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_CONSUMED, 380 }; 381 typedef enum gro_result gro_result_t; 382 383 /* 384 * enum rx_handler_result - Possible return values for rx_handlers. 385 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 386 * further. 387 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 388 * case skb->dev was changed by rx_handler. 389 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 390 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 391 * 392 * rx_handlers are functions called from inside __netif_receive_skb(), to do 393 * special processing of the skb, prior to delivery to protocol handlers. 394 * 395 * Currently, a net_device can only have a single rx_handler registered. Trying 396 * to register a second rx_handler will return -EBUSY. 397 * 398 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 399 * To unregister a rx_handler on a net_device, use 400 * netdev_rx_handler_unregister(). 401 * 402 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 403 * do with the skb. 404 * 405 * If the rx_handler consumed the skb in some way, it should return 406 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 407 * the skb to be delivered in some other way. 408 * 409 * If the rx_handler changed skb->dev, to divert the skb to another 410 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 411 * new device will be called if it exists. 412 * 413 * If the rx_handler decides the skb should be ignored, it should return 414 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 415 * are registered on exact device (ptype->dev == skb->dev). 416 * 417 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 418 * delivered, it should return RX_HANDLER_PASS. 419 * 420 * A device without a registered rx_handler will behave as if rx_handler 421 * returned RX_HANDLER_PASS. 422 */ 423 424 enum rx_handler_result { 425 RX_HANDLER_CONSUMED, 426 RX_HANDLER_ANOTHER, 427 RX_HANDLER_EXACT, 428 RX_HANDLER_PASS, 429 }; 430 typedef enum rx_handler_result rx_handler_result_t; 431 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 432 433 void __napi_schedule(struct napi_struct *n); 434 void __napi_schedule_irqoff(struct napi_struct *n); 435 436 static inline bool napi_disable_pending(struct napi_struct *n) 437 { 438 return test_bit(NAPI_STATE_DISABLE, &n->state); 439 } 440 441 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 442 { 443 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 444 } 445 446 bool napi_schedule_prep(struct napi_struct *n); 447 448 /** 449 * napi_schedule - schedule NAPI poll 450 * @n: NAPI context 451 * 452 * Schedule NAPI poll routine to be called if it is not already 453 * running. 454 */ 455 static inline void napi_schedule(struct napi_struct *n) 456 { 457 if (napi_schedule_prep(n)) 458 __napi_schedule(n); 459 } 460 461 /** 462 * napi_schedule_irqoff - schedule NAPI poll 463 * @n: NAPI context 464 * 465 * Variant of napi_schedule(), assuming hard irqs are masked. 466 */ 467 static inline void napi_schedule_irqoff(struct napi_struct *n) 468 { 469 if (napi_schedule_prep(n)) 470 __napi_schedule_irqoff(n); 471 } 472 473 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 474 static inline bool napi_reschedule(struct napi_struct *napi) 475 { 476 if (napi_schedule_prep(napi)) { 477 __napi_schedule(napi); 478 return true; 479 } 480 return false; 481 } 482 483 bool napi_complete_done(struct napi_struct *n, int work_done); 484 /** 485 * napi_complete - NAPI processing complete 486 * @n: NAPI context 487 * 488 * Mark NAPI processing as complete. 489 * Consider using napi_complete_done() instead. 490 * Return false if device should avoid rearming interrupts. 491 */ 492 static inline bool napi_complete(struct napi_struct *n) 493 { 494 return napi_complete_done(n, 0); 495 } 496 497 /** 498 * napi_disable - prevent NAPI from scheduling 499 * @n: NAPI context 500 * 501 * Stop NAPI from being scheduled on this context. 502 * Waits till any outstanding processing completes. 503 */ 504 void napi_disable(struct napi_struct *n); 505 506 /** 507 * napi_enable - enable NAPI scheduling 508 * @n: NAPI context 509 * 510 * Resume NAPI from being scheduled on this context. 511 * Must be paired with napi_disable. 512 */ 513 static inline void napi_enable(struct napi_struct *n) 514 { 515 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 516 smp_mb__before_atomic(); 517 clear_bit(NAPI_STATE_SCHED, &n->state); 518 clear_bit(NAPI_STATE_NPSVC, &n->state); 519 } 520 521 /** 522 * napi_synchronize - wait until NAPI is not running 523 * @n: NAPI context 524 * 525 * Wait until NAPI is done being scheduled on this context. 526 * Waits till any outstanding processing completes but 527 * does not disable future activations. 528 */ 529 static inline void napi_synchronize(const struct napi_struct *n) 530 { 531 if (IS_ENABLED(CONFIG_SMP)) 532 while (test_bit(NAPI_STATE_SCHED, &n->state)) 533 msleep(1); 534 else 535 barrier(); 536 } 537 538 /** 539 * napi_if_scheduled_mark_missed - if napi is running, set the 540 * NAPIF_STATE_MISSED 541 * @n: NAPI context 542 * 543 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 544 * NAPI is scheduled. 545 **/ 546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 547 { 548 unsigned long val, new; 549 550 do { 551 val = READ_ONCE(n->state); 552 if (val & NAPIF_STATE_DISABLE) 553 return true; 554 555 if (!(val & NAPIF_STATE_SCHED)) 556 return false; 557 558 new = val | NAPIF_STATE_MISSED; 559 } while (cmpxchg(&n->state, val, new) != val); 560 561 return true; 562 } 563 564 enum netdev_queue_state_t { 565 __QUEUE_STATE_DRV_XOFF, 566 __QUEUE_STATE_STACK_XOFF, 567 __QUEUE_STATE_FROZEN, 568 }; 569 570 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 571 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 572 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 573 574 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 576 QUEUE_STATE_FROZEN) 577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 578 QUEUE_STATE_FROZEN) 579 580 /* 581 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 582 * netif_tx_* functions below are used to manipulate this flag. The 583 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 584 * queue independently. The netif_xmit_*stopped functions below are called 585 * to check if the queue has been stopped by the driver or stack (either 586 * of the XOFF bits are set in the state). Drivers should not need to call 587 * netif_xmit*stopped functions, they should only be using netif_tx_*. 588 */ 589 590 struct netdev_queue { 591 /* 592 * read-mostly part 593 */ 594 struct net_device *dev; 595 struct Qdisc __rcu *qdisc; 596 struct Qdisc *qdisc_sleeping; 597 #ifdef CONFIG_SYSFS 598 struct kobject kobj; 599 #endif 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 int numa_node; 602 #endif 603 unsigned long tx_maxrate; 604 /* 605 * Number of TX timeouts for this queue 606 * (/sys/class/net/DEV/Q/trans_timeout) 607 */ 608 unsigned long trans_timeout; 609 610 /* Subordinate device that the queue has been assigned to */ 611 struct net_device *sb_dev; 612 #ifdef CONFIG_XDP_SOCKETS 613 struct xsk_buff_pool *pool; 614 #endif 615 /* 616 * write-mostly part 617 */ 618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 619 int xmit_lock_owner; 620 /* 621 * Time (in jiffies) of last Tx 622 */ 623 unsigned long trans_start; 624 625 unsigned long state; 626 627 #ifdef CONFIG_BQL 628 struct dql dql; 629 #endif 630 } ____cacheline_aligned_in_smp; 631 632 extern int sysctl_fb_tunnels_only_for_init_net; 633 extern int sysctl_devconf_inherit_init_net; 634 635 /* 636 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 637 * == 1 : For initns only 638 * == 2 : For none. 639 */ 640 static inline bool net_has_fallback_tunnels(const struct net *net) 641 { 642 return !IS_ENABLED(CONFIG_SYSCTL) || 643 !sysctl_fb_tunnels_only_for_init_net || 644 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 645 } 646 647 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 648 { 649 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 650 return q->numa_node; 651 #else 652 return NUMA_NO_NODE; 653 #endif 654 } 655 656 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 657 { 658 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 659 q->numa_node = node; 660 #endif 661 } 662 663 #ifdef CONFIG_RPS 664 /* 665 * This structure holds an RPS map which can be of variable length. The 666 * map is an array of CPUs. 667 */ 668 struct rps_map { 669 unsigned int len; 670 struct rcu_head rcu; 671 u16 cpus[]; 672 }; 673 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 674 675 /* 676 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 677 * tail pointer for that CPU's input queue at the time of last enqueue, and 678 * a hardware filter index. 679 */ 680 struct rps_dev_flow { 681 u16 cpu; 682 u16 filter; 683 unsigned int last_qtail; 684 }; 685 #define RPS_NO_FILTER 0xffff 686 687 /* 688 * The rps_dev_flow_table structure contains a table of flow mappings. 689 */ 690 struct rps_dev_flow_table { 691 unsigned int mask; 692 struct rcu_head rcu; 693 struct rps_dev_flow flows[]; 694 }; 695 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 696 ((_num) * sizeof(struct rps_dev_flow))) 697 698 /* 699 * The rps_sock_flow_table contains mappings of flows to the last CPU 700 * on which they were processed by the application (set in recvmsg). 701 * Each entry is a 32bit value. Upper part is the high-order bits 702 * of flow hash, lower part is CPU number. 703 * rps_cpu_mask is used to partition the space, depending on number of 704 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 705 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 706 * meaning we use 32-6=26 bits for the hash. 707 */ 708 struct rps_sock_flow_table { 709 u32 mask; 710 711 u32 ents[] ____cacheline_aligned_in_smp; 712 }; 713 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 714 715 #define RPS_NO_CPU 0xffff 716 717 extern u32 rps_cpu_mask; 718 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 719 720 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 721 u32 hash) 722 { 723 if (table && hash) { 724 unsigned int index = hash & table->mask; 725 u32 val = hash & ~rps_cpu_mask; 726 727 /* We only give a hint, preemption can change CPU under us */ 728 val |= raw_smp_processor_id(); 729 730 if (table->ents[index] != val) 731 table->ents[index] = val; 732 } 733 } 734 735 #ifdef CONFIG_RFS_ACCEL 736 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 737 u16 filter_id); 738 #endif 739 #endif /* CONFIG_RPS */ 740 741 /* This structure contains an instance of an RX queue. */ 742 struct netdev_rx_queue { 743 #ifdef CONFIG_RPS 744 struct rps_map __rcu *rps_map; 745 struct rps_dev_flow_table __rcu *rps_flow_table; 746 #endif 747 struct kobject kobj; 748 struct net_device *dev; 749 struct xdp_rxq_info xdp_rxq; 750 #ifdef CONFIG_XDP_SOCKETS 751 struct xsk_buff_pool *pool; 752 #endif 753 } ____cacheline_aligned_in_smp; 754 755 /* 756 * RX queue sysfs structures and functions. 757 */ 758 struct rx_queue_attribute { 759 struct attribute attr; 760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 761 ssize_t (*store)(struct netdev_rx_queue *queue, 762 const char *buf, size_t len); 763 }; 764 765 #ifdef CONFIG_XPS 766 /* 767 * This structure holds an XPS map which can be of variable length. The 768 * map is an array of queues. 769 */ 770 struct xps_map { 771 unsigned int len; 772 unsigned int alloc_len; 773 struct rcu_head rcu; 774 u16 queues[]; 775 }; 776 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 777 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 778 - sizeof(struct xps_map)) / sizeof(u16)) 779 780 /* 781 * This structure holds all XPS maps for device. Maps are indexed by CPU. 782 */ 783 struct xps_dev_maps { 784 struct rcu_head rcu; 785 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 786 }; 787 788 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 789 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 790 791 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 792 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 793 794 #endif /* CONFIG_XPS */ 795 796 #define TC_MAX_QUEUE 16 797 #define TC_BITMASK 15 798 /* HW offloaded queuing disciplines txq count and offset maps */ 799 struct netdev_tc_txq { 800 u16 count; 801 u16 offset; 802 }; 803 804 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 805 /* 806 * This structure is to hold information about the device 807 * configured to run FCoE protocol stack. 808 */ 809 struct netdev_fcoe_hbainfo { 810 char manufacturer[64]; 811 char serial_number[64]; 812 char hardware_version[64]; 813 char driver_version[64]; 814 char optionrom_version[64]; 815 char firmware_version[64]; 816 char model[256]; 817 char model_description[256]; 818 }; 819 #endif 820 821 #define MAX_PHYS_ITEM_ID_LEN 32 822 823 /* This structure holds a unique identifier to identify some 824 * physical item (port for example) used by a netdevice. 825 */ 826 struct netdev_phys_item_id { 827 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 828 unsigned char id_len; 829 }; 830 831 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 832 struct netdev_phys_item_id *b) 833 { 834 return a->id_len == b->id_len && 835 memcmp(a->id, b->id, a->id_len) == 0; 836 } 837 838 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 839 struct sk_buff *skb, 840 struct net_device *sb_dev); 841 842 enum tc_setup_type { 843 TC_SETUP_QDISC_MQPRIO, 844 TC_SETUP_CLSU32, 845 TC_SETUP_CLSFLOWER, 846 TC_SETUP_CLSMATCHALL, 847 TC_SETUP_CLSBPF, 848 TC_SETUP_BLOCK, 849 TC_SETUP_QDISC_CBS, 850 TC_SETUP_QDISC_RED, 851 TC_SETUP_QDISC_PRIO, 852 TC_SETUP_QDISC_MQ, 853 TC_SETUP_QDISC_ETF, 854 TC_SETUP_ROOT_QDISC, 855 TC_SETUP_QDISC_GRED, 856 TC_SETUP_QDISC_TAPRIO, 857 TC_SETUP_FT, 858 TC_SETUP_QDISC_ETS, 859 TC_SETUP_QDISC_TBF, 860 TC_SETUP_QDISC_FIFO, 861 }; 862 863 /* These structures hold the attributes of bpf state that are being passed 864 * to the netdevice through the bpf op. 865 */ 866 enum bpf_netdev_command { 867 /* Set or clear a bpf program used in the earliest stages of packet 868 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 869 * is responsible for calling bpf_prog_put on any old progs that are 870 * stored. In case of error, the callee need not release the new prog 871 * reference, but on success it takes ownership and must bpf_prog_put 872 * when it is no longer used. 873 */ 874 XDP_SETUP_PROG, 875 XDP_SETUP_PROG_HW, 876 /* BPF program for offload callbacks, invoked at program load time. */ 877 BPF_OFFLOAD_MAP_ALLOC, 878 BPF_OFFLOAD_MAP_FREE, 879 XDP_SETUP_XSK_POOL, 880 }; 881 882 struct bpf_prog_offload_ops; 883 struct netlink_ext_ack; 884 struct xdp_umem; 885 struct xdp_dev_bulk_queue; 886 struct bpf_xdp_link; 887 888 enum bpf_xdp_mode { 889 XDP_MODE_SKB = 0, 890 XDP_MODE_DRV = 1, 891 XDP_MODE_HW = 2, 892 __MAX_XDP_MODE 893 }; 894 895 struct bpf_xdp_entity { 896 struct bpf_prog *prog; 897 struct bpf_xdp_link *link; 898 }; 899 900 struct netdev_bpf { 901 enum bpf_netdev_command command; 902 union { 903 /* XDP_SETUP_PROG */ 904 struct { 905 u32 flags; 906 struct bpf_prog *prog; 907 struct netlink_ext_ack *extack; 908 }; 909 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 910 struct { 911 struct bpf_offloaded_map *offmap; 912 }; 913 /* XDP_SETUP_XSK_POOL */ 914 struct { 915 struct xsk_buff_pool *pool; 916 u16 queue_id; 917 } xsk; 918 }; 919 }; 920 921 /* Flags for ndo_xsk_wakeup. */ 922 #define XDP_WAKEUP_RX (1 << 0) 923 #define XDP_WAKEUP_TX (1 << 1) 924 925 #ifdef CONFIG_XFRM_OFFLOAD 926 struct xfrmdev_ops { 927 int (*xdo_dev_state_add) (struct xfrm_state *x); 928 void (*xdo_dev_state_delete) (struct xfrm_state *x); 929 void (*xdo_dev_state_free) (struct xfrm_state *x); 930 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 931 struct xfrm_state *x); 932 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 933 }; 934 #endif 935 936 struct dev_ifalias { 937 struct rcu_head rcuhead; 938 char ifalias[]; 939 }; 940 941 struct devlink; 942 struct tlsdev_ops; 943 944 struct netdev_name_node { 945 struct hlist_node hlist; 946 struct list_head list; 947 struct net_device *dev; 948 const char *name; 949 }; 950 951 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 952 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 953 954 struct netdev_net_notifier { 955 struct list_head list; 956 struct notifier_block *nb; 957 }; 958 959 /* 960 * This structure defines the management hooks for network devices. 961 * The following hooks can be defined; unless noted otherwise, they are 962 * optional and can be filled with a null pointer. 963 * 964 * int (*ndo_init)(struct net_device *dev); 965 * This function is called once when a network device is registered. 966 * The network device can use this for any late stage initialization 967 * or semantic validation. It can fail with an error code which will 968 * be propagated back to register_netdev. 969 * 970 * void (*ndo_uninit)(struct net_device *dev); 971 * This function is called when device is unregistered or when registration 972 * fails. It is not called if init fails. 973 * 974 * int (*ndo_open)(struct net_device *dev); 975 * This function is called when a network device transitions to the up 976 * state. 977 * 978 * int (*ndo_stop)(struct net_device *dev); 979 * This function is called when a network device transitions to the down 980 * state. 981 * 982 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 983 * struct net_device *dev); 984 * Called when a packet needs to be transmitted. 985 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 986 * the queue before that can happen; it's for obsolete devices and weird 987 * corner cases, but the stack really does a non-trivial amount 988 * of useless work if you return NETDEV_TX_BUSY. 989 * Required; cannot be NULL. 990 * 991 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 992 * struct net_device *dev 993 * netdev_features_t features); 994 * Called by core transmit path to determine if device is capable of 995 * performing offload operations on a given packet. This is to give 996 * the device an opportunity to implement any restrictions that cannot 997 * be otherwise expressed by feature flags. The check is called with 998 * the set of features that the stack has calculated and it returns 999 * those the driver believes to be appropriate. 1000 * 1001 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1002 * struct net_device *sb_dev); 1003 * Called to decide which queue to use when device supports multiple 1004 * transmit queues. 1005 * 1006 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1007 * This function is called to allow device receiver to make 1008 * changes to configuration when multicast or promiscuous is enabled. 1009 * 1010 * void (*ndo_set_rx_mode)(struct net_device *dev); 1011 * This function is called device changes address list filtering. 1012 * If driver handles unicast address filtering, it should set 1013 * IFF_UNICAST_FLT in its priv_flags. 1014 * 1015 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1016 * This function is called when the Media Access Control address 1017 * needs to be changed. If this interface is not defined, the 1018 * MAC address can not be changed. 1019 * 1020 * int (*ndo_validate_addr)(struct net_device *dev); 1021 * Test if Media Access Control address is valid for the device. 1022 * 1023 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1024 * Called when a user requests an ioctl which can't be handled by 1025 * the generic interface code. If not defined ioctls return 1026 * not supported error code. 1027 * 1028 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1029 * Used to set network devices bus interface parameters. This interface 1030 * is retained for legacy reasons; new devices should use the bus 1031 * interface (PCI) for low level management. 1032 * 1033 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1034 * Called when a user wants to change the Maximum Transfer Unit 1035 * of a device. 1036 * 1037 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1038 * Callback used when the transmitter has not made any progress 1039 * for dev->watchdog ticks. 1040 * 1041 * void (*ndo_get_stats64)(struct net_device *dev, 1042 * struct rtnl_link_stats64 *storage); 1043 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1044 * Called when a user wants to get the network device usage 1045 * statistics. Drivers must do one of the following: 1046 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1047 * rtnl_link_stats64 structure passed by the caller. 1048 * 2. Define @ndo_get_stats to update a net_device_stats structure 1049 * (which should normally be dev->stats) and return a pointer to 1050 * it. The structure may be changed asynchronously only if each 1051 * field is written atomically. 1052 * 3. Update dev->stats asynchronously and atomically, and define 1053 * neither operation. 1054 * 1055 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1056 * Return true if this device supports offload stats of this attr_id. 1057 * 1058 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1059 * void *attr_data) 1060 * Get statistics for offload operations by attr_id. Write it into the 1061 * attr_data pointer. 1062 * 1063 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1064 * If device supports VLAN filtering this function is called when a 1065 * VLAN id is registered. 1066 * 1067 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1068 * If device supports VLAN filtering this function is called when a 1069 * VLAN id is unregistered. 1070 * 1071 * void (*ndo_poll_controller)(struct net_device *dev); 1072 * 1073 * SR-IOV management functions. 1074 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1075 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1076 * u8 qos, __be16 proto); 1077 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1078 * int max_tx_rate); 1079 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1080 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1081 * int (*ndo_get_vf_config)(struct net_device *dev, 1082 * int vf, struct ifla_vf_info *ivf); 1083 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1084 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1085 * struct nlattr *port[]); 1086 * 1087 * Enable or disable the VF ability to query its RSS Redirection Table and 1088 * Hash Key. This is needed since on some devices VF share this information 1089 * with PF and querying it may introduce a theoretical security risk. 1090 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1091 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1092 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1093 * void *type_data); 1094 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1095 * This is always called from the stack with the rtnl lock held and netif 1096 * tx queues stopped. This allows the netdevice to perform queue 1097 * management safely. 1098 * 1099 * Fiber Channel over Ethernet (FCoE) offload functions. 1100 * int (*ndo_fcoe_enable)(struct net_device *dev); 1101 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1102 * so the underlying device can perform whatever needed configuration or 1103 * initialization to support acceleration of FCoE traffic. 1104 * 1105 * int (*ndo_fcoe_disable)(struct net_device *dev); 1106 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1107 * so the underlying device can perform whatever needed clean-ups to 1108 * stop supporting acceleration of FCoE traffic. 1109 * 1110 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1111 * struct scatterlist *sgl, unsigned int sgc); 1112 * Called when the FCoE Initiator wants to initialize an I/O that 1113 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1114 * perform necessary setup and returns 1 to indicate the device is set up 1115 * successfully to perform DDP on this I/O, otherwise this returns 0. 1116 * 1117 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1118 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1119 * indicated by the FC exchange id 'xid', so the underlying device can 1120 * clean up and reuse resources for later DDP requests. 1121 * 1122 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1123 * struct scatterlist *sgl, unsigned int sgc); 1124 * Called when the FCoE Target wants to initialize an I/O that 1125 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1126 * perform necessary setup and returns 1 to indicate the device is set up 1127 * successfully to perform DDP on this I/O, otherwise this returns 0. 1128 * 1129 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1130 * struct netdev_fcoe_hbainfo *hbainfo); 1131 * Called when the FCoE Protocol stack wants information on the underlying 1132 * device. This information is utilized by the FCoE protocol stack to 1133 * register attributes with Fiber Channel management service as per the 1134 * FC-GS Fabric Device Management Information(FDMI) specification. 1135 * 1136 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1137 * Called when the underlying device wants to override default World Wide 1138 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1139 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1140 * protocol stack to use. 1141 * 1142 * RFS acceleration. 1143 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1144 * u16 rxq_index, u32 flow_id); 1145 * Set hardware filter for RFS. rxq_index is the target queue index; 1146 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1147 * Return the filter ID on success, or a negative error code. 1148 * 1149 * Slave management functions (for bridge, bonding, etc). 1150 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1151 * Called to make another netdev an underling. 1152 * 1153 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1154 * Called to release previously enslaved netdev. 1155 * 1156 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1157 * struct sk_buff *skb, 1158 * bool all_slaves); 1159 * Get the xmit slave of master device. If all_slaves is true, function 1160 * assume all the slaves can transmit. 1161 * 1162 * Feature/offload setting functions. 1163 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1164 * netdev_features_t features); 1165 * Adjusts the requested feature flags according to device-specific 1166 * constraints, and returns the resulting flags. Must not modify 1167 * the device state. 1168 * 1169 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1170 * Called to update device configuration to new features. Passed 1171 * feature set might be less than what was returned by ndo_fix_features()). 1172 * Must return >0 or -errno if it changed dev->features itself. 1173 * 1174 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1175 * struct net_device *dev, 1176 * const unsigned char *addr, u16 vid, u16 flags, 1177 * struct netlink_ext_ack *extack); 1178 * Adds an FDB entry to dev for addr. 1179 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1180 * struct net_device *dev, 1181 * const unsigned char *addr, u16 vid) 1182 * Deletes the FDB entry from dev coresponding to addr. 1183 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1184 * struct net_device *dev, struct net_device *filter_dev, 1185 * int *idx) 1186 * Used to add FDB entries to dump requests. Implementers should add 1187 * entries to skb and update idx with the number of entries. 1188 * 1189 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1190 * u16 flags, struct netlink_ext_ack *extack) 1191 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1192 * struct net_device *dev, u32 filter_mask, 1193 * int nlflags) 1194 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1195 * u16 flags); 1196 * 1197 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1198 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1199 * which do not represent real hardware may define this to allow their 1200 * userspace components to manage their virtual carrier state. Devices 1201 * that determine carrier state from physical hardware properties (eg 1202 * network cables) or protocol-dependent mechanisms (eg 1203 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1204 * 1205 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1206 * struct netdev_phys_item_id *ppid); 1207 * Called to get ID of physical port of this device. If driver does 1208 * not implement this, it is assumed that the hw is not able to have 1209 * multiple net devices on single physical port. 1210 * 1211 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1212 * struct netdev_phys_item_id *ppid) 1213 * Called to get the parent ID of the physical port of this device. 1214 * 1215 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1216 * struct net_device *dev) 1217 * Called by upper layer devices to accelerate switching or other 1218 * station functionality into hardware. 'pdev is the lowerdev 1219 * to use for the offload and 'dev' is the net device that will 1220 * back the offload. Returns a pointer to the private structure 1221 * the upper layer will maintain. 1222 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1223 * Called by upper layer device to delete the station created 1224 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1225 * the station and priv is the structure returned by the add 1226 * operation. 1227 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1228 * int queue_index, u32 maxrate); 1229 * Called when a user wants to set a max-rate limitation of specific 1230 * TX queue. 1231 * int (*ndo_get_iflink)(const struct net_device *dev); 1232 * Called to get the iflink value of this device. 1233 * void (*ndo_change_proto_down)(struct net_device *dev, 1234 * bool proto_down); 1235 * This function is used to pass protocol port error state information 1236 * to the switch driver. The switch driver can react to the proto_down 1237 * by doing a phys down on the associated switch port. 1238 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1239 * This function is used to get egress tunnel information for given skb. 1240 * This is useful for retrieving outer tunnel header parameters while 1241 * sampling packet. 1242 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1243 * This function is used to specify the headroom that the skb must 1244 * consider when allocation skb during packet reception. Setting 1245 * appropriate rx headroom value allows avoiding skb head copy on 1246 * forward. Setting a negative value resets the rx headroom to the 1247 * default value. 1248 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1249 * This function is used to set or query state related to XDP on the 1250 * netdevice and manage BPF offload. See definition of 1251 * enum bpf_netdev_command for details. 1252 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1253 * u32 flags); 1254 * This function is used to submit @n XDP packets for transmit on a 1255 * netdevice. Returns number of frames successfully transmitted, frames 1256 * that got dropped are freed/returned via xdp_return_frame(). 1257 * Returns negative number, means general error invoking ndo, meaning 1258 * no frames were xmit'ed and core-caller will free all frames. 1259 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1260 * This function is used to wake up the softirq, ksoftirqd or kthread 1261 * responsible for sending and/or receiving packets on a specific 1262 * queue id bound to an AF_XDP socket. The flags field specifies if 1263 * only RX, only Tx, or both should be woken up using the flags 1264 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1265 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1266 * Get devlink port instance associated with a given netdev. 1267 * Called with a reference on the netdevice and devlink locks only, 1268 * rtnl_lock is not held. 1269 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1270 * int cmd); 1271 * Add, change, delete or get information on an IPv4 tunnel. 1272 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1273 * If a device is paired with a peer device, return the peer instance. 1274 * The caller must be under RCU read context. 1275 */ 1276 struct net_device_ops { 1277 int (*ndo_init)(struct net_device *dev); 1278 void (*ndo_uninit)(struct net_device *dev); 1279 int (*ndo_open)(struct net_device *dev); 1280 int (*ndo_stop)(struct net_device *dev); 1281 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1282 struct net_device *dev); 1283 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1284 struct net_device *dev, 1285 netdev_features_t features); 1286 u16 (*ndo_select_queue)(struct net_device *dev, 1287 struct sk_buff *skb, 1288 struct net_device *sb_dev); 1289 void (*ndo_change_rx_flags)(struct net_device *dev, 1290 int flags); 1291 void (*ndo_set_rx_mode)(struct net_device *dev); 1292 int (*ndo_set_mac_address)(struct net_device *dev, 1293 void *addr); 1294 int (*ndo_validate_addr)(struct net_device *dev); 1295 int (*ndo_do_ioctl)(struct net_device *dev, 1296 struct ifreq *ifr, int cmd); 1297 int (*ndo_set_config)(struct net_device *dev, 1298 struct ifmap *map); 1299 int (*ndo_change_mtu)(struct net_device *dev, 1300 int new_mtu); 1301 int (*ndo_neigh_setup)(struct net_device *dev, 1302 struct neigh_parms *); 1303 void (*ndo_tx_timeout) (struct net_device *dev, 1304 unsigned int txqueue); 1305 1306 void (*ndo_get_stats64)(struct net_device *dev, 1307 struct rtnl_link_stats64 *storage); 1308 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1309 int (*ndo_get_offload_stats)(int attr_id, 1310 const struct net_device *dev, 1311 void *attr_data); 1312 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1313 1314 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1315 __be16 proto, u16 vid); 1316 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1317 __be16 proto, u16 vid); 1318 #ifdef CONFIG_NET_POLL_CONTROLLER 1319 void (*ndo_poll_controller)(struct net_device *dev); 1320 int (*ndo_netpoll_setup)(struct net_device *dev, 1321 struct netpoll_info *info); 1322 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1323 #endif 1324 int (*ndo_set_vf_mac)(struct net_device *dev, 1325 int queue, u8 *mac); 1326 int (*ndo_set_vf_vlan)(struct net_device *dev, 1327 int queue, u16 vlan, 1328 u8 qos, __be16 proto); 1329 int (*ndo_set_vf_rate)(struct net_device *dev, 1330 int vf, int min_tx_rate, 1331 int max_tx_rate); 1332 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1333 int vf, bool setting); 1334 int (*ndo_set_vf_trust)(struct net_device *dev, 1335 int vf, bool setting); 1336 int (*ndo_get_vf_config)(struct net_device *dev, 1337 int vf, 1338 struct ifla_vf_info *ivf); 1339 int (*ndo_set_vf_link_state)(struct net_device *dev, 1340 int vf, int link_state); 1341 int (*ndo_get_vf_stats)(struct net_device *dev, 1342 int vf, 1343 struct ifla_vf_stats 1344 *vf_stats); 1345 int (*ndo_set_vf_port)(struct net_device *dev, 1346 int vf, 1347 struct nlattr *port[]); 1348 int (*ndo_get_vf_port)(struct net_device *dev, 1349 int vf, struct sk_buff *skb); 1350 int (*ndo_get_vf_guid)(struct net_device *dev, 1351 int vf, 1352 struct ifla_vf_guid *node_guid, 1353 struct ifla_vf_guid *port_guid); 1354 int (*ndo_set_vf_guid)(struct net_device *dev, 1355 int vf, u64 guid, 1356 int guid_type); 1357 int (*ndo_set_vf_rss_query_en)( 1358 struct net_device *dev, 1359 int vf, bool setting); 1360 int (*ndo_setup_tc)(struct net_device *dev, 1361 enum tc_setup_type type, 1362 void *type_data); 1363 #if IS_ENABLED(CONFIG_FCOE) 1364 int (*ndo_fcoe_enable)(struct net_device *dev); 1365 int (*ndo_fcoe_disable)(struct net_device *dev); 1366 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1367 u16 xid, 1368 struct scatterlist *sgl, 1369 unsigned int sgc); 1370 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1371 u16 xid); 1372 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1373 u16 xid, 1374 struct scatterlist *sgl, 1375 unsigned int sgc); 1376 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1377 struct netdev_fcoe_hbainfo *hbainfo); 1378 #endif 1379 1380 #if IS_ENABLED(CONFIG_LIBFCOE) 1381 #define NETDEV_FCOE_WWNN 0 1382 #define NETDEV_FCOE_WWPN 1 1383 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1384 u64 *wwn, int type); 1385 #endif 1386 1387 #ifdef CONFIG_RFS_ACCEL 1388 int (*ndo_rx_flow_steer)(struct net_device *dev, 1389 const struct sk_buff *skb, 1390 u16 rxq_index, 1391 u32 flow_id); 1392 #endif 1393 int (*ndo_add_slave)(struct net_device *dev, 1394 struct net_device *slave_dev, 1395 struct netlink_ext_ack *extack); 1396 int (*ndo_del_slave)(struct net_device *dev, 1397 struct net_device *slave_dev); 1398 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1399 struct sk_buff *skb, 1400 bool all_slaves); 1401 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1402 netdev_features_t features); 1403 int (*ndo_set_features)(struct net_device *dev, 1404 netdev_features_t features); 1405 int (*ndo_neigh_construct)(struct net_device *dev, 1406 struct neighbour *n); 1407 void (*ndo_neigh_destroy)(struct net_device *dev, 1408 struct neighbour *n); 1409 1410 int (*ndo_fdb_add)(struct ndmsg *ndm, 1411 struct nlattr *tb[], 1412 struct net_device *dev, 1413 const unsigned char *addr, 1414 u16 vid, 1415 u16 flags, 1416 struct netlink_ext_ack *extack); 1417 int (*ndo_fdb_del)(struct ndmsg *ndm, 1418 struct nlattr *tb[], 1419 struct net_device *dev, 1420 const unsigned char *addr, 1421 u16 vid); 1422 int (*ndo_fdb_dump)(struct sk_buff *skb, 1423 struct netlink_callback *cb, 1424 struct net_device *dev, 1425 struct net_device *filter_dev, 1426 int *idx); 1427 int (*ndo_fdb_get)(struct sk_buff *skb, 1428 struct nlattr *tb[], 1429 struct net_device *dev, 1430 const unsigned char *addr, 1431 u16 vid, u32 portid, u32 seq, 1432 struct netlink_ext_ack *extack); 1433 int (*ndo_bridge_setlink)(struct net_device *dev, 1434 struct nlmsghdr *nlh, 1435 u16 flags, 1436 struct netlink_ext_ack *extack); 1437 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1438 u32 pid, u32 seq, 1439 struct net_device *dev, 1440 u32 filter_mask, 1441 int nlflags); 1442 int (*ndo_bridge_dellink)(struct net_device *dev, 1443 struct nlmsghdr *nlh, 1444 u16 flags); 1445 int (*ndo_change_carrier)(struct net_device *dev, 1446 bool new_carrier); 1447 int (*ndo_get_phys_port_id)(struct net_device *dev, 1448 struct netdev_phys_item_id *ppid); 1449 int (*ndo_get_port_parent_id)(struct net_device *dev, 1450 struct netdev_phys_item_id *ppid); 1451 int (*ndo_get_phys_port_name)(struct net_device *dev, 1452 char *name, size_t len); 1453 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1454 struct net_device *dev); 1455 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1456 void *priv); 1457 1458 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1459 int queue_index, 1460 u32 maxrate); 1461 int (*ndo_get_iflink)(const struct net_device *dev); 1462 int (*ndo_change_proto_down)(struct net_device *dev, 1463 bool proto_down); 1464 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1465 struct sk_buff *skb); 1466 void (*ndo_set_rx_headroom)(struct net_device *dev, 1467 int needed_headroom); 1468 int (*ndo_bpf)(struct net_device *dev, 1469 struct netdev_bpf *bpf); 1470 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1471 struct xdp_frame **xdp, 1472 u32 flags); 1473 int (*ndo_xsk_wakeup)(struct net_device *dev, 1474 u32 queue_id, u32 flags); 1475 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1476 int (*ndo_tunnel_ctl)(struct net_device *dev, 1477 struct ip_tunnel_parm *p, int cmd); 1478 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1479 }; 1480 1481 /** 1482 * enum netdev_priv_flags - &struct net_device priv_flags 1483 * 1484 * These are the &struct net_device, they are only set internally 1485 * by drivers and used in the kernel. These flags are invisible to 1486 * userspace; this means that the order of these flags can change 1487 * during any kernel release. 1488 * 1489 * You should have a pretty good reason to be extending these flags. 1490 * 1491 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1492 * @IFF_EBRIDGE: Ethernet bridging device 1493 * @IFF_BONDING: bonding master or slave 1494 * @IFF_ISATAP: ISATAP interface (RFC4214) 1495 * @IFF_WAN_HDLC: WAN HDLC device 1496 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1497 * release skb->dst 1498 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1499 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1500 * @IFF_MACVLAN_PORT: device used as macvlan port 1501 * @IFF_BRIDGE_PORT: device used as bridge port 1502 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1503 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1504 * @IFF_UNICAST_FLT: Supports unicast filtering 1505 * @IFF_TEAM_PORT: device used as team port 1506 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1507 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1508 * change when it's running 1509 * @IFF_MACVLAN: Macvlan device 1510 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1511 * underlying stacked devices 1512 * @IFF_L3MDEV_MASTER: device is an L3 master device 1513 * @IFF_NO_QUEUE: device can run without qdisc attached 1514 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1515 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1516 * @IFF_TEAM: device is a team device 1517 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1518 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1519 * entity (i.e. the master device for bridged veth) 1520 * @IFF_MACSEC: device is a MACsec device 1521 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1522 * @IFF_FAILOVER: device is a failover master device 1523 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1524 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1525 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1526 */ 1527 enum netdev_priv_flags { 1528 IFF_802_1Q_VLAN = 1<<0, 1529 IFF_EBRIDGE = 1<<1, 1530 IFF_BONDING = 1<<2, 1531 IFF_ISATAP = 1<<3, 1532 IFF_WAN_HDLC = 1<<4, 1533 IFF_XMIT_DST_RELEASE = 1<<5, 1534 IFF_DONT_BRIDGE = 1<<6, 1535 IFF_DISABLE_NETPOLL = 1<<7, 1536 IFF_MACVLAN_PORT = 1<<8, 1537 IFF_BRIDGE_PORT = 1<<9, 1538 IFF_OVS_DATAPATH = 1<<10, 1539 IFF_TX_SKB_SHARING = 1<<11, 1540 IFF_UNICAST_FLT = 1<<12, 1541 IFF_TEAM_PORT = 1<<13, 1542 IFF_SUPP_NOFCS = 1<<14, 1543 IFF_LIVE_ADDR_CHANGE = 1<<15, 1544 IFF_MACVLAN = 1<<16, 1545 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1546 IFF_L3MDEV_MASTER = 1<<18, 1547 IFF_NO_QUEUE = 1<<19, 1548 IFF_OPENVSWITCH = 1<<20, 1549 IFF_L3MDEV_SLAVE = 1<<21, 1550 IFF_TEAM = 1<<22, 1551 IFF_RXFH_CONFIGURED = 1<<23, 1552 IFF_PHONY_HEADROOM = 1<<24, 1553 IFF_MACSEC = 1<<25, 1554 IFF_NO_RX_HANDLER = 1<<26, 1555 IFF_FAILOVER = 1<<27, 1556 IFF_FAILOVER_SLAVE = 1<<28, 1557 IFF_L3MDEV_RX_HANDLER = 1<<29, 1558 IFF_LIVE_RENAME_OK = 1<<30, 1559 }; 1560 1561 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1562 #define IFF_EBRIDGE IFF_EBRIDGE 1563 #define IFF_BONDING IFF_BONDING 1564 #define IFF_ISATAP IFF_ISATAP 1565 #define IFF_WAN_HDLC IFF_WAN_HDLC 1566 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1567 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1568 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1569 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1570 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1571 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1572 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1573 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1574 #define IFF_TEAM_PORT IFF_TEAM_PORT 1575 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1576 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1577 #define IFF_MACVLAN IFF_MACVLAN 1578 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1579 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1580 #define IFF_NO_QUEUE IFF_NO_QUEUE 1581 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1582 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1583 #define IFF_TEAM IFF_TEAM 1584 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1585 #define IFF_MACSEC IFF_MACSEC 1586 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1587 #define IFF_FAILOVER IFF_FAILOVER 1588 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1589 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1590 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1591 1592 /** 1593 * struct net_device - The DEVICE structure. 1594 * 1595 * Actually, this whole structure is a big mistake. It mixes I/O 1596 * data with strictly "high-level" data, and it has to know about 1597 * almost every data structure used in the INET module. 1598 * 1599 * @name: This is the first field of the "visible" part of this structure 1600 * (i.e. as seen by users in the "Space.c" file). It is the name 1601 * of the interface. 1602 * 1603 * @name_node: Name hashlist node 1604 * @ifalias: SNMP alias 1605 * @mem_end: Shared memory end 1606 * @mem_start: Shared memory start 1607 * @base_addr: Device I/O address 1608 * @irq: Device IRQ number 1609 * 1610 * @state: Generic network queuing layer state, see netdev_state_t 1611 * @dev_list: The global list of network devices 1612 * @napi_list: List entry used for polling NAPI devices 1613 * @unreg_list: List entry when we are unregistering the 1614 * device; see the function unregister_netdev 1615 * @close_list: List entry used when we are closing the device 1616 * @ptype_all: Device-specific packet handlers for all protocols 1617 * @ptype_specific: Device-specific, protocol-specific packet handlers 1618 * 1619 * @adj_list: Directly linked devices, like slaves for bonding 1620 * @features: Currently active device features 1621 * @hw_features: User-changeable features 1622 * 1623 * @wanted_features: User-requested features 1624 * @vlan_features: Mask of features inheritable by VLAN devices 1625 * 1626 * @hw_enc_features: Mask of features inherited by encapsulating devices 1627 * This field indicates what encapsulation 1628 * offloads the hardware is capable of doing, 1629 * and drivers will need to set them appropriately. 1630 * 1631 * @mpls_features: Mask of features inheritable by MPLS 1632 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1633 * 1634 * @ifindex: interface index 1635 * @group: The group the device belongs to 1636 * 1637 * @stats: Statistics struct, which was left as a legacy, use 1638 * rtnl_link_stats64 instead 1639 * 1640 * @rx_dropped: Dropped packets by core network, 1641 * do not use this in drivers 1642 * @tx_dropped: Dropped packets by core network, 1643 * do not use this in drivers 1644 * @rx_nohandler: nohandler dropped packets by core network on 1645 * inactive devices, do not use this in drivers 1646 * @carrier_up_count: Number of times the carrier has been up 1647 * @carrier_down_count: Number of times the carrier has been down 1648 * 1649 * @wireless_handlers: List of functions to handle Wireless Extensions, 1650 * instead of ioctl, 1651 * see <net/iw_handler.h> for details. 1652 * @wireless_data: Instance data managed by the core of wireless extensions 1653 * 1654 * @netdev_ops: Includes several pointers to callbacks, 1655 * if one wants to override the ndo_*() functions 1656 * @ethtool_ops: Management operations 1657 * @l3mdev_ops: Layer 3 master device operations 1658 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1659 * discovery handling. Necessary for e.g. 6LoWPAN. 1660 * @xfrmdev_ops: Transformation offload operations 1661 * @tlsdev_ops: Transport Layer Security offload operations 1662 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1663 * of Layer 2 headers. 1664 * 1665 * @flags: Interface flags (a la BSD) 1666 * @priv_flags: Like 'flags' but invisible to userspace, 1667 * see if.h for the definitions 1668 * @gflags: Global flags ( kept as legacy ) 1669 * @padded: How much padding added by alloc_netdev() 1670 * @operstate: RFC2863 operstate 1671 * @link_mode: Mapping policy to operstate 1672 * @if_port: Selectable AUI, TP, ... 1673 * @dma: DMA channel 1674 * @mtu: Interface MTU value 1675 * @min_mtu: Interface Minimum MTU value 1676 * @max_mtu: Interface Maximum MTU value 1677 * @type: Interface hardware type 1678 * @hard_header_len: Maximum hardware header length. 1679 * @min_header_len: Minimum hardware header length 1680 * 1681 * @needed_headroom: Extra headroom the hardware may need, but not in all 1682 * cases can this be guaranteed 1683 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1684 * cases can this be guaranteed. Some cases also use 1685 * LL_MAX_HEADER instead to allocate the skb 1686 * 1687 * interface address info: 1688 * 1689 * @perm_addr: Permanent hw address 1690 * @addr_assign_type: Hw address assignment type 1691 * @addr_len: Hardware address length 1692 * @upper_level: Maximum depth level of upper devices. 1693 * @lower_level: Maximum depth level of lower devices. 1694 * @neigh_priv_len: Used in neigh_alloc() 1695 * @dev_id: Used to differentiate devices that share 1696 * the same link layer address 1697 * @dev_port: Used to differentiate devices that share 1698 * the same function 1699 * @addr_list_lock: XXX: need comments on this one 1700 * @name_assign_type: network interface name assignment type 1701 * @uc_promisc: Counter that indicates promiscuous mode 1702 * has been enabled due to the need to listen to 1703 * additional unicast addresses in a device that 1704 * does not implement ndo_set_rx_mode() 1705 * @uc: unicast mac addresses 1706 * @mc: multicast mac addresses 1707 * @dev_addrs: list of device hw addresses 1708 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1709 * @promiscuity: Number of times the NIC is told to work in 1710 * promiscuous mode; if it becomes 0 the NIC will 1711 * exit promiscuous mode 1712 * @allmulti: Counter, enables or disables allmulticast mode 1713 * 1714 * @vlan_info: VLAN info 1715 * @dsa_ptr: dsa specific data 1716 * @tipc_ptr: TIPC specific data 1717 * @atalk_ptr: AppleTalk link 1718 * @ip_ptr: IPv4 specific data 1719 * @dn_ptr: DECnet specific data 1720 * @ip6_ptr: IPv6 specific data 1721 * @ax25_ptr: AX.25 specific data 1722 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1723 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1724 * device struct 1725 * @mpls_ptr: mpls_dev struct pointer 1726 * 1727 * @dev_addr: Hw address (before bcast, 1728 * because most packets are unicast) 1729 * 1730 * @_rx: Array of RX queues 1731 * @num_rx_queues: Number of RX queues 1732 * allocated at register_netdev() time 1733 * @real_num_rx_queues: Number of RX queues currently active in device 1734 * @xdp_prog: XDP sockets filter program pointer 1735 * @gro_flush_timeout: timeout for GRO layer in NAPI 1736 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1737 * allow to avoid NIC hard IRQ, on busy queues. 1738 * 1739 * @rx_handler: handler for received packets 1740 * @rx_handler_data: XXX: need comments on this one 1741 * @miniq_ingress: ingress/clsact qdisc specific data for 1742 * ingress processing 1743 * @ingress_queue: XXX: need comments on this one 1744 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1745 * @broadcast: hw bcast address 1746 * 1747 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1748 * indexed by RX queue number. Assigned by driver. 1749 * This must only be set if the ndo_rx_flow_steer 1750 * operation is defined 1751 * @index_hlist: Device index hash chain 1752 * 1753 * @_tx: Array of TX queues 1754 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1755 * @real_num_tx_queues: Number of TX queues currently active in device 1756 * @qdisc: Root qdisc from userspace point of view 1757 * @tx_queue_len: Max frames per queue allowed 1758 * @tx_global_lock: XXX: need comments on this one 1759 * @xdp_bulkq: XDP device bulk queue 1760 * @xps_cpus_map: all CPUs map for XPS device 1761 * @xps_rxqs_map: all RXQs map for XPS device 1762 * 1763 * @xps_maps: XXX: need comments on this one 1764 * @miniq_egress: clsact qdisc specific data for 1765 * egress processing 1766 * @qdisc_hash: qdisc hash table 1767 * @watchdog_timeo: Represents the timeout that is used by 1768 * the watchdog (see dev_watchdog()) 1769 * @watchdog_timer: List of timers 1770 * 1771 * @proto_down_reason: reason a netdev interface is held down 1772 * @pcpu_refcnt: Number of references to this device 1773 * @todo_list: Delayed register/unregister 1774 * @link_watch_list: XXX: need comments on this one 1775 * 1776 * @reg_state: Register/unregister state machine 1777 * @dismantle: Device is going to be freed 1778 * @rtnl_link_state: This enum represents the phases of creating 1779 * a new link 1780 * 1781 * @needs_free_netdev: Should unregister perform free_netdev? 1782 * @priv_destructor: Called from unregister 1783 * @npinfo: XXX: need comments on this one 1784 * @nd_net: Network namespace this network device is inside 1785 * 1786 * @ml_priv: Mid-layer private 1787 * @lstats: Loopback statistics 1788 * @tstats: Tunnel statistics 1789 * @dstats: Dummy statistics 1790 * @vstats: Virtual ethernet statistics 1791 * 1792 * @garp_port: GARP 1793 * @mrp_port: MRP 1794 * 1795 * @dev: Class/net/name entry 1796 * @sysfs_groups: Space for optional device, statistics and wireless 1797 * sysfs groups 1798 * 1799 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1800 * @rtnl_link_ops: Rtnl_link_ops 1801 * 1802 * @gso_max_size: Maximum size of generic segmentation offload 1803 * @gso_max_segs: Maximum number of segments that can be passed to the 1804 * NIC for GSO 1805 * 1806 * @dcbnl_ops: Data Center Bridging netlink ops 1807 * @num_tc: Number of traffic classes in the net device 1808 * @tc_to_txq: XXX: need comments on this one 1809 * @prio_tc_map: XXX: need comments on this one 1810 * 1811 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1812 * 1813 * @priomap: XXX: need comments on this one 1814 * @phydev: Physical device may attach itself 1815 * for hardware timestamping 1816 * @sfp_bus: attached &struct sfp_bus structure. 1817 * 1818 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1819 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1820 * 1821 * @proto_down: protocol port state information can be sent to the 1822 * switch driver and used to set the phys state of the 1823 * switch port. 1824 * 1825 * @wol_enabled: Wake-on-LAN is enabled 1826 * 1827 * @net_notifier_list: List of per-net netdev notifier block 1828 * that follow this device when it is moved 1829 * to another network namespace. 1830 * 1831 * @macsec_ops: MACsec offloading ops 1832 * 1833 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1834 * offload capabilities of the device 1835 * @udp_tunnel_nic: UDP tunnel offload state 1836 * @xdp_state: stores info on attached XDP BPF programs 1837 * 1838 * @nested_level: Used as as a parameter of spin_lock_nested() of 1839 * dev->addr_list_lock. 1840 * @unlink_list: As netif_addr_lock() can be called recursively, 1841 * keep a list of interfaces to be deleted. 1842 * 1843 * FIXME: cleanup struct net_device such that network protocol info 1844 * moves out. 1845 */ 1846 1847 struct net_device { 1848 char name[IFNAMSIZ]; 1849 struct netdev_name_node *name_node; 1850 struct dev_ifalias __rcu *ifalias; 1851 /* 1852 * I/O specific fields 1853 * FIXME: Merge these and struct ifmap into one 1854 */ 1855 unsigned long mem_end; 1856 unsigned long mem_start; 1857 unsigned long base_addr; 1858 int irq; 1859 1860 /* 1861 * Some hardware also needs these fields (state,dev_list, 1862 * napi_list,unreg_list,close_list) but they are not 1863 * part of the usual set specified in Space.c. 1864 */ 1865 1866 unsigned long state; 1867 1868 struct list_head dev_list; 1869 struct list_head napi_list; 1870 struct list_head unreg_list; 1871 struct list_head close_list; 1872 struct list_head ptype_all; 1873 struct list_head ptype_specific; 1874 1875 struct { 1876 struct list_head upper; 1877 struct list_head lower; 1878 } adj_list; 1879 1880 netdev_features_t features; 1881 netdev_features_t hw_features; 1882 netdev_features_t wanted_features; 1883 netdev_features_t vlan_features; 1884 netdev_features_t hw_enc_features; 1885 netdev_features_t mpls_features; 1886 netdev_features_t gso_partial_features; 1887 1888 int ifindex; 1889 int group; 1890 1891 struct net_device_stats stats; 1892 1893 atomic_long_t rx_dropped; 1894 atomic_long_t tx_dropped; 1895 atomic_long_t rx_nohandler; 1896 1897 /* Stats to monitor link on/off, flapping */ 1898 atomic_t carrier_up_count; 1899 atomic_t carrier_down_count; 1900 1901 #ifdef CONFIG_WIRELESS_EXT 1902 const struct iw_handler_def *wireless_handlers; 1903 struct iw_public_data *wireless_data; 1904 #endif 1905 const struct net_device_ops *netdev_ops; 1906 const struct ethtool_ops *ethtool_ops; 1907 #ifdef CONFIG_NET_L3_MASTER_DEV 1908 const struct l3mdev_ops *l3mdev_ops; 1909 #endif 1910 #if IS_ENABLED(CONFIG_IPV6) 1911 const struct ndisc_ops *ndisc_ops; 1912 #endif 1913 1914 #ifdef CONFIG_XFRM_OFFLOAD 1915 const struct xfrmdev_ops *xfrmdev_ops; 1916 #endif 1917 1918 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1919 const struct tlsdev_ops *tlsdev_ops; 1920 #endif 1921 1922 const struct header_ops *header_ops; 1923 1924 unsigned int flags; 1925 unsigned int priv_flags; 1926 1927 unsigned short gflags; 1928 unsigned short padded; 1929 1930 unsigned char operstate; 1931 unsigned char link_mode; 1932 1933 unsigned char if_port; 1934 unsigned char dma; 1935 1936 /* Note : dev->mtu is often read without holding a lock. 1937 * Writers usually hold RTNL. 1938 * It is recommended to use READ_ONCE() to annotate the reads, 1939 * and to use WRITE_ONCE() to annotate the writes. 1940 */ 1941 unsigned int mtu; 1942 unsigned int min_mtu; 1943 unsigned int max_mtu; 1944 unsigned short type; 1945 unsigned short hard_header_len; 1946 unsigned char min_header_len; 1947 unsigned char name_assign_type; 1948 1949 unsigned short needed_headroom; 1950 unsigned short needed_tailroom; 1951 1952 /* Interface address info. */ 1953 unsigned char perm_addr[MAX_ADDR_LEN]; 1954 unsigned char addr_assign_type; 1955 unsigned char addr_len; 1956 unsigned char upper_level; 1957 unsigned char lower_level; 1958 1959 unsigned short neigh_priv_len; 1960 unsigned short dev_id; 1961 unsigned short dev_port; 1962 spinlock_t addr_list_lock; 1963 1964 struct netdev_hw_addr_list uc; 1965 struct netdev_hw_addr_list mc; 1966 struct netdev_hw_addr_list dev_addrs; 1967 1968 #ifdef CONFIG_SYSFS 1969 struct kset *queues_kset; 1970 #endif 1971 #ifdef CONFIG_LOCKDEP 1972 struct list_head unlink_list; 1973 #endif 1974 unsigned int promiscuity; 1975 unsigned int allmulti; 1976 bool uc_promisc; 1977 #ifdef CONFIG_LOCKDEP 1978 unsigned char nested_level; 1979 #endif 1980 1981 1982 /* Protocol-specific pointers */ 1983 1984 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1985 struct vlan_info __rcu *vlan_info; 1986 #endif 1987 #if IS_ENABLED(CONFIG_NET_DSA) 1988 struct dsa_port *dsa_ptr; 1989 #endif 1990 #if IS_ENABLED(CONFIG_TIPC) 1991 struct tipc_bearer __rcu *tipc_ptr; 1992 #endif 1993 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1994 void *atalk_ptr; 1995 #endif 1996 struct in_device __rcu *ip_ptr; 1997 #if IS_ENABLED(CONFIG_DECNET) 1998 struct dn_dev __rcu *dn_ptr; 1999 #endif 2000 struct inet6_dev __rcu *ip6_ptr; 2001 #if IS_ENABLED(CONFIG_AX25) 2002 void *ax25_ptr; 2003 #endif 2004 struct wireless_dev *ieee80211_ptr; 2005 struct wpan_dev *ieee802154_ptr; 2006 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2007 struct mpls_dev __rcu *mpls_ptr; 2008 #endif 2009 2010 /* 2011 * Cache lines mostly used on receive path (including eth_type_trans()) 2012 */ 2013 /* Interface address info used in eth_type_trans() */ 2014 unsigned char *dev_addr; 2015 2016 struct netdev_rx_queue *_rx; 2017 unsigned int num_rx_queues; 2018 unsigned int real_num_rx_queues; 2019 2020 struct bpf_prog __rcu *xdp_prog; 2021 unsigned long gro_flush_timeout; 2022 int napi_defer_hard_irqs; 2023 rx_handler_func_t __rcu *rx_handler; 2024 void __rcu *rx_handler_data; 2025 2026 #ifdef CONFIG_NET_CLS_ACT 2027 struct mini_Qdisc __rcu *miniq_ingress; 2028 #endif 2029 struct netdev_queue __rcu *ingress_queue; 2030 #ifdef CONFIG_NETFILTER_INGRESS 2031 struct nf_hook_entries __rcu *nf_hooks_ingress; 2032 #endif 2033 2034 unsigned char broadcast[MAX_ADDR_LEN]; 2035 #ifdef CONFIG_RFS_ACCEL 2036 struct cpu_rmap *rx_cpu_rmap; 2037 #endif 2038 struct hlist_node index_hlist; 2039 2040 /* 2041 * Cache lines mostly used on transmit path 2042 */ 2043 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2044 unsigned int num_tx_queues; 2045 unsigned int real_num_tx_queues; 2046 struct Qdisc *qdisc; 2047 unsigned int tx_queue_len; 2048 spinlock_t tx_global_lock; 2049 2050 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2051 2052 #ifdef CONFIG_XPS 2053 struct xps_dev_maps __rcu *xps_cpus_map; 2054 struct xps_dev_maps __rcu *xps_rxqs_map; 2055 #endif 2056 #ifdef CONFIG_NET_CLS_ACT 2057 struct mini_Qdisc __rcu *miniq_egress; 2058 #endif 2059 2060 #ifdef CONFIG_NET_SCHED 2061 DECLARE_HASHTABLE (qdisc_hash, 4); 2062 #endif 2063 /* These may be needed for future network-power-down code. */ 2064 struct timer_list watchdog_timer; 2065 int watchdog_timeo; 2066 2067 u32 proto_down_reason; 2068 2069 struct list_head todo_list; 2070 int __percpu *pcpu_refcnt; 2071 2072 struct list_head link_watch_list; 2073 2074 enum { NETREG_UNINITIALIZED=0, 2075 NETREG_REGISTERED, /* completed register_netdevice */ 2076 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2077 NETREG_UNREGISTERED, /* completed unregister todo */ 2078 NETREG_RELEASED, /* called free_netdev */ 2079 NETREG_DUMMY, /* dummy device for NAPI poll */ 2080 } reg_state:8; 2081 2082 bool dismantle; 2083 2084 enum { 2085 RTNL_LINK_INITIALIZED, 2086 RTNL_LINK_INITIALIZING, 2087 } rtnl_link_state:16; 2088 2089 bool needs_free_netdev; 2090 void (*priv_destructor)(struct net_device *dev); 2091 2092 #ifdef CONFIG_NETPOLL 2093 struct netpoll_info __rcu *npinfo; 2094 #endif 2095 2096 possible_net_t nd_net; 2097 2098 /* mid-layer private */ 2099 union { 2100 void *ml_priv; 2101 struct pcpu_lstats __percpu *lstats; 2102 struct pcpu_sw_netstats __percpu *tstats; 2103 struct pcpu_dstats __percpu *dstats; 2104 }; 2105 2106 #if IS_ENABLED(CONFIG_GARP) 2107 struct garp_port __rcu *garp_port; 2108 #endif 2109 #if IS_ENABLED(CONFIG_MRP) 2110 struct mrp_port __rcu *mrp_port; 2111 #endif 2112 2113 struct device dev; 2114 const struct attribute_group *sysfs_groups[4]; 2115 const struct attribute_group *sysfs_rx_queue_group; 2116 2117 const struct rtnl_link_ops *rtnl_link_ops; 2118 2119 /* for setting kernel sock attribute on TCP connection setup */ 2120 #define GSO_MAX_SIZE 65536 2121 unsigned int gso_max_size; 2122 #define GSO_MAX_SEGS 65535 2123 u16 gso_max_segs; 2124 2125 #ifdef CONFIG_DCB 2126 const struct dcbnl_rtnl_ops *dcbnl_ops; 2127 #endif 2128 s16 num_tc; 2129 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2130 u8 prio_tc_map[TC_BITMASK + 1]; 2131 2132 #if IS_ENABLED(CONFIG_FCOE) 2133 unsigned int fcoe_ddp_xid; 2134 #endif 2135 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2136 struct netprio_map __rcu *priomap; 2137 #endif 2138 struct phy_device *phydev; 2139 struct sfp_bus *sfp_bus; 2140 struct lock_class_key *qdisc_tx_busylock; 2141 struct lock_class_key *qdisc_running_key; 2142 bool proto_down; 2143 unsigned wol_enabled:1; 2144 2145 struct list_head net_notifier_list; 2146 2147 #if IS_ENABLED(CONFIG_MACSEC) 2148 /* MACsec management functions */ 2149 const struct macsec_ops *macsec_ops; 2150 #endif 2151 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2152 struct udp_tunnel_nic *udp_tunnel_nic; 2153 2154 /* protected by rtnl_lock */ 2155 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2156 }; 2157 #define to_net_dev(d) container_of(d, struct net_device, dev) 2158 2159 static inline bool netif_elide_gro(const struct net_device *dev) 2160 { 2161 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2162 return true; 2163 return false; 2164 } 2165 2166 #define NETDEV_ALIGN 32 2167 2168 static inline 2169 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2170 { 2171 return dev->prio_tc_map[prio & TC_BITMASK]; 2172 } 2173 2174 static inline 2175 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2176 { 2177 if (tc >= dev->num_tc) 2178 return -EINVAL; 2179 2180 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2181 return 0; 2182 } 2183 2184 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2185 void netdev_reset_tc(struct net_device *dev); 2186 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2187 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2188 2189 static inline 2190 int netdev_get_num_tc(struct net_device *dev) 2191 { 2192 return dev->num_tc; 2193 } 2194 2195 static inline void net_prefetch(void *p) 2196 { 2197 prefetch(p); 2198 #if L1_CACHE_BYTES < 128 2199 prefetch((u8 *)p + L1_CACHE_BYTES); 2200 #endif 2201 } 2202 2203 static inline void net_prefetchw(void *p) 2204 { 2205 prefetchw(p); 2206 #if L1_CACHE_BYTES < 128 2207 prefetchw((u8 *)p + L1_CACHE_BYTES); 2208 #endif 2209 } 2210 2211 void netdev_unbind_sb_channel(struct net_device *dev, 2212 struct net_device *sb_dev); 2213 int netdev_bind_sb_channel_queue(struct net_device *dev, 2214 struct net_device *sb_dev, 2215 u8 tc, u16 count, u16 offset); 2216 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2217 static inline int netdev_get_sb_channel(struct net_device *dev) 2218 { 2219 return max_t(int, -dev->num_tc, 0); 2220 } 2221 2222 static inline 2223 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2224 unsigned int index) 2225 { 2226 return &dev->_tx[index]; 2227 } 2228 2229 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2230 const struct sk_buff *skb) 2231 { 2232 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2233 } 2234 2235 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2236 void (*f)(struct net_device *, 2237 struct netdev_queue *, 2238 void *), 2239 void *arg) 2240 { 2241 unsigned int i; 2242 2243 for (i = 0; i < dev->num_tx_queues; i++) 2244 f(dev, &dev->_tx[i], arg); 2245 } 2246 2247 #define netdev_lockdep_set_classes(dev) \ 2248 { \ 2249 static struct lock_class_key qdisc_tx_busylock_key; \ 2250 static struct lock_class_key qdisc_running_key; \ 2251 static struct lock_class_key qdisc_xmit_lock_key; \ 2252 static struct lock_class_key dev_addr_list_lock_key; \ 2253 unsigned int i; \ 2254 \ 2255 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2256 (dev)->qdisc_running_key = &qdisc_running_key; \ 2257 lockdep_set_class(&(dev)->addr_list_lock, \ 2258 &dev_addr_list_lock_key); \ 2259 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2260 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2261 &qdisc_xmit_lock_key); \ 2262 } 2263 2264 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2265 struct net_device *sb_dev); 2266 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2267 struct sk_buff *skb, 2268 struct net_device *sb_dev); 2269 2270 /* returns the headroom that the master device needs to take in account 2271 * when forwarding to this dev 2272 */ 2273 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2274 { 2275 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2276 } 2277 2278 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2279 { 2280 if (dev->netdev_ops->ndo_set_rx_headroom) 2281 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2282 } 2283 2284 /* set the device rx headroom to the dev's default */ 2285 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2286 { 2287 netdev_set_rx_headroom(dev, -1); 2288 } 2289 2290 /* 2291 * Net namespace inlines 2292 */ 2293 static inline 2294 struct net *dev_net(const struct net_device *dev) 2295 { 2296 return read_pnet(&dev->nd_net); 2297 } 2298 2299 static inline 2300 void dev_net_set(struct net_device *dev, struct net *net) 2301 { 2302 write_pnet(&dev->nd_net, net); 2303 } 2304 2305 /** 2306 * netdev_priv - access network device private data 2307 * @dev: network device 2308 * 2309 * Get network device private data 2310 */ 2311 static inline void *netdev_priv(const struct net_device *dev) 2312 { 2313 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2314 } 2315 2316 /* Set the sysfs physical device reference for the network logical device 2317 * if set prior to registration will cause a symlink during initialization. 2318 */ 2319 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2320 2321 /* Set the sysfs device type for the network logical device to allow 2322 * fine-grained identification of different network device types. For 2323 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2324 */ 2325 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2326 2327 /* Default NAPI poll() weight 2328 * Device drivers are strongly advised to not use bigger value 2329 */ 2330 #define NAPI_POLL_WEIGHT 64 2331 2332 /** 2333 * netif_napi_add - initialize a NAPI context 2334 * @dev: network device 2335 * @napi: NAPI context 2336 * @poll: polling function 2337 * @weight: default weight 2338 * 2339 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2340 * *any* of the other NAPI-related functions. 2341 */ 2342 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2343 int (*poll)(struct napi_struct *, int), int weight); 2344 2345 /** 2346 * netif_tx_napi_add - initialize a NAPI context 2347 * @dev: network device 2348 * @napi: NAPI context 2349 * @poll: polling function 2350 * @weight: default weight 2351 * 2352 * This variant of netif_napi_add() should be used from drivers using NAPI 2353 * to exclusively poll a TX queue. 2354 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2355 */ 2356 static inline void netif_tx_napi_add(struct net_device *dev, 2357 struct napi_struct *napi, 2358 int (*poll)(struct napi_struct *, int), 2359 int weight) 2360 { 2361 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2362 netif_napi_add(dev, napi, poll, weight); 2363 } 2364 2365 /** 2366 * __netif_napi_del - remove a NAPI context 2367 * @napi: NAPI context 2368 * 2369 * Warning: caller must observe RCU grace period before freeing memory 2370 * containing @napi. Drivers might want to call this helper to combine 2371 * all the needed RCU grace periods into a single one. 2372 */ 2373 void __netif_napi_del(struct napi_struct *napi); 2374 2375 /** 2376 * netif_napi_del - remove a NAPI context 2377 * @napi: NAPI context 2378 * 2379 * netif_napi_del() removes a NAPI context from the network device NAPI list 2380 */ 2381 static inline void netif_napi_del(struct napi_struct *napi) 2382 { 2383 __netif_napi_del(napi); 2384 synchronize_net(); 2385 } 2386 2387 struct napi_gro_cb { 2388 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2389 void *frag0; 2390 2391 /* Length of frag0. */ 2392 unsigned int frag0_len; 2393 2394 /* This indicates where we are processing relative to skb->data. */ 2395 int data_offset; 2396 2397 /* This is non-zero if the packet cannot be merged with the new skb. */ 2398 u16 flush; 2399 2400 /* Save the IP ID here and check when we get to the transport layer */ 2401 u16 flush_id; 2402 2403 /* Number of segments aggregated. */ 2404 u16 count; 2405 2406 /* Start offset for remote checksum offload */ 2407 u16 gro_remcsum_start; 2408 2409 /* jiffies when first packet was created/queued */ 2410 unsigned long age; 2411 2412 /* Used in ipv6_gro_receive() and foo-over-udp */ 2413 u16 proto; 2414 2415 /* This is non-zero if the packet may be of the same flow. */ 2416 u8 same_flow:1; 2417 2418 /* Used in tunnel GRO receive */ 2419 u8 encap_mark:1; 2420 2421 /* GRO checksum is valid */ 2422 u8 csum_valid:1; 2423 2424 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2425 u8 csum_cnt:3; 2426 2427 /* Free the skb? */ 2428 u8 free:2; 2429 #define NAPI_GRO_FREE 1 2430 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2431 2432 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2433 u8 is_ipv6:1; 2434 2435 /* Used in GRE, set in fou/gue_gro_receive */ 2436 u8 is_fou:1; 2437 2438 /* Used to determine if flush_id can be ignored */ 2439 u8 is_atomic:1; 2440 2441 /* Number of gro_receive callbacks this packet already went through */ 2442 u8 recursion_counter:4; 2443 2444 /* GRO is done by frag_list pointer chaining. */ 2445 u8 is_flist:1; 2446 2447 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2448 __wsum csum; 2449 2450 /* used in skb_gro_receive() slow path */ 2451 struct sk_buff *last; 2452 }; 2453 2454 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2455 2456 #define GRO_RECURSION_LIMIT 15 2457 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2458 { 2459 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2460 } 2461 2462 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2463 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2464 struct list_head *head, 2465 struct sk_buff *skb) 2466 { 2467 if (unlikely(gro_recursion_inc_test(skb))) { 2468 NAPI_GRO_CB(skb)->flush |= 1; 2469 return NULL; 2470 } 2471 2472 return cb(head, skb); 2473 } 2474 2475 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2476 struct sk_buff *); 2477 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2478 struct sock *sk, 2479 struct list_head *head, 2480 struct sk_buff *skb) 2481 { 2482 if (unlikely(gro_recursion_inc_test(skb))) { 2483 NAPI_GRO_CB(skb)->flush |= 1; 2484 return NULL; 2485 } 2486 2487 return cb(sk, head, skb); 2488 } 2489 2490 struct packet_type { 2491 __be16 type; /* This is really htons(ether_type). */ 2492 bool ignore_outgoing; 2493 struct net_device *dev; /* NULL is wildcarded here */ 2494 int (*func) (struct sk_buff *, 2495 struct net_device *, 2496 struct packet_type *, 2497 struct net_device *); 2498 void (*list_func) (struct list_head *, 2499 struct packet_type *, 2500 struct net_device *); 2501 bool (*id_match)(struct packet_type *ptype, 2502 struct sock *sk); 2503 void *af_packet_priv; 2504 struct list_head list; 2505 }; 2506 2507 struct offload_callbacks { 2508 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2509 netdev_features_t features); 2510 struct sk_buff *(*gro_receive)(struct list_head *head, 2511 struct sk_buff *skb); 2512 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2513 }; 2514 2515 struct packet_offload { 2516 __be16 type; /* This is really htons(ether_type). */ 2517 u16 priority; 2518 struct offload_callbacks callbacks; 2519 struct list_head list; 2520 }; 2521 2522 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2523 struct pcpu_sw_netstats { 2524 u64 rx_packets; 2525 u64 rx_bytes; 2526 u64 tx_packets; 2527 u64 tx_bytes; 2528 struct u64_stats_sync syncp; 2529 } __aligned(4 * sizeof(u64)); 2530 2531 struct pcpu_lstats { 2532 u64_stats_t packets; 2533 u64_stats_t bytes; 2534 struct u64_stats_sync syncp; 2535 } __aligned(2 * sizeof(u64)); 2536 2537 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2538 2539 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2540 { 2541 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2542 2543 u64_stats_update_begin(&tstats->syncp); 2544 tstats->rx_bytes += len; 2545 tstats->rx_packets++; 2546 u64_stats_update_end(&tstats->syncp); 2547 } 2548 2549 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2550 unsigned int packets, 2551 unsigned int len) 2552 { 2553 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2554 2555 u64_stats_update_begin(&tstats->syncp); 2556 tstats->tx_bytes += len; 2557 tstats->tx_packets += packets; 2558 u64_stats_update_end(&tstats->syncp); 2559 } 2560 2561 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2562 { 2563 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2564 2565 u64_stats_update_begin(&lstats->syncp); 2566 u64_stats_add(&lstats->bytes, len); 2567 u64_stats_inc(&lstats->packets); 2568 u64_stats_update_end(&lstats->syncp); 2569 } 2570 2571 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2572 ({ \ 2573 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2574 if (pcpu_stats) { \ 2575 int __cpu; \ 2576 for_each_possible_cpu(__cpu) { \ 2577 typeof(type) *stat; \ 2578 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2579 u64_stats_init(&stat->syncp); \ 2580 } \ 2581 } \ 2582 pcpu_stats; \ 2583 }) 2584 2585 #define netdev_alloc_pcpu_stats(type) \ 2586 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2587 2588 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2589 ({ \ 2590 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2591 if (pcpu_stats) { \ 2592 int __cpu; \ 2593 for_each_possible_cpu(__cpu) { \ 2594 typeof(type) *stat; \ 2595 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2596 u64_stats_init(&stat->syncp); \ 2597 } \ 2598 } \ 2599 pcpu_stats; \ 2600 }) 2601 2602 enum netdev_lag_tx_type { 2603 NETDEV_LAG_TX_TYPE_UNKNOWN, 2604 NETDEV_LAG_TX_TYPE_RANDOM, 2605 NETDEV_LAG_TX_TYPE_BROADCAST, 2606 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2607 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2608 NETDEV_LAG_TX_TYPE_HASH, 2609 }; 2610 2611 enum netdev_lag_hash { 2612 NETDEV_LAG_HASH_NONE, 2613 NETDEV_LAG_HASH_L2, 2614 NETDEV_LAG_HASH_L34, 2615 NETDEV_LAG_HASH_L23, 2616 NETDEV_LAG_HASH_E23, 2617 NETDEV_LAG_HASH_E34, 2618 NETDEV_LAG_HASH_UNKNOWN, 2619 }; 2620 2621 struct netdev_lag_upper_info { 2622 enum netdev_lag_tx_type tx_type; 2623 enum netdev_lag_hash hash_type; 2624 }; 2625 2626 struct netdev_lag_lower_state_info { 2627 u8 link_up : 1, 2628 tx_enabled : 1; 2629 }; 2630 2631 #include <linux/notifier.h> 2632 2633 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2634 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2635 * adding new types. 2636 */ 2637 enum netdev_cmd { 2638 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2639 NETDEV_DOWN, 2640 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2641 detected a hardware crash and restarted 2642 - we can use this eg to kick tcp sessions 2643 once done */ 2644 NETDEV_CHANGE, /* Notify device state change */ 2645 NETDEV_REGISTER, 2646 NETDEV_UNREGISTER, 2647 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2648 NETDEV_CHANGEADDR, /* notify after the address change */ 2649 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2650 NETDEV_GOING_DOWN, 2651 NETDEV_CHANGENAME, 2652 NETDEV_FEAT_CHANGE, 2653 NETDEV_BONDING_FAILOVER, 2654 NETDEV_PRE_UP, 2655 NETDEV_PRE_TYPE_CHANGE, 2656 NETDEV_POST_TYPE_CHANGE, 2657 NETDEV_POST_INIT, 2658 NETDEV_RELEASE, 2659 NETDEV_NOTIFY_PEERS, 2660 NETDEV_JOIN, 2661 NETDEV_CHANGEUPPER, 2662 NETDEV_RESEND_IGMP, 2663 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2664 NETDEV_CHANGEINFODATA, 2665 NETDEV_BONDING_INFO, 2666 NETDEV_PRECHANGEUPPER, 2667 NETDEV_CHANGELOWERSTATE, 2668 NETDEV_UDP_TUNNEL_PUSH_INFO, 2669 NETDEV_UDP_TUNNEL_DROP_INFO, 2670 NETDEV_CHANGE_TX_QUEUE_LEN, 2671 NETDEV_CVLAN_FILTER_PUSH_INFO, 2672 NETDEV_CVLAN_FILTER_DROP_INFO, 2673 NETDEV_SVLAN_FILTER_PUSH_INFO, 2674 NETDEV_SVLAN_FILTER_DROP_INFO, 2675 }; 2676 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2677 2678 int register_netdevice_notifier(struct notifier_block *nb); 2679 int unregister_netdevice_notifier(struct notifier_block *nb); 2680 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2681 int unregister_netdevice_notifier_net(struct net *net, 2682 struct notifier_block *nb); 2683 int register_netdevice_notifier_dev_net(struct net_device *dev, 2684 struct notifier_block *nb, 2685 struct netdev_net_notifier *nn); 2686 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2687 struct notifier_block *nb, 2688 struct netdev_net_notifier *nn); 2689 2690 struct netdev_notifier_info { 2691 struct net_device *dev; 2692 struct netlink_ext_ack *extack; 2693 }; 2694 2695 struct netdev_notifier_info_ext { 2696 struct netdev_notifier_info info; /* must be first */ 2697 union { 2698 u32 mtu; 2699 } ext; 2700 }; 2701 2702 struct netdev_notifier_change_info { 2703 struct netdev_notifier_info info; /* must be first */ 2704 unsigned int flags_changed; 2705 }; 2706 2707 struct netdev_notifier_changeupper_info { 2708 struct netdev_notifier_info info; /* must be first */ 2709 struct net_device *upper_dev; /* new upper dev */ 2710 bool master; /* is upper dev master */ 2711 bool linking; /* is the notification for link or unlink */ 2712 void *upper_info; /* upper dev info */ 2713 }; 2714 2715 struct netdev_notifier_changelowerstate_info { 2716 struct netdev_notifier_info info; /* must be first */ 2717 void *lower_state_info; /* is lower dev state */ 2718 }; 2719 2720 struct netdev_notifier_pre_changeaddr_info { 2721 struct netdev_notifier_info info; /* must be first */ 2722 const unsigned char *dev_addr; 2723 }; 2724 2725 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2726 struct net_device *dev) 2727 { 2728 info->dev = dev; 2729 info->extack = NULL; 2730 } 2731 2732 static inline struct net_device * 2733 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2734 { 2735 return info->dev; 2736 } 2737 2738 static inline struct netlink_ext_ack * 2739 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2740 { 2741 return info->extack; 2742 } 2743 2744 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2745 2746 2747 extern rwlock_t dev_base_lock; /* Device list lock */ 2748 2749 #define for_each_netdev(net, d) \ 2750 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2751 #define for_each_netdev_reverse(net, d) \ 2752 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2753 #define for_each_netdev_rcu(net, d) \ 2754 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2755 #define for_each_netdev_safe(net, d, n) \ 2756 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2757 #define for_each_netdev_continue(net, d) \ 2758 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2759 #define for_each_netdev_continue_reverse(net, d) \ 2760 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2761 dev_list) 2762 #define for_each_netdev_continue_rcu(net, d) \ 2763 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2764 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2765 for_each_netdev_rcu(&init_net, slave) \ 2766 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2767 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2768 2769 static inline struct net_device *next_net_device(struct net_device *dev) 2770 { 2771 struct list_head *lh; 2772 struct net *net; 2773 2774 net = dev_net(dev); 2775 lh = dev->dev_list.next; 2776 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2777 } 2778 2779 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2780 { 2781 struct list_head *lh; 2782 struct net *net; 2783 2784 net = dev_net(dev); 2785 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2786 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2787 } 2788 2789 static inline struct net_device *first_net_device(struct net *net) 2790 { 2791 return list_empty(&net->dev_base_head) ? NULL : 2792 net_device_entry(net->dev_base_head.next); 2793 } 2794 2795 static inline struct net_device *first_net_device_rcu(struct net *net) 2796 { 2797 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2798 2799 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2800 } 2801 2802 int netdev_boot_setup_check(struct net_device *dev); 2803 unsigned long netdev_boot_base(const char *prefix, int unit); 2804 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2805 const char *hwaddr); 2806 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2807 void dev_add_pack(struct packet_type *pt); 2808 void dev_remove_pack(struct packet_type *pt); 2809 void __dev_remove_pack(struct packet_type *pt); 2810 void dev_add_offload(struct packet_offload *po); 2811 void dev_remove_offload(struct packet_offload *po); 2812 2813 int dev_get_iflink(const struct net_device *dev); 2814 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2815 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2816 unsigned short mask); 2817 struct net_device *dev_get_by_name(struct net *net, const char *name); 2818 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2819 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2820 int dev_alloc_name(struct net_device *dev, const char *name); 2821 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2822 void dev_close(struct net_device *dev); 2823 void dev_close_many(struct list_head *head, bool unlink); 2824 void dev_disable_lro(struct net_device *dev); 2825 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2826 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2827 struct net_device *sb_dev); 2828 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2829 struct net_device *sb_dev); 2830 2831 int dev_queue_xmit(struct sk_buff *skb); 2832 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2833 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2834 2835 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2836 { 2837 int ret; 2838 2839 ret = __dev_direct_xmit(skb, queue_id); 2840 if (!dev_xmit_complete(ret)) 2841 kfree_skb(skb); 2842 return ret; 2843 } 2844 2845 int register_netdevice(struct net_device *dev); 2846 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2847 void unregister_netdevice_many(struct list_head *head); 2848 static inline void unregister_netdevice(struct net_device *dev) 2849 { 2850 unregister_netdevice_queue(dev, NULL); 2851 } 2852 2853 int netdev_refcnt_read(const struct net_device *dev); 2854 void free_netdev(struct net_device *dev); 2855 void netdev_freemem(struct net_device *dev); 2856 int init_dummy_netdev(struct net_device *dev); 2857 2858 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2859 struct sk_buff *skb, 2860 bool all_slaves); 2861 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2862 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2863 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2864 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2865 int netdev_get_name(struct net *net, char *name, int ifindex); 2866 int dev_restart(struct net_device *dev); 2867 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2868 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2869 2870 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2871 { 2872 return NAPI_GRO_CB(skb)->data_offset; 2873 } 2874 2875 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2876 { 2877 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2878 } 2879 2880 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2881 { 2882 NAPI_GRO_CB(skb)->data_offset += len; 2883 } 2884 2885 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2886 unsigned int offset) 2887 { 2888 return NAPI_GRO_CB(skb)->frag0 + offset; 2889 } 2890 2891 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2892 { 2893 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2894 } 2895 2896 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2897 { 2898 NAPI_GRO_CB(skb)->frag0 = NULL; 2899 NAPI_GRO_CB(skb)->frag0_len = 0; 2900 } 2901 2902 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2903 unsigned int offset) 2904 { 2905 if (!pskb_may_pull(skb, hlen)) 2906 return NULL; 2907 2908 skb_gro_frag0_invalidate(skb); 2909 return skb->data + offset; 2910 } 2911 2912 static inline void *skb_gro_network_header(struct sk_buff *skb) 2913 { 2914 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2915 skb_network_offset(skb); 2916 } 2917 2918 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2919 const void *start, unsigned int len) 2920 { 2921 if (NAPI_GRO_CB(skb)->csum_valid) 2922 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2923 csum_partial(start, len, 0)); 2924 } 2925 2926 /* GRO checksum functions. These are logical equivalents of the normal 2927 * checksum functions (in skbuff.h) except that they operate on the GRO 2928 * offsets and fields in sk_buff. 2929 */ 2930 2931 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2932 2933 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2934 { 2935 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2936 } 2937 2938 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2939 bool zero_okay, 2940 __sum16 check) 2941 { 2942 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2943 skb_checksum_start_offset(skb) < 2944 skb_gro_offset(skb)) && 2945 !skb_at_gro_remcsum_start(skb) && 2946 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2947 (!zero_okay || check)); 2948 } 2949 2950 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2951 __wsum psum) 2952 { 2953 if (NAPI_GRO_CB(skb)->csum_valid && 2954 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2955 return 0; 2956 2957 NAPI_GRO_CB(skb)->csum = psum; 2958 2959 return __skb_gro_checksum_complete(skb); 2960 } 2961 2962 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2963 { 2964 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2965 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2966 NAPI_GRO_CB(skb)->csum_cnt--; 2967 } else { 2968 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2969 * verified a new top level checksum or an encapsulated one 2970 * during GRO. This saves work if we fallback to normal path. 2971 */ 2972 __skb_incr_checksum_unnecessary(skb); 2973 } 2974 } 2975 2976 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2977 compute_pseudo) \ 2978 ({ \ 2979 __sum16 __ret = 0; \ 2980 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2981 __ret = __skb_gro_checksum_validate_complete(skb, \ 2982 compute_pseudo(skb, proto)); \ 2983 if (!__ret) \ 2984 skb_gro_incr_csum_unnecessary(skb); \ 2985 __ret; \ 2986 }) 2987 2988 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2989 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2990 2991 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2992 compute_pseudo) \ 2993 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2994 2995 #define skb_gro_checksum_simple_validate(skb) \ 2996 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2997 2998 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2999 { 3000 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3001 !NAPI_GRO_CB(skb)->csum_valid); 3002 } 3003 3004 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3005 __wsum pseudo) 3006 { 3007 NAPI_GRO_CB(skb)->csum = ~pseudo; 3008 NAPI_GRO_CB(skb)->csum_valid = 1; 3009 } 3010 3011 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3012 do { \ 3013 if (__skb_gro_checksum_convert_check(skb)) \ 3014 __skb_gro_checksum_convert(skb, \ 3015 compute_pseudo(skb, proto)); \ 3016 } while (0) 3017 3018 struct gro_remcsum { 3019 int offset; 3020 __wsum delta; 3021 }; 3022 3023 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3024 { 3025 grc->offset = 0; 3026 grc->delta = 0; 3027 } 3028 3029 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3030 unsigned int off, size_t hdrlen, 3031 int start, int offset, 3032 struct gro_remcsum *grc, 3033 bool nopartial) 3034 { 3035 __wsum delta; 3036 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3037 3038 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3039 3040 if (!nopartial) { 3041 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3042 return ptr; 3043 } 3044 3045 ptr = skb_gro_header_fast(skb, off); 3046 if (skb_gro_header_hard(skb, off + plen)) { 3047 ptr = skb_gro_header_slow(skb, off + plen, off); 3048 if (!ptr) 3049 return NULL; 3050 } 3051 3052 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3053 start, offset); 3054 3055 /* Adjust skb->csum since we changed the packet */ 3056 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3057 3058 grc->offset = off + hdrlen + offset; 3059 grc->delta = delta; 3060 3061 return ptr; 3062 } 3063 3064 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3065 struct gro_remcsum *grc) 3066 { 3067 void *ptr; 3068 size_t plen = grc->offset + sizeof(u16); 3069 3070 if (!grc->delta) 3071 return; 3072 3073 ptr = skb_gro_header_fast(skb, grc->offset); 3074 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3075 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3076 if (!ptr) 3077 return; 3078 } 3079 3080 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3081 } 3082 3083 #ifdef CONFIG_XFRM_OFFLOAD 3084 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3085 { 3086 if (PTR_ERR(pp) != -EINPROGRESS) 3087 NAPI_GRO_CB(skb)->flush |= flush; 3088 } 3089 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3090 struct sk_buff *pp, 3091 int flush, 3092 struct gro_remcsum *grc) 3093 { 3094 if (PTR_ERR(pp) != -EINPROGRESS) { 3095 NAPI_GRO_CB(skb)->flush |= flush; 3096 skb_gro_remcsum_cleanup(skb, grc); 3097 skb->remcsum_offload = 0; 3098 } 3099 } 3100 #else 3101 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3102 { 3103 NAPI_GRO_CB(skb)->flush |= flush; 3104 } 3105 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3106 struct sk_buff *pp, 3107 int flush, 3108 struct gro_remcsum *grc) 3109 { 3110 NAPI_GRO_CB(skb)->flush |= flush; 3111 skb_gro_remcsum_cleanup(skb, grc); 3112 skb->remcsum_offload = 0; 3113 } 3114 #endif 3115 3116 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3117 unsigned short type, 3118 const void *daddr, const void *saddr, 3119 unsigned int len) 3120 { 3121 if (!dev->header_ops || !dev->header_ops->create) 3122 return 0; 3123 3124 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3125 } 3126 3127 static inline int dev_parse_header(const struct sk_buff *skb, 3128 unsigned char *haddr) 3129 { 3130 const struct net_device *dev = skb->dev; 3131 3132 if (!dev->header_ops || !dev->header_ops->parse) 3133 return 0; 3134 return dev->header_ops->parse(skb, haddr); 3135 } 3136 3137 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3138 { 3139 const struct net_device *dev = skb->dev; 3140 3141 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3142 return 0; 3143 return dev->header_ops->parse_protocol(skb); 3144 } 3145 3146 /* ll_header must have at least hard_header_len allocated */ 3147 static inline bool dev_validate_header(const struct net_device *dev, 3148 char *ll_header, int len) 3149 { 3150 if (likely(len >= dev->hard_header_len)) 3151 return true; 3152 if (len < dev->min_header_len) 3153 return false; 3154 3155 if (capable(CAP_SYS_RAWIO)) { 3156 memset(ll_header + len, 0, dev->hard_header_len - len); 3157 return true; 3158 } 3159 3160 if (dev->header_ops && dev->header_ops->validate) 3161 return dev->header_ops->validate(ll_header, len); 3162 3163 return false; 3164 } 3165 3166 static inline bool dev_has_header(const struct net_device *dev) 3167 { 3168 return dev->header_ops && dev->header_ops->create; 3169 } 3170 3171 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3172 int len, int size); 3173 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3174 static inline int unregister_gifconf(unsigned int family) 3175 { 3176 return register_gifconf(family, NULL); 3177 } 3178 3179 #ifdef CONFIG_NET_FLOW_LIMIT 3180 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3181 struct sd_flow_limit { 3182 u64 count; 3183 unsigned int num_buckets; 3184 unsigned int history_head; 3185 u16 history[FLOW_LIMIT_HISTORY]; 3186 u8 buckets[]; 3187 }; 3188 3189 extern int netdev_flow_limit_table_len; 3190 #endif /* CONFIG_NET_FLOW_LIMIT */ 3191 3192 /* 3193 * Incoming packets are placed on per-CPU queues 3194 */ 3195 struct softnet_data { 3196 struct list_head poll_list; 3197 struct sk_buff_head process_queue; 3198 3199 /* stats */ 3200 unsigned int processed; 3201 unsigned int time_squeeze; 3202 unsigned int received_rps; 3203 #ifdef CONFIG_RPS 3204 struct softnet_data *rps_ipi_list; 3205 #endif 3206 #ifdef CONFIG_NET_FLOW_LIMIT 3207 struct sd_flow_limit __rcu *flow_limit; 3208 #endif 3209 struct Qdisc *output_queue; 3210 struct Qdisc **output_queue_tailp; 3211 struct sk_buff *completion_queue; 3212 #ifdef CONFIG_XFRM_OFFLOAD 3213 struct sk_buff_head xfrm_backlog; 3214 #endif 3215 /* written and read only by owning cpu: */ 3216 struct { 3217 u16 recursion; 3218 u8 more; 3219 } xmit; 3220 #ifdef CONFIG_RPS 3221 /* input_queue_head should be written by cpu owning this struct, 3222 * and only read by other cpus. Worth using a cache line. 3223 */ 3224 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3225 3226 /* Elements below can be accessed between CPUs for RPS/RFS */ 3227 call_single_data_t csd ____cacheline_aligned_in_smp; 3228 struct softnet_data *rps_ipi_next; 3229 unsigned int cpu; 3230 unsigned int input_queue_tail; 3231 #endif 3232 unsigned int dropped; 3233 struct sk_buff_head input_pkt_queue; 3234 struct napi_struct backlog; 3235 3236 }; 3237 3238 static inline void input_queue_head_incr(struct softnet_data *sd) 3239 { 3240 #ifdef CONFIG_RPS 3241 sd->input_queue_head++; 3242 #endif 3243 } 3244 3245 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3246 unsigned int *qtail) 3247 { 3248 #ifdef CONFIG_RPS 3249 *qtail = ++sd->input_queue_tail; 3250 #endif 3251 } 3252 3253 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3254 3255 static inline int dev_recursion_level(void) 3256 { 3257 return this_cpu_read(softnet_data.xmit.recursion); 3258 } 3259 3260 #define XMIT_RECURSION_LIMIT 8 3261 static inline bool dev_xmit_recursion(void) 3262 { 3263 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3264 XMIT_RECURSION_LIMIT); 3265 } 3266 3267 static inline void dev_xmit_recursion_inc(void) 3268 { 3269 __this_cpu_inc(softnet_data.xmit.recursion); 3270 } 3271 3272 static inline void dev_xmit_recursion_dec(void) 3273 { 3274 __this_cpu_dec(softnet_data.xmit.recursion); 3275 } 3276 3277 void __netif_schedule(struct Qdisc *q); 3278 void netif_schedule_queue(struct netdev_queue *txq); 3279 3280 static inline void netif_tx_schedule_all(struct net_device *dev) 3281 { 3282 unsigned int i; 3283 3284 for (i = 0; i < dev->num_tx_queues; i++) 3285 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3286 } 3287 3288 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3289 { 3290 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3291 } 3292 3293 /** 3294 * netif_start_queue - allow transmit 3295 * @dev: network device 3296 * 3297 * Allow upper layers to call the device hard_start_xmit routine. 3298 */ 3299 static inline void netif_start_queue(struct net_device *dev) 3300 { 3301 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3302 } 3303 3304 static inline void netif_tx_start_all_queues(struct net_device *dev) 3305 { 3306 unsigned int i; 3307 3308 for (i = 0; i < dev->num_tx_queues; i++) { 3309 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3310 netif_tx_start_queue(txq); 3311 } 3312 } 3313 3314 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3315 3316 /** 3317 * netif_wake_queue - restart transmit 3318 * @dev: network device 3319 * 3320 * Allow upper layers to call the device hard_start_xmit routine. 3321 * Used for flow control when transmit resources are available. 3322 */ 3323 static inline void netif_wake_queue(struct net_device *dev) 3324 { 3325 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3326 } 3327 3328 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3329 { 3330 unsigned int i; 3331 3332 for (i = 0; i < dev->num_tx_queues; i++) { 3333 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3334 netif_tx_wake_queue(txq); 3335 } 3336 } 3337 3338 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3339 { 3340 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3341 } 3342 3343 /** 3344 * netif_stop_queue - stop transmitted packets 3345 * @dev: network device 3346 * 3347 * Stop upper layers calling the device hard_start_xmit routine. 3348 * Used for flow control when transmit resources are unavailable. 3349 */ 3350 static inline void netif_stop_queue(struct net_device *dev) 3351 { 3352 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3353 } 3354 3355 void netif_tx_stop_all_queues(struct net_device *dev); 3356 3357 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3358 { 3359 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3360 } 3361 3362 /** 3363 * netif_queue_stopped - test if transmit queue is flowblocked 3364 * @dev: network device 3365 * 3366 * Test if transmit queue on device is currently unable to send. 3367 */ 3368 static inline bool netif_queue_stopped(const struct net_device *dev) 3369 { 3370 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3371 } 3372 3373 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3374 { 3375 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3376 } 3377 3378 static inline bool 3379 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3380 { 3381 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3382 } 3383 3384 static inline bool 3385 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3386 { 3387 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3388 } 3389 3390 /** 3391 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3392 * @dev_queue: pointer to transmit queue 3393 * 3394 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3395 * to give appropriate hint to the CPU. 3396 */ 3397 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3398 { 3399 #ifdef CONFIG_BQL 3400 prefetchw(&dev_queue->dql.num_queued); 3401 #endif 3402 } 3403 3404 /** 3405 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3406 * @dev_queue: pointer to transmit queue 3407 * 3408 * BQL enabled drivers might use this helper in their TX completion path, 3409 * to give appropriate hint to the CPU. 3410 */ 3411 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3412 { 3413 #ifdef CONFIG_BQL 3414 prefetchw(&dev_queue->dql.limit); 3415 #endif 3416 } 3417 3418 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3419 unsigned int bytes) 3420 { 3421 #ifdef CONFIG_BQL 3422 dql_queued(&dev_queue->dql, bytes); 3423 3424 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3425 return; 3426 3427 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3428 3429 /* 3430 * The XOFF flag must be set before checking the dql_avail below, 3431 * because in netdev_tx_completed_queue we update the dql_completed 3432 * before checking the XOFF flag. 3433 */ 3434 smp_mb(); 3435 3436 /* check again in case another CPU has just made room avail */ 3437 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3438 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3439 #endif 3440 } 3441 3442 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3443 * that they should not test BQL status themselves. 3444 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3445 * skb of a batch. 3446 * Returns true if the doorbell must be used to kick the NIC. 3447 */ 3448 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3449 unsigned int bytes, 3450 bool xmit_more) 3451 { 3452 if (xmit_more) { 3453 #ifdef CONFIG_BQL 3454 dql_queued(&dev_queue->dql, bytes); 3455 #endif 3456 return netif_tx_queue_stopped(dev_queue); 3457 } 3458 netdev_tx_sent_queue(dev_queue, bytes); 3459 return true; 3460 } 3461 3462 /** 3463 * netdev_sent_queue - report the number of bytes queued to hardware 3464 * @dev: network device 3465 * @bytes: number of bytes queued to the hardware device queue 3466 * 3467 * Report the number of bytes queued for sending/completion to the network 3468 * device hardware queue. @bytes should be a good approximation and should 3469 * exactly match netdev_completed_queue() @bytes 3470 */ 3471 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3472 { 3473 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3474 } 3475 3476 static inline bool __netdev_sent_queue(struct net_device *dev, 3477 unsigned int bytes, 3478 bool xmit_more) 3479 { 3480 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3481 xmit_more); 3482 } 3483 3484 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3485 unsigned int pkts, unsigned int bytes) 3486 { 3487 #ifdef CONFIG_BQL 3488 if (unlikely(!bytes)) 3489 return; 3490 3491 dql_completed(&dev_queue->dql, bytes); 3492 3493 /* 3494 * Without the memory barrier there is a small possiblity that 3495 * netdev_tx_sent_queue will miss the update and cause the queue to 3496 * be stopped forever 3497 */ 3498 smp_mb(); 3499 3500 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3501 return; 3502 3503 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3504 netif_schedule_queue(dev_queue); 3505 #endif 3506 } 3507 3508 /** 3509 * netdev_completed_queue - report bytes and packets completed by device 3510 * @dev: network device 3511 * @pkts: actual number of packets sent over the medium 3512 * @bytes: actual number of bytes sent over the medium 3513 * 3514 * Report the number of bytes and packets transmitted by the network device 3515 * hardware queue over the physical medium, @bytes must exactly match the 3516 * @bytes amount passed to netdev_sent_queue() 3517 */ 3518 static inline void netdev_completed_queue(struct net_device *dev, 3519 unsigned int pkts, unsigned int bytes) 3520 { 3521 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3522 } 3523 3524 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3525 { 3526 #ifdef CONFIG_BQL 3527 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3528 dql_reset(&q->dql); 3529 #endif 3530 } 3531 3532 /** 3533 * netdev_reset_queue - reset the packets and bytes count of a network device 3534 * @dev_queue: network device 3535 * 3536 * Reset the bytes and packet count of a network device and clear the 3537 * software flow control OFF bit for this network device 3538 */ 3539 static inline void netdev_reset_queue(struct net_device *dev_queue) 3540 { 3541 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3542 } 3543 3544 /** 3545 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3546 * @dev: network device 3547 * @queue_index: given tx queue index 3548 * 3549 * Returns 0 if given tx queue index >= number of device tx queues, 3550 * otherwise returns the originally passed tx queue index. 3551 */ 3552 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3553 { 3554 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3555 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3556 dev->name, queue_index, 3557 dev->real_num_tx_queues); 3558 return 0; 3559 } 3560 3561 return queue_index; 3562 } 3563 3564 /** 3565 * netif_running - test if up 3566 * @dev: network device 3567 * 3568 * Test if the device has been brought up. 3569 */ 3570 static inline bool netif_running(const struct net_device *dev) 3571 { 3572 return test_bit(__LINK_STATE_START, &dev->state); 3573 } 3574 3575 /* 3576 * Routines to manage the subqueues on a device. We only need start, 3577 * stop, and a check if it's stopped. All other device management is 3578 * done at the overall netdevice level. 3579 * Also test the device if we're multiqueue. 3580 */ 3581 3582 /** 3583 * netif_start_subqueue - allow sending packets on subqueue 3584 * @dev: network device 3585 * @queue_index: sub queue index 3586 * 3587 * Start individual transmit queue of a device with multiple transmit queues. 3588 */ 3589 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3590 { 3591 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3592 3593 netif_tx_start_queue(txq); 3594 } 3595 3596 /** 3597 * netif_stop_subqueue - stop sending packets on subqueue 3598 * @dev: network device 3599 * @queue_index: sub queue index 3600 * 3601 * Stop individual transmit queue of a device with multiple transmit queues. 3602 */ 3603 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3604 { 3605 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3606 netif_tx_stop_queue(txq); 3607 } 3608 3609 /** 3610 * __netif_subqueue_stopped - test status of subqueue 3611 * @dev: network device 3612 * @queue_index: sub queue index 3613 * 3614 * Check individual transmit queue of a device with multiple transmit queues. 3615 */ 3616 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3617 u16 queue_index) 3618 { 3619 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3620 3621 return netif_tx_queue_stopped(txq); 3622 } 3623 3624 /** 3625 * netif_subqueue_stopped - test status of subqueue 3626 * @dev: network device 3627 * @skb: sub queue buffer pointer 3628 * 3629 * Check individual transmit queue of a device with multiple transmit queues. 3630 */ 3631 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3632 struct sk_buff *skb) 3633 { 3634 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3635 } 3636 3637 /** 3638 * netif_wake_subqueue - allow sending packets on subqueue 3639 * @dev: network device 3640 * @queue_index: sub queue index 3641 * 3642 * Resume individual transmit queue of a device with multiple transmit queues. 3643 */ 3644 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3645 { 3646 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3647 3648 netif_tx_wake_queue(txq); 3649 } 3650 3651 #ifdef CONFIG_XPS 3652 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3653 u16 index); 3654 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3655 u16 index, bool is_rxqs_map); 3656 3657 /** 3658 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3659 * @j: CPU/Rx queue index 3660 * @mask: bitmask of all cpus/rx queues 3661 * @nr_bits: number of bits in the bitmask 3662 * 3663 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3664 */ 3665 static inline bool netif_attr_test_mask(unsigned long j, 3666 const unsigned long *mask, 3667 unsigned int nr_bits) 3668 { 3669 cpu_max_bits_warn(j, nr_bits); 3670 return test_bit(j, mask); 3671 } 3672 3673 /** 3674 * netif_attr_test_online - Test for online CPU/Rx queue 3675 * @j: CPU/Rx queue index 3676 * @online_mask: bitmask for CPUs/Rx queues that are online 3677 * @nr_bits: number of bits in the bitmask 3678 * 3679 * Returns true if a CPU/Rx queue is online. 3680 */ 3681 static inline bool netif_attr_test_online(unsigned long j, 3682 const unsigned long *online_mask, 3683 unsigned int nr_bits) 3684 { 3685 cpu_max_bits_warn(j, nr_bits); 3686 3687 if (online_mask) 3688 return test_bit(j, online_mask); 3689 3690 return (j < nr_bits); 3691 } 3692 3693 /** 3694 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3695 * @n: CPU/Rx queue index 3696 * @srcp: the cpumask/Rx queue mask pointer 3697 * @nr_bits: number of bits in the bitmask 3698 * 3699 * Returns >= nr_bits if no further CPUs/Rx queues set. 3700 */ 3701 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3702 unsigned int nr_bits) 3703 { 3704 /* -1 is a legal arg here. */ 3705 if (n != -1) 3706 cpu_max_bits_warn(n, nr_bits); 3707 3708 if (srcp) 3709 return find_next_bit(srcp, nr_bits, n + 1); 3710 3711 return n + 1; 3712 } 3713 3714 /** 3715 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3716 * @n: CPU/Rx queue index 3717 * @src1p: the first CPUs/Rx queues mask pointer 3718 * @src2p: the second CPUs/Rx queues mask pointer 3719 * @nr_bits: number of bits in the bitmask 3720 * 3721 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3722 */ 3723 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3724 const unsigned long *src2p, 3725 unsigned int nr_bits) 3726 { 3727 /* -1 is a legal arg here. */ 3728 if (n != -1) 3729 cpu_max_bits_warn(n, nr_bits); 3730 3731 if (src1p && src2p) 3732 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3733 else if (src1p) 3734 return find_next_bit(src1p, nr_bits, n + 1); 3735 else if (src2p) 3736 return find_next_bit(src2p, nr_bits, n + 1); 3737 3738 return n + 1; 3739 } 3740 #else 3741 static inline int netif_set_xps_queue(struct net_device *dev, 3742 const struct cpumask *mask, 3743 u16 index) 3744 { 3745 return 0; 3746 } 3747 3748 static inline int __netif_set_xps_queue(struct net_device *dev, 3749 const unsigned long *mask, 3750 u16 index, bool is_rxqs_map) 3751 { 3752 return 0; 3753 } 3754 #endif 3755 3756 /** 3757 * netif_is_multiqueue - test if device has multiple transmit queues 3758 * @dev: network device 3759 * 3760 * Check if device has multiple transmit queues 3761 */ 3762 static inline bool netif_is_multiqueue(const struct net_device *dev) 3763 { 3764 return dev->num_tx_queues > 1; 3765 } 3766 3767 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3768 3769 #ifdef CONFIG_SYSFS 3770 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3771 #else 3772 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3773 unsigned int rxqs) 3774 { 3775 dev->real_num_rx_queues = rxqs; 3776 return 0; 3777 } 3778 #endif 3779 3780 static inline struct netdev_rx_queue * 3781 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3782 { 3783 return dev->_rx + rxq; 3784 } 3785 3786 #ifdef CONFIG_SYSFS 3787 static inline unsigned int get_netdev_rx_queue_index( 3788 struct netdev_rx_queue *queue) 3789 { 3790 struct net_device *dev = queue->dev; 3791 int index = queue - dev->_rx; 3792 3793 BUG_ON(index >= dev->num_rx_queues); 3794 return index; 3795 } 3796 #endif 3797 3798 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3799 int netif_get_num_default_rss_queues(void); 3800 3801 enum skb_free_reason { 3802 SKB_REASON_CONSUMED, 3803 SKB_REASON_DROPPED, 3804 }; 3805 3806 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3807 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3808 3809 /* 3810 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3811 * interrupt context or with hardware interrupts being disabled. 3812 * (in_irq() || irqs_disabled()) 3813 * 3814 * We provide four helpers that can be used in following contexts : 3815 * 3816 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3817 * replacing kfree_skb(skb) 3818 * 3819 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3820 * Typically used in place of consume_skb(skb) in TX completion path 3821 * 3822 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3823 * replacing kfree_skb(skb) 3824 * 3825 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3826 * and consumed a packet. Used in place of consume_skb(skb) 3827 */ 3828 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3829 { 3830 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3831 } 3832 3833 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3834 { 3835 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3836 } 3837 3838 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3839 { 3840 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3841 } 3842 3843 static inline void dev_consume_skb_any(struct sk_buff *skb) 3844 { 3845 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3846 } 3847 3848 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3849 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3850 int netif_rx(struct sk_buff *skb); 3851 int netif_rx_ni(struct sk_buff *skb); 3852 int netif_rx_any_context(struct sk_buff *skb); 3853 int netif_receive_skb(struct sk_buff *skb); 3854 int netif_receive_skb_core(struct sk_buff *skb); 3855 void netif_receive_skb_list(struct list_head *head); 3856 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3857 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3858 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3859 gro_result_t napi_gro_frags(struct napi_struct *napi); 3860 struct packet_offload *gro_find_receive_by_type(__be16 type); 3861 struct packet_offload *gro_find_complete_by_type(__be16 type); 3862 3863 static inline void napi_free_frags(struct napi_struct *napi) 3864 { 3865 kfree_skb(napi->skb); 3866 napi->skb = NULL; 3867 } 3868 3869 bool netdev_is_rx_handler_busy(struct net_device *dev); 3870 int netdev_rx_handler_register(struct net_device *dev, 3871 rx_handler_func_t *rx_handler, 3872 void *rx_handler_data); 3873 void netdev_rx_handler_unregister(struct net_device *dev); 3874 3875 bool dev_valid_name(const char *name); 3876 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3877 bool *need_copyout); 3878 int dev_ifconf(struct net *net, struct ifconf *, int); 3879 int dev_ethtool(struct net *net, struct ifreq *); 3880 unsigned int dev_get_flags(const struct net_device *); 3881 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3882 struct netlink_ext_ack *extack); 3883 int dev_change_flags(struct net_device *dev, unsigned int flags, 3884 struct netlink_ext_ack *extack); 3885 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3886 unsigned int gchanges); 3887 int dev_change_name(struct net_device *, const char *); 3888 int dev_set_alias(struct net_device *, const char *, size_t); 3889 int dev_get_alias(const struct net_device *, char *, size_t); 3890 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3891 int __dev_set_mtu(struct net_device *, int); 3892 int dev_validate_mtu(struct net_device *dev, int mtu, 3893 struct netlink_ext_ack *extack); 3894 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3895 struct netlink_ext_ack *extack); 3896 int dev_set_mtu(struct net_device *, int); 3897 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3898 void dev_set_group(struct net_device *, int); 3899 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3900 struct netlink_ext_ack *extack); 3901 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3902 struct netlink_ext_ack *extack); 3903 int dev_change_carrier(struct net_device *, bool new_carrier); 3904 int dev_get_phys_port_id(struct net_device *dev, 3905 struct netdev_phys_item_id *ppid); 3906 int dev_get_phys_port_name(struct net_device *dev, 3907 char *name, size_t len); 3908 int dev_get_port_parent_id(struct net_device *dev, 3909 struct netdev_phys_item_id *ppid, bool recurse); 3910 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3911 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3912 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3913 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3914 u32 value); 3915 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3916 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3917 struct netdev_queue *txq, int *ret); 3918 3919 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3920 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3921 int fd, int expected_fd, u32 flags); 3922 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3923 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3924 3925 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3926 3927 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3928 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3929 bool is_skb_forwardable(const struct net_device *dev, 3930 const struct sk_buff *skb); 3931 3932 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3933 struct sk_buff *skb) 3934 { 3935 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3936 unlikely(!is_skb_forwardable(dev, skb))) { 3937 atomic_long_inc(&dev->rx_dropped); 3938 kfree_skb(skb); 3939 return NET_RX_DROP; 3940 } 3941 3942 skb_scrub_packet(skb, true); 3943 skb->priority = 0; 3944 return 0; 3945 } 3946 3947 bool dev_nit_active(struct net_device *dev); 3948 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3949 3950 extern int netdev_budget; 3951 extern unsigned int netdev_budget_usecs; 3952 3953 /* Called by rtnetlink.c:rtnl_unlock() */ 3954 void netdev_run_todo(void); 3955 3956 /** 3957 * dev_put - release reference to device 3958 * @dev: network device 3959 * 3960 * Release reference to device to allow it to be freed. 3961 */ 3962 static inline void dev_put(struct net_device *dev) 3963 { 3964 this_cpu_dec(*dev->pcpu_refcnt); 3965 } 3966 3967 /** 3968 * dev_hold - get reference to device 3969 * @dev: network device 3970 * 3971 * Hold reference to device to keep it from being freed. 3972 */ 3973 static inline void dev_hold(struct net_device *dev) 3974 { 3975 this_cpu_inc(*dev->pcpu_refcnt); 3976 } 3977 3978 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3979 * and _off may be called from IRQ context, but it is caller 3980 * who is responsible for serialization of these calls. 3981 * 3982 * The name carrier is inappropriate, these functions should really be 3983 * called netif_lowerlayer_*() because they represent the state of any 3984 * kind of lower layer not just hardware media. 3985 */ 3986 3987 void linkwatch_init_dev(struct net_device *dev); 3988 void linkwatch_fire_event(struct net_device *dev); 3989 void linkwatch_forget_dev(struct net_device *dev); 3990 3991 /** 3992 * netif_carrier_ok - test if carrier present 3993 * @dev: network device 3994 * 3995 * Check if carrier is present on device 3996 */ 3997 static inline bool netif_carrier_ok(const struct net_device *dev) 3998 { 3999 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4000 } 4001 4002 unsigned long dev_trans_start(struct net_device *dev); 4003 4004 void __netdev_watchdog_up(struct net_device *dev); 4005 4006 void netif_carrier_on(struct net_device *dev); 4007 4008 void netif_carrier_off(struct net_device *dev); 4009 4010 /** 4011 * netif_dormant_on - mark device as dormant. 4012 * @dev: network device 4013 * 4014 * Mark device as dormant (as per RFC2863). 4015 * 4016 * The dormant state indicates that the relevant interface is not 4017 * actually in a condition to pass packets (i.e., it is not 'up') but is 4018 * in a "pending" state, waiting for some external event. For "on- 4019 * demand" interfaces, this new state identifies the situation where the 4020 * interface is waiting for events to place it in the up state. 4021 */ 4022 static inline void netif_dormant_on(struct net_device *dev) 4023 { 4024 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4025 linkwatch_fire_event(dev); 4026 } 4027 4028 /** 4029 * netif_dormant_off - set device as not dormant. 4030 * @dev: network device 4031 * 4032 * Device is not in dormant state. 4033 */ 4034 static inline void netif_dormant_off(struct net_device *dev) 4035 { 4036 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4037 linkwatch_fire_event(dev); 4038 } 4039 4040 /** 4041 * netif_dormant - test if device is dormant 4042 * @dev: network device 4043 * 4044 * Check if device is dormant. 4045 */ 4046 static inline bool netif_dormant(const struct net_device *dev) 4047 { 4048 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4049 } 4050 4051 4052 /** 4053 * netif_testing_on - mark device as under test. 4054 * @dev: network device 4055 * 4056 * Mark device as under test (as per RFC2863). 4057 * 4058 * The testing state indicates that some test(s) must be performed on 4059 * the interface. After completion, of the test, the interface state 4060 * will change to up, dormant, or down, as appropriate. 4061 */ 4062 static inline void netif_testing_on(struct net_device *dev) 4063 { 4064 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4065 linkwatch_fire_event(dev); 4066 } 4067 4068 /** 4069 * netif_testing_off - set device as not under test. 4070 * @dev: network device 4071 * 4072 * Device is not in testing state. 4073 */ 4074 static inline void netif_testing_off(struct net_device *dev) 4075 { 4076 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4077 linkwatch_fire_event(dev); 4078 } 4079 4080 /** 4081 * netif_testing - test if device is under test 4082 * @dev: network device 4083 * 4084 * Check if device is under test 4085 */ 4086 static inline bool netif_testing(const struct net_device *dev) 4087 { 4088 return test_bit(__LINK_STATE_TESTING, &dev->state); 4089 } 4090 4091 4092 /** 4093 * netif_oper_up - test if device is operational 4094 * @dev: network device 4095 * 4096 * Check if carrier is operational 4097 */ 4098 static inline bool netif_oper_up(const struct net_device *dev) 4099 { 4100 return (dev->operstate == IF_OPER_UP || 4101 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4102 } 4103 4104 /** 4105 * netif_device_present - is device available or removed 4106 * @dev: network device 4107 * 4108 * Check if device has not been removed from system. 4109 */ 4110 static inline bool netif_device_present(struct net_device *dev) 4111 { 4112 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4113 } 4114 4115 void netif_device_detach(struct net_device *dev); 4116 4117 void netif_device_attach(struct net_device *dev); 4118 4119 /* 4120 * Network interface message level settings 4121 */ 4122 4123 enum { 4124 NETIF_MSG_DRV_BIT, 4125 NETIF_MSG_PROBE_BIT, 4126 NETIF_MSG_LINK_BIT, 4127 NETIF_MSG_TIMER_BIT, 4128 NETIF_MSG_IFDOWN_BIT, 4129 NETIF_MSG_IFUP_BIT, 4130 NETIF_MSG_RX_ERR_BIT, 4131 NETIF_MSG_TX_ERR_BIT, 4132 NETIF_MSG_TX_QUEUED_BIT, 4133 NETIF_MSG_INTR_BIT, 4134 NETIF_MSG_TX_DONE_BIT, 4135 NETIF_MSG_RX_STATUS_BIT, 4136 NETIF_MSG_PKTDATA_BIT, 4137 NETIF_MSG_HW_BIT, 4138 NETIF_MSG_WOL_BIT, 4139 4140 /* When you add a new bit above, update netif_msg_class_names array 4141 * in net/ethtool/common.c 4142 */ 4143 NETIF_MSG_CLASS_COUNT, 4144 }; 4145 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4146 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4147 4148 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4149 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4150 4151 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4152 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4153 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4154 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4155 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4156 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4157 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4158 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4159 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4160 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4161 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4162 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4163 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4164 #define NETIF_MSG_HW __NETIF_MSG(HW) 4165 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4166 4167 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4168 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4169 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4170 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4171 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4172 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4173 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4174 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4175 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4176 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4177 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4178 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4179 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4180 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4181 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4182 4183 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4184 { 4185 /* use default */ 4186 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4187 return default_msg_enable_bits; 4188 if (debug_value == 0) /* no output */ 4189 return 0; 4190 /* set low N bits */ 4191 return (1U << debug_value) - 1; 4192 } 4193 4194 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4195 { 4196 spin_lock(&txq->_xmit_lock); 4197 txq->xmit_lock_owner = cpu; 4198 } 4199 4200 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4201 { 4202 __acquire(&txq->_xmit_lock); 4203 return true; 4204 } 4205 4206 static inline void __netif_tx_release(struct netdev_queue *txq) 4207 { 4208 __release(&txq->_xmit_lock); 4209 } 4210 4211 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4212 { 4213 spin_lock_bh(&txq->_xmit_lock); 4214 txq->xmit_lock_owner = smp_processor_id(); 4215 } 4216 4217 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4218 { 4219 bool ok = spin_trylock(&txq->_xmit_lock); 4220 if (likely(ok)) 4221 txq->xmit_lock_owner = smp_processor_id(); 4222 return ok; 4223 } 4224 4225 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4226 { 4227 txq->xmit_lock_owner = -1; 4228 spin_unlock(&txq->_xmit_lock); 4229 } 4230 4231 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4232 { 4233 txq->xmit_lock_owner = -1; 4234 spin_unlock_bh(&txq->_xmit_lock); 4235 } 4236 4237 static inline void txq_trans_update(struct netdev_queue *txq) 4238 { 4239 if (txq->xmit_lock_owner != -1) 4240 txq->trans_start = jiffies; 4241 } 4242 4243 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4244 static inline void netif_trans_update(struct net_device *dev) 4245 { 4246 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4247 4248 if (txq->trans_start != jiffies) 4249 txq->trans_start = jiffies; 4250 } 4251 4252 /** 4253 * netif_tx_lock - grab network device transmit lock 4254 * @dev: network device 4255 * 4256 * Get network device transmit lock 4257 */ 4258 static inline void netif_tx_lock(struct net_device *dev) 4259 { 4260 unsigned int i; 4261 int cpu; 4262 4263 spin_lock(&dev->tx_global_lock); 4264 cpu = smp_processor_id(); 4265 for (i = 0; i < dev->num_tx_queues; i++) { 4266 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4267 4268 /* We are the only thread of execution doing a 4269 * freeze, but we have to grab the _xmit_lock in 4270 * order to synchronize with threads which are in 4271 * the ->hard_start_xmit() handler and already 4272 * checked the frozen bit. 4273 */ 4274 __netif_tx_lock(txq, cpu); 4275 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4276 __netif_tx_unlock(txq); 4277 } 4278 } 4279 4280 static inline void netif_tx_lock_bh(struct net_device *dev) 4281 { 4282 local_bh_disable(); 4283 netif_tx_lock(dev); 4284 } 4285 4286 static inline void netif_tx_unlock(struct net_device *dev) 4287 { 4288 unsigned int i; 4289 4290 for (i = 0; i < dev->num_tx_queues; i++) { 4291 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4292 4293 /* No need to grab the _xmit_lock here. If the 4294 * queue is not stopped for another reason, we 4295 * force a schedule. 4296 */ 4297 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4298 netif_schedule_queue(txq); 4299 } 4300 spin_unlock(&dev->tx_global_lock); 4301 } 4302 4303 static inline void netif_tx_unlock_bh(struct net_device *dev) 4304 { 4305 netif_tx_unlock(dev); 4306 local_bh_enable(); 4307 } 4308 4309 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4310 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4311 __netif_tx_lock(txq, cpu); \ 4312 } else { \ 4313 __netif_tx_acquire(txq); \ 4314 } \ 4315 } 4316 4317 #define HARD_TX_TRYLOCK(dev, txq) \ 4318 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4319 __netif_tx_trylock(txq) : \ 4320 __netif_tx_acquire(txq)) 4321 4322 #define HARD_TX_UNLOCK(dev, txq) { \ 4323 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4324 __netif_tx_unlock(txq); \ 4325 } else { \ 4326 __netif_tx_release(txq); \ 4327 } \ 4328 } 4329 4330 static inline void netif_tx_disable(struct net_device *dev) 4331 { 4332 unsigned int i; 4333 int cpu; 4334 4335 local_bh_disable(); 4336 cpu = smp_processor_id(); 4337 for (i = 0; i < dev->num_tx_queues; i++) { 4338 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4339 4340 __netif_tx_lock(txq, cpu); 4341 netif_tx_stop_queue(txq); 4342 __netif_tx_unlock(txq); 4343 } 4344 local_bh_enable(); 4345 } 4346 4347 static inline void netif_addr_lock(struct net_device *dev) 4348 { 4349 unsigned char nest_level = 0; 4350 4351 #ifdef CONFIG_LOCKDEP 4352 nest_level = dev->nested_level; 4353 #endif 4354 spin_lock_nested(&dev->addr_list_lock, nest_level); 4355 } 4356 4357 static inline void netif_addr_lock_bh(struct net_device *dev) 4358 { 4359 unsigned char nest_level = 0; 4360 4361 #ifdef CONFIG_LOCKDEP 4362 nest_level = dev->nested_level; 4363 #endif 4364 local_bh_disable(); 4365 spin_lock_nested(&dev->addr_list_lock, nest_level); 4366 } 4367 4368 static inline void netif_addr_unlock(struct net_device *dev) 4369 { 4370 spin_unlock(&dev->addr_list_lock); 4371 } 4372 4373 static inline void netif_addr_unlock_bh(struct net_device *dev) 4374 { 4375 spin_unlock_bh(&dev->addr_list_lock); 4376 } 4377 4378 /* 4379 * dev_addrs walker. Should be used only for read access. Call with 4380 * rcu_read_lock held. 4381 */ 4382 #define for_each_dev_addr(dev, ha) \ 4383 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4384 4385 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4386 4387 void ether_setup(struct net_device *dev); 4388 4389 /* Support for loadable net-drivers */ 4390 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4391 unsigned char name_assign_type, 4392 void (*setup)(struct net_device *), 4393 unsigned int txqs, unsigned int rxqs); 4394 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4395 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4396 4397 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4398 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4399 count) 4400 4401 int register_netdev(struct net_device *dev); 4402 void unregister_netdev(struct net_device *dev); 4403 4404 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4405 4406 /* General hardware address lists handling functions */ 4407 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4408 struct netdev_hw_addr_list *from_list, int addr_len); 4409 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4410 struct netdev_hw_addr_list *from_list, int addr_len); 4411 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4412 struct net_device *dev, 4413 int (*sync)(struct net_device *, const unsigned char *), 4414 int (*unsync)(struct net_device *, 4415 const unsigned char *)); 4416 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4417 struct net_device *dev, 4418 int (*sync)(struct net_device *, 4419 const unsigned char *, int), 4420 int (*unsync)(struct net_device *, 4421 const unsigned char *, int)); 4422 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4423 struct net_device *dev, 4424 int (*unsync)(struct net_device *, 4425 const unsigned char *, int)); 4426 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4427 struct net_device *dev, 4428 int (*unsync)(struct net_device *, 4429 const unsigned char *)); 4430 void __hw_addr_init(struct netdev_hw_addr_list *list); 4431 4432 /* Functions used for device addresses handling */ 4433 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4434 unsigned char addr_type); 4435 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4436 unsigned char addr_type); 4437 void dev_addr_flush(struct net_device *dev); 4438 int dev_addr_init(struct net_device *dev); 4439 4440 /* Functions used for unicast addresses handling */ 4441 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4442 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4443 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4444 int dev_uc_sync(struct net_device *to, struct net_device *from); 4445 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4446 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4447 void dev_uc_flush(struct net_device *dev); 4448 void dev_uc_init(struct net_device *dev); 4449 4450 /** 4451 * __dev_uc_sync - Synchonize device's unicast list 4452 * @dev: device to sync 4453 * @sync: function to call if address should be added 4454 * @unsync: function to call if address should be removed 4455 * 4456 * Add newly added addresses to the interface, and release 4457 * addresses that have been deleted. 4458 */ 4459 static inline int __dev_uc_sync(struct net_device *dev, 4460 int (*sync)(struct net_device *, 4461 const unsigned char *), 4462 int (*unsync)(struct net_device *, 4463 const unsigned char *)) 4464 { 4465 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4466 } 4467 4468 /** 4469 * __dev_uc_unsync - Remove synchronized addresses from device 4470 * @dev: device to sync 4471 * @unsync: function to call if address should be removed 4472 * 4473 * Remove all addresses that were added to the device by dev_uc_sync(). 4474 */ 4475 static inline void __dev_uc_unsync(struct net_device *dev, 4476 int (*unsync)(struct net_device *, 4477 const unsigned char *)) 4478 { 4479 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4480 } 4481 4482 /* Functions used for multicast addresses handling */ 4483 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4484 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4485 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4486 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4487 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4488 int dev_mc_sync(struct net_device *to, struct net_device *from); 4489 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4490 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4491 void dev_mc_flush(struct net_device *dev); 4492 void dev_mc_init(struct net_device *dev); 4493 4494 /** 4495 * __dev_mc_sync - Synchonize device's multicast list 4496 * @dev: device to sync 4497 * @sync: function to call if address should be added 4498 * @unsync: function to call if address should be removed 4499 * 4500 * Add newly added addresses to the interface, and release 4501 * addresses that have been deleted. 4502 */ 4503 static inline int __dev_mc_sync(struct net_device *dev, 4504 int (*sync)(struct net_device *, 4505 const unsigned char *), 4506 int (*unsync)(struct net_device *, 4507 const unsigned char *)) 4508 { 4509 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4510 } 4511 4512 /** 4513 * __dev_mc_unsync - Remove synchronized addresses from device 4514 * @dev: device to sync 4515 * @unsync: function to call if address should be removed 4516 * 4517 * Remove all addresses that were added to the device by dev_mc_sync(). 4518 */ 4519 static inline void __dev_mc_unsync(struct net_device *dev, 4520 int (*unsync)(struct net_device *, 4521 const unsigned char *)) 4522 { 4523 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4524 } 4525 4526 /* Functions used for secondary unicast and multicast support */ 4527 void dev_set_rx_mode(struct net_device *dev); 4528 void __dev_set_rx_mode(struct net_device *dev); 4529 int dev_set_promiscuity(struct net_device *dev, int inc); 4530 int dev_set_allmulti(struct net_device *dev, int inc); 4531 void netdev_state_change(struct net_device *dev); 4532 void __netdev_notify_peers(struct net_device *dev); 4533 void netdev_notify_peers(struct net_device *dev); 4534 void netdev_features_change(struct net_device *dev); 4535 /* Load a device via the kmod */ 4536 void dev_load(struct net *net, const char *name); 4537 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4538 struct rtnl_link_stats64 *storage); 4539 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4540 const struct net_device_stats *netdev_stats); 4541 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4542 const struct pcpu_sw_netstats __percpu *netstats); 4543 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4544 4545 extern int netdev_max_backlog; 4546 extern int netdev_tstamp_prequeue; 4547 extern int weight_p; 4548 extern int dev_weight_rx_bias; 4549 extern int dev_weight_tx_bias; 4550 extern int dev_rx_weight; 4551 extern int dev_tx_weight; 4552 extern int gro_normal_batch; 4553 4554 enum { 4555 NESTED_SYNC_IMM_BIT, 4556 NESTED_SYNC_TODO_BIT, 4557 }; 4558 4559 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4560 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4561 4562 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4563 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4564 4565 struct netdev_nested_priv { 4566 unsigned char flags; 4567 void *data; 4568 }; 4569 4570 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4571 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4572 struct list_head **iter); 4573 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4574 struct list_head **iter); 4575 4576 #ifdef CONFIG_LOCKDEP 4577 static LIST_HEAD(net_unlink_list); 4578 4579 static inline void net_unlink_todo(struct net_device *dev) 4580 { 4581 if (list_empty(&dev->unlink_list)) 4582 list_add_tail(&dev->unlink_list, &net_unlink_list); 4583 } 4584 #endif 4585 4586 /* iterate through upper list, must be called under RCU read lock */ 4587 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4588 for (iter = &(dev)->adj_list.upper, \ 4589 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4590 updev; \ 4591 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4592 4593 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4594 int (*fn)(struct net_device *upper_dev, 4595 struct netdev_nested_priv *priv), 4596 struct netdev_nested_priv *priv); 4597 4598 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4599 struct net_device *upper_dev); 4600 4601 bool netdev_has_any_upper_dev(struct net_device *dev); 4602 4603 void *netdev_lower_get_next_private(struct net_device *dev, 4604 struct list_head **iter); 4605 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4606 struct list_head **iter); 4607 4608 #define netdev_for_each_lower_private(dev, priv, iter) \ 4609 for (iter = (dev)->adj_list.lower.next, \ 4610 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4611 priv; \ 4612 priv = netdev_lower_get_next_private(dev, &(iter))) 4613 4614 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4615 for (iter = &(dev)->adj_list.lower, \ 4616 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4617 priv; \ 4618 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4619 4620 void *netdev_lower_get_next(struct net_device *dev, 4621 struct list_head **iter); 4622 4623 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4624 for (iter = (dev)->adj_list.lower.next, \ 4625 ldev = netdev_lower_get_next(dev, &(iter)); \ 4626 ldev; \ 4627 ldev = netdev_lower_get_next(dev, &(iter))) 4628 4629 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4630 struct list_head **iter); 4631 int netdev_walk_all_lower_dev(struct net_device *dev, 4632 int (*fn)(struct net_device *lower_dev, 4633 struct netdev_nested_priv *priv), 4634 struct netdev_nested_priv *priv); 4635 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4636 int (*fn)(struct net_device *lower_dev, 4637 struct netdev_nested_priv *priv), 4638 struct netdev_nested_priv *priv); 4639 4640 void *netdev_adjacent_get_private(struct list_head *adj_list); 4641 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4642 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4643 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4644 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4645 struct netlink_ext_ack *extack); 4646 int netdev_master_upper_dev_link(struct net_device *dev, 4647 struct net_device *upper_dev, 4648 void *upper_priv, void *upper_info, 4649 struct netlink_ext_ack *extack); 4650 void netdev_upper_dev_unlink(struct net_device *dev, 4651 struct net_device *upper_dev); 4652 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4653 struct net_device *new_dev, 4654 struct net_device *dev, 4655 struct netlink_ext_ack *extack); 4656 void netdev_adjacent_change_commit(struct net_device *old_dev, 4657 struct net_device *new_dev, 4658 struct net_device *dev); 4659 void netdev_adjacent_change_abort(struct net_device *old_dev, 4660 struct net_device *new_dev, 4661 struct net_device *dev); 4662 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4663 void *netdev_lower_dev_get_private(struct net_device *dev, 4664 struct net_device *lower_dev); 4665 void netdev_lower_state_changed(struct net_device *lower_dev, 4666 void *lower_state_info); 4667 4668 /* RSS keys are 40 or 52 bytes long */ 4669 #define NETDEV_RSS_KEY_LEN 52 4670 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4671 void netdev_rss_key_fill(void *buffer, size_t len); 4672 4673 int skb_checksum_help(struct sk_buff *skb); 4674 int skb_crc32c_csum_help(struct sk_buff *skb); 4675 int skb_csum_hwoffload_help(struct sk_buff *skb, 4676 const netdev_features_t features); 4677 4678 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4679 netdev_features_t features, bool tx_path); 4680 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4681 netdev_features_t features); 4682 4683 struct netdev_bonding_info { 4684 ifslave slave; 4685 ifbond master; 4686 }; 4687 4688 struct netdev_notifier_bonding_info { 4689 struct netdev_notifier_info info; /* must be first */ 4690 struct netdev_bonding_info bonding_info; 4691 }; 4692 4693 void netdev_bonding_info_change(struct net_device *dev, 4694 struct netdev_bonding_info *bonding_info); 4695 4696 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4697 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4698 #else 4699 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4700 const void *data) 4701 { 4702 } 4703 #endif 4704 4705 static inline 4706 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4707 { 4708 return __skb_gso_segment(skb, features, true); 4709 } 4710 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4711 4712 static inline bool can_checksum_protocol(netdev_features_t features, 4713 __be16 protocol) 4714 { 4715 if (protocol == htons(ETH_P_FCOE)) 4716 return !!(features & NETIF_F_FCOE_CRC); 4717 4718 /* Assume this is an IP checksum (not SCTP CRC) */ 4719 4720 if (features & NETIF_F_HW_CSUM) { 4721 /* Can checksum everything */ 4722 return true; 4723 } 4724 4725 switch (protocol) { 4726 case htons(ETH_P_IP): 4727 return !!(features & NETIF_F_IP_CSUM); 4728 case htons(ETH_P_IPV6): 4729 return !!(features & NETIF_F_IPV6_CSUM); 4730 default: 4731 return false; 4732 } 4733 } 4734 4735 #ifdef CONFIG_BUG 4736 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4737 #else 4738 static inline void netdev_rx_csum_fault(struct net_device *dev, 4739 struct sk_buff *skb) 4740 { 4741 } 4742 #endif 4743 /* rx skb timestamps */ 4744 void net_enable_timestamp(void); 4745 void net_disable_timestamp(void); 4746 4747 #ifdef CONFIG_PROC_FS 4748 int __init dev_proc_init(void); 4749 #else 4750 #define dev_proc_init() 0 4751 #endif 4752 4753 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4754 struct sk_buff *skb, struct net_device *dev, 4755 bool more) 4756 { 4757 __this_cpu_write(softnet_data.xmit.more, more); 4758 return ops->ndo_start_xmit(skb, dev); 4759 } 4760 4761 static inline bool netdev_xmit_more(void) 4762 { 4763 return __this_cpu_read(softnet_data.xmit.more); 4764 } 4765 4766 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4767 struct netdev_queue *txq, bool more) 4768 { 4769 const struct net_device_ops *ops = dev->netdev_ops; 4770 netdev_tx_t rc; 4771 4772 rc = __netdev_start_xmit(ops, skb, dev, more); 4773 if (rc == NETDEV_TX_OK) 4774 txq_trans_update(txq); 4775 4776 return rc; 4777 } 4778 4779 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4780 const void *ns); 4781 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4782 const void *ns); 4783 4784 extern const struct kobj_ns_type_operations net_ns_type_operations; 4785 4786 const char *netdev_drivername(const struct net_device *dev); 4787 4788 void linkwatch_run_queue(void); 4789 4790 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4791 netdev_features_t f2) 4792 { 4793 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4794 if (f1 & NETIF_F_HW_CSUM) 4795 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4796 else 4797 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4798 } 4799 4800 return f1 & f2; 4801 } 4802 4803 static inline netdev_features_t netdev_get_wanted_features( 4804 struct net_device *dev) 4805 { 4806 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4807 } 4808 netdev_features_t netdev_increment_features(netdev_features_t all, 4809 netdev_features_t one, netdev_features_t mask); 4810 4811 /* Allow TSO being used on stacked device : 4812 * Performing the GSO segmentation before last device 4813 * is a performance improvement. 4814 */ 4815 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4816 netdev_features_t mask) 4817 { 4818 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4819 } 4820 4821 int __netdev_update_features(struct net_device *dev); 4822 void netdev_update_features(struct net_device *dev); 4823 void netdev_change_features(struct net_device *dev); 4824 4825 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4826 struct net_device *dev); 4827 4828 netdev_features_t passthru_features_check(struct sk_buff *skb, 4829 struct net_device *dev, 4830 netdev_features_t features); 4831 netdev_features_t netif_skb_features(struct sk_buff *skb); 4832 4833 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4834 { 4835 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4836 4837 /* check flags correspondence */ 4838 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4839 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4840 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4841 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4842 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4843 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4844 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4845 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4846 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4847 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4848 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4849 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4850 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4851 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4852 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4853 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4854 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4855 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4856 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4857 4858 return (features & feature) == feature; 4859 } 4860 4861 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4862 { 4863 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4864 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4865 } 4866 4867 static inline bool netif_needs_gso(struct sk_buff *skb, 4868 netdev_features_t features) 4869 { 4870 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4871 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4872 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4873 } 4874 4875 static inline void netif_set_gso_max_size(struct net_device *dev, 4876 unsigned int size) 4877 { 4878 dev->gso_max_size = size; 4879 } 4880 4881 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4882 int pulled_hlen, u16 mac_offset, 4883 int mac_len) 4884 { 4885 skb->protocol = protocol; 4886 skb->encapsulation = 1; 4887 skb_push(skb, pulled_hlen); 4888 skb_reset_transport_header(skb); 4889 skb->mac_header = mac_offset; 4890 skb->network_header = skb->mac_header + mac_len; 4891 skb->mac_len = mac_len; 4892 } 4893 4894 static inline bool netif_is_macsec(const struct net_device *dev) 4895 { 4896 return dev->priv_flags & IFF_MACSEC; 4897 } 4898 4899 static inline bool netif_is_macvlan(const struct net_device *dev) 4900 { 4901 return dev->priv_flags & IFF_MACVLAN; 4902 } 4903 4904 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4905 { 4906 return dev->priv_flags & IFF_MACVLAN_PORT; 4907 } 4908 4909 static inline bool netif_is_bond_master(const struct net_device *dev) 4910 { 4911 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4912 } 4913 4914 static inline bool netif_is_bond_slave(const struct net_device *dev) 4915 { 4916 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4917 } 4918 4919 static inline bool netif_supports_nofcs(struct net_device *dev) 4920 { 4921 return dev->priv_flags & IFF_SUPP_NOFCS; 4922 } 4923 4924 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4925 { 4926 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4927 } 4928 4929 static inline bool netif_is_l3_master(const struct net_device *dev) 4930 { 4931 return dev->priv_flags & IFF_L3MDEV_MASTER; 4932 } 4933 4934 static inline bool netif_is_l3_slave(const struct net_device *dev) 4935 { 4936 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4937 } 4938 4939 static inline bool netif_is_bridge_master(const struct net_device *dev) 4940 { 4941 return dev->priv_flags & IFF_EBRIDGE; 4942 } 4943 4944 static inline bool netif_is_bridge_port(const struct net_device *dev) 4945 { 4946 return dev->priv_flags & IFF_BRIDGE_PORT; 4947 } 4948 4949 static inline bool netif_is_ovs_master(const struct net_device *dev) 4950 { 4951 return dev->priv_flags & IFF_OPENVSWITCH; 4952 } 4953 4954 static inline bool netif_is_ovs_port(const struct net_device *dev) 4955 { 4956 return dev->priv_flags & IFF_OVS_DATAPATH; 4957 } 4958 4959 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4960 { 4961 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4962 } 4963 4964 static inline bool netif_is_team_master(const struct net_device *dev) 4965 { 4966 return dev->priv_flags & IFF_TEAM; 4967 } 4968 4969 static inline bool netif_is_team_port(const struct net_device *dev) 4970 { 4971 return dev->priv_flags & IFF_TEAM_PORT; 4972 } 4973 4974 static inline bool netif_is_lag_master(const struct net_device *dev) 4975 { 4976 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4977 } 4978 4979 static inline bool netif_is_lag_port(const struct net_device *dev) 4980 { 4981 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4982 } 4983 4984 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4985 { 4986 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4987 } 4988 4989 static inline bool netif_is_failover(const struct net_device *dev) 4990 { 4991 return dev->priv_flags & IFF_FAILOVER; 4992 } 4993 4994 static inline bool netif_is_failover_slave(const struct net_device *dev) 4995 { 4996 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4997 } 4998 4999 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5000 static inline void netif_keep_dst(struct net_device *dev) 5001 { 5002 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5003 } 5004 5005 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5006 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5007 { 5008 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5009 return dev->priv_flags & IFF_MACSEC; 5010 } 5011 5012 extern struct pernet_operations __net_initdata loopback_net_ops; 5013 5014 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5015 5016 /* netdev_printk helpers, similar to dev_printk */ 5017 5018 static inline const char *netdev_name(const struct net_device *dev) 5019 { 5020 if (!dev->name[0] || strchr(dev->name, '%')) 5021 return "(unnamed net_device)"; 5022 return dev->name; 5023 } 5024 5025 static inline bool netdev_unregistering(const struct net_device *dev) 5026 { 5027 return dev->reg_state == NETREG_UNREGISTERING; 5028 } 5029 5030 static inline const char *netdev_reg_state(const struct net_device *dev) 5031 { 5032 switch (dev->reg_state) { 5033 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5034 case NETREG_REGISTERED: return ""; 5035 case NETREG_UNREGISTERING: return " (unregistering)"; 5036 case NETREG_UNREGISTERED: return " (unregistered)"; 5037 case NETREG_RELEASED: return " (released)"; 5038 case NETREG_DUMMY: return " (dummy)"; 5039 } 5040 5041 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5042 return " (unknown)"; 5043 } 5044 5045 __printf(3, 4) __cold 5046 void netdev_printk(const char *level, const struct net_device *dev, 5047 const char *format, ...); 5048 __printf(2, 3) __cold 5049 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5050 __printf(2, 3) __cold 5051 void netdev_alert(const struct net_device *dev, const char *format, ...); 5052 __printf(2, 3) __cold 5053 void netdev_crit(const struct net_device *dev, const char *format, ...); 5054 __printf(2, 3) __cold 5055 void netdev_err(const struct net_device *dev, const char *format, ...); 5056 __printf(2, 3) __cold 5057 void netdev_warn(const struct net_device *dev, const char *format, ...); 5058 __printf(2, 3) __cold 5059 void netdev_notice(const struct net_device *dev, const char *format, ...); 5060 __printf(2, 3) __cold 5061 void netdev_info(const struct net_device *dev, const char *format, ...); 5062 5063 #define netdev_level_once(level, dev, fmt, ...) \ 5064 do { \ 5065 static bool __print_once __read_mostly; \ 5066 \ 5067 if (!__print_once) { \ 5068 __print_once = true; \ 5069 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5070 } \ 5071 } while (0) 5072 5073 #define netdev_emerg_once(dev, fmt, ...) \ 5074 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5075 #define netdev_alert_once(dev, fmt, ...) \ 5076 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5077 #define netdev_crit_once(dev, fmt, ...) \ 5078 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5079 #define netdev_err_once(dev, fmt, ...) \ 5080 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5081 #define netdev_warn_once(dev, fmt, ...) \ 5082 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5083 #define netdev_notice_once(dev, fmt, ...) \ 5084 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5085 #define netdev_info_once(dev, fmt, ...) \ 5086 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5087 5088 #define MODULE_ALIAS_NETDEV(device) \ 5089 MODULE_ALIAS("netdev-" device) 5090 5091 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5092 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5093 #define netdev_dbg(__dev, format, args...) \ 5094 do { \ 5095 dynamic_netdev_dbg(__dev, format, ##args); \ 5096 } while (0) 5097 #elif defined(DEBUG) 5098 #define netdev_dbg(__dev, format, args...) \ 5099 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5100 #else 5101 #define netdev_dbg(__dev, format, args...) \ 5102 ({ \ 5103 if (0) \ 5104 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5105 }) 5106 #endif 5107 5108 #if defined(VERBOSE_DEBUG) 5109 #define netdev_vdbg netdev_dbg 5110 #else 5111 5112 #define netdev_vdbg(dev, format, args...) \ 5113 ({ \ 5114 if (0) \ 5115 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5116 0; \ 5117 }) 5118 #endif 5119 5120 /* 5121 * netdev_WARN() acts like dev_printk(), but with the key difference 5122 * of using a WARN/WARN_ON to get the message out, including the 5123 * file/line information and a backtrace. 5124 */ 5125 #define netdev_WARN(dev, format, args...) \ 5126 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5127 netdev_reg_state(dev), ##args) 5128 5129 #define netdev_WARN_ONCE(dev, format, args...) \ 5130 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5131 netdev_reg_state(dev), ##args) 5132 5133 /* netif printk helpers, similar to netdev_printk */ 5134 5135 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5136 do { \ 5137 if (netif_msg_##type(priv)) \ 5138 netdev_printk(level, (dev), fmt, ##args); \ 5139 } while (0) 5140 5141 #define netif_level(level, priv, type, dev, fmt, args...) \ 5142 do { \ 5143 if (netif_msg_##type(priv)) \ 5144 netdev_##level(dev, fmt, ##args); \ 5145 } while (0) 5146 5147 #define netif_emerg(priv, type, dev, fmt, args...) \ 5148 netif_level(emerg, priv, type, dev, fmt, ##args) 5149 #define netif_alert(priv, type, dev, fmt, args...) \ 5150 netif_level(alert, priv, type, dev, fmt, ##args) 5151 #define netif_crit(priv, type, dev, fmt, args...) \ 5152 netif_level(crit, priv, type, dev, fmt, ##args) 5153 #define netif_err(priv, type, dev, fmt, args...) \ 5154 netif_level(err, priv, type, dev, fmt, ##args) 5155 #define netif_warn(priv, type, dev, fmt, args...) \ 5156 netif_level(warn, priv, type, dev, fmt, ##args) 5157 #define netif_notice(priv, type, dev, fmt, args...) \ 5158 netif_level(notice, priv, type, dev, fmt, ##args) 5159 #define netif_info(priv, type, dev, fmt, args...) \ 5160 netif_level(info, priv, type, dev, fmt, ##args) 5161 5162 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5163 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5164 #define netif_dbg(priv, type, netdev, format, args...) \ 5165 do { \ 5166 if (netif_msg_##type(priv)) \ 5167 dynamic_netdev_dbg(netdev, format, ##args); \ 5168 } while (0) 5169 #elif defined(DEBUG) 5170 #define netif_dbg(priv, type, dev, format, args...) \ 5171 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5172 #else 5173 #define netif_dbg(priv, type, dev, format, args...) \ 5174 ({ \ 5175 if (0) \ 5176 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5177 0; \ 5178 }) 5179 #endif 5180 5181 /* if @cond then downgrade to debug, else print at @level */ 5182 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5183 do { \ 5184 if (cond) \ 5185 netif_dbg(priv, type, netdev, fmt, ##args); \ 5186 else \ 5187 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5188 } while (0) 5189 5190 #if defined(VERBOSE_DEBUG) 5191 #define netif_vdbg netif_dbg 5192 #else 5193 #define netif_vdbg(priv, type, dev, format, args...) \ 5194 ({ \ 5195 if (0) \ 5196 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5197 0; \ 5198 }) 5199 #endif 5200 5201 /* 5202 * The list of packet types we will receive (as opposed to discard) 5203 * and the routines to invoke. 5204 * 5205 * Why 16. Because with 16 the only overlap we get on a hash of the 5206 * low nibble of the protocol value is RARP/SNAP/X.25. 5207 * 5208 * 0800 IP 5209 * 0001 802.3 5210 * 0002 AX.25 5211 * 0004 802.2 5212 * 8035 RARP 5213 * 0005 SNAP 5214 * 0805 X.25 5215 * 0806 ARP 5216 * 8137 IPX 5217 * 0009 Localtalk 5218 * 86DD IPv6 5219 */ 5220 #define PTYPE_HASH_SIZE (16) 5221 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5222 5223 extern struct net_device *blackhole_netdev; 5224 5225 #endif /* _LINUX_NETDEVICE_H */ 5226