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