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