572 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4573 } 4574 4575 return f1 & f2; 4576 } 4577 4578 static inline netdev_features_t netdev_get_wanted_features( 4579 struct net_device *dev) 4580 { 4581 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4582 } 4583 netdev_features_t netdev_increment_features(netdev_features_t all, 4584 netdev_features_t one, netdev_features_t mask); 4585 4586 /* Allow TSO being used on stacked device : 4587 * Performing the GSO segmentation before last device 4588 * is a performance improvement. 4589 */ 4590 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4591 netdev_features_t mask) 4592 { 4593 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4594 } 4595 4596 int __netdev_update_features(struct net_device *dev); 4597 void netdev_update_features(struct net_device *dev); 4598 void netdev_change_features(struct net_device *dev); 4599 4600 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4601 struct net_device *dev); 4602 4603 netdev_features_t passthru_features_check(struct sk_buff *skb, 4604 struct net_device *dev, 4605 netdev_features_t features); 4606 netdev_features_t netif_skb_features(struct sk_buff *skb); 4607 4608 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4609 { 4610 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4611 4612 /* check flags correspondence */ 4613 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4614 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4615 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4616 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4617 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4618 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4619 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4620 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4621 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4622 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4623 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4624 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4625 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4626 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4627 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4628 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4629 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4630 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4631 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4632 4633 return (features & feature) == feature; 4634 } 4635 4636 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4637 { 4638 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4639 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4640 } 4641 4642 static inline bool netif_needs_gso(struct sk_buff *skb, 4643 netdev_features_t features) 4644 { 4645 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4646 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4647 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4648 } 4649 4650 static inline void netif_set_gso_max_size(struct net_device *dev, 4651 unsigned int size) 4652 { 4653 dev->gso_max_size = size; 4654 } 4655 4656 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4657 int pulled_hlen, u16 mac_offset, 4658 int mac_len) 4659 { 4660 skb->protocol = protocol; 4661 skb->encapsulation = 1; 4662 skb_push(skb, pulled_hlen); 4663 skb_reset_transport_header(skb); 4664 skb->mac_header = mac_offset; 4665 skb->network_header = skb->mac_header + mac_len; 4666 skb->mac_len = mac_len; 4667 } 4668 4669 static inline bool netif_is_macsec(const struct net_device *dev) 4670 { 4671 return dev->priv_flags & IFF_MACSEC; 4672 } 4673 4674 static inline bool netif_is_macvlan(const struct net_device *dev) 4675 { 4676 return dev->priv_flags & IFF_MACVLAN; 4677 } 4678 4679 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4680 { 4681 return dev->priv_flags & IFF_MACVLAN_PORT; 4682 } 4683 4684 static inline bool netif_is_bond_master(const struct net_device *dev) 4685 { 4686 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4687 } 4688 4689 static inline bool netif_is_bond_slave(const struct net_device *dev) 4690 { 4691 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4692 } 4693 4694 static inline bool netif_supports_nofcs(struct net_device *dev) 4695 { 4696 return dev->priv_flags & IFF_SUPP_NOFCS; 4697 } 4698 4699 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4700 { 4701 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4702 } 4703 4704 static inline bool netif_is_l3_master(const struct net_device *dev) 4705 { 4706 return dev->priv_flags & IFF_L3MDEV_MASTER; 4707 } 4708 4709 static inline bool netif_is_l3_slave(const struct net_device *dev) 4710 { 4711 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4712 } 4713 4714 static inline bool netif_is_bridge_master(const struct net_device *dev) 4715 { 4716 return dev->priv_flags & IFF_EBRIDGE; 4717 } 4718 4719 static inline bool netif_is_bridge_port(const struct net_device *dev) 4720 { 4721 return dev->priv_flags & IFF_BRIDGE_PORT; 4722 } 4723 4724 static inline bool netif_is_ovs_master(const struct net_device *dev) 4725 { 4726 return dev->priv_flags & IFF_OPENVSWITCH; 4727 } 4728 4729 static inline bool netif_is_ovs_port(const struct net_device *dev) 4730 { 4731 return dev->priv_flags & IFF_OVS_DATAPATH; 4732 } 4733 4734 static inline bool netif_is_team_master(const struct net_device *dev) 4735 { 4736 return dev->priv_flags & IFF_TEAM; 4737 } 4738 4739 static inline bool netif_is_team_port(const struct net_device *dev) 4740 { 4741 return dev->priv_flags & IFF_TEAM_PORT; 4742 } 4743 4744 static inline bool netif_is_lag_master(const struct net_device *dev) 4745 { 4746 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4747 } 4748 4749 static inline bool netif_is_lag_port(const struct net_device *dev) 4750 { 4751 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4752 } 4753 4754 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4755 { 4756 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4757 } 4758 4759 static inline bool netif_is_failover(const struct net_device *dev) 4760 { 4761 return dev->priv_flags & IFF_FAILOVER; 4762 } 4763 4764 static inline bool netif_is_failover_slave(const struct net_device *dev) 4765 { 4766 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4767 } 4768 4769 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4770 static inline void netif_keep_dst(struct net_device *dev) 4771 { 4772 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4773 } 4774 4775 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4776 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4777 { 4778 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4779 return dev->priv_flags & IFF_MACSEC; 4780 } 4781 4782 extern struct pernet_operations __net_initdata loopback_net_ops; 4783 4784 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4785 4786 /* netdev_printk helpers, similar to dev_printk */ 4787 4788 static inline const char *netdev_name(const struct net_device *dev) 4789 { 4790 if (!dev->name[0] || strchr(dev->name, '%')) 4791 return "(unnamed net_device)"; 4792 return dev->name; 4793 } 4794 4795 static inline bool netdev_unregistering(const struct net_device *dev) 4796 { 4797 return dev->reg_state == NETREG_UNREGISTERING; 4798 } 4799 4800 static inline const char *netdev_reg_state(const struct net_device *dev) 4801 { 4802 switch (dev->reg_state) { 4803 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4804 case NETREG_REGISTERED: return ""; 4805 case NETREG_UNREGISTERING: return " (unregistering)"; 4806 case NETREG_UNREGISTERED: return " (unregistered)"; 4807 case NETREG_RELEASED: return " (released)"; 4808 case NETREG_DUMMY: return " (dummy)"; 4809 } 4810 4811 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4812 return " (unknown)"; 4813 } 4814 4815 __printf(3, 4) __cold 4816 void netdev_printk(const char *level, const struct net_device *dev, 4817 const char *format, ...); 4818 __printf(2, 3) __cold 4819 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4820 __printf(2, 3) __cold 4821 void netdev_alert(const struct net_device *dev, const char *format, ...); 4822 __printf(2, 3) __cold 4823 void netdev_crit(const struct net_device *dev, const char *format, ...); 4824 __printf(2, 3) __cold 4825 void netdev_err(const struct net_device *dev, const char *format, ...); 4826 __printf(2, 3) __cold 4827 void netdev_warn(const struct net_device *dev, const char *format, ...); 4828 __printf(2, 3) __cold 4829 void netdev_notice(const struct net_device *dev, const char *format, ...); 4830 __printf(2, 3) __cold 4831 void netdev_info(const struct net_device *dev, const char *format, ...); 4832 4833 #define netdev_level_once(level, dev, fmt, ...) \ 4834 do { \ 4835 static bool __print_once __read_mostly; \ 4836 \ 4837 if (!__print_once) { \ 4838 __print_once = true; \ 4839 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4840 } \ 4841 } while (0) 4842 4843 #define netdev_emerg_once(dev, fmt, ...) \ 4844 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4845 #define netdev_alert_once(dev, fmt, ...) \ 4846 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4847 #define netdev_crit_once(dev, fmt, ...) \ 4848 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4849 #define netdev_err_once(dev, fmt, ...) \ 4850 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4851 #define netdev_warn_once(dev, fmt, ...) \ 4852 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4853 #define netdev_notice_once(dev, fmt, ...) \ 4854 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4855 #define netdev_info_once(dev, fmt, ...) \ 4856 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4857 4858 #define MODULE_ALIAS_NETDEV(device) \ 4859 MODULE_ALIAS("netdev-" device) 4860 4861 #if defined(CONFIG_DYNAMIC_DEBUG) 4862 #define netdev_dbg(__dev, format, args...) \ 4863 do { \ 4864 dynamic_netdev_dbg(__dev, format, ##args); \ 4865 } while (0) 4866 #elif defined(DEBUG) 4867 #define netdev_dbg(__dev, format, args...) \ 4868 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4869 #else 4870 #define netdev_dbg(__dev, format, args...) \ 4871 ({ \ 4872 if (0) \ 4873 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4874 }) 4875 #endif 4876 4877 #if defined(VERBOSE_DEBUG) 4878 #define netdev_vdbg netdev_dbg 4879 #else 4880 4881 #define netdev_vdbg(dev, format, args...) \ 4882 ({ \ 4883 if (0) \ 4884 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4885 0; \ 4886 }) 4887 #endif 4888 4889 /* 4890 * netdev_WARN() acts like dev_printk(), but with the key difference 4891 * of using a WARN/WARN_ON to get the message out, including the 4892 * file/line information and a backtrace. 4893 */ 4894 #define netdev_WARN(dev, format, args...) \ 4895 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4896 netdev_reg_state(dev), ##args) 4897 4898 #define netdev_WARN_ONCE(dev, format, args...) \ 4899 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4900 netdev_reg_state(dev), ##args) 4901 4902 /* netif printk helpers, similar to netdev_printk */ 4903 4904 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4905 do { \ 4906 if (netif_msg_##type(priv)) \ 4907 netdev_printk(level, (dev), fmt, ##args); \ 4908 } while (0) 4909 4910 #define netif_level(level, priv, type, dev, fmt, args...) \ 4911 do { \ 4912 if (netif_msg_##type(priv)) \ 4913 netdev_##level(dev, fmt, ##args); \ 4914 } while (0) 4915 4916 #define netif_emerg(priv, type, dev, fmt, args...) \ 4917 netif_level(emerg, priv, type, dev, fmt, ##args) 4918 #define netif_alert(priv, type, dev, fmt, args...) \ 4919 netif_level(alert, priv, type, dev, fmt, ##args) 4920 #define netif_crit(priv, type, dev, fmt, args...) \ 4921 netif_level(crit, priv, type, dev, fmt, ##args) 4922 #define netif_err(priv, type, dev, fmt, args...) \ 4923 netif_level(err, priv, type, dev, fmt, ##args) 4924 #define netif_warn(priv, type, dev, fmt, args...) \ 4925 netif_level(warn, priv, type, dev, fmt, ##args) 4926 #define netif_notice(priv, type, dev, fmt, args...) \ 4927 netif_level(notice, priv, type, dev, fmt, ##args) 4928 #define netif_info(priv, type, dev, fmt, args...) \ 4929 netif_level(info, priv, type, dev, fmt, ##args) 4930 4931 #if defined(CONFIG_DYNAMIC_DEBUG) 4932 #define netif_dbg(priv, type, netdev, format, args...) \ 4933 do { \ 4934 if (netif_msg_##type(priv)) \ 4935 dynamic_netdev_dbg(netdev, format, ##args); \ 4936 } while (0) 4937 #elif defined(DEBUG) 4938 #define netif_dbg(priv, type, dev, format, args...) \ 4939 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4940 #else 4941 #define netif_dbg(priv, type, dev, format, args...) \ 4942 ({ \ 4943 if (0) \ 4944 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4945 0; \ 4946 }) 4947 #endif 4948 4949 /* if @cond then downgrade to debug, else print at @level */ 4950 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4951 do { \ 4952 if (cond) \ 4953 netif_dbg(priv, type, netdev, fmt, ##args); \ 4954 else \ 4955 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4956 } while (0) 4957 4958 #if defined(VERBOSE_DEBUG) 4959 #define netif_vdbg netif_dbg 4960 #else 4961 #define netif_vdbg(priv, type, dev, format, args...) \ 4962 ({ \ 4963 if (0) \ 4964 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4965 0; \ 4966 }) 4967 #endif 4968 4969 /* 4970 * The list of packet types we will receive (as opposed to discard) 4971 * and the routines to invoke. 4972 * 4973 * Why 16. Because with 16 the only overlap we get on a hash of the 4974 * low nibble of the protocol value is RARP/SNAP/X.25. 4975 * 4976 * 0800 IP 4977 * 0001 802.3 4978 * 0002 AX.25 4979 * 0004 802.2 4980 * 8035 RARP 4981 * 0005 SNAP 4982 * 0805 X.25 4983 * 0806 ARP 4984 * 8137 IPX 4985 * 0009 Localtalk 4986 * 86DD IPv6 4987 */ 4988 #define PTYPE_HASH_SIZE (16) 4989 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4990 4991 extern struct net_device *blackhole_netdev; 4992 4993 #endif /* _LINUX_NETDEVICE_H */ 4994