1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <net/net_namespace.h> 38 #ifdef CONFIG_DCB 39 #include <net/dcbnl.h> 40 #endif 41 #include <net/netprio_cgroup.h> 42 #include <net/xdp.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <linux/hashtable.h> 50 51 struct netpoll_info; 52 struct device; 53 struct ethtool_ops; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void synchronize_net(void); 74 void netdev_set_default_ethtool_ops(struct net_device *dev, 75 const struct ethtool_ops *ops); 76 77 /* Backlog congestion levels */ 78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 79 #define NET_RX_DROP 1 /* packet dropped */ 80 81 #define MAX_NEST_DEV 8 82 83 /* 84 * Transmit return codes: transmit return codes originate from three different 85 * namespaces: 86 * 87 * - qdisc return codes 88 * - driver transmit return codes 89 * - errno values 90 * 91 * Drivers are allowed to return any one of those in their hard_start_xmit() 92 * function. Real network devices commonly used with qdiscs should only return 93 * the driver transmit return codes though - when qdiscs are used, the actual 94 * transmission happens asynchronously, so the value is not propagated to 95 * higher layers. Virtual network devices transmit synchronously; in this case 96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 97 * others are propagated to higher layers. 98 */ 99 100 /* qdisc ->enqueue() return codes. */ 101 #define NET_XMIT_SUCCESS 0x00 102 #define NET_XMIT_DROP 0x01 /* skb dropped */ 103 #define NET_XMIT_CN 0x02 /* congestion notification */ 104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 105 106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 107 * indicates that the device will soon be dropping packets, or already drops 108 * some packets of the same priority; prompting us to send less aggressively. */ 109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 111 112 /* Driver transmit return codes */ 113 #define NETDEV_TX_MASK 0xf0 114 115 enum netdev_tx { 116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 117 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 119 }; 120 typedef enum netdev_tx netdev_tx_t; 121 122 /* 123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 125 */ 126 static inline bool dev_xmit_complete(int rc) 127 { 128 /* 129 * Positive cases with an skb consumed by a driver: 130 * - successful transmission (rc == NETDEV_TX_OK) 131 * - error while transmitting (rc < 0) 132 * - error while queueing to a different device (rc & NET_XMIT_MASK) 133 */ 134 if (likely(rc < NET_XMIT_MASK)) 135 return true; 136 137 return false; 138 } 139 140 /* 141 * Compute the worst-case header length according to the protocols 142 * used. 143 */ 144 145 #if defined(CONFIG_HYPERV_NET) 146 # define LL_MAX_HEADER 128 147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 148 # if defined(CONFIG_MAC80211_MESH) 149 # define LL_MAX_HEADER 128 150 # else 151 # define LL_MAX_HEADER 96 152 # endif 153 #else 154 # define LL_MAX_HEADER 32 155 #endif 156 157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 159 #define MAX_HEADER LL_MAX_HEADER 160 #else 161 #define MAX_HEADER (LL_MAX_HEADER + 48) 162 #endif 163 164 /* 165 * Old network device statistics. Fields are native words 166 * (unsigned long) so they can be read and written atomically. 167 */ 168 169 struct net_device_stats { 170 unsigned long rx_packets; 171 unsigned long tx_packets; 172 unsigned long rx_bytes; 173 unsigned long tx_bytes; 174 unsigned long rx_errors; 175 unsigned long tx_errors; 176 unsigned long rx_dropped; 177 unsigned long tx_dropped; 178 unsigned long multicast; 179 unsigned long collisions; 180 unsigned long rx_length_errors; 181 unsigned long rx_over_errors; 182 unsigned long rx_crc_errors; 183 unsigned long rx_frame_errors; 184 unsigned long rx_fifo_errors; 185 unsigned long rx_missed_errors; 186 unsigned long tx_aborted_errors; 187 unsigned long tx_carrier_errors; 188 unsigned long tx_fifo_errors; 189 unsigned long tx_heartbeat_errors; 190 unsigned long tx_window_errors; 191 unsigned long rx_compressed; 192 unsigned long tx_compressed; 193 }; 194 195 196 #include <linux/cache.h> 197 #include <linux/skbuff.h> 198 199 #ifdef CONFIG_RPS 200 #include <linux/static_key.h> 201 extern struct static_key_false rps_needed; 202 extern struct static_key_false rfs_needed; 203 #endif 204 205 struct neighbour; 206 struct neigh_parms; 207 struct sk_buff; 208 209 struct netdev_hw_addr { 210 struct list_head list; 211 unsigned char addr[MAX_ADDR_LEN]; 212 unsigned char type; 213 #define NETDEV_HW_ADDR_T_LAN 1 214 #define NETDEV_HW_ADDR_T_SAN 2 215 #define NETDEV_HW_ADDR_T_UNICAST 3 216 #define NETDEV_HW_ADDR_T_MULTICAST 4 217 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_LISTED, /* NAPI added to system lists */ 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 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 361 }; 362 363 enum { 364 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 365 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 366 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 367 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 368 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 369 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 370 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 371 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 372 }; 373 374 enum gro_result { 375 GRO_MERGED, 376 GRO_MERGED_FREE, 377 GRO_HELD, 378 GRO_NORMAL, 379 GRO_DROP, 380 GRO_CONSUMED, 381 }; 382 typedef enum gro_result gro_result_t; 383 384 /* 385 * enum rx_handler_result - Possible return values for rx_handlers. 386 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 387 * further. 388 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 389 * case skb->dev was changed by rx_handler. 390 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 391 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 392 * 393 * rx_handlers are functions called from inside __netif_receive_skb(), to do 394 * special processing of the skb, prior to delivery to protocol handlers. 395 * 396 * Currently, a net_device can only have a single rx_handler registered. Trying 397 * to register a second rx_handler will return -EBUSY. 398 * 399 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 400 * To unregister a rx_handler on a net_device, use 401 * netdev_rx_handler_unregister(). 402 * 403 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 404 * do with the skb. 405 * 406 * If the rx_handler consumed the skb in some way, it should return 407 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 408 * the skb to be delivered in some other way. 409 * 410 * If the rx_handler changed skb->dev, to divert the skb to another 411 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 412 * new device will be called if it exists. 413 * 414 * If the rx_handler decides the skb should be ignored, it should return 415 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 416 * are registered on exact device (ptype->dev == skb->dev). 417 * 418 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 419 * delivered, it should return RX_HANDLER_PASS. 420 * 421 * A device without a registered rx_handler will behave as if rx_handler 422 * returned RX_HANDLER_PASS. 423 */ 424 425 enum rx_handler_result { 426 RX_HANDLER_CONSUMED, 427 RX_HANDLER_ANOTHER, 428 RX_HANDLER_EXACT, 429 RX_HANDLER_PASS, 430 }; 431 typedef enum rx_handler_result rx_handler_result_t; 432 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 433 434 void __napi_schedule(struct napi_struct *n); 435 void __napi_schedule_irqoff(struct napi_struct *n); 436 437 static inline bool napi_disable_pending(struct napi_struct *n) 438 { 439 return test_bit(NAPI_STATE_DISABLE, &n->state); 440 } 441 442 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 443 { 444 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 445 } 446 447 bool napi_schedule_prep(struct napi_struct *n); 448 449 /** 450 * napi_schedule - schedule NAPI poll 451 * @n: NAPI context 452 * 453 * Schedule NAPI poll routine to be called if it is not already 454 * running. 455 */ 456 static inline void napi_schedule(struct napi_struct *n) 457 { 458 if (napi_schedule_prep(n)) 459 __napi_schedule(n); 460 } 461 462 /** 463 * napi_schedule_irqoff - schedule NAPI poll 464 * @n: NAPI context 465 * 466 * Variant of napi_schedule(), assuming hard irqs are masked. 467 */ 468 static inline void napi_schedule_irqoff(struct napi_struct *n) 469 { 470 if (napi_schedule_prep(n)) 471 __napi_schedule_irqoff(n); 472 } 473 474 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 475 static inline bool napi_reschedule(struct napi_struct *napi) 476 { 477 if (napi_schedule_prep(napi)) { 478 __napi_schedule(napi); 479 return true; 480 } 481 return false; 482 } 483 484 bool napi_complete_done(struct napi_struct *n, int work_done); 485 /** 486 * napi_complete - NAPI processing complete 487 * @n: NAPI context 488 * 489 * Mark NAPI processing as complete. 490 * Consider using napi_complete_done() instead. 491 * Return false if device should avoid rearming interrupts. 492 */ 493 static inline bool napi_complete(struct napi_struct *n) 494 { 495 return napi_complete_done(n, 0); 496 } 497 498 /** 499 * napi_disable - prevent NAPI from scheduling 500 * @n: NAPI context 501 * 502 * Stop NAPI from being scheduled on this context. 503 * Waits till any outstanding processing completes. 504 */ 505 void napi_disable(struct napi_struct *n); 506 507 /** 508 * napi_enable - enable NAPI scheduling 509 * @n: NAPI context 510 * 511 * Resume NAPI from being scheduled on this context. 512 * Must be paired with napi_disable. 513 */ 514 static inline void napi_enable(struct napi_struct *n) 515 { 516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 517 smp_mb__before_atomic(); 518 clear_bit(NAPI_STATE_SCHED, &n->state); 519 clear_bit(NAPI_STATE_NPSVC, &n->state); 520 } 521 522 /** 523 * napi_synchronize - wait until NAPI is not running 524 * @n: NAPI context 525 * 526 * Wait until NAPI is done being scheduled on this context. 527 * Waits till any outstanding processing completes but 528 * does not disable future activations. 529 */ 530 static inline void napi_synchronize(const struct napi_struct *n) 531 { 532 if (IS_ENABLED(CONFIG_SMP)) 533 while (test_bit(NAPI_STATE_SCHED, &n->state)) 534 msleep(1); 535 else 536 barrier(); 537 } 538 539 /** 540 * napi_if_scheduled_mark_missed - if napi is running, set the 541 * NAPIF_STATE_MISSED 542 * @n: NAPI context 543 * 544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 545 * NAPI is scheduled. 546 **/ 547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 548 { 549 unsigned long val, new; 550 551 do { 552 val = READ_ONCE(n->state); 553 if (val & NAPIF_STATE_DISABLE) 554 return true; 555 556 if (!(val & NAPIF_STATE_SCHED)) 557 return false; 558 559 new = val | NAPIF_STATE_MISSED; 560 } while (cmpxchg(&n->state, val, new) != val); 561 562 return true; 563 } 564 565 enum netdev_queue_state_t { 566 __QUEUE_STATE_DRV_XOFF, 567 __QUEUE_STATE_STACK_XOFF, 568 __QUEUE_STATE_FROZEN, 569 }; 570 571 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 572 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 573 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 574 575 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 579 QUEUE_STATE_FROZEN) 580 581 /* 582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 583 * netif_tx_* functions below are used to manipulate this flag. The 584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 585 * queue independently. The netif_xmit_*stopped functions below are called 586 * to check if the queue has been stopped by the driver or stack (either 587 * of the XOFF bits are set in the state). Drivers should not need to call 588 * netif_xmit*stopped functions, they should only be using netif_tx_*. 589 */ 590 591 struct netdev_queue { 592 /* 593 * read-mostly part 594 */ 595 struct net_device *dev; 596 struct Qdisc __rcu *qdisc; 597 struct Qdisc *qdisc_sleeping; 598 #ifdef CONFIG_SYSFS 599 struct kobject kobj; 600 #endif 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 602 int numa_node; 603 #endif 604 unsigned long tx_maxrate; 605 /* 606 * Number of TX timeouts for this queue 607 * (/sys/class/net/DEV/Q/trans_timeout) 608 */ 609 unsigned long trans_timeout; 610 611 /* Subordinate device that the queue has been assigned to */ 612 struct net_device *sb_dev; 613 #ifdef CONFIG_XDP_SOCKETS 614 struct xsk_buff_pool *pool; 615 #endif 616 /* 617 * write-mostly part 618 */ 619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 620 int xmit_lock_owner; 621 /* 622 * Time (in jiffies) of last Tx 623 */ 624 unsigned long trans_start; 625 626 unsigned long state; 627 628 #ifdef CONFIG_BQL 629 struct dql dql; 630 #endif 631 } ____cacheline_aligned_in_smp; 632 633 extern int sysctl_fb_tunnels_only_for_init_net; 634 extern int sysctl_devconf_inherit_init_net; 635 636 /* 637 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 638 * == 1 : For initns only 639 * == 2 : For none. 640 */ 641 static inline bool net_has_fallback_tunnels(const struct net *net) 642 { 643 return !IS_ENABLED(CONFIG_SYSCTL) || 644 !sysctl_fb_tunnels_only_for_init_net || 645 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 646 } 647 648 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 649 { 650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 651 return q->numa_node; 652 #else 653 return NUMA_NO_NODE; 654 #endif 655 } 656 657 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 658 { 659 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 660 q->numa_node = node; 661 #endif 662 } 663 664 #ifdef CONFIG_RPS 665 /* 666 * This structure holds an RPS map which can be of variable length. The 667 * map is an array of CPUs. 668 */ 669 struct rps_map { 670 unsigned int len; 671 struct rcu_head rcu; 672 u16 cpus[]; 673 }; 674 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 675 676 /* 677 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 678 * tail pointer for that CPU's input queue at the time of last enqueue, and 679 * a hardware filter index. 680 */ 681 struct rps_dev_flow { 682 u16 cpu; 683 u16 filter; 684 unsigned int last_qtail; 685 }; 686 #define RPS_NO_FILTER 0xffff 687 688 /* 689 * The rps_dev_flow_table structure contains a table of flow mappings. 690 */ 691 struct rps_dev_flow_table { 692 unsigned int mask; 693 struct rcu_head rcu; 694 struct rps_dev_flow flows[]; 695 }; 696 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 697 ((_num) * sizeof(struct rps_dev_flow))) 698 699 /* 700 * The rps_sock_flow_table contains mappings of flows to the last CPU 701 * on which they were processed by the application (set in recvmsg). 702 * Each entry is a 32bit value. Upper part is the high-order bits 703 * of flow hash, lower part is CPU number. 704 * rps_cpu_mask is used to partition the space, depending on number of 705 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 706 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 707 * meaning we use 32-6=26 bits for the hash. 708 */ 709 struct rps_sock_flow_table { 710 u32 mask; 711 712 u32 ents[] ____cacheline_aligned_in_smp; 713 }; 714 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 715 716 #define RPS_NO_CPU 0xffff 717 718 extern u32 rps_cpu_mask; 719 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 720 721 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 722 u32 hash) 723 { 724 if (table && hash) { 725 unsigned int index = hash & table->mask; 726 u32 val = hash & ~rps_cpu_mask; 727 728 /* We only give a hint, preemption can change CPU under us */ 729 val |= raw_smp_processor_id(); 730 731 if (table->ents[index] != val) 732 table->ents[index] = val; 733 } 734 } 735 736 #ifdef CONFIG_RFS_ACCEL 737 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 738 u16 filter_id); 739 #endif 740 #endif /* CONFIG_RPS */ 741 742 /* This structure contains an instance of an RX queue. */ 743 struct netdev_rx_queue { 744 #ifdef CONFIG_RPS 745 struct rps_map __rcu *rps_map; 746 struct rps_dev_flow_table __rcu *rps_flow_table; 747 #endif 748 struct kobject kobj; 749 struct net_device *dev; 750 struct xdp_rxq_info xdp_rxq; 751 #ifdef CONFIG_XDP_SOCKETS 752 struct xsk_buff_pool *pool; 753 #endif 754 } ____cacheline_aligned_in_smp; 755 756 /* 757 * RX queue sysfs structures and functions. 758 */ 759 struct rx_queue_attribute { 760 struct attribute attr; 761 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 762 ssize_t (*store)(struct netdev_rx_queue *queue, 763 const char *buf, size_t len); 764 }; 765 766 #ifdef CONFIG_XPS 767 /* 768 * This structure holds an XPS map which can be of variable length. The 769 * map is an array of queues. 770 */ 771 struct xps_map { 772 unsigned int len; 773 unsigned int alloc_len; 774 struct rcu_head rcu; 775 u16 queues[]; 776 }; 777 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 778 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 779 - sizeof(struct xps_map)) / sizeof(u16)) 780 781 /* 782 * This structure holds all XPS maps for device. Maps are indexed by CPU. 783 */ 784 struct xps_dev_maps { 785 struct rcu_head rcu; 786 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 787 }; 788 789 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 790 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 791 792 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 793 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 794 795 #endif /* CONFIG_XPS */ 796 797 #define TC_MAX_QUEUE 16 798 #define TC_BITMASK 15 799 /* HW offloaded queuing disciplines txq count and offset maps */ 800 struct netdev_tc_txq { 801 u16 count; 802 u16 offset; 803 }; 804 805 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 806 /* 807 * This structure is to hold information about the device 808 * configured to run FCoE protocol stack. 809 */ 810 struct netdev_fcoe_hbainfo { 811 char manufacturer[64]; 812 char serial_number[64]; 813 char hardware_version[64]; 814 char driver_version[64]; 815 char optionrom_version[64]; 816 char firmware_version[64]; 817 char model[256]; 818 char model_description[256]; 819 }; 820 #endif 821 822 #define MAX_PHYS_ITEM_ID_LEN 32 823 824 /* This structure holds a unique identifier to identify some 825 * physical item (port for example) used by a netdevice. 826 */ 827 struct netdev_phys_item_id { 828 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 829 unsigned char id_len; 830 }; 831 832 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 833 struct netdev_phys_item_id *b) 834 { 835 return a->id_len == b->id_len && 836 memcmp(a->id, b->id, a->id_len) == 0; 837 } 838 839 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 840 struct sk_buff *skb, 841 struct net_device *sb_dev); 842 843 enum tc_setup_type { 844 TC_SETUP_QDISC_MQPRIO, 845 TC_SETUP_CLSU32, 846 TC_SETUP_CLSFLOWER, 847 TC_SETUP_CLSMATCHALL, 848 TC_SETUP_CLSBPF, 849 TC_SETUP_BLOCK, 850 TC_SETUP_QDISC_CBS, 851 TC_SETUP_QDISC_RED, 852 TC_SETUP_QDISC_PRIO, 853 TC_SETUP_QDISC_MQ, 854 TC_SETUP_QDISC_ETF, 855 TC_SETUP_ROOT_QDISC, 856 TC_SETUP_QDISC_GRED, 857 TC_SETUP_QDISC_TAPRIO, 858 TC_SETUP_FT, 859 TC_SETUP_QDISC_ETS, 860 TC_SETUP_QDISC_TBF, 861 TC_SETUP_QDISC_FIFO, 862 }; 863 864 /* These structures hold the attributes of bpf state that are being passed 865 * to the netdevice through the bpf op. 866 */ 867 enum bpf_netdev_command { 868 /* Set or clear a bpf program used in the earliest stages of packet 869 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 870 * is responsible for calling bpf_prog_put on any old progs that are 871 * stored. In case of error, the callee need not release the new prog 872 * reference, but on success it takes ownership and must bpf_prog_put 873 * when it is no longer used. 874 */ 875 XDP_SETUP_PROG, 876 XDP_SETUP_PROG_HW, 877 /* BPF program for offload callbacks, invoked at program load time. */ 878 BPF_OFFLOAD_MAP_ALLOC, 879 BPF_OFFLOAD_MAP_FREE, 880 XDP_SETUP_XSK_POOL, 881 }; 882 883 struct bpf_prog_offload_ops; 884 struct netlink_ext_ack; 885 struct xdp_umem; 886 struct xdp_dev_bulk_queue; 887 struct bpf_xdp_link; 888 889 enum bpf_xdp_mode { 890 XDP_MODE_SKB = 0, 891 XDP_MODE_DRV = 1, 892 XDP_MODE_HW = 2, 893 __MAX_XDP_MODE 894 }; 895 896 struct bpf_xdp_entity { 897 struct bpf_prog *prog; 898 struct bpf_xdp_link *link; 899 }; 900 901 struct netdev_bpf { 902 enum bpf_netdev_command command; 903 union { 904 /* XDP_SETUP_PROG */ 905 struct { 906 u32 flags; 907 struct bpf_prog *prog; 908 struct netlink_ext_ack *extack; 909 }; 910 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 911 struct { 912 struct bpf_offloaded_map *offmap; 913 }; 914 /* XDP_SETUP_XSK_POOL */ 915 struct { 916 struct xsk_buff_pool *pool; 917 u16 queue_id; 918 } xsk; 919 }; 920 }; 921 922 /* Flags for ndo_xsk_wakeup. */ 923 #define XDP_WAKEUP_RX (1 << 0) 924 #define XDP_WAKEUP_TX (1 << 1) 925 926 #ifdef CONFIG_XFRM_OFFLOAD 927 struct xfrmdev_ops { 928 int (*xdo_dev_state_add) (struct xfrm_state *x); 929 void (*xdo_dev_state_delete) (struct xfrm_state *x); 930 void (*xdo_dev_state_free) (struct xfrm_state *x); 931 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 932 struct xfrm_state *x); 933 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 934 }; 935 #endif 936 937 struct dev_ifalias { 938 struct rcu_head rcuhead; 939 char ifalias[]; 940 }; 941 942 struct devlink; 943 struct tlsdev_ops; 944 945 struct netdev_name_node { 946 struct hlist_node hlist; 947 struct list_head list; 948 struct net_device *dev; 949 const char *name; 950 }; 951 952 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 953 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 954 955 struct netdev_net_notifier { 956 struct list_head list; 957 struct notifier_block *nb; 958 }; 959 960 /* 961 * This structure defines the management hooks for network devices. 962 * The following hooks can be defined; unless noted otherwise, they are 963 * optional and can be filled with a null pointer. 964 * 965 * int (*ndo_init)(struct net_device *dev); 966 * This function is called once when a network device is registered. 967 * The network device can use this for any late stage initialization 968 * or semantic validation. It can fail with an error code which will 969 * be propagated back to register_netdev. 970 * 971 * void (*ndo_uninit)(struct net_device *dev); 972 * This function is called when device is unregistered or when registration 973 * fails. It is not called if init fails. 974 * 975 * int (*ndo_open)(struct net_device *dev); 976 * This function is called when a network device transitions to the up 977 * state. 978 * 979 * int (*ndo_stop)(struct net_device *dev); 980 * This function is called when a network device transitions to the down 981 * state. 982 * 983 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 984 * struct net_device *dev); 985 * Called when a packet needs to be transmitted. 986 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 987 * the queue before that can happen; it's for obsolete devices and weird 988 * corner cases, but the stack really does a non-trivial amount 989 * of useless work if you return NETDEV_TX_BUSY. 990 * Required; cannot be NULL. 991 * 992 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 993 * struct net_device *dev 994 * netdev_features_t features); 995 * Called by core transmit path to determine if device is capable of 996 * performing offload operations on a given packet. This is to give 997 * the device an opportunity to implement any restrictions that cannot 998 * be otherwise expressed by feature flags. The check is called with 999 * the set of features that the stack has calculated and it returns 1000 * those the driver believes to be appropriate. 1001 * 1002 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1003 * struct net_device *sb_dev); 1004 * Called to decide which queue to use when device supports multiple 1005 * transmit queues. 1006 * 1007 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1008 * This function is called to allow device receiver to make 1009 * changes to configuration when multicast or promiscuous is enabled. 1010 * 1011 * void (*ndo_set_rx_mode)(struct net_device *dev); 1012 * This function is called device changes address list filtering. 1013 * If driver handles unicast address filtering, it should set 1014 * IFF_UNICAST_FLT in its priv_flags. 1015 * 1016 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1017 * This function is called when the Media Access Control address 1018 * needs to be changed. If this interface is not defined, the 1019 * MAC address can not be changed. 1020 * 1021 * int (*ndo_validate_addr)(struct net_device *dev); 1022 * Test if Media Access Control address is valid for the device. 1023 * 1024 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1025 * Called when a user requests an ioctl which can't be handled by 1026 * the generic interface code. If not defined ioctls return 1027 * not supported error code. 1028 * 1029 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1030 * Used to set network devices bus interface parameters. This interface 1031 * is retained for legacy reasons; new devices should use the bus 1032 * interface (PCI) for low level management. 1033 * 1034 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1035 * Called when a user wants to change the Maximum Transfer Unit 1036 * of a device. 1037 * 1038 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1039 * Callback used when the transmitter has not made any progress 1040 * for dev->watchdog ticks. 1041 * 1042 * void (*ndo_get_stats64)(struct net_device *dev, 1043 * struct rtnl_link_stats64 *storage); 1044 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1045 * Called when a user wants to get the network device usage 1046 * statistics. Drivers must do one of the following: 1047 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1048 * rtnl_link_stats64 structure passed by the caller. 1049 * 2. Define @ndo_get_stats to update a net_device_stats structure 1050 * (which should normally be dev->stats) and return a pointer to 1051 * it. The structure may be changed asynchronously only if each 1052 * field is written atomically. 1053 * 3. Update dev->stats asynchronously and atomically, and define 1054 * neither operation. 1055 * 1056 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1057 * Return true if this device supports offload stats of this attr_id. 1058 * 1059 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1060 * void *attr_data) 1061 * Get statistics for offload operations by attr_id. Write it into the 1062 * attr_data pointer. 1063 * 1064 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1065 * If device supports VLAN filtering this function is called when a 1066 * VLAN id is registered. 1067 * 1068 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1069 * If device supports VLAN filtering this function is called when a 1070 * VLAN id is unregistered. 1071 * 1072 * void (*ndo_poll_controller)(struct net_device *dev); 1073 * 1074 * SR-IOV management functions. 1075 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1076 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1077 * u8 qos, __be16 proto); 1078 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1079 * int max_tx_rate); 1080 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1081 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1082 * int (*ndo_get_vf_config)(struct net_device *dev, 1083 * int vf, struct ifla_vf_info *ivf); 1084 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1085 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1086 * struct nlattr *port[]); 1087 * 1088 * Enable or disable the VF ability to query its RSS Redirection Table and 1089 * Hash Key. This is needed since on some devices VF share this information 1090 * with PF and querying it may introduce a theoretical security risk. 1091 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1092 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1093 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1094 * void *type_data); 1095 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1096 * This is always called from the stack with the rtnl lock held and netif 1097 * tx queues stopped. This allows the netdevice to perform queue 1098 * management safely. 1099 * 1100 * Fiber Channel over Ethernet (FCoE) offload functions. 1101 * int (*ndo_fcoe_enable)(struct net_device *dev); 1102 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1103 * so the underlying device can perform whatever needed configuration or 1104 * initialization to support acceleration of FCoE traffic. 1105 * 1106 * int (*ndo_fcoe_disable)(struct net_device *dev); 1107 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1108 * so the underlying device can perform whatever needed clean-ups to 1109 * stop supporting acceleration of FCoE traffic. 1110 * 1111 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1112 * struct scatterlist *sgl, unsigned int sgc); 1113 * Called when the FCoE Initiator wants to initialize an I/O that 1114 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1115 * perform necessary setup and returns 1 to indicate the device is set up 1116 * successfully to perform DDP on this I/O, otherwise this returns 0. 1117 * 1118 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1119 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1120 * indicated by the FC exchange id 'xid', so the underlying device can 1121 * clean up and reuse resources for later DDP requests. 1122 * 1123 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1124 * struct scatterlist *sgl, unsigned int sgc); 1125 * Called when the FCoE Target wants to initialize an I/O that 1126 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1127 * perform necessary setup and returns 1 to indicate the device is set up 1128 * successfully to perform DDP on this I/O, otherwise this returns 0. 1129 * 1130 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1131 * struct netdev_fcoe_hbainfo *hbainfo); 1132 * Called when the FCoE Protocol stack wants information on the underlying 1133 * device. This information is utilized by the FCoE protocol stack to 1134 * register attributes with Fiber Channel management service as per the 1135 * FC-GS Fabric Device Management Information(FDMI) specification. 1136 * 1137 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1138 * Called when the underlying device wants to override default World Wide 1139 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1140 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1141 * protocol stack to use. 1142 * 1143 * RFS acceleration. 1144 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1145 * u16 rxq_index, u32 flow_id); 1146 * Set hardware filter for RFS. rxq_index is the target queue index; 1147 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1148 * Return the filter ID on success, or a negative error code. 1149 * 1150 * Slave management functions (for bridge, bonding, etc). 1151 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1152 * Called to make another netdev an underling. 1153 * 1154 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1155 * Called to release previously enslaved netdev. 1156 * 1157 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1158 * struct sk_buff *skb, 1159 * bool all_slaves); 1160 * Get the xmit slave of master device. If all_slaves is true, function 1161 * assume all the slaves can transmit. 1162 * 1163 * Feature/offload setting functions. 1164 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1165 * netdev_features_t features); 1166 * Adjusts the requested feature flags according to device-specific 1167 * constraints, and returns the resulting flags. Must not modify 1168 * the device state. 1169 * 1170 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1171 * Called to update device configuration to new features. Passed 1172 * feature set might be less than what was returned by ndo_fix_features()). 1173 * Must return >0 or -errno if it changed dev->features itself. 1174 * 1175 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1176 * struct net_device *dev, 1177 * const unsigned char *addr, u16 vid, u16 flags, 1178 * struct netlink_ext_ack *extack); 1179 * Adds an FDB entry to dev for addr. 1180 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1181 * struct net_device *dev, 1182 * const unsigned char *addr, u16 vid) 1183 * Deletes the FDB entry from dev coresponding to addr. 1184 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1185 * struct net_device *dev, struct net_device *filter_dev, 1186 * int *idx) 1187 * Used to add FDB entries to dump requests. Implementers should add 1188 * entries to skb and update idx with the number of entries. 1189 * 1190 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1191 * u16 flags, struct netlink_ext_ack *extack) 1192 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1193 * struct net_device *dev, u32 filter_mask, 1194 * int nlflags) 1195 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1196 * u16 flags); 1197 * 1198 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1199 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1200 * which do not represent real hardware may define this to allow their 1201 * userspace components to manage their virtual carrier state. Devices 1202 * that determine carrier state from physical hardware properties (eg 1203 * network cables) or protocol-dependent mechanisms (eg 1204 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1205 * 1206 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1207 * struct netdev_phys_item_id *ppid); 1208 * Called to get ID of physical port of this device. If driver does 1209 * not implement this, it is assumed that the hw is not able to have 1210 * multiple net devices on single physical port. 1211 * 1212 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1213 * struct netdev_phys_item_id *ppid) 1214 * Called to get the parent ID of the physical port of this device. 1215 * 1216 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1217 * struct net_device *dev) 1218 * Called by upper layer devices to accelerate switching or other 1219 * station functionality into hardware. 'pdev is the lowerdev 1220 * to use for the offload and 'dev' is the net device that will 1221 * back the offload. Returns a pointer to the private structure 1222 * the upper layer will maintain. 1223 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1224 * Called by upper layer device to delete the station created 1225 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1226 * the station and priv is the structure returned by the add 1227 * operation. 1228 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1229 * int queue_index, u32 maxrate); 1230 * Called when a user wants to set a max-rate limitation of specific 1231 * TX queue. 1232 * int (*ndo_get_iflink)(const struct net_device *dev); 1233 * Called to get the iflink value of this device. 1234 * void (*ndo_change_proto_down)(struct net_device *dev, 1235 * bool proto_down); 1236 * This function is used to pass protocol port error state information 1237 * to the switch driver. The switch driver can react to the proto_down 1238 * by doing a phys down on the associated switch port. 1239 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1240 * This function is used to get egress tunnel information for given skb. 1241 * This is useful for retrieving outer tunnel header parameters while 1242 * sampling packet. 1243 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1244 * This function is used to specify the headroom that the skb must 1245 * consider when allocation skb during packet reception. Setting 1246 * appropriate rx headroom value allows avoiding skb head copy on 1247 * forward. Setting a negative value resets the rx headroom to the 1248 * default value. 1249 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1250 * This function is used to set or query state related to XDP on the 1251 * netdevice and manage BPF offload. See definition of 1252 * enum bpf_netdev_command for details. 1253 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1254 * u32 flags); 1255 * This function is used to submit @n XDP packets for transmit on a 1256 * netdevice. Returns number of frames successfully transmitted, frames 1257 * that got dropped are freed/returned via xdp_return_frame(). 1258 * Returns negative number, means general error invoking ndo, meaning 1259 * no frames were xmit'ed and core-caller will free all frames. 1260 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1261 * This function is used to wake up the softirq, ksoftirqd or kthread 1262 * responsible for sending and/or receiving packets on a specific 1263 * queue id bound to an AF_XDP socket. The flags field specifies if 1264 * only RX, only Tx, or both should be woken up using the flags 1265 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1266 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1267 * Get devlink port instance associated with a given netdev. 1268 * Called with a reference on the netdevice and devlink locks only, 1269 * rtnl_lock is not held. 1270 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1271 * int cmd); 1272 * Add, change, delete or get information on an IPv4 tunnel. 1273 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1274 * If a device is paired with a peer device, return the peer instance. 1275 * The caller must be under RCU read context. 1276 */ 1277 struct net_device_ops { 1278 int (*ndo_init)(struct net_device *dev); 1279 void (*ndo_uninit)(struct net_device *dev); 1280 int (*ndo_open)(struct net_device *dev); 1281 int (*ndo_stop)(struct net_device *dev); 1282 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1283 struct net_device *dev); 1284 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1285 struct net_device *dev, 1286 netdev_features_t features); 1287 u16 (*ndo_select_queue)(struct net_device *dev, 1288 struct sk_buff *skb, 1289 struct net_device *sb_dev); 1290 void (*ndo_change_rx_flags)(struct net_device *dev, 1291 int flags); 1292 void (*ndo_set_rx_mode)(struct net_device *dev); 1293 int (*ndo_set_mac_address)(struct net_device *dev, 1294 void *addr); 1295 int (*ndo_validate_addr)(struct net_device *dev); 1296 int (*ndo_do_ioctl)(struct net_device *dev, 1297 struct ifreq *ifr, int cmd); 1298 int (*ndo_set_config)(struct net_device *dev, 1299 struct ifmap *map); 1300 int (*ndo_change_mtu)(struct net_device *dev, 1301 int new_mtu); 1302 int (*ndo_neigh_setup)(struct net_device *dev, 1303 struct neigh_parms *); 1304 void (*ndo_tx_timeout) (struct net_device *dev, 1305 unsigned int txqueue); 1306 1307 void (*ndo_get_stats64)(struct net_device *dev, 1308 struct rtnl_link_stats64 *storage); 1309 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1310 int (*ndo_get_offload_stats)(int attr_id, 1311 const struct net_device *dev, 1312 void *attr_data); 1313 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1314 1315 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1316 __be16 proto, u16 vid); 1317 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1318 __be16 proto, u16 vid); 1319 #ifdef CONFIG_NET_POLL_CONTROLLER 1320 void (*ndo_poll_controller)(struct net_device *dev); 1321 int (*ndo_netpoll_setup)(struct net_device *dev, 1322 struct netpoll_info *info); 1323 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1324 #endif 1325 int (*ndo_set_vf_mac)(struct net_device *dev, 1326 int queue, u8 *mac); 1327 int (*ndo_set_vf_vlan)(struct net_device *dev, 1328 int queue, u16 vlan, 1329 u8 qos, __be16 proto); 1330 int (*ndo_set_vf_rate)(struct net_device *dev, 1331 int vf, int min_tx_rate, 1332 int max_tx_rate); 1333 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1334 int vf, bool setting); 1335 int (*ndo_set_vf_trust)(struct net_device *dev, 1336 int vf, bool setting); 1337 int (*ndo_get_vf_config)(struct net_device *dev, 1338 int vf, 1339 struct ifla_vf_info *ivf); 1340 int (*ndo_set_vf_link_state)(struct net_device *dev, 1341 int vf, int link_state); 1342 int (*ndo_get_vf_stats)(struct net_device *dev, 1343 int vf, 1344 struct ifla_vf_stats 1345 *vf_stats); 1346 int (*ndo_set_vf_port)(struct net_device *dev, 1347 int vf, 1348 struct nlattr *port[]); 1349 int (*ndo_get_vf_port)(struct net_device *dev, 1350 int vf, struct sk_buff *skb); 1351 int (*ndo_get_vf_guid)(struct net_device *dev, 1352 int vf, 1353 struct ifla_vf_guid *node_guid, 1354 struct ifla_vf_guid *port_guid); 1355 int (*ndo_set_vf_guid)(struct net_device *dev, 1356 int vf, u64 guid, 1357 int guid_type); 1358 int (*ndo_set_vf_rss_query_en)( 1359 struct net_device *dev, 1360 int vf, bool setting); 1361 int (*ndo_setup_tc)(struct net_device *dev, 1362 enum tc_setup_type type, 1363 void *type_data); 1364 #if IS_ENABLED(CONFIG_FCOE) 1365 int (*ndo_fcoe_enable)(struct net_device *dev); 1366 int (*ndo_fcoe_disable)(struct net_device *dev); 1367 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1368 u16 xid, 1369 struct scatterlist *sgl, 1370 unsigned int sgc); 1371 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1372 u16 xid); 1373 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1374 u16 xid, 1375 struct scatterlist *sgl, 1376 unsigned int sgc); 1377 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1378 struct netdev_fcoe_hbainfo *hbainfo); 1379 #endif 1380 1381 #if IS_ENABLED(CONFIG_LIBFCOE) 1382 #define NETDEV_FCOE_WWNN 0 1383 #define NETDEV_FCOE_WWPN 1 1384 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1385 u64 *wwn, int type); 1386 #endif 1387 1388 #ifdef CONFIG_RFS_ACCEL 1389 int (*ndo_rx_flow_steer)(struct net_device *dev, 1390 const struct sk_buff *skb, 1391 u16 rxq_index, 1392 u32 flow_id); 1393 #endif 1394 int (*ndo_add_slave)(struct net_device *dev, 1395 struct net_device *slave_dev, 1396 struct netlink_ext_ack *extack); 1397 int (*ndo_del_slave)(struct net_device *dev, 1398 struct net_device *slave_dev); 1399 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1400 struct sk_buff *skb, 1401 bool all_slaves); 1402 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1403 netdev_features_t features); 1404 int (*ndo_set_features)(struct net_device *dev, 1405 netdev_features_t features); 1406 int (*ndo_neigh_construct)(struct net_device *dev, 1407 struct neighbour *n); 1408 void (*ndo_neigh_destroy)(struct net_device *dev, 1409 struct neighbour *n); 1410 1411 int (*ndo_fdb_add)(struct ndmsg *ndm, 1412 struct nlattr *tb[], 1413 struct net_device *dev, 1414 const unsigned char *addr, 1415 u16 vid, 1416 u16 flags, 1417 struct netlink_ext_ack *extack); 1418 int (*ndo_fdb_del)(struct ndmsg *ndm, 1419 struct nlattr *tb[], 1420 struct net_device *dev, 1421 const unsigned char *addr, 1422 u16 vid); 1423 int (*ndo_fdb_dump)(struct sk_buff *skb, 1424 struct netlink_callback *cb, 1425 struct net_device *dev, 1426 struct net_device *filter_dev, 1427 int *idx); 1428 int (*ndo_fdb_get)(struct sk_buff *skb, 1429 struct nlattr *tb[], 1430 struct net_device *dev, 1431 const unsigned char *addr, 1432 u16 vid, u32 portid, u32 seq, 1433 struct netlink_ext_ack *extack); 1434 int (*ndo_bridge_setlink)(struct net_device *dev, 1435 struct nlmsghdr *nlh, 1436 u16 flags, 1437 struct netlink_ext_ack *extack); 1438 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1439 u32 pid, u32 seq, 1440 struct net_device *dev, 1441 u32 filter_mask, 1442 int nlflags); 1443 int (*ndo_bridge_dellink)(struct net_device *dev, 1444 struct nlmsghdr *nlh, 1445 u16 flags); 1446 int (*ndo_change_carrier)(struct net_device *dev, 1447 bool new_carrier); 1448 int (*ndo_get_phys_port_id)(struct net_device *dev, 1449 struct netdev_phys_item_id *ppid); 1450 int (*ndo_get_port_parent_id)(struct net_device *dev, 1451 struct netdev_phys_item_id *ppid); 1452 int (*ndo_get_phys_port_name)(struct net_device *dev, 1453 char *name, size_t len); 1454 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1455 struct net_device *dev); 1456 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1457 void *priv); 1458 1459 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1460 int queue_index, 1461 u32 maxrate); 1462 int (*ndo_get_iflink)(const struct net_device *dev); 1463 int (*ndo_change_proto_down)(struct net_device *dev, 1464 bool proto_down); 1465 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1466 struct sk_buff *skb); 1467 void (*ndo_set_rx_headroom)(struct net_device *dev, 1468 int needed_headroom); 1469 int (*ndo_bpf)(struct net_device *dev, 1470 struct netdev_bpf *bpf); 1471 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1472 struct xdp_frame **xdp, 1473 u32 flags); 1474 int (*ndo_xsk_wakeup)(struct net_device *dev, 1475 u32 queue_id, u32 flags); 1476 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1477 int (*ndo_tunnel_ctl)(struct net_device *dev, 1478 struct ip_tunnel_parm *p, int cmd); 1479 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1480 }; 1481 1482 /** 1483 * enum netdev_priv_flags - &struct net_device priv_flags 1484 * 1485 * These are the &struct net_device, they are only set internally 1486 * by drivers and used in the kernel. These flags are invisible to 1487 * userspace; this means that the order of these flags can change 1488 * during any kernel release. 1489 * 1490 * You should have a pretty good reason to be extending these flags. 1491 * 1492 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1493 * @IFF_EBRIDGE: Ethernet bridging device 1494 * @IFF_BONDING: bonding master or slave 1495 * @IFF_ISATAP: ISATAP interface (RFC4214) 1496 * @IFF_WAN_HDLC: WAN HDLC device 1497 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1498 * release skb->dst 1499 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1500 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1501 * @IFF_MACVLAN_PORT: device used as macvlan port 1502 * @IFF_BRIDGE_PORT: device used as bridge port 1503 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1504 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1505 * @IFF_UNICAST_FLT: Supports unicast filtering 1506 * @IFF_TEAM_PORT: device used as team port 1507 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1508 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1509 * change when it's running 1510 * @IFF_MACVLAN: Macvlan device 1511 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1512 * underlying stacked devices 1513 * @IFF_L3MDEV_MASTER: device is an L3 master device 1514 * @IFF_NO_QUEUE: device can run without qdisc attached 1515 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1516 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1517 * @IFF_TEAM: device is a team device 1518 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1519 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1520 * entity (i.e. the master device for bridged veth) 1521 * @IFF_MACSEC: device is a MACsec device 1522 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1523 * @IFF_FAILOVER: device is a failover master device 1524 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1525 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1526 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1527 */ 1528 enum netdev_priv_flags { 1529 IFF_802_1Q_VLAN = 1<<0, 1530 IFF_EBRIDGE = 1<<1, 1531 IFF_BONDING = 1<<2, 1532 IFF_ISATAP = 1<<3, 1533 IFF_WAN_HDLC = 1<<4, 1534 IFF_XMIT_DST_RELEASE = 1<<5, 1535 IFF_DONT_BRIDGE = 1<<6, 1536 IFF_DISABLE_NETPOLL = 1<<7, 1537 IFF_MACVLAN_PORT = 1<<8, 1538 IFF_BRIDGE_PORT = 1<<9, 1539 IFF_OVS_DATAPATH = 1<<10, 1540 IFF_TX_SKB_SHARING = 1<<11, 1541 IFF_UNICAST_FLT = 1<<12, 1542 IFF_TEAM_PORT = 1<<13, 1543 IFF_SUPP_NOFCS = 1<<14, 1544 IFF_LIVE_ADDR_CHANGE = 1<<15, 1545 IFF_MACVLAN = 1<<16, 1546 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1547 IFF_L3MDEV_MASTER = 1<<18, 1548 IFF_NO_QUEUE = 1<<19, 1549 IFF_OPENVSWITCH = 1<<20, 1550 IFF_L3MDEV_SLAVE = 1<<21, 1551 IFF_TEAM = 1<<22, 1552 IFF_RXFH_CONFIGURED = 1<<23, 1553 IFF_PHONY_HEADROOM = 1<<24, 1554 IFF_MACSEC = 1<<25, 1555 IFF_NO_RX_HANDLER = 1<<26, 1556 IFF_FAILOVER = 1<<27, 1557 IFF_FAILOVER_SLAVE = 1<<28, 1558 IFF_L3MDEV_RX_HANDLER = 1<<29, 1559 IFF_LIVE_RENAME_OK = 1<<30, 1560 }; 1561 1562 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1563 #define IFF_EBRIDGE IFF_EBRIDGE 1564 #define IFF_BONDING IFF_BONDING 1565 #define IFF_ISATAP IFF_ISATAP 1566 #define IFF_WAN_HDLC IFF_WAN_HDLC 1567 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1568 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1569 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1570 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1571 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1572 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1573 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1574 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1575 #define IFF_TEAM_PORT IFF_TEAM_PORT 1576 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1577 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1578 #define IFF_MACVLAN IFF_MACVLAN 1579 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1580 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1581 #define IFF_NO_QUEUE IFF_NO_QUEUE 1582 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1583 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1584 #define IFF_TEAM IFF_TEAM 1585 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1586 #define IFF_MACSEC IFF_MACSEC 1587 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1588 #define IFF_FAILOVER IFF_FAILOVER 1589 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1590 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1591 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1592 1593 /** 1594 * struct net_device - The DEVICE structure. 1595 * 1596 * Actually, this whole structure is a big mistake. It mixes I/O 1597 * data with strictly "high-level" data, and it has to know about 1598 * almost every data structure used in the INET module. 1599 * 1600 * @name: This is the first field of the "visible" part of this structure 1601 * (i.e. as seen by users in the "Space.c" file). It is the name 1602 * of the interface. 1603 * 1604 * @name_node: Name hashlist node 1605 * @ifalias: SNMP alias 1606 * @mem_end: Shared memory end 1607 * @mem_start: Shared memory start 1608 * @base_addr: Device I/O address 1609 * @irq: Device IRQ number 1610 * 1611 * @state: Generic network queuing layer state, see netdev_state_t 1612 * @dev_list: The global list of network devices 1613 * @napi_list: List entry used for polling NAPI devices 1614 * @unreg_list: List entry when we are unregistering the 1615 * device; see the function unregister_netdev 1616 * @close_list: List entry used when we are closing the device 1617 * @ptype_all: Device-specific packet handlers for all protocols 1618 * @ptype_specific: Device-specific, protocol-specific packet handlers 1619 * 1620 * @adj_list: Directly linked devices, like slaves for bonding 1621 * @features: Currently active device features 1622 * @hw_features: User-changeable features 1623 * 1624 * @wanted_features: User-requested features 1625 * @vlan_features: Mask of features inheritable by VLAN devices 1626 * 1627 * @hw_enc_features: Mask of features inherited by encapsulating devices 1628 * This field indicates what encapsulation 1629 * offloads the hardware is capable of doing, 1630 * and drivers will need to set them appropriately. 1631 * 1632 * @mpls_features: Mask of features inheritable by MPLS 1633 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1634 * 1635 * @ifindex: interface index 1636 * @group: The group the device belongs to 1637 * 1638 * @stats: Statistics struct, which was left as a legacy, use 1639 * rtnl_link_stats64 instead 1640 * 1641 * @rx_dropped: Dropped packets by core network, 1642 * do not use this in drivers 1643 * @tx_dropped: Dropped packets by core network, 1644 * do not use this in drivers 1645 * @rx_nohandler: nohandler dropped packets by core network on 1646 * inactive devices, do not use this in drivers 1647 * @carrier_up_count: Number of times the carrier has been up 1648 * @carrier_down_count: Number of times the carrier has been down 1649 * 1650 * @wireless_handlers: List of functions to handle Wireless Extensions, 1651 * instead of ioctl, 1652 * see <net/iw_handler.h> for details. 1653 * @wireless_data: Instance data managed by the core of wireless extensions 1654 * 1655 * @netdev_ops: Includes several pointers to callbacks, 1656 * if one wants to override the ndo_*() functions 1657 * @ethtool_ops: Management operations 1658 * @l3mdev_ops: Layer 3 master device operations 1659 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1660 * discovery handling. Necessary for e.g. 6LoWPAN. 1661 * @xfrmdev_ops: Transformation offload operations 1662 * @tlsdev_ops: Transport Layer Security offload operations 1663 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1664 * of Layer 2 headers. 1665 * 1666 * @flags: Interface flags (a la BSD) 1667 * @priv_flags: Like 'flags' but invisible to userspace, 1668 * see if.h for the definitions 1669 * @gflags: Global flags ( kept as legacy ) 1670 * @padded: How much padding added by alloc_netdev() 1671 * @operstate: RFC2863 operstate 1672 * @link_mode: Mapping policy to operstate 1673 * @if_port: Selectable AUI, TP, ... 1674 * @dma: DMA channel 1675 * @mtu: Interface MTU value 1676 * @min_mtu: Interface Minimum MTU value 1677 * @max_mtu: Interface Maximum MTU value 1678 * @type: Interface hardware type 1679 * @hard_header_len: Maximum hardware header length. 1680 * @min_header_len: Minimum hardware header length 1681 * 1682 * @needed_headroom: Extra headroom the hardware may need, but not in all 1683 * cases can this be guaranteed 1684 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1685 * cases can this be guaranteed. Some cases also use 1686 * LL_MAX_HEADER instead to allocate the skb 1687 * 1688 * interface address info: 1689 * 1690 * @perm_addr: Permanent hw address 1691 * @addr_assign_type: Hw address assignment type 1692 * @addr_len: Hardware address length 1693 * @upper_level: Maximum depth level of upper devices. 1694 * @lower_level: Maximum depth level of lower devices. 1695 * @neigh_priv_len: Used in neigh_alloc() 1696 * @dev_id: Used to differentiate devices that share 1697 * the same link layer address 1698 * @dev_port: Used to differentiate devices that share 1699 * the same function 1700 * @addr_list_lock: XXX: need comments on this one 1701 * @name_assign_type: network interface name assignment type 1702 * @uc_promisc: Counter that indicates promiscuous mode 1703 * has been enabled due to the need to listen to 1704 * additional unicast addresses in a device that 1705 * does not implement ndo_set_rx_mode() 1706 * @uc: unicast mac addresses 1707 * @mc: multicast mac addresses 1708 * @dev_addrs: list of device hw addresses 1709 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1710 * @promiscuity: Number of times the NIC is told to work in 1711 * promiscuous mode; if it becomes 0 the NIC will 1712 * exit promiscuous mode 1713 * @allmulti: Counter, enables or disables allmulticast mode 1714 * 1715 * @vlan_info: VLAN info 1716 * @dsa_ptr: dsa specific data 1717 * @tipc_ptr: TIPC specific data 1718 * @atalk_ptr: AppleTalk link 1719 * @ip_ptr: IPv4 specific data 1720 * @dn_ptr: DECnet specific data 1721 * @ip6_ptr: IPv6 specific data 1722 * @ax25_ptr: AX.25 specific data 1723 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1724 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1725 * device struct 1726 * @mpls_ptr: mpls_dev struct pointer 1727 * 1728 * @dev_addr: Hw address (before bcast, 1729 * because most packets are unicast) 1730 * 1731 * @_rx: Array of RX queues 1732 * @num_rx_queues: Number of RX queues 1733 * allocated at register_netdev() time 1734 * @real_num_rx_queues: Number of RX queues currently active in device 1735 * @xdp_prog: XDP sockets filter program pointer 1736 * @gro_flush_timeout: timeout for GRO layer in NAPI 1737 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1738 * allow to avoid NIC hard IRQ, on busy queues. 1739 * 1740 * @rx_handler: handler for received packets 1741 * @rx_handler_data: XXX: need comments on this one 1742 * @miniq_ingress: ingress/clsact qdisc specific data for 1743 * ingress processing 1744 * @ingress_queue: XXX: need comments on this one 1745 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1746 * @broadcast: hw bcast address 1747 * 1748 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1749 * indexed by RX queue number. Assigned by driver. 1750 * This must only be set if the ndo_rx_flow_steer 1751 * operation is defined 1752 * @index_hlist: Device index hash chain 1753 * 1754 * @_tx: Array of TX queues 1755 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1756 * @real_num_tx_queues: Number of TX queues currently active in device 1757 * @qdisc: Root qdisc from userspace point of view 1758 * @tx_queue_len: Max frames per queue allowed 1759 * @tx_global_lock: XXX: need comments on this one 1760 * @xdp_bulkq: XDP device bulk queue 1761 * @xps_cpus_map: all CPUs map for XPS device 1762 * @xps_rxqs_map: all RXQs map for XPS device 1763 * 1764 * @xps_maps: XXX: need comments on this one 1765 * @miniq_egress: clsact qdisc specific data for 1766 * egress processing 1767 * @qdisc_hash: qdisc hash table 1768 * @watchdog_timeo: Represents the timeout that is used by 1769 * the watchdog (see dev_watchdog()) 1770 * @watchdog_timer: List of timers 1771 * 1772 * @proto_down_reason: reason a netdev interface is held down 1773 * @pcpu_refcnt: Number of references to this device 1774 * @todo_list: Delayed register/unregister 1775 * @link_watch_list: XXX: need comments on this one 1776 * 1777 * @reg_state: Register/unregister state machine 1778 * @dismantle: Device is going to be freed 1779 * @rtnl_link_state: This enum represents the phases of creating 1780 * a new link 1781 * 1782 * @needs_free_netdev: Should unregister perform free_netdev? 1783 * @priv_destructor: Called from unregister 1784 * @npinfo: XXX: need comments on this one 1785 * @nd_net: Network namespace this network device is inside 1786 * 1787 * @ml_priv: Mid-layer private 1788 * @lstats: Loopback statistics 1789 * @tstats: Tunnel statistics 1790 * @dstats: Dummy statistics 1791 * @vstats: Virtual ethernet statistics 1792 * 1793 * @garp_port: GARP 1794 * @mrp_port: MRP 1795 * 1796 * @dev: Class/net/name entry 1797 * @sysfs_groups: Space for optional device, statistics and wireless 1798 * sysfs groups 1799 * 1800 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1801 * @rtnl_link_ops: Rtnl_link_ops 1802 * 1803 * @gso_max_size: Maximum size of generic segmentation offload 1804 * @gso_max_segs: Maximum number of segments that can be passed to the 1805 * NIC for GSO 1806 * 1807 * @dcbnl_ops: Data Center Bridging netlink ops 1808 * @num_tc: Number of traffic classes in the net device 1809 * @tc_to_txq: XXX: need comments on this one 1810 * @prio_tc_map: XXX: need comments on this one 1811 * 1812 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1813 * 1814 * @priomap: XXX: need comments on this one 1815 * @phydev: Physical device may attach itself 1816 * for hardware timestamping 1817 * @sfp_bus: attached &struct sfp_bus structure. 1818 * 1819 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1820 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1821 * 1822 * @proto_down: protocol port state information can be sent to the 1823 * switch driver and used to set the phys state of the 1824 * switch port. 1825 * 1826 * @wol_enabled: Wake-on-LAN is enabled 1827 * 1828 * @net_notifier_list: List of per-net netdev notifier block 1829 * that follow this device when it is moved 1830 * to another network namespace. 1831 * 1832 * @macsec_ops: MACsec offloading ops 1833 * 1834 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1835 * offload capabilities of the device 1836 * @udp_tunnel_nic: UDP tunnel offload state 1837 * @xdp_state: stores info on attached XDP BPF programs 1838 * 1839 * @nested_level: Used as as a parameter of spin_lock_nested() of 1840 * dev->addr_list_lock. 1841 * @unlink_list: As netif_addr_lock() can be called recursively, 1842 * keep a list of interfaces to be deleted. 1843 * 1844 * FIXME: cleanup struct net_device such that network protocol info 1845 * moves out. 1846 */ 1847 1848 struct net_device { 1849 char name[IFNAMSIZ]; 1850 struct netdev_name_node *name_node; 1851 struct dev_ifalias __rcu *ifalias; 1852 /* 1853 * I/O specific fields 1854 * FIXME: Merge these and struct ifmap into one 1855 */ 1856 unsigned long mem_end; 1857 unsigned long mem_start; 1858 unsigned long base_addr; 1859 int irq; 1860 1861 /* 1862 * Some hardware also needs these fields (state,dev_list, 1863 * napi_list,unreg_list,close_list) but they are not 1864 * part of the usual set specified in Space.c. 1865 */ 1866 1867 unsigned long state; 1868 1869 struct list_head dev_list; 1870 struct list_head napi_list; 1871 struct list_head unreg_list; 1872 struct list_head close_list; 1873 struct list_head ptype_all; 1874 struct list_head ptype_specific; 1875 1876 struct { 1877 struct list_head upper; 1878 struct list_head lower; 1879 } adj_list; 1880 1881 netdev_features_t features; 1882 netdev_features_t hw_features; 1883 netdev_features_t wanted_features; 1884 netdev_features_t vlan_features; 1885 netdev_features_t hw_enc_features; 1886 netdev_features_t mpls_features; 1887 netdev_features_t gso_partial_features; 1888 1889 int ifindex; 1890 int group; 1891 1892 struct net_device_stats stats; 1893 1894 atomic_long_t rx_dropped; 1895 atomic_long_t tx_dropped; 1896 atomic_long_t rx_nohandler; 1897 1898 /* Stats to monitor link on/off, flapping */ 1899 atomic_t carrier_up_count; 1900 atomic_t carrier_down_count; 1901 1902 #ifdef CONFIG_WIRELESS_EXT 1903 const struct iw_handler_def *wireless_handlers; 1904 struct iw_public_data *wireless_data; 1905 #endif 1906 const struct net_device_ops *netdev_ops; 1907 const struct ethtool_ops *ethtool_ops; 1908 #ifdef CONFIG_NET_L3_MASTER_DEV 1909 const struct l3mdev_ops *l3mdev_ops; 1910 #endif 1911 #if IS_ENABLED(CONFIG_IPV6) 1912 const struct ndisc_ops *ndisc_ops; 1913 #endif 1914 1915 #ifdef CONFIG_XFRM_OFFLOAD 1916 const struct xfrmdev_ops *xfrmdev_ops; 1917 #endif 1918 1919 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1920 const struct tlsdev_ops *tlsdev_ops; 1921 #endif 1922 1923 const struct header_ops *header_ops; 1924 1925 unsigned int flags; 1926 unsigned int priv_flags; 1927 1928 unsigned short gflags; 1929 unsigned short padded; 1930 1931 unsigned char operstate; 1932 unsigned char link_mode; 1933 1934 unsigned char if_port; 1935 unsigned char dma; 1936 1937 /* Note : dev->mtu is often read without holding a lock. 1938 * Writers usually hold RTNL. 1939 * It is recommended to use READ_ONCE() to annotate the reads, 1940 * and to use WRITE_ONCE() to annotate the writes. 1941 */ 1942 unsigned int mtu; 1943 unsigned int min_mtu; 1944 unsigned int max_mtu; 1945 unsigned short type; 1946 unsigned short hard_header_len; 1947 unsigned char min_header_len; 1948 unsigned char name_assign_type; 1949 1950 unsigned short needed_headroom; 1951 unsigned short needed_tailroom; 1952 1953 /* Interface address info. */ 1954 unsigned char perm_addr[MAX_ADDR_LEN]; 1955 unsigned char addr_assign_type; 1956 unsigned char addr_len; 1957 unsigned char upper_level; 1958 unsigned char lower_level; 1959 1960 unsigned short neigh_priv_len; 1961 unsigned short dev_id; 1962 unsigned short dev_port; 1963 spinlock_t addr_list_lock; 1964 1965 struct netdev_hw_addr_list uc; 1966 struct netdev_hw_addr_list mc; 1967 struct netdev_hw_addr_list dev_addrs; 1968 1969 #ifdef CONFIG_SYSFS 1970 struct kset *queues_kset; 1971 #endif 1972 #ifdef CONFIG_LOCKDEP 1973 struct list_head unlink_list; 1974 #endif 1975 unsigned int promiscuity; 1976 unsigned int allmulti; 1977 bool uc_promisc; 1978 #ifdef CONFIG_LOCKDEP 1979 unsigned char nested_level; 1980 #endif 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 static inline void net_prefetch(void *p) 2197 { 2198 prefetch(p); 2199 #if L1_CACHE_BYTES < 128 2200 prefetch((u8 *)p + L1_CACHE_BYTES); 2201 #endif 2202 } 2203 2204 static inline void net_prefetchw(void *p) 2205 { 2206 prefetchw(p); 2207 #if L1_CACHE_BYTES < 128 2208 prefetchw((u8 *)p + L1_CACHE_BYTES); 2209 #endif 2210 } 2211 2212 void netdev_unbind_sb_channel(struct net_device *dev, 2213 struct net_device *sb_dev); 2214 int netdev_bind_sb_channel_queue(struct net_device *dev, 2215 struct net_device *sb_dev, 2216 u8 tc, u16 count, u16 offset); 2217 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2218 static inline int netdev_get_sb_channel(struct net_device *dev) 2219 { 2220 return max_t(int, -dev->num_tc, 0); 2221 } 2222 2223 static inline 2224 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2225 unsigned int index) 2226 { 2227 return &dev->_tx[index]; 2228 } 2229 2230 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2231 const struct sk_buff *skb) 2232 { 2233 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2234 } 2235 2236 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2237 void (*f)(struct net_device *, 2238 struct netdev_queue *, 2239 void *), 2240 void *arg) 2241 { 2242 unsigned int i; 2243 2244 for (i = 0; i < dev->num_tx_queues; i++) 2245 f(dev, &dev->_tx[i], arg); 2246 } 2247 2248 #define netdev_lockdep_set_classes(dev) \ 2249 { \ 2250 static struct lock_class_key qdisc_tx_busylock_key; \ 2251 static struct lock_class_key qdisc_running_key; \ 2252 static struct lock_class_key qdisc_xmit_lock_key; \ 2253 static struct lock_class_key dev_addr_list_lock_key; \ 2254 unsigned int i; \ 2255 \ 2256 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2257 (dev)->qdisc_running_key = &qdisc_running_key; \ 2258 lockdep_set_class(&(dev)->addr_list_lock, \ 2259 &dev_addr_list_lock_key); \ 2260 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2261 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2262 &qdisc_xmit_lock_key); \ 2263 } 2264 2265 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2266 struct net_device *sb_dev); 2267 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2268 struct sk_buff *skb, 2269 struct net_device *sb_dev); 2270 2271 /* returns the headroom that the master device needs to take in account 2272 * when forwarding to this dev 2273 */ 2274 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2275 { 2276 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2277 } 2278 2279 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2280 { 2281 if (dev->netdev_ops->ndo_set_rx_headroom) 2282 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2283 } 2284 2285 /* set the device rx headroom to the dev's default */ 2286 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2287 { 2288 netdev_set_rx_headroom(dev, -1); 2289 } 2290 2291 /* 2292 * Net namespace inlines 2293 */ 2294 static inline 2295 struct net *dev_net(const struct net_device *dev) 2296 { 2297 return read_pnet(&dev->nd_net); 2298 } 2299 2300 static inline 2301 void dev_net_set(struct net_device *dev, struct net *net) 2302 { 2303 write_pnet(&dev->nd_net, net); 2304 } 2305 2306 /** 2307 * netdev_priv - access network device private data 2308 * @dev: network device 2309 * 2310 * Get network device private data 2311 */ 2312 static inline void *netdev_priv(const struct net_device *dev) 2313 { 2314 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2315 } 2316 2317 /* Set the sysfs physical device reference for the network logical device 2318 * if set prior to registration will cause a symlink during initialization. 2319 */ 2320 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2321 2322 /* Set the sysfs device type for the network logical device to allow 2323 * fine-grained identification of different network device types. For 2324 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2325 */ 2326 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2327 2328 /* Default NAPI poll() weight 2329 * Device drivers are strongly advised to not use bigger value 2330 */ 2331 #define NAPI_POLL_WEIGHT 64 2332 2333 /** 2334 * netif_napi_add - initialize a NAPI context 2335 * @dev: network device 2336 * @napi: NAPI context 2337 * @poll: polling function 2338 * @weight: default weight 2339 * 2340 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2341 * *any* of the other NAPI-related functions. 2342 */ 2343 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2344 int (*poll)(struct napi_struct *, int), int weight); 2345 2346 /** 2347 * netif_tx_napi_add - initialize a NAPI context 2348 * @dev: network device 2349 * @napi: NAPI context 2350 * @poll: polling function 2351 * @weight: default weight 2352 * 2353 * This variant of netif_napi_add() should be used from drivers using NAPI 2354 * to exclusively poll a TX queue. 2355 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2356 */ 2357 static inline void netif_tx_napi_add(struct net_device *dev, 2358 struct napi_struct *napi, 2359 int (*poll)(struct napi_struct *, int), 2360 int weight) 2361 { 2362 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2363 netif_napi_add(dev, napi, poll, weight); 2364 } 2365 2366 /** 2367 * __netif_napi_del - remove a NAPI context 2368 * @napi: NAPI context 2369 * 2370 * Warning: caller must observe RCU grace period before freeing memory 2371 * containing @napi. Drivers might want to call this helper to combine 2372 * all the needed RCU grace periods into a single one. 2373 */ 2374 void __netif_napi_del(struct napi_struct *napi); 2375 2376 /** 2377 * netif_napi_del - remove a NAPI context 2378 * @napi: NAPI context 2379 * 2380 * netif_napi_del() removes a NAPI context from the network device NAPI list 2381 */ 2382 static inline void netif_napi_del(struct napi_struct *napi) 2383 { 2384 __netif_napi_del(napi); 2385 synchronize_net(); 2386 } 2387 2388 struct napi_gro_cb { 2389 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2390 void *frag0; 2391 2392 /* Length of frag0. */ 2393 unsigned int frag0_len; 2394 2395 /* This indicates where we are processing relative to skb->data. */ 2396 int data_offset; 2397 2398 /* This is non-zero if the packet cannot be merged with the new skb. */ 2399 u16 flush; 2400 2401 /* Save the IP ID here and check when we get to the transport layer */ 2402 u16 flush_id; 2403 2404 /* Number of segments aggregated. */ 2405 u16 count; 2406 2407 /* Start offset for remote checksum offload */ 2408 u16 gro_remcsum_start; 2409 2410 /* jiffies when first packet was created/queued */ 2411 unsigned long age; 2412 2413 /* Used in ipv6_gro_receive() and foo-over-udp */ 2414 u16 proto; 2415 2416 /* This is non-zero if the packet may be of the same flow. */ 2417 u8 same_flow:1; 2418 2419 /* Used in tunnel GRO receive */ 2420 u8 encap_mark:1; 2421 2422 /* GRO checksum is valid */ 2423 u8 csum_valid:1; 2424 2425 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2426 u8 csum_cnt:3; 2427 2428 /* Free the skb? */ 2429 u8 free:2; 2430 #define NAPI_GRO_FREE 1 2431 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2432 2433 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2434 u8 is_ipv6:1; 2435 2436 /* Used in GRE, set in fou/gue_gro_receive */ 2437 u8 is_fou:1; 2438 2439 /* Used to determine if flush_id can be ignored */ 2440 u8 is_atomic:1; 2441 2442 /* Number of gro_receive callbacks this packet already went through */ 2443 u8 recursion_counter:4; 2444 2445 /* GRO is done by frag_list pointer chaining. */ 2446 u8 is_flist:1; 2447 2448 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2449 __wsum csum; 2450 2451 /* used in skb_gro_receive() slow path */ 2452 struct sk_buff *last; 2453 }; 2454 2455 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2456 2457 #define GRO_RECURSION_LIMIT 15 2458 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2459 { 2460 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2461 } 2462 2463 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2464 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2465 struct list_head *head, 2466 struct sk_buff *skb) 2467 { 2468 if (unlikely(gro_recursion_inc_test(skb))) { 2469 NAPI_GRO_CB(skb)->flush |= 1; 2470 return NULL; 2471 } 2472 2473 return cb(head, skb); 2474 } 2475 2476 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2477 struct sk_buff *); 2478 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2479 struct sock *sk, 2480 struct list_head *head, 2481 struct sk_buff *skb) 2482 { 2483 if (unlikely(gro_recursion_inc_test(skb))) { 2484 NAPI_GRO_CB(skb)->flush |= 1; 2485 return NULL; 2486 } 2487 2488 return cb(sk, head, skb); 2489 } 2490 2491 struct packet_type { 2492 __be16 type; /* This is really htons(ether_type). */ 2493 bool ignore_outgoing; 2494 struct net_device *dev; /* NULL is wildcarded here */ 2495 int (*func) (struct sk_buff *, 2496 struct net_device *, 2497 struct packet_type *, 2498 struct net_device *); 2499 void (*list_func) (struct list_head *, 2500 struct packet_type *, 2501 struct net_device *); 2502 bool (*id_match)(struct packet_type *ptype, 2503 struct sock *sk); 2504 void *af_packet_priv; 2505 struct list_head list; 2506 }; 2507 2508 struct offload_callbacks { 2509 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2510 netdev_features_t features); 2511 struct sk_buff *(*gro_receive)(struct list_head *head, 2512 struct sk_buff *skb); 2513 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2514 }; 2515 2516 struct packet_offload { 2517 __be16 type; /* This is really htons(ether_type). */ 2518 u16 priority; 2519 struct offload_callbacks callbacks; 2520 struct list_head list; 2521 }; 2522 2523 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2524 struct pcpu_sw_netstats { 2525 u64 rx_packets; 2526 u64 rx_bytes; 2527 u64 tx_packets; 2528 u64 tx_bytes; 2529 struct u64_stats_sync syncp; 2530 } __aligned(4 * sizeof(u64)); 2531 2532 struct pcpu_lstats { 2533 u64_stats_t packets; 2534 u64_stats_t bytes; 2535 struct u64_stats_sync syncp; 2536 } __aligned(2 * sizeof(u64)); 2537 2538 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2539 2540 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2541 { 2542 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2543 2544 u64_stats_update_begin(&tstats->syncp); 2545 tstats->rx_bytes += len; 2546 tstats->rx_packets++; 2547 u64_stats_update_end(&tstats->syncp); 2548 } 2549 2550 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2551 unsigned int packets, 2552 unsigned int len) 2553 { 2554 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2555 2556 u64_stats_update_begin(&tstats->syncp); 2557 tstats->tx_bytes += len; 2558 tstats->tx_packets += packets; 2559 u64_stats_update_end(&tstats->syncp); 2560 } 2561 2562 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2563 { 2564 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2565 2566 u64_stats_update_begin(&lstats->syncp); 2567 u64_stats_add(&lstats->bytes, len); 2568 u64_stats_inc(&lstats->packets); 2569 u64_stats_update_end(&lstats->syncp); 2570 } 2571 2572 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2573 ({ \ 2574 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2575 if (pcpu_stats) { \ 2576 int __cpu; \ 2577 for_each_possible_cpu(__cpu) { \ 2578 typeof(type) *stat; \ 2579 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2580 u64_stats_init(&stat->syncp); \ 2581 } \ 2582 } \ 2583 pcpu_stats; \ 2584 }) 2585 2586 #define netdev_alloc_pcpu_stats(type) \ 2587 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2588 2589 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2590 ({ \ 2591 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2592 if (pcpu_stats) { \ 2593 int __cpu; \ 2594 for_each_possible_cpu(__cpu) { \ 2595 typeof(type) *stat; \ 2596 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2597 u64_stats_init(&stat->syncp); \ 2598 } \ 2599 } \ 2600 pcpu_stats; \ 2601 }) 2602 2603 enum netdev_lag_tx_type { 2604 NETDEV_LAG_TX_TYPE_UNKNOWN, 2605 NETDEV_LAG_TX_TYPE_RANDOM, 2606 NETDEV_LAG_TX_TYPE_BROADCAST, 2607 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2608 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2609 NETDEV_LAG_TX_TYPE_HASH, 2610 }; 2611 2612 enum netdev_lag_hash { 2613 NETDEV_LAG_HASH_NONE, 2614 NETDEV_LAG_HASH_L2, 2615 NETDEV_LAG_HASH_L34, 2616 NETDEV_LAG_HASH_L23, 2617 NETDEV_LAG_HASH_E23, 2618 NETDEV_LAG_HASH_E34, 2619 NETDEV_LAG_HASH_UNKNOWN, 2620 }; 2621 2622 struct netdev_lag_upper_info { 2623 enum netdev_lag_tx_type tx_type; 2624 enum netdev_lag_hash hash_type; 2625 }; 2626 2627 struct netdev_lag_lower_state_info { 2628 u8 link_up : 1, 2629 tx_enabled : 1; 2630 }; 2631 2632 #include <linux/notifier.h> 2633 2634 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2635 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2636 * adding new types. 2637 */ 2638 enum netdev_cmd { 2639 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2640 NETDEV_DOWN, 2641 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2642 detected a hardware crash and restarted 2643 - we can use this eg to kick tcp sessions 2644 once done */ 2645 NETDEV_CHANGE, /* Notify device state change */ 2646 NETDEV_REGISTER, 2647 NETDEV_UNREGISTER, 2648 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2649 NETDEV_CHANGEADDR, /* notify after the address change */ 2650 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2651 NETDEV_GOING_DOWN, 2652 NETDEV_CHANGENAME, 2653 NETDEV_FEAT_CHANGE, 2654 NETDEV_BONDING_FAILOVER, 2655 NETDEV_PRE_UP, 2656 NETDEV_PRE_TYPE_CHANGE, 2657 NETDEV_POST_TYPE_CHANGE, 2658 NETDEV_POST_INIT, 2659 NETDEV_RELEASE, 2660 NETDEV_NOTIFY_PEERS, 2661 NETDEV_JOIN, 2662 NETDEV_CHANGEUPPER, 2663 NETDEV_RESEND_IGMP, 2664 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2665 NETDEV_CHANGEINFODATA, 2666 NETDEV_BONDING_INFO, 2667 NETDEV_PRECHANGEUPPER, 2668 NETDEV_CHANGELOWERSTATE, 2669 NETDEV_UDP_TUNNEL_PUSH_INFO, 2670 NETDEV_UDP_TUNNEL_DROP_INFO, 2671 NETDEV_CHANGE_TX_QUEUE_LEN, 2672 NETDEV_CVLAN_FILTER_PUSH_INFO, 2673 NETDEV_CVLAN_FILTER_DROP_INFO, 2674 NETDEV_SVLAN_FILTER_PUSH_INFO, 2675 NETDEV_SVLAN_FILTER_DROP_INFO, 2676 }; 2677 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2678 2679 int register_netdevice_notifier(struct notifier_block *nb); 2680 int unregister_netdevice_notifier(struct notifier_block *nb); 2681 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2682 int unregister_netdevice_notifier_net(struct net *net, 2683 struct notifier_block *nb); 2684 int register_netdevice_notifier_dev_net(struct net_device *dev, 2685 struct notifier_block *nb, 2686 struct netdev_net_notifier *nn); 2687 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2688 struct notifier_block *nb, 2689 struct netdev_net_notifier *nn); 2690 2691 struct netdev_notifier_info { 2692 struct net_device *dev; 2693 struct netlink_ext_ack *extack; 2694 }; 2695 2696 struct netdev_notifier_info_ext { 2697 struct netdev_notifier_info info; /* must be first */ 2698 union { 2699 u32 mtu; 2700 } ext; 2701 }; 2702 2703 struct netdev_notifier_change_info { 2704 struct netdev_notifier_info info; /* must be first */ 2705 unsigned int flags_changed; 2706 }; 2707 2708 struct netdev_notifier_changeupper_info { 2709 struct netdev_notifier_info info; /* must be first */ 2710 struct net_device *upper_dev; /* new upper dev */ 2711 bool master; /* is upper dev master */ 2712 bool linking; /* is the notification for link or unlink */ 2713 void *upper_info; /* upper dev info */ 2714 }; 2715 2716 struct netdev_notifier_changelowerstate_info { 2717 struct netdev_notifier_info info; /* must be first */ 2718 void *lower_state_info; /* is lower dev state */ 2719 }; 2720 2721 struct netdev_notifier_pre_changeaddr_info { 2722 struct netdev_notifier_info info; /* must be first */ 2723 const unsigned char *dev_addr; 2724 }; 2725 2726 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2727 struct net_device *dev) 2728 { 2729 info->dev = dev; 2730 info->extack = NULL; 2731 } 2732 2733 static inline struct net_device * 2734 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2735 { 2736 return info->dev; 2737 } 2738 2739 static inline struct netlink_ext_ack * 2740 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2741 { 2742 return info->extack; 2743 } 2744 2745 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2746 2747 2748 extern rwlock_t dev_base_lock; /* Device list lock */ 2749 2750 #define for_each_netdev(net, d) \ 2751 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2752 #define for_each_netdev_reverse(net, d) \ 2753 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2754 #define for_each_netdev_rcu(net, d) \ 2755 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2756 #define for_each_netdev_safe(net, d, n) \ 2757 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2758 #define for_each_netdev_continue(net, d) \ 2759 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2760 #define for_each_netdev_continue_reverse(net, d) \ 2761 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2762 dev_list) 2763 #define for_each_netdev_continue_rcu(net, d) \ 2764 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2765 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2766 for_each_netdev_rcu(&init_net, slave) \ 2767 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2768 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2769 2770 static inline struct net_device *next_net_device(struct net_device *dev) 2771 { 2772 struct list_head *lh; 2773 struct net *net; 2774 2775 net = dev_net(dev); 2776 lh = dev->dev_list.next; 2777 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2778 } 2779 2780 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2781 { 2782 struct list_head *lh; 2783 struct net *net; 2784 2785 net = dev_net(dev); 2786 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2787 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2788 } 2789 2790 static inline struct net_device *first_net_device(struct net *net) 2791 { 2792 return list_empty(&net->dev_base_head) ? NULL : 2793 net_device_entry(net->dev_base_head.next); 2794 } 2795 2796 static inline struct net_device *first_net_device_rcu(struct net *net) 2797 { 2798 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2799 2800 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2801 } 2802 2803 int netdev_boot_setup_check(struct net_device *dev); 2804 unsigned long netdev_boot_base(const char *prefix, int unit); 2805 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2806 const char *hwaddr); 2807 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2808 void dev_add_pack(struct packet_type *pt); 2809 void dev_remove_pack(struct packet_type *pt); 2810 void __dev_remove_pack(struct packet_type *pt); 2811 void dev_add_offload(struct packet_offload *po); 2812 void dev_remove_offload(struct packet_offload *po); 2813 2814 int dev_get_iflink(const struct net_device *dev); 2815 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2816 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2817 unsigned short mask); 2818 struct net_device *dev_get_by_name(struct net *net, const char *name); 2819 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2820 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2821 int dev_alloc_name(struct net_device *dev, const char *name); 2822 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2823 void dev_close(struct net_device *dev); 2824 void dev_close_many(struct list_head *head, bool unlink); 2825 void dev_disable_lro(struct net_device *dev); 2826 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2827 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2828 struct net_device *sb_dev); 2829 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2830 struct net_device *sb_dev); 2831 2832 int dev_queue_xmit(struct sk_buff *skb); 2833 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2834 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2835 2836 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2837 { 2838 int ret; 2839 2840 ret = __dev_direct_xmit(skb, queue_id); 2841 if (!dev_xmit_complete(ret)) 2842 kfree_skb(skb); 2843 return ret; 2844 } 2845 2846 int register_netdevice(struct net_device *dev); 2847 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2848 void unregister_netdevice_many(struct list_head *head); 2849 static inline void unregister_netdevice(struct net_device *dev) 2850 { 2851 unregister_netdevice_queue(dev, NULL); 2852 } 2853 2854 int netdev_refcnt_read(const struct net_device *dev); 2855 void free_netdev(struct net_device *dev); 2856 void netdev_freemem(struct net_device *dev); 2857 int init_dummy_netdev(struct net_device *dev); 2858 2859 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2860 struct sk_buff *skb, 2861 bool all_slaves); 2862 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2863 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2864 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2865 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2866 int netdev_get_name(struct net *net, char *name, int ifindex); 2867 int dev_restart(struct net_device *dev); 2868 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2869 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2870 2871 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2872 { 2873 return NAPI_GRO_CB(skb)->data_offset; 2874 } 2875 2876 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2877 { 2878 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2879 } 2880 2881 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2882 { 2883 NAPI_GRO_CB(skb)->data_offset += len; 2884 } 2885 2886 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2887 unsigned int offset) 2888 { 2889 return NAPI_GRO_CB(skb)->frag0 + offset; 2890 } 2891 2892 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2893 { 2894 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2895 } 2896 2897 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2898 { 2899 NAPI_GRO_CB(skb)->frag0 = NULL; 2900 NAPI_GRO_CB(skb)->frag0_len = 0; 2901 } 2902 2903 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2904 unsigned int offset) 2905 { 2906 if (!pskb_may_pull(skb, hlen)) 2907 return NULL; 2908 2909 skb_gro_frag0_invalidate(skb); 2910 return skb->data + offset; 2911 } 2912 2913 static inline void *skb_gro_network_header(struct sk_buff *skb) 2914 { 2915 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2916 skb_network_offset(skb); 2917 } 2918 2919 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2920 const void *start, unsigned int len) 2921 { 2922 if (NAPI_GRO_CB(skb)->csum_valid) 2923 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2924 csum_partial(start, len, 0)); 2925 } 2926 2927 /* GRO checksum functions. These are logical equivalents of the normal 2928 * checksum functions (in skbuff.h) except that they operate on the GRO 2929 * offsets and fields in sk_buff. 2930 */ 2931 2932 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2933 2934 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2935 { 2936 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2937 } 2938 2939 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2940 bool zero_okay, 2941 __sum16 check) 2942 { 2943 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2944 skb_checksum_start_offset(skb) < 2945 skb_gro_offset(skb)) && 2946 !skb_at_gro_remcsum_start(skb) && 2947 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2948 (!zero_okay || check)); 2949 } 2950 2951 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2952 __wsum psum) 2953 { 2954 if (NAPI_GRO_CB(skb)->csum_valid && 2955 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2956 return 0; 2957 2958 NAPI_GRO_CB(skb)->csum = psum; 2959 2960 return __skb_gro_checksum_complete(skb); 2961 } 2962 2963 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2964 { 2965 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2966 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2967 NAPI_GRO_CB(skb)->csum_cnt--; 2968 } else { 2969 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2970 * verified a new top level checksum or an encapsulated one 2971 * during GRO. This saves work if we fallback to normal path. 2972 */ 2973 __skb_incr_checksum_unnecessary(skb); 2974 } 2975 } 2976 2977 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2978 compute_pseudo) \ 2979 ({ \ 2980 __sum16 __ret = 0; \ 2981 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2982 __ret = __skb_gro_checksum_validate_complete(skb, \ 2983 compute_pseudo(skb, proto)); \ 2984 if (!__ret) \ 2985 skb_gro_incr_csum_unnecessary(skb); \ 2986 __ret; \ 2987 }) 2988 2989 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2990 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2991 2992 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2993 compute_pseudo) \ 2994 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2995 2996 #define skb_gro_checksum_simple_validate(skb) \ 2997 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2998 2999 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3000 { 3001 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3002 !NAPI_GRO_CB(skb)->csum_valid); 3003 } 3004 3005 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3006 __wsum pseudo) 3007 { 3008 NAPI_GRO_CB(skb)->csum = ~pseudo; 3009 NAPI_GRO_CB(skb)->csum_valid = 1; 3010 } 3011 3012 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3013 do { \ 3014 if (__skb_gro_checksum_convert_check(skb)) \ 3015 __skb_gro_checksum_convert(skb, \ 3016 compute_pseudo(skb, proto)); \ 3017 } while (0) 3018 3019 struct gro_remcsum { 3020 int offset; 3021 __wsum delta; 3022 }; 3023 3024 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3025 { 3026 grc->offset = 0; 3027 grc->delta = 0; 3028 } 3029 3030 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3031 unsigned int off, size_t hdrlen, 3032 int start, int offset, 3033 struct gro_remcsum *grc, 3034 bool nopartial) 3035 { 3036 __wsum delta; 3037 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3038 3039 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3040 3041 if (!nopartial) { 3042 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3043 return ptr; 3044 } 3045 3046 ptr = skb_gro_header_fast(skb, off); 3047 if (skb_gro_header_hard(skb, off + plen)) { 3048 ptr = skb_gro_header_slow(skb, off + plen, off); 3049 if (!ptr) 3050 return NULL; 3051 } 3052 3053 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3054 start, offset); 3055 3056 /* Adjust skb->csum since we changed the packet */ 3057 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3058 3059 grc->offset = off + hdrlen + offset; 3060 grc->delta = delta; 3061 3062 return ptr; 3063 } 3064 3065 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3066 struct gro_remcsum *grc) 3067 { 3068 void *ptr; 3069 size_t plen = grc->offset + sizeof(u16); 3070 3071 if (!grc->delta) 3072 return; 3073 3074 ptr = skb_gro_header_fast(skb, grc->offset); 3075 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3076 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3077 if (!ptr) 3078 return; 3079 } 3080 3081 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3082 } 3083 3084 #ifdef CONFIG_XFRM_OFFLOAD 3085 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3086 { 3087 if (PTR_ERR(pp) != -EINPROGRESS) 3088 NAPI_GRO_CB(skb)->flush |= flush; 3089 } 3090 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3091 struct sk_buff *pp, 3092 int flush, 3093 struct gro_remcsum *grc) 3094 { 3095 if (PTR_ERR(pp) != -EINPROGRESS) { 3096 NAPI_GRO_CB(skb)->flush |= flush; 3097 skb_gro_remcsum_cleanup(skb, grc); 3098 skb->remcsum_offload = 0; 3099 } 3100 } 3101 #else 3102 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3103 { 3104 NAPI_GRO_CB(skb)->flush |= flush; 3105 } 3106 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3107 struct sk_buff *pp, 3108 int flush, 3109 struct gro_remcsum *grc) 3110 { 3111 NAPI_GRO_CB(skb)->flush |= flush; 3112 skb_gro_remcsum_cleanup(skb, grc); 3113 skb->remcsum_offload = 0; 3114 } 3115 #endif 3116 3117 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3118 unsigned short type, 3119 const void *daddr, const void *saddr, 3120 unsigned int len) 3121 { 3122 if (!dev->header_ops || !dev->header_ops->create) 3123 return 0; 3124 3125 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3126 } 3127 3128 static inline int dev_parse_header(const struct sk_buff *skb, 3129 unsigned char *haddr) 3130 { 3131 const struct net_device *dev = skb->dev; 3132 3133 if (!dev->header_ops || !dev->header_ops->parse) 3134 return 0; 3135 return dev->header_ops->parse(skb, haddr); 3136 } 3137 3138 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3139 { 3140 const struct net_device *dev = skb->dev; 3141 3142 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3143 return 0; 3144 return dev->header_ops->parse_protocol(skb); 3145 } 3146 3147 /* ll_header must have at least hard_header_len allocated */ 3148 static inline bool dev_validate_header(const struct net_device *dev, 3149 char *ll_header, int len) 3150 { 3151 if (likely(len >= dev->hard_header_len)) 3152 return true; 3153 if (len < dev->min_header_len) 3154 return false; 3155 3156 if (capable(CAP_SYS_RAWIO)) { 3157 memset(ll_header + len, 0, dev->hard_header_len - len); 3158 return true; 3159 } 3160 3161 if (dev->header_ops && dev->header_ops->validate) 3162 return dev->header_ops->validate(ll_header, len); 3163 3164 return false; 3165 } 3166 3167 static inline bool dev_has_header(const struct net_device *dev) 3168 { 3169 return dev->header_ops && dev->header_ops->create; 3170 } 3171 3172 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3173 int len, int size); 3174 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3175 static inline int unregister_gifconf(unsigned int family) 3176 { 3177 return register_gifconf(family, NULL); 3178 } 3179 3180 #ifdef CONFIG_NET_FLOW_LIMIT 3181 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3182 struct sd_flow_limit { 3183 u64 count; 3184 unsigned int num_buckets; 3185 unsigned int history_head; 3186 u16 history[FLOW_LIMIT_HISTORY]; 3187 u8 buckets[]; 3188 }; 3189 3190 extern int netdev_flow_limit_table_len; 3191 #endif /* CONFIG_NET_FLOW_LIMIT */ 3192 3193 /* 3194 * Incoming packets are placed on per-CPU queues 3195 */ 3196 struct softnet_data { 3197 struct list_head poll_list; 3198 struct sk_buff_head process_queue; 3199 3200 /* stats */ 3201 unsigned int processed; 3202 unsigned int time_squeeze; 3203 unsigned int received_rps; 3204 #ifdef CONFIG_RPS 3205 struct softnet_data *rps_ipi_list; 3206 #endif 3207 #ifdef CONFIG_NET_FLOW_LIMIT 3208 struct sd_flow_limit __rcu *flow_limit; 3209 #endif 3210 struct Qdisc *output_queue; 3211 struct Qdisc **output_queue_tailp; 3212 struct sk_buff *completion_queue; 3213 #ifdef CONFIG_XFRM_OFFLOAD 3214 struct sk_buff_head xfrm_backlog; 3215 #endif 3216 /* written and read only by owning cpu: */ 3217 struct { 3218 u16 recursion; 3219 u8 more; 3220 } xmit; 3221 #ifdef CONFIG_RPS 3222 /* input_queue_head should be written by cpu owning this struct, 3223 * and only read by other cpus. Worth using a cache line. 3224 */ 3225 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3226 3227 /* Elements below can be accessed between CPUs for RPS/RFS */ 3228 call_single_data_t csd ____cacheline_aligned_in_smp; 3229 struct softnet_data *rps_ipi_next; 3230 unsigned int cpu; 3231 unsigned int input_queue_tail; 3232 #endif 3233 unsigned int dropped; 3234 struct sk_buff_head input_pkt_queue; 3235 struct napi_struct backlog; 3236 3237 }; 3238 3239 static inline void input_queue_head_incr(struct softnet_data *sd) 3240 { 3241 #ifdef CONFIG_RPS 3242 sd->input_queue_head++; 3243 #endif 3244 } 3245 3246 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3247 unsigned int *qtail) 3248 { 3249 #ifdef CONFIG_RPS 3250 *qtail = ++sd->input_queue_tail; 3251 #endif 3252 } 3253 3254 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3255 3256 static inline int dev_recursion_level(void) 3257 { 3258 return this_cpu_read(softnet_data.xmit.recursion); 3259 } 3260 3261 #define XMIT_RECURSION_LIMIT 8 3262 static inline bool dev_xmit_recursion(void) 3263 { 3264 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3265 XMIT_RECURSION_LIMIT); 3266 } 3267 3268 static inline void dev_xmit_recursion_inc(void) 3269 { 3270 __this_cpu_inc(softnet_data.xmit.recursion); 3271 } 3272 3273 static inline void dev_xmit_recursion_dec(void) 3274 { 3275 __this_cpu_dec(softnet_data.xmit.recursion); 3276 } 3277 3278 void __netif_schedule(struct Qdisc *q); 3279 void netif_schedule_queue(struct netdev_queue *txq); 3280 3281 static inline void netif_tx_schedule_all(struct net_device *dev) 3282 { 3283 unsigned int i; 3284 3285 for (i = 0; i < dev->num_tx_queues; i++) 3286 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3287 } 3288 3289 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3290 { 3291 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3292 } 3293 3294 /** 3295 * netif_start_queue - allow transmit 3296 * @dev: network device 3297 * 3298 * Allow upper layers to call the device hard_start_xmit routine. 3299 */ 3300 static inline void netif_start_queue(struct net_device *dev) 3301 { 3302 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3303 } 3304 3305 static inline void netif_tx_start_all_queues(struct net_device *dev) 3306 { 3307 unsigned int i; 3308 3309 for (i = 0; i < dev->num_tx_queues; i++) { 3310 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3311 netif_tx_start_queue(txq); 3312 } 3313 } 3314 3315 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3316 3317 /** 3318 * netif_wake_queue - restart transmit 3319 * @dev: network device 3320 * 3321 * Allow upper layers to call the device hard_start_xmit routine. 3322 * Used for flow control when transmit resources are available. 3323 */ 3324 static inline void netif_wake_queue(struct net_device *dev) 3325 { 3326 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3327 } 3328 3329 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3330 { 3331 unsigned int i; 3332 3333 for (i = 0; i < dev->num_tx_queues; i++) { 3334 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3335 netif_tx_wake_queue(txq); 3336 } 3337 } 3338 3339 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3340 { 3341 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3342 } 3343 3344 /** 3345 * netif_stop_queue - stop transmitted packets 3346 * @dev: network device 3347 * 3348 * Stop upper layers calling the device hard_start_xmit routine. 3349 * Used for flow control when transmit resources are unavailable. 3350 */ 3351 static inline void netif_stop_queue(struct net_device *dev) 3352 { 3353 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3354 } 3355 3356 void netif_tx_stop_all_queues(struct net_device *dev); 3357 3358 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3359 { 3360 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3361 } 3362 3363 /** 3364 * netif_queue_stopped - test if transmit queue is flowblocked 3365 * @dev: network device 3366 * 3367 * Test if transmit queue on device is currently unable to send. 3368 */ 3369 static inline bool netif_queue_stopped(const struct net_device *dev) 3370 { 3371 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3372 } 3373 3374 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3375 { 3376 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3377 } 3378 3379 static inline bool 3380 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3381 { 3382 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3383 } 3384 3385 static inline bool 3386 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3387 { 3388 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3389 } 3390 3391 /** 3392 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3393 * @dev_queue: pointer to transmit queue 3394 * 3395 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3396 * to give appropriate hint to the CPU. 3397 */ 3398 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3399 { 3400 #ifdef CONFIG_BQL 3401 prefetchw(&dev_queue->dql.num_queued); 3402 #endif 3403 } 3404 3405 /** 3406 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3407 * @dev_queue: pointer to transmit queue 3408 * 3409 * BQL enabled drivers might use this helper in their TX completion path, 3410 * to give appropriate hint to the CPU. 3411 */ 3412 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3413 { 3414 #ifdef CONFIG_BQL 3415 prefetchw(&dev_queue->dql.limit); 3416 #endif 3417 } 3418 3419 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3420 unsigned int bytes) 3421 { 3422 #ifdef CONFIG_BQL 3423 dql_queued(&dev_queue->dql, bytes); 3424 3425 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3426 return; 3427 3428 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3429 3430 /* 3431 * The XOFF flag must be set before checking the dql_avail below, 3432 * because in netdev_tx_completed_queue we update the dql_completed 3433 * before checking the XOFF flag. 3434 */ 3435 smp_mb(); 3436 3437 /* check again in case another CPU has just made room avail */ 3438 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3439 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3440 #endif 3441 } 3442 3443 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3444 * that they should not test BQL status themselves. 3445 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3446 * skb of a batch. 3447 * Returns true if the doorbell must be used to kick the NIC. 3448 */ 3449 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3450 unsigned int bytes, 3451 bool xmit_more) 3452 { 3453 if (xmit_more) { 3454 #ifdef CONFIG_BQL 3455 dql_queued(&dev_queue->dql, bytes); 3456 #endif 3457 return netif_tx_queue_stopped(dev_queue); 3458 } 3459 netdev_tx_sent_queue(dev_queue, bytes); 3460 return true; 3461 } 3462 3463 /** 3464 * netdev_sent_queue - report the number of bytes queued to hardware 3465 * @dev: network device 3466 * @bytes: number of bytes queued to the hardware device queue 3467 * 3468 * Report the number of bytes queued for sending/completion to the network 3469 * device hardware queue. @bytes should be a good approximation and should 3470 * exactly match netdev_completed_queue() @bytes 3471 */ 3472 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3473 { 3474 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3475 } 3476 3477 static inline bool __netdev_sent_queue(struct net_device *dev, 3478 unsigned int bytes, 3479 bool xmit_more) 3480 { 3481 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3482 xmit_more); 3483 } 3484 3485 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3486 unsigned int pkts, unsigned int bytes) 3487 { 3488 #ifdef CONFIG_BQL 3489 if (unlikely(!bytes)) 3490 return; 3491 3492 dql_completed(&dev_queue->dql, bytes); 3493 3494 /* 3495 * Without the memory barrier there is a small possiblity that 3496 * netdev_tx_sent_queue will miss the update and cause the queue to 3497 * be stopped forever 3498 */ 3499 smp_mb(); 3500 3501 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3502 return; 3503 3504 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3505 netif_schedule_queue(dev_queue); 3506 #endif 3507 } 3508 3509 /** 3510 * netdev_completed_queue - report bytes and packets completed by device 3511 * @dev: network device 3512 * @pkts: actual number of packets sent over the medium 3513 * @bytes: actual number of bytes sent over the medium 3514 * 3515 * Report the number of bytes and packets transmitted by the network device 3516 * hardware queue over the physical medium, @bytes must exactly match the 3517 * @bytes amount passed to netdev_sent_queue() 3518 */ 3519 static inline void netdev_completed_queue(struct net_device *dev, 3520 unsigned int pkts, unsigned int bytes) 3521 { 3522 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3523 } 3524 3525 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3526 { 3527 #ifdef CONFIG_BQL 3528 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3529 dql_reset(&q->dql); 3530 #endif 3531 } 3532 3533 /** 3534 * netdev_reset_queue - reset the packets and bytes count of a network device 3535 * @dev_queue: network device 3536 * 3537 * Reset the bytes and packet count of a network device and clear the 3538 * software flow control OFF bit for this network device 3539 */ 3540 static inline void netdev_reset_queue(struct net_device *dev_queue) 3541 { 3542 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3543 } 3544 3545 /** 3546 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3547 * @dev: network device 3548 * @queue_index: given tx queue index 3549 * 3550 * Returns 0 if given tx queue index >= number of device tx queues, 3551 * otherwise returns the originally passed tx queue index. 3552 */ 3553 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3554 { 3555 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3556 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3557 dev->name, queue_index, 3558 dev->real_num_tx_queues); 3559 return 0; 3560 } 3561 3562 return queue_index; 3563 } 3564 3565 /** 3566 * netif_running - test if up 3567 * @dev: network device 3568 * 3569 * Test if the device has been brought up. 3570 */ 3571 static inline bool netif_running(const struct net_device *dev) 3572 { 3573 return test_bit(__LINK_STATE_START, &dev->state); 3574 } 3575 3576 /* 3577 * Routines to manage the subqueues on a device. We only need start, 3578 * stop, and a check if it's stopped. All other device management is 3579 * done at the overall netdevice level. 3580 * Also test the device if we're multiqueue. 3581 */ 3582 3583 /** 3584 * netif_start_subqueue - allow sending packets on subqueue 3585 * @dev: network device 3586 * @queue_index: sub queue index 3587 * 3588 * Start individual transmit queue of a device with multiple transmit queues. 3589 */ 3590 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3591 { 3592 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3593 3594 netif_tx_start_queue(txq); 3595 } 3596 3597 /** 3598 * netif_stop_subqueue - stop sending packets on subqueue 3599 * @dev: network device 3600 * @queue_index: sub queue index 3601 * 3602 * Stop individual transmit queue of a device with multiple transmit queues. 3603 */ 3604 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3605 { 3606 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3607 netif_tx_stop_queue(txq); 3608 } 3609 3610 /** 3611 * __netif_subqueue_stopped - test status of subqueue 3612 * @dev: network device 3613 * @queue_index: sub queue index 3614 * 3615 * Check individual transmit queue of a device with multiple transmit queues. 3616 */ 3617 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3618 u16 queue_index) 3619 { 3620 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3621 3622 return netif_tx_queue_stopped(txq); 3623 } 3624 3625 /** 3626 * netif_subqueue_stopped - test status of subqueue 3627 * @dev: network device 3628 * @skb: sub queue buffer pointer 3629 * 3630 * Check individual transmit queue of a device with multiple transmit queues. 3631 */ 3632 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3633 struct sk_buff *skb) 3634 { 3635 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3636 } 3637 3638 /** 3639 * netif_wake_subqueue - allow sending packets on subqueue 3640 * @dev: network device 3641 * @queue_index: sub queue index 3642 * 3643 * Resume individual transmit queue of a device with multiple transmit queues. 3644 */ 3645 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3646 { 3647 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3648 3649 netif_tx_wake_queue(txq); 3650 } 3651 3652 #ifdef CONFIG_XPS 3653 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3654 u16 index); 3655 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3656 u16 index, bool is_rxqs_map); 3657 3658 /** 3659 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3660 * @j: CPU/Rx queue index 3661 * @mask: bitmask of all cpus/rx queues 3662 * @nr_bits: number of bits in the bitmask 3663 * 3664 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3665 */ 3666 static inline bool netif_attr_test_mask(unsigned long j, 3667 const unsigned long *mask, 3668 unsigned int nr_bits) 3669 { 3670 cpu_max_bits_warn(j, nr_bits); 3671 return test_bit(j, mask); 3672 } 3673 3674 /** 3675 * netif_attr_test_online - Test for online CPU/Rx queue 3676 * @j: CPU/Rx queue index 3677 * @online_mask: bitmask for CPUs/Rx queues that are online 3678 * @nr_bits: number of bits in the bitmask 3679 * 3680 * Returns true if a CPU/Rx queue is online. 3681 */ 3682 static inline bool netif_attr_test_online(unsigned long j, 3683 const unsigned long *online_mask, 3684 unsigned int nr_bits) 3685 { 3686 cpu_max_bits_warn(j, nr_bits); 3687 3688 if (online_mask) 3689 return test_bit(j, online_mask); 3690 3691 return (j < nr_bits); 3692 } 3693 3694 /** 3695 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3696 * @n: CPU/Rx queue index 3697 * @srcp: the cpumask/Rx queue mask pointer 3698 * @nr_bits: number of bits in the bitmask 3699 * 3700 * Returns >= nr_bits if no further CPUs/Rx queues set. 3701 */ 3702 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3703 unsigned int nr_bits) 3704 { 3705 /* -1 is a legal arg here. */ 3706 if (n != -1) 3707 cpu_max_bits_warn(n, nr_bits); 3708 3709 if (srcp) 3710 return find_next_bit(srcp, nr_bits, n + 1); 3711 3712 return n + 1; 3713 } 3714 3715 /** 3716 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3717 * @n: CPU/Rx queue index 3718 * @src1p: the first CPUs/Rx queues mask pointer 3719 * @src2p: the second CPUs/Rx queues mask pointer 3720 * @nr_bits: number of bits in the bitmask 3721 * 3722 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3723 */ 3724 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3725 const unsigned long *src2p, 3726 unsigned int nr_bits) 3727 { 3728 /* -1 is a legal arg here. */ 3729 if (n != -1) 3730 cpu_max_bits_warn(n, nr_bits); 3731 3732 if (src1p && src2p) 3733 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3734 else if (src1p) 3735 return find_next_bit(src1p, nr_bits, n + 1); 3736 else if (src2p) 3737 return find_next_bit(src2p, nr_bits, n + 1); 3738 3739 return n + 1; 3740 } 3741 #else 3742 static inline int netif_set_xps_queue(struct net_device *dev, 3743 const struct cpumask *mask, 3744 u16 index) 3745 { 3746 return 0; 3747 } 3748 3749 static inline int __netif_set_xps_queue(struct net_device *dev, 3750 const unsigned long *mask, 3751 u16 index, bool is_rxqs_map) 3752 { 3753 return 0; 3754 } 3755 #endif 3756 3757 /** 3758 * netif_is_multiqueue - test if device has multiple transmit queues 3759 * @dev: network device 3760 * 3761 * Check if device has multiple transmit queues 3762 */ 3763 static inline bool netif_is_multiqueue(const struct net_device *dev) 3764 { 3765 return dev->num_tx_queues > 1; 3766 } 3767 3768 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3769 3770 #ifdef CONFIG_SYSFS 3771 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3772 #else 3773 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3774 unsigned int rxqs) 3775 { 3776 dev->real_num_rx_queues = rxqs; 3777 return 0; 3778 } 3779 #endif 3780 3781 static inline struct netdev_rx_queue * 3782 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3783 { 3784 return dev->_rx + rxq; 3785 } 3786 3787 #ifdef CONFIG_SYSFS 3788 static inline unsigned int get_netdev_rx_queue_index( 3789 struct netdev_rx_queue *queue) 3790 { 3791 struct net_device *dev = queue->dev; 3792 int index = queue - dev->_rx; 3793 3794 BUG_ON(index >= dev->num_rx_queues); 3795 return index; 3796 } 3797 #endif 3798 3799 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3800 int netif_get_num_default_rss_queues(void); 3801 3802 enum skb_free_reason { 3803 SKB_REASON_CONSUMED, 3804 SKB_REASON_DROPPED, 3805 }; 3806 3807 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3808 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3809 3810 /* 3811 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3812 * interrupt context or with hardware interrupts being disabled. 3813 * (in_irq() || irqs_disabled()) 3814 * 3815 * We provide four helpers that can be used in following contexts : 3816 * 3817 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3818 * replacing kfree_skb(skb) 3819 * 3820 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3821 * Typically used in place of consume_skb(skb) in TX completion path 3822 * 3823 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3824 * replacing kfree_skb(skb) 3825 * 3826 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3827 * and consumed a packet. Used in place of consume_skb(skb) 3828 */ 3829 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3830 { 3831 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3832 } 3833 3834 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3835 { 3836 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3837 } 3838 3839 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3840 { 3841 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3842 } 3843 3844 static inline void dev_consume_skb_any(struct sk_buff *skb) 3845 { 3846 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3847 } 3848 3849 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3850 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3851 int netif_rx(struct sk_buff *skb); 3852 int netif_rx_ni(struct sk_buff *skb); 3853 int netif_rx_any_context(struct sk_buff *skb); 3854 int netif_receive_skb(struct sk_buff *skb); 3855 int netif_receive_skb_core(struct sk_buff *skb); 3856 void netif_receive_skb_list(struct list_head *head); 3857 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3858 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3859 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3860 gro_result_t napi_gro_frags(struct napi_struct *napi); 3861 struct packet_offload *gro_find_receive_by_type(__be16 type); 3862 struct packet_offload *gro_find_complete_by_type(__be16 type); 3863 3864 static inline void napi_free_frags(struct napi_struct *napi) 3865 { 3866 kfree_skb(napi->skb); 3867 napi->skb = NULL; 3868 } 3869 3870 bool netdev_is_rx_handler_busy(struct net_device *dev); 3871 int netdev_rx_handler_register(struct net_device *dev, 3872 rx_handler_func_t *rx_handler, 3873 void *rx_handler_data); 3874 void netdev_rx_handler_unregister(struct net_device *dev); 3875 3876 bool dev_valid_name(const char *name); 3877 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3878 bool *need_copyout); 3879 int dev_ifconf(struct net *net, struct ifconf *, int); 3880 int dev_ethtool(struct net *net, struct ifreq *); 3881 unsigned int dev_get_flags(const struct net_device *); 3882 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3883 struct netlink_ext_ack *extack); 3884 int dev_change_flags(struct net_device *dev, unsigned int flags, 3885 struct netlink_ext_ack *extack); 3886 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3887 unsigned int gchanges); 3888 int dev_change_name(struct net_device *, const char *); 3889 int dev_set_alias(struct net_device *, const char *, size_t); 3890 int dev_get_alias(const struct net_device *, char *, size_t); 3891 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3892 int __dev_set_mtu(struct net_device *, int); 3893 int dev_validate_mtu(struct net_device *dev, int mtu, 3894 struct netlink_ext_ack *extack); 3895 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3896 struct netlink_ext_ack *extack); 3897 int dev_set_mtu(struct net_device *, int); 3898 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3899 void dev_set_group(struct net_device *, int); 3900 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3901 struct netlink_ext_ack *extack); 3902 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3903 struct netlink_ext_ack *extack); 3904 int dev_change_carrier(struct net_device *, bool new_carrier); 3905 int dev_get_phys_port_id(struct net_device *dev, 3906 struct netdev_phys_item_id *ppid); 3907 int dev_get_phys_port_name(struct net_device *dev, 3908 char *name, size_t len); 3909 int dev_get_port_parent_id(struct net_device *dev, 3910 struct netdev_phys_item_id *ppid, bool recurse); 3911 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3912 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3913 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3914 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3915 u32 value); 3916 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3917 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3918 struct netdev_queue *txq, int *ret); 3919 3920 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3921 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3922 int fd, int expected_fd, u32 flags); 3923 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3924 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3925 3926 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3927 3928 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3929 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3930 bool is_skb_forwardable(const struct net_device *dev, 3931 const struct sk_buff *skb); 3932 3933 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3934 struct sk_buff *skb) 3935 { 3936 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3937 unlikely(!is_skb_forwardable(dev, skb))) { 3938 atomic_long_inc(&dev->rx_dropped); 3939 kfree_skb(skb); 3940 return NET_RX_DROP; 3941 } 3942 3943 skb_scrub_packet(skb, true); 3944 skb->priority = 0; 3945 return 0; 3946 } 3947 3948 bool dev_nit_active(struct net_device *dev); 3949 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3950 3951 extern int netdev_budget; 3952 extern unsigned int netdev_budget_usecs; 3953 3954 /* Called by rtnetlink.c:rtnl_unlock() */ 3955 void netdev_run_todo(void); 3956 3957 /** 3958 * dev_put - release reference to device 3959 * @dev: network device 3960 * 3961 * Release reference to device to allow it to be freed. 3962 */ 3963 static inline void dev_put(struct net_device *dev) 3964 { 3965 this_cpu_dec(*dev->pcpu_refcnt); 3966 } 3967 3968 /** 3969 * dev_hold - get reference to device 3970 * @dev: network device 3971 * 3972 * Hold reference to device to keep it from being freed. 3973 */ 3974 static inline void dev_hold(struct net_device *dev) 3975 { 3976 this_cpu_inc(*dev->pcpu_refcnt); 3977 } 3978 3979 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3980 * and _off may be called from IRQ context, but it is caller 3981 * who is responsible for serialization of these calls. 3982 * 3983 * The name carrier is inappropriate, these functions should really be 3984 * called netif_lowerlayer_*() because they represent the state of any 3985 * kind of lower layer not just hardware media. 3986 */ 3987 3988 void linkwatch_init_dev(struct net_device *dev); 3989 void linkwatch_fire_event(struct net_device *dev); 3990 void linkwatch_forget_dev(struct net_device *dev); 3991 3992 /** 3993 * netif_carrier_ok - test if carrier present 3994 * @dev: network device 3995 * 3996 * Check if carrier is present on device 3997 */ 3998 static inline bool netif_carrier_ok(const struct net_device *dev) 3999 { 4000 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4001 } 4002 4003 unsigned long dev_trans_start(struct net_device *dev); 4004 4005 void __netdev_watchdog_up(struct net_device *dev); 4006 4007 void netif_carrier_on(struct net_device *dev); 4008 4009 void netif_carrier_off(struct net_device *dev); 4010 4011 /** 4012 * netif_dormant_on - mark device as dormant. 4013 * @dev: network device 4014 * 4015 * Mark device as dormant (as per RFC2863). 4016 * 4017 * The dormant state indicates that the relevant interface is not 4018 * actually in a condition to pass packets (i.e., it is not 'up') but is 4019 * in a "pending" state, waiting for some external event. For "on- 4020 * demand" interfaces, this new state identifies the situation where the 4021 * interface is waiting for events to place it in the up state. 4022 */ 4023 static inline void netif_dormant_on(struct net_device *dev) 4024 { 4025 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4026 linkwatch_fire_event(dev); 4027 } 4028 4029 /** 4030 * netif_dormant_off - set device as not dormant. 4031 * @dev: network device 4032 * 4033 * Device is not in dormant state. 4034 */ 4035 static inline void netif_dormant_off(struct net_device *dev) 4036 { 4037 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4038 linkwatch_fire_event(dev); 4039 } 4040 4041 /** 4042 * netif_dormant - test if device is dormant 4043 * @dev: network device 4044 * 4045 * Check if device is dormant. 4046 */ 4047 static inline bool netif_dormant(const struct net_device *dev) 4048 { 4049 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4050 } 4051 4052 4053 /** 4054 * netif_testing_on - mark device as under test. 4055 * @dev: network device 4056 * 4057 * Mark device as under test (as per RFC2863). 4058 * 4059 * The testing state indicates that some test(s) must be performed on 4060 * the interface. After completion, of the test, the interface state 4061 * will change to up, dormant, or down, as appropriate. 4062 */ 4063 static inline void netif_testing_on(struct net_device *dev) 4064 { 4065 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4066 linkwatch_fire_event(dev); 4067 } 4068 4069 /** 4070 * netif_testing_off - set device as not under test. 4071 * @dev: network device 4072 * 4073 * Device is not in testing state. 4074 */ 4075 static inline void netif_testing_off(struct net_device *dev) 4076 { 4077 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4078 linkwatch_fire_event(dev); 4079 } 4080 4081 /** 4082 * netif_testing - test if device is under test 4083 * @dev: network device 4084 * 4085 * Check if device is under test 4086 */ 4087 static inline bool netif_testing(const struct net_device *dev) 4088 { 4089 return test_bit(__LINK_STATE_TESTING, &dev->state); 4090 } 4091 4092 4093 /** 4094 * netif_oper_up - test if device is operational 4095 * @dev: network device 4096 * 4097 * Check if carrier is operational 4098 */ 4099 static inline bool netif_oper_up(const struct net_device *dev) 4100 { 4101 return (dev->operstate == IF_OPER_UP || 4102 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4103 } 4104 4105 /** 4106 * netif_device_present - is device available or removed 4107 * @dev: network device 4108 * 4109 * Check if device has not been removed from system. 4110 */ 4111 static inline bool netif_device_present(struct net_device *dev) 4112 { 4113 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4114 } 4115 4116 void netif_device_detach(struct net_device *dev); 4117 4118 void netif_device_attach(struct net_device *dev); 4119 4120 /* 4121 * Network interface message level settings 4122 */ 4123 4124 enum { 4125 NETIF_MSG_DRV_BIT, 4126 NETIF_MSG_PROBE_BIT, 4127 NETIF_MSG_LINK_BIT, 4128 NETIF_MSG_TIMER_BIT, 4129 NETIF_MSG_IFDOWN_BIT, 4130 NETIF_MSG_IFUP_BIT, 4131 NETIF_MSG_RX_ERR_BIT, 4132 NETIF_MSG_TX_ERR_BIT, 4133 NETIF_MSG_TX_QUEUED_BIT, 4134 NETIF_MSG_INTR_BIT, 4135 NETIF_MSG_TX_DONE_BIT, 4136 NETIF_MSG_RX_STATUS_BIT, 4137 NETIF_MSG_PKTDATA_BIT, 4138 NETIF_MSG_HW_BIT, 4139 NETIF_MSG_WOL_BIT, 4140 4141 /* When you add a new bit above, update netif_msg_class_names array 4142 * in net/ethtool/common.c 4143 */ 4144 NETIF_MSG_CLASS_COUNT, 4145 }; 4146 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4147 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4148 4149 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4150 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4151 4152 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4153 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4154 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4155 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4156 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4157 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4158 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4159 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4160 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4161 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4162 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4163 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4164 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4165 #define NETIF_MSG_HW __NETIF_MSG(HW) 4166 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4167 4168 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4169 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4170 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4171 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4172 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4173 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4174 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4175 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4176 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4177 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4178 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4179 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4180 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4181 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4182 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4183 4184 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4185 { 4186 /* use default */ 4187 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4188 return default_msg_enable_bits; 4189 if (debug_value == 0) /* no output */ 4190 return 0; 4191 /* set low N bits */ 4192 return (1U << debug_value) - 1; 4193 } 4194 4195 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4196 { 4197 spin_lock(&txq->_xmit_lock); 4198 txq->xmit_lock_owner = cpu; 4199 } 4200 4201 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4202 { 4203 __acquire(&txq->_xmit_lock); 4204 return true; 4205 } 4206 4207 static inline void __netif_tx_release(struct netdev_queue *txq) 4208 { 4209 __release(&txq->_xmit_lock); 4210 } 4211 4212 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4213 { 4214 spin_lock_bh(&txq->_xmit_lock); 4215 txq->xmit_lock_owner = smp_processor_id(); 4216 } 4217 4218 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4219 { 4220 bool ok = spin_trylock(&txq->_xmit_lock); 4221 if (likely(ok)) 4222 txq->xmit_lock_owner = smp_processor_id(); 4223 return ok; 4224 } 4225 4226 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4227 { 4228 txq->xmit_lock_owner = -1; 4229 spin_unlock(&txq->_xmit_lock); 4230 } 4231 4232 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4233 { 4234 txq->xmit_lock_owner = -1; 4235 spin_unlock_bh(&txq->_xmit_lock); 4236 } 4237 4238 static inline void txq_trans_update(struct netdev_queue *txq) 4239 { 4240 if (txq->xmit_lock_owner != -1) 4241 txq->trans_start = jiffies; 4242 } 4243 4244 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4245 static inline void netif_trans_update(struct net_device *dev) 4246 { 4247 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4248 4249 if (txq->trans_start != jiffies) 4250 txq->trans_start = jiffies; 4251 } 4252 4253 /** 4254 * netif_tx_lock - grab network device transmit lock 4255 * @dev: network device 4256 * 4257 * Get network device transmit lock 4258 */ 4259 static inline void netif_tx_lock(struct net_device *dev) 4260 { 4261 unsigned int i; 4262 int cpu; 4263 4264 spin_lock(&dev->tx_global_lock); 4265 cpu = smp_processor_id(); 4266 for (i = 0; i < dev->num_tx_queues; i++) { 4267 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4268 4269 /* We are the only thread of execution doing a 4270 * freeze, but we have to grab the _xmit_lock in 4271 * order to synchronize with threads which are in 4272 * the ->hard_start_xmit() handler and already 4273 * checked the frozen bit. 4274 */ 4275 __netif_tx_lock(txq, cpu); 4276 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4277 __netif_tx_unlock(txq); 4278 } 4279 } 4280 4281 static inline void netif_tx_lock_bh(struct net_device *dev) 4282 { 4283 local_bh_disable(); 4284 netif_tx_lock(dev); 4285 } 4286 4287 static inline void netif_tx_unlock(struct net_device *dev) 4288 { 4289 unsigned int i; 4290 4291 for (i = 0; i < dev->num_tx_queues; i++) { 4292 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4293 4294 /* No need to grab the _xmit_lock here. If the 4295 * queue is not stopped for another reason, we 4296 * force a schedule. 4297 */ 4298 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4299 netif_schedule_queue(txq); 4300 } 4301 spin_unlock(&dev->tx_global_lock); 4302 } 4303 4304 static inline void netif_tx_unlock_bh(struct net_device *dev) 4305 { 4306 netif_tx_unlock(dev); 4307 local_bh_enable(); 4308 } 4309 4310 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4311 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4312 __netif_tx_lock(txq, cpu); \ 4313 } else { \ 4314 __netif_tx_acquire(txq); \ 4315 } \ 4316 } 4317 4318 #define HARD_TX_TRYLOCK(dev, txq) \ 4319 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4320 __netif_tx_trylock(txq) : \ 4321 __netif_tx_acquire(txq)) 4322 4323 #define HARD_TX_UNLOCK(dev, txq) { \ 4324 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4325 __netif_tx_unlock(txq); \ 4326 } else { \ 4327 __netif_tx_release(txq); \ 4328 } \ 4329 } 4330 4331 static inline void netif_tx_disable(struct net_device *dev) 4332 { 4333 unsigned int i; 4334 int cpu; 4335 4336 local_bh_disable(); 4337 cpu = smp_processor_id(); 4338 for (i = 0; i < dev->num_tx_queues; i++) { 4339 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4340 4341 __netif_tx_lock(txq, cpu); 4342 netif_tx_stop_queue(txq); 4343 __netif_tx_unlock(txq); 4344 } 4345 local_bh_enable(); 4346 } 4347 4348 static inline void netif_addr_lock(struct net_device *dev) 4349 { 4350 unsigned char nest_level = 0; 4351 4352 #ifdef CONFIG_LOCKDEP 4353 nest_level = dev->nested_level; 4354 #endif 4355 spin_lock_nested(&dev->addr_list_lock, nest_level); 4356 } 4357 4358 static inline void netif_addr_lock_bh(struct net_device *dev) 4359 { 4360 unsigned char nest_level = 0; 4361 4362 #ifdef CONFIG_LOCKDEP 4363 nest_level = dev->nested_level; 4364 #endif 4365 local_bh_disable(); 4366 spin_lock_nested(&dev->addr_list_lock, nest_level); 4367 } 4368 4369 static inline void netif_addr_unlock(struct net_device *dev) 4370 { 4371 spin_unlock(&dev->addr_list_lock); 4372 } 4373 4374 static inline void netif_addr_unlock_bh(struct net_device *dev) 4375 { 4376 spin_unlock_bh(&dev->addr_list_lock); 4377 } 4378 4379 /* 4380 * dev_addrs walker. Should be used only for read access. Call with 4381 * rcu_read_lock held. 4382 */ 4383 #define for_each_dev_addr(dev, ha) \ 4384 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4385 4386 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4387 4388 void ether_setup(struct net_device *dev); 4389 4390 /* Support for loadable net-drivers */ 4391 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4392 unsigned char name_assign_type, 4393 void (*setup)(struct net_device *), 4394 unsigned int txqs, unsigned int rxqs); 4395 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4396 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4397 4398 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4399 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4400 count) 4401 4402 int register_netdev(struct net_device *dev); 4403 void unregister_netdev(struct net_device *dev); 4404 4405 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4406 4407 /* General hardware address lists handling functions */ 4408 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4409 struct netdev_hw_addr_list *from_list, int addr_len); 4410 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4411 struct netdev_hw_addr_list *from_list, int addr_len); 4412 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4413 struct net_device *dev, 4414 int (*sync)(struct net_device *, const unsigned char *), 4415 int (*unsync)(struct net_device *, 4416 const unsigned char *)); 4417 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4418 struct net_device *dev, 4419 int (*sync)(struct net_device *, 4420 const unsigned char *, int), 4421 int (*unsync)(struct net_device *, 4422 const unsigned char *, int)); 4423 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4424 struct net_device *dev, 4425 int (*unsync)(struct net_device *, 4426 const unsigned char *, int)); 4427 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4428 struct net_device *dev, 4429 int (*unsync)(struct net_device *, 4430 const unsigned char *)); 4431 void __hw_addr_init(struct netdev_hw_addr_list *list); 4432 4433 /* Functions used for device addresses handling */ 4434 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4435 unsigned char addr_type); 4436 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4437 unsigned char addr_type); 4438 void dev_addr_flush(struct net_device *dev); 4439 int dev_addr_init(struct net_device *dev); 4440 4441 /* Functions used for unicast addresses handling */ 4442 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4443 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4444 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4445 int dev_uc_sync(struct net_device *to, struct net_device *from); 4446 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4447 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4448 void dev_uc_flush(struct net_device *dev); 4449 void dev_uc_init(struct net_device *dev); 4450 4451 /** 4452 * __dev_uc_sync - Synchonize device's unicast list 4453 * @dev: device to sync 4454 * @sync: function to call if address should be added 4455 * @unsync: function to call if address should be removed 4456 * 4457 * Add newly added addresses to the interface, and release 4458 * addresses that have been deleted. 4459 */ 4460 static inline int __dev_uc_sync(struct net_device *dev, 4461 int (*sync)(struct net_device *, 4462 const unsigned char *), 4463 int (*unsync)(struct net_device *, 4464 const unsigned char *)) 4465 { 4466 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4467 } 4468 4469 /** 4470 * __dev_uc_unsync - Remove synchronized addresses from device 4471 * @dev: device to sync 4472 * @unsync: function to call if address should be removed 4473 * 4474 * Remove all addresses that were added to the device by dev_uc_sync(). 4475 */ 4476 static inline void __dev_uc_unsync(struct net_device *dev, 4477 int (*unsync)(struct net_device *, 4478 const unsigned char *)) 4479 { 4480 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4481 } 4482 4483 /* Functions used for multicast addresses handling */ 4484 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4485 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4486 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4487 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4488 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4489 int dev_mc_sync(struct net_device *to, struct net_device *from); 4490 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4491 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4492 void dev_mc_flush(struct net_device *dev); 4493 void dev_mc_init(struct net_device *dev); 4494 4495 /** 4496 * __dev_mc_sync - Synchonize device's multicast list 4497 * @dev: device to sync 4498 * @sync: function to call if address should be added 4499 * @unsync: function to call if address should be removed 4500 * 4501 * Add newly added addresses to the interface, and release 4502 * addresses that have been deleted. 4503 */ 4504 static inline int __dev_mc_sync(struct net_device *dev, 4505 int (*sync)(struct net_device *, 4506 const unsigned char *), 4507 int (*unsync)(struct net_device *, 4508 const unsigned char *)) 4509 { 4510 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4511 } 4512 4513 /** 4514 * __dev_mc_unsync - Remove synchronized addresses from device 4515 * @dev: device to sync 4516 * @unsync: function to call if address should be removed 4517 * 4518 * Remove all addresses that were added to the device by dev_mc_sync(). 4519 */ 4520 static inline void __dev_mc_unsync(struct net_device *dev, 4521 int (*unsync)(struct net_device *, 4522 const unsigned char *)) 4523 { 4524 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4525 } 4526 4527 /* Functions used for secondary unicast and multicast support */ 4528 void dev_set_rx_mode(struct net_device *dev); 4529 void __dev_set_rx_mode(struct net_device *dev); 4530 int dev_set_promiscuity(struct net_device *dev, int inc); 4531 int dev_set_allmulti(struct net_device *dev, int inc); 4532 void netdev_state_change(struct net_device *dev); 4533 void __netdev_notify_peers(struct net_device *dev); 4534 void netdev_notify_peers(struct net_device *dev); 4535 void netdev_features_change(struct net_device *dev); 4536 /* Load a device via the kmod */ 4537 void dev_load(struct net *net, const char *name); 4538 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4539 struct rtnl_link_stats64 *storage); 4540 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4541 const struct net_device_stats *netdev_stats); 4542 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4543 const struct pcpu_sw_netstats __percpu *netstats); 4544 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4545 4546 extern int netdev_max_backlog; 4547 extern int netdev_tstamp_prequeue; 4548 extern int weight_p; 4549 extern int dev_weight_rx_bias; 4550 extern int dev_weight_tx_bias; 4551 extern int dev_rx_weight; 4552 extern int dev_tx_weight; 4553 extern int gro_normal_batch; 4554 4555 enum { 4556 NESTED_SYNC_IMM_BIT, 4557 NESTED_SYNC_TODO_BIT, 4558 }; 4559 4560 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4561 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4562 4563 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4564 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4565 4566 struct netdev_nested_priv { 4567 unsigned char flags; 4568 void *data; 4569 }; 4570 4571 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4572 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4573 struct list_head **iter); 4574 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4575 struct list_head **iter); 4576 4577 #ifdef CONFIG_LOCKDEP 4578 static LIST_HEAD(net_unlink_list); 4579 4580 static inline void net_unlink_todo(struct net_device *dev) 4581 { 4582 if (list_empty(&dev->unlink_list)) 4583 list_add_tail(&dev->unlink_list, &net_unlink_list); 4584 } 4585 #endif 4586 4587 /* iterate through upper list, must be called under RCU read lock */ 4588 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4589 for (iter = &(dev)->adj_list.upper, \ 4590 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4591 updev; \ 4592 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4593 4594 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4595 int (*fn)(struct net_device *upper_dev, 4596 struct netdev_nested_priv *priv), 4597 struct netdev_nested_priv *priv); 4598 4599 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4600 struct net_device *upper_dev); 4601 4602 bool netdev_has_any_upper_dev(struct net_device *dev); 4603 4604 void *netdev_lower_get_next_private(struct net_device *dev, 4605 struct list_head **iter); 4606 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4607 struct list_head **iter); 4608 4609 #define netdev_for_each_lower_private(dev, priv, iter) \ 4610 for (iter = (dev)->adj_list.lower.next, \ 4611 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4612 priv; \ 4613 priv = netdev_lower_get_next_private(dev, &(iter))) 4614 4615 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4616 for (iter = &(dev)->adj_list.lower, \ 4617 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4618 priv; \ 4619 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4620 4621 void *netdev_lower_get_next(struct net_device *dev, 4622 struct list_head **iter); 4623 4624 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4625 for (iter = (dev)->adj_list.lower.next, \ 4626 ldev = netdev_lower_get_next(dev, &(iter)); \ 4627 ldev; \ 4628 ldev = netdev_lower_get_next(dev, &(iter))) 4629 4630 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4631 struct list_head **iter); 4632 int netdev_walk_all_lower_dev(struct net_device *dev, 4633 int (*fn)(struct net_device *lower_dev, 4634 struct netdev_nested_priv *priv), 4635 struct netdev_nested_priv *priv); 4636 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4637 int (*fn)(struct net_device *lower_dev, 4638 struct netdev_nested_priv *priv), 4639 struct netdev_nested_priv *priv); 4640 4641 void *netdev_adjacent_get_private(struct list_head *adj_list); 4642 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4643 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4644 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4645 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4646 struct netlink_ext_ack *extack); 4647 int netdev_master_upper_dev_link(struct net_device *dev, 4648 struct net_device *upper_dev, 4649 void *upper_priv, void *upper_info, 4650 struct netlink_ext_ack *extack); 4651 void netdev_upper_dev_unlink(struct net_device *dev, 4652 struct net_device *upper_dev); 4653 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4654 struct net_device *new_dev, 4655 struct net_device *dev, 4656 struct netlink_ext_ack *extack); 4657 void netdev_adjacent_change_commit(struct net_device *old_dev, 4658 struct net_device *new_dev, 4659 struct net_device *dev); 4660 void netdev_adjacent_change_abort(struct net_device *old_dev, 4661 struct net_device *new_dev, 4662 struct net_device *dev); 4663 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4664 void *netdev_lower_dev_get_private(struct net_device *dev, 4665 struct net_device *lower_dev); 4666 void netdev_lower_state_changed(struct net_device *lower_dev, 4667 void *lower_state_info); 4668 4669 /* RSS keys are 40 or 52 bytes long */ 4670 #define NETDEV_RSS_KEY_LEN 52 4671 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4672 void netdev_rss_key_fill(void *buffer, size_t len); 4673 4674 int skb_checksum_help(struct sk_buff *skb); 4675 int skb_crc32c_csum_help(struct sk_buff *skb); 4676 int skb_csum_hwoffload_help(struct sk_buff *skb, 4677 const netdev_features_t features); 4678 4679 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4680 netdev_features_t features, bool tx_path); 4681 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4682 netdev_features_t features); 4683 4684 struct netdev_bonding_info { 4685 ifslave slave; 4686 ifbond master; 4687 }; 4688 4689 struct netdev_notifier_bonding_info { 4690 struct netdev_notifier_info info; /* must be first */ 4691 struct netdev_bonding_info bonding_info; 4692 }; 4693 4694 void netdev_bonding_info_change(struct net_device *dev, 4695 struct netdev_bonding_info *bonding_info); 4696 4697 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4698 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4699 #else 4700 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4701 const void *data) 4702 { 4703 } 4704 #endif 4705 4706 static inline 4707 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4708 { 4709 return __skb_gso_segment(skb, features, true); 4710 } 4711 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4712 4713 static inline bool can_checksum_protocol(netdev_features_t features, 4714 __be16 protocol) 4715 { 4716 if (protocol == htons(ETH_P_FCOE)) 4717 return !!(features & NETIF_F_FCOE_CRC); 4718 4719 /* Assume this is an IP checksum (not SCTP CRC) */ 4720 4721 if (features & NETIF_F_HW_CSUM) { 4722 /* Can checksum everything */ 4723 return true; 4724 } 4725 4726 switch (protocol) { 4727 case htons(ETH_P_IP): 4728 return !!(features & NETIF_F_IP_CSUM); 4729 case htons(ETH_P_IPV6): 4730 return !!(features & NETIF_F_IPV6_CSUM); 4731 default: 4732 return false; 4733 } 4734 } 4735 4736 #ifdef CONFIG_BUG 4737 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4738 #else 4739 static inline void netdev_rx_csum_fault(struct net_device *dev, 4740 struct sk_buff *skb) 4741 { 4742 } 4743 #endif 4744 /* rx skb timestamps */ 4745 void net_enable_timestamp(void); 4746 void net_disable_timestamp(void); 4747 4748 #ifdef CONFIG_PROC_FS 4749 int __init dev_proc_init(void); 4750 #else 4751 #define dev_proc_init() 0 4752 #endif 4753 4754 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4755 struct sk_buff *skb, struct net_device *dev, 4756 bool more) 4757 { 4758 __this_cpu_write(softnet_data.xmit.more, more); 4759 return ops->ndo_start_xmit(skb, dev); 4760 } 4761 4762 static inline bool netdev_xmit_more(void) 4763 { 4764 return __this_cpu_read(softnet_data.xmit.more); 4765 } 4766 4767 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4768 struct netdev_queue *txq, bool more) 4769 { 4770 const struct net_device_ops *ops = dev->netdev_ops; 4771 netdev_tx_t rc; 4772 4773 rc = __netdev_start_xmit(ops, skb, dev, more); 4774 if (rc == NETDEV_TX_OK) 4775 txq_trans_update(txq); 4776 4777 return rc; 4778 } 4779 4780 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4781 const void *ns); 4782 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4783 const void *ns); 4784 4785 extern const struct kobj_ns_type_operations net_ns_type_operations; 4786 4787 const char *netdev_drivername(const struct net_device *dev); 4788 4789 void linkwatch_run_queue(void); 4790 4791 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4792 netdev_features_t f2) 4793 { 4794 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4795 if (f1 & NETIF_F_HW_CSUM) 4796 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4797 else 4798 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4799 } 4800 4801 return f1 & f2; 4802 } 4803 4804 static inline netdev_features_t netdev_get_wanted_features( 4805 struct net_device *dev) 4806 { 4807 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4808 } 4809 netdev_features_t netdev_increment_features(netdev_features_t all, 4810 netdev_features_t one, netdev_features_t mask); 4811 4812 /* Allow TSO being used on stacked device : 4813 * Performing the GSO segmentation before last device 4814 * is a performance improvement. 4815 */ 4816 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4817 netdev_features_t mask) 4818 { 4819 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4820 } 4821 4822 int __netdev_update_features(struct net_device *dev); 4823 void netdev_update_features(struct net_device *dev); 4824 void netdev_change_features(struct net_device *dev); 4825 4826 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4827 struct net_device *dev); 4828 4829 netdev_features_t passthru_features_check(struct sk_buff *skb, 4830 struct net_device *dev, 4831 netdev_features_t features); 4832 netdev_features_t netif_skb_features(struct sk_buff *skb); 4833 4834 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4835 { 4836 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4837 4838 /* check flags correspondence */ 4839 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4840 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4841 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4842 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4843 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4844 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4845 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4846 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4847 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4848 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4849 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4850 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4851 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4852 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4853 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4854 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4855 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4856 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4857 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4858 4859 return (features & feature) == feature; 4860 } 4861 4862 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4863 { 4864 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4865 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4866 } 4867 4868 static inline bool netif_needs_gso(struct sk_buff *skb, 4869 netdev_features_t features) 4870 { 4871 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4872 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4873 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4874 } 4875 4876 static inline void netif_set_gso_max_size(struct net_device *dev, 4877 unsigned int size) 4878 { 4879 dev->gso_max_size = size; 4880 } 4881 4882 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4883 int pulled_hlen, u16 mac_offset, 4884 int mac_len) 4885 { 4886 skb->protocol = protocol; 4887 skb->encapsulation = 1; 4888 skb_push(skb, pulled_hlen); 4889 skb_reset_transport_header(skb); 4890 skb->mac_header = mac_offset; 4891 skb->network_header = skb->mac_header + mac_len; 4892 skb->mac_len = mac_len; 4893 } 4894 4895 static inline bool netif_is_macsec(const struct net_device *dev) 4896 { 4897 return dev->priv_flags & IFF_MACSEC; 4898 } 4899 4900 static inline bool netif_is_macvlan(const struct net_device *dev) 4901 { 4902 return dev->priv_flags & IFF_MACVLAN; 4903 } 4904 4905 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4906 { 4907 return dev->priv_flags & IFF_MACVLAN_PORT; 4908 } 4909 4910 static inline bool netif_is_bond_master(const struct net_device *dev) 4911 { 4912 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4913 } 4914 4915 static inline bool netif_is_bond_slave(const struct net_device *dev) 4916 { 4917 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4918 } 4919 4920 static inline bool netif_supports_nofcs(struct net_device *dev) 4921 { 4922 return dev->priv_flags & IFF_SUPP_NOFCS; 4923 } 4924 4925 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4926 { 4927 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4928 } 4929 4930 static inline bool netif_is_l3_master(const struct net_device *dev) 4931 { 4932 return dev->priv_flags & IFF_L3MDEV_MASTER; 4933 } 4934 4935 static inline bool netif_is_l3_slave(const struct net_device *dev) 4936 { 4937 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4938 } 4939 4940 static inline bool netif_is_bridge_master(const struct net_device *dev) 4941 { 4942 return dev->priv_flags & IFF_EBRIDGE; 4943 } 4944 4945 static inline bool netif_is_bridge_port(const struct net_device *dev) 4946 { 4947 return dev->priv_flags & IFF_BRIDGE_PORT; 4948 } 4949 4950 static inline bool netif_is_ovs_master(const struct net_device *dev) 4951 { 4952 return dev->priv_flags & IFF_OPENVSWITCH; 4953 } 4954 4955 static inline bool netif_is_ovs_port(const struct net_device *dev) 4956 { 4957 return dev->priv_flags & IFF_OVS_DATAPATH; 4958 } 4959 4960 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4961 { 4962 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4963 } 4964 4965 static inline bool netif_is_team_master(const struct net_device *dev) 4966 { 4967 return dev->priv_flags & IFF_TEAM; 4968 } 4969 4970 static inline bool netif_is_team_port(const struct net_device *dev) 4971 { 4972 return dev->priv_flags & IFF_TEAM_PORT; 4973 } 4974 4975 static inline bool netif_is_lag_master(const struct net_device *dev) 4976 { 4977 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4978 } 4979 4980 static inline bool netif_is_lag_port(const struct net_device *dev) 4981 { 4982 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4983 } 4984 4985 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4986 { 4987 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4988 } 4989 4990 static inline bool netif_is_failover(const struct net_device *dev) 4991 { 4992 return dev->priv_flags & IFF_FAILOVER; 4993 } 4994 4995 static inline bool netif_is_failover_slave(const struct net_device *dev) 4996 { 4997 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4998 } 4999 5000 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5001 static inline void netif_keep_dst(struct net_device *dev) 5002 { 5003 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5004 } 5005 5006 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5007 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5008 { 5009 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5010 return dev->priv_flags & IFF_MACSEC; 5011 } 5012 5013 extern struct pernet_operations __net_initdata loopback_net_ops; 5014 5015 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5016 5017 /* netdev_printk helpers, similar to dev_printk */ 5018 5019 static inline const char *netdev_name(const struct net_device *dev) 5020 { 5021 if (!dev->name[0] || strchr(dev->name, '%')) 5022 return "(unnamed net_device)"; 5023 return dev->name; 5024 } 5025 5026 static inline bool netdev_unregistering(const struct net_device *dev) 5027 { 5028 return dev->reg_state == NETREG_UNREGISTERING; 5029 } 5030 5031 static inline const char *netdev_reg_state(const struct net_device *dev) 5032 { 5033 switch (dev->reg_state) { 5034 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5035 case NETREG_REGISTERED: return ""; 5036 case NETREG_UNREGISTERING: return " (unregistering)"; 5037 case NETREG_UNREGISTERED: return " (unregistered)"; 5038 case NETREG_RELEASED: return " (released)"; 5039 case NETREG_DUMMY: return " (dummy)"; 5040 } 5041 5042 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5043 return " (unknown)"; 5044 } 5045 5046 __printf(3, 4) __cold 5047 void netdev_printk(const char *level, const struct net_device *dev, 5048 const char *format, ...); 5049 __printf(2, 3) __cold 5050 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5051 __printf(2, 3) __cold 5052 void netdev_alert(const struct net_device *dev, const char *format, ...); 5053 __printf(2, 3) __cold 5054 void netdev_crit(const struct net_device *dev, const char *format, ...); 5055 __printf(2, 3) __cold 5056 void netdev_err(const struct net_device *dev, const char *format, ...); 5057 __printf(2, 3) __cold 5058 void netdev_warn(const struct net_device *dev, const char *format, ...); 5059 __printf(2, 3) __cold 5060 void netdev_notice(const struct net_device *dev, const char *format, ...); 5061 __printf(2, 3) __cold 5062 void netdev_info(const struct net_device *dev, const char *format, ...); 5063 5064 #define netdev_level_once(level, dev, fmt, ...) \ 5065 do { \ 5066 static bool __print_once __read_mostly; \ 5067 \ 5068 if (!__print_once) { \ 5069 __print_once = true; \ 5070 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5071 } \ 5072 } while (0) 5073 5074 #define netdev_emerg_once(dev, fmt, ...) \ 5075 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5076 #define netdev_alert_once(dev, fmt, ...) \ 5077 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5078 #define netdev_crit_once(dev, fmt, ...) \ 5079 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5080 #define netdev_err_once(dev, fmt, ...) \ 5081 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5082 #define netdev_warn_once(dev, fmt, ...) \ 5083 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5084 #define netdev_notice_once(dev, fmt, ...) \ 5085 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5086 #define netdev_info_once(dev, fmt, ...) \ 5087 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5088 5089 #define MODULE_ALIAS_NETDEV(device) \ 5090 MODULE_ALIAS("netdev-" device) 5091 5092 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5093 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5094 #define netdev_dbg(__dev, format, args...) \ 5095 do { \ 5096 dynamic_netdev_dbg(__dev, format, ##args); \ 5097 } while (0) 5098 #elif defined(DEBUG) 5099 #define netdev_dbg(__dev, format, args...) \ 5100 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5101 #else 5102 #define netdev_dbg(__dev, format, args...) \ 5103 ({ \ 5104 if (0) \ 5105 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5106 }) 5107 #endif 5108 5109 #if defined(VERBOSE_DEBUG) 5110 #define netdev_vdbg netdev_dbg 5111 #else 5112 5113 #define netdev_vdbg(dev, format, args...) \ 5114 ({ \ 5115 if (0) \ 5116 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5117 0; \ 5118 }) 5119 #endif 5120 5121 /* 5122 * netdev_WARN() acts like dev_printk(), but with the key difference 5123 * of using a WARN/WARN_ON to get the message out, including the 5124 * file/line information and a backtrace. 5125 */ 5126 #define netdev_WARN(dev, format, args...) \ 5127 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5128 netdev_reg_state(dev), ##args) 5129 5130 #define netdev_WARN_ONCE(dev, format, args...) \ 5131 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5132 netdev_reg_state(dev), ##args) 5133 5134 /* netif printk helpers, similar to netdev_printk */ 5135 5136 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5137 do { \ 5138 if (netif_msg_##type(priv)) \ 5139 netdev_printk(level, (dev), fmt, ##args); \ 5140 } while (0) 5141 5142 #define netif_level(level, priv, type, dev, fmt, args...) \ 5143 do { \ 5144 if (netif_msg_##type(priv)) \ 5145 netdev_##level(dev, fmt, ##args); \ 5146 } while (0) 5147 5148 #define netif_emerg(priv, type, dev, fmt, args...) \ 5149 netif_level(emerg, priv, type, dev, fmt, ##args) 5150 #define netif_alert(priv, type, dev, fmt, args...) \ 5151 netif_level(alert, priv, type, dev, fmt, ##args) 5152 #define netif_crit(priv, type, dev, fmt, args...) \ 5153 netif_level(crit, priv, type, dev, fmt, ##args) 5154 #define netif_err(priv, type, dev, fmt, args...) \ 5155 netif_level(err, priv, type, dev, fmt, ##args) 5156 #define netif_warn(priv, type, dev, fmt, args...) \ 5157 netif_level(warn, priv, type, dev, fmt, ##args) 5158 #define netif_notice(priv, type, dev, fmt, args...) \ 5159 netif_level(notice, priv, type, dev, fmt, ##args) 5160 #define netif_info(priv, type, dev, fmt, args...) \ 5161 netif_level(info, priv, type, dev, fmt, ##args) 5162 5163 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5164 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5165 #define netif_dbg(priv, type, netdev, format, args...) \ 5166 do { \ 5167 if (netif_msg_##type(priv)) \ 5168 dynamic_netdev_dbg(netdev, format, ##args); \ 5169 } while (0) 5170 #elif defined(DEBUG) 5171 #define netif_dbg(priv, type, dev, format, args...) \ 5172 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5173 #else 5174 #define netif_dbg(priv, type, dev, format, args...) \ 5175 ({ \ 5176 if (0) \ 5177 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5178 0; \ 5179 }) 5180 #endif 5181 5182 /* if @cond then downgrade to debug, else print at @level */ 5183 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5184 do { \ 5185 if (cond) \ 5186 netif_dbg(priv, type, netdev, fmt, ##args); \ 5187 else \ 5188 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5189 } while (0) 5190 5191 #if defined(VERBOSE_DEBUG) 5192 #define netif_vdbg netif_dbg 5193 #else 5194 #define netif_vdbg(priv, type, dev, format, args...) \ 5195 ({ \ 5196 if (0) \ 5197 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5198 0; \ 5199 }) 5200 #endif 5201 5202 /* 5203 * The list of packet types we will receive (as opposed to discard) 5204 * and the routines to invoke. 5205 * 5206 * Why 16. Because with 16 the only overlap we get on a hash of the 5207 * low nibble of the protocol value is RARP/SNAP/X.25. 5208 * 5209 * 0800 IP 5210 * 0001 802.3 5211 * 0002 AX.25 5212 * 0004 802.2 5213 * 8035 RARP 5214 * 0005 SNAP 5215 * 0805 X.25 5216 * 0806 ARP 5217 * 8137 IPX 5218 * 0009 Localtalk 5219 * 86DD IPv6 5220 */ 5221 #define PTYPE_HASH_SIZE (16) 5222 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5223 5224 extern struct net_device *blackhole_netdev; 5225 5226 #endif /* _LINUX_NETDEVICE_H */ 5227