1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void netdev_set_default_ethtool_ops(struct net_device *dev, 74 const struct ethtool_ops *ops); 75 76 /* Backlog congestion levels */ 77 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 78 #define NET_RX_DROP 1 /* packet dropped */ 79 80 #define MAX_NEST_DEV 8 81 82 /* 83 * Transmit return codes: transmit return codes originate from three different 84 * namespaces: 85 * 86 * - qdisc return codes 87 * - driver transmit return codes 88 * - errno values 89 * 90 * Drivers are allowed to return any one of those in their hard_start_xmit() 91 * function. Real network devices commonly used with qdiscs should only return 92 * the driver transmit return codes though - when qdiscs are used, the actual 93 * transmission happens asynchronously, so the value is not propagated to 94 * higher layers. Virtual network devices transmit synchronously; in this case 95 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 96 * others are propagated to higher layers. 97 */ 98 99 /* qdisc ->enqueue() return codes. */ 100 #define NET_XMIT_SUCCESS 0x00 101 #define NET_XMIT_DROP 0x01 /* skb dropped */ 102 #define NET_XMIT_CN 0x02 /* congestion notification */ 103 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 104 105 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 106 * indicates that the device will soon be dropping packets, or already drops 107 * some packets of the same priority; prompting us to send less aggressively. */ 108 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 109 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 110 111 /* Driver transmit return codes */ 112 #define NETDEV_TX_MASK 0xf0 113 114 enum netdev_tx { 115 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 116 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 117 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 118 }; 119 typedef enum netdev_tx netdev_tx_t; 120 121 /* 122 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 123 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 124 */ 125 static inline bool dev_xmit_complete(int rc) 126 { 127 /* 128 * Positive cases with an skb consumed by a driver: 129 * - successful transmission (rc == NETDEV_TX_OK) 130 * - error while transmitting (rc < 0) 131 * - error while queueing to a different device (rc & NET_XMIT_MASK) 132 */ 133 if (likely(rc < NET_XMIT_MASK)) 134 return true; 135 136 return false; 137 } 138 139 /* 140 * Compute the worst-case header length according to the protocols 141 * used. 142 */ 143 144 #if defined(CONFIG_HYPERV_NET) 145 # define LL_MAX_HEADER 128 146 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 147 # if defined(CONFIG_MAC80211_MESH) 148 # define LL_MAX_HEADER 128 149 # else 150 # define LL_MAX_HEADER 96 151 # endif 152 #else 153 # define LL_MAX_HEADER 32 154 #endif 155 156 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 157 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 158 #define MAX_HEADER LL_MAX_HEADER 159 #else 160 #define MAX_HEADER (LL_MAX_HEADER + 48) 161 #endif 162 163 /* 164 * Old network device statistics. Fields are native words 165 * (unsigned long) so they can be read and written atomically. 166 */ 167 168 struct net_device_stats { 169 unsigned long rx_packets; 170 unsigned long tx_packets; 171 unsigned long rx_bytes; 172 unsigned long tx_bytes; 173 unsigned long rx_errors; 174 unsigned long tx_errors; 175 unsigned long rx_dropped; 176 unsigned long tx_dropped; 177 unsigned long multicast; 178 unsigned long collisions; 179 unsigned long rx_length_errors; 180 unsigned long rx_over_errors; 181 unsigned long rx_crc_errors; 182 unsigned long rx_frame_errors; 183 unsigned long rx_fifo_errors; 184 unsigned long rx_missed_errors; 185 unsigned long tx_aborted_errors; 186 unsigned long tx_carrier_errors; 187 unsigned long tx_fifo_errors; 188 unsigned long tx_heartbeat_errors; 189 unsigned long tx_window_errors; 190 unsigned long rx_compressed; 191 unsigned long tx_compressed; 192 }; 193 194 195 #include <linux/cache.h> 196 #include <linux/skbuff.h> 197 198 #ifdef CONFIG_RPS 199 #include <linux/static_key.h> 200 extern struct static_key_false rps_needed; 201 extern struct static_key_false rfs_needed; 202 #endif 203 204 struct neighbour; 205 struct neigh_parms; 206 struct sk_buff; 207 208 struct netdev_hw_addr { 209 struct list_head list; 210 unsigned char addr[MAX_ADDR_LEN]; 211 unsigned char type; 212 #define NETDEV_HW_ADDR_T_LAN 1 213 #define NETDEV_HW_ADDR_T_SAN 2 214 #define NETDEV_HW_ADDR_T_SLAVE 3 215 #define NETDEV_HW_ADDR_T_UNICAST 4 216 #define NETDEV_HW_ADDR_T_MULTICAST 5 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 }; 351 352 enum { 353 NAPI_STATE_SCHED, /* Poll is scheduled */ 354 NAPI_STATE_MISSED, /* reschedule a napi */ 355 NAPI_STATE_DISABLE, /* Disable pending */ 356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 357 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 358 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 359 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 360 }; 361 362 enum { 363 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 364 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 365 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 366 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 367 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 368 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 369 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 370 }; 371 372 enum gro_result { 373 GRO_MERGED, 374 GRO_MERGED_FREE, 375 GRO_HELD, 376 GRO_NORMAL, 377 GRO_DROP, 378 GRO_CONSUMED, 379 }; 380 typedef enum gro_result gro_result_t; 381 382 /* 383 * enum rx_handler_result - Possible return values for rx_handlers. 384 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 385 * further. 386 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 387 * case skb->dev was changed by rx_handler. 388 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 389 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 390 * 391 * rx_handlers are functions called from inside __netif_receive_skb(), to do 392 * special processing of the skb, prior to delivery to protocol handlers. 393 * 394 * Currently, a net_device can only have a single rx_handler registered. Trying 395 * to register a second rx_handler will return -EBUSY. 396 * 397 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 398 * To unregister a rx_handler on a net_device, use 399 * netdev_rx_handler_unregister(). 400 * 401 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 402 * do with the skb. 403 * 404 * If the rx_handler consumed the skb in some way, it should return 405 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 406 * the skb to be delivered in some other way. 407 * 408 * If the rx_handler changed skb->dev, to divert the skb to another 409 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 410 * new device will be called if it exists. 411 * 412 * If the rx_handler decides the skb should be ignored, it should return 413 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 414 * are registered on exact device (ptype->dev == skb->dev). 415 * 416 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 417 * delivered, it should return RX_HANDLER_PASS. 418 * 419 * A device without a registered rx_handler will behave as if rx_handler 420 * returned RX_HANDLER_PASS. 421 */ 422 423 enum rx_handler_result { 424 RX_HANDLER_CONSUMED, 425 RX_HANDLER_ANOTHER, 426 RX_HANDLER_EXACT, 427 RX_HANDLER_PASS, 428 }; 429 typedef enum rx_handler_result rx_handler_result_t; 430 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 431 432 void __napi_schedule(struct napi_struct *n); 433 void __napi_schedule_irqoff(struct napi_struct *n); 434 435 static inline bool napi_disable_pending(struct napi_struct *n) 436 { 437 return test_bit(NAPI_STATE_DISABLE, &n->state); 438 } 439 440 bool napi_schedule_prep(struct napi_struct *n); 441 442 /** 443 * napi_schedule - schedule NAPI poll 444 * @n: NAPI context 445 * 446 * Schedule NAPI poll routine to be called if it is not already 447 * running. 448 */ 449 static inline void napi_schedule(struct napi_struct *n) 450 { 451 if (napi_schedule_prep(n)) 452 __napi_schedule(n); 453 } 454 455 /** 456 * napi_schedule_irqoff - schedule NAPI poll 457 * @n: NAPI context 458 * 459 * Variant of napi_schedule(), assuming hard irqs are masked. 460 */ 461 static inline void napi_schedule_irqoff(struct napi_struct *n) 462 { 463 if (napi_schedule_prep(n)) 464 __napi_schedule_irqoff(n); 465 } 466 467 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 468 static inline bool napi_reschedule(struct napi_struct *napi) 469 { 470 if (napi_schedule_prep(napi)) { 471 __napi_schedule(napi); 472 return true; 473 } 474 return false; 475 } 476 477 bool napi_complete_done(struct napi_struct *n, int work_done); 478 /** 479 * napi_complete - NAPI processing complete 480 * @n: NAPI context 481 * 482 * Mark NAPI processing as complete. 483 * Consider using napi_complete_done() instead. 484 * Return false if device should avoid rearming interrupts. 485 */ 486 static inline bool napi_complete(struct napi_struct *n) 487 { 488 return napi_complete_done(n, 0); 489 } 490 491 /** 492 * napi_hash_del - remove a NAPI from global table 493 * @napi: NAPI context 494 * 495 * Warning: caller must observe RCU grace period 496 * before freeing memory containing @napi, if 497 * this function returns true. 498 * Note: core networking stack automatically calls it 499 * from netif_napi_del(). 500 * Drivers might want to call this helper to combine all 501 * the needed RCU grace periods into a single one. 502 */ 503 bool napi_hash_del(struct napi_struct *napi); 504 505 /** 506 * napi_disable - prevent NAPI from scheduling 507 * @n: NAPI context 508 * 509 * Stop NAPI from being scheduled on this context. 510 * Waits till any outstanding processing completes. 511 */ 512 void napi_disable(struct napi_struct *n); 513 514 /** 515 * napi_enable - enable NAPI scheduling 516 * @n: NAPI context 517 * 518 * Resume NAPI from being scheduled on this context. 519 * Must be paired with napi_disable. 520 */ 521 static inline void napi_enable(struct napi_struct *n) 522 { 523 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 524 smp_mb__before_atomic(); 525 clear_bit(NAPI_STATE_SCHED, &n->state); 526 clear_bit(NAPI_STATE_NPSVC, &n->state); 527 } 528 529 /** 530 * napi_synchronize - wait until NAPI is not running 531 * @n: NAPI context 532 * 533 * Wait until NAPI is done being scheduled on this context. 534 * Waits till any outstanding processing completes but 535 * does not disable future activations. 536 */ 537 static inline void napi_synchronize(const struct napi_struct *n) 538 { 539 if (IS_ENABLED(CONFIG_SMP)) 540 while (test_bit(NAPI_STATE_SCHED, &n->state)) 541 msleep(1); 542 else 543 barrier(); 544 } 545 546 /** 547 * napi_if_scheduled_mark_missed - if napi is running, set the 548 * NAPIF_STATE_MISSED 549 * @n: NAPI context 550 * 551 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 552 * NAPI is scheduled. 553 **/ 554 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 555 { 556 unsigned long val, new; 557 558 do { 559 val = READ_ONCE(n->state); 560 if (val & NAPIF_STATE_DISABLE) 561 return true; 562 563 if (!(val & NAPIF_STATE_SCHED)) 564 return false; 565 566 new = val | NAPIF_STATE_MISSED; 567 } while (cmpxchg(&n->state, val, new) != val); 568 569 return true; 570 } 571 572 enum netdev_queue_state_t { 573 __QUEUE_STATE_DRV_XOFF, 574 __QUEUE_STATE_STACK_XOFF, 575 __QUEUE_STATE_FROZEN, 576 }; 577 578 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 579 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 580 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 581 582 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 583 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 584 QUEUE_STATE_FROZEN) 585 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 586 QUEUE_STATE_FROZEN) 587 588 /* 589 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 590 * netif_tx_* functions below are used to manipulate this flag. The 591 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 592 * queue independently. The netif_xmit_*stopped functions below are called 593 * to check if the queue has been stopped by the driver or stack (either 594 * of the XOFF bits are set in the state). Drivers should not need to call 595 * netif_xmit*stopped functions, they should only be using netif_tx_*. 596 */ 597 598 struct netdev_queue { 599 /* 600 * read-mostly part 601 */ 602 struct net_device *dev; 603 struct Qdisc __rcu *qdisc; 604 struct Qdisc *qdisc_sleeping; 605 #ifdef CONFIG_SYSFS 606 struct kobject kobj; 607 #endif 608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 609 int numa_node; 610 #endif 611 unsigned long tx_maxrate; 612 /* 613 * Number of TX timeouts for this queue 614 * (/sys/class/net/DEV/Q/trans_timeout) 615 */ 616 unsigned long trans_timeout; 617 618 /* Subordinate device that the queue has been assigned to */ 619 struct net_device *sb_dev; 620 #ifdef CONFIG_XDP_SOCKETS 621 struct xdp_umem *umem; 622 #endif 623 /* 624 * write-mostly part 625 */ 626 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 627 int xmit_lock_owner; 628 /* 629 * Time (in jiffies) of last Tx 630 */ 631 unsigned long trans_start; 632 633 unsigned long state; 634 635 #ifdef CONFIG_BQL 636 struct dql dql; 637 #endif 638 } ____cacheline_aligned_in_smp; 639 640 extern int sysctl_fb_tunnels_only_for_init_net; 641 extern int sysctl_devconf_inherit_init_net; 642 643 static inline bool net_has_fallback_tunnels(const struct net *net) 644 { 645 return net == &init_net || 646 !IS_ENABLED(CONFIG_SYSCTL) || 647 !sysctl_fb_tunnels_only_for_init_net; 648 } 649 650 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 return q->numa_node; 654 #else 655 return NUMA_NO_NODE; 656 #endif 657 } 658 659 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 660 { 661 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 662 q->numa_node = node; 663 #endif 664 } 665 666 #ifdef CONFIG_RPS 667 /* 668 * This structure holds an RPS map which can be of variable length. The 669 * map is an array of CPUs. 670 */ 671 struct rps_map { 672 unsigned int len; 673 struct rcu_head rcu; 674 u16 cpus[]; 675 }; 676 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 677 678 /* 679 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 680 * tail pointer for that CPU's input queue at the time of last enqueue, and 681 * a hardware filter index. 682 */ 683 struct rps_dev_flow { 684 u16 cpu; 685 u16 filter; 686 unsigned int last_qtail; 687 }; 688 #define RPS_NO_FILTER 0xffff 689 690 /* 691 * The rps_dev_flow_table structure contains a table of flow mappings. 692 */ 693 struct rps_dev_flow_table { 694 unsigned int mask; 695 struct rcu_head rcu; 696 struct rps_dev_flow flows[]; 697 }; 698 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 699 ((_num) * sizeof(struct rps_dev_flow))) 700 701 /* 702 * The rps_sock_flow_table contains mappings of flows to the last CPU 703 * on which they were processed by the application (set in recvmsg). 704 * Each entry is a 32bit value. Upper part is the high-order bits 705 * of flow hash, lower part is CPU number. 706 * rps_cpu_mask is used to partition the space, depending on number of 707 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 708 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 709 * meaning we use 32-6=26 bits for the hash. 710 */ 711 struct rps_sock_flow_table { 712 u32 mask; 713 714 u32 ents[] ____cacheline_aligned_in_smp; 715 }; 716 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 717 718 #define RPS_NO_CPU 0xffff 719 720 extern u32 rps_cpu_mask; 721 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 722 723 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 724 u32 hash) 725 { 726 if (table && hash) { 727 unsigned int index = hash & table->mask; 728 u32 val = hash & ~rps_cpu_mask; 729 730 /* We only give a hint, preemption can change CPU under us */ 731 val |= raw_smp_processor_id(); 732 733 if (table->ents[index] != val) 734 table->ents[index] = val; 735 } 736 } 737 738 #ifdef CONFIG_RFS_ACCEL 739 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 740 u16 filter_id); 741 #endif 742 #endif /* CONFIG_RPS */ 743 744 /* This structure contains an instance of an RX queue. */ 745 struct netdev_rx_queue { 746 #ifdef CONFIG_RPS 747 struct rps_map __rcu *rps_map; 748 struct rps_dev_flow_table __rcu *rps_flow_table; 749 #endif 750 struct kobject kobj; 751 struct net_device *dev; 752 struct xdp_rxq_info xdp_rxq; 753 #ifdef CONFIG_XDP_SOCKETS 754 struct xdp_umem *umem; 755 #endif 756 } ____cacheline_aligned_in_smp; 757 758 /* 759 * RX queue sysfs structures and functions. 760 */ 761 struct rx_queue_attribute { 762 struct attribute attr; 763 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 764 ssize_t (*store)(struct netdev_rx_queue *queue, 765 const char *buf, size_t len); 766 }; 767 768 #ifdef CONFIG_XPS 769 /* 770 * This structure holds an XPS map which can be of variable length. The 771 * map is an array of queues. 772 */ 773 struct xps_map { 774 unsigned int len; 775 unsigned int alloc_len; 776 struct rcu_head rcu; 777 u16 queues[]; 778 }; 779 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 780 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 781 - sizeof(struct xps_map)) / sizeof(u16)) 782 783 /* 784 * This structure holds all XPS maps for device. Maps are indexed by CPU. 785 */ 786 struct xps_dev_maps { 787 struct rcu_head rcu; 788 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 789 }; 790 791 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 792 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 793 794 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 795 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 796 797 #endif /* CONFIG_XPS */ 798 799 #define TC_MAX_QUEUE 16 800 #define TC_BITMASK 15 801 /* HW offloaded queuing disciplines txq count and offset maps */ 802 struct netdev_tc_txq { 803 u16 count; 804 u16 offset; 805 }; 806 807 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 808 /* 809 * This structure is to hold information about the device 810 * configured to run FCoE protocol stack. 811 */ 812 struct netdev_fcoe_hbainfo { 813 char manufacturer[64]; 814 char serial_number[64]; 815 char hardware_version[64]; 816 char driver_version[64]; 817 char optionrom_version[64]; 818 char firmware_version[64]; 819 char model[256]; 820 char model_description[256]; 821 }; 822 #endif 823 824 #define MAX_PHYS_ITEM_ID_LEN 32 825 826 /* This structure holds a unique identifier to identify some 827 * physical item (port for example) used by a netdevice. 828 */ 829 struct netdev_phys_item_id { 830 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 831 unsigned char id_len; 832 }; 833 834 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 835 struct netdev_phys_item_id *b) 836 { 837 return a->id_len == b->id_len && 838 memcmp(a->id, b->id, a->id_len) == 0; 839 } 840 841 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 842 struct sk_buff *skb, 843 struct net_device *sb_dev); 844 845 enum tc_setup_type { 846 TC_SETUP_QDISC_MQPRIO, 847 TC_SETUP_CLSU32, 848 TC_SETUP_CLSFLOWER, 849 TC_SETUP_CLSMATCHALL, 850 TC_SETUP_CLSBPF, 851 TC_SETUP_BLOCK, 852 TC_SETUP_QDISC_CBS, 853 TC_SETUP_QDISC_RED, 854 TC_SETUP_QDISC_PRIO, 855 TC_SETUP_QDISC_MQ, 856 TC_SETUP_QDISC_ETF, 857 TC_SETUP_ROOT_QDISC, 858 TC_SETUP_QDISC_GRED, 859 TC_SETUP_QDISC_TAPRIO, 860 TC_SETUP_FT, 861 TC_SETUP_QDISC_ETS, 862 TC_SETUP_QDISC_TBF, 863 TC_SETUP_QDISC_FIFO, 864 }; 865 866 /* These structures hold the attributes of bpf state that are being passed 867 * to the netdevice through the bpf op. 868 */ 869 enum bpf_netdev_command { 870 /* Set or clear a bpf program used in the earliest stages of packet 871 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 872 * is responsible for calling bpf_prog_put on any old progs that are 873 * stored. In case of error, the callee need not release the new prog 874 * reference, but on success it takes ownership and must bpf_prog_put 875 * when it is no longer used. 876 */ 877 XDP_SETUP_PROG, 878 XDP_SETUP_PROG_HW, 879 /* BPF program for offload callbacks, invoked at program load time. */ 880 BPF_OFFLOAD_MAP_ALLOC, 881 BPF_OFFLOAD_MAP_FREE, 882 XDP_SETUP_XSK_UMEM, 883 }; 884 885 struct bpf_prog_offload_ops; 886 struct netlink_ext_ack; 887 struct xdp_umem; 888 struct xdp_dev_bulk_queue; 889 struct bpf_xdp_link; 890 891 enum bpf_xdp_mode { 892 XDP_MODE_SKB = 0, 893 XDP_MODE_DRV = 1, 894 XDP_MODE_HW = 2, 895 __MAX_XDP_MODE 896 }; 897 898 struct bpf_xdp_entity { 899 struct bpf_prog *prog; 900 struct bpf_xdp_link *link; 901 }; 902 903 struct netdev_bpf { 904 enum bpf_netdev_command command; 905 union { 906 /* XDP_SETUP_PROG */ 907 struct { 908 u32 flags; 909 struct bpf_prog *prog; 910 struct netlink_ext_ack *extack; 911 }; 912 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 913 struct { 914 struct bpf_offloaded_map *offmap; 915 }; 916 /* XDP_SETUP_XSK_UMEM */ 917 struct { 918 struct xdp_umem *umem; 919 u16 queue_id; 920 } xsk; 921 }; 922 }; 923 924 /* Flags for ndo_xsk_wakeup. */ 925 #define XDP_WAKEUP_RX (1 << 0) 926 #define XDP_WAKEUP_TX (1 << 1) 927 928 #ifdef CONFIG_XFRM_OFFLOAD 929 struct xfrmdev_ops { 930 int (*xdo_dev_state_add) (struct xfrm_state *x); 931 void (*xdo_dev_state_delete) (struct xfrm_state *x); 932 void (*xdo_dev_state_free) (struct xfrm_state *x); 933 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 934 struct xfrm_state *x); 935 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 936 }; 937 #endif 938 939 struct dev_ifalias { 940 struct rcu_head rcuhead; 941 char ifalias[]; 942 }; 943 944 struct devlink; 945 struct tlsdev_ops; 946 947 struct netdev_name_node { 948 struct hlist_node hlist; 949 struct list_head list; 950 struct net_device *dev; 951 const char *name; 952 }; 953 954 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 955 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 956 957 struct netdev_net_notifier { 958 struct list_head list; 959 struct notifier_block *nb; 960 }; 961 962 /* 963 * This structure defines the management hooks for network devices. 964 * The following hooks can be defined; unless noted otherwise, they are 965 * optional and can be filled with a null pointer. 966 * 967 * int (*ndo_init)(struct net_device *dev); 968 * This function is called once when a network device is registered. 969 * The network device can use this for any late stage initialization 970 * or semantic validation. It can fail with an error code which will 971 * be propagated back to register_netdev. 972 * 973 * void (*ndo_uninit)(struct net_device *dev); 974 * This function is called when device is unregistered or when registration 975 * fails. It is not called if init fails. 976 * 977 * int (*ndo_open)(struct net_device *dev); 978 * This function is called when a network device transitions to the up 979 * state. 980 * 981 * int (*ndo_stop)(struct net_device *dev); 982 * This function is called when a network device transitions to the down 983 * state. 984 * 985 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 986 * struct net_device *dev); 987 * Called when a packet needs to be transmitted. 988 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 989 * the queue before that can happen; it's for obsolete devices and weird 990 * corner cases, but the stack really does a non-trivial amount 991 * of useless work if you return NETDEV_TX_BUSY. 992 * Required; cannot be NULL. 993 * 994 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 995 * struct net_device *dev 996 * netdev_features_t features); 997 * Called by core transmit path to determine if device is capable of 998 * performing offload operations on a given packet. This is to give 999 * the device an opportunity to implement any restrictions that cannot 1000 * be otherwise expressed by feature flags. The check is called with 1001 * the set of features that the stack has calculated and it returns 1002 * those the driver believes to be appropriate. 1003 * 1004 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1005 * struct net_device *sb_dev); 1006 * Called to decide which queue to use when device supports multiple 1007 * transmit queues. 1008 * 1009 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1010 * This function is called to allow device receiver to make 1011 * changes to configuration when multicast or promiscuous is enabled. 1012 * 1013 * void (*ndo_set_rx_mode)(struct net_device *dev); 1014 * This function is called device changes address list filtering. 1015 * If driver handles unicast address filtering, it should set 1016 * IFF_UNICAST_FLT in its priv_flags. 1017 * 1018 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1019 * This function is called when the Media Access Control address 1020 * needs to be changed. If this interface is not defined, the 1021 * MAC address can not be changed. 1022 * 1023 * int (*ndo_validate_addr)(struct net_device *dev); 1024 * Test if Media Access Control address is valid for the device. 1025 * 1026 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1027 * Called when a user requests an ioctl which can't be handled by 1028 * the generic interface code. If not defined ioctls return 1029 * not supported error code. 1030 * 1031 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1032 * Used to set network devices bus interface parameters. This interface 1033 * is retained for legacy reasons; new devices should use the bus 1034 * interface (PCI) for low level management. 1035 * 1036 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1037 * Called when a user wants to change the Maximum Transfer Unit 1038 * of a device. 1039 * 1040 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1041 * Callback used when the transmitter has not made any progress 1042 * for dev->watchdog ticks. 1043 * 1044 * void (*ndo_get_stats64)(struct net_device *dev, 1045 * struct rtnl_link_stats64 *storage); 1046 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1047 * Called when a user wants to get the network device usage 1048 * statistics. Drivers must do one of the following: 1049 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1050 * rtnl_link_stats64 structure passed by the caller. 1051 * 2. Define @ndo_get_stats to update a net_device_stats structure 1052 * (which should normally be dev->stats) and return a pointer to 1053 * it. The structure may be changed asynchronously only if each 1054 * field is written atomically. 1055 * 3. Update dev->stats asynchronously and atomically, and define 1056 * neither operation. 1057 * 1058 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1059 * Return true if this device supports offload stats of this attr_id. 1060 * 1061 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1062 * void *attr_data) 1063 * Get statistics for offload operations by attr_id. Write it into the 1064 * attr_data pointer. 1065 * 1066 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1067 * If device supports VLAN filtering this function is called when a 1068 * VLAN id is registered. 1069 * 1070 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1071 * If device supports VLAN filtering this function is called when a 1072 * VLAN id is unregistered. 1073 * 1074 * void (*ndo_poll_controller)(struct net_device *dev); 1075 * 1076 * SR-IOV management functions. 1077 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1078 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1079 * u8 qos, __be16 proto); 1080 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1081 * int max_tx_rate); 1082 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1083 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1084 * int (*ndo_get_vf_config)(struct net_device *dev, 1085 * int vf, struct ifla_vf_info *ivf); 1086 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1087 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1088 * struct nlattr *port[]); 1089 * 1090 * Enable or disable the VF ability to query its RSS Redirection Table and 1091 * Hash Key. This is needed since on some devices VF share this information 1092 * with PF and querying it may introduce a theoretical security risk. 1093 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1094 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1095 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1096 * void *type_data); 1097 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1098 * This is always called from the stack with the rtnl lock held and netif 1099 * tx queues stopped. This allows the netdevice to perform queue 1100 * management safely. 1101 * 1102 * Fiber Channel over Ethernet (FCoE) offload functions. 1103 * int (*ndo_fcoe_enable)(struct net_device *dev); 1104 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1105 * so the underlying device can perform whatever needed configuration or 1106 * initialization to support acceleration of FCoE traffic. 1107 * 1108 * int (*ndo_fcoe_disable)(struct net_device *dev); 1109 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1110 * so the underlying device can perform whatever needed clean-ups to 1111 * stop supporting acceleration of FCoE traffic. 1112 * 1113 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1114 * struct scatterlist *sgl, unsigned int sgc); 1115 * Called when the FCoE Initiator wants to initialize an I/O that 1116 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1117 * perform necessary setup and returns 1 to indicate the device is set up 1118 * successfully to perform DDP on this I/O, otherwise this returns 0. 1119 * 1120 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1121 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1122 * indicated by the FC exchange id 'xid', so the underlying device can 1123 * clean up and reuse resources for later DDP requests. 1124 * 1125 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1126 * struct scatterlist *sgl, unsigned int sgc); 1127 * Called when the FCoE Target wants to initialize an I/O that 1128 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1129 * perform necessary setup and returns 1 to indicate the device is set up 1130 * successfully to perform DDP on this I/O, otherwise this returns 0. 1131 * 1132 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1133 * struct netdev_fcoe_hbainfo *hbainfo); 1134 * Called when the FCoE Protocol stack wants information on the underlying 1135 * device. This information is utilized by the FCoE protocol stack to 1136 * register attributes with Fiber Channel management service as per the 1137 * FC-GS Fabric Device Management Information(FDMI) specification. 1138 * 1139 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1140 * Called when the underlying device wants to override default World Wide 1141 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1142 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1143 * protocol stack to use. 1144 * 1145 * RFS acceleration. 1146 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1147 * u16 rxq_index, u32 flow_id); 1148 * Set hardware filter for RFS. rxq_index is the target queue index; 1149 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1150 * Return the filter ID on success, or a negative error code. 1151 * 1152 * Slave management functions (for bridge, bonding, etc). 1153 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1154 * Called to make another netdev an underling. 1155 * 1156 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1157 * Called to release previously enslaved netdev. 1158 * 1159 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1160 * struct sk_buff *skb, 1161 * bool all_slaves); 1162 * Get the xmit slave of master device. If all_slaves is true, function 1163 * assume all the slaves can transmit. 1164 * 1165 * Feature/offload setting functions. 1166 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1167 * netdev_features_t features); 1168 * Adjusts the requested feature flags according to device-specific 1169 * constraints, and returns the resulting flags. Must not modify 1170 * the device state. 1171 * 1172 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1173 * Called to update device configuration to new features. Passed 1174 * feature set might be less than what was returned by ndo_fix_features()). 1175 * Must return >0 or -errno if it changed dev->features itself. 1176 * 1177 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1178 * struct net_device *dev, 1179 * const unsigned char *addr, u16 vid, u16 flags, 1180 * struct netlink_ext_ack *extack); 1181 * Adds an FDB entry to dev for addr. 1182 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1183 * struct net_device *dev, 1184 * const unsigned char *addr, u16 vid) 1185 * Deletes the FDB entry from dev coresponding to addr. 1186 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1187 * struct net_device *dev, struct net_device *filter_dev, 1188 * int *idx) 1189 * Used to add FDB entries to dump requests. Implementers should add 1190 * entries to skb and update idx with the number of entries. 1191 * 1192 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1193 * u16 flags, struct netlink_ext_ack *extack) 1194 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1195 * struct net_device *dev, u32 filter_mask, 1196 * int nlflags) 1197 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1198 * u16 flags); 1199 * 1200 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1201 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1202 * which do not represent real hardware may define this to allow their 1203 * userspace components to manage their virtual carrier state. Devices 1204 * that determine carrier state from physical hardware properties (eg 1205 * network cables) or protocol-dependent mechanisms (eg 1206 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1207 * 1208 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1209 * struct netdev_phys_item_id *ppid); 1210 * Called to get ID of physical port of this device. If driver does 1211 * not implement this, it is assumed that the hw is not able to have 1212 * multiple net devices on single physical port. 1213 * 1214 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1215 * struct netdev_phys_item_id *ppid) 1216 * Called to get the parent ID of the physical port of this device. 1217 * 1218 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1219 * struct udp_tunnel_info *ti); 1220 * Called by UDP tunnel to notify a driver about the UDP port and socket 1221 * address family that a UDP tunnel is listnening to. It is called only 1222 * when a new port starts listening. The operation is protected by the 1223 * RTNL. 1224 * 1225 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1226 * struct udp_tunnel_info *ti); 1227 * Called by UDP tunnel to notify the driver about a UDP port and socket 1228 * address family that the UDP tunnel is not listening to anymore. The 1229 * operation is protected by the RTNL. 1230 * 1231 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1232 * struct net_device *dev) 1233 * Called by upper layer devices to accelerate switching or other 1234 * station functionality into hardware. 'pdev is the lowerdev 1235 * to use for the offload and 'dev' is the net device that will 1236 * back the offload. Returns a pointer to the private structure 1237 * the upper layer will maintain. 1238 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1239 * Called by upper layer device to delete the station created 1240 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1241 * the station and priv is the structure returned by the add 1242 * operation. 1243 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1244 * int queue_index, u32 maxrate); 1245 * Called when a user wants to set a max-rate limitation of specific 1246 * TX queue. 1247 * int (*ndo_get_iflink)(const struct net_device *dev); 1248 * Called to get the iflink value of this device. 1249 * void (*ndo_change_proto_down)(struct net_device *dev, 1250 * bool proto_down); 1251 * This function is used to pass protocol port error state information 1252 * to the switch driver. The switch driver can react to the proto_down 1253 * by doing a phys down on the associated switch port. 1254 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1255 * This function is used to get egress tunnel information for given skb. 1256 * This is useful for retrieving outer tunnel header parameters while 1257 * sampling packet. 1258 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1259 * This function is used to specify the headroom that the skb must 1260 * consider when allocation skb during packet reception. Setting 1261 * appropriate rx headroom value allows avoiding skb head copy on 1262 * forward. Setting a negative value resets the rx headroom to the 1263 * default value. 1264 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1265 * This function is used to set or query state related to XDP on the 1266 * netdevice and manage BPF offload. See definition of 1267 * enum bpf_netdev_command for details. 1268 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1269 * u32 flags); 1270 * This function is used to submit @n XDP packets for transmit on a 1271 * netdevice. Returns number of frames successfully transmitted, frames 1272 * that got dropped are freed/returned via xdp_return_frame(). 1273 * Returns negative number, means general error invoking ndo, meaning 1274 * no frames were xmit'ed and core-caller will free all frames. 1275 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1276 * This function is used to wake up the softirq, ksoftirqd or kthread 1277 * responsible for sending and/or receiving packets on a specific 1278 * queue id bound to an AF_XDP socket. The flags field specifies if 1279 * only RX, only Tx, or both should be woken up using the flags 1280 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1281 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1282 * Get devlink port instance associated with a given netdev. 1283 * Called with a reference on the netdevice and devlink locks only, 1284 * rtnl_lock is not held. 1285 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1286 * int cmd); 1287 * Add, change, delete or get information on an IPv4 tunnel. 1288 */ 1289 struct net_device_ops { 1290 int (*ndo_init)(struct net_device *dev); 1291 void (*ndo_uninit)(struct net_device *dev); 1292 int (*ndo_open)(struct net_device *dev); 1293 int (*ndo_stop)(struct net_device *dev); 1294 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1295 struct net_device *dev); 1296 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1297 struct net_device *dev, 1298 netdev_features_t features); 1299 u16 (*ndo_select_queue)(struct net_device *dev, 1300 struct sk_buff *skb, 1301 struct net_device *sb_dev); 1302 void (*ndo_change_rx_flags)(struct net_device *dev, 1303 int flags); 1304 void (*ndo_set_rx_mode)(struct net_device *dev); 1305 int (*ndo_set_mac_address)(struct net_device *dev, 1306 void *addr); 1307 int (*ndo_validate_addr)(struct net_device *dev); 1308 int (*ndo_do_ioctl)(struct net_device *dev, 1309 struct ifreq *ifr, int cmd); 1310 int (*ndo_set_config)(struct net_device *dev, 1311 struct ifmap *map); 1312 int (*ndo_change_mtu)(struct net_device *dev, 1313 int new_mtu); 1314 int (*ndo_neigh_setup)(struct net_device *dev, 1315 struct neigh_parms *); 1316 void (*ndo_tx_timeout) (struct net_device *dev, 1317 unsigned int txqueue); 1318 1319 void (*ndo_get_stats64)(struct net_device *dev, 1320 struct rtnl_link_stats64 *storage); 1321 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1322 int (*ndo_get_offload_stats)(int attr_id, 1323 const struct net_device *dev, 1324 void *attr_data); 1325 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1326 1327 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1328 __be16 proto, u16 vid); 1329 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1330 __be16 proto, u16 vid); 1331 #ifdef CONFIG_NET_POLL_CONTROLLER 1332 void (*ndo_poll_controller)(struct net_device *dev); 1333 int (*ndo_netpoll_setup)(struct net_device *dev, 1334 struct netpoll_info *info); 1335 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1336 #endif 1337 int (*ndo_set_vf_mac)(struct net_device *dev, 1338 int queue, u8 *mac); 1339 int (*ndo_set_vf_vlan)(struct net_device *dev, 1340 int queue, u16 vlan, 1341 u8 qos, __be16 proto); 1342 int (*ndo_set_vf_rate)(struct net_device *dev, 1343 int vf, int min_tx_rate, 1344 int max_tx_rate); 1345 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1346 int vf, bool setting); 1347 int (*ndo_set_vf_trust)(struct net_device *dev, 1348 int vf, bool setting); 1349 int (*ndo_get_vf_config)(struct net_device *dev, 1350 int vf, 1351 struct ifla_vf_info *ivf); 1352 int (*ndo_set_vf_link_state)(struct net_device *dev, 1353 int vf, int link_state); 1354 int (*ndo_get_vf_stats)(struct net_device *dev, 1355 int vf, 1356 struct ifla_vf_stats 1357 *vf_stats); 1358 int (*ndo_set_vf_port)(struct net_device *dev, 1359 int vf, 1360 struct nlattr *port[]); 1361 int (*ndo_get_vf_port)(struct net_device *dev, 1362 int vf, struct sk_buff *skb); 1363 int (*ndo_get_vf_guid)(struct net_device *dev, 1364 int vf, 1365 struct ifla_vf_guid *node_guid, 1366 struct ifla_vf_guid *port_guid); 1367 int (*ndo_set_vf_guid)(struct net_device *dev, 1368 int vf, u64 guid, 1369 int guid_type); 1370 int (*ndo_set_vf_rss_query_en)( 1371 struct net_device *dev, 1372 int vf, bool setting); 1373 int (*ndo_setup_tc)(struct net_device *dev, 1374 enum tc_setup_type type, 1375 void *type_data); 1376 #if IS_ENABLED(CONFIG_FCOE) 1377 int (*ndo_fcoe_enable)(struct net_device *dev); 1378 int (*ndo_fcoe_disable)(struct net_device *dev); 1379 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1380 u16 xid, 1381 struct scatterlist *sgl, 1382 unsigned int sgc); 1383 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1384 u16 xid); 1385 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1386 u16 xid, 1387 struct scatterlist *sgl, 1388 unsigned int sgc); 1389 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1390 struct netdev_fcoe_hbainfo *hbainfo); 1391 #endif 1392 1393 #if IS_ENABLED(CONFIG_LIBFCOE) 1394 #define NETDEV_FCOE_WWNN 0 1395 #define NETDEV_FCOE_WWPN 1 1396 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1397 u64 *wwn, int type); 1398 #endif 1399 1400 #ifdef CONFIG_RFS_ACCEL 1401 int (*ndo_rx_flow_steer)(struct net_device *dev, 1402 const struct sk_buff *skb, 1403 u16 rxq_index, 1404 u32 flow_id); 1405 #endif 1406 int (*ndo_add_slave)(struct net_device *dev, 1407 struct net_device *slave_dev, 1408 struct netlink_ext_ack *extack); 1409 int (*ndo_del_slave)(struct net_device *dev, 1410 struct net_device *slave_dev); 1411 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1412 struct sk_buff *skb, 1413 bool all_slaves); 1414 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1415 netdev_features_t features); 1416 int (*ndo_set_features)(struct net_device *dev, 1417 netdev_features_t features); 1418 int (*ndo_neigh_construct)(struct net_device *dev, 1419 struct neighbour *n); 1420 void (*ndo_neigh_destroy)(struct net_device *dev, 1421 struct neighbour *n); 1422 1423 int (*ndo_fdb_add)(struct ndmsg *ndm, 1424 struct nlattr *tb[], 1425 struct net_device *dev, 1426 const unsigned char *addr, 1427 u16 vid, 1428 u16 flags, 1429 struct netlink_ext_ack *extack); 1430 int (*ndo_fdb_del)(struct ndmsg *ndm, 1431 struct nlattr *tb[], 1432 struct net_device *dev, 1433 const unsigned char *addr, 1434 u16 vid); 1435 int (*ndo_fdb_dump)(struct sk_buff *skb, 1436 struct netlink_callback *cb, 1437 struct net_device *dev, 1438 struct net_device *filter_dev, 1439 int *idx); 1440 int (*ndo_fdb_get)(struct sk_buff *skb, 1441 struct nlattr *tb[], 1442 struct net_device *dev, 1443 const unsigned char *addr, 1444 u16 vid, u32 portid, u32 seq, 1445 struct netlink_ext_ack *extack); 1446 int (*ndo_bridge_setlink)(struct net_device *dev, 1447 struct nlmsghdr *nlh, 1448 u16 flags, 1449 struct netlink_ext_ack *extack); 1450 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1451 u32 pid, u32 seq, 1452 struct net_device *dev, 1453 u32 filter_mask, 1454 int nlflags); 1455 int (*ndo_bridge_dellink)(struct net_device *dev, 1456 struct nlmsghdr *nlh, 1457 u16 flags); 1458 int (*ndo_change_carrier)(struct net_device *dev, 1459 bool new_carrier); 1460 int (*ndo_get_phys_port_id)(struct net_device *dev, 1461 struct netdev_phys_item_id *ppid); 1462 int (*ndo_get_port_parent_id)(struct net_device *dev, 1463 struct netdev_phys_item_id *ppid); 1464 int (*ndo_get_phys_port_name)(struct net_device *dev, 1465 char *name, size_t len); 1466 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1467 struct udp_tunnel_info *ti); 1468 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1469 struct udp_tunnel_info *ti); 1470 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1471 struct net_device *dev); 1472 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1473 void *priv); 1474 1475 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1476 int queue_index, 1477 u32 maxrate); 1478 int (*ndo_get_iflink)(const struct net_device *dev); 1479 int (*ndo_change_proto_down)(struct net_device *dev, 1480 bool proto_down); 1481 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1482 struct sk_buff *skb); 1483 void (*ndo_set_rx_headroom)(struct net_device *dev, 1484 int needed_headroom); 1485 int (*ndo_bpf)(struct net_device *dev, 1486 struct netdev_bpf *bpf); 1487 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1488 struct xdp_frame **xdp, 1489 u32 flags); 1490 int (*ndo_xsk_wakeup)(struct net_device *dev, 1491 u32 queue_id, u32 flags); 1492 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1493 int (*ndo_tunnel_ctl)(struct net_device *dev, 1494 struct ip_tunnel_parm *p, int cmd); 1495 }; 1496 1497 /** 1498 * enum net_device_priv_flags - &struct net_device priv_flags 1499 * 1500 * These are the &struct net_device, they are only set internally 1501 * by drivers and used in the kernel. These flags are invisible to 1502 * userspace; this means that the order of these flags can change 1503 * during any kernel release. 1504 * 1505 * You should have a pretty good reason to be extending these flags. 1506 * 1507 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1508 * @IFF_EBRIDGE: Ethernet bridging device 1509 * @IFF_BONDING: bonding master or slave 1510 * @IFF_ISATAP: ISATAP interface (RFC4214) 1511 * @IFF_WAN_HDLC: WAN HDLC device 1512 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1513 * release skb->dst 1514 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1515 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1516 * @IFF_MACVLAN_PORT: device used as macvlan port 1517 * @IFF_BRIDGE_PORT: device used as bridge port 1518 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1519 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1520 * @IFF_UNICAST_FLT: Supports unicast filtering 1521 * @IFF_TEAM_PORT: device used as team port 1522 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1523 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1524 * change when it's running 1525 * @IFF_MACVLAN: Macvlan device 1526 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1527 * underlying stacked devices 1528 * @IFF_L3MDEV_MASTER: device is an L3 master device 1529 * @IFF_NO_QUEUE: device can run without qdisc attached 1530 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1531 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1532 * @IFF_TEAM: device is a team device 1533 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1534 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1535 * entity (i.e. the master device for bridged veth) 1536 * @IFF_MACSEC: device is a MACsec device 1537 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1538 * @IFF_FAILOVER: device is a failover master device 1539 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1540 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1541 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1542 */ 1543 enum netdev_priv_flags { 1544 IFF_802_1Q_VLAN = 1<<0, 1545 IFF_EBRIDGE = 1<<1, 1546 IFF_BONDING = 1<<2, 1547 IFF_ISATAP = 1<<3, 1548 IFF_WAN_HDLC = 1<<4, 1549 IFF_XMIT_DST_RELEASE = 1<<5, 1550 IFF_DONT_BRIDGE = 1<<6, 1551 IFF_DISABLE_NETPOLL = 1<<7, 1552 IFF_MACVLAN_PORT = 1<<8, 1553 IFF_BRIDGE_PORT = 1<<9, 1554 IFF_OVS_DATAPATH = 1<<10, 1555 IFF_TX_SKB_SHARING = 1<<11, 1556 IFF_UNICAST_FLT = 1<<12, 1557 IFF_TEAM_PORT = 1<<13, 1558 IFF_SUPP_NOFCS = 1<<14, 1559 IFF_LIVE_ADDR_CHANGE = 1<<15, 1560 IFF_MACVLAN = 1<<16, 1561 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1562 IFF_L3MDEV_MASTER = 1<<18, 1563 IFF_NO_QUEUE = 1<<19, 1564 IFF_OPENVSWITCH = 1<<20, 1565 IFF_L3MDEV_SLAVE = 1<<21, 1566 IFF_TEAM = 1<<22, 1567 IFF_RXFH_CONFIGURED = 1<<23, 1568 IFF_PHONY_HEADROOM = 1<<24, 1569 IFF_MACSEC = 1<<25, 1570 IFF_NO_RX_HANDLER = 1<<26, 1571 IFF_FAILOVER = 1<<27, 1572 IFF_FAILOVER_SLAVE = 1<<28, 1573 IFF_L3MDEV_RX_HANDLER = 1<<29, 1574 IFF_LIVE_RENAME_OK = 1<<30, 1575 }; 1576 1577 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1578 #define IFF_EBRIDGE IFF_EBRIDGE 1579 #define IFF_BONDING IFF_BONDING 1580 #define IFF_ISATAP IFF_ISATAP 1581 #define IFF_WAN_HDLC IFF_WAN_HDLC 1582 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1583 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1584 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1585 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1586 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1587 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1588 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1589 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1590 #define IFF_TEAM_PORT IFF_TEAM_PORT 1591 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1592 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1593 #define IFF_MACVLAN IFF_MACVLAN 1594 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1595 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1596 #define IFF_NO_QUEUE IFF_NO_QUEUE 1597 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1598 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1599 #define IFF_TEAM IFF_TEAM 1600 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1601 #define IFF_MACSEC IFF_MACSEC 1602 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1603 #define IFF_FAILOVER IFF_FAILOVER 1604 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1605 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1606 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1607 1608 /** 1609 * struct net_device - The DEVICE structure. 1610 * 1611 * Actually, this whole structure is a big mistake. It mixes I/O 1612 * data with strictly "high-level" data, and it has to know about 1613 * almost every data structure used in the INET module. 1614 * 1615 * @name: This is the first field of the "visible" part of this structure 1616 * (i.e. as seen by users in the "Space.c" file). It is the name 1617 * of the interface. 1618 * 1619 * @name_node: Name hashlist node 1620 * @ifalias: SNMP alias 1621 * @mem_end: Shared memory end 1622 * @mem_start: Shared memory start 1623 * @base_addr: Device I/O address 1624 * @irq: Device IRQ number 1625 * 1626 * @state: Generic network queuing layer state, see netdev_state_t 1627 * @dev_list: The global list of network devices 1628 * @napi_list: List entry used for polling NAPI devices 1629 * @unreg_list: List entry when we are unregistering the 1630 * device; see the function unregister_netdev 1631 * @close_list: List entry used when we are closing the device 1632 * @ptype_all: Device-specific packet handlers for all protocols 1633 * @ptype_specific: Device-specific, protocol-specific packet handlers 1634 * 1635 * @adj_list: Directly linked devices, like slaves for bonding 1636 * @features: Currently active device features 1637 * @hw_features: User-changeable features 1638 * 1639 * @wanted_features: User-requested features 1640 * @vlan_features: Mask of features inheritable by VLAN devices 1641 * 1642 * @hw_enc_features: Mask of features inherited by encapsulating devices 1643 * This field indicates what encapsulation 1644 * offloads the hardware is capable of doing, 1645 * and drivers will need to set them appropriately. 1646 * 1647 * @mpls_features: Mask of features inheritable by MPLS 1648 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1649 * 1650 * @ifindex: interface index 1651 * @group: The group the device belongs to 1652 * 1653 * @stats: Statistics struct, which was left as a legacy, use 1654 * rtnl_link_stats64 instead 1655 * 1656 * @rx_dropped: Dropped packets by core network, 1657 * do not use this in drivers 1658 * @tx_dropped: Dropped packets by core network, 1659 * do not use this in drivers 1660 * @rx_nohandler: nohandler dropped packets by core network on 1661 * inactive devices, do not use this in drivers 1662 * @carrier_up_count: Number of times the carrier has been up 1663 * @carrier_down_count: Number of times the carrier has been down 1664 * 1665 * @wireless_handlers: List of functions to handle Wireless Extensions, 1666 * instead of ioctl, 1667 * see <net/iw_handler.h> for details. 1668 * @wireless_data: Instance data managed by the core of wireless extensions 1669 * 1670 * @netdev_ops: Includes several pointers to callbacks, 1671 * if one wants to override the ndo_*() functions 1672 * @ethtool_ops: Management operations 1673 * @l3mdev_ops: Layer 3 master device operations 1674 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1675 * discovery handling. Necessary for e.g. 6LoWPAN. 1676 * @xfrmdev_ops: Transformation offload operations 1677 * @tlsdev_ops: Transport Layer Security offload operations 1678 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1679 * of Layer 2 headers. 1680 * 1681 * @flags: Interface flags (a la BSD) 1682 * @priv_flags: Like 'flags' but invisible to userspace, 1683 * see if.h for the definitions 1684 * @gflags: Global flags ( kept as legacy ) 1685 * @padded: How much padding added by alloc_netdev() 1686 * @operstate: RFC2863 operstate 1687 * @link_mode: Mapping policy to operstate 1688 * @if_port: Selectable AUI, TP, ... 1689 * @dma: DMA channel 1690 * @mtu: Interface MTU value 1691 * @min_mtu: Interface Minimum MTU value 1692 * @max_mtu: Interface Maximum MTU value 1693 * @type: Interface hardware type 1694 * @hard_header_len: Maximum hardware header length. 1695 * @min_header_len: Minimum hardware header length 1696 * 1697 * @needed_headroom: Extra headroom the hardware may need, but not in all 1698 * cases can this be guaranteed 1699 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1700 * cases can this be guaranteed. Some cases also use 1701 * LL_MAX_HEADER instead to allocate the skb 1702 * 1703 * interface address info: 1704 * 1705 * @perm_addr: Permanent hw address 1706 * @addr_assign_type: Hw address assignment type 1707 * @addr_len: Hardware address length 1708 * @upper_level: Maximum depth level of upper devices. 1709 * @lower_level: Maximum depth level of lower devices. 1710 * @neigh_priv_len: Used in neigh_alloc() 1711 * @dev_id: Used to differentiate devices that share 1712 * the same link layer address 1713 * @dev_port: Used to differentiate devices that share 1714 * the same function 1715 * @addr_list_lock: XXX: need comments on this one 1716 * @name_assign_type: network interface name assignment type 1717 * @uc_promisc: Counter that indicates promiscuous mode 1718 * has been enabled due to the need to listen to 1719 * additional unicast addresses in a device that 1720 * does not implement ndo_set_rx_mode() 1721 * @uc: unicast mac addresses 1722 * @mc: multicast mac addresses 1723 * @dev_addrs: list of device hw addresses 1724 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1725 * @promiscuity: Number of times the NIC is told to work in 1726 * promiscuous mode; if it becomes 0 the NIC will 1727 * exit promiscuous mode 1728 * @allmulti: Counter, enables or disables allmulticast mode 1729 * 1730 * @vlan_info: VLAN info 1731 * @dsa_ptr: dsa specific data 1732 * @tipc_ptr: TIPC specific data 1733 * @atalk_ptr: AppleTalk link 1734 * @ip_ptr: IPv4 specific data 1735 * @dn_ptr: DECnet specific data 1736 * @ip6_ptr: IPv6 specific data 1737 * @ax25_ptr: AX.25 specific data 1738 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1739 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1740 * device struct 1741 * @mpls_ptr: mpls_dev struct pointer 1742 * 1743 * @dev_addr: Hw address (before bcast, 1744 * because most packets are unicast) 1745 * 1746 * @_rx: Array of RX queues 1747 * @num_rx_queues: Number of RX queues 1748 * allocated at register_netdev() time 1749 * @real_num_rx_queues: Number of RX queues currently active in device 1750 * @xdp_prog: XDP sockets filter program pointer 1751 * @gro_flush_timeout: timeout for GRO layer in NAPI 1752 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1753 * allow to avoid NIC hard IRQ, on busy queues. 1754 * 1755 * @rx_handler: handler for received packets 1756 * @rx_handler_data: XXX: need comments on this one 1757 * @miniq_ingress: ingress/clsact qdisc specific data for 1758 * ingress processing 1759 * @ingress_queue: XXX: need comments on this one 1760 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1761 * @broadcast: hw bcast address 1762 * 1763 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1764 * indexed by RX queue number. Assigned by driver. 1765 * This must only be set if the ndo_rx_flow_steer 1766 * operation is defined 1767 * @index_hlist: Device index hash chain 1768 * 1769 * @_tx: Array of TX queues 1770 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1771 * @real_num_tx_queues: Number of TX queues currently active in device 1772 * @qdisc: Root qdisc from userspace point of view 1773 * @tx_queue_len: Max frames per queue allowed 1774 * @tx_global_lock: XXX: need comments on this one 1775 * @xdp_bulkq: XDP device bulk queue 1776 * @xps_cpus_map: all CPUs map for XPS device 1777 * @xps_rxqs_map: all RXQs map for XPS device 1778 * 1779 * @xps_maps: XXX: need comments on this one 1780 * @miniq_egress: clsact qdisc specific data for 1781 * egress processing 1782 * @qdisc_hash: qdisc hash table 1783 * @watchdog_timeo: Represents the timeout that is used by 1784 * the watchdog (see dev_watchdog()) 1785 * @watchdog_timer: List of timers 1786 * 1787 * @pcpu_refcnt: Number of references to this device 1788 * @todo_list: Delayed register/unregister 1789 * @link_watch_list: XXX: need comments on this one 1790 * 1791 * @reg_state: Register/unregister state machine 1792 * @dismantle: Device is going to be freed 1793 * @rtnl_link_state: This enum represents the phases of creating 1794 * a new link 1795 * 1796 * @needs_free_netdev: Should unregister perform free_netdev? 1797 * @priv_destructor: Called from unregister 1798 * @npinfo: XXX: need comments on this one 1799 * @nd_net: Network namespace this network device is inside 1800 * 1801 * @ml_priv: Mid-layer private 1802 * @lstats: Loopback statistics 1803 * @tstats: Tunnel statistics 1804 * @dstats: Dummy statistics 1805 * @vstats: Virtual ethernet statistics 1806 * 1807 * @garp_port: GARP 1808 * @mrp_port: MRP 1809 * 1810 * @dev: Class/net/name entry 1811 * @sysfs_groups: Space for optional device, statistics and wireless 1812 * sysfs groups 1813 * 1814 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1815 * @rtnl_link_ops: Rtnl_link_ops 1816 * 1817 * @gso_max_size: Maximum size of generic segmentation offload 1818 * @gso_max_segs: Maximum number of segments that can be passed to the 1819 * NIC for GSO 1820 * 1821 * @dcbnl_ops: Data Center Bridging netlink ops 1822 * @num_tc: Number of traffic classes in the net device 1823 * @tc_to_txq: XXX: need comments on this one 1824 * @prio_tc_map: XXX: need comments on this one 1825 * 1826 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1827 * 1828 * @priomap: XXX: need comments on this one 1829 * @phydev: Physical device may attach itself 1830 * for hardware timestamping 1831 * @sfp_bus: attached &struct sfp_bus structure. 1832 * 1833 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1834 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1835 * 1836 * @proto_down: protocol port state information can be sent to the 1837 * switch driver and used to set the phys state of the 1838 * switch port. 1839 * 1840 * @wol_enabled: Wake-on-LAN is enabled 1841 * 1842 * @net_notifier_list: List of per-net netdev notifier block 1843 * that follow this device when it is moved 1844 * to another network namespace. 1845 * 1846 * @macsec_ops: MACsec offloading ops 1847 * 1848 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1849 * offload capabilities of the device 1850 * @udp_tunnel_nic: UDP tunnel offload state 1851 * 1852 * FIXME: cleanup struct net_device such that network protocol info 1853 * moves out. 1854 */ 1855 1856 struct net_device { 1857 char name[IFNAMSIZ]; 1858 struct netdev_name_node *name_node; 1859 struct dev_ifalias __rcu *ifalias; 1860 /* 1861 * I/O specific fields 1862 * FIXME: Merge these and struct ifmap into one 1863 */ 1864 unsigned long mem_end; 1865 unsigned long mem_start; 1866 unsigned long base_addr; 1867 int irq; 1868 1869 /* 1870 * Some hardware also needs these fields (state,dev_list, 1871 * napi_list,unreg_list,close_list) but they are not 1872 * part of the usual set specified in Space.c. 1873 */ 1874 1875 unsigned long state; 1876 1877 struct list_head dev_list; 1878 struct list_head napi_list; 1879 struct list_head unreg_list; 1880 struct list_head close_list; 1881 struct list_head ptype_all; 1882 struct list_head ptype_specific; 1883 1884 struct { 1885 struct list_head upper; 1886 struct list_head lower; 1887 } adj_list; 1888 1889 netdev_features_t features; 1890 netdev_features_t hw_features; 1891 netdev_features_t wanted_features; 1892 netdev_features_t vlan_features; 1893 netdev_features_t hw_enc_features; 1894 netdev_features_t mpls_features; 1895 netdev_features_t gso_partial_features; 1896 1897 int ifindex; 1898 int group; 1899 1900 struct net_device_stats stats; 1901 1902 atomic_long_t rx_dropped; 1903 atomic_long_t tx_dropped; 1904 atomic_long_t rx_nohandler; 1905 1906 /* Stats to monitor link on/off, flapping */ 1907 atomic_t carrier_up_count; 1908 atomic_t carrier_down_count; 1909 1910 #ifdef CONFIG_WIRELESS_EXT 1911 const struct iw_handler_def *wireless_handlers; 1912 struct iw_public_data *wireless_data; 1913 #endif 1914 const struct net_device_ops *netdev_ops; 1915 const struct ethtool_ops *ethtool_ops; 1916 #ifdef CONFIG_NET_L3_MASTER_DEV 1917 const struct l3mdev_ops *l3mdev_ops; 1918 #endif 1919 #if IS_ENABLED(CONFIG_IPV6) 1920 const struct ndisc_ops *ndisc_ops; 1921 #endif 1922 1923 #ifdef CONFIG_XFRM_OFFLOAD 1924 const struct xfrmdev_ops *xfrmdev_ops; 1925 #endif 1926 1927 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1928 const struct tlsdev_ops *tlsdev_ops; 1929 #endif 1930 1931 const struct header_ops *header_ops; 1932 1933 unsigned int flags; 1934 unsigned int priv_flags; 1935 1936 unsigned short gflags; 1937 unsigned short padded; 1938 1939 unsigned char operstate; 1940 unsigned char link_mode; 1941 1942 unsigned char if_port; 1943 unsigned char dma; 1944 1945 /* Note : dev->mtu is often read without holding a lock. 1946 * Writers usually hold RTNL. 1947 * It is recommended to use READ_ONCE() to annotate the reads, 1948 * and to use WRITE_ONCE() to annotate the writes. 1949 */ 1950 unsigned int mtu; 1951 unsigned int min_mtu; 1952 unsigned int max_mtu; 1953 unsigned short type; 1954 unsigned short hard_header_len; 1955 unsigned char min_header_len; 1956 1957 unsigned short needed_headroom; 1958 unsigned short needed_tailroom; 1959 1960 /* Interface address info. */ 1961 unsigned char perm_addr[MAX_ADDR_LEN]; 1962 unsigned char addr_assign_type; 1963 unsigned char addr_len; 1964 unsigned char upper_level; 1965 unsigned char lower_level; 1966 unsigned short neigh_priv_len; 1967 unsigned short dev_id; 1968 unsigned short dev_port; 1969 spinlock_t addr_list_lock; 1970 unsigned char name_assign_type; 1971 bool uc_promisc; 1972 struct netdev_hw_addr_list uc; 1973 struct netdev_hw_addr_list mc; 1974 struct netdev_hw_addr_list dev_addrs; 1975 1976 #ifdef CONFIG_SYSFS 1977 struct kset *queues_kset; 1978 #endif 1979 unsigned int promiscuity; 1980 unsigned int allmulti; 1981 1982 1983 /* Protocol-specific pointers */ 1984 1985 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1986 struct vlan_info __rcu *vlan_info; 1987 #endif 1988 #if IS_ENABLED(CONFIG_NET_DSA) 1989 struct dsa_port *dsa_ptr; 1990 #endif 1991 #if IS_ENABLED(CONFIG_TIPC) 1992 struct tipc_bearer __rcu *tipc_ptr; 1993 #endif 1994 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1995 void *atalk_ptr; 1996 #endif 1997 struct in_device __rcu *ip_ptr; 1998 #if IS_ENABLED(CONFIG_DECNET) 1999 struct dn_dev __rcu *dn_ptr; 2000 #endif 2001 struct inet6_dev __rcu *ip6_ptr; 2002 #if IS_ENABLED(CONFIG_AX25) 2003 void *ax25_ptr; 2004 #endif 2005 struct wireless_dev *ieee80211_ptr; 2006 struct wpan_dev *ieee802154_ptr; 2007 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2008 struct mpls_dev __rcu *mpls_ptr; 2009 #endif 2010 2011 /* 2012 * Cache lines mostly used on receive path (including eth_type_trans()) 2013 */ 2014 /* Interface address info used in eth_type_trans() */ 2015 unsigned char *dev_addr; 2016 2017 struct netdev_rx_queue *_rx; 2018 unsigned int num_rx_queues; 2019 unsigned int real_num_rx_queues; 2020 2021 struct bpf_prog __rcu *xdp_prog; 2022 unsigned long gro_flush_timeout; 2023 int napi_defer_hard_irqs; 2024 rx_handler_func_t __rcu *rx_handler; 2025 void __rcu *rx_handler_data; 2026 2027 #ifdef CONFIG_NET_CLS_ACT 2028 struct mini_Qdisc __rcu *miniq_ingress; 2029 #endif 2030 struct netdev_queue __rcu *ingress_queue; 2031 #ifdef CONFIG_NETFILTER_INGRESS 2032 struct nf_hook_entries __rcu *nf_hooks_ingress; 2033 #endif 2034 2035 unsigned char broadcast[MAX_ADDR_LEN]; 2036 #ifdef CONFIG_RFS_ACCEL 2037 struct cpu_rmap *rx_cpu_rmap; 2038 #endif 2039 struct hlist_node index_hlist; 2040 2041 /* 2042 * Cache lines mostly used on transmit path 2043 */ 2044 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2045 unsigned int num_tx_queues; 2046 unsigned int real_num_tx_queues; 2047 struct Qdisc *qdisc; 2048 unsigned int tx_queue_len; 2049 spinlock_t tx_global_lock; 2050 2051 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2052 2053 #ifdef CONFIG_XPS 2054 struct xps_dev_maps __rcu *xps_cpus_map; 2055 struct xps_dev_maps __rcu *xps_rxqs_map; 2056 #endif 2057 #ifdef CONFIG_NET_CLS_ACT 2058 struct mini_Qdisc __rcu *miniq_egress; 2059 #endif 2060 2061 #ifdef CONFIG_NET_SCHED 2062 DECLARE_HASHTABLE (qdisc_hash, 4); 2063 #endif 2064 /* These may be needed for future network-power-down code. */ 2065 struct timer_list watchdog_timer; 2066 int watchdog_timeo; 2067 2068 u32 proto_down_reason; 2069 2070 struct list_head todo_list; 2071 int __percpu *pcpu_refcnt; 2072 2073 struct list_head link_watch_list; 2074 2075 enum { NETREG_UNINITIALIZED=0, 2076 NETREG_REGISTERED, /* completed register_netdevice */ 2077 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2078 NETREG_UNREGISTERED, /* completed unregister todo */ 2079 NETREG_RELEASED, /* called free_netdev */ 2080 NETREG_DUMMY, /* dummy device for NAPI poll */ 2081 } reg_state:8; 2082 2083 bool dismantle; 2084 2085 enum { 2086 RTNL_LINK_INITIALIZED, 2087 RTNL_LINK_INITIALIZING, 2088 } rtnl_link_state:16; 2089 2090 bool needs_free_netdev; 2091 void (*priv_destructor)(struct net_device *dev); 2092 2093 #ifdef CONFIG_NETPOLL 2094 struct netpoll_info __rcu *npinfo; 2095 #endif 2096 2097 possible_net_t nd_net; 2098 2099 /* mid-layer private */ 2100 union { 2101 void *ml_priv; 2102 struct pcpu_lstats __percpu *lstats; 2103 struct pcpu_sw_netstats __percpu *tstats; 2104 struct pcpu_dstats __percpu *dstats; 2105 }; 2106 2107 #if IS_ENABLED(CONFIG_GARP) 2108 struct garp_port __rcu *garp_port; 2109 #endif 2110 #if IS_ENABLED(CONFIG_MRP) 2111 struct mrp_port __rcu *mrp_port; 2112 #endif 2113 2114 struct device dev; 2115 const struct attribute_group *sysfs_groups[4]; 2116 const struct attribute_group *sysfs_rx_queue_group; 2117 2118 const struct rtnl_link_ops *rtnl_link_ops; 2119 2120 /* for setting kernel sock attribute on TCP connection setup */ 2121 #define GSO_MAX_SIZE 65536 2122 unsigned int gso_max_size; 2123 #define GSO_MAX_SEGS 65535 2124 u16 gso_max_segs; 2125 2126 #ifdef CONFIG_DCB 2127 const struct dcbnl_rtnl_ops *dcbnl_ops; 2128 #endif 2129 s16 num_tc; 2130 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2131 u8 prio_tc_map[TC_BITMASK + 1]; 2132 2133 #if IS_ENABLED(CONFIG_FCOE) 2134 unsigned int fcoe_ddp_xid; 2135 #endif 2136 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2137 struct netprio_map __rcu *priomap; 2138 #endif 2139 struct phy_device *phydev; 2140 struct sfp_bus *sfp_bus; 2141 struct lock_class_key *qdisc_tx_busylock; 2142 struct lock_class_key *qdisc_running_key; 2143 bool proto_down; 2144 unsigned wol_enabled:1; 2145 2146 struct list_head net_notifier_list; 2147 2148 #if IS_ENABLED(CONFIG_MACSEC) 2149 /* MACsec management functions */ 2150 const struct macsec_ops *macsec_ops; 2151 #endif 2152 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2153 struct udp_tunnel_nic *udp_tunnel_nic; 2154 2155 /* protected by rtnl_lock */ 2156 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2157 }; 2158 #define to_net_dev(d) container_of(d, struct net_device, dev) 2159 2160 static inline bool netif_elide_gro(const struct net_device *dev) 2161 { 2162 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2163 return true; 2164 return false; 2165 } 2166 2167 #define NETDEV_ALIGN 32 2168 2169 static inline 2170 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2171 { 2172 return dev->prio_tc_map[prio & TC_BITMASK]; 2173 } 2174 2175 static inline 2176 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2177 { 2178 if (tc >= dev->num_tc) 2179 return -EINVAL; 2180 2181 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2182 return 0; 2183 } 2184 2185 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2186 void netdev_reset_tc(struct net_device *dev); 2187 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2188 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2189 2190 static inline 2191 int netdev_get_num_tc(struct net_device *dev) 2192 { 2193 return dev->num_tc; 2194 } 2195 2196 void netdev_unbind_sb_channel(struct net_device *dev, 2197 struct net_device *sb_dev); 2198 int netdev_bind_sb_channel_queue(struct net_device *dev, 2199 struct net_device *sb_dev, 2200 u8 tc, u16 count, u16 offset); 2201 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2202 static inline int netdev_get_sb_channel(struct net_device *dev) 2203 { 2204 return max_t(int, -dev->num_tc, 0); 2205 } 2206 2207 static inline 2208 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2209 unsigned int index) 2210 { 2211 return &dev->_tx[index]; 2212 } 2213 2214 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2215 const struct sk_buff *skb) 2216 { 2217 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2218 } 2219 2220 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2221 void (*f)(struct net_device *, 2222 struct netdev_queue *, 2223 void *), 2224 void *arg) 2225 { 2226 unsigned int i; 2227 2228 for (i = 0; i < dev->num_tx_queues; i++) 2229 f(dev, &dev->_tx[i], arg); 2230 } 2231 2232 #define netdev_lockdep_set_classes(dev) \ 2233 { \ 2234 static struct lock_class_key qdisc_tx_busylock_key; \ 2235 static struct lock_class_key qdisc_running_key; \ 2236 static struct lock_class_key qdisc_xmit_lock_key; \ 2237 static struct lock_class_key dev_addr_list_lock_key; \ 2238 unsigned int i; \ 2239 \ 2240 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2241 (dev)->qdisc_running_key = &qdisc_running_key; \ 2242 lockdep_set_class(&(dev)->addr_list_lock, \ 2243 &dev_addr_list_lock_key); \ 2244 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2245 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2246 &qdisc_xmit_lock_key); \ 2247 } 2248 2249 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2250 struct net_device *sb_dev); 2251 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2252 struct sk_buff *skb, 2253 struct net_device *sb_dev); 2254 2255 /* returns the headroom that the master device needs to take in account 2256 * when forwarding to this dev 2257 */ 2258 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2259 { 2260 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2261 } 2262 2263 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2264 { 2265 if (dev->netdev_ops->ndo_set_rx_headroom) 2266 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2267 } 2268 2269 /* set the device rx headroom to the dev's default */ 2270 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2271 { 2272 netdev_set_rx_headroom(dev, -1); 2273 } 2274 2275 /* 2276 * Net namespace inlines 2277 */ 2278 static inline 2279 struct net *dev_net(const struct net_device *dev) 2280 { 2281 return read_pnet(&dev->nd_net); 2282 } 2283 2284 static inline 2285 void dev_net_set(struct net_device *dev, struct net *net) 2286 { 2287 write_pnet(&dev->nd_net, net); 2288 } 2289 2290 /** 2291 * netdev_priv - access network device private data 2292 * @dev: network device 2293 * 2294 * Get network device private data 2295 */ 2296 static inline void *netdev_priv(const struct net_device *dev) 2297 { 2298 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2299 } 2300 2301 /* Set the sysfs physical device reference for the network logical device 2302 * if set prior to registration will cause a symlink during initialization. 2303 */ 2304 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2305 2306 /* Set the sysfs device type for the network logical device to allow 2307 * fine-grained identification of different network device types. For 2308 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2309 */ 2310 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2311 2312 /* Default NAPI poll() weight 2313 * Device drivers are strongly advised to not use bigger value 2314 */ 2315 #define NAPI_POLL_WEIGHT 64 2316 2317 /** 2318 * netif_napi_add - initialize a NAPI context 2319 * @dev: network device 2320 * @napi: NAPI context 2321 * @poll: polling function 2322 * @weight: default weight 2323 * 2324 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2325 * *any* of the other NAPI-related functions. 2326 */ 2327 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2328 int (*poll)(struct napi_struct *, int), int weight); 2329 2330 /** 2331 * netif_tx_napi_add - initialize a NAPI context 2332 * @dev: network device 2333 * @napi: NAPI context 2334 * @poll: polling function 2335 * @weight: default weight 2336 * 2337 * This variant of netif_napi_add() should be used from drivers using NAPI 2338 * to exclusively poll a TX queue. 2339 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2340 */ 2341 static inline void netif_tx_napi_add(struct net_device *dev, 2342 struct napi_struct *napi, 2343 int (*poll)(struct napi_struct *, int), 2344 int weight) 2345 { 2346 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2347 netif_napi_add(dev, napi, poll, weight); 2348 } 2349 2350 /** 2351 * netif_napi_del - remove a NAPI context 2352 * @napi: NAPI context 2353 * 2354 * netif_napi_del() removes a NAPI context from the network device NAPI list 2355 */ 2356 void netif_napi_del(struct napi_struct *napi); 2357 2358 struct napi_gro_cb { 2359 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2360 void *frag0; 2361 2362 /* Length of frag0. */ 2363 unsigned int frag0_len; 2364 2365 /* This indicates where we are processing relative to skb->data. */ 2366 int data_offset; 2367 2368 /* This is non-zero if the packet cannot be merged with the new skb. */ 2369 u16 flush; 2370 2371 /* Save the IP ID here and check when we get to the transport layer */ 2372 u16 flush_id; 2373 2374 /* Number of segments aggregated. */ 2375 u16 count; 2376 2377 /* Start offset for remote checksum offload */ 2378 u16 gro_remcsum_start; 2379 2380 /* jiffies when first packet was created/queued */ 2381 unsigned long age; 2382 2383 /* Used in ipv6_gro_receive() and foo-over-udp */ 2384 u16 proto; 2385 2386 /* This is non-zero if the packet may be of the same flow. */ 2387 u8 same_flow:1; 2388 2389 /* Used in tunnel GRO receive */ 2390 u8 encap_mark:1; 2391 2392 /* GRO checksum is valid */ 2393 u8 csum_valid:1; 2394 2395 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2396 u8 csum_cnt:3; 2397 2398 /* Free the skb? */ 2399 u8 free:2; 2400 #define NAPI_GRO_FREE 1 2401 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2402 2403 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2404 u8 is_ipv6:1; 2405 2406 /* Used in GRE, set in fou/gue_gro_receive */ 2407 u8 is_fou:1; 2408 2409 /* Used to determine if flush_id can be ignored */ 2410 u8 is_atomic:1; 2411 2412 /* Number of gro_receive callbacks this packet already went through */ 2413 u8 recursion_counter:4; 2414 2415 /* GRO is done by frag_list pointer chaining. */ 2416 u8 is_flist:1; 2417 2418 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2419 __wsum csum; 2420 2421 /* used in skb_gro_receive() slow path */ 2422 struct sk_buff *last; 2423 }; 2424 2425 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2426 2427 #define GRO_RECURSION_LIMIT 15 2428 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2429 { 2430 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2431 } 2432 2433 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2434 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2435 struct list_head *head, 2436 struct sk_buff *skb) 2437 { 2438 if (unlikely(gro_recursion_inc_test(skb))) { 2439 NAPI_GRO_CB(skb)->flush |= 1; 2440 return NULL; 2441 } 2442 2443 return cb(head, skb); 2444 } 2445 2446 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2447 struct sk_buff *); 2448 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2449 struct sock *sk, 2450 struct list_head *head, 2451 struct sk_buff *skb) 2452 { 2453 if (unlikely(gro_recursion_inc_test(skb))) { 2454 NAPI_GRO_CB(skb)->flush |= 1; 2455 return NULL; 2456 } 2457 2458 return cb(sk, head, skb); 2459 } 2460 2461 struct packet_type { 2462 __be16 type; /* This is really htons(ether_type). */ 2463 bool ignore_outgoing; 2464 struct net_device *dev; /* NULL is wildcarded here */ 2465 int (*func) (struct sk_buff *, 2466 struct net_device *, 2467 struct packet_type *, 2468 struct net_device *); 2469 void (*list_func) (struct list_head *, 2470 struct packet_type *, 2471 struct net_device *); 2472 bool (*id_match)(struct packet_type *ptype, 2473 struct sock *sk); 2474 void *af_packet_priv; 2475 struct list_head list; 2476 }; 2477 2478 struct offload_callbacks { 2479 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2480 netdev_features_t features); 2481 struct sk_buff *(*gro_receive)(struct list_head *head, 2482 struct sk_buff *skb); 2483 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2484 }; 2485 2486 struct packet_offload { 2487 __be16 type; /* This is really htons(ether_type). */ 2488 u16 priority; 2489 struct offload_callbacks callbacks; 2490 struct list_head list; 2491 }; 2492 2493 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2494 struct pcpu_sw_netstats { 2495 u64 rx_packets; 2496 u64 rx_bytes; 2497 u64 tx_packets; 2498 u64 tx_bytes; 2499 struct u64_stats_sync syncp; 2500 } __aligned(4 * sizeof(u64)); 2501 2502 struct pcpu_lstats { 2503 u64_stats_t packets; 2504 u64_stats_t bytes; 2505 struct u64_stats_sync syncp; 2506 } __aligned(2 * sizeof(u64)); 2507 2508 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2509 2510 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2511 { 2512 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2513 2514 u64_stats_update_begin(&lstats->syncp); 2515 u64_stats_add(&lstats->bytes, len); 2516 u64_stats_inc(&lstats->packets); 2517 u64_stats_update_end(&lstats->syncp); 2518 } 2519 2520 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2521 ({ \ 2522 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2523 if (pcpu_stats) { \ 2524 int __cpu; \ 2525 for_each_possible_cpu(__cpu) { \ 2526 typeof(type) *stat; \ 2527 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2528 u64_stats_init(&stat->syncp); \ 2529 } \ 2530 } \ 2531 pcpu_stats; \ 2532 }) 2533 2534 #define netdev_alloc_pcpu_stats(type) \ 2535 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2536 2537 enum netdev_lag_tx_type { 2538 NETDEV_LAG_TX_TYPE_UNKNOWN, 2539 NETDEV_LAG_TX_TYPE_RANDOM, 2540 NETDEV_LAG_TX_TYPE_BROADCAST, 2541 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2542 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2543 NETDEV_LAG_TX_TYPE_HASH, 2544 }; 2545 2546 enum netdev_lag_hash { 2547 NETDEV_LAG_HASH_NONE, 2548 NETDEV_LAG_HASH_L2, 2549 NETDEV_LAG_HASH_L34, 2550 NETDEV_LAG_HASH_L23, 2551 NETDEV_LAG_HASH_E23, 2552 NETDEV_LAG_HASH_E34, 2553 NETDEV_LAG_HASH_UNKNOWN, 2554 }; 2555 2556 struct netdev_lag_upper_info { 2557 enum netdev_lag_tx_type tx_type; 2558 enum netdev_lag_hash hash_type; 2559 }; 2560 2561 struct netdev_lag_lower_state_info { 2562 u8 link_up : 1, 2563 tx_enabled : 1; 2564 }; 2565 2566 #include <linux/notifier.h> 2567 2568 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2569 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2570 * adding new types. 2571 */ 2572 enum netdev_cmd { 2573 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2574 NETDEV_DOWN, 2575 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2576 detected a hardware crash and restarted 2577 - we can use this eg to kick tcp sessions 2578 once done */ 2579 NETDEV_CHANGE, /* Notify device state change */ 2580 NETDEV_REGISTER, 2581 NETDEV_UNREGISTER, 2582 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2583 NETDEV_CHANGEADDR, /* notify after the address change */ 2584 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2585 NETDEV_GOING_DOWN, 2586 NETDEV_CHANGENAME, 2587 NETDEV_FEAT_CHANGE, 2588 NETDEV_BONDING_FAILOVER, 2589 NETDEV_PRE_UP, 2590 NETDEV_PRE_TYPE_CHANGE, 2591 NETDEV_POST_TYPE_CHANGE, 2592 NETDEV_POST_INIT, 2593 NETDEV_RELEASE, 2594 NETDEV_NOTIFY_PEERS, 2595 NETDEV_JOIN, 2596 NETDEV_CHANGEUPPER, 2597 NETDEV_RESEND_IGMP, 2598 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2599 NETDEV_CHANGEINFODATA, 2600 NETDEV_BONDING_INFO, 2601 NETDEV_PRECHANGEUPPER, 2602 NETDEV_CHANGELOWERSTATE, 2603 NETDEV_UDP_TUNNEL_PUSH_INFO, 2604 NETDEV_UDP_TUNNEL_DROP_INFO, 2605 NETDEV_CHANGE_TX_QUEUE_LEN, 2606 NETDEV_CVLAN_FILTER_PUSH_INFO, 2607 NETDEV_CVLAN_FILTER_DROP_INFO, 2608 NETDEV_SVLAN_FILTER_PUSH_INFO, 2609 NETDEV_SVLAN_FILTER_DROP_INFO, 2610 }; 2611 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2612 2613 int register_netdevice_notifier(struct notifier_block *nb); 2614 int unregister_netdevice_notifier(struct notifier_block *nb); 2615 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2616 int unregister_netdevice_notifier_net(struct net *net, 2617 struct notifier_block *nb); 2618 int register_netdevice_notifier_dev_net(struct net_device *dev, 2619 struct notifier_block *nb, 2620 struct netdev_net_notifier *nn); 2621 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2622 struct notifier_block *nb, 2623 struct netdev_net_notifier *nn); 2624 2625 struct netdev_notifier_info { 2626 struct net_device *dev; 2627 struct netlink_ext_ack *extack; 2628 }; 2629 2630 struct netdev_notifier_info_ext { 2631 struct netdev_notifier_info info; /* must be first */ 2632 union { 2633 u32 mtu; 2634 } ext; 2635 }; 2636 2637 struct netdev_notifier_change_info { 2638 struct netdev_notifier_info info; /* must be first */ 2639 unsigned int flags_changed; 2640 }; 2641 2642 struct netdev_notifier_changeupper_info { 2643 struct netdev_notifier_info info; /* must be first */ 2644 struct net_device *upper_dev; /* new upper dev */ 2645 bool master; /* is upper dev master */ 2646 bool linking; /* is the notification for link or unlink */ 2647 void *upper_info; /* upper dev info */ 2648 }; 2649 2650 struct netdev_notifier_changelowerstate_info { 2651 struct netdev_notifier_info info; /* must be first */ 2652 void *lower_state_info; /* is lower dev state */ 2653 }; 2654 2655 struct netdev_notifier_pre_changeaddr_info { 2656 struct netdev_notifier_info info; /* must be first */ 2657 const unsigned char *dev_addr; 2658 }; 2659 2660 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2661 struct net_device *dev) 2662 { 2663 info->dev = dev; 2664 info->extack = NULL; 2665 } 2666 2667 static inline struct net_device * 2668 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2669 { 2670 return info->dev; 2671 } 2672 2673 static inline struct netlink_ext_ack * 2674 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2675 { 2676 return info->extack; 2677 } 2678 2679 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2680 2681 2682 extern rwlock_t dev_base_lock; /* Device list lock */ 2683 2684 #define for_each_netdev(net, d) \ 2685 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2686 #define for_each_netdev_reverse(net, d) \ 2687 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2688 #define for_each_netdev_rcu(net, d) \ 2689 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2690 #define for_each_netdev_safe(net, d, n) \ 2691 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2692 #define for_each_netdev_continue(net, d) \ 2693 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2694 #define for_each_netdev_continue_reverse(net, d) \ 2695 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2696 dev_list) 2697 #define for_each_netdev_continue_rcu(net, d) \ 2698 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2699 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2700 for_each_netdev_rcu(&init_net, slave) \ 2701 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2702 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2703 2704 static inline struct net_device *next_net_device(struct net_device *dev) 2705 { 2706 struct list_head *lh; 2707 struct net *net; 2708 2709 net = dev_net(dev); 2710 lh = dev->dev_list.next; 2711 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2712 } 2713 2714 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2715 { 2716 struct list_head *lh; 2717 struct net *net; 2718 2719 net = dev_net(dev); 2720 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2721 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2722 } 2723 2724 static inline struct net_device *first_net_device(struct net *net) 2725 { 2726 return list_empty(&net->dev_base_head) ? NULL : 2727 net_device_entry(net->dev_base_head.next); 2728 } 2729 2730 static inline struct net_device *first_net_device_rcu(struct net *net) 2731 { 2732 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2733 2734 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2735 } 2736 2737 int netdev_boot_setup_check(struct net_device *dev); 2738 unsigned long netdev_boot_base(const char *prefix, int unit); 2739 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2740 const char *hwaddr); 2741 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2742 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2743 void dev_add_pack(struct packet_type *pt); 2744 void dev_remove_pack(struct packet_type *pt); 2745 void __dev_remove_pack(struct packet_type *pt); 2746 void dev_add_offload(struct packet_offload *po); 2747 void dev_remove_offload(struct packet_offload *po); 2748 2749 int dev_get_iflink(const struct net_device *dev); 2750 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2751 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2752 unsigned short mask); 2753 struct net_device *dev_get_by_name(struct net *net, const char *name); 2754 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2755 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2756 int dev_alloc_name(struct net_device *dev, const char *name); 2757 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2758 void dev_close(struct net_device *dev); 2759 void dev_close_many(struct list_head *head, bool unlink); 2760 void dev_disable_lro(struct net_device *dev); 2761 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2762 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2763 struct net_device *sb_dev); 2764 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2765 struct net_device *sb_dev); 2766 int dev_queue_xmit(struct sk_buff *skb); 2767 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2768 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2769 int register_netdevice(struct net_device *dev); 2770 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2771 void unregister_netdevice_many(struct list_head *head); 2772 static inline void unregister_netdevice(struct net_device *dev) 2773 { 2774 unregister_netdevice_queue(dev, NULL); 2775 } 2776 2777 int netdev_refcnt_read(const struct net_device *dev); 2778 void free_netdev(struct net_device *dev); 2779 void netdev_freemem(struct net_device *dev); 2780 void synchronize_net(void); 2781 int init_dummy_netdev(struct net_device *dev); 2782 2783 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2784 struct sk_buff *skb, 2785 bool all_slaves); 2786 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2787 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2788 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2789 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2790 int netdev_get_name(struct net *net, char *name, int ifindex); 2791 int dev_restart(struct net_device *dev); 2792 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2793 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2794 2795 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2796 { 2797 return NAPI_GRO_CB(skb)->data_offset; 2798 } 2799 2800 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2801 { 2802 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2803 } 2804 2805 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2806 { 2807 NAPI_GRO_CB(skb)->data_offset += len; 2808 } 2809 2810 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2811 unsigned int offset) 2812 { 2813 return NAPI_GRO_CB(skb)->frag0 + offset; 2814 } 2815 2816 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2817 { 2818 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2819 } 2820 2821 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2822 { 2823 NAPI_GRO_CB(skb)->frag0 = NULL; 2824 NAPI_GRO_CB(skb)->frag0_len = 0; 2825 } 2826 2827 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2828 unsigned int offset) 2829 { 2830 if (!pskb_may_pull(skb, hlen)) 2831 return NULL; 2832 2833 skb_gro_frag0_invalidate(skb); 2834 return skb->data + offset; 2835 } 2836 2837 static inline void *skb_gro_network_header(struct sk_buff *skb) 2838 { 2839 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2840 skb_network_offset(skb); 2841 } 2842 2843 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2844 const void *start, unsigned int len) 2845 { 2846 if (NAPI_GRO_CB(skb)->csum_valid) 2847 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2848 csum_partial(start, len, 0)); 2849 } 2850 2851 /* GRO checksum functions. These are logical equivalents of the normal 2852 * checksum functions (in skbuff.h) except that they operate on the GRO 2853 * offsets and fields in sk_buff. 2854 */ 2855 2856 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2857 2858 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2859 { 2860 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2861 } 2862 2863 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2864 bool zero_okay, 2865 __sum16 check) 2866 { 2867 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2868 skb_checksum_start_offset(skb) < 2869 skb_gro_offset(skb)) && 2870 !skb_at_gro_remcsum_start(skb) && 2871 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2872 (!zero_okay || check)); 2873 } 2874 2875 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2876 __wsum psum) 2877 { 2878 if (NAPI_GRO_CB(skb)->csum_valid && 2879 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2880 return 0; 2881 2882 NAPI_GRO_CB(skb)->csum = psum; 2883 2884 return __skb_gro_checksum_complete(skb); 2885 } 2886 2887 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2888 { 2889 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2890 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2891 NAPI_GRO_CB(skb)->csum_cnt--; 2892 } else { 2893 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2894 * verified a new top level checksum or an encapsulated one 2895 * during GRO. This saves work if we fallback to normal path. 2896 */ 2897 __skb_incr_checksum_unnecessary(skb); 2898 } 2899 } 2900 2901 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2902 compute_pseudo) \ 2903 ({ \ 2904 __sum16 __ret = 0; \ 2905 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2906 __ret = __skb_gro_checksum_validate_complete(skb, \ 2907 compute_pseudo(skb, proto)); \ 2908 if (!__ret) \ 2909 skb_gro_incr_csum_unnecessary(skb); \ 2910 __ret; \ 2911 }) 2912 2913 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2914 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2915 2916 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2917 compute_pseudo) \ 2918 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2919 2920 #define skb_gro_checksum_simple_validate(skb) \ 2921 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2922 2923 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2924 { 2925 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2926 !NAPI_GRO_CB(skb)->csum_valid); 2927 } 2928 2929 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2930 __wsum pseudo) 2931 { 2932 NAPI_GRO_CB(skb)->csum = ~pseudo; 2933 NAPI_GRO_CB(skb)->csum_valid = 1; 2934 } 2935 2936 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2937 do { \ 2938 if (__skb_gro_checksum_convert_check(skb)) \ 2939 __skb_gro_checksum_convert(skb, \ 2940 compute_pseudo(skb, proto)); \ 2941 } while (0) 2942 2943 struct gro_remcsum { 2944 int offset; 2945 __wsum delta; 2946 }; 2947 2948 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2949 { 2950 grc->offset = 0; 2951 grc->delta = 0; 2952 } 2953 2954 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2955 unsigned int off, size_t hdrlen, 2956 int start, int offset, 2957 struct gro_remcsum *grc, 2958 bool nopartial) 2959 { 2960 __wsum delta; 2961 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2962 2963 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2964 2965 if (!nopartial) { 2966 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2967 return ptr; 2968 } 2969 2970 ptr = skb_gro_header_fast(skb, off); 2971 if (skb_gro_header_hard(skb, off + plen)) { 2972 ptr = skb_gro_header_slow(skb, off + plen, off); 2973 if (!ptr) 2974 return NULL; 2975 } 2976 2977 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2978 start, offset); 2979 2980 /* Adjust skb->csum since we changed the packet */ 2981 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2982 2983 grc->offset = off + hdrlen + offset; 2984 grc->delta = delta; 2985 2986 return ptr; 2987 } 2988 2989 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2990 struct gro_remcsum *grc) 2991 { 2992 void *ptr; 2993 size_t plen = grc->offset + sizeof(u16); 2994 2995 if (!grc->delta) 2996 return; 2997 2998 ptr = skb_gro_header_fast(skb, grc->offset); 2999 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3000 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3001 if (!ptr) 3002 return; 3003 } 3004 3005 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3006 } 3007 3008 #ifdef CONFIG_XFRM_OFFLOAD 3009 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3010 { 3011 if (PTR_ERR(pp) != -EINPROGRESS) 3012 NAPI_GRO_CB(skb)->flush |= flush; 3013 } 3014 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3015 struct sk_buff *pp, 3016 int flush, 3017 struct gro_remcsum *grc) 3018 { 3019 if (PTR_ERR(pp) != -EINPROGRESS) { 3020 NAPI_GRO_CB(skb)->flush |= flush; 3021 skb_gro_remcsum_cleanup(skb, grc); 3022 skb->remcsum_offload = 0; 3023 } 3024 } 3025 #else 3026 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3027 { 3028 NAPI_GRO_CB(skb)->flush |= flush; 3029 } 3030 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3031 struct sk_buff *pp, 3032 int flush, 3033 struct gro_remcsum *grc) 3034 { 3035 NAPI_GRO_CB(skb)->flush |= flush; 3036 skb_gro_remcsum_cleanup(skb, grc); 3037 skb->remcsum_offload = 0; 3038 } 3039 #endif 3040 3041 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3042 unsigned short type, 3043 const void *daddr, const void *saddr, 3044 unsigned int len) 3045 { 3046 if (!dev->header_ops || !dev->header_ops->create) 3047 return 0; 3048 3049 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3050 } 3051 3052 static inline int dev_parse_header(const struct sk_buff *skb, 3053 unsigned char *haddr) 3054 { 3055 const struct net_device *dev = skb->dev; 3056 3057 if (!dev->header_ops || !dev->header_ops->parse) 3058 return 0; 3059 return dev->header_ops->parse(skb, haddr); 3060 } 3061 3062 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3063 { 3064 const struct net_device *dev = skb->dev; 3065 3066 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3067 return 0; 3068 return dev->header_ops->parse_protocol(skb); 3069 } 3070 3071 /* ll_header must have at least hard_header_len allocated */ 3072 static inline bool dev_validate_header(const struct net_device *dev, 3073 char *ll_header, int len) 3074 { 3075 if (likely(len >= dev->hard_header_len)) 3076 return true; 3077 if (len < dev->min_header_len) 3078 return false; 3079 3080 if (capable(CAP_SYS_RAWIO)) { 3081 memset(ll_header + len, 0, dev->hard_header_len - len); 3082 return true; 3083 } 3084 3085 if (dev->header_ops && dev->header_ops->validate) 3086 return dev->header_ops->validate(ll_header, len); 3087 3088 return false; 3089 } 3090 3091 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3092 int len, int size); 3093 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3094 static inline int unregister_gifconf(unsigned int family) 3095 { 3096 return register_gifconf(family, NULL); 3097 } 3098 3099 #ifdef CONFIG_NET_FLOW_LIMIT 3100 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3101 struct sd_flow_limit { 3102 u64 count; 3103 unsigned int num_buckets; 3104 unsigned int history_head; 3105 u16 history[FLOW_LIMIT_HISTORY]; 3106 u8 buckets[]; 3107 }; 3108 3109 extern int netdev_flow_limit_table_len; 3110 #endif /* CONFIG_NET_FLOW_LIMIT */ 3111 3112 /* 3113 * Incoming packets are placed on per-CPU queues 3114 */ 3115 struct softnet_data { 3116 struct list_head poll_list; 3117 struct sk_buff_head process_queue; 3118 3119 /* stats */ 3120 unsigned int processed; 3121 unsigned int time_squeeze; 3122 unsigned int received_rps; 3123 #ifdef CONFIG_RPS 3124 struct softnet_data *rps_ipi_list; 3125 #endif 3126 #ifdef CONFIG_NET_FLOW_LIMIT 3127 struct sd_flow_limit __rcu *flow_limit; 3128 #endif 3129 struct Qdisc *output_queue; 3130 struct Qdisc **output_queue_tailp; 3131 struct sk_buff *completion_queue; 3132 #ifdef CONFIG_XFRM_OFFLOAD 3133 struct sk_buff_head xfrm_backlog; 3134 #endif 3135 /* written and read only by owning cpu: */ 3136 struct { 3137 u16 recursion; 3138 u8 more; 3139 } xmit; 3140 #ifdef CONFIG_RPS 3141 /* input_queue_head should be written by cpu owning this struct, 3142 * and only read by other cpus. Worth using a cache line. 3143 */ 3144 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3145 3146 /* Elements below can be accessed between CPUs for RPS/RFS */ 3147 call_single_data_t csd ____cacheline_aligned_in_smp; 3148 struct softnet_data *rps_ipi_next; 3149 unsigned int cpu; 3150 unsigned int input_queue_tail; 3151 #endif 3152 unsigned int dropped; 3153 struct sk_buff_head input_pkt_queue; 3154 struct napi_struct backlog; 3155 3156 }; 3157 3158 static inline void input_queue_head_incr(struct softnet_data *sd) 3159 { 3160 #ifdef CONFIG_RPS 3161 sd->input_queue_head++; 3162 #endif 3163 } 3164 3165 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3166 unsigned int *qtail) 3167 { 3168 #ifdef CONFIG_RPS 3169 *qtail = ++sd->input_queue_tail; 3170 #endif 3171 } 3172 3173 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3174 3175 static inline int dev_recursion_level(void) 3176 { 3177 return this_cpu_read(softnet_data.xmit.recursion); 3178 } 3179 3180 #define XMIT_RECURSION_LIMIT 8 3181 static inline bool dev_xmit_recursion(void) 3182 { 3183 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3184 XMIT_RECURSION_LIMIT); 3185 } 3186 3187 static inline void dev_xmit_recursion_inc(void) 3188 { 3189 __this_cpu_inc(softnet_data.xmit.recursion); 3190 } 3191 3192 static inline void dev_xmit_recursion_dec(void) 3193 { 3194 __this_cpu_dec(softnet_data.xmit.recursion); 3195 } 3196 3197 void __netif_schedule(struct Qdisc *q); 3198 void netif_schedule_queue(struct netdev_queue *txq); 3199 3200 static inline void netif_tx_schedule_all(struct net_device *dev) 3201 { 3202 unsigned int i; 3203 3204 for (i = 0; i < dev->num_tx_queues; i++) 3205 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3206 } 3207 3208 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3209 { 3210 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3211 } 3212 3213 /** 3214 * netif_start_queue - allow transmit 3215 * @dev: network device 3216 * 3217 * Allow upper layers to call the device hard_start_xmit routine. 3218 */ 3219 static inline void netif_start_queue(struct net_device *dev) 3220 { 3221 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3222 } 3223 3224 static inline void netif_tx_start_all_queues(struct net_device *dev) 3225 { 3226 unsigned int i; 3227 3228 for (i = 0; i < dev->num_tx_queues; i++) { 3229 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3230 netif_tx_start_queue(txq); 3231 } 3232 } 3233 3234 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3235 3236 /** 3237 * netif_wake_queue - restart transmit 3238 * @dev: network device 3239 * 3240 * Allow upper layers to call the device hard_start_xmit routine. 3241 * Used for flow control when transmit resources are available. 3242 */ 3243 static inline void netif_wake_queue(struct net_device *dev) 3244 { 3245 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3246 } 3247 3248 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3249 { 3250 unsigned int i; 3251 3252 for (i = 0; i < dev->num_tx_queues; i++) { 3253 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3254 netif_tx_wake_queue(txq); 3255 } 3256 } 3257 3258 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3259 { 3260 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3261 } 3262 3263 /** 3264 * netif_stop_queue - stop transmitted packets 3265 * @dev: network device 3266 * 3267 * Stop upper layers calling the device hard_start_xmit routine. 3268 * Used for flow control when transmit resources are unavailable. 3269 */ 3270 static inline void netif_stop_queue(struct net_device *dev) 3271 { 3272 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3273 } 3274 3275 void netif_tx_stop_all_queues(struct net_device *dev); 3276 3277 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3278 { 3279 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3280 } 3281 3282 /** 3283 * netif_queue_stopped - test if transmit queue is flowblocked 3284 * @dev: network device 3285 * 3286 * Test if transmit queue on device is currently unable to send. 3287 */ 3288 static inline bool netif_queue_stopped(const struct net_device *dev) 3289 { 3290 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3291 } 3292 3293 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3294 { 3295 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3296 } 3297 3298 static inline bool 3299 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3300 { 3301 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3302 } 3303 3304 static inline bool 3305 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3306 { 3307 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3308 } 3309 3310 /** 3311 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3312 * @dev_queue: pointer to transmit queue 3313 * 3314 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3315 * to give appropriate hint to the CPU. 3316 */ 3317 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3318 { 3319 #ifdef CONFIG_BQL 3320 prefetchw(&dev_queue->dql.num_queued); 3321 #endif 3322 } 3323 3324 /** 3325 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3326 * @dev_queue: pointer to transmit queue 3327 * 3328 * BQL enabled drivers might use this helper in their TX completion path, 3329 * to give appropriate hint to the CPU. 3330 */ 3331 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3332 { 3333 #ifdef CONFIG_BQL 3334 prefetchw(&dev_queue->dql.limit); 3335 #endif 3336 } 3337 3338 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3339 unsigned int bytes) 3340 { 3341 #ifdef CONFIG_BQL 3342 dql_queued(&dev_queue->dql, bytes); 3343 3344 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3345 return; 3346 3347 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3348 3349 /* 3350 * The XOFF flag must be set before checking the dql_avail below, 3351 * because in netdev_tx_completed_queue we update the dql_completed 3352 * before checking the XOFF flag. 3353 */ 3354 smp_mb(); 3355 3356 /* check again in case another CPU has just made room avail */ 3357 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3358 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3359 #endif 3360 } 3361 3362 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3363 * that they should not test BQL status themselves. 3364 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3365 * skb of a batch. 3366 * Returns true if the doorbell must be used to kick the NIC. 3367 */ 3368 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3369 unsigned int bytes, 3370 bool xmit_more) 3371 { 3372 if (xmit_more) { 3373 #ifdef CONFIG_BQL 3374 dql_queued(&dev_queue->dql, bytes); 3375 #endif 3376 return netif_tx_queue_stopped(dev_queue); 3377 } 3378 netdev_tx_sent_queue(dev_queue, bytes); 3379 return true; 3380 } 3381 3382 /** 3383 * netdev_sent_queue - report the number of bytes queued to hardware 3384 * @dev: network device 3385 * @bytes: number of bytes queued to the hardware device queue 3386 * 3387 * Report the number of bytes queued for sending/completion to the network 3388 * device hardware queue. @bytes should be a good approximation and should 3389 * exactly match netdev_completed_queue() @bytes 3390 */ 3391 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3392 { 3393 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3394 } 3395 3396 static inline bool __netdev_sent_queue(struct net_device *dev, 3397 unsigned int bytes, 3398 bool xmit_more) 3399 { 3400 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3401 xmit_more); 3402 } 3403 3404 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3405 unsigned int pkts, unsigned int bytes) 3406 { 3407 #ifdef CONFIG_BQL 3408 if (unlikely(!bytes)) 3409 return; 3410 3411 dql_completed(&dev_queue->dql, bytes); 3412 3413 /* 3414 * Without the memory barrier there is a small possiblity that 3415 * netdev_tx_sent_queue will miss the update and cause the queue to 3416 * be stopped forever 3417 */ 3418 smp_mb(); 3419 3420 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3421 return; 3422 3423 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3424 netif_schedule_queue(dev_queue); 3425 #endif 3426 } 3427 3428 /** 3429 * netdev_completed_queue - report bytes and packets completed by device 3430 * @dev: network device 3431 * @pkts: actual number of packets sent over the medium 3432 * @bytes: actual number of bytes sent over the medium 3433 * 3434 * Report the number of bytes and packets transmitted by the network device 3435 * hardware queue over the physical medium, @bytes must exactly match the 3436 * @bytes amount passed to netdev_sent_queue() 3437 */ 3438 static inline void netdev_completed_queue(struct net_device *dev, 3439 unsigned int pkts, unsigned int bytes) 3440 { 3441 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3442 } 3443 3444 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3445 { 3446 #ifdef CONFIG_BQL 3447 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3448 dql_reset(&q->dql); 3449 #endif 3450 } 3451 3452 /** 3453 * netdev_reset_queue - reset the packets and bytes count of a network device 3454 * @dev_queue: network device 3455 * 3456 * Reset the bytes and packet count of a network device and clear the 3457 * software flow control OFF bit for this network device 3458 */ 3459 static inline void netdev_reset_queue(struct net_device *dev_queue) 3460 { 3461 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3462 } 3463 3464 /** 3465 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3466 * @dev: network device 3467 * @queue_index: given tx queue index 3468 * 3469 * Returns 0 if given tx queue index >= number of device tx queues, 3470 * otherwise returns the originally passed tx queue index. 3471 */ 3472 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3473 { 3474 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3475 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3476 dev->name, queue_index, 3477 dev->real_num_tx_queues); 3478 return 0; 3479 } 3480 3481 return queue_index; 3482 } 3483 3484 /** 3485 * netif_running - test if up 3486 * @dev: network device 3487 * 3488 * Test if the device has been brought up. 3489 */ 3490 static inline bool netif_running(const struct net_device *dev) 3491 { 3492 return test_bit(__LINK_STATE_START, &dev->state); 3493 } 3494 3495 /* 3496 * Routines to manage the subqueues on a device. We only need start, 3497 * stop, and a check if it's stopped. All other device management is 3498 * done at the overall netdevice level. 3499 * Also test the device if we're multiqueue. 3500 */ 3501 3502 /** 3503 * netif_start_subqueue - allow sending packets on subqueue 3504 * @dev: network device 3505 * @queue_index: sub queue index 3506 * 3507 * Start individual transmit queue of a device with multiple transmit queues. 3508 */ 3509 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3510 { 3511 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3512 3513 netif_tx_start_queue(txq); 3514 } 3515 3516 /** 3517 * netif_stop_subqueue - stop sending packets on subqueue 3518 * @dev: network device 3519 * @queue_index: sub queue index 3520 * 3521 * Stop individual transmit queue of a device with multiple transmit queues. 3522 */ 3523 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3524 { 3525 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3526 netif_tx_stop_queue(txq); 3527 } 3528 3529 /** 3530 * netif_subqueue_stopped - test status of subqueue 3531 * @dev: network device 3532 * @queue_index: sub queue index 3533 * 3534 * Check individual transmit queue of a device with multiple transmit queues. 3535 */ 3536 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3537 u16 queue_index) 3538 { 3539 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3540 3541 return netif_tx_queue_stopped(txq); 3542 } 3543 3544 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3545 struct sk_buff *skb) 3546 { 3547 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3548 } 3549 3550 /** 3551 * netif_wake_subqueue - allow sending packets on subqueue 3552 * @dev: network device 3553 * @queue_index: sub queue index 3554 * 3555 * Resume individual transmit queue of a device with multiple transmit queues. 3556 */ 3557 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3558 { 3559 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3560 3561 netif_tx_wake_queue(txq); 3562 } 3563 3564 #ifdef CONFIG_XPS 3565 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3566 u16 index); 3567 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3568 u16 index, bool is_rxqs_map); 3569 3570 /** 3571 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3572 * @j: CPU/Rx queue index 3573 * @mask: bitmask of all cpus/rx queues 3574 * @nr_bits: number of bits in the bitmask 3575 * 3576 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3577 */ 3578 static inline bool netif_attr_test_mask(unsigned long j, 3579 const unsigned long *mask, 3580 unsigned int nr_bits) 3581 { 3582 cpu_max_bits_warn(j, nr_bits); 3583 return test_bit(j, mask); 3584 } 3585 3586 /** 3587 * netif_attr_test_online - Test for online CPU/Rx queue 3588 * @j: CPU/Rx queue index 3589 * @online_mask: bitmask for CPUs/Rx queues that are online 3590 * @nr_bits: number of bits in the bitmask 3591 * 3592 * Returns true if a CPU/Rx queue is online. 3593 */ 3594 static inline bool netif_attr_test_online(unsigned long j, 3595 const unsigned long *online_mask, 3596 unsigned int nr_bits) 3597 { 3598 cpu_max_bits_warn(j, nr_bits); 3599 3600 if (online_mask) 3601 return test_bit(j, online_mask); 3602 3603 return (j < nr_bits); 3604 } 3605 3606 /** 3607 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3608 * @n: CPU/Rx queue index 3609 * @srcp: the cpumask/Rx queue mask pointer 3610 * @nr_bits: number of bits in the bitmask 3611 * 3612 * Returns >= nr_bits if no further CPUs/Rx queues set. 3613 */ 3614 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3615 unsigned int nr_bits) 3616 { 3617 /* -1 is a legal arg here. */ 3618 if (n != -1) 3619 cpu_max_bits_warn(n, nr_bits); 3620 3621 if (srcp) 3622 return find_next_bit(srcp, nr_bits, n + 1); 3623 3624 return n + 1; 3625 } 3626 3627 /** 3628 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3629 * @n: CPU/Rx queue index 3630 * @src1p: the first CPUs/Rx queues mask pointer 3631 * @src2p: the second CPUs/Rx queues mask pointer 3632 * @nr_bits: number of bits in the bitmask 3633 * 3634 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3635 */ 3636 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3637 const unsigned long *src2p, 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 (src1p && src2p) 3645 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3646 else if (src1p) 3647 return find_next_bit(src1p, nr_bits, n + 1); 3648 else if (src2p) 3649 return find_next_bit(src2p, nr_bits, n + 1); 3650 3651 return n + 1; 3652 } 3653 #else 3654 static inline int netif_set_xps_queue(struct net_device *dev, 3655 const struct cpumask *mask, 3656 u16 index) 3657 { 3658 return 0; 3659 } 3660 3661 static inline int __netif_set_xps_queue(struct net_device *dev, 3662 const unsigned long *mask, 3663 u16 index, bool is_rxqs_map) 3664 { 3665 return 0; 3666 } 3667 #endif 3668 3669 /** 3670 * netif_is_multiqueue - test if device has multiple transmit queues 3671 * @dev: network device 3672 * 3673 * Check if device has multiple transmit queues 3674 */ 3675 static inline bool netif_is_multiqueue(const struct net_device *dev) 3676 { 3677 return dev->num_tx_queues > 1; 3678 } 3679 3680 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3681 3682 #ifdef CONFIG_SYSFS 3683 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3684 #else 3685 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3686 unsigned int rxqs) 3687 { 3688 dev->real_num_rx_queues = rxqs; 3689 return 0; 3690 } 3691 #endif 3692 3693 static inline struct netdev_rx_queue * 3694 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3695 { 3696 return dev->_rx + rxq; 3697 } 3698 3699 #ifdef CONFIG_SYSFS 3700 static inline unsigned int get_netdev_rx_queue_index( 3701 struct netdev_rx_queue *queue) 3702 { 3703 struct net_device *dev = queue->dev; 3704 int index = queue - dev->_rx; 3705 3706 BUG_ON(index >= dev->num_rx_queues); 3707 return index; 3708 } 3709 #endif 3710 3711 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3712 int netif_get_num_default_rss_queues(void); 3713 3714 enum skb_free_reason { 3715 SKB_REASON_CONSUMED, 3716 SKB_REASON_DROPPED, 3717 }; 3718 3719 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3720 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3721 3722 /* 3723 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3724 * interrupt context or with hardware interrupts being disabled. 3725 * (in_irq() || irqs_disabled()) 3726 * 3727 * We provide four helpers that can be used in following contexts : 3728 * 3729 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3730 * replacing kfree_skb(skb) 3731 * 3732 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3733 * Typically used in place of consume_skb(skb) in TX completion path 3734 * 3735 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3736 * replacing kfree_skb(skb) 3737 * 3738 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3739 * and consumed a packet. Used in place of consume_skb(skb) 3740 */ 3741 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3742 { 3743 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3744 } 3745 3746 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3747 { 3748 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3749 } 3750 3751 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3752 { 3753 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3754 } 3755 3756 static inline void dev_consume_skb_any(struct sk_buff *skb) 3757 { 3758 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3759 } 3760 3761 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3762 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3763 int netif_rx(struct sk_buff *skb); 3764 int netif_rx_ni(struct sk_buff *skb); 3765 int netif_receive_skb(struct sk_buff *skb); 3766 int netif_receive_skb_core(struct sk_buff *skb); 3767 void netif_receive_skb_list(struct list_head *head); 3768 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3769 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3770 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3771 gro_result_t napi_gro_frags(struct napi_struct *napi); 3772 struct packet_offload *gro_find_receive_by_type(__be16 type); 3773 struct packet_offload *gro_find_complete_by_type(__be16 type); 3774 3775 static inline void napi_free_frags(struct napi_struct *napi) 3776 { 3777 kfree_skb(napi->skb); 3778 napi->skb = NULL; 3779 } 3780 3781 bool netdev_is_rx_handler_busy(struct net_device *dev); 3782 int netdev_rx_handler_register(struct net_device *dev, 3783 rx_handler_func_t *rx_handler, 3784 void *rx_handler_data); 3785 void netdev_rx_handler_unregister(struct net_device *dev); 3786 3787 bool dev_valid_name(const char *name); 3788 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3789 bool *need_copyout); 3790 int dev_ifconf(struct net *net, struct ifconf *, int); 3791 int dev_ethtool(struct net *net, struct ifreq *); 3792 unsigned int dev_get_flags(const struct net_device *); 3793 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3794 struct netlink_ext_ack *extack); 3795 int dev_change_flags(struct net_device *dev, unsigned int flags, 3796 struct netlink_ext_ack *extack); 3797 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3798 unsigned int gchanges); 3799 int dev_change_name(struct net_device *, const char *); 3800 int dev_set_alias(struct net_device *, const char *, size_t); 3801 int dev_get_alias(const struct net_device *, char *, size_t); 3802 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3803 int __dev_set_mtu(struct net_device *, int); 3804 int dev_validate_mtu(struct net_device *dev, int mtu, 3805 struct netlink_ext_ack *extack); 3806 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3807 struct netlink_ext_ack *extack); 3808 int dev_set_mtu(struct net_device *, int); 3809 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3810 void dev_set_group(struct net_device *, int); 3811 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3812 struct netlink_ext_ack *extack); 3813 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3814 struct netlink_ext_ack *extack); 3815 int dev_change_carrier(struct net_device *, bool new_carrier); 3816 int dev_get_phys_port_id(struct net_device *dev, 3817 struct netdev_phys_item_id *ppid); 3818 int dev_get_phys_port_name(struct net_device *dev, 3819 char *name, size_t len); 3820 int dev_get_port_parent_id(struct net_device *dev, 3821 struct netdev_phys_item_id *ppid, bool recurse); 3822 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3823 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3824 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3825 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3826 u32 value); 3827 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3828 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3829 struct netdev_queue *txq, int *ret); 3830 3831 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3832 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3833 int fd, int expected_fd, u32 flags); 3834 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3835 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3836 3837 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3838 3839 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3840 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3841 bool is_skb_forwardable(const struct net_device *dev, 3842 const struct sk_buff *skb); 3843 3844 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3845 struct sk_buff *skb) 3846 { 3847 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3848 unlikely(!is_skb_forwardable(dev, skb))) { 3849 atomic_long_inc(&dev->rx_dropped); 3850 kfree_skb(skb); 3851 return NET_RX_DROP; 3852 } 3853 3854 skb_scrub_packet(skb, true); 3855 skb->priority = 0; 3856 return 0; 3857 } 3858 3859 bool dev_nit_active(struct net_device *dev); 3860 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3861 3862 extern int netdev_budget; 3863 extern unsigned int netdev_budget_usecs; 3864 3865 /* Called by rtnetlink.c:rtnl_unlock() */ 3866 void netdev_run_todo(void); 3867 3868 /** 3869 * dev_put - release reference to device 3870 * @dev: network device 3871 * 3872 * Release reference to device to allow it to be freed. 3873 */ 3874 static inline void dev_put(struct net_device *dev) 3875 { 3876 this_cpu_dec(*dev->pcpu_refcnt); 3877 } 3878 3879 /** 3880 * dev_hold - get reference to device 3881 * @dev: network device 3882 * 3883 * Hold reference to device to keep it from being freed. 3884 */ 3885 static inline void dev_hold(struct net_device *dev) 3886 { 3887 this_cpu_inc(*dev->pcpu_refcnt); 3888 } 3889 3890 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3891 * and _off may be called from IRQ context, but it is caller 3892 * who is responsible for serialization of these calls. 3893 * 3894 * The name carrier is inappropriate, these functions should really be 3895 * called netif_lowerlayer_*() because they represent the state of any 3896 * kind of lower layer not just hardware media. 3897 */ 3898 3899 void linkwatch_init_dev(struct net_device *dev); 3900 void linkwatch_fire_event(struct net_device *dev); 3901 void linkwatch_forget_dev(struct net_device *dev); 3902 3903 /** 3904 * netif_carrier_ok - test if carrier present 3905 * @dev: network device 3906 * 3907 * Check if carrier is present on device 3908 */ 3909 static inline bool netif_carrier_ok(const struct net_device *dev) 3910 { 3911 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3912 } 3913 3914 unsigned long dev_trans_start(struct net_device *dev); 3915 3916 void __netdev_watchdog_up(struct net_device *dev); 3917 3918 void netif_carrier_on(struct net_device *dev); 3919 3920 void netif_carrier_off(struct net_device *dev); 3921 3922 /** 3923 * netif_dormant_on - mark device as dormant. 3924 * @dev: network device 3925 * 3926 * Mark device as dormant (as per RFC2863). 3927 * 3928 * The dormant state indicates that the relevant interface is not 3929 * actually in a condition to pass packets (i.e., it is not 'up') but is 3930 * in a "pending" state, waiting for some external event. For "on- 3931 * demand" interfaces, this new state identifies the situation where the 3932 * interface is waiting for events to place it in the up state. 3933 */ 3934 static inline void netif_dormant_on(struct net_device *dev) 3935 { 3936 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3937 linkwatch_fire_event(dev); 3938 } 3939 3940 /** 3941 * netif_dormant_off - set device as not dormant. 3942 * @dev: network device 3943 * 3944 * Device is not in dormant state. 3945 */ 3946 static inline void netif_dormant_off(struct net_device *dev) 3947 { 3948 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3949 linkwatch_fire_event(dev); 3950 } 3951 3952 /** 3953 * netif_dormant - test if device is dormant 3954 * @dev: network device 3955 * 3956 * Check if device is dormant. 3957 */ 3958 static inline bool netif_dormant(const struct net_device *dev) 3959 { 3960 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3961 } 3962 3963 3964 /** 3965 * netif_testing_on - mark device as under test. 3966 * @dev: network device 3967 * 3968 * Mark device as under test (as per RFC2863). 3969 * 3970 * The testing state indicates that some test(s) must be performed on 3971 * the interface. After completion, of the test, the interface state 3972 * will change to up, dormant, or down, as appropriate. 3973 */ 3974 static inline void netif_testing_on(struct net_device *dev) 3975 { 3976 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 3977 linkwatch_fire_event(dev); 3978 } 3979 3980 /** 3981 * netif_testing_off - set device as not under test. 3982 * @dev: network device 3983 * 3984 * Device is not in testing state. 3985 */ 3986 static inline void netif_testing_off(struct net_device *dev) 3987 { 3988 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 3989 linkwatch_fire_event(dev); 3990 } 3991 3992 /** 3993 * netif_testing - test if device is under test 3994 * @dev: network device 3995 * 3996 * Check if device is under test 3997 */ 3998 static inline bool netif_testing(const struct net_device *dev) 3999 { 4000 return test_bit(__LINK_STATE_TESTING, &dev->state); 4001 } 4002 4003 4004 /** 4005 * netif_oper_up - test if device is operational 4006 * @dev: network device 4007 * 4008 * Check if carrier is operational 4009 */ 4010 static inline bool netif_oper_up(const struct net_device *dev) 4011 { 4012 return (dev->operstate == IF_OPER_UP || 4013 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4014 } 4015 4016 /** 4017 * netif_device_present - is device available or removed 4018 * @dev: network device 4019 * 4020 * Check if device has not been removed from system. 4021 */ 4022 static inline bool netif_device_present(struct net_device *dev) 4023 { 4024 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4025 } 4026 4027 void netif_device_detach(struct net_device *dev); 4028 4029 void netif_device_attach(struct net_device *dev); 4030 4031 /* 4032 * Network interface message level settings 4033 */ 4034 4035 enum { 4036 NETIF_MSG_DRV_BIT, 4037 NETIF_MSG_PROBE_BIT, 4038 NETIF_MSG_LINK_BIT, 4039 NETIF_MSG_TIMER_BIT, 4040 NETIF_MSG_IFDOWN_BIT, 4041 NETIF_MSG_IFUP_BIT, 4042 NETIF_MSG_RX_ERR_BIT, 4043 NETIF_MSG_TX_ERR_BIT, 4044 NETIF_MSG_TX_QUEUED_BIT, 4045 NETIF_MSG_INTR_BIT, 4046 NETIF_MSG_TX_DONE_BIT, 4047 NETIF_MSG_RX_STATUS_BIT, 4048 NETIF_MSG_PKTDATA_BIT, 4049 NETIF_MSG_HW_BIT, 4050 NETIF_MSG_WOL_BIT, 4051 4052 /* When you add a new bit above, update netif_msg_class_names array 4053 * in net/ethtool/common.c 4054 */ 4055 NETIF_MSG_CLASS_COUNT, 4056 }; 4057 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4058 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4059 4060 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4061 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4062 4063 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4064 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4065 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4066 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4067 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4068 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4069 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4070 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4071 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4072 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4073 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4074 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4075 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4076 #define NETIF_MSG_HW __NETIF_MSG(HW) 4077 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4078 4079 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4080 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4081 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4082 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4083 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4084 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4085 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4086 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4087 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4088 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4089 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4090 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4091 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4092 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4093 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4094 4095 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4096 { 4097 /* use default */ 4098 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4099 return default_msg_enable_bits; 4100 if (debug_value == 0) /* no output */ 4101 return 0; 4102 /* set low N bits */ 4103 return (1U << debug_value) - 1; 4104 } 4105 4106 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4107 { 4108 spin_lock(&txq->_xmit_lock); 4109 txq->xmit_lock_owner = cpu; 4110 } 4111 4112 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4113 { 4114 __acquire(&txq->_xmit_lock); 4115 return true; 4116 } 4117 4118 static inline void __netif_tx_release(struct netdev_queue *txq) 4119 { 4120 __release(&txq->_xmit_lock); 4121 } 4122 4123 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4124 { 4125 spin_lock_bh(&txq->_xmit_lock); 4126 txq->xmit_lock_owner = smp_processor_id(); 4127 } 4128 4129 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4130 { 4131 bool ok = spin_trylock(&txq->_xmit_lock); 4132 if (likely(ok)) 4133 txq->xmit_lock_owner = smp_processor_id(); 4134 return ok; 4135 } 4136 4137 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4138 { 4139 txq->xmit_lock_owner = -1; 4140 spin_unlock(&txq->_xmit_lock); 4141 } 4142 4143 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4144 { 4145 txq->xmit_lock_owner = -1; 4146 spin_unlock_bh(&txq->_xmit_lock); 4147 } 4148 4149 static inline void txq_trans_update(struct netdev_queue *txq) 4150 { 4151 if (txq->xmit_lock_owner != -1) 4152 txq->trans_start = jiffies; 4153 } 4154 4155 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4156 static inline void netif_trans_update(struct net_device *dev) 4157 { 4158 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4159 4160 if (txq->trans_start != jiffies) 4161 txq->trans_start = jiffies; 4162 } 4163 4164 /** 4165 * netif_tx_lock - grab network device transmit lock 4166 * @dev: network device 4167 * 4168 * Get network device transmit lock 4169 */ 4170 static inline void netif_tx_lock(struct net_device *dev) 4171 { 4172 unsigned int i; 4173 int cpu; 4174 4175 spin_lock(&dev->tx_global_lock); 4176 cpu = smp_processor_id(); 4177 for (i = 0; i < dev->num_tx_queues; i++) { 4178 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4179 4180 /* We are the only thread of execution doing a 4181 * freeze, but we have to grab the _xmit_lock in 4182 * order to synchronize with threads which are in 4183 * the ->hard_start_xmit() handler and already 4184 * checked the frozen bit. 4185 */ 4186 __netif_tx_lock(txq, cpu); 4187 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4188 __netif_tx_unlock(txq); 4189 } 4190 } 4191 4192 static inline void netif_tx_lock_bh(struct net_device *dev) 4193 { 4194 local_bh_disable(); 4195 netif_tx_lock(dev); 4196 } 4197 4198 static inline void netif_tx_unlock(struct net_device *dev) 4199 { 4200 unsigned int i; 4201 4202 for (i = 0; i < dev->num_tx_queues; i++) { 4203 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4204 4205 /* No need to grab the _xmit_lock here. If the 4206 * queue is not stopped for another reason, we 4207 * force a schedule. 4208 */ 4209 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4210 netif_schedule_queue(txq); 4211 } 4212 spin_unlock(&dev->tx_global_lock); 4213 } 4214 4215 static inline void netif_tx_unlock_bh(struct net_device *dev) 4216 { 4217 netif_tx_unlock(dev); 4218 local_bh_enable(); 4219 } 4220 4221 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4222 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4223 __netif_tx_lock(txq, cpu); \ 4224 } else { \ 4225 __netif_tx_acquire(txq); \ 4226 } \ 4227 } 4228 4229 #define HARD_TX_TRYLOCK(dev, txq) \ 4230 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4231 __netif_tx_trylock(txq) : \ 4232 __netif_tx_acquire(txq)) 4233 4234 #define HARD_TX_UNLOCK(dev, txq) { \ 4235 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4236 __netif_tx_unlock(txq); \ 4237 } else { \ 4238 __netif_tx_release(txq); \ 4239 } \ 4240 } 4241 4242 static inline void netif_tx_disable(struct net_device *dev) 4243 { 4244 unsigned int i; 4245 int cpu; 4246 4247 local_bh_disable(); 4248 cpu = smp_processor_id(); 4249 for (i = 0; i < dev->num_tx_queues; i++) { 4250 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4251 4252 __netif_tx_lock(txq, cpu); 4253 netif_tx_stop_queue(txq); 4254 __netif_tx_unlock(txq); 4255 } 4256 local_bh_enable(); 4257 } 4258 4259 static inline void netif_addr_lock(struct net_device *dev) 4260 { 4261 spin_lock(&dev->addr_list_lock); 4262 } 4263 4264 static inline void netif_addr_lock_nested(struct net_device *dev) 4265 { 4266 spin_lock_nested(&dev->addr_list_lock, dev->lower_level); 4267 } 4268 4269 static inline void netif_addr_lock_bh(struct net_device *dev) 4270 { 4271 spin_lock_bh(&dev->addr_list_lock); 4272 } 4273 4274 static inline void netif_addr_unlock(struct net_device *dev) 4275 { 4276 spin_unlock(&dev->addr_list_lock); 4277 } 4278 4279 static inline void netif_addr_unlock_bh(struct net_device *dev) 4280 { 4281 spin_unlock_bh(&dev->addr_list_lock); 4282 } 4283 4284 /* 4285 * dev_addrs walker. Should be used only for read access. Call with 4286 * rcu_read_lock held. 4287 */ 4288 #define for_each_dev_addr(dev, ha) \ 4289 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4290 4291 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4292 4293 void ether_setup(struct net_device *dev); 4294 4295 /* Support for loadable net-drivers */ 4296 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4297 unsigned char name_assign_type, 4298 void (*setup)(struct net_device *), 4299 unsigned int txqs, unsigned int rxqs); 4300 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4301 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4302 4303 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4304 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4305 count) 4306 4307 int register_netdev(struct net_device *dev); 4308 void unregister_netdev(struct net_device *dev); 4309 4310 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4311 4312 /* General hardware address lists handling functions */ 4313 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4314 struct netdev_hw_addr_list *from_list, int addr_len); 4315 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4316 struct netdev_hw_addr_list *from_list, int addr_len); 4317 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4318 struct net_device *dev, 4319 int (*sync)(struct net_device *, const unsigned char *), 4320 int (*unsync)(struct net_device *, 4321 const unsigned char *)); 4322 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4323 struct net_device *dev, 4324 int (*sync)(struct net_device *, 4325 const unsigned char *, int), 4326 int (*unsync)(struct net_device *, 4327 const unsigned char *, int)); 4328 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4329 struct net_device *dev, 4330 int (*unsync)(struct net_device *, 4331 const unsigned char *, int)); 4332 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4333 struct net_device *dev, 4334 int (*unsync)(struct net_device *, 4335 const unsigned char *)); 4336 void __hw_addr_init(struct netdev_hw_addr_list *list); 4337 4338 /* Functions used for device addresses handling */ 4339 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4340 unsigned char addr_type); 4341 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4342 unsigned char addr_type); 4343 void dev_addr_flush(struct net_device *dev); 4344 int dev_addr_init(struct net_device *dev); 4345 4346 /* Functions used for unicast addresses handling */ 4347 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4348 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4349 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4350 int dev_uc_sync(struct net_device *to, struct net_device *from); 4351 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4352 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4353 void dev_uc_flush(struct net_device *dev); 4354 void dev_uc_init(struct net_device *dev); 4355 4356 /** 4357 * __dev_uc_sync - Synchonize device's unicast list 4358 * @dev: device to sync 4359 * @sync: function to call if address should be added 4360 * @unsync: function to call if address should be removed 4361 * 4362 * Add newly added addresses to the interface, and release 4363 * addresses that have been deleted. 4364 */ 4365 static inline int __dev_uc_sync(struct net_device *dev, 4366 int (*sync)(struct net_device *, 4367 const unsigned char *), 4368 int (*unsync)(struct net_device *, 4369 const unsigned char *)) 4370 { 4371 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4372 } 4373 4374 /** 4375 * __dev_uc_unsync - Remove synchronized addresses from device 4376 * @dev: device to sync 4377 * @unsync: function to call if address should be removed 4378 * 4379 * Remove all addresses that were added to the device by dev_uc_sync(). 4380 */ 4381 static inline void __dev_uc_unsync(struct net_device *dev, 4382 int (*unsync)(struct net_device *, 4383 const unsigned char *)) 4384 { 4385 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4386 } 4387 4388 /* Functions used for multicast addresses handling */ 4389 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4390 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4391 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4392 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4393 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4394 int dev_mc_sync(struct net_device *to, struct net_device *from); 4395 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4396 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4397 void dev_mc_flush(struct net_device *dev); 4398 void dev_mc_init(struct net_device *dev); 4399 4400 /** 4401 * __dev_mc_sync - Synchonize device's multicast list 4402 * @dev: device to sync 4403 * @sync: function to call if address should be added 4404 * @unsync: function to call if address should be removed 4405 * 4406 * Add newly added addresses to the interface, and release 4407 * addresses that have been deleted. 4408 */ 4409 static inline int __dev_mc_sync(struct net_device *dev, 4410 int (*sync)(struct net_device *, 4411 const unsigned char *), 4412 int (*unsync)(struct net_device *, 4413 const unsigned char *)) 4414 { 4415 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4416 } 4417 4418 /** 4419 * __dev_mc_unsync - Remove synchronized addresses from device 4420 * @dev: device to sync 4421 * @unsync: function to call if address should be removed 4422 * 4423 * Remove all addresses that were added to the device by dev_mc_sync(). 4424 */ 4425 static inline void __dev_mc_unsync(struct net_device *dev, 4426 int (*unsync)(struct net_device *, 4427 const unsigned char *)) 4428 { 4429 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4430 } 4431 4432 /* Functions used for secondary unicast and multicast support */ 4433 void dev_set_rx_mode(struct net_device *dev); 4434 void __dev_set_rx_mode(struct net_device *dev); 4435 int dev_set_promiscuity(struct net_device *dev, int inc); 4436 int dev_set_allmulti(struct net_device *dev, int inc); 4437 void netdev_state_change(struct net_device *dev); 4438 void netdev_notify_peers(struct net_device *dev); 4439 void netdev_features_change(struct net_device *dev); 4440 /* Load a device via the kmod */ 4441 void dev_load(struct net *net, const char *name); 4442 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4443 struct rtnl_link_stats64 *storage); 4444 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4445 const struct net_device_stats *netdev_stats); 4446 4447 extern int netdev_max_backlog; 4448 extern int netdev_tstamp_prequeue; 4449 extern int weight_p; 4450 extern int dev_weight_rx_bias; 4451 extern int dev_weight_tx_bias; 4452 extern int dev_rx_weight; 4453 extern int dev_tx_weight; 4454 extern int gro_normal_batch; 4455 4456 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4457 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4458 struct list_head **iter); 4459 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4460 struct list_head **iter); 4461 4462 /* iterate through upper list, must be called under RCU read lock */ 4463 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4464 for (iter = &(dev)->adj_list.upper, \ 4465 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4466 updev; \ 4467 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4468 4469 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4470 int (*fn)(struct net_device *upper_dev, 4471 void *data), 4472 void *data); 4473 4474 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4475 struct net_device *upper_dev); 4476 4477 bool netdev_has_any_upper_dev(struct net_device *dev); 4478 4479 void *netdev_lower_get_next_private(struct net_device *dev, 4480 struct list_head **iter); 4481 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4482 struct list_head **iter); 4483 4484 #define netdev_for_each_lower_private(dev, priv, iter) \ 4485 for (iter = (dev)->adj_list.lower.next, \ 4486 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4487 priv; \ 4488 priv = netdev_lower_get_next_private(dev, &(iter))) 4489 4490 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4491 for (iter = &(dev)->adj_list.lower, \ 4492 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4493 priv; \ 4494 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4495 4496 void *netdev_lower_get_next(struct net_device *dev, 4497 struct list_head **iter); 4498 4499 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4500 for (iter = (dev)->adj_list.lower.next, \ 4501 ldev = netdev_lower_get_next(dev, &(iter)); \ 4502 ldev; \ 4503 ldev = netdev_lower_get_next(dev, &(iter))) 4504 4505 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4506 struct list_head **iter); 4507 int netdev_walk_all_lower_dev(struct net_device *dev, 4508 int (*fn)(struct net_device *lower_dev, 4509 void *data), 4510 void *data); 4511 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4512 int (*fn)(struct net_device *lower_dev, 4513 void *data), 4514 void *data); 4515 4516 void *netdev_adjacent_get_private(struct list_head *adj_list); 4517 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4518 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4519 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4520 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4521 struct netlink_ext_ack *extack); 4522 int netdev_master_upper_dev_link(struct net_device *dev, 4523 struct net_device *upper_dev, 4524 void *upper_priv, void *upper_info, 4525 struct netlink_ext_ack *extack); 4526 void netdev_upper_dev_unlink(struct net_device *dev, 4527 struct net_device *upper_dev); 4528 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4529 struct net_device *new_dev, 4530 struct net_device *dev, 4531 struct netlink_ext_ack *extack); 4532 void netdev_adjacent_change_commit(struct net_device *old_dev, 4533 struct net_device *new_dev, 4534 struct net_device *dev); 4535 void netdev_adjacent_change_abort(struct net_device *old_dev, 4536 struct net_device *new_dev, 4537 struct net_device *dev); 4538 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4539 void *netdev_lower_dev_get_private(struct net_device *dev, 4540 struct net_device *lower_dev); 4541 void netdev_lower_state_changed(struct net_device *lower_dev, 4542 void *lower_state_info); 4543 4544 /* RSS keys are 40 or 52 bytes long */ 4545 #define NETDEV_RSS_KEY_LEN 52 4546 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4547 void netdev_rss_key_fill(void *buffer, size_t len); 4548 4549 int skb_checksum_help(struct sk_buff *skb); 4550 int skb_crc32c_csum_help(struct sk_buff *skb); 4551 int skb_csum_hwoffload_help(struct sk_buff *skb, 4552 const netdev_features_t features); 4553 4554 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4555 netdev_features_t features, bool tx_path); 4556 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4557 netdev_features_t features); 4558 4559 struct netdev_bonding_info { 4560 ifslave slave; 4561 ifbond master; 4562 }; 4563 4564 struct netdev_notifier_bonding_info { 4565 struct netdev_notifier_info info; /* must be first */ 4566 struct netdev_bonding_info bonding_info; 4567 }; 4568 4569 void netdev_bonding_info_change(struct net_device *dev, 4570 struct netdev_bonding_info *bonding_info); 4571 4572 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4573 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4574 #else 4575 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4576 const void *data) 4577 { 4578 } 4579 #endif 4580 4581 static inline 4582 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4583 { 4584 return __skb_gso_segment(skb, features, true); 4585 } 4586 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4587 4588 static inline bool can_checksum_protocol(netdev_features_t features, 4589 __be16 protocol) 4590 { 4591 if (protocol == htons(ETH_P_FCOE)) 4592 return !!(features & NETIF_F_FCOE_CRC); 4593 4594 /* Assume this is an IP checksum (not SCTP CRC) */ 4595 4596 if (features & NETIF_F_HW_CSUM) { 4597 /* Can checksum everything */ 4598 return true; 4599 } 4600 4601 switch (protocol) { 4602 case htons(ETH_P_IP): 4603 return !!(features & NETIF_F_IP_CSUM); 4604 case htons(ETH_P_IPV6): 4605 return !!(features & NETIF_F_IPV6_CSUM); 4606 default: 4607 return false; 4608 } 4609 } 4610 4611 #ifdef CONFIG_BUG 4612 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4613 #else 4614 static inline void netdev_rx_csum_fault(struct net_device *dev, 4615 struct sk_buff *skb) 4616 { 4617 } 4618 #endif 4619 /* rx skb timestamps */ 4620 void net_enable_timestamp(void); 4621 void net_disable_timestamp(void); 4622 4623 #ifdef CONFIG_PROC_FS 4624 int __init dev_proc_init(void); 4625 #else 4626 #define dev_proc_init() 0 4627 #endif 4628 4629 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4630 struct sk_buff *skb, struct net_device *dev, 4631 bool more) 4632 { 4633 __this_cpu_write(softnet_data.xmit.more, more); 4634 return ops->ndo_start_xmit(skb, dev); 4635 } 4636 4637 static inline bool netdev_xmit_more(void) 4638 { 4639 return __this_cpu_read(softnet_data.xmit.more); 4640 } 4641 4642 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4643 struct netdev_queue *txq, bool more) 4644 { 4645 const struct net_device_ops *ops = dev->netdev_ops; 4646 netdev_tx_t rc; 4647 4648 rc = __netdev_start_xmit(ops, skb, dev, more); 4649 if (rc == NETDEV_TX_OK) 4650 txq_trans_update(txq); 4651 4652 return rc; 4653 } 4654 4655 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4656 const void *ns); 4657 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4658 const void *ns); 4659 4660 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4661 { 4662 return netdev_class_create_file_ns(class_attr, NULL); 4663 } 4664 4665 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4666 { 4667 netdev_class_remove_file_ns(class_attr, NULL); 4668 } 4669 4670 extern const struct kobj_ns_type_operations net_ns_type_operations; 4671 4672 const char *netdev_drivername(const struct net_device *dev); 4673 4674 void linkwatch_run_queue(void); 4675 4676 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4677 netdev_features_t f2) 4678 { 4679 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4680 if (f1 & NETIF_F_HW_CSUM) 4681 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4682 else 4683 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4684 } 4685 4686 return f1 & f2; 4687 } 4688 4689 static inline netdev_features_t netdev_get_wanted_features( 4690 struct net_device *dev) 4691 { 4692 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4693 } 4694 netdev_features_t netdev_increment_features(netdev_features_t all, 4695 netdev_features_t one, netdev_features_t mask); 4696 4697 /* Allow TSO being used on stacked device : 4698 * Performing the GSO segmentation before last device 4699 * is a performance improvement. 4700 */ 4701 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4702 netdev_features_t mask) 4703 { 4704 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4705 } 4706 4707 int __netdev_update_features(struct net_device *dev); 4708 void netdev_update_features(struct net_device *dev); 4709 void netdev_change_features(struct net_device *dev); 4710 4711 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4712 struct net_device *dev); 4713 4714 netdev_features_t passthru_features_check(struct sk_buff *skb, 4715 struct net_device *dev, 4716 netdev_features_t features); 4717 netdev_features_t netif_skb_features(struct sk_buff *skb); 4718 4719 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4720 { 4721 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4722 4723 /* check flags correspondence */ 4724 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4725 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4726 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4727 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4728 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4729 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4730 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4731 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4732 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4733 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4734 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4735 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4736 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4737 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4738 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4739 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4740 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4741 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4742 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4743 4744 return (features & feature) == feature; 4745 } 4746 4747 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4748 { 4749 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4750 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4751 } 4752 4753 static inline bool netif_needs_gso(struct sk_buff *skb, 4754 netdev_features_t features) 4755 { 4756 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4757 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4758 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4759 } 4760 4761 static inline void netif_set_gso_max_size(struct net_device *dev, 4762 unsigned int size) 4763 { 4764 dev->gso_max_size = size; 4765 } 4766 4767 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4768 int pulled_hlen, u16 mac_offset, 4769 int mac_len) 4770 { 4771 skb->protocol = protocol; 4772 skb->encapsulation = 1; 4773 skb_push(skb, pulled_hlen); 4774 skb_reset_transport_header(skb); 4775 skb->mac_header = mac_offset; 4776 skb->network_header = skb->mac_header + mac_len; 4777 skb->mac_len = mac_len; 4778 } 4779 4780 static inline bool netif_is_macsec(const struct net_device *dev) 4781 { 4782 return dev->priv_flags & IFF_MACSEC; 4783 } 4784 4785 static inline bool netif_is_macvlan(const struct net_device *dev) 4786 { 4787 return dev->priv_flags & IFF_MACVLAN; 4788 } 4789 4790 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4791 { 4792 return dev->priv_flags & IFF_MACVLAN_PORT; 4793 } 4794 4795 static inline bool netif_is_bond_master(const struct net_device *dev) 4796 { 4797 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4798 } 4799 4800 static inline bool netif_is_bond_slave(const struct net_device *dev) 4801 { 4802 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4803 } 4804 4805 static inline bool netif_supports_nofcs(struct net_device *dev) 4806 { 4807 return dev->priv_flags & IFF_SUPP_NOFCS; 4808 } 4809 4810 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4811 { 4812 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4813 } 4814 4815 static inline bool netif_is_l3_master(const struct net_device *dev) 4816 { 4817 return dev->priv_flags & IFF_L3MDEV_MASTER; 4818 } 4819 4820 static inline bool netif_is_l3_slave(const struct net_device *dev) 4821 { 4822 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4823 } 4824 4825 static inline bool netif_is_bridge_master(const struct net_device *dev) 4826 { 4827 return dev->priv_flags & IFF_EBRIDGE; 4828 } 4829 4830 static inline bool netif_is_bridge_port(const struct net_device *dev) 4831 { 4832 return dev->priv_flags & IFF_BRIDGE_PORT; 4833 } 4834 4835 static inline bool netif_is_ovs_master(const struct net_device *dev) 4836 { 4837 return dev->priv_flags & IFF_OPENVSWITCH; 4838 } 4839 4840 static inline bool netif_is_ovs_port(const struct net_device *dev) 4841 { 4842 return dev->priv_flags & IFF_OVS_DATAPATH; 4843 } 4844 4845 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4846 { 4847 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4848 } 4849 4850 static inline bool netif_is_team_master(const struct net_device *dev) 4851 { 4852 return dev->priv_flags & IFF_TEAM; 4853 } 4854 4855 static inline bool netif_is_team_port(const struct net_device *dev) 4856 { 4857 return dev->priv_flags & IFF_TEAM_PORT; 4858 } 4859 4860 static inline bool netif_is_lag_master(const struct net_device *dev) 4861 { 4862 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4863 } 4864 4865 static inline bool netif_is_lag_port(const struct net_device *dev) 4866 { 4867 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4868 } 4869 4870 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4871 { 4872 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4873 } 4874 4875 static inline bool netif_is_failover(const struct net_device *dev) 4876 { 4877 return dev->priv_flags & IFF_FAILOVER; 4878 } 4879 4880 static inline bool netif_is_failover_slave(const struct net_device *dev) 4881 { 4882 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4883 } 4884 4885 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4886 static inline void netif_keep_dst(struct net_device *dev) 4887 { 4888 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4889 } 4890 4891 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4892 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4893 { 4894 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4895 return dev->priv_flags & IFF_MACSEC; 4896 } 4897 4898 extern struct pernet_operations __net_initdata loopback_net_ops; 4899 4900 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4901 4902 /* netdev_printk helpers, similar to dev_printk */ 4903 4904 static inline const char *netdev_name(const struct net_device *dev) 4905 { 4906 if (!dev->name[0] || strchr(dev->name, '%')) 4907 return "(unnamed net_device)"; 4908 return dev->name; 4909 } 4910 4911 static inline bool netdev_unregistering(const struct net_device *dev) 4912 { 4913 return dev->reg_state == NETREG_UNREGISTERING; 4914 } 4915 4916 static inline const char *netdev_reg_state(const struct net_device *dev) 4917 { 4918 switch (dev->reg_state) { 4919 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4920 case NETREG_REGISTERED: return ""; 4921 case NETREG_UNREGISTERING: return " (unregistering)"; 4922 case NETREG_UNREGISTERED: return " (unregistered)"; 4923 case NETREG_RELEASED: return " (released)"; 4924 case NETREG_DUMMY: return " (dummy)"; 4925 } 4926 4927 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4928 return " (unknown)"; 4929 } 4930 4931 __printf(3, 4) __cold 4932 void netdev_printk(const char *level, const struct net_device *dev, 4933 const char *format, ...); 4934 __printf(2, 3) __cold 4935 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4936 __printf(2, 3) __cold 4937 void netdev_alert(const struct net_device *dev, const char *format, ...); 4938 __printf(2, 3) __cold 4939 void netdev_crit(const struct net_device *dev, const char *format, ...); 4940 __printf(2, 3) __cold 4941 void netdev_err(const struct net_device *dev, const char *format, ...); 4942 __printf(2, 3) __cold 4943 void netdev_warn(const struct net_device *dev, const char *format, ...); 4944 __printf(2, 3) __cold 4945 void netdev_notice(const struct net_device *dev, const char *format, ...); 4946 __printf(2, 3) __cold 4947 void netdev_info(const struct net_device *dev, const char *format, ...); 4948 4949 #define netdev_level_once(level, dev, fmt, ...) \ 4950 do { \ 4951 static bool __print_once __read_mostly; \ 4952 \ 4953 if (!__print_once) { \ 4954 __print_once = true; \ 4955 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4956 } \ 4957 } while (0) 4958 4959 #define netdev_emerg_once(dev, fmt, ...) \ 4960 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4961 #define netdev_alert_once(dev, fmt, ...) \ 4962 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4963 #define netdev_crit_once(dev, fmt, ...) \ 4964 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4965 #define netdev_err_once(dev, fmt, ...) \ 4966 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4967 #define netdev_warn_once(dev, fmt, ...) \ 4968 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4969 #define netdev_notice_once(dev, fmt, ...) \ 4970 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4971 #define netdev_info_once(dev, fmt, ...) \ 4972 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4973 4974 #define MODULE_ALIAS_NETDEV(device) \ 4975 MODULE_ALIAS("netdev-" device) 4976 4977 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 4978 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 4979 #define netdev_dbg(__dev, format, args...) \ 4980 do { \ 4981 dynamic_netdev_dbg(__dev, format, ##args); \ 4982 } while (0) 4983 #elif defined(DEBUG) 4984 #define netdev_dbg(__dev, format, args...) \ 4985 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4986 #else 4987 #define netdev_dbg(__dev, format, args...) \ 4988 ({ \ 4989 if (0) \ 4990 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4991 }) 4992 #endif 4993 4994 #if defined(VERBOSE_DEBUG) 4995 #define netdev_vdbg netdev_dbg 4996 #else 4997 4998 #define netdev_vdbg(dev, format, args...) \ 4999 ({ \ 5000 if (0) \ 5001 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5002 0; \ 5003 }) 5004 #endif 5005 5006 /* 5007 * netdev_WARN() acts like dev_printk(), but with the key difference 5008 * of using a WARN/WARN_ON to get the message out, including the 5009 * file/line information and a backtrace. 5010 */ 5011 #define netdev_WARN(dev, format, args...) \ 5012 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5013 netdev_reg_state(dev), ##args) 5014 5015 #define netdev_WARN_ONCE(dev, format, args...) \ 5016 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5017 netdev_reg_state(dev), ##args) 5018 5019 /* netif printk helpers, similar to netdev_printk */ 5020 5021 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5022 do { \ 5023 if (netif_msg_##type(priv)) \ 5024 netdev_printk(level, (dev), fmt, ##args); \ 5025 } while (0) 5026 5027 #define netif_level(level, priv, type, dev, fmt, args...) \ 5028 do { \ 5029 if (netif_msg_##type(priv)) \ 5030 netdev_##level(dev, fmt, ##args); \ 5031 } while (0) 5032 5033 #define netif_emerg(priv, type, dev, fmt, args...) \ 5034 netif_level(emerg, priv, type, dev, fmt, ##args) 5035 #define netif_alert(priv, type, dev, fmt, args...) \ 5036 netif_level(alert, priv, type, dev, fmt, ##args) 5037 #define netif_crit(priv, type, dev, fmt, args...) \ 5038 netif_level(crit, priv, type, dev, fmt, ##args) 5039 #define netif_err(priv, type, dev, fmt, args...) \ 5040 netif_level(err, priv, type, dev, fmt, ##args) 5041 #define netif_warn(priv, type, dev, fmt, args...) \ 5042 netif_level(warn, priv, type, dev, fmt, ##args) 5043 #define netif_notice(priv, type, dev, fmt, args...) \ 5044 netif_level(notice, priv, type, dev, fmt, ##args) 5045 #define netif_info(priv, type, dev, fmt, args...) \ 5046 netif_level(info, priv, type, dev, fmt, ##args) 5047 5048 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5049 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5050 #define netif_dbg(priv, type, netdev, format, args...) \ 5051 do { \ 5052 if (netif_msg_##type(priv)) \ 5053 dynamic_netdev_dbg(netdev, format, ##args); \ 5054 } while (0) 5055 #elif defined(DEBUG) 5056 #define netif_dbg(priv, type, dev, format, args...) \ 5057 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5058 #else 5059 #define netif_dbg(priv, type, dev, format, args...) \ 5060 ({ \ 5061 if (0) \ 5062 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5063 0; \ 5064 }) 5065 #endif 5066 5067 /* if @cond then downgrade to debug, else print at @level */ 5068 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5069 do { \ 5070 if (cond) \ 5071 netif_dbg(priv, type, netdev, fmt, ##args); \ 5072 else \ 5073 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5074 } while (0) 5075 5076 #if defined(VERBOSE_DEBUG) 5077 #define netif_vdbg netif_dbg 5078 #else 5079 #define netif_vdbg(priv, type, dev, format, args...) \ 5080 ({ \ 5081 if (0) \ 5082 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5083 0; \ 5084 }) 5085 #endif 5086 5087 /* 5088 * The list of packet types we will receive (as opposed to discard) 5089 * and the routines to invoke. 5090 * 5091 * Why 16. Because with 16 the only overlap we get on a hash of the 5092 * low nibble of the protocol value is RARP/SNAP/X.25. 5093 * 5094 * 0800 IP 5095 * 0001 802.3 5096 * 0002 AX.25 5097 * 0004 802.2 5098 * 8035 RARP 5099 * 0005 SNAP 5100 * 0805 X.25 5101 * 0806 ARP 5102 * 8137 IPX 5103 * 0009 Localtalk 5104 * 86DD IPv6 5105 */ 5106 #define PTYPE_HASH_SIZE (16) 5107 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5108 5109 extern struct net_device *blackhole_netdev; 5110 5111 #endif /* _LINUX_NETDEVICE_H */ 5112