1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 struct macsec_context; 57 struct macsec_ops; 58 59 struct sfp_bus; 60 /* 802.11 specific */ 61 struct wireless_dev; 62 /* 802.15.4 specific */ 63 struct wpan_dev; 64 struct mpls_dev; 65 /* UDP Tunnel offloads */ 66 struct udp_tunnel_info; 67 struct bpf_prog; 68 struct xdp_buff; 69 70 void netdev_set_default_ethtool_ops(struct net_device *dev, 71 const struct ethtool_ops *ops); 72 73 /* Backlog congestion levels */ 74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 75 #define NET_RX_DROP 1 /* packet dropped */ 76 77 #define MAX_NEST_DEV 8 78 79 /* 80 * Transmit return codes: transmit return codes originate from three different 81 * namespaces: 82 * 83 * - qdisc return codes 84 * - driver transmit return codes 85 * - errno values 86 * 87 * Drivers are allowed to return any one of those in their hard_start_xmit() 88 * function. Real network devices commonly used with qdiscs should only return 89 * the driver transmit return codes though - when qdiscs are used, the actual 90 * transmission happens asynchronously, so the value is not propagated to 91 * higher layers. Virtual network devices transmit synchronously; in this case 92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 93 * others are propagated to higher layers. 94 */ 95 96 /* qdisc ->enqueue() return codes. */ 97 #define NET_XMIT_SUCCESS 0x00 98 #define NET_XMIT_DROP 0x01 /* skb dropped */ 99 #define NET_XMIT_CN 0x02 /* congestion notification */ 100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 101 102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 103 * indicates that the device will soon be dropping packets, or already drops 104 * some packets of the same priority; prompting us to send less aggressively. */ 105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 107 108 /* Driver transmit return codes */ 109 #define NETDEV_TX_MASK 0xf0 110 111 enum netdev_tx { 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 113 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 115 }; 116 typedef enum netdev_tx netdev_tx_t; 117 118 /* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122 static inline bool dev_xmit_complete(int rc) 123 { 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134 } 135 136 /* 137 * Compute the worst-case header length according to the protocols 138 * used. 139 */ 140 141 #if defined(CONFIG_HYPERV_NET) 142 # define LL_MAX_HEADER 128 143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 144 # if defined(CONFIG_MAC80211_MESH) 145 # define LL_MAX_HEADER 128 146 # else 147 # define LL_MAX_HEADER 96 148 # endif 149 #else 150 # define LL_MAX_HEADER 32 151 #endif 152 153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 155 #define MAX_HEADER LL_MAX_HEADER 156 #else 157 #define MAX_HEADER (LL_MAX_HEADER + 48) 158 #endif 159 160 /* 161 * Old network device statistics. Fields are native words 162 * (unsigned long) so they can be read and written atomically. 163 */ 164 165 struct net_device_stats { 166 unsigned long rx_packets; 167 unsigned long tx_packets; 168 unsigned long rx_bytes; 169 unsigned long tx_bytes; 170 unsigned long rx_errors; 171 unsigned long tx_errors; 172 unsigned long rx_dropped; 173 unsigned long tx_dropped; 174 unsigned long multicast; 175 unsigned long collisions; 176 unsigned long rx_length_errors; 177 unsigned long rx_over_errors; 178 unsigned long rx_crc_errors; 179 unsigned long rx_frame_errors; 180 unsigned long rx_fifo_errors; 181 unsigned long rx_missed_errors; 182 unsigned long tx_aborted_errors; 183 unsigned long tx_carrier_errors; 184 unsigned long tx_fifo_errors; 185 unsigned long tx_heartbeat_errors; 186 unsigned long tx_window_errors; 187 unsigned long rx_compressed; 188 unsigned long tx_compressed; 189 }; 190 191 192 #include <linux/cache.h> 193 #include <linux/skbuff.h> 194 195 #ifdef CONFIG_RPS 196 #include <linux/static_key.h> 197 extern struct static_key_false rps_needed; 198 extern struct static_key_false rfs_needed; 199 #endif 200 201 struct neighbour; 202 struct neigh_parms; 203 struct sk_buff; 204 205 struct netdev_hw_addr { 206 struct list_head list; 207 unsigned char addr[MAX_ADDR_LEN]; 208 unsigned char type; 209 #define NETDEV_HW_ADDR_T_LAN 1 210 #define NETDEV_HW_ADDR_T_SAN 2 211 #define NETDEV_HW_ADDR_T_SLAVE 3 212 #define NETDEV_HW_ADDR_T_UNICAST 4 213 #define NETDEV_HW_ADDR_T_MULTICAST 5 214 bool global_use; 215 int sync_cnt; 216 int refcount; 217 int synced; 218 struct rcu_head rcu_head; 219 }; 220 221 struct netdev_hw_addr_list { 222 struct list_head list; 223 int count; 224 }; 225 226 #define netdev_hw_addr_list_count(l) ((l)->count) 227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 228 #define netdev_hw_addr_list_for_each(ha, l) \ 229 list_for_each_entry(ha, &(l)->list, list) 230 231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 233 #define netdev_for_each_uc_addr(ha, dev) \ 234 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 235 236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 238 #define netdev_for_each_mc_addr(ha, dev) \ 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 240 241 struct hh_cache { 242 unsigned int hh_len; 243 seqlock_t hh_lock; 244 245 /* cached hardware header; allow for machine alignment needs. */ 246 #define HH_DATA_MOD 16 247 #define HH_DATA_OFF(__len) \ 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 249 #define HH_DATA_ALIGN(__len) \ 250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 252 }; 253 254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 255 * Alternative is: 256 * dev->hard_header_len ? (dev->hard_header_len + 257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 258 * 259 * We could use other alignment values, but we must maintain the 260 * relationship HH alignment <= LL alignment. 261 */ 262 #define LL_RESERVED_SPACE(dev) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 266 267 struct header_ops { 268 int (*create) (struct sk_buff *skb, struct net_device *dev, 269 unsigned short type, const void *daddr, 270 const void *saddr, unsigned int len); 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 273 void (*cache_update)(struct hh_cache *hh, 274 const struct net_device *dev, 275 const unsigned char *haddr); 276 bool (*validate)(const char *ll_header, unsigned int len); 277 __be16 (*parse_protocol)(const struct sk_buff *skb); 278 }; 279 280 /* These flag bits are private to the generic network queueing 281 * layer; they may not be explicitly referenced by any other 282 * code. 283 */ 284 285 enum netdev_state_t { 286 __LINK_STATE_START, 287 __LINK_STATE_PRESENT, 288 __LINK_STATE_NOCARRIER, 289 __LINK_STATE_LINKWATCH_PENDING, 290 __LINK_STATE_DORMANT, 291 }; 292 293 294 /* 295 * This structure holds boot-time configured netdevice settings. They 296 * are then used in the device probing. 297 */ 298 struct netdev_boot_setup { 299 char name[IFNAMSIZ]; 300 struct ifmap map; 301 }; 302 #define NETDEV_BOOT_SETUP_MAX 8 303 304 int __init netdev_boot_setup(char *str); 305 306 struct gro_list { 307 struct list_head list; 308 int count; 309 }; 310 311 /* 312 * size of gro hash buckets, must less than bit number of 313 * napi_struct::gro_bitmask 314 */ 315 #define GRO_HASH_BUCKETS 8 316 317 /* 318 * Structure for NAPI scheduling similar to tasklet but with weighting 319 */ 320 struct napi_struct { 321 /* The poll_list must only be managed by the entity which 322 * changes the state of the NAPI_STATE_SCHED bit. This means 323 * whoever atomically sets that bit can add this napi_struct 324 * to the per-CPU poll_list, and whoever clears that bit 325 * can remove from the list right before clearing the bit. 326 */ 327 struct list_head poll_list; 328 329 unsigned long state; 330 int weight; 331 unsigned long gro_bitmask; 332 int (*poll)(struct napi_struct *, int); 333 #ifdef CONFIG_NETPOLL 334 int poll_owner; 335 #endif 336 struct net_device *dev; 337 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 338 struct sk_buff *skb; 339 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 340 int rx_count; /* length of rx_list */ 341 struct hrtimer timer; 342 struct list_head dev_list; 343 struct hlist_node napi_hash_node; 344 unsigned int napi_id; 345 }; 346 347 enum { 348 NAPI_STATE_SCHED, /* Poll is scheduled */ 349 NAPI_STATE_MISSED, /* reschedule a napi */ 350 NAPI_STATE_DISABLE, /* Disable pending */ 351 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 352 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 353 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 354 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 355 }; 356 357 enum { 358 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 359 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 360 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 361 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 362 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 363 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 364 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 365 }; 366 367 enum gro_result { 368 GRO_MERGED, 369 GRO_MERGED_FREE, 370 GRO_HELD, 371 GRO_NORMAL, 372 GRO_DROP, 373 GRO_CONSUMED, 374 }; 375 typedef enum gro_result gro_result_t; 376 377 /* 378 * enum rx_handler_result - Possible return values for rx_handlers. 379 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 380 * further. 381 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 382 * case skb->dev was changed by rx_handler. 383 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 384 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 385 * 386 * rx_handlers are functions called from inside __netif_receive_skb(), to do 387 * special processing of the skb, prior to delivery to protocol handlers. 388 * 389 * Currently, a net_device can only have a single rx_handler registered. Trying 390 * to register a second rx_handler will return -EBUSY. 391 * 392 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 393 * To unregister a rx_handler on a net_device, use 394 * netdev_rx_handler_unregister(). 395 * 396 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 397 * do with the skb. 398 * 399 * If the rx_handler consumed the skb in some way, it should return 400 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 401 * the skb to be delivered in some other way. 402 * 403 * If the rx_handler changed skb->dev, to divert the skb to another 404 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 405 * new device will be called if it exists. 406 * 407 * If the rx_handler decides the skb should be ignored, it should return 408 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 409 * are registered on exact device (ptype->dev == skb->dev). 410 * 411 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 412 * delivered, it should return RX_HANDLER_PASS. 413 * 414 * A device without a registered rx_handler will behave as if rx_handler 415 * returned RX_HANDLER_PASS. 416 */ 417 418 enum rx_handler_result { 419 RX_HANDLER_CONSUMED, 420 RX_HANDLER_ANOTHER, 421 RX_HANDLER_EXACT, 422 RX_HANDLER_PASS, 423 }; 424 typedef enum rx_handler_result rx_handler_result_t; 425 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 426 427 void __napi_schedule(struct napi_struct *n); 428 void __napi_schedule_irqoff(struct napi_struct *n); 429 430 static inline bool napi_disable_pending(struct napi_struct *n) 431 { 432 return test_bit(NAPI_STATE_DISABLE, &n->state); 433 } 434 435 bool napi_schedule_prep(struct napi_struct *n); 436 437 /** 438 * napi_schedule - schedule NAPI poll 439 * @n: NAPI context 440 * 441 * Schedule NAPI poll routine to be called if it is not already 442 * running. 443 */ 444 static inline void napi_schedule(struct napi_struct *n) 445 { 446 if (napi_schedule_prep(n)) 447 __napi_schedule(n); 448 } 449 450 /** 451 * napi_schedule_irqoff - schedule NAPI poll 452 * @n: NAPI context 453 * 454 * Variant of napi_schedule(), assuming hard irqs are masked. 455 */ 456 static inline void napi_schedule_irqoff(struct napi_struct *n) 457 { 458 if (napi_schedule_prep(n)) 459 __napi_schedule_irqoff(n); 460 } 461 462 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 463 static inline bool napi_reschedule(struct napi_struct *napi) 464 { 465 if (napi_schedule_prep(napi)) { 466 __napi_schedule(napi); 467 return true; 468 } 469 return false; 470 } 471 472 bool napi_complete_done(struct napi_struct *n, int work_done); 473 /** 474 * napi_complete - NAPI processing complete 475 * @n: NAPI context 476 * 477 * Mark NAPI processing as complete. 478 * Consider using napi_complete_done() instead. 479 * Return false if device should avoid rearming interrupts. 480 */ 481 static inline bool napi_complete(struct napi_struct *n) 482 { 483 return napi_complete_done(n, 0); 484 } 485 486 /** 487 * napi_hash_del - remove a NAPI from global table 488 * @napi: NAPI context 489 * 490 * Warning: caller must observe RCU grace period 491 * before freeing memory containing @napi, if 492 * this function returns true. 493 * Note: core networking stack automatically calls it 494 * from netif_napi_del(). 495 * Drivers might want to call this helper to combine all 496 * the needed RCU grace periods into a single one. 497 */ 498 bool napi_hash_del(struct napi_struct *napi); 499 500 /** 501 * napi_disable - prevent NAPI from scheduling 502 * @n: NAPI context 503 * 504 * Stop NAPI from being scheduled on this context. 505 * Waits till any outstanding processing completes. 506 */ 507 void napi_disable(struct napi_struct *n); 508 509 /** 510 * napi_enable - enable NAPI scheduling 511 * @n: NAPI context 512 * 513 * Resume NAPI from being scheduled on this context. 514 * Must be paired with napi_disable. 515 */ 516 static inline void napi_enable(struct napi_struct *n) 517 { 518 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 519 smp_mb__before_atomic(); 520 clear_bit(NAPI_STATE_SCHED, &n->state); 521 clear_bit(NAPI_STATE_NPSVC, &n->state); 522 } 523 524 /** 525 * napi_synchronize - wait until NAPI is not running 526 * @n: NAPI context 527 * 528 * Wait until NAPI is done being scheduled on this context. 529 * Waits till any outstanding processing completes but 530 * does not disable future activations. 531 */ 532 static inline void napi_synchronize(const struct napi_struct *n) 533 { 534 if (IS_ENABLED(CONFIG_SMP)) 535 while (test_bit(NAPI_STATE_SCHED, &n->state)) 536 msleep(1); 537 else 538 barrier(); 539 } 540 541 /** 542 * napi_if_scheduled_mark_missed - if napi is running, set the 543 * NAPIF_STATE_MISSED 544 * @n: NAPI context 545 * 546 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 547 * NAPI is scheduled. 548 **/ 549 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 550 { 551 unsigned long val, new; 552 553 do { 554 val = READ_ONCE(n->state); 555 if (val & NAPIF_STATE_DISABLE) 556 return true; 557 558 if (!(val & NAPIF_STATE_SCHED)) 559 return false; 560 561 new = val | NAPIF_STATE_MISSED; 562 } while (cmpxchg(&n->state, val, new) != val); 563 564 return true; 565 } 566 567 enum netdev_queue_state_t { 568 __QUEUE_STATE_DRV_XOFF, 569 __QUEUE_STATE_STACK_XOFF, 570 __QUEUE_STATE_FROZEN, 571 }; 572 573 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 574 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 575 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 576 577 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 578 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 579 QUEUE_STATE_FROZEN) 580 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 581 QUEUE_STATE_FROZEN) 582 583 /* 584 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 585 * netif_tx_* functions below are used to manipulate this flag. The 586 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 587 * queue independently. The netif_xmit_*stopped functions below are called 588 * to check if the queue has been stopped by the driver or stack (either 589 * of the XOFF bits are set in the state). Drivers should not need to call 590 * netif_xmit*stopped functions, they should only be using netif_tx_*. 591 */ 592 593 struct netdev_queue { 594 /* 595 * read-mostly part 596 */ 597 struct net_device *dev; 598 struct Qdisc __rcu *qdisc; 599 struct Qdisc *qdisc_sleeping; 600 #ifdef CONFIG_SYSFS 601 struct kobject kobj; 602 #endif 603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 604 int numa_node; 605 #endif 606 unsigned long tx_maxrate; 607 /* 608 * Number of TX timeouts for this queue 609 * (/sys/class/net/DEV/Q/trans_timeout) 610 */ 611 unsigned long trans_timeout; 612 613 /* Subordinate device that the queue has been assigned to */ 614 struct net_device *sb_dev; 615 #ifdef CONFIG_XDP_SOCKETS 616 struct xdp_umem *umem; 617 #endif 618 /* 619 * write-mostly part 620 */ 621 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 622 int xmit_lock_owner; 623 /* 624 * Time (in jiffies) of last Tx 625 */ 626 unsigned long trans_start; 627 628 unsigned long state; 629 630 #ifdef CONFIG_BQL 631 struct dql dql; 632 #endif 633 } ____cacheline_aligned_in_smp; 634 635 extern int sysctl_fb_tunnels_only_for_init_net; 636 extern int sysctl_devconf_inherit_init_net; 637 638 static inline bool net_has_fallback_tunnels(const struct net *net) 639 { 640 return net == &init_net || 641 !IS_ENABLED(CONFIG_SYSCTL) || 642 !sysctl_fb_tunnels_only_for_init_net; 643 } 644 645 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 646 { 647 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 648 return q->numa_node; 649 #else 650 return NUMA_NO_NODE; 651 #endif 652 } 653 654 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 655 { 656 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 657 q->numa_node = node; 658 #endif 659 } 660 661 #ifdef CONFIG_RPS 662 /* 663 * This structure holds an RPS map which can be of variable length. The 664 * map is an array of CPUs. 665 */ 666 struct rps_map { 667 unsigned int len; 668 struct rcu_head rcu; 669 u16 cpus[]; 670 }; 671 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 672 673 /* 674 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 675 * tail pointer for that CPU's input queue at the time of last enqueue, and 676 * a hardware filter index. 677 */ 678 struct rps_dev_flow { 679 u16 cpu; 680 u16 filter; 681 unsigned int last_qtail; 682 }; 683 #define RPS_NO_FILTER 0xffff 684 685 /* 686 * The rps_dev_flow_table structure contains a table of flow mappings. 687 */ 688 struct rps_dev_flow_table { 689 unsigned int mask; 690 struct rcu_head rcu; 691 struct rps_dev_flow flows[]; 692 }; 693 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 694 ((_num) * sizeof(struct rps_dev_flow))) 695 696 /* 697 * The rps_sock_flow_table contains mappings of flows to the last CPU 698 * on which they were processed by the application (set in recvmsg). 699 * Each entry is a 32bit value. Upper part is the high-order bits 700 * of flow hash, lower part is CPU number. 701 * rps_cpu_mask is used to partition the space, depending on number of 702 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 703 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 704 * meaning we use 32-6=26 bits for the hash. 705 */ 706 struct rps_sock_flow_table { 707 u32 mask; 708 709 u32 ents[] ____cacheline_aligned_in_smp; 710 }; 711 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 712 713 #define RPS_NO_CPU 0xffff 714 715 extern u32 rps_cpu_mask; 716 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 717 718 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 719 u32 hash) 720 { 721 if (table && hash) { 722 unsigned int index = hash & table->mask; 723 u32 val = hash & ~rps_cpu_mask; 724 725 /* We only give a hint, preemption can change CPU under us */ 726 val |= raw_smp_processor_id(); 727 728 if (table->ents[index] != val) 729 table->ents[index] = val; 730 } 731 } 732 733 #ifdef CONFIG_RFS_ACCEL 734 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 735 u16 filter_id); 736 #endif 737 #endif /* CONFIG_RPS */ 738 739 /* This structure contains an instance of an RX queue. */ 740 struct netdev_rx_queue { 741 #ifdef CONFIG_RPS 742 struct rps_map __rcu *rps_map; 743 struct rps_dev_flow_table __rcu *rps_flow_table; 744 #endif 745 struct kobject kobj; 746 struct net_device *dev; 747 struct xdp_rxq_info xdp_rxq; 748 #ifdef CONFIG_XDP_SOCKETS 749 struct xdp_umem *umem; 750 #endif 751 } ____cacheline_aligned_in_smp; 752 753 /* 754 * RX queue sysfs structures and functions. 755 */ 756 struct rx_queue_attribute { 757 struct attribute attr; 758 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 759 ssize_t (*store)(struct netdev_rx_queue *queue, 760 const char *buf, size_t len); 761 }; 762 763 #ifdef CONFIG_XPS 764 /* 765 * This structure holds an XPS map which can be of variable length. The 766 * map is an array of queues. 767 */ 768 struct xps_map { 769 unsigned int len; 770 unsigned int alloc_len; 771 struct rcu_head rcu; 772 u16 queues[]; 773 }; 774 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 775 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 776 - sizeof(struct xps_map)) / sizeof(u16)) 777 778 /* 779 * This structure holds all XPS maps for device. Maps are indexed by CPU. 780 */ 781 struct xps_dev_maps { 782 struct rcu_head rcu; 783 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 784 }; 785 786 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 787 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 788 789 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 790 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 791 792 #endif /* CONFIG_XPS */ 793 794 #define TC_MAX_QUEUE 16 795 #define TC_BITMASK 15 796 /* HW offloaded queuing disciplines txq count and offset maps */ 797 struct netdev_tc_txq { 798 u16 count; 799 u16 offset; 800 }; 801 802 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 803 /* 804 * This structure is to hold information about the device 805 * configured to run FCoE protocol stack. 806 */ 807 struct netdev_fcoe_hbainfo { 808 char manufacturer[64]; 809 char serial_number[64]; 810 char hardware_version[64]; 811 char driver_version[64]; 812 char optionrom_version[64]; 813 char firmware_version[64]; 814 char model[256]; 815 char model_description[256]; 816 }; 817 #endif 818 819 #define MAX_PHYS_ITEM_ID_LEN 32 820 821 /* This structure holds a unique identifier to identify some 822 * physical item (port for example) used by a netdevice. 823 */ 824 struct netdev_phys_item_id { 825 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 826 unsigned char id_len; 827 }; 828 829 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 830 struct netdev_phys_item_id *b) 831 { 832 return a->id_len == b->id_len && 833 memcmp(a->id, b->id, a->id_len) == 0; 834 } 835 836 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 837 struct sk_buff *skb, 838 struct net_device *sb_dev); 839 840 enum tc_setup_type { 841 TC_SETUP_QDISC_MQPRIO, 842 TC_SETUP_CLSU32, 843 TC_SETUP_CLSFLOWER, 844 TC_SETUP_CLSMATCHALL, 845 TC_SETUP_CLSBPF, 846 TC_SETUP_BLOCK, 847 TC_SETUP_QDISC_CBS, 848 TC_SETUP_QDISC_RED, 849 TC_SETUP_QDISC_PRIO, 850 TC_SETUP_QDISC_MQ, 851 TC_SETUP_QDISC_ETF, 852 TC_SETUP_ROOT_QDISC, 853 TC_SETUP_QDISC_GRED, 854 TC_SETUP_QDISC_TAPRIO, 855 TC_SETUP_FT, 856 TC_SETUP_QDISC_ETS, 857 TC_SETUP_QDISC_TBF, 858 TC_SETUP_QDISC_FIFO, 859 }; 860 861 /* These structures hold the attributes of bpf state that are being passed 862 * to the netdevice through the bpf op. 863 */ 864 enum bpf_netdev_command { 865 /* Set or clear a bpf program used in the earliest stages of packet 866 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 867 * is responsible for calling bpf_prog_put on any old progs that are 868 * stored. In case of error, the callee need not release the new prog 869 * reference, but on success it takes ownership and must bpf_prog_put 870 * when it is no longer used. 871 */ 872 XDP_SETUP_PROG, 873 XDP_SETUP_PROG_HW, 874 XDP_QUERY_PROG, 875 XDP_QUERY_PROG_HW, 876 /* BPF program for offload callbacks, invoked at program load time. */ 877 BPF_OFFLOAD_MAP_ALLOC, 878 BPF_OFFLOAD_MAP_FREE, 879 XDP_SETUP_XSK_UMEM, 880 }; 881 882 struct bpf_prog_offload_ops; 883 struct netlink_ext_ack; 884 struct xdp_umem; 885 struct xdp_dev_bulk_queue; 886 887 struct netdev_bpf { 888 enum bpf_netdev_command command; 889 union { 890 /* XDP_SETUP_PROG */ 891 struct { 892 u32 flags; 893 struct bpf_prog *prog; 894 struct netlink_ext_ack *extack; 895 }; 896 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 897 struct { 898 u32 prog_id; 899 /* flags with which program was installed */ 900 u32 prog_flags; 901 }; 902 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 903 struct { 904 struct bpf_offloaded_map *offmap; 905 }; 906 /* XDP_SETUP_XSK_UMEM */ 907 struct { 908 struct xdp_umem *umem; 909 u16 queue_id; 910 } xsk; 911 }; 912 }; 913 914 /* Flags for ndo_xsk_wakeup. */ 915 #define XDP_WAKEUP_RX (1 << 0) 916 #define XDP_WAKEUP_TX (1 << 1) 917 918 #ifdef CONFIG_XFRM_OFFLOAD 919 struct xfrmdev_ops { 920 int (*xdo_dev_state_add) (struct xfrm_state *x); 921 void (*xdo_dev_state_delete) (struct xfrm_state *x); 922 void (*xdo_dev_state_free) (struct xfrm_state *x); 923 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 924 struct xfrm_state *x); 925 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 926 }; 927 #endif 928 929 struct dev_ifalias { 930 struct rcu_head rcuhead; 931 char ifalias[]; 932 }; 933 934 struct devlink; 935 struct tlsdev_ops; 936 937 struct netdev_name_node { 938 struct hlist_node hlist; 939 struct list_head list; 940 struct net_device *dev; 941 const char *name; 942 }; 943 944 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 945 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 946 947 struct netdev_net_notifier { 948 struct list_head list; 949 struct notifier_block *nb; 950 }; 951 952 /* 953 * This structure defines the management hooks for network devices. 954 * The following hooks can be defined; unless noted otherwise, they are 955 * optional and can be filled with a null pointer. 956 * 957 * int (*ndo_init)(struct net_device *dev); 958 * This function is called once when a network device is registered. 959 * The network device can use this for any late stage initialization 960 * or semantic validation. It can fail with an error code which will 961 * be propagated back to register_netdev. 962 * 963 * void (*ndo_uninit)(struct net_device *dev); 964 * This function is called when device is unregistered or when registration 965 * fails. It is not called if init fails. 966 * 967 * int (*ndo_open)(struct net_device *dev); 968 * This function is called when a network device transitions to the up 969 * state. 970 * 971 * int (*ndo_stop)(struct net_device *dev); 972 * This function is called when a network device transitions to the down 973 * state. 974 * 975 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 976 * struct net_device *dev); 977 * Called when a packet needs to be transmitted. 978 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 979 * the queue before that can happen; it's for obsolete devices and weird 980 * corner cases, but the stack really does a non-trivial amount 981 * of useless work if you return NETDEV_TX_BUSY. 982 * Required; cannot be NULL. 983 * 984 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 985 * struct net_device *dev 986 * netdev_features_t features); 987 * Called by core transmit path to determine if device is capable of 988 * performing offload operations on a given packet. This is to give 989 * the device an opportunity to implement any restrictions that cannot 990 * be otherwise expressed by feature flags. The check is called with 991 * the set of features that the stack has calculated and it returns 992 * those the driver believes to be appropriate. 993 * 994 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 995 * struct net_device *sb_dev); 996 * Called to decide which queue to use when device supports multiple 997 * transmit queues. 998 * 999 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1000 * This function is called to allow device receiver to make 1001 * changes to configuration when multicast or promiscuous is enabled. 1002 * 1003 * void (*ndo_set_rx_mode)(struct net_device *dev); 1004 * This function is called device changes address list filtering. 1005 * If driver handles unicast address filtering, it should set 1006 * IFF_UNICAST_FLT in its priv_flags. 1007 * 1008 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1009 * This function is called when the Media Access Control address 1010 * needs to be changed. If this interface is not defined, the 1011 * MAC address can not be changed. 1012 * 1013 * int (*ndo_validate_addr)(struct net_device *dev); 1014 * Test if Media Access Control address is valid for the device. 1015 * 1016 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1017 * Called when a user requests an ioctl which can't be handled by 1018 * the generic interface code. If not defined ioctls return 1019 * not supported error code. 1020 * 1021 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1022 * Used to set network devices bus interface parameters. This interface 1023 * is retained for legacy reasons; new devices should use the bus 1024 * interface (PCI) for low level management. 1025 * 1026 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1027 * Called when a user wants to change the Maximum Transfer Unit 1028 * of a device. 1029 * 1030 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1031 * Callback used when the transmitter has not made any progress 1032 * for dev->watchdog ticks. 1033 * 1034 * void (*ndo_get_stats64)(struct net_device *dev, 1035 * struct rtnl_link_stats64 *storage); 1036 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1037 * Called when a user wants to get the network device usage 1038 * statistics. Drivers must do one of the following: 1039 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1040 * rtnl_link_stats64 structure passed by the caller. 1041 * 2. Define @ndo_get_stats to update a net_device_stats structure 1042 * (which should normally be dev->stats) and return a pointer to 1043 * it. The structure may be changed asynchronously only if each 1044 * field is written atomically. 1045 * 3. Update dev->stats asynchronously and atomically, and define 1046 * neither operation. 1047 * 1048 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1049 * Return true if this device supports offload stats of this attr_id. 1050 * 1051 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1052 * void *attr_data) 1053 * Get statistics for offload operations by attr_id. Write it into the 1054 * attr_data pointer. 1055 * 1056 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1057 * If device supports VLAN filtering this function is called when a 1058 * VLAN id is registered. 1059 * 1060 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1061 * If device supports VLAN filtering this function is called when a 1062 * VLAN id is unregistered. 1063 * 1064 * void (*ndo_poll_controller)(struct net_device *dev); 1065 * 1066 * SR-IOV management functions. 1067 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1068 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1069 * u8 qos, __be16 proto); 1070 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1071 * int max_tx_rate); 1072 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1073 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1074 * int (*ndo_get_vf_config)(struct net_device *dev, 1075 * int vf, struct ifla_vf_info *ivf); 1076 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1077 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1078 * struct nlattr *port[]); 1079 * 1080 * Enable or disable the VF ability to query its RSS Redirection Table and 1081 * Hash Key. This is needed since on some devices VF share this information 1082 * with PF and querying it may introduce a theoretical security risk. 1083 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1084 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1085 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1086 * void *type_data); 1087 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1088 * This is always called from the stack with the rtnl lock held and netif 1089 * tx queues stopped. This allows the netdevice to perform queue 1090 * management safely. 1091 * 1092 * Fiber Channel over Ethernet (FCoE) offload functions. 1093 * int (*ndo_fcoe_enable)(struct net_device *dev); 1094 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1095 * so the underlying device can perform whatever needed configuration or 1096 * initialization to support acceleration of FCoE traffic. 1097 * 1098 * int (*ndo_fcoe_disable)(struct net_device *dev); 1099 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1100 * so the underlying device can perform whatever needed clean-ups to 1101 * stop supporting acceleration of FCoE traffic. 1102 * 1103 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1104 * struct scatterlist *sgl, unsigned int sgc); 1105 * Called when the FCoE Initiator wants to initialize an I/O that 1106 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1107 * perform necessary setup and returns 1 to indicate the device is set up 1108 * successfully to perform DDP on this I/O, otherwise this returns 0. 1109 * 1110 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1111 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1112 * indicated by the FC exchange id 'xid', so the underlying device can 1113 * clean up and reuse resources for later DDP requests. 1114 * 1115 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1116 * struct scatterlist *sgl, unsigned int sgc); 1117 * Called when the FCoE Target wants to initialize an I/O that 1118 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1119 * perform necessary setup and returns 1 to indicate the device is set up 1120 * successfully to perform DDP on this I/O, otherwise this returns 0. 1121 * 1122 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1123 * struct netdev_fcoe_hbainfo *hbainfo); 1124 * Called when the FCoE Protocol stack wants information on the underlying 1125 * device. This information is utilized by the FCoE protocol stack to 1126 * register attributes with Fiber Channel management service as per the 1127 * FC-GS Fabric Device Management Information(FDMI) specification. 1128 * 1129 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1130 * Called when the underlying device wants to override default World Wide 1131 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1132 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1133 * protocol stack to use. 1134 * 1135 * RFS acceleration. 1136 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1137 * u16 rxq_index, u32 flow_id); 1138 * Set hardware filter for RFS. rxq_index is the target queue index; 1139 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1140 * Return the filter ID on success, or a negative error code. 1141 * 1142 * Slave management functions (for bridge, bonding, etc). 1143 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1144 * Called to make another netdev an underling. 1145 * 1146 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1147 * Called to release previously enslaved netdev. 1148 * 1149 * Feature/offload setting functions. 1150 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1151 * netdev_features_t features); 1152 * Adjusts the requested feature flags according to device-specific 1153 * constraints, and returns the resulting flags. Must not modify 1154 * the device state. 1155 * 1156 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1157 * Called to update device configuration to new features. Passed 1158 * feature set might be less than what was returned by ndo_fix_features()). 1159 * Must return >0 or -errno if it changed dev->features itself. 1160 * 1161 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1162 * struct net_device *dev, 1163 * const unsigned char *addr, u16 vid, u16 flags, 1164 * struct netlink_ext_ack *extack); 1165 * Adds an FDB entry to dev for addr. 1166 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1167 * struct net_device *dev, 1168 * const unsigned char *addr, u16 vid) 1169 * Deletes the FDB entry from dev coresponding to addr. 1170 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1171 * struct net_device *dev, struct net_device *filter_dev, 1172 * int *idx) 1173 * Used to add FDB entries to dump requests. Implementers should add 1174 * entries to skb and update idx with the number of entries. 1175 * 1176 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1177 * u16 flags, struct netlink_ext_ack *extack) 1178 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1179 * struct net_device *dev, u32 filter_mask, 1180 * int nlflags) 1181 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1182 * u16 flags); 1183 * 1184 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1185 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1186 * which do not represent real hardware may define this to allow their 1187 * userspace components to manage their virtual carrier state. Devices 1188 * that determine carrier state from physical hardware properties (eg 1189 * network cables) or protocol-dependent mechanisms (eg 1190 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1191 * 1192 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1193 * struct netdev_phys_item_id *ppid); 1194 * Called to get ID of physical port of this device. If driver does 1195 * not implement this, it is assumed that the hw is not able to have 1196 * multiple net devices on single physical port. 1197 * 1198 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1199 * struct netdev_phys_item_id *ppid) 1200 * Called to get the parent ID of the physical port of this device. 1201 * 1202 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1203 * struct udp_tunnel_info *ti); 1204 * Called by UDP tunnel to notify a driver about the UDP port and socket 1205 * address family that a UDP tunnel is listnening to. It is called only 1206 * when a new port starts listening. The operation is protected by the 1207 * RTNL. 1208 * 1209 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1210 * struct udp_tunnel_info *ti); 1211 * Called by UDP tunnel to notify the driver about a UDP port and socket 1212 * address family that the UDP tunnel is not listening to anymore. The 1213 * operation is protected by the RTNL. 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 */ 1270 struct net_device_ops { 1271 int (*ndo_init)(struct net_device *dev); 1272 void (*ndo_uninit)(struct net_device *dev); 1273 int (*ndo_open)(struct net_device *dev); 1274 int (*ndo_stop)(struct net_device *dev); 1275 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1276 struct net_device *dev); 1277 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1278 struct net_device *dev, 1279 netdev_features_t features); 1280 u16 (*ndo_select_queue)(struct net_device *dev, 1281 struct sk_buff *skb, 1282 struct net_device *sb_dev); 1283 void (*ndo_change_rx_flags)(struct net_device *dev, 1284 int flags); 1285 void (*ndo_set_rx_mode)(struct net_device *dev); 1286 int (*ndo_set_mac_address)(struct net_device *dev, 1287 void *addr); 1288 int (*ndo_validate_addr)(struct net_device *dev); 1289 int (*ndo_do_ioctl)(struct net_device *dev, 1290 struct ifreq *ifr, int cmd); 1291 int (*ndo_set_config)(struct net_device *dev, 1292 struct ifmap *map); 1293 int (*ndo_change_mtu)(struct net_device *dev, 1294 int new_mtu); 1295 int (*ndo_neigh_setup)(struct net_device *dev, 1296 struct neigh_parms *); 1297 void (*ndo_tx_timeout) (struct net_device *dev, 1298 unsigned int txqueue); 1299 1300 void (*ndo_get_stats64)(struct net_device *dev, 1301 struct rtnl_link_stats64 *storage); 1302 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1303 int (*ndo_get_offload_stats)(int attr_id, 1304 const struct net_device *dev, 1305 void *attr_data); 1306 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1307 1308 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1309 __be16 proto, u16 vid); 1310 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1311 __be16 proto, u16 vid); 1312 #ifdef CONFIG_NET_POLL_CONTROLLER 1313 void (*ndo_poll_controller)(struct net_device *dev); 1314 int (*ndo_netpoll_setup)(struct net_device *dev, 1315 struct netpoll_info *info); 1316 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1317 #endif 1318 int (*ndo_set_vf_mac)(struct net_device *dev, 1319 int queue, u8 *mac); 1320 int (*ndo_set_vf_vlan)(struct net_device *dev, 1321 int queue, u16 vlan, 1322 u8 qos, __be16 proto); 1323 int (*ndo_set_vf_rate)(struct net_device *dev, 1324 int vf, int min_tx_rate, 1325 int max_tx_rate); 1326 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1327 int vf, bool setting); 1328 int (*ndo_set_vf_trust)(struct net_device *dev, 1329 int vf, bool setting); 1330 int (*ndo_get_vf_config)(struct net_device *dev, 1331 int vf, 1332 struct ifla_vf_info *ivf); 1333 int (*ndo_set_vf_link_state)(struct net_device *dev, 1334 int vf, int link_state); 1335 int (*ndo_get_vf_stats)(struct net_device *dev, 1336 int vf, 1337 struct ifla_vf_stats 1338 *vf_stats); 1339 int (*ndo_set_vf_port)(struct net_device *dev, 1340 int vf, 1341 struct nlattr *port[]); 1342 int (*ndo_get_vf_port)(struct net_device *dev, 1343 int vf, struct sk_buff *skb); 1344 int (*ndo_get_vf_guid)(struct net_device *dev, 1345 int vf, 1346 struct ifla_vf_guid *node_guid, 1347 struct ifla_vf_guid *port_guid); 1348 int (*ndo_set_vf_guid)(struct net_device *dev, 1349 int vf, u64 guid, 1350 int guid_type); 1351 int (*ndo_set_vf_rss_query_en)( 1352 struct net_device *dev, 1353 int vf, bool setting); 1354 int (*ndo_setup_tc)(struct net_device *dev, 1355 enum tc_setup_type type, 1356 void *type_data); 1357 #if IS_ENABLED(CONFIG_FCOE) 1358 int (*ndo_fcoe_enable)(struct net_device *dev); 1359 int (*ndo_fcoe_disable)(struct net_device *dev); 1360 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1361 u16 xid, 1362 struct scatterlist *sgl, 1363 unsigned int sgc); 1364 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1365 u16 xid); 1366 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1367 u16 xid, 1368 struct scatterlist *sgl, 1369 unsigned int sgc); 1370 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1371 struct netdev_fcoe_hbainfo *hbainfo); 1372 #endif 1373 1374 #if IS_ENABLED(CONFIG_LIBFCOE) 1375 #define NETDEV_FCOE_WWNN 0 1376 #define NETDEV_FCOE_WWPN 1 1377 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1378 u64 *wwn, int type); 1379 #endif 1380 1381 #ifdef CONFIG_RFS_ACCEL 1382 int (*ndo_rx_flow_steer)(struct net_device *dev, 1383 const struct sk_buff *skb, 1384 u16 rxq_index, 1385 u32 flow_id); 1386 #endif 1387 int (*ndo_add_slave)(struct net_device *dev, 1388 struct net_device *slave_dev, 1389 struct netlink_ext_ack *extack); 1390 int (*ndo_del_slave)(struct net_device *dev, 1391 struct net_device *slave_dev); 1392 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1393 netdev_features_t features); 1394 int (*ndo_set_features)(struct net_device *dev, 1395 netdev_features_t features); 1396 int (*ndo_neigh_construct)(struct net_device *dev, 1397 struct neighbour *n); 1398 void (*ndo_neigh_destroy)(struct net_device *dev, 1399 struct neighbour *n); 1400 1401 int (*ndo_fdb_add)(struct ndmsg *ndm, 1402 struct nlattr *tb[], 1403 struct net_device *dev, 1404 const unsigned char *addr, 1405 u16 vid, 1406 u16 flags, 1407 struct netlink_ext_ack *extack); 1408 int (*ndo_fdb_del)(struct ndmsg *ndm, 1409 struct nlattr *tb[], 1410 struct net_device *dev, 1411 const unsigned char *addr, 1412 u16 vid); 1413 int (*ndo_fdb_dump)(struct sk_buff *skb, 1414 struct netlink_callback *cb, 1415 struct net_device *dev, 1416 struct net_device *filter_dev, 1417 int *idx); 1418 int (*ndo_fdb_get)(struct sk_buff *skb, 1419 struct nlattr *tb[], 1420 struct net_device *dev, 1421 const unsigned char *addr, 1422 u16 vid, u32 portid, u32 seq, 1423 struct netlink_ext_ack *extack); 1424 int (*ndo_bridge_setlink)(struct net_device *dev, 1425 struct nlmsghdr *nlh, 1426 u16 flags, 1427 struct netlink_ext_ack *extack); 1428 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1429 u32 pid, u32 seq, 1430 struct net_device *dev, 1431 u32 filter_mask, 1432 int nlflags); 1433 int (*ndo_bridge_dellink)(struct net_device *dev, 1434 struct nlmsghdr *nlh, 1435 u16 flags); 1436 int (*ndo_change_carrier)(struct net_device *dev, 1437 bool new_carrier); 1438 int (*ndo_get_phys_port_id)(struct net_device *dev, 1439 struct netdev_phys_item_id *ppid); 1440 int (*ndo_get_port_parent_id)(struct net_device *dev, 1441 struct netdev_phys_item_id *ppid); 1442 int (*ndo_get_phys_port_name)(struct net_device *dev, 1443 char *name, size_t len); 1444 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1445 struct udp_tunnel_info *ti); 1446 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1447 struct udp_tunnel_info *ti); 1448 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1449 struct net_device *dev); 1450 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1451 void *priv); 1452 1453 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1454 int queue_index, 1455 u32 maxrate); 1456 int (*ndo_get_iflink)(const struct net_device *dev); 1457 int (*ndo_change_proto_down)(struct net_device *dev, 1458 bool proto_down); 1459 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1460 struct sk_buff *skb); 1461 void (*ndo_set_rx_headroom)(struct net_device *dev, 1462 int needed_headroom); 1463 int (*ndo_bpf)(struct net_device *dev, 1464 struct netdev_bpf *bpf); 1465 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1466 struct xdp_frame **xdp, 1467 u32 flags); 1468 int (*ndo_xsk_wakeup)(struct net_device *dev, 1469 u32 queue_id, u32 flags); 1470 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1471 }; 1472 1473 /** 1474 * enum net_device_priv_flags - &struct net_device priv_flags 1475 * 1476 * These are the &struct net_device, they are only set internally 1477 * by drivers and used in the kernel. These flags are invisible to 1478 * userspace; this means that the order of these flags can change 1479 * during any kernel release. 1480 * 1481 * You should have a pretty good reason to be extending these flags. 1482 * 1483 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1484 * @IFF_EBRIDGE: Ethernet bridging device 1485 * @IFF_BONDING: bonding master or slave 1486 * @IFF_ISATAP: ISATAP interface (RFC4214) 1487 * @IFF_WAN_HDLC: WAN HDLC device 1488 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1489 * release skb->dst 1490 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1491 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1492 * @IFF_MACVLAN_PORT: device used as macvlan port 1493 * @IFF_BRIDGE_PORT: device used as bridge port 1494 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1495 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1496 * @IFF_UNICAST_FLT: Supports unicast filtering 1497 * @IFF_TEAM_PORT: device used as team port 1498 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1499 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1500 * change when it's running 1501 * @IFF_MACVLAN: Macvlan device 1502 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1503 * underlying stacked devices 1504 * @IFF_L3MDEV_MASTER: device is an L3 master device 1505 * @IFF_NO_QUEUE: device can run without qdisc attached 1506 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1507 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1508 * @IFF_TEAM: device is a team device 1509 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1510 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1511 * entity (i.e. the master device for bridged veth) 1512 * @IFF_MACSEC: device is a MACsec device 1513 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1514 * @IFF_FAILOVER: device is a failover master device 1515 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1516 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1517 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1518 */ 1519 enum netdev_priv_flags { 1520 IFF_802_1Q_VLAN = 1<<0, 1521 IFF_EBRIDGE = 1<<1, 1522 IFF_BONDING = 1<<2, 1523 IFF_ISATAP = 1<<3, 1524 IFF_WAN_HDLC = 1<<4, 1525 IFF_XMIT_DST_RELEASE = 1<<5, 1526 IFF_DONT_BRIDGE = 1<<6, 1527 IFF_DISABLE_NETPOLL = 1<<7, 1528 IFF_MACVLAN_PORT = 1<<8, 1529 IFF_BRIDGE_PORT = 1<<9, 1530 IFF_OVS_DATAPATH = 1<<10, 1531 IFF_TX_SKB_SHARING = 1<<11, 1532 IFF_UNICAST_FLT = 1<<12, 1533 IFF_TEAM_PORT = 1<<13, 1534 IFF_SUPP_NOFCS = 1<<14, 1535 IFF_LIVE_ADDR_CHANGE = 1<<15, 1536 IFF_MACVLAN = 1<<16, 1537 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1538 IFF_L3MDEV_MASTER = 1<<18, 1539 IFF_NO_QUEUE = 1<<19, 1540 IFF_OPENVSWITCH = 1<<20, 1541 IFF_L3MDEV_SLAVE = 1<<21, 1542 IFF_TEAM = 1<<22, 1543 IFF_RXFH_CONFIGURED = 1<<23, 1544 IFF_PHONY_HEADROOM = 1<<24, 1545 IFF_MACSEC = 1<<25, 1546 IFF_NO_RX_HANDLER = 1<<26, 1547 IFF_FAILOVER = 1<<27, 1548 IFF_FAILOVER_SLAVE = 1<<28, 1549 IFF_L3MDEV_RX_HANDLER = 1<<29, 1550 IFF_LIVE_RENAME_OK = 1<<30, 1551 }; 1552 1553 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1554 #define IFF_EBRIDGE IFF_EBRIDGE 1555 #define IFF_BONDING IFF_BONDING 1556 #define IFF_ISATAP IFF_ISATAP 1557 #define IFF_WAN_HDLC IFF_WAN_HDLC 1558 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1559 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1560 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1561 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1562 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1563 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1564 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1565 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1566 #define IFF_TEAM_PORT IFF_TEAM_PORT 1567 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1568 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1569 #define IFF_MACVLAN IFF_MACVLAN 1570 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1571 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1572 #define IFF_NO_QUEUE IFF_NO_QUEUE 1573 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1574 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1575 #define IFF_TEAM IFF_TEAM 1576 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1577 #define IFF_MACSEC IFF_MACSEC 1578 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1579 #define IFF_FAILOVER IFF_FAILOVER 1580 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1581 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1582 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1583 1584 /** 1585 * struct net_device - The DEVICE structure. 1586 * 1587 * Actually, this whole structure is a big mistake. It mixes I/O 1588 * data with strictly "high-level" data, and it has to know about 1589 * almost every data structure used in the INET module. 1590 * 1591 * @name: This is the first field of the "visible" part of this structure 1592 * (i.e. as seen by users in the "Space.c" file). It is the name 1593 * of the interface. 1594 * 1595 * @name_node: Name hashlist node 1596 * @ifalias: SNMP alias 1597 * @mem_end: Shared memory end 1598 * @mem_start: Shared memory start 1599 * @base_addr: Device I/O address 1600 * @irq: Device IRQ number 1601 * 1602 * @state: Generic network queuing layer state, see netdev_state_t 1603 * @dev_list: The global list of network devices 1604 * @napi_list: List entry used for polling NAPI devices 1605 * @unreg_list: List entry when we are unregistering the 1606 * device; see the function unregister_netdev 1607 * @close_list: List entry used when we are closing the device 1608 * @ptype_all: Device-specific packet handlers for all protocols 1609 * @ptype_specific: Device-specific, protocol-specific packet handlers 1610 * 1611 * @adj_list: Directly linked devices, like slaves for bonding 1612 * @features: Currently active device features 1613 * @hw_features: User-changeable features 1614 * 1615 * @wanted_features: User-requested features 1616 * @vlan_features: Mask of features inheritable by VLAN devices 1617 * 1618 * @hw_enc_features: Mask of features inherited by encapsulating devices 1619 * This field indicates what encapsulation 1620 * offloads the hardware is capable of doing, 1621 * and drivers will need to set them appropriately. 1622 * 1623 * @mpls_features: Mask of features inheritable by MPLS 1624 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1625 * 1626 * @ifindex: interface index 1627 * @group: The group the device belongs to 1628 * 1629 * @stats: Statistics struct, which was left as a legacy, use 1630 * rtnl_link_stats64 instead 1631 * 1632 * @rx_dropped: Dropped packets by core network, 1633 * do not use this in drivers 1634 * @tx_dropped: Dropped packets by core network, 1635 * do not use this in drivers 1636 * @rx_nohandler: nohandler dropped packets by core network on 1637 * inactive devices, do not use this in drivers 1638 * @carrier_up_count: Number of times the carrier has been up 1639 * @carrier_down_count: Number of times the carrier has been down 1640 * 1641 * @wireless_handlers: List of functions to handle Wireless Extensions, 1642 * instead of ioctl, 1643 * see <net/iw_handler.h> for details. 1644 * @wireless_data: Instance data managed by the core of wireless extensions 1645 * 1646 * @netdev_ops: Includes several pointers to callbacks, 1647 * if one wants to override the ndo_*() functions 1648 * @ethtool_ops: Management operations 1649 * @l3mdev_ops: Layer 3 master device operations 1650 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1651 * discovery handling. Necessary for e.g. 6LoWPAN. 1652 * @xfrmdev_ops: Transformation offload operations 1653 * @tlsdev_ops: Transport Layer Security offload operations 1654 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1655 * of Layer 2 headers. 1656 * 1657 * @flags: Interface flags (a la BSD) 1658 * @priv_flags: Like 'flags' but invisible to userspace, 1659 * see if.h for the definitions 1660 * @gflags: Global flags ( kept as legacy ) 1661 * @padded: How much padding added by alloc_netdev() 1662 * @operstate: RFC2863 operstate 1663 * @link_mode: Mapping policy to operstate 1664 * @if_port: Selectable AUI, TP, ... 1665 * @dma: DMA channel 1666 * @mtu: Interface MTU value 1667 * @min_mtu: Interface Minimum MTU value 1668 * @max_mtu: Interface Maximum MTU value 1669 * @type: Interface hardware type 1670 * @hard_header_len: Maximum hardware header length. 1671 * @min_header_len: Minimum hardware header length 1672 * 1673 * @needed_headroom: Extra headroom the hardware may need, but not in all 1674 * cases can this be guaranteed 1675 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1676 * cases can this be guaranteed. Some cases also use 1677 * LL_MAX_HEADER instead to allocate the skb 1678 * 1679 * interface address info: 1680 * 1681 * @perm_addr: Permanent hw address 1682 * @addr_assign_type: Hw address assignment type 1683 * @addr_len: Hardware address length 1684 * @upper_level: Maximum depth level of upper devices. 1685 * @lower_level: Maximum depth level of lower devices. 1686 * @neigh_priv_len: Used in neigh_alloc() 1687 * @dev_id: Used to differentiate devices that share 1688 * the same link layer address 1689 * @dev_port: Used to differentiate devices that share 1690 * the same function 1691 * @addr_list_lock: XXX: need comments on this one 1692 * @name_assign_type: network interface name assignment type 1693 * @uc_promisc: Counter that indicates promiscuous mode 1694 * has been enabled due to the need to listen to 1695 * additional unicast addresses in a device that 1696 * does not implement ndo_set_rx_mode() 1697 * @uc: unicast mac addresses 1698 * @mc: multicast mac addresses 1699 * @dev_addrs: list of device hw addresses 1700 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1701 * @promiscuity: Number of times the NIC is told to work in 1702 * promiscuous mode; if it becomes 0 the NIC will 1703 * exit promiscuous mode 1704 * @allmulti: Counter, enables or disables allmulticast mode 1705 * 1706 * @vlan_info: VLAN info 1707 * @dsa_ptr: dsa specific data 1708 * @tipc_ptr: TIPC specific data 1709 * @atalk_ptr: AppleTalk link 1710 * @ip_ptr: IPv4 specific data 1711 * @dn_ptr: DECnet specific data 1712 * @ip6_ptr: IPv6 specific data 1713 * @ax25_ptr: AX.25 specific data 1714 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1715 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1716 * device struct 1717 * @mpls_ptr: mpls_dev struct pointer 1718 * 1719 * @dev_addr: Hw address (before bcast, 1720 * because most packets are unicast) 1721 * 1722 * @_rx: Array of RX queues 1723 * @num_rx_queues: Number of RX queues 1724 * allocated at register_netdev() time 1725 * @real_num_rx_queues: Number of RX queues currently active in device 1726 * @xdp_prog: XDP sockets filter program pointer 1727 * @gro_flush_timeout: timeout for GRO layer in NAPI 1728 * 1729 * @rx_handler: handler for received packets 1730 * @rx_handler_data: XXX: need comments on this one 1731 * @miniq_ingress: ingress/clsact qdisc specific data for 1732 * ingress processing 1733 * @ingress_queue: XXX: need comments on this one 1734 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1735 * @broadcast: hw bcast address 1736 * 1737 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1738 * indexed by RX queue number. Assigned by driver. 1739 * This must only be set if the ndo_rx_flow_steer 1740 * operation is defined 1741 * @index_hlist: Device index hash chain 1742 * 1743 * @_tx: Array of TX queues 1744 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1745 * @real_num_tx_queues: Number of TX queues currently active in device 1746 * @qdisc: Root qdisc from userspace point of view 1747 * @tx_queue_len: Max frames per queue allowed 1748 * @tx_global_lock: XXX: need comments on this one 1749 * @xdp_bulkq: XDP device bulk queue 1750 * @xps_cpus_map: all CPUs map for XPS device 1751 * @xps_rxqs_map: all RXQs map for XPS device 1752 * 1753 * @xps_maps: XXX: need comments on this one 1754 * @miniq_egress: clsact qdisc specific data for 1755 * egress processing 1756 * @qdisc_hash: qdisc hash table 1757 * @watchdog_timeo: Represents the timeout that is used by 1758 * the watchdog (see dev_watchdog()) 1759 * @watchdog_timer: List of timers 1760 * 1761 * @pcpu_refcnt: Number of references to this device 1762 * @todo_list: Delayed register/unregister 1763 * @link_watch_list: XXX: need comments on this one 1764 * 1765 * @reg_state: Register/unregister state machine 1766 * @dismantle: Device is going to be freed 1767 * @rtnl_link_state: This enum represents the phases of creating 1768 * a new link 1769 * 1770 * @needs_free_netdev: Should unregister perform free_netdev? 1771 * @priv_destructor: Called from unregister 1772 * @npinfo: XXX: need comments on this one 1773 * @nd_net: Network namespace this network device is inside 1774 * 1775 * @ml_priv: Mid-layer private 1776 * @lstats: Loopback statistics 1777 * @tstats: Tunnel statistics 1778 * @dstats: Dummy statistics 1779 * @vstats: Virtual ethernet statistics 1780 * 1781 * @garp_port: GARP 1782 * @mrp_port: MRP 1783 * 1784 * @dev: Class/net/name entry 1785 * @sysfs_groups: Space for optional device, statistics and wireless 1786 * sysfs groups 1787 * 1788 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1789 * @rtnl_link_ops: Rtnl_link_ops 1790 * 1791 * @gso_max_size: Maximum size of generic segmentation offload 1792 * @gso_max_segs: Maximum number of segments that can be passed to the 1793 * NIC for GSO 1794 * 1795 * @dcbnl_ops: Data Center Bridging netlink ops 1796 * @num_tc: Number of traffic classes in the net device 1797 * @tc_to_txq: XXX: need comments on this one 1798 * @prio_tc_map: XXX: need comments on this one 1799 * 1800 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1801 * 1802 * @priomap: XXX: need comments on this one 1803 * @phydev: Physical device may attach itself 1804 * for hardware timestamping 1805 * @sfp_bus: attached &struct sfp_bus structure. 1806 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock 1807 * spinlock 1808 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1809 * @qdisc_xmit_lock_key: lockdep class annotating 1810 * netdev_queue->_xmit_lock spinlock 1811 * @addr_list_lock_key: lockdep class annotating 1812 * net_device->addr_list_lock spinlock 1813 * 1814 * @proto_down: protocol port state information can be sent to the 1815 * switch driver and used to set the phys state of the 1816 * switch port. 1817 * 1818 * @wol_enabled: Wake-on-LAN is enabled 1819 * 1820 * @net_notifier_list: List of per-net netdev notifier block 1821 * that follow this device when it is moved 1822 * to another network namespace. 1823 * 1824 * @macsec_ops: MACsec offloading ops 1825 * 1826 * FIXME: cleanup struct net_device such that network protocol info 1827 * moves out. 1828 */ 1829 1830 struct net_device { 1831 char name[IFNAMSIZ]; 1832 struct netdev_name_node *name_node; 1833 struct dev_ifalias __rcu *ifalias; 1834 /* 1835 * I/O specific fields 1836 * FIXME: Merge these and struct ifmap into one 1837 */ 1838 unsigned long mem_end; 1839 unsigned long mem_start; 1840 unsigned long base_addr; 1841 int irq; 1842 1843 /* 1844 * Some hardware also needs these fields (state,dev_list, 1845 * napi_list,unreg_list,close_list) but they are not 1846 * part of the usual set specified in Space.c. 1847 */ 1848 1849 unsigned long state; 1850 1851 struct list_head dev_list; 1852 struct list_head napi_list; 1853 struct list_head unreg_list; 1854 struct list_head close_list; 1855 struct list_head ptype_all; 1856 struct list_head ptype_specific; 1857 1858 struct { 1859 struct list_head upper; 1860 struct list_head lower; 1861 } adj_list; 1862 1863 netdev_features_t features; 1864 netdev_features_t hw_features; 1865 netdev_features_t wanted_features; 1866 netdev_features_t vlan_features; 1867 netdev_features_t hw_enc_features; 1868 netdev_features_t mpls_features; 1869 netdev_features_t gso_partial_features; 1870 1871 int ifindex; 1872 int group; 1873 1874 struct net_device_stats stats; 1875 1876 atomic_long_t rx_dropped; 1877 atomic_long_t tx_dropped; 1878 atomic_long_t rx_nohandler; 1879 1880 /* Stats to monitor link on/off, flapping */ 1881 atomic_t carrier_up_count; 1882 atomic_t carrier_down_count; 1883 1884 #ifdef CONFIG_WIRELESS_EXT 1885 const struct iw_handler_def *wireless_handlers; 1886 struct iw_public_data *wireless_data; 1887 #endif 1888 const struct net_device_ops *netdev_ops; 1889 const struct ethtool_ops *ethtool_ops; 1890 #ifdef CONFIG_NET_L3_MASTER_DEV 1891 const struct l3mdev_ops *l3mdev_ops; 1892 #endif 1893 #if IS_ENABLED(CONFIG_IPV6) 1894 const struct ndisc_ops *ndisc_ops; 1895 #endif 1896 1897 #ifdef CONFIG_XFRM_OFFLOAD 1898 const struct xfrmdev_ops *xfrmdev_ops; 1899 #endif 1900 1901 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1902 const struct tlsdev_ops *tlsdev_ops; 1903 #endif 1904 1905 const struct header_ops *header_ops; 1906 1907 unsigned int flags; 1908 unsigned int priv_flags; 1909 1910 unsigned short gflags; 1911 unsigned short padded; 1912 1913 unsigned char operstate; 1914 unsigned char link_mode; 1915 1916 unsigned char if_port; 1917 unsigned char dma; 1918 1919 /* Note : dev->mtu is often read without holding a lock. 1920 * Writers usually hold RTNL. 1921 * It is recommended to use READ_ONCE() to annotate the reads, 1922 * and to use WRITE_ONCE() to annotate the writes. 1923 */ 1924 unsigned int mtu; 1925 unsigned int min_mtu; 1926 unsigned int max_mtu; 1927 unsigned short type; 1928 unsigned short hard_header_len; 1929 unsigned char min_header_len; 1930 1931 unsigned short needed_headroom; 1932 unsigned short needed_tailroom; 1933 1934 /* Interface address info. */ 1935 unsigned char perm_addr[MAX_ADDR_LEN]; 1936 unsigned char addr_assign_type; 1937 unsigned char addr_len; 1938 unsigned char upper_level; 1939 unsigned char lower_level; 1940 unsigned short neigh_priv_len; 1941 unsigned short dev_id; 1942 unsigned short dev_port; 1943 spinlock_t addr_list_lock; 1944 unsigned char name_assign_type; 1945 bool uc_promisc; 1946 struct netdev_hw_addr_list uc; 1947 struct netdev_hw_addr_list mc; 1948 struct netdev_hw_addr_list dev_addrs; 1949 1950 #ifdef CONFIG_SYSFS 1951 struct kset *queues_kset; 1952 #endif 1953 unsigned int promiscuity; 1954 unsigned int allmulti; 1955 1956 1957 /* Protocol-specific pointers */ 1958 1959 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1960 struct vlan_info __rcu *vlan_info; 1961 #endif 1962 #if IS_ENABLED(CONFIG_NET_DSA) 1963 struct dsa_port *dsa_ptr; 1964 #endif 1965 #if IS_ENABLED(CONFIG_TIPC) 1966 struct tipc_bearer __rcu *tipc_ptr; 1967 #endif 1968 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1969 void *atalk_ptr; 1970 #endif 1971 struct in_device __rcu *ip_ptr; 1972 #if IS_ENABLED(CONFIG_DECNET) 1973 struct dn_dev __rcu *dn_ptr; 1974 #endif 1975 struct inet6_dev __rcu *ip6_ptr; 1976 #if IS_ENABLED(CONFIG_AX25) 1977 void *ax25_ptr; 1978 #endif 1979 struct wireless_dev *ieee80211_ptr; 1980 struct wpan_dev *ieee802154_ptr; 1981 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1982 struct mpls_dev __rcu *mpls_ptr; 1983 #endif 1984 1985 /* 1986 * Cache lines mostly used on receive path (including eth_type_trans()) 1987 */ 1988 /* Interface address info used in eth_type_trans() */ 1989 unsigned char *dev_addr; 1990 1991 struct netdev_rx_queue *_rx; 1992 unsigned int num_rx_queues; 1993 unsigned int real_num_rx_queues; 1994 1995 struct bpf_prog __rcu *xdp_prog; 1996 unsigned long gro_flush_timeout; 1997 rx_handler_func_t __rcu *rx_handler; 1998 void __rcu *rx_handler_data; 1999 2000 #ifdef CONFIG_NET_CLS_ACT 2001 struct mini_Qdisc __rcu *miniq_ingress; 2002 #endif 2003 struct netdev_queue __rcu *ingress_queue; 2004 #ifdef CONFIG_NETFILTER_INGRESS 2005 struct nf_hook_entries __rcu *nf_hooks_ingress; 2006 #endif 2007 2008 unsigned char broadcast[MAX_ADDR_LEN]; 2009 #ifdef CONFIG_RFS_ACCEL 2010 struct cpu_rmap *rx_cpu_rmap; 2011 #endif 2012 struct hlist_node index_hlist; 2013 2014 /* 2015 * Cache lines mostly used on transmit path 2016 */ 2017 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2018 unsigned int num_tx_queues; 2019 unsigned int real_num_tx_queues; 2020 struct Qdisc *qdisc; 2021 unsigned int tx_queue_len; 2022 spinlock_t tx_global_lock; 2023 2024 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2025 2026 #ifdef CONFIG_XPS 2027 struct xps_dev_maps __rcu *xps_cpus_map; 2028 struct xps_dev_maps __rcu *xps_rxqs_map; 2029 #endif 2030 #ifdef CONFIG_NET_CLS_ACT 2031 struct mini_Qdisc __rcu *miniq_egress; 2032 #endif 2033 2034 #ifdef CONFIG_NET_SCHED 2035 DECLARE_HASHTABLE (qdisc_hash, 4); 2036 #endif 2037 /* These may be needed for future network-power-down code. */ 2038 struct timer_list watchdog_timer; 2039 int watchdog_timeo; 2040 2041 struct list_head todo_list; 2042 int __percpu *pcpu_refcnt; 2043 2044 struct list_head link_watch_list; 2045 2046 enum { NETREG_UNINITIALIZED=0, 2047 NETREG_REGISTERED, /* completed register_netdevice */ 2048 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2049 NETREG_UNREGISTERED, /* completed unregister todo */ 2050 NETREG_RELEASED, /* called free_netdev */ 2051 NETREG_DUMMY, /* dummy device for NAPI poll */ 2052 } reg_state:8; 2053 2054 bool dismantle; 2055 2056 enum { 2057 RTNL_LINK_INITIALIZED, 2058 RTNL_LINK_INITIALIZING, 2059 } rtnl_link_state:16; 2060 2061 bool needs_free_netdev; 2062 void (*priv_destructor)(struct net_device *dev); 2063 2064 #ifdef CONFIG_NETPOLL 2065 struct netpoll_info __rcu *npinfo; 2066 #endif 2067 2068 possible_net_t nd_net; 2069 2070 /* mid-layer private */ 2071 union { 2072 void *ml_priv; 2073 struct pcpu_lstats __percpu *lstats; 2074 struct pcpu_sw_netstats __percpu *tstats; 2075 struct pcpu_dstats __percpu *dstats; 2076 }; 2077 2078 #if IS_ENABLED(CONFIG_GARP) 2079 struct garp_port __rcu *garp_port; 2080 #endif 2081 #if IS_ENABLED(CONFIG_MRP) 2082 struct mrp_port __rcu *mrp_port; 2083 #endif 2084 2085 struct device dev; 2086 const struct attribute_group *sysfs_groups[4]; 2087 const struct attribute_group *sysfs_rx_queue_group; 2088 2089 const struct rtnl_link_ops *rtnl_link_ops; 2090 2091 /* for setting kernel sock attribute on TCP connection setup */ 2092 #define GSO_MAX_SIZE 65536 2093 unsigned int gso_max_size; 2094 #define GSO_MAX_SEGS 65535 2095 u16 gso_max_segs; 2096 2097 #ifdef CONFIG_DCB 2098 const struct dcbnl_rtnl_ops *dcbnl_ops; 2099 #endif 2100 s16 num_tc; 2101 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2102 u8 prio_tc_map[TC_BITMASK + 1]; 2103 2104 #if IS_ENABLED(CONFIG_FCOE) 2105 unsigned int fcoe_ddp_xid; 2106 #endif 2107 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2108 struct netprio_map __rcu *priomap; 2109 #endif 2110 struct phy_device *phydev; 2111 struct sfp_bus *sfp_bus; 2112 struct lock_class_key qdisc_tx_busylock_key; 2113 struct lock_class_key qdisc_running_key; 2114 struct lock_class_key qdisc_xmit_lock_key; 2115 struct lock_class_key addr_list_lock_key; 2116 bool proto_down; 2117 unsigned wol_enabled:1; 2118 2119 struct list_head net_notifier_list; 2120 2121 #if IS_ENABLED(CONFIG_MACSEC) 2122 /* MACsec management functions */ 2123 const struct macsec_ops *macsec_ops; 2124 #endif 2125 }; 2126 #define to_net_dev(d) container_of(d, struct net_device, dev) 2127 2128 static inline bool netif_elide_gro(const struct net_device *dev) 2129 { 2130 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2131 return true; 2132 return false; 2133 } 2134 2135 #define NETDEV_ALIGN 32 2136 2137 static inline 2138 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2139 { 2140 return dev->prio_tc_map[prio & TC_BITMASK]; 2141 } 2142 2143 static inline 2144 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2145 { 2146 if (tc >= dev->num_tc) 2147 return -EINVAL; 2148 2149 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2150 return 0; 2151 } 2152 2153 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2154 void netdev_reset_tc(struct net_device *dev); 2155 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2156 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2157 2158 static inline 2159 int netdev_get_num_tc(struct net_device *dev) 2160 { 2161 return dev->num_tc; 2162 } 2163 2164 void netdev_unbind_sb_channel(struct net_device *dev, 2165 struct net_device *sb_dev); 2166 int netdev_bind_sb_channel_queue(struct net_device *dev, 2167 struct net_device *sb_dev, 2168 u8 tc, u16 count, u16 offset); 2169 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2170 static inline int netdev_get_sb_channel(struct net_device *dev) 2171 { 2172 return max_t(int, -dev->num_tc, 0); 2173 } 2174 2175 static inline 2176 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2177 unsigned int index) 2178 { 2179 return &dev->_tx[index]; 2180 } 2181 2182 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2183 const struct sk_buff *skb) 2184 { 2185 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2186 } 2187 2188 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2189 void (*f)(struct net_device *, 2190 struct netdev_queue *, 2191 void *), 2192 void *arg) 2193 { 2194 unsigned int i; 2195 2196 for (i = 0; i < dev->num_tx_queues; i++) 2197 f(dev, &dev->_tx[i], arg); 2198 } 2199 2200 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2201 struct net_device *sb_dev); 2202 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2203 struct sk_buff *skb, 2204 struct net_device *sb_dev); 2205 2206 /* returns the headroom that the master device needs to take in account 2207 * when forwarding to this dev 2208 */ 2209 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2210 { 2211 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2212 } 2213 2214 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2215 { 2216 if (dev->netdev_ops->ndo_set_rx_headroom) 2217 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2218 } 2219 2220 /* set the device rx headroom to the dev's default */ 2221 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2222 { 2223 netdev_set_rx_headroom(dev, -1); 2224 } 2225 2226 /* 2227 * Net namespace inlines 2228 */ 2229 static inline 2230 struct net *dev_net(const struct net_device *dev) 2231 { 2232 return read_pnet(&dev->nd_net); 2233 } 2234 2235 static inline 2236 void dev_net_set(struct net_device *dev, struct net *net) 2237 { 2238 write_pnet(&dev->nd_net, net); 2239 } 2240 2241 /** 2242 * netdev_priv - access network device private data 2243 * @dev: network device 2244 * 2245 * Get network device private data 2246 */ 2247 static inline void *netdev_priv(const struct net_device *dev) 2248 { 2249 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2250 } 2251 2252 /* Set the sysfs physical device reference for the network logical device 2253 * if set prior to registration will cause a symlink during initialization. 2254 */ 2255 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2256 2257 /* Set the sysfs device type for the network logical device to allow 2258 * fine-grained identification of different network device types. For 2259 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2260 */ 2261 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2262 2263 /* Default NAPI poll() weight 2264 * Device drivers are strongly advised to not use bigger value 2265 */ 2266 #define NAPI_POLL_WEIGHT 64 2267 2268 /** 2269 * netif_napi_add - initialize a NAPI context 2270 * @dev: network device 2271 * @napi: NAPI context 2272 * @poll: polling function 2273 * @weight: default weight 2274 * 2275 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2276 * *any* of the other NAPI-related functions. 2277 */ 2278 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2279 int (*poll)(struct napi_struct *, int), int weight); 2280 2281 /** 2282 * netif_tx_napi_add - initialize a NAPI context 2283 * @dev: network device 2284 * @napi: NAPI context 2285 * @poll: polling function 2286 * @weight: default weight 2287 * 2288 * This variant of netif_napi_add() should be used from drivers using NAPI 2289 * to exclusively poll a TX queue. 2290 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2291 */ 2292 static inline void netif_tx_napi_add(struct net_device *dev, 2293 struct napi_struct *napi, 2294 int (*poll)(struct napi_struct *, int), 2295 int weight) 2296 { 2297 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2298 netif_napi_add(dev, napi, poll, weight); 2299 } 2300 2301 /** 2302 * netif_napi_del - remove a NAPI context 2303 * @napi: NAPI context 2304 * 2305 * netif_napi_del() removes a NAPI context from the network device NAPI list 2306 */ 2307 void netif_napi_del(struct napi_struct *napi); 2308 2309 struct napi_gro_cb { 2310 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2311 void *frag0; 2312 2313 /* Length of frag0. */ 2314 unsigned int frag0_len; 2315 2316 /* This indicates where we are processing relative to skb->data. */ 2317 int data_offset; 2318 2319 /* This is non-zero if the packet cannot be merged with the new skb. */ 2320 u16 flush; 2321 2322 /* Save the IP ID here and check when we get to the transport layer */ 2323 u16 flush_id; 2324 2325 /* Number of segments aggregated. */ 2326 u16 count; 2327 2328 /* Start offset for remote checksum offload */ 2329 u16 gro_remcsum_start; 2330 2331 /* jiffies when first packet was created/queued */ 2332 unsigned long age; 2333 2334 /* Used in ipv6_gro_receive() and foo-over-udp */ 2335 u16 proto; 2336 2337 /* This is non-zero if the packet may be of the same flow. */ 2338 u8 same_flow:1; 2339 2340 /* Used in tunnel GRO receive */ 2341 u8 encap_mark:1; 2342 2343 /* GRO checksum is valid */ 2344 u8 csum_valid:1; 2345 2346 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2347 u8 csum_cnt:3; 2348 2349 /* Free the skb? */ 2350 u8 free:2; 2351 #define NAPI_GRO_FREE 1 2352 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2353 2354 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2355 u8 is_ipv6:1; 2356 2357 /* Used in GRE, set in fou/gue_gro_receive */ 2358 u8 is_fou:1; 2359 2360 /* Used to determine if flush_id can be ignored */ 2361 u8 is_atomic:1; 2362 2363 /* Number of gro_receive callbacks this packet already went through */ 2364 u8 recursion_counter:4; 2365 2366 /* GRO is done by frag_list pointer chaining. */ 2367 u8 is_flist:1; 2368 2369 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2370 __wsum csum; 2371 2372 /* used in skb_gro_receive() slow path */ 2373 struct sk_buff *last; 2374 }; 2375 2376 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2377 2378 #define GRO_RECURSION_LIMIT 15 2379 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2380 { 2381 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2382 } 2383 2384 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2385 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2386 struct list_head *head, 2387 struct sk_buff *skb) 2388 { 2389 if (unlikely(gro_recursion_inc_test(skb))) { 2390 NAPI_GRO_CB(skb)->flush |= 1; 2391 return NULL; 2392 } 2393 2394 return cb(head, skb); 2395 } 2396 2397 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2398 struct sk_buff *); 2399 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2400 struct sock *sk, 2401 struct list_head *head, 2402 struct sk_buff *skb) 2403 { 2404 if (unlikely(gro_recursion_inc_test(skb))) { 2405 NAPI_GRO_CB(skb)->flush |= 1; 2406 return NULL; 2407 } 2408 2409 return cb(sk, head, skb); 2410 } 2411 2412 struct packet_type { 2413 __be16 type; /* This is really htons(ether_type). */ 2414 bool ignore_outgoing; 2415 struct net_device *dev; /* NULL is wildcarded here */ 2416 int (*func) (struct sk_buff *, 2417 struct net_device *, 2418 struct packet_type *, 2419 struct net_device *); 2420 void (*list_func) (struct list_head *, 2421 struct packet_type *, 2422 struct net_device *); 2423 bool (*id_match)(struct packet_type *ptype, 2424 struct sock *sk); 2425 void *af_packet_priv; 2426 struct list_head list; 2427 }; 2428 2429 struct offload_callbacks { 2430 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2431 netdev_features_t features); 2432 struct sk_buff *(*gro_receive)(struct list_head *head, 2433 struct sk_buff *skb); 2434 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2435 }; 2436 2437 struct packet_offload { 2438 __be16 type; /* This is really htons(ether_type). */ 2439 u16 priority; 2440 struct offload_callbacks callbacks; 2441 struct list_head list; 2442 }; 2443 2444 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2445 struct pcpu_sw_netstats { 2446 u64 rx_packets; 2447 u64 rx_bytes; 2448 u64 tx_packets; 2449 u64 tx_bytes; 2450 struct u64_stats_sync syncp; 2451 } __aligned(4 * sizeof(u64)); 2452 2453 struct pcpu_lstats { 2454 u64_stats_t packets; 2455 u64_stats_t bytes; 2456 struct u64_stats_sync syncp; 2457 } __aligned(2 * sizeof(u64)); 2458 2459 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2460 2461 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2462 { 2463 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2464 2465 u64_stats_update_begin(&lstats->syncp); 2466 u64_stats_add(&lstats->bytes, len); 2467 u64_stats_inc(&lstats->packets); 2468 u64_stats_update_end(&lstats->syncp); 2469 } 2470 2471 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2472 ({ \ 2473 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2474 if (pcpu_stats) { \ 2475 int __cpu; \ 2476 for_each_possible_cpu(__cpu) { \ 2477 typeof(type) *stat; \ 2478 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2479 u64_stats_init(&stat->syncp); \ 2480 } \ 2481 } \ 2482 pcpu_stats; \ 2483 }) 2484 2485 #define netdev_alloc_pcpu_stats(type) \ 2486 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2487 2488 enum netdev_lag_tx_type { 2489 NETDEV_LAG_TX_TYPE_UNKNOWN, 2490 NETDEV_LAG_TX_TYPE_RANDOM, 2491 NETDEV_LAG_TX_TYPE_BROADCAST, 2492 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2493 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2494 NETDEV_LAG_TX_TYPE_HASH, 2495 }; 2496 2497 enum netdev_lag_hash { 2498 NETDEV_LAG_HASH_NONE, 2499 NETDEV_LAG_HASH_L2, 2500 NETDEV_LAG_HASH_L34, 2501 NETDEV_LAG_HASH_L23, 2502 NETDEV_LAG_HASH_E23, 2503 NETDEV_LAG_HASH_E34, 2504 NETDEV_LAG_HASH_UNKNOWN, 2505 }; 2506 2507 struct netdev_lag_upper_info { 2508 enum netdev_lag_tx_type tx_type; 2509 enum netdev_lag_hash hash_type; 2510 }; 2511 2512 struct netdev_lag_lower_state_info { 2513 u8 link_up : 1, 2514 tx_enabled : 1; 2515 }; 2516 2517 #include <linux/notifier.h> 2518 2519 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2520 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2521 * adding new types. 2522 */ 2523 enum netdev_cmd { 2524 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2525 NETDEV_DOWN, 2526 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2527 detected a hardware crash and restarted 2528 - we can use this eg to kick tcp sessions 2529 once done */ 2530 NETDEV_CHANGE, /* Notify device state change */ 2531 NETDEV_REGISTER, 2532 NETDEV_UNREGISTER, 2533 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2534 NETDEV_CHANGEADDR, /* notify after the address change */ 2535 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2536 NETDEV_GOING_DOWN, 2537 NETDEV_CHANGENAME, 2538 NETDEV_FEAT_CHANGE, 2539 NETDEV_BONDING_FAILOVER, 2540 NETDEV_PRE_UP, 2541 NETDEV_PRE_TYPE_CHANGE, 2542 NETDEV_POST_TYPE_CHANGE, 2543 NETDEV_POST_INIT, 2544 NETDEV_RELEASE, 2545 NETDEV_NOTIFY_PEERS, 2546 NETDEV_JOIN, 2547 NETDEV_CHANGEUPPER, 2548 NETDEV_RESEND_IGMP, 2549 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2550 NETDEV_CHANGEINFODATA, 2551 NETDEV_BONDING_INFO, 2552 NETDEV_PRECHANGEUPPER, 2553 NETDEV_CHANGELOWERSTATE, 2554 NETDEV_UDP_TUNNEL_PUSH_INFO, 2555 NETDEV_UDP_TUNNEL_DROP_INFO, 2556 NETDEV_CHANGE_TX_QUEUE_LEN, 2557 NETDEV_CVLAN_FILTER_PUSH_INFO, 2558 NETDEV_CVLAN_FILTER_DROP_INFO, 2559 NETDEV_SVLAN_FILTER_PUSH_INFO, 2560 NETDEV_SVLAN_FILTER_DROP_INFO, 2561 }; 2562 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2563 2564 int register_netdevice_notifier(struct notifier_block *nb); 2565 int unregister_netdevice_notifier(struct notifier_block *nb); 2566 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2567 int unregister_netdevice_notifier_net(struct net *net, 2568 struct notifier_block *nb); 2569 int register_netdevice_notifier_dev_net(struct net_device *dev, 2570 struct notifier_block *nb, 2571 struct netdev_net_notifier *nn); 2572 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2573 struct notifier_block *nb, 2574 struct netdev_net_notifier *nn); 2575 2576 struct netdev_notifier_info { 2577 struct net_device *dev; 2578 struct netlink_ext_ack *extack; 2579 }; 2580 2581 struct netdev_notifier_info_ext { 2582 struct netdev_notifier_info info; /* must be first */ 2583 union { 2584 u32 mtu; 2585 } ext; 2586 }; 2587 2588 struct netdev_notifier_change_info { 2589 struct netdev_notifier_info info; /* must be first */ 2590 unsigned int flags_changed; 2591 }; 2592 2593 struct netdev_notifier_changeupper_info { 2594 struct netdev_notifier_info info; /* must be first */ 2595 struct net_device *upper_dev; /* new upper dev */ 2596 bool master; /* is upper dev master */ 2597 bool linking; /* is the notification for link or unlink */ 2598 void *upper_info; /* upper dev info */ 2599 }; 2600 2601 struct netdev_notifier_changelowerstate_info { 2602 struct netdev_notifier_info info; /* must be first */ 2603 void *lower_state_info; /* is lower dev state */ 2604 }; 2605 2606 struct netdev_notifier_pre_changeaddr_info { 2607 struct netdev_notifier_info info; /* must be first */ 2608 const unsigned char *dev_addr; 2609 }; 2610 2611 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2612 struct net_device *dev) 2613 { 2614 info->dev = dev; 2615 info->extack = NULL; 2616 } 2617 2618 static inline struct net_device * 2619 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2620 { 2621 return info->dev; 2622 } 2623 2624 static inline struct netlink_ext_ack * 2625 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2626 { 2627 return info->extack; 2628 } 2629 2630 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2631 2632 2633 extern rwlock_t dev_base_lock; /* Device list lock */ 2634 2635 #define for_each_netdev(net, d) \ 2636 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2637 #define for_each_netdev_reverse(net, d) \ 2638 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2639 #define for_each_netdev_rcu(net, d) \ 2640 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2641 #define for_each_netdev_safe(net, d, n) \ 2642 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2643 #define for_each_netdev_continue(net, d) \ 2644 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2645 #define for_each_netdev_continue_reverse(net, d) \ 2646 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2647 dev_list) 2648 #define for_each_netdev_continue_rcu(net, d) \ 2649 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2650 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2651 for_each_netdev_rcu(&init_net, slave) \ 2652 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2653 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2654 2655 static inline struct net_device *next_net_device(struct net_device *dev) 2656 { 2657 struct list_head *lh; 2658 struct net *net; 2659 2660 net = dev_net(dev); 2661 lh = dev->dev_list.next; 2662 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2663 } 2664 2665 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2666 { 2667 struct list_head *lh; 2668 struct net *net; 2669 2670 net = dev_net(dev); 2671 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2672 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2673 } 2674 2675 static inline struct net_device *first_net_device(struct net *net) 2676 { 2677 return list_empty(&net->dev_base_head) ? NULL : 2678 net_device_entry(net->dev_base_head.next); 2679 } 2680 2681 static inline struct net_device *first_net_device_rcu(struct net *net) 2682 { 2683 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2684 2685 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2686 } 2687 2688 int netdev_boot_setup_check(struct net_device *dev); 2689 unsigned long netdev_boot_base(const char *prefix, int unit); 2690 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2691 const char *hwaddr); 2692 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2693 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2694 void dev_add_pack(struct packet_type *pt); 2695 void dev_remove_pack(struct packet_type *pt); 2696 void __dev_remove_pack(struct packet_type *pt); 2697 void dev_add_offload(struct packet_offload *po); 2698 void dev_remove_offload(struct packet_offload *po); 2699 2700 int dev_get_iflink(const struct net_device *dev); 2701 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2702 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2703 unsigned short mask); 2704 struct net_device *dev_get_by_name(struct net *net, const char *name); 2705 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2706 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2707 int dev_alloc_name(struct net_device *dev, const char *name); 2708 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2709 void dev_close(struct net_device *dev); 2710 void dev_close_many(struct list_head *head, bool unlink); 2711 void dev_disable_lro(struct net_device *dev); 2712 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2713 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2714 struct net_device *sb_dev); 2715 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2716 struct net_device *sb_dev); 2717 int dev_queue_xmit(struct sk_buff *skb); 2718 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2719 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2720 int register_netdevice(struct net_device *dev); 2721 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2722 void unregister_netdevice_many(struct list_head *head); 2723 static inline void unregister_netdevice(struct net_device *dev) 2724 { 2725 unregister_netdevice_queue(dev, NULL); 2726 } 2727 2728 int netdev_refcnt_read(const struct net_device *dev); 2729 void free_netdev(struct net_device *dev); 2730 void netdev_freemem(struct net_device *dev); 2731 void synchronize_net(void); 2732 int init_dummy_netdev(struct net_device *dev); 2733 2734 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2735 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2736 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2737 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2738 int netdev_get_name(struct net *net, char *name, int ifindex); 2739 int dev_restart(struct net_device *dev); 2740 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2741 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2742 2743 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2744 { 2745 return NAPI_GRO_CB(skb)->data_offset; 2746 } 2747 2748 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2749 { 2750 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2751 } 2752 2753 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2754 { 2755 NAPI_GRO_CB(skb)->data_offset += len; 2756 } 2757 2758 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2759 unsigned int offset) 2760 { 2761 return NAPI_GRO_CB(skb)->frag0 + offset; 2762 } 2763 2764 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2765 { 2766 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2767 } 2768 2769 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2770 { 2771 NAPI_GRO_CB(skb)->frag0 = NULL; 2772 NAPI_GRO_CB(skb)->frag0_len = 0; 2773 } 2774 2775 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2776 unsigned int offset) 2777 { 2778 if (!pskb_may_pull(skb, hlen)) 2779 return NULL; 2780 2781 skb_gro_frag0_invalidate(skb); 2782 return skb->data + offset; 2783 } 2784 2785 static inline void *skb_gro_network_header(struct sk_buff *skb) 2786 { 2787 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2788 skb_network_offset(skb); 2789 } 2790 2791 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2792 const void *start, unsigned int len) 2793 { 2794 if (NAPI_GRO_CB(skb)->csum_valid) 2795 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2796 csum_partial(start, len, 0)); 2797 } 2798 2799 /* GRO checksum functions. These are logical equivalents of the normal 2800 * checksum functions (in skbuff.h) except that they operate on the GRO 2801 * offsets and fields in sk_buff. 2802 */ 2803 2804 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2805 2806 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2807 { 2808 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2809 } 2810 2811 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2812 bool zero_okay, 2813 __sum16 check) 2814 { 2815 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2816 skb_checksum_start_offset(skb) < 2817 skb_gro_offset(skb)) && 2818 !skb_at_gro_remcsum_start(skb) && 2819 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2820 (!zero_okay || check)); 2821 } 2822 2823 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2824 __wsum psum) 2825 { 2826 if (NAPI_GRO_CB(skb)->csum_valid && 2827 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2828 return 0; 2829 2830 NAPI_GRO_CB(skb)->csum = psum; 2831 2832 return __skb_gro_checksum_complete(skb); 2833 } 2834 2835 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2836 { 2837 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2838 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2839 NAPI_GRO_CB(skb)->csum_cnt--; 2840 } else { 2841 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2842 * verified a new top level checksum or an encapsulated one 2843 * during GRO. This saves work if we fallback to normal path. 2844 */ 2845 __skb_incr_checksum_unnecessary(skb); 2846 } 2847 } 2848 2849 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2850 compute_pseudo) \ 2851 ({ \ 2852 __sum16 __ret = 0; \ 2853 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2854 __ret = __skb_gro_checksum_validate_complete(skb, \ 2855 compute_pseudo(skb, proto)); \ 2856 if (!__ret) \ 2857 skb_gro_incr_csum_unnecessary(skb); \ 2858 __ret; \ 2859 }) 2860 2861 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2862 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2863 2864 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2865 compute_pseudo) \ 2866 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2867 2868 #define skb_gro_checksum_simple_validate(skb) \ 2869 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2870 2871 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2872 { 2873 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2874 !NAPI_GRO_CB(skb)->csum_valid); 2875 } 2876 2877 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2878 __wsum pseudo) 2879 { 2880 NAPI_GRO_CB(skb)->csum = ~pseudo; 2881 NAPI_GRO_CB(skb)->csum_valid = 1; 2882 } 2883 2884 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2885 do { \ 2886 if (__skb_gro_checksum_convert_check(skb)) \ 2887 __skb_gro_checksum_convert(skb, \ 2888 compute_pseudo(skb, proto)); \ 2889 } while (0) 2890 2891 struct gro_remcsum { 2892 int offset; 2893 __wsum delta; 2894 }; 2895 2896 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2897 { 2898 grc->offset = 0; 2899 grc->delta = 0; 2900 } 2901 2902 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2903 unsigned int off, size_t hdrlen, 2904 int start, int offset, 2905 struct gro_remcsum *grc, 2906 bool nopartial) 2907 { 2908 __wsum delta; 2909 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2910 2911 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2912 2913 if (!nopartial) { 2914 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2915 return ptr; 2916 } 2917 2918 ptr = skb_gro_header_fast(skb, off); 2919 if (skb_gro_header_hard(skb, off + plen)) { 2920 ptr = skb_gro_header_slow(skb, off + plen, off); 2921 if (!ptr) 2922 return NULL; 2923 } 2924 2925 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2926 start, offset); 2927 2928 /* Adjust skb->csum since we changed the packet */ 2929 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2930 2931 grc->offset = off + hdrlen + offset; 2932 grc->delta = delta; 2933 2934 return ptr; 2935 } 2936 2937 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2938 struct gro_remcsum *grc) 2939 { 2940 void *ptr; 2941 size_t plen = grc->offset + sizeof(u16); 2942 2943 if (!grc->delta) 2944 return; 2945 2946 ptr = skb_gro_header_fast(skb, grc->offset); 2947 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2948 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2949 if (!ptr) 2950 return; 2951 } 2952 2953 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2954 } 2955 2956 #ifdef CONFIG_XFRM_OFFLOAD 2957 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2958 { 2959 if (PTR_ERR(pp) != -EINPROGRESS) 2960 NAPI_GRO_CB(skb)->flush |= flush; 2961 } 2962 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2963 struct sk_buff *pp, 2964 int flush, 2965 struct gro_remcsum *grc) 2966 { 2967 if (PTR_ERR(pp) != -EINPROGRESS) { 2968 NAPI_GRO_CB(skb)->flush |= flush; 2969 skb_gro_remcsum_cleanup(skb, grc); 2970 skb->remcsum_offload = 0; 2971 } 2972 } 2973 #else 2974 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2975 { 2976 NAPI_GRO_CB(skb)->flush |= flush; 2977 } 2978 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2979 struct sk_buff *pp, 2980 int flush, 2981 struct gro_remcsum *grc) 2982 { 2983 NAPI_GRO_CB(skb)->flush |= flush; 2984 skb_gro_remcsum_cleanup(skb, grc); 2985 skb->remcsum_offload = 0; 2986 } 2987 #endif 2988 2989 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2990 unsigned short type, 2991 const void *daddr, const void *saddr, 2992 unsigned int len) 2993 { 2994 if (!dev->header_ops || !dev->header_ops->create) 2995 return 0; 2996 2997 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2998 } 2999 3000 static inline int dev_parse_header(const struct sk_buff *skb, 3001 unsigned char *haddr) 3002 { 3003 const struct net_device *dev = skb->dev; 3004 3005 if (!dev->header_ops || !dev->header_ops->parse) 3006 return 0; 3007 return dev->header_ops->parse(skb, haddr); 3008 } 3009 3010 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3011 { 3012 const struct net_device *dev = skb->dev; 3013 3014 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3015 return 0; 3016 return dev->header_ops->parse_protocol(skb); 3017 } 3018 3019 /* ll_header must have at least hard_header_len allocated */ 3020 static inline bool dev_validate_header(const struct net_device *dev, 3021 char *ll_header, int len) 3022 { 3023 if (likely(len >= dev->hard_header_len)) 3024 return true; 3025 if (len < dev->min_header_len) 3026 return false; 3027 3028 if (capable(CAP_SYS_RAWIO)) { 3029 memset(ll_header + len, 0, dev->hard_header_len - len); 3030 return true; 3031 } 3032 3033 if (dev->header_ops && dev->header_ops->validate) 3034 return dev->header_ops->validate(ll_header, len); 3035 3036 return false; 3037 } 3038 3039 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3040 int len, int size); 3041 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3042 static inline int unregister_gifconf(unsigned int family) 3043 { 3044 return register_gifconf(family, NULL); 3045 } 3046 3047 #ifdef CONFIG_NET_FLOW_LIMIT 3048 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3049 struct sd_flow_limit { 3050 u64 count; 3051 unsigned int num_buckets; 3052 unsigned int history_head; 3053 u16 history[FLOW_LIMIT_HISTORY]; 3054 u8 buckets[]; 3055 }; 3056 3057 extern int netdev_flow_limit_table_len; 3058 #endif /* CONFIG_NET_FLOW_LIMIT */ 3059 3060 /* 3061 * Incoming packets are placed on per-CPU queues 3062 */ 3063 struct softnet_data { 3064 struct list_head poll_list; 3065 struct sk_buff_head process_queue; 3066 3067 /* stats */ 3068 unsigned int processed; 3069 unsigned int time_squeeze; 3070 unsigned int received_rps; 3071 #ifdef CONFIG_RPS 3072 struct softnet_data *rps_ipi_list; 3073 #endif 3074 #ifdef CONFIG_NET_FLOW_LIMIT 3075 struct sd_flow_limit __rcu *flow_limit; 3076 #endif 3077 struct Qdisc *output_queue; 3078 struct Qdisc **output_queue_tailp; 3079 struct sk_buff *completion_queue; 3080 #ifdef CONFIG_XFRM_OFFLOAD 3081 struct sk_buff_head xfrm_backlog; 3082 #endif 3083 /* written and read only by owning cpu: */ 3084 struct { 3085 u16 recursion; 3086 u8 more; 3087 } xmit; 3088 #ifdef CONFIG_RPS 3089 /* input_queue_head should be written by cpu owning this struct, 3090 * and only read by other cpus. Worth using a cache line. 3091 */ 3092 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3093 3094 /* Elements below can be accessed between CPUs for RPS/RFS */ 3095 call_single_data_t csd ____cacheline_aligned_in_smp; 3096 struct softnet_data *rps_ipi_next; 3097 unsigned int cpu; 3098 unsigned int input_queue_tail; 3099 #endif 3100 unsigned int dropped; 3101 struct sk_buff_head input_pkt_queue; 3102 struct napi_struct backlog; 3103 3104 }; 3105 3106 static inline void input_queue_head_incr(struct softnet_data *sd) 3107 { 3108 #ifdef CONFIG_RPS 3109 sd->input_queue_head++; 3110 #endif 3111 } 3112 3113 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3114 unsigned int *qtail) 3115 { 3116 #ifdef CONFIG_RPS 3117 *qtail = ++sd->input_queue_tail; 3118 #endif 3119 } 3120 3121 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3122 3123 static inline int dev_recursion_level(void) 3124 { 3125 return this_cpu_read(softnet_data.xmit.recursion); 3126 } 3127 3128 #define XMIT_RECURSION_LIMIT 10 3129 static inline bool dev_xmit_recursion(void) 3130 { 3131 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3132 XMIT_RECURSION_LIMIT); 3133 } 3134 3135 static inline void dev_xmit_recursion_inc(void) 3136 { 3137 __this_cpu_inc(softnet_data.xmit.recursion); 3138 } 3139 3140 static inline void dev_xmit_recursion_dec(void) 3141 { 3142 __this_cpu_dec(softnet_data.xmit.recursion); 3143 } 3144 3145 void __netif_schedule(struct Qdisc *q); 3146 void netif_schedule_queue(struct netdev_queue *txq); 3147 3148 static inline void netif_tx_schedule_all(struct net_device *dev) 3149 { 3150 unsigned int i; 3151 3152 for (i = 0; i < dev->num_tx_queues; i++) 3153 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3154 } 3155 3156 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3157 { 3158 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3159 } 3160 3161 /** 3162 * netif_start_queue - allow transmit 3163 * @dev: network device 3164 * 3165 * Allow upper layers to call the device hard_start_xmit routine. 3166 */ 3167 static inline void netif_start_queue(struct net_device *dev) 3168 { 3169 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3170 } 3171 3172 static inline void netif_tx_start_all_queues(struct net_device *dev) 3173 { 3174 unsigned int i; 3175 3176 for (i = 0; i < dev->num_tx_queues; i++) { 3177 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3178 netif_tx_start_queue(txq); 3179 } 3180 } 3181 3182 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3183 3184 /** 3185 * netif_wake_queue - restart transmit 3186 * @dev: network device 3187 * 3188 * Allow upper layers to call the device hard_start_xmit routine. 3189 * Used for flow control when transmit resources are available. 3190 */ 3191 static inline void netif_wake_queue(struct net_device *dev) 3192 { 3193 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3194 } 3195 3196 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3197 { 3198 unsigned int i; 3199 3200 for (i = 0; i < dev->num_tx_queues; i++) { 3201 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3202 netif_tx_wake_queue(txq); 3203 } 3204 } 3205 3206 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3207 { 3208 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3209 } 3210 3211 /** 3212 * netif_stop_queue - stop transmitted packets 3213 * @dev: network device 3214 * 3215 * Stop upper layers calling the device hard_start_xmit routine. 3216 * Used for flow control when transmit resources are unavailable. 3217 */ 3218 static inline void netif_stop_queue(struct net_device *dev) 3219 { 3220 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3221 } 3222 3223 void netif_tx_stop_all_queues(struct net_device *dev); 3224 void netdev_update_lockdep_key(struct net_device *dev); 3225 3226 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3227 { 3228 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3229 } 3230 3231 /** 3232 * netif_queue_stopped - test if transmit queue is flowblocked 3233 * @dev: network device 3234 * 3235 * Test if transmit queue on device is currently unable to send. 3236 */ 3237 static inline bool netif_queue_stopped(const struct net_device *dev) 3238 { 3239 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3240 } 3241 3242 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3243 { 3244 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3245 } 3246 3247 static inline bool 3248 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3249 { 3250 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3251 } 3252 3253 static inline bool 3254 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3255 { 3256 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3257 } 3258 3259 /** 3260 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3261 * @dev_queue: pointer to transmit queue 3262 * 3263 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3264 * to give appropriate hint to the CPU. 3265 */ 3266 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3267 { 3268 #ifdef CONFIG_BQL 3269 prefetchw(&dev_queue->dql.num_queued); 3270 #endif 3271 } 3272 3273 /** 3274 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3275 * @dev_queue: pointer to transmit queue 3276 * 3277 * BQL enabled drivers might use this helper in their TX completion path, 3278 * to give appropriate hint to the CPU. 3279 */ 3280 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3281 { 3282 #ifdef CONFIG_BQL 3283 prefetchw(&dev_queue->dql.limit); 3284 #endif 3285 } 3286 3287 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3288 unsigned int bytes) 3289 { 3290 #ifdef CONFIG_BQL 3291 dql_queued(&dev_queue->dql, bytes); 3292 3293 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3294 return; 3295 3296 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3297 3298 /* 3299 * The XOFF flag must be set before checking the dql_avail below, 3300 * because in netdev_tx_completed_queue we update the dql_completed 3301 * before checking the XOFF flag. 3302 */ 3303 smp_mb(); 3304 3305 /* check again in case another CPU has just made room avail */ 3306 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3307 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3308 #endif 3309 } 3310 3311 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3312 * that they should not test BQL status themselves. 3313 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3314 * skb of a batch. 3315 * Returns true if the doorbell must be used to kick the NIC. 3316 */ 3317 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3318 unsigned int bytes, 3319 bool xmit_more) 3320 { 3321 if (xmit_more) { 3322 #ifdef CONFIG_BQL 3323 dql_queued(&dev_queue->dql, bytes); 3324 #endif 3325 return netif_tx_queue_stopped(dev_queue); 3326 } 3327 netdev_tx_sent_queue(dev_queue, bytes); 3328 return true; 3329 } 3330 3331 /** 3332 * netdev_sent_queue - report the number of bytes queued to hardware 3333 * @dev: network device 3334 * @bytes: number of bytes queued to the hardware device queue 3335 * 3336 * Report the number of bytes queued for sending/completion to the network 3337 * device hardware queue. @bytes should be a good approximation and should 3338 * exactly match netdev_completed_queue() @bytes 3339 */ 3340 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3341 { 3342 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3343 } 3344 3345 static inline bool __netdev_sent_queue(struct net_device *dev, 3346 unsigned int bytes, 3347 bool xmit_more) 3348 { 3349 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3350 xmit_more); 3351 } 3352 3353 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3354 unsigned int pkts, unsigned int bytes) 3355 { 3356 #ifdef CONFIG_BQL 3357 if (unlikely(!bytes)) 3358 return; 3359 3360 dql_completed(&dev_queue->dql, bytes); 3361 3362 /* 3363 * Without the memory barrier there is a small possiblity that 3364 * netdev_tx_sent_queue will miss the update and cause the queue to 3365 * be stopped forever 3366 */ 3367 smp_mb(); 3368 3369 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3370 return; 3371 3372 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3373 netif_schedule_queue(dev_queue); 3374 #endif 3375 } 3376 3377 /** 3378 * netdev_completed_queue - report bytes and packets completed by device 3379 * @dev: network device 3380 * @pkts: actual number of packets sent over the medium 3381 * @bytes: actual number of bytes sent over the medium 3382 * 3383 * Report the number of bytes and packets transmitted by the network device 3384 * hardware queue over the physical medium, @bytes must exactly match the 3385 * @bytes amount passed to netdev_sent_queue() 3386 */ 3387 static inline void netdev_completed_queue(struct net_device *dev, 3388 unsigned int pkts, unsigned int bytes) 3389 { 3390 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3391 } 3392 3393 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3394 { 3395 #ifdef CONFIG_BQL 3396 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3397 dql_reset(&q->dql); 3398 #endif 3399 } 3400 3401 /** 3402 * netdev_reset_queue - reset the packets and bytes count of a network device 3403 * @dev_queue: network device 3404 * 3405 * Reset the bytes and packet count of a network device and clear the 3406 * software flow control OFF bit for this network device 3407 */ 3408 static inline void netdev_reset_queue(struct net_device *dev_queue) 3409 { 3410 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3411 } 3412 3413 /** 3414 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3415 * @dev: network device 3416 * @queue_index: given tx queue index 3417 * 3418 * Returns 0 if given tx queue index >= number of device tx queues, 3419 * otherwise returns the originally passed tx queue index. 3420 */ 3421 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3422 { 3423 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3424 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3425 dev->name, queue_index, 3426 dev->real_num_tx_queues); 3427 return 0; 3428 } 3429 3430 return queue_index; 3431 } 3432 3433 /** 3434 * netif_running - test if up 3435 * @dev: network device 3436 * 3437 * Test if the device has been brought up. 3438 */ 3439 static inline bool netif_running(const struct net_device *dev) 3440 { 3441 return test_bit(__LINK_STATE_START, &dev->state); 3442 } 3443 3444 /* 3445 * Routines to manage the subqueues on a device. We only need start, 3446 * stop, and a check if it's stopped. All other device management is 3447 * done at the overall netdevice level. 3448 * Also test the device if we're multiqueue. 3449 */ 3450 3451 /** 3452 * netif_start_subqueue - allow sending packets on subqueue 3453 * @dev: network device 3454 * @queue_index: sub queue index 3455 * 3456 * Start individual transmit queue of a device with multiple transmit queues. 3457 */ 3458 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3459 { 3460 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3461 3462 netif_tx_start_queue(txq); 3463 } 3464 3465 /** 3466 * netif_stop_subqueue - stop sending packets on subqueue 3467 * @dev: network device 3468 * @queue_index: sub queue index 3469 * 3470 * Stop individual transmit queue of a device with multiple transmit queues. 3471 */ 3472 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3473 { 3474 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3475 netif_tx_stop_queue(txq); 3476 } 3477 3478 /** 3479 * netif_subqueue_stopped - test status of subqueue 3480 * @dev: network device 3481 * @queue_index: sub queue index 3482 * 3483 * Check individual transmit queue of a device with multiple transmit queues. 3484 */ 3485 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3486 u16 queue_index) 3487 { 3488 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3489 3490 return netif_tx_queue_stopped(txq); 3491 } 3492 3493 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3494 struct sk_buff *skb) 3495 { 3496 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3497 } 3498 3499 /** 3500 * netif_wake_subqueue - allow sending packets on subqueue 3501 * @dev: network device 3502 * @queue_index: sub queue index 3503 * 3504 * Resume individual transmit queue of a device with multiple transmit queues. 3505 */ 3506 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3507 { 3508 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3509 3510 netif_tx_wake_queue(txq); 3511 } 3512 3513 #ifdef CONFIG_XPS 3514 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3515 u16 index); 3516 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3517 u16 index, bool is_rxqs_map); 3518 3519 /** 3520 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3521 * @j: CPU/Rx queue index 3522 * @mask: bitmask of all cpus/rx queues 3523 * @nr_bits: number of bits in the bitmask 3524 * 3525 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3526 */ 3527 static inline bool netif_attr_test_mask(unsigned long j, 3528 const unsigned long *mask, 3529 unsigned int nr_bits) 3530 { 3531 cpu_max_bits_warn(j, nr_bits); 3532 return test_bit(j, mask); 3533 } 3534 3535 /** 3536 * netif_attr_test_online - Test for online CPU/Rx queue 3537 * @j: CPU/Rx queue index 3538 * @online_mask: bitmask for CPUs/Rx queues that are online 3539 * @nr_bits: number of bits in the bitmask 3540 * 3541 * Returns true if a CPU/Rx queue is online. 3542 */ 3543 static inline bool netif_attr_test_online(unsigned long j, 3544 const unsigned long *online_mask, 3545 unsigned int nr_bits) 3546 { 3547 cpu_max_bits_warn(j, nr_bits); 3548 3549 if (online_mask) 3550 return test_bit(j, online_mask); 3551 3552 return (j < nr_bits); 3553 } 3554 3555 /** 3556 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3557 * @n: CPU/Rx queue index 3558 * @srcp: the cpumask/Rx queue mask pointer 3559 * @nr_bits: number of bits in the bitmask 3560 * 3561 * Returns >= nr_bits if no further CPUs/Rx queues set. 3562 */ 3563 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3564 unsigned int nr_bits) 3565 { 3566 /* -1 is a legal arg here. */ 3567 if (n != -1) 3568 cpu_max_bits_warn(n, nr_bits); 3569 3570 if (srcp) 3571 return find_next_bit(srcp, nr_bits, n + 1); 3572 3573 return n + 1; 3574 } 3575 3576 /** 3577 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3578 * @n: CPU/Rx queue index 3579 * @src1p: the first CPUs/Rx queues mask pointer 3580 * @src2p: the second CPUs/Rx queues mask pointer 3581 * @nr_bits: number of bits in the bitmask 3582 * 3583 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3584 */ 3585 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3586 const unsigned long *src2p, 3587 unsigned int nr_bits) 3588 { 3589 /* -1 is a legal arg here. */ 3590 if (n != -1) 3591 cpu_max_bits_warn(n, nr_bits); 3592 3593 if (src1p && src2p) 3594 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3595 else if (src1p) 3596 return find_next_bit(src1p, nr_bits, n + 1); 3597 else if (src2p) 3598 return find_next_bit(src2p, nr_bits, n + 1); 3599 3600 return n + 1; 3601 } 3602 #else 3603 static inline int netif_set_xps_queue(struct net_device *dev, 3604 const struct cpumask *mask, 3605 u16 index) 3606 { 3607 return 0; 3608 } 3609 3610 static inline int __netif_set_xps_queue(struct net_device *dev, 3611 const unsigned long *mask, 3612 u16 index, bool is_rxqs_map) 3613 { 3614 return 0; 3615 } 3616 #endif 3617 3618 /** 3619 * netif_is_multiqueue - test if device has multiple transmit queues 3620 * @dev: network device 3621 * 3622 * Check if device has multiple transmit queues 3623 */ 3624 static inline bool netif_is_multiqueue(const struct net_device *dev) 3625 { 3626 return dev->num_tx_queues > 1; 3627 } 3628 3629 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3630 3631 #ifdef CONFIG_SYSFS 3632 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3633 #else 3634 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3635 unsigned int rxqs) 3636 { 3637 dev->real_num_rx_queues = rxqs; 3638 return 0; 3639 } 3640 #endif 3641 3642 static inline struct netdev_rx_queue * 3643 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3644 { 3645 return dev->_rx + rxq; 3646 } 3647 3648 #ifdef CONFIG_SYSFS 3649 static inline unsigned int get_netdev_rx_queue_index( 3650 struct netdev_rx_queue *queue) 3651 { 3652 struct net_device *dev = queue->dev; 3653 int index = queue - dev->_rx; 3654 3655 BUG_ON(index >= dev->num_rx_queues); 3656 return index; 3657 } 3658 #endif 3659 3660 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3661 int netif_get_num_default_rss_queues(void); 3662 3663 enum skb_free_reason { 3664 SKB_REASON_CONSUMED, 3665 SKB_REASON_DROPPED, 3666 }; 3667 3668 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3669 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3670 3671 /* 3672 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3673 * interrupt context or with hardware interrupts being disabled. 3674 * (in_irq() || irqs_disabled()) 3675 * 3676 * We provide four helpers that can be used in following contexts : 3677 * 3678 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3679 * replacing kfree_skb(skb) 3680 * 3681 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3682 * Typically used in place of consume_skb(skb) in TX completion path 3683 * 3684 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3685 * replacing kfree_skb(skb) 3686 * 3687 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3688 * and consumed a packet. Used in place of consume_skb(skb) 3689 */ 3690 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3691 { 3692 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3693 } 3694 3695 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3696 { 3697 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3698 } 3699 3700 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3701 { 3702 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3703 } 3704 3705 static inline void dev_consume_skb_any(struct sk_buff *skb) 3706 { 3707 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3708 } 3709 3710 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3711 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3712 int netif_rx(struct sk_buff *skb); 3713 int netif_rx_ni(struct sk_buff *skb); 3714 int netif_receive_skb(struct sk_buff *skb); 3715 int netif_receive_skb_core(struct sk_buff *skb); 3716 void netif_receive_skb_list(struct list_head *head); 3717 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3718 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3719 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3720 gro_result_t napi_gro_frags(struct napi_struct *napi); 3721 struct packet_offload *gro_find_receive_by_type(__be16 type); 3722 struct packet_offload *gro_find_complete_by_type(__be16 type); 3723 3724 static inline void napi_free_frags(struct napi_struct *napi) 3725 { 3726 kfree_skb(napi->skb); 3727 napi->skb = NULL; 3728 } 3729 3730 bool netdev_is_rx_handler_busy(struct net_device *dev); 3731 int netdev_rx_handler_register(struct net_device *dev, 3732 rx_handler_func_t *rx_handler, 3733 void *rx_handler_data); 3734 void netdev_rx_handler_unregister(struct net_device *dev); 3735 3736 bool dev_valid_name(const char *name); 3737 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3738 bool *need_copyout); 3739 int dev_ifconf(struct net *net, struct ifconf *, int); 3740 int dev_ethtool(struct net *net, struct ifreq *); 3741 unsigned int dev_get_flags(const struct net_device *); 3742 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3743 struct netlink_ext_ack *extack); 3744 int dev_change_flags(struct net_device *dev, unsigned int flags, 3745 struct netlink_ext_ack *extack); 3746 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3747 unsigned int gchanges); 3748 int dev_change_name(struct net_device *, const char *); 3749 int dev_set_alias(struct net_device *, const char *, size_t); 3750 int dev_get_alias(const struct net_device *, char *, size_t); 3751 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3752 int __dev_set_mtu(struct net_device *, int); 3753 int dev_validate_mtu(struct net_device *dev, int mtu, 3754 struct netlink_ext_ack *extack); 3755 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3756 struct netlink_ext_ack *extack); 3757 int dev_set_mtu(struct net_device *, int); 3758 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3759 void dev_set_group(struct net_device *, int); 3760 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3761 struct netlink_ext_ack *extack); 3762 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3763 struct netlink_ext_ack *extack); 3764 int dev_change_carrier(struct net_device *, bool new_carrier); 3765 int dev_get_phys_port_id(struct net_device *dev, 3766 struct netdev_phys_item_id *ppid); 3767 int dev_get_phys_port_name(struct net_device *dev, 3768 char *name, size_t len); 3769 int dev_get_port_parent_id(struct net_device *dev, 3770 struct netdev_phys_item_id *ppid, bool recurse); 3771 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3772 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3773 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3774 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3775 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3776 struct netdev_queue *txq, int *ret); 3777 3778 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3779 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3780 int fd, int expected_fd, u32 flags); 3781 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3782 enum bpf_netdev_command cmd); 3783 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3784 3785 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3786 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3787 bool is_skb_forwardable(const struct net_device *dev, 3788 const struct sk_buff *skb); 3789 3790 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3791 struct sk_buff *skb) 3792 { 3793 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3794 unlikely(!is_skb_forwardable(dev, skb))) { 3795 atomic_long_inc(&dev->rx_dropped); 3796 kfree_skb(skb); 3797 return NET_RX_DROP; 3798 } 3799 3800 skb_scrub_packet(skb, true); 3801 skb->priority = 0; 3802 return 0; 3803 } 3804 3805 bool dev_nit_active(struct net_device *dev); 3806 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3807 3808 extern int netdev_budget; 3809 extern unsigned int netdev_budget_usecs; 3810 3811 /* Called by rtnetlink.c:rtnl_unlock() */ 3812 void netdev_run_todo(void); 3813 3814 /** 3815 * dev_put - release reference to device 3816 * @dev: network device 3817 * 3818 * Release reference to device to allow it to be freed. 3819 */ 3820 static inline void dev_put(struct net_device *dev) 3821 { 3822 this_cpu_dec(*dev->pcpu_refcnt); 3823 } 3824 3825 /** 3826 * dev_hold - get reference to device 3827 * @dev: network device 3828 * 3829 * Hold reference to device to keep it from being freed. 3830 */ 3831 static inline void dev_hold(struct net_device *dev) 3832 { 3833 this_cpu_inc(*dev->pcpu_refcnt); 3834 } 3835 3836 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3837 * and _off may be called from IRQ context, but it is caller 3838 * who is responsible for serialization of these calls. 3839 * 3840 * The name carrier is inappropriate, these functions should really be 3841 * called netif_lowerlayer_*() because they represent the state of any 3842 * kind of lower layer not just hardware media. 3843 */ 3844 3845 void linkwatch_init_dev(struct net_device *dev); 3846 void linkwatch_fire_event(struct net_device *dev); 3847 void linkwatch_forget_dev(struct net_device *dev); 3848 3849 /** 3850 * netif_carrier_ok - test if carrier present 3851 * @dev: network device 3852 * 3853 * Check if carrier is present on device 3854 */ 3855 static inline bool netif_carrier_ok(const struct net_device *dev) 3856 { 3857 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3858 } 3859 3860 unsigned long dev_trans_start(struct net_device *dev); 3861 3862 void __netdev_watchdog_up(struct net_device *dev); 3863 3864 void netif_carrier_on(struct net_device *dev); 3865 3866 void netif_carrier_off(struct net_device *dev); 3867 3868 /** 3869 * netif_dormant_on - mark device as dormant. 3870 * @dev: network device 3871 * 3872 * Mark device as dormant (as per RFC2863). 3873 * 3874 * The dormant state indicates that the relevant interface is not 3875 * actually in a condition to pass packets (i.e., it is not 'up') but is 3876 * in a "pending" state, waiting for some external event. For "on- 3877 * demand" interfaces, this new state identifies the situation where the 3878 * interface is waiting for events to place it in the up state. 3879 */ 3880 static inline void netif_dormant_on(struct net_device *dev) 3881 { 3882 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3883 linkwatch_fire_event(dev); 3884 } 3885 3886 /** 3887 * netif_dormant_off - set device as not dormant. 3888 * @dev: network device 3889 * 3890 * Device is not in dormant state. 3891 */ 3892 static inline void netif_dormant_off(struct net_device *dev) 3893 { 3894 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3895 linkwatch_fire_event(dev); 3896 } 3897 3898 /** 3899 * netif_dormant - test if device is dormant 3900 * @dev: network device 3901 * 3902 * Check if device is dormant. 3903 */ 3904 static inline bool netif_dormant(const struct net_device *dev) 3905 { 3906 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3907 } 3908 3909 3910 /** 3911 * netif_oper_up - test if device is operational 3912 * @dev: network device 3913 * 3914 * Check if carrier is operational 3915 */ 3916 static inline bool netif_oper_up(const struct net_device *dev) 3917 { 3918 return (dev->operstate == IF_OPER_UP || 3919 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3920 } 3921 3922 /** 3923 * netif_device_present - is device available or removed 3924 * @dev: network device 3925 * 3926 * Check if device has not been removed from system. 3927 */ 3928 static inline bool netif_device_present(struct net_device *dev) 3929 { 3930 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3931 } 3932 3933 void netif_device_detach(struct net_device *dev); 3934 3935 void netif_device_attach(struct net_device *dev); 3936 3937 /* 3938 * Network interface message level settings 3939 */ 3940 3941 enum { 3942 NETIF_MSG_DRV_BIT, 3943 NETIF_MSG_PROBE_BIT, 3944 NETIF_MSG_LINK_BIT, 3945 NETIF_MSG_TIMER_BIT, 3946 NETIF_MSG_IFDOWN_BIT, 3947 NETIF_MSG_IFUP_BIT, 3948 NETIF_MSG_RX_ERR_BIT, 3949 NETIF_MSG_TX_ERR_BIT, 3950 NETIF_MSG_TX_QUEUED_BIT, 3951 NETIF_MSG_INTR_BIT, 3952 NETIF_MSG_TX_DONE_BIT, 3953 NETIF_MSG_RX_STATUS_BIT, 3954 NETIF_MSG_PKTDATA_BIT, 3955 NETIF_MSG_HW_BIT, 3956 NETIF_MSG_WOL_BIT, 3957 3958 /* When you add a new bit above, update netif_msg_class_names array 3959 * in net/ethtool/common.c 3960 */ 3961 NETIF_MSG_CLASS_COUNT, 3962 }; 3963 /* Both ethtool_ops interface and internal driver implementation use u32 */ 3964 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 3965 3966 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 3967 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 3968 3969 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 3970 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 3971 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 3972 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 3973 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 3974 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 3975 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 3976 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 3977 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 3978 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 3979 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 3980 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 3981 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 3982 #define NETIF_MSG_HW __NETIF_MSG(HW) 3983 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 3984 3985 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3986 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3987 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3988 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3989 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3990 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3991 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3992 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3993 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3994 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3995 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3996 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3997 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3998 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3999 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4000 4001 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4002 { 4003 /* use default */ 4004 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4005 return default_msg_enable_bits; 4006 if (debug_value == 0) /* no output */ 4007 return 0; 4008 /* set low N bits */ 4009 return (1U << debug_value) - 1; 4010 } 4011 4012 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4013 { 4014 spin_lock(&txq->_xmit_lock); 4015 txq->xmit_lock_owner = cpu; 4016 } 4017 4018 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4019 { 4020 __acquire(&txq->_xmit_lock); 4021 return true; 4022 } 4023 4024 static inline void __netif_tx_release(struct netdev_queue *txq) 4025 { 4026 __release(&txq->_xmit_lock); 4027 } 4028 4029 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4030 { 4031 spin_lock_bh(&txq->_xmit_lock); 4032 txq->xmit_lock_owner = smp_processor_id(); 4033 } 4034 4035 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4036 { 4037 bool ok = spin_trylock(&txq->_xmit_lock); 4038 if (likely(ok)) 4039 txq->xmit_lock_owner = smp_processor_id(); 4040 return ok; 4041 } 4042 4043 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4044 { 4045 txq->xmit_lock_owner = -1; 4046 spin_unlock(&txq->_xmit_lock); 4047 } 4048 4049 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4050 { 4051 txq->xmit_lock_owner = -1; 4052 spin_unlock_bh(&txq->_xmit_lock); 4053 } 4054 4055 static inline void txq_trans_update(struct netdev_queue *txq) 4056 { 4057 if (txq->xmit_lock_owner != -1) 4058 txq->trans_start = jiffies; 4059 } 4060 4061 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4062 static inline void netif_trans_update(struct net_device *dev) 4063 { 4064 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4065 4066 if (txq->trans_start != jiffies) 4067 txq->trans_start = jiffies; 4068 } 4069 4070 /** 4071 * netif_tx_lock - grab network device transmit lock 4072 * @dev: network device 4073 * 4074 * Get network device transmit lock 4075 */ 4076 static inline void netif_tx_lock(struct net_device *dev) 4077 { 4078 unsigned int i; 4079 int cpu; 4080 4081 spin_lock(&dev->tx_global_lock); 4082 cpu = smp_processor_id(); 4083 for (i = 0; i < dev->num_tx_queues; i++) { 4084 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4085 4086 /* We are the only thread of execution doing a 4087 * freeze, but we have to grab the _xmit_lock in 4088 * order to synchronize with threads which are in 4089 * the ->hard_start_xmit() handler and already 4090 * checked the frozen bit. 4091 */ 4092 __netif_tx_lock(txq, cpu); 4093 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4094 __netif_tx_unlock(txq); 4095 } 4096 } 4097 4098 static inline void netif_tx_lock_bh(struct net_device *dev) 4099 { 4100 local_bh_disable(); 4101 netif_tx_lock(dev); 4102 } 4103 4104 static inline void netif_tx_unlock(struct net_device *dev) 4105 { 4106 unsigned int i; 4107 4108 for (i = 0; i < dev->num_tx_queues; i++) { 4109 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4110 4111 /* No need to grab the _xmit_lock here. If the 4112 * queue is not stopped for another reason, we 4113 * force a schedule. 4114 */ 4115 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4116 netif_schedule_queue(txq); 4117 } 4118 spin_unlock(&dev->tx_global_lock); 4119 } 4120 4121 static inline void netif_tx_unlock_bh(struct net_device *dev) 4122 { 4123 netif_tx_unlock(dev); 4124 local_bh_enable(); 4125 } 4126 4127 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4128 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4129 __netif_tx_lock(txq, cpu); \ 4130 } else { \ 4131 __netif_tx_acquire(txq); \ 4132 } \ 4133 } 4134 4135 #define HARD_TX_TRYLOCK(dev, txq) \ 4136 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4137 __netif_tx_trylock(txq) : \ 4138 __netif_tx_acquire(txq)) 4139 4140 #define HARD_TX_UNLOCK(dev, txq) { \ 4141 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4142 __netif_tx_unlock(txq); \ 4143 } else { \ 4144 __netif_tx_release(txq); \ 4145 } \ 4146 } 4147 4148 static inline void netif_tx_disable(struct net_device *dev) 4149 { 4150 unsigned int i; 4151 int cpu; 4152 4153 local_bh_disable(); 4154 cpu = smp_processor_id(); 4155 for (i = 0; i < dev->num_tx_queues; i++) { 4156 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4157 4158 __netif_tx_lock(txq, cpu); 4159 netif_tx_stop_queue(txq); 4160 __netif_tx_unlock(txq); 4161 } 4162 local_bh_enable(); 4163 } 4164 4165 static inline void netif_addr_lock(struct net_device *dev) 4166 { 4167 spin_lock(&dev->addr_list_lock); 4168 } 4169 4170 static inline void netif_addr_lock_bh(struct net_device *dev) 4171 { 4172 spin_lock_bh(&dev->addr_list_lock); 4173 } 4174 4175 static inline void netif_addr_unlock(struct net_device *dev) 4176 { 4177 spin_unlock(&dev->addr_list_lock); 4178 } 4179 4180 static inline void netif_addr_unlock_bh(struct net_device *dev) 4181 { 4182 spin_unlock_bh(&dev->addr_list_lock); 4183 } 4184 4185 /* 4186 * dev_addrs walker. Should be used only for read access. Call with 4187 * rcu_read_lock held. 4188 */ 4189 #define for_each_dev_addr(dev, ha) \ 4190 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4191 4192 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4193 4194 void ether_setup(struct net_device *dev); 4195 4196 /* Support for loadable net-drivers */ 4197 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4198 unsigned char name_assign_type, 4199 void (*setup)(struct net_device *), 4200 unsigned int txqs, unsigned int rxqs); 4201 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4202 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4203 4204 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4205 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4206 count) 4207 4208 int register_netdev(struct net_device *dev); 4209 void unregister_netdev(struct net_device *dev); 4210 4211 /* General hardware address lists handling functions */ 4212 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4213 struct netdev_hw_addr_list *from_list, int addr_len); 4214 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4215 struct netdev_hw_addr_list *from_list, int addr_len); 4216 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4217 struct net_device *dev, 4218 int (*sync)(struct net_device *, const unsigned char *), 4219 int (*unsync)(struct net_device *, 4220 const unsigned char *)); 4221 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4222 struct net_device *dev, 4223 int (*sync)(struct net_device *, 4224 const unsigned char *, int), 4225 int (*unsync)(struct net_device *, 4226 const unsigned char *, int)); 4227 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4228 struct net_device *dev, 4229 int (*unsync)(struct net_device *, 4230 const unsigned char *, int)); 4231 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4232 struct net_device *dev, 4233 int (*unsync)(struct net_device *, 4234 const unsigned char *)); 4235 void __hw_addr_init(struct netdev_hw_addr_list *list); 4236 4237 /* Functions used for device addresses handling */ 4238 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4239 unsigned char addr_type); 4240 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4241 unsigned char addr_type); 4242 void dev_addr_flush(struct net_device *dev); 4243 int dev_addr_init(struct net_device *dev); 4244 4245 /* Functions used for unicast addresses handling */ 4246 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4247 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4248 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4249 int dev_uc_sync(struct net_device *to, struct net_device *from); 4250 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4251 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4252 void dev_uc_flush(struct net_device *dev); 4253 void dev_uc_init(struct net_device *dev); 4254 4255 /** 4256 * __dev_uc_sync - Synchonize device's unicast list 4257 * @dev: device to sync 4258 * @sync: function to call if address should be added 4259 * @unsync: function to call if address should be removed 4260 * 4261 * Add newly added addresses to the interface, and release 4262 * addresses that have been deleted. 4263 */ 4264 static inline int __dev_uc_sync(struct net_device *dev, 4265 int (*sync)(struct net_device *, 4266 const unsigned char *), 4267 int (*unsync)(struct net_device *, 4268 const unsigned char *)) 4269 { 4270 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4271 } 4272 4273 /** 4274 * __dev_uc_unsync - Remove synchronized addresses from device 4275 * @dev: device to sync 4276 * @unsync: function to call if address should be removed 4277 * 4278 * Remove all addresses that were added to the device by dev_uc_sync(). 4279 */ 4280 static inline void __dev_uc_unsync(struct net_device *dev, 4281 int (*unsync)(struct net_device *, 4282 const unsigned char *)) 4283 { 4284 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4285 } 4286 4287 /* Functions used for multicast addresses handling */ 4288 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4289 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4290 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4291 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4292 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4293 int dev_mc_sync(struct net_device *to, struct net_device *from); 4294 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4295 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4296 void dev_mc_flush(struct net_device *dev); 4297 void dev_mc_init(struct net_device *dev); 4298 4299 /** 4300 * __dev_mc_sync - Synchonize device's multicast list 4301 * @dev: device to sync 4302 * @sync: function to call if address should be added 4303 * @unsync: function to call if address should be removed 4304 * 4305 * Add newly added addresses to the interface, and release 4306 * addresses that have been deleted. 4307 */ 4308 static inline int __dev_mc_sync(struct net_device *dev, 4309 int (*sync)(struct net_device *, 4310 const unsigned char *), 4311 int (*unsync)(struct net_device *, 4312 const unsigned char *)) 4313 { 4314 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4315 } 4316 4317 /** 4318 * __dev_mc_unsync - Remove synchronized addresses from device 4319 * @dev: device to sync 4320 * @unsync: function to call if address should be removed 4321 * 4322 * Remove all addresses that were added to the device by dev_mc_sync(). 4323 */ 4324 static inline void __dev_mc_unsync(struct net_device *dev, 4325 int (*unsync)(struct net_device *, 4326 const unsigned char *)) 4327 { 4328 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4329 } 4330 4331 /* Functions used for secondary unicast and multicast support */ 4332 void dev_set_rx_mode(struct net_device *dev); 4333 void __dev_set_rx_mode(struct net_device *dev); 4334 int dev_set_promiscuity(struct net_device *dev, int inc); 4335 int dev_set_allmulti(struct net_device *dev, int inc); 4336 void netdev_state_change(struct net_device *dev); 4337 void netdev_notify_peers(struct net_device *dev); 4338 void netdev_features_change(struct net_device *dev); 4339 /* Load a device via the kmod */ 4340 void dev_load(struct net *net, const char *name); 4341 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4342 struct rtnl_link_stats64 *storage); 4343 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4344 const struct net_device_stats *netdev_stats); 4345 4346 extern int netdev_max_backlog; 4347 extern int netdev_tstamp_prequeue; 4348 extern int weight_p; 4349 extern int dev_weight_rx_bias; 4350 extern int dev_weight_tx_bias; 4351 extern int dev_rx_weight; 4352 extern int dev_tx_weight; 4353 extern int gro_normal_batch; 4354 4355 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4356 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4357 struct list_head **iter); 4358 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4359 struct list_head **iter); 4360 4361 /* iterate through upper list, must be called under RCU read lock */ 4362 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4363 for (iter = &(dev)->adj_list.upper, \ 4364 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4365 updev; \ 4366 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4367 4368 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4369 int (*fn)(struct net_device *upper_dev, 4370 void *data), 4371 void *data); 4372 4373 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4374 struct net_device *upper_dev); 4375 4376 bool netdev_has_any_upper_dev(struct net_device *dev); 4377 4378 void *netdev_lower_get_next_private(struct net_device *dev, 4379 struct list_head **iter); 4380 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4381 struct list_head **iter); 4382 4383 #define netdev_for_each_lower_private(dev, priv, iter) \ 4384 for (iter = (dev)->adj_list.lower.next, \ 4385 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4386 priv; \ 4387 priv = netdev_lower_get_next_private(dev, &(iter))) 4388 4389 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4390 for (iter = &(dev)->adj_list.lower, \ 4391 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4392 priv; \ 4393 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4394 4395 void *netdev_lower_get_next(struct net_device *dev, 4396 struct list_head **iter); 4397 4398 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4399 for (iter = (dev)->adj_list.lower.next, \ 4400 ldev = netdev_lower_get_next(dev, &(iter)); \ 4401 ldev; \ 4402 ldev = netdev_lower_get_next(dev, &(iter))) 4403 4404 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4405 struct list_head **iter); 4406 int netdev_walk_all_lower_dev(struct net_device *dev, 4407 int (*fn)(struct net_device *lower_dev, 4408 void *data), 4409 void *data); 4410 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4411 int (*fn)(struct net_device *lower_dev, 4412 void *data), 4413 void *data); 4414 4415 void *netdev_adjacent_get_private(struct list_head *adj_list); 4416 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4417 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4418 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4419 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4420 struct netlink_ext_ack *extack); 4421 int netdev_master_upper_dev_link(struct net_device *dev, 4422 struct net_device *upper_dev, 4423 void *upper_priv, void *upper_info, 4424 struct netlink_ext_ack *extack); 4425 void netdev_upper_dev_unlink(struct net_device *dev, 4426 struct net_device *upper_dev); 4427 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4428 struct net_device *new_dev, 4429 struct net_device *dev, 4430 struct netlink_ext_ack *extack); 4431 void netdev_adjacent_change_commit(struct net_device *old_dev, 4432 struct net_device *new_dev, 4433 struct net_device *dev); 4434 void netdev_adjacent_change_abort(struct net_device *old_dev, 4435 struct net_device *new_dev, 4436 struct net_device *dev); 4437 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4438 void *netdev_lower_dev_get_private(struct net_device *dev, 4439 struct net_device *lower_dev); 4440 void netdev_lower_state_changed(struct net_device *lower_dev, 4441 void *lower_state_info); 4442 4443 /* RSS keys are 40 or 52 bytes long */ 4444 #define NETDEV_RSS_KEY_LEN 52 4445 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4446 void netdev_rss_key_fill(void *buffer, size_t len); 4447 4448 int skb_checksum_help(struct sk_buff *skb); 4449 int skb_crc32c_csum_help(struct sk_buff *skb); 4450 int skb_csum_hwoffload_help(struct sk_buff *skb, 4451 const netdev_features_t features); 4452 4453 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4454 netdev_features_t features, bool tx_path); 4455 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4456 netdev_features_t features); 4457 4458 struct netdev_bonding_info { 4459 ifslave slave; 4460 ifbond master; 4461 }; 4462 4463 struct netdev_notifier_bonding_info { 4464 struct netdev_notifier_info info; /* must be first */ 4465 struct netdev_bonding_info bonding_info; 4466 }; 4467 4468 void netdev_bonding_info_change(struct net_device *dev, 4469 struct netdev_bonding_info *bonding_info); 4470 4471 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4472 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4473 #else 4474 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4475 const void *data) 4476 { 4477 } 4478 #endif 4479 4480 static inline 4481 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4482 { 4483 return __skb_gso_segment(skb, features, true); 4484 } 4485 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4486 4487 static inline bool can_checksum_protocol(netdev_features_t features, 4488 __be16 protocol) 4489 { 4490 if (protocol == htons(ETH_P_FCOE)) 4491 return !!(features & NETIF_F_FCOE_CRC); 4492 4493 /* Assume this is an IP checksum (not SCTP CRC) */ 4494 4495 if (features & NETIF_F_HW_CSUM) { 4496 /* Can checksum everything */ 4497 return true; 4498 } 4499 4500 switch (protocol) { 4501 case htons(ETH_P_IP): 4502 return !!(features & NETIF_F_IP_CSUM); 4503 case htons(ETH_P_IPV6): 4504 return !!(features & NETIF_F_IPV6_CSUM); 4505 default: 4506 return false; 4507 } 4508 } 4509 4510 #ifdef CONFIG_BUG 4511 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4512 #else 4513 static inline void netdev_rx_csum_fault(struct net_device *dev, 4514 struct sk_buff *skb) 4515 { 4516 } 4517 #endif 4518 /* rx skb timestamps */ 4519 void net_enable_timestamp(void); 4520 void net_disable_timestamp(void); 4521 4522 #ifdef CONFIG_PROC_FS 4523 int __init dev_proc_init(void); 4524 #else 4525 #define dev_proc_init() 0 4526 #endif 4527 4528 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4529 struct sk_buff *skb, struct net_device *dev, 4530 bool more) 4531 { 4532 __this_cpu_write(softnet_data.xmit.more, more); 4533 return ops->ndo_start_xmit(skb, dev); 4534 } 4535 4536 static inline bool netdev_xmit_more(void) 4537 { 4538 return __this_cpu_read(softnet_data.xmit.more); 4539 } 4540 4541 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4542 struct netdev_queue *txq, bool more) 4543 { 4544 const struct net_device_ops *ops = dev->netdev_ops; 4545 netdev_tx_t rc; 4546 4547 rc = __netdev_start_xmit(ops, skb, dev, more); 4548 if (rc == NETDEV_TX_OK) 4549 txq_trans_update(txq); 4550 4551 return rc; 4552 } 4553 4554 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4555 const void *ns); 4556 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4557 const void *ns); 4558 4559 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4560 { 4561 return netdev_class_create_file_ns(class_attr, NULL); 4562 } 4563 4564 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4565 { 4566 netdev_class_remove_file_ns(class_attr, NULL); 4567 } 4568 4569 extern const struct kobj_ns_type_operations net_ns_type_operations; 4570 4571 const char *netdev_drivername(const struct net_device *dev); 4572 4573 void linkwatch_run_queue(void); 4574 4575 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4576 netdev_features_t f2) 4577 { 4578 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4579 if (f1 & NETIF_F_HW_CSUM) 4580 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4581 else 4582 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4583 } 4584 4585 return f1 & f2; 4586 } 4587 4588 static inline netdev_features_t netdev_get_wanted_features( 4589 struct net_device *dev) 4590 { 4591 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4592 } 4593 netdev_features_t netdev_increment_features(netdev_features_t all, 4594 netdev_features_t one, netdev_features_t mask); 4595 4596 /* Allow TSO being used on stacked device : 4597 * Performing the GSO segmentation before last device 4598 * is a performance improvement. 4599 */ 4600 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4601 netdev_features_t mask) 4602 { 4603 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4604 } 4605 4606 int __netdev_update_features(struct net_device *dev); 4607 void netdev_update_features(struct net_device *dev); 4608 void netdev_change_features(struct net_device *dev); 4609 4610 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4611 struct net_device *dev); 4612 4613 netdev_features_t passthru_features_check(struct sk_buff *skb, 4614 struct net_device *dev, 4615 netdev_features_t features); 4616 netdev_features_t netif_skb_features(struct sk_buff *skb); 4617 4618 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4619 { 4620 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4621 4622 /* check flags correspondence */ 4623 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4624 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4625 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4626 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4627 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4628 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4629 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4630 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4631 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4632 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4633 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4634 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4635 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4636 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4637 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4638 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4639 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4640 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4641 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4642 4643 return (features & feature) == feature; 4644 } 4645 4646 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4647 { 4648 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4649 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4650 } 4651 4652 static inline bool netif_needs_gso(struct sk_buff *skb, 4653 netdev_features_t features) 4654 { 4655 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4656 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4657 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4658 } 4659 4660 static inline void netif_set_gso_max_size(struct net_device *dev, 4661 unsigned int size) 4662 { 4663 dev->gso_max_size = size; 4664 } 4665 4666 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4667 int pulled_hlen, u16 mac_offset, 4668 int mac_len) 4669 { 4670 skb->protocol = protocol; 4671 skb->encapsulation = 1; 4672 skb_push(skb, pulled_hlen); 4673 skb_reset_transport_header(skb); 4674 skb->mac_header = mac_offset; 4675 skb->network_header = skb->mac_header + mac_len; 4676 skb->mac_len = mac_len; 4677 } 4678 4679 static inline bool netif_is_macsec(const struct net_device *dev) 4680 { 4681 return dev->priv_flags & IFF_MACSEC; 4682 } 4683 4684 static inline bool netif_is_macvlan(const struct net_device *dev) 4685 { 4686 return dev->priv_flags & IFF_MACVLAN; 4687 } 4688 4689 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4690 { 4691 return dev->priv_flags & IFF_MACVLAN_PORT; 4692 } 4693 4694 static inline bool netif_is_bond_master(const struct net_device *dev) 4695 { 4696 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4697 } 4698 4699 static inline bool netif_is_bond_slave(const struct net_device *dev) 4700 { 4701 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4702 } 4703 4704 static inline bool netif_supports_nofcs(struct net_device *dev) 4705 { 4706 return dev->priv_flags & IFF_SUPP_NOFCS; 4707 } 4708 4709 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4710 { 4711 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4712 } 4713 4714 static inline bool netif_is_l3_master(const struct net_device *dev) 4715 { 4716 return dev->priv_flags & IFF_L3MDEV_MASTER; 4717 } 4718 4719 static inline bool netif_is_l3_slave(const struct net_device *dev) 4720 { 4721 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4722 } 4723 4724 static inline bool netif_is_bridge_master(const struct net_device *dev) 4725 { 4726 return dev->priv_flags & IFF_EBRIDGE; 4727 } 4728 4729 static inline bool netif_is_bridge_port(const struct net_device *dev) 4730 { 4731 return dev->priv_flags & IFF_BRIDGE_PORT; 4732 } 4733 4734 static inline bool netif_is_ovs_master(const struct net_device *dev) 4735 { 4736 return dev->priv_flags & IFF_OPENVSWITCH; 4737 } 4738 4739 static inline bool netif_is_ovs_port(const struct net_device *dev) 4740 { 4741 return dev->priv_flags & IFF_OVS_DATAPATH; 4742 } 4743 4744 static inline bool netif_is_team_master(const struct net_device *dev) 4745 { 4746 return dev->priv_flags & IFF_TEAM; 4747 } 4748 4749 static inline bool netif_is_team_port(const struct net_device *dev) 4750 { 4751 return dev->priv_flags & IFF_TEAM_PORT; 4752 } 4753 4754 static inline bool netif_is_lag_master(const struct net_device *dev) 4755 { 4756 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4757 } 4758 4759 static inline bool netif_is_lag_port(const struct net_device *dev) 4760 { 4761 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4762 } 4763 4764 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4765 { 4766 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4767 } 4768 4769 static inline bool netif_is_failover(const struct net_device *dev) 4770 { 4771 return dev->priv_flags & IFF_FAILOVER; 4772 } 4773 4774 static inline bool netif_is_failover_slave(const struct net_device *dev) 4775 { 4776 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4777 } 4778 4779 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4780 static inline void netif_keep_dst(struct net_device *dev) 4781 { 4782 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4783 } 4784 4785 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4786 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4787 { 4788 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4789 return dev->priv_flags & IFF_MACSEC; 4790 } 4791 4792 extern struct pernet_operations __net_initdata loopback_net_ops; 4793 4794 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4795 4796 /* netdev_printk helpers, similar to dev_printk */ 4797 4798 static inline const char *netdev_name(const struct net_device *dev) 4799 { 4800 if (!dev->name[0] || strchr(dev->name, '%')) 4801 return "(unnamed net_device)"; 4802 return dev->name; 4803 } 4804 4805 static inline bool netdev_unregistering(const struct net_device *dev) 4806 { 4807 return dev->reg_state == NETREG_UNREGISTERING; 4808 } 4809 4810 static inline const char *netdev_reg_state(const struct net_device *dev) 4811 { 4812 switch (dev->reg_state) { 4813 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4814 case NETREG_REGISTERED: return ""; 4815 case NETREG_UNREGISTERING: return " (unregistering)"; 4816 case NETREG_UNREGISTERED: return " (unregistered)"; 4817 case NETREG_RELEASED: return " (released)"; 4818 case NETREG_DUMMY: return " (dummy)"; 4819 } 4820 4821 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4822 return " (unknown)"; 4823 } 4824 4825 __printf(3, 4) __cold 4826 void netdev_printk(const char *level, const struct net_device *dev, 4827 const char *format, ...); 4828 __printf(2, 3) __cold 4829 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4830 __printf(2, 3) __cold 4831 void netdev_alert(const struct net_device *dev, const char *format, ...); 4832 __printf(2, 3) __cold 4833 void netdev_crit(const struct net_device *dev, const char *format, ...); 4834 __printf(2, 3) __cold 4835 void netdev_err(const struct net_device *dev, const char *format, ...); 4836 __printf(2, 3) __cold 4837 void netdev_warn(const struct net_device *dev, const char *format, ...); 4838 __printf(2, 3) __cold 4839 void netdev_notice(const struct net_device *dev, const char *format, ...); 4840 __printf(2, 3) __cold 4841 void netdev_info(const struct net_device *dev, const char *format, ...); 4842 4843 #define netdev_level_once(level, dev, fmt, ...) \ 4844 do { \ 4845 static bool __print_once __read_mostly; \ 4846 \ 4847 if (!__print_once) { \ 4848 __print_once = true; \ 4849 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4850 } \ 4851 } while (0) 4852 4853 #define netdev_emerg_once(dev, fmt, ...) \ 4854 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4855 #define netdev_alert_once(dev, fmt, ...) \ 4856 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4857 #define netdev_crit_once(dev, fmt, ...) \ 4858 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4859 #define netdev_err_once(dev, fmt, ...) \ 4860 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4861 #define netdev_warn_once(dev, fmt, ...) \ 4862 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4863 #define netdev_notice_once(dev, fmt, ...) \ 4864 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4865 #define netdev_info_once(dev, fmt, ...) \ 4866 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4867 4868 #define MODULE_ALIAS_NETDEV(device) \ 4869 MODULE_ALIAS("netdev-" device) 4870 4871 #if defined(CONFIG_DYNAMIC_DEBUG) 4872 #define netdev_dbg(__dev, format, args...) \ 4873 do { \ 4874 dynamic_netdev_dbg(__dev, format, ##args); \ 4875 } while (0) 4876 #elif defined(DEBUG) 4877 #define netdev_dbg(__dev, format, args...) \ 4878 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4879 #else 4880 #define netdev_dbg(__dev, format, args...) \ 4881 ({ \ 4882 if (0) \ 4883 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4884 }) 4885 #endif 4886 4887 #if defined(VERBOSE_DEBUG) 4888 #define netdev_vdbg netdev_dbg 4889 #else 4890 4891 #define netdev_vdbg(dev, format, args...) \ 4892 ({ \ 4893 if (0) \ 4894 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4895 0; \ 4896 }) 4897 #endif 4898 4899 /* 4900 * netdev_WARN() acts like dev_printk(), but with the key difference 4901 * of using a WARN/WARN_ON to get the message out, including the 4902 * file/line information and a backtrace. 4903 */ 4904 #define netdev_WARN(dev, format, args...) \ 4905 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4906 netdev_reg_state(dev), ##args) 4907 4908 #define netdev_WARN_ONCE(dev, format, args...) \ 4909 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4910 netdev_reg_state(dev), ##args) 4911 4912 /* netif printk helpers, similar to netdev_printk */ 4913 4914 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4915 do { \ 4916 if (netif_msg_##type(priv)) \ 4917 netdev_printk(level, (dev), fmt, ##args); \ 4918 } while (0) 4919 4920 #define netif_level(level, priv, type, dev, fmt, args...) \ 4921 do { \ 4922 if (netif_msg_##type(priv)) \ 4923 netdev_##level(dev, fmt, ##args); \ 4924 } while (0) 4925 4926 #define netif_emerg(priv, type, dev, fmt, args...) \ 4927 netif_level(emerg, priv, type, dev, fmt, ##args) 4928 #define netif_alert(priv, type, dev, fmt, args...) \ 4929 netif_level(alert, priv, type, dev, fmt, ##args) 4930 #define netif_crit(priv, type, dev, fmt, args...) \ 4931 netif_level(crit, priv, type, dev, fmt, ##args) 4932 #define netif_err(priv, type, dev, fmt, args...) \ 4933 netif_level(err, priv, type, dev, fmt, ##args) 4934 #define netif_warn(priv, type, dev, fmt, args...) \ 4935 netif_level(warn, priv, type, dev, fmt, ##args) 4936 #define netif_notice(priv, type, dev, fmt, args...) \ 4937 netif_level(notice, priv, type, dev, fmt, ##args) 4938 #define netif_info(priv, type, dev, fmt, args...) \ 4939 netif_level(info, priv, type, dev, fmt, ##args) 4940 4941 #if defined(CONFIG_DYNAMIC_DEBUG) 4942 #define netif_dbg(priv, type, netdev, format, args...) \ 4943 do { \ 4944 if (netif_msg_##type(priv)) \ 4945 dynamic_netdev_dbg(netdev, format, ##args); \ 4946 } while (0) 4947 #elif defined(DEBUG) 4948 #define netif_dbg(priv, type, dev, format, args...) \ 4949 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4950 #else 4951 #define netif_dbg(priv, type, dev, format, args...) \ 4952 ({ \ 4953 if (0) \ 4954 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4955 0; \ 4956 }) 4957 #endif 4958 4959 /* if @cond then downgrade to debug, else print at @level */ 4960 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4961 do { \ 4962 if (cond) \ 4963 netif_dbg(priv, type, netdev, fmt, ##args); \ 4964 else \ 4965 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4966 } while (0) 4967 4968 #if defined(VERBOSE_DEBUG) 4969 #define netif_vdbg netif_dbg 4970 #else 4971 #define netif_vdbg(priv, type, dev, format, args...) \ 4972 ({ \ 4973 if (0) \ 4974 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4975 0; \ 4976 }) 4977 #endif 4978 4979 /* 4980 * The list of packet types we will receive (as opposed to discard) 4981 * and the routines to invoke. 4982 * 4983 * Why 16. Because with 16 the only overlap we get on a hash of the 4984 * low nibble of the protocol value is RARP/SNAP/X.25. 4985 * 4986 * 0800 IP 4987 * 0001 802.3 4988 * 0002 AX.25 4989 * 0004 802.2 4990 * 8035 RARP 4991 * 0005 SNAP 4992 * 0805 X.25 4993 * 0806 ARP 4994 * 8137 IPX 4995 * 0009 Localtalk 4996 * 86DD IPv6 4997 */ 4998 #define PTYPE_HASH_SIZE (16) 4999 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5000 5001 extern struct net_device *blackhole_netdev; 5002 5003 #endif /* _LINUX_NETDEVICE_H */ 5004