1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void synchronize_net(void); 74 void netdev_set_default_ethtool_ops(struct net_device *dev, 75 const struct ethtool_ops *ops); 76 77 /* Backlog congestion levels */ 78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 79 #define NET_RX_DROP 1 /* packet dropped */ 80 81 #define MAX_NEST_DEV 8 82 83 /* 84 * Transmit return codes: transmit return codes originate from three different 85 * namespaces: 86 * 87 * - qdisc return codes 88 * - driver transmit return codes 89 * - errno values 90 * 91 * Drivers are allowed to return any one of those in their hard_start_xmit() 92 * function. Real network devices commonly used with qdiscs should only return 93 * the driver transmit return codes though - when qdiscs are used, the actual 94 * transmission happens asynchronously, so the value is not propagated to 95 * higher layers. Virtual network devices transmit synchronously; in this case 96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 97 * others are propagated to higher layers. 98 */ 99 100 /* qdisc ->enqueue() return codes. */ 101 #define NET_XMIT_SUCCESS 0x00 102 #define NET_XMIT_DROP 0x01 /* skb dropped */ 103 #define NET_XMIT_CN 0x02 /* congestion notification */ 104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 105 106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 107 * indicates that the device will soon be dropping packets, or already drops 108 * some packets of the same priority; prompting us to send less aggressively. */ 109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 111 112 /* Driver transmit return codes */ 113 #define NETDEV_TX_MASK 0xf0 114 115 enum netdev_tx { 116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 117 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 119 }; 120 typedef enum netdev_tx netdev_tx_t; 121 122 /* 123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 125 */ 126 static inline bool dev_xmit_complete(int rc) 127 { 128 /* 129 * Positive cases with an skb consumed by a driver: 130 * - successful transmission (rc == NETDEV_TX_OK) 131 * - error while transmitting (rc < 0) 132 * - error while queueing to a different device (rc & NET_XMIT_MASK) 133 */ 134 if (likely(rc < NET_XMIT_MASK)) 135 return true; 136 137 return false; 138 } 139 140 /* 141 * Compute the worst-case header length according to the protocols 142 * used. 143 */ 144 145 #if defined(CONFIG_HYPERV_NET) 146 # define LL_MAX_HEADER 128 147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 148 # if defined(CONFIG_MAC80211_MESH) 149 # define LL_MAX_HEADER 128 150 # else 151 # define LL_MAX_HEADER 96 152 # endif 153 #else 154 # define LL_MAX_HEADER 32 155 #endif 156 157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 159 #define MAX_HEADER LL_MAX_HEADER 160 #else 161 #define MAX_HEADER (LL_MAX_HEADER + 48) 162 #endif 163 164 /* 165 * Old network device statistics. Fields are native words 166 * (unsigned long) so they can be read and written atomically. 167 */ 168 169 struct net_device_stats { 170 unsigned long rx_packets; 171 unsigned long tx_packets; 172 unsigned long rx_bytes; 173 unsigned long tx_bytes; 174 unsigned long rx_errors; 175 unsigned long tx_errors; 176 unsigned long rx_dropped; 177 unsigned long tx_dropped; 178 unsigned long multicast; 179 unsigned long collisions; 180 unsigned long rx_length_errors; 181 unsigned long rx_over_errors; 182 unsigned long rx_crc_errors; 183 unsigned long rx_frame_errors; 184 unsigned long rx_fifo_errors; 185 unsigned long rx_missed_errors; 186 unsigned long tx_aborted_errors; 187 unsigned long tx_carrier_errors; 188 unsigned long tx_fifo_errors; 189 unsigned long tx_heartbeat_errors; 190 unsigned long tx_window_errors; 191 unsigned long rx_compressed; 192 unsigned long tx_compressed; 193 }; 194 195 196 #include <linux/cache.h> 197 #include <linux/skbuff.h> 198 199 #ifdef CONFIG_RPS 200 #include <linux/static_key.h> 201 extern struct static_key_false rps_needed; 202 extern struct static_key_false rfs_needed; 203 #endif 204 205 struct neighbour; 206 struct neigh_parms; 207 struct sk_buff; 208 209 struct netdev_hw_addr { 210 struct list_head list; 211 unsigned char addr[MAX_ADDR_LEN]; 212 unsigned char type; 213 #define NETDEV_HW_ADDR_T_LAN 1 214 #define NETDEV_HW_ADDR_T_SAN 2 215 #define NETDEV_HW_ADDR_T_UNICAST 3 216 #define NETDEV_HW_ADDR_T_MULTICAST 4 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 }; 351 352 enum { 353 NAPI_STATE_SCHED, /* Poll is scheduled */ 354 NAPI_STATE_MISSED, /* reschedule a napi */ 355 NAPI_STATE_DISABLE, /* Disable pending */ 356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 357 NAPI_STATE_LISTED, /* NAPI added to system lists */ 358 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 359 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 360 }; 361 362 enum { 363 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 364 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 365 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 366 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 367 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 368 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 369 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 370 }; 371 372 enum gro_result { 373 GRO_MERGED, 374 GRO_MERGED_FREE, 375 GRO_HELD, 376 GRO_NORMAL, 377 GRO_DROP, 378 GRO_CONSUMED, 379 }; 380 typedef enum gro_result gro_result_t; 381 382 /* 383 * enum rx_handler_result - Possible return values for rx_handlers. 384 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 385 * further. 386 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 387 * case skb->dev was changed by rx_handler. 388 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 389 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 390 * 391 * rx_handlers are functions called from inside __netif_receive_skb(), to do 392 * special processing of the skb, prior to delivery to protocol handlers. 393 * 394 * Currently, a net_device can only have a single rx_handler registered. Trying 395 * to register a second rx_handler will return -EBUSY. 396 * 397 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 398 * To unregister a rx_handler on a net_device, use 399 * netdev_rx_handler_unregister(). 400 * 401 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 402 * do with the skb. 403 * 404 * If the rx_handler consumed the skb in some way, it should return 405 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 406 * the skb to be delivered in some other way. 407 * 408 * If the rx_handler changed skb->dev, to divert the skb to another 409 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 410 * new device will be called if it exists. 411 * 412 * If the rx_handler decides the skb should be ignored, it should return 413 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 414 * are registered on exact device (ptype->dev == skb->dev). 415 * 416 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 417 * delivered, it should return RX_HANDLER_PASS. 418 * 419 * A device without a registered rx_handler will behave as if rx_handler 420 * returned RX_HANDLER_PASS. 421 */ 422 423 enum rx_handler_result { 424 RX_HANDLER_CONSUMED, 425 RX_HANDLER_ANOTHER, 426 RX_HANDLER_EXACT, 427 RX_HANDLER_PASS, 428 }; 429 typedef enum rx_handler_result rx_handler_result_t; 430 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 431 432 void __napi_schedule(struct napi_struct *n); 433 void __napi_schedule_irqoff(struct napi_struct *n); 434 435 static inline bool napi_disable_pending(struct napi_struct *n) 436 { 437 return test_bit(NAPI_STATE_DISABLE, &n->state); 438 } 439 440 bool napi_schedule_prep(struct napi_struct *n); 441 442 /** 443 * napi_schedule - schedule NAPI poll 444 * @n: NAPI context 445 * 446 * Schedule NAPI poll routine to be called if it is not already 447 * running. 448 */ 449 static inline void napi_schedule(struct napi_struct *n) 450 { 451 if (napi_schedule_prep(n)) 452 __napi_schedule(n); 453 } 454 455 /** 456 * napi_schedule_irqoff - schedule NAPI poll 457 * @n: NAPI context 458 * 459 * Variant of napi_schedule(), assuming hard irqs are masked. 460 */ 461 static inline void napi_schedule_irqoff(struct napi_struct *n) 462 { 463 if (napi_schedule_prep(n)) 464 __napi_schedule_irqoff(n); 465 } 466 467 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 468 static inline bool napi_reschedule(struct napi_struct *napi) 469 { 470 if (napi_schedule_prep(napi)) { 471 __napi_schedule(napi); 472 return true; 473 } 474 return false; 475 } 476 477 bool napi_complete_done(struct napi_struct *n, int work_done); 478 /** 479 * napi_complete - NAPI processing complete 480 * @n: NAPI context 481 * 482 * Mark NAPI processing as complete. 483 * Consider using napi_complete_done() instead. 484 * Return false if device should avoid rearming interrupts. 485 */ 486 static inline bool napi_complete(struct napi_struct *n) 487 { 488 return napi_complete_done(n, 0); 489 } 490 491 /** 492 * napi_disable - prevent NAPI from scheduling 493 * @n: NAPI context 494 * 495 * Stop NAPI from being scheduled on this context. 496 * Waits till any outstanding processing completes. 497 */ 498 void napi_disable(struct napi_struct *n); 499 500 /** 501 * napi_enable - enable NAPI scheduling 502 * @n: NAPI context 503 * 504 * Resume NAPI from being scheduled on this context. 505 * Must be paired with napi_disable. 506 */ 507 static inline void napi_enable(struct napi_struct *n) 508 { 509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 510 smp_mb__before_atomic(); 511 clear_bit(NAPI_STATE_SCHED, &n->state); 512 clear_bit(NAPI_STATE_NPSVC, &n->state); 513 } 514 515 /** 516 * napi_synchronize - wait until NAPI is not running 517 * @n: NAPI context 518 * 519 * Wait until NAPI is done being scheduled on this context. 520 * Waits till any outstanding processing completes but 521 * does not disable future activations. 522 */ 523 static inline void napi_synchronize(const struct napi_struct *n) 524 { 525 if (IS_ENABLED(CONFIG_SMP)) 526 while (test_bit(NAPI_STATE_SCHED, &n->state)) 527 msleep(1); 528 else 529 barrier(); 530 } 531 532 /** 533 * napi_if_scheduled_mark_missed - if napi is running, set the 534 * NAPIF_STATE_MISSED 535 * @n: NAPI context 536 * 537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 538 * NAPI is scheduled. 539 **/ 540 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 541 { 542 unsigned long val, new; 543 544 do { 545 val = READ_ONCE(n->state); 546 if (val & NAPIF_STATE_DISABLE) 547 return true; 548 549 if (!(val & NAPIF_STATE_SCHED)) 550 return false; 551 552 new = val | NAPIF_STATE_MISSED; 553 } while (cmpxchg(&n->state, val, new) != val); 554 555 return true; 556 } 557 558 enum netdev_queue_state_t { 559 __QUEUE_STATE_DRV_XOFF, 560 __QUEUE_STATE_STACK_XOFF, 561 __QUEUE_STATE_FROZEN, 562 }; 563 564 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 565 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 566 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 567 568 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 569 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 570 QUEUE_STATE_FROZEN) 571 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 572 QUEUE_STATE_FROZEN) 573 574 /* 575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 576 * netif_tx_* functions below are used to manipulate this flag. The 577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 578 * queue independently. The netif_xmit_*stopped functions below are called 579 * to check if the queue has been stopped by the driver or stack (either 580 * of the XOFF bits are set in the state). Drivers should not need to call 581 * netif_xmit*stopped functions, they should only be using netif_tx_*. 582 */ 583 584 struct netdev_queue { 585 /* 586 * read-mostly part 587 */ 588 struct net_device *dev; 589 struct Qdisc __rcu *qdisc; 590 struct Qdisc *qdisc_sleeping; 591 #ifdef CONFIG_SYSFS 592 struct kobject kobj; 593 #endif 594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 595 int numa_node; 596 #endif 597 unsigned long tx_maxrate; 598 /* 599 * Number of TX timeouts for this queue 600 * (/sys/class/net/DEV/Q/trans_timeout) 601 */ 602 unsigned long trans_timeout; 603 604 /* Subordinate device that the queue has been assigned to */ 605 struct net_device *sb_dev; 606 #ifdef CONFIG_XDP_SOCKETS 607 struct xsk_buff_pool *pool; 608 #endif 609 /* 610 * write-mostly part 611 */ 612 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 613 int xmit_lock_owner; 614 /* 615 * Time (in jiffies) of last Tx 616 */ 617 unsigned long trans_start; 618 619 unsigned long state; 620 621 #ifdef CONFIG_BQL 622 struct dql dql; 623 #endif 624 } ____cacheline_aligned_in_smp; 625 626 extern int sysctl_fb_tunnels_only_for_init_net; 627 extern int sysctl_devconf_inherit_init_net; 628 629 /* 630 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 631 * == 1 : For initns only 632 * == 2 : For none. 633 */ 634 static inline bool net_has_fallback_tunnels(const struct net *net) 635 { 636 return !IS_ENABLED(CONFIG_SYSCTL) || 637 !sysctl_fb_tunnels_only_for_init_net || 638 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 639 } 640 641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 642 { 643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 644 return q->numa_node; 645 #else 646 return NUMA_NO_NODE; 647 #endif 648 } 649 650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 q->numa_node = node; 654 #endif 655 } 656 657 #ifdef CONFIG_RPS 658 /* 659 * This structure holds an RPS map which can be of variable length. The 660 * map is an array of CPUs. 661 */ 662 struct rps_map { 663 unsigned int len; 664 struct rcu_head rcu; 665 u16 cpus[]; 666 }; 667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 668 669 /* 670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 671 * tail pointer for that CPU's input queue at the time of last enqueue, and 672 * a hardware filter index. 673 */ 674 struct rps_dev_flow { 675 u16 cpu; 676 u16 filter; 677 unsigned int last_qtail; 678 }; 679 #define RPS_NO_FILTER 0xffff 680 681 /* 682 * The rps_dev_flow_table structure contains a table of flow mappings. 683 */ 684 struct rps_dev_flow_table { 685 unsigned int mask; 686 struct rcu_head rcu; 687 struct rps_dev_flow flows[]; 688 }; 689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 690 ((_num) * sizeof(struct rps_dev_flow))) 691 692 /* 693 * The rps_sock_flow_table contains mappings of flows to the last CPU 694 * on which they were processed by the application (set in recvmsg). 695 * Each entry is a 32bit value. Upper part is the high-order bits 696 * of flow hash, lower part is CPU number. 697 * rps_cpu_mask is used to partition the space, depending on number of 698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 700 * meaning we use 32-6=26 bits for the hash. 701 */ 702 struct rps_sock_flow_table { 703 u32 mask; 704 705 u32 ents[] ____cacheline_aligned_in_smp; 706 }; 707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 708 709 #define RPS_NO_CPU 0xffff 710 711 extern u32 rps_cpu_mask; 712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 713 714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 715 u32 hash) 716 { 717 if (table && hash) { 718 unsigned int index = hash & table->mask; 719 u32 val = hash & ~rps_cpu_mask; 720 721 /* We only give a hint, preemption can change CPU under us */ 722 val |= raw_smp_processor_id(); 723 724 if (table->ents[index] != val) 725 table->ents[index] = val; 726 } 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 #endif /* CONFIG_RPS */ 734 735 /* This structure contains an instance of an RX queue. */ 736 struct netdev_rx_queue { 737 #ifdef CONFIG_RPS 738 struct rps_map __rcu *rps_map; 739 struct rps_dev_flow_table __rcu *rps_flow_table; 740 #endif 741 struct kobject kobj; 742 struct net_device *dev; 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_XDP_SOCKETS 745 struct xsk_buff_pool *pool; 746 #endif 747 } ____cacheline_aligned_in_smp; 748 749 /* 750 * RX queue sysfs structures and functions. 751 */ 752 struct rx_queue_attribute { 753 struct attribute attr; 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 755 ssize_t (*store)(struct netdev_rx_queue *queue, 756 const char *buf, size_t len); 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 */ 777 struct xps_dev_maps { 778 struct rcu_head rcu; 779 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 780 }; 781 782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 784 785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 786 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 787 788 #endif /* CONFIG_XPS */ 789 790 #define TC_MAX_QUEUE 16 791 #define TC_BITMASK 15 792 /* HW offloaded queuing disciplines txq count and offset maps */ 793 struct netdev_tc_txq { 794 u16 count; 795 u16 offset; 796 }; 797 798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 799 /* 800 * This structure is to hold information about the device 801 * configured to run FCoE protocol stack. 802 */ 803 struct netdev_fcoe_hbainfo { 804 char manufacturer[64]; 805 char serial_number[64]; 806 char hardware_version[64]; 807 char driver_version[64]; 808 char optionrom_version[64]; 809 char firmware_version[64]; 810 char model[256]; 811 char model_description[256]; 812 }; 813 #endif 814 815 #define MAX_PHYS_ITEM_ID_LEN 32 816 817 /* This structure holds a unique identifier to identify some 818 * physical item (port for example) used by a netdevice. 819 */ 820 struct netdev_phys_item_id { 821 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 822 unsigned char id_len; 823 }; 824 825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 826 struct netdev_phys_item_id *b) 827 { 828 return a->id_len == b->id_len && 829 memcmp(a->id, b->id, a->id_len) == 0; 830 } 831 832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 833 struct sk_buff *skb, 834 struct net_device *sb_dev); 835 836 enum tc_setup_type { 837 TC_SETUP_QDISC_MQPRIO, 838 TC_SETUP_CLSU32, 839 TC_SETUP_CLSFLOWER, 840 TC_SETUP_CLSMATCHALL, 841 TC_SETUP_CLSBPF, 842 TC_SETUP_BLOCK, 843 TC_SETUP_QDISC_CBS, 844 TC_SETUP_QDISC_RED, 845 TC_SETUP_QDISC_PRIO, 846 TC_SETUP_QDISC_MQ, 847 TC_SETUP_QDISC_ETF, 848 TC_SETUP_ROOT_QDISC, 849 TC_SETUP_QDISC_GRED, 850 TC_SETUP_QDISC_TAPRIO, 851 TC_SETUP_FT, 852 TC_SETUP_QDISC_ETS, 853 TC_SETUP_QDISC_TBF, 854 TC_SETUP_QDISC_FIFO, 855 }; 856 857 /* These structures hold the attributes of bpf state that are being passed 858 * to the netdevice through the bpf op. 859 */ 860 enum bpf_netdev_command { 861 /* Set or clear a bpf program used in the earliest stages of packet 862 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 863 * is responsible for calling bpf_prog_put on any old progs that are 864 * stored. In case of error, the callee need not release the new prog 865 * reference, but on success it takes ownership and must bpf_prog_put 866 * when it is no longer used. 867 */ 868 XDP_SETUP_PROG, 869 XDP_SETUP_PROG_HW, 870 /* BPF program for offload callbacks, invoked at program load time. */ 871 BPF_OFFLOAD_MAP_ALLOC, 872 BPF_OFFLOAD_MAP_FREE, 873 XDP_SETUP_XSK_POOL, 874 }; 875 876 struct bpf_prog_offload_ops; 877 struct netlink_ext_ack; 878 struct xdp_umem; 879 struct xdp_dev_bulk_queue; 880 struct bpf_xdp_link; 881 882 enum bpf_xdp_mode { 883 XDP_MODE_SKB = 0, 884 XDP_MODE_DRV = 1, 885 XDP_MODE_HW = 2, 886 __MAX_XDP_MODE 887 }; 888 889 struct bpf_xdp_entity { 890 struct bpf_prog *prog; 891 struct bpf_xdp_link *link; 892 }; 893 894 struct netdev_bpf { 895 enum bpf_netdev_command command; 896 union { 897 /* XDP_SETUP_PROG */ 898 struct { 899 u32 flags; 900 struct bpf_prog *prog; 901 struct netlink_ext_ack *extack; 902 }; 903 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 904 struct { 905 struct bpf_offloaded_map *offmap; 906 }; 907 /* XDP_SETUP_XSK_POOL */ 908 struct { 909 struct xsk_buff_pool *pool; 910 u16 queue_id; 911 } xsk; 912 }; 913 }; 914 915 /* Flags for ndo_xsk_wakeup. */ 916 #define XDP_WAKEUP_RX (1 << 0) 917 #define XDP_WAKEUP_TX (1 << 1) 918 919 #ifdef CONFIG_XFRM_OFFLOAD 920 struct xfrmdev_ops { 921 int (*xdo_dev_state_add) (struct xfrm_state *x); 922 void (*xdo_dev_state_delete) (struct xfrm_state *x); 923 void (*xdo_dev_state_free) (struct xfrm_state *x); 924 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 925 struct xfrm_state *x); 926 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 927 }; 928 #endif 929 930 struct dev_ifalias { 931 struct rcu_head rcuhead; 932 char ifalias[]; 933 }; 934 935 struct devlink; 936 struct tlsdev_ops; 937 938 struct netdev_name_node { 939 struct hlist_node hlist; 940 struct list_head list; 941 struct net_device *dev; 942 const char *name; 943 }; 944 945 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 946 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 947 948 struct netdev_net_notifier { 949 struct list_head list; 950 struct notifier_block *nb; 951 }; 952 953 /* 954 * This structure defines the management hooks for network devices. 955 * The following hooks can be defined; unless noted otherwise, they are 956 * optional and can be filled with a null pointer. 957 * 958 * int (*ndo_init)(struct net_device *dev); 959 * This function is called once when a network device is registered. 960 * The network device can use this for any late stage initialization 961 * or semantic validation. It can fail with an error code which will 962 * be propagated back to register_netdev. 963 * 964 * void (*ndo_uninit)(struct net_device *dev); 965 * This function is called when device is unregistered or when registration 966 * fails. It is not called if init fails. 967 * 968 * int (*ndo_open)(struct net_device *dev); 969 * This function is called when a network device transitions to the up 970 * state. 971 * 972 * int (*ndo_stop)(struct net_device *dev); 973 * This function is called when a network device transitions to the down 974 * state. 975 * 976 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 977 * struct net_device *dev); 978 * Called when a packet needs to be transmitted. 979 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 980 * the queue before that can happen; it's for obsolete devices and weird 981 * corner cases, but the stack really does a non-trivial amount 982 * of useless work if you return NETDEV_TX_BUSY. 983 * Required; cannot be NULL. 984 * 985 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 986 * struct net_device *dev 987 * netdev_features_t features); 988 * Called by core transmit path to determine if device is capable of 989 * performing offload operations on a given packet. This is to give 990 * the device an opportunity to implement any restrictions that cannot 991 * be otherwise expressed by feature flags. The check is called with 992 * the set of features that the stack has calculated and it returns 993 * those the driver believes to be appropriate. 994 * 995 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 996 * struct net_device *sb_dev); 997 * Called to decide which queue to use when device supports multiple 998 * transmit queues. 999 * 1000 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1001 * This function is called to allow device receiver to make 1002 * changes to configuration when multicast or promiscuous is enabled. 1003 * 1004 * void (*ndo_set_rx_mode)(struct net_device *dev); 1005 * This function is called device changes address list filtering. 1006 * If driver handles unicast address filtering, it should set 1007 * IFF_UNICAST_FLT in its priv_flags. 1008 * 1009 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1010 * This function is called when the Media Access Control address 1011 * needs to be changed. If this interface is not defined, the 1012 * MAC address can not be changed. 1013 * 1014 * int (*ndo_validate_addr)(struct net_device *dev); 1015 * Test if Media Access Control address is valid for the device. 1016 * 1017 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1018 * Called when a user requests an ioctl which can't be handled by 1019 * the generic interface code. If not defined ioctls return 1020 * not supported error code. 1021 * 1022 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1023 * Used to set network devices bus interface parameters. This interface 1024 * is retained for legacy reasons; new devices should use the bus 1025 * interface (PCI) for low level management. 1026 * 1027 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1028 * Called when a user wants to change the Maximum Transfer Unit 1029 * of a device. 1030 * 1031 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1032 * Callback used when the transmitter has not made any progress 1033 * for dev->watchdog ticks. 1034 * 1035 * void (*ndo_get_stats64)(struct net_device *dev, 1036 * struct rtnl_link_stats64 *storage); 1037 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1038 * Called when a user wants to get the network device usage 1039 * statistics. Drivers must do one of the following: 1040 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1041 * rtnl_link_stats64 structure passed by the caller. 1042 * 2. Define @ndo_get_stats to update a net_device_stats structure 1043 * (which should normally be dev->stats) and return a pointer to 1044 * it. The structure may be changed asynchronously only if each 1045 * field is written atomically. 1046 * 3. Update dev->stats asynchronously and atomically, and define 1047 * neither operation. 1048 * 1049 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1050 * Return true if this device supports offload stats of this attr_id. 1051 * 1052 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1053 * void *attr_data) 1054 * Get statistics for offload operations by attr_id. Write it into the 1055 * attr_data pointer. 1056 * 1057 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1058 * If device supports VLAN filtering this function is called when a 1059 * VLAN id is registered. 1060 * 1061 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1062 * If device supports VLAN filtering this function is called when a 1063 * VLAN id is unregistered. 1064 * 1065 * void (*ndo_poll_controller)(struct net_device *dev); 1066 * 1067 * SR-IOV management functions. 1068 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1069 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1070 * u8 qos, __be16 proto); 1071 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1072 * int max_tx_rate); 1073 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1074 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1075 * int (*ndo_get_vf_config)(struct net_device *dev, 1076 * int vf, struct ifla_vf_info *ivf); 1077 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1078 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1079 * struct nlattr *port[]); 1080 * 1081 * Enable or disable the VF ability to query its RSS Redirection Table and 1082 * Hash Key. This is needed since on some devices VF share this information 1083 * with PF and querying it may introduce a theoretical security risk. 1084 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1085 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1086 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1087 * void *type_data); 1088 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1089 * This is always called from the stack with the rtnl lock held and netif 1090 * tx queues stopped. This allows the netdevice to perform queue 1091 * management safely. 1092 * 1093 * Fiber Channel over Ethernet (FCoE) offload functions. 1094 * int (*ndo_fcoe_enable)(struct net_device *dev); 1095 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1096 * so the underlying device can perform whatever needed configuration or 1097 * initialization to support acceleration of FCoE traffic. 1098 * 1099 * int (*ndo_fcoe_disable)(struct net_device *dev); 1100 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1101 * so the underlying device can perform whatever needed clean-ups to 1102 * stop supporting acceleration of FCoE traffic. 1103 * 1104 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1105 * struct scatterlist *sgl, unsigned int sgc); 1106 * Called when the FCoE Initiator wants to initialize an I/O that 1107 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1108 * perform necessary setup and returns 1 to indicate the device is set up 1109 * successfully to perform DDP on this I/O, otherwise this returns 0. 1110 * 1111 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1112 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1113 * indicated by the FC exchange id 'xid', so the underlying device can 1114 * clean up and reuse resources for later DDP requests. 1115 * 1116 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1117 * struct scatterlist *sgl, unsigned int sgc); 1118 * Called when the FCoE Target wants to initialize an I/O that 1119 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1120 * perform necessary setup and returns 1 to indicate the device is set up 1121 * successfully to perform DDP on this I/O, otherwise this returns 0. 1122 * 1123 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1124 * struct netdev_fcoe_hbainfo *hbainfo); 1125 * Called when the FCoE Protocol stack wants information on the underlying 1126 * device. This information is utilized by the FCoE protocol stack to 1127 * register attributes with Fiber Channel management service as per the 1128 * FC-GS Fabric Device Management Information(FDMI) specification. 1129 * 1130 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1131 * Called when the underlying device wants to override default World Wide 1132 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1133 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1134 * protocol stack to use. 1135 * 1136 * RFS acceleration. 1137 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1138 * u16 rxq_index, u32 flow_id); 1139 * Set hardware filter for RFS. rxq_index is the target queue index; 1140 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1141 * Return the filter ID on success, or a negative error code. 1142 * 1143 * Slave management functions (for bridge, bonding, etc). 1144 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1145 * Called to make another netdev an underling. 1146 * 1147 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1148 * Called to release previously enslaved netdev. 1149 * 1150 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1151 * struct sk_buff *skb, 1152 * bool all_slaves); 1153 * Get the xmit slave of master device. If all_slaves is true, function 1154 * assume all the slaves can transmit. 1155 * 1156 * Feature/offload setting functions. 1157 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1158 * netdev_features_t features); 1159 * Adjusts the requested feature flags according to device-specific 1160 * constraints, and returns the resulting flags. Must not modify 1161 * the device state. 1162 * 1163 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1164 * Called to update device configuration to new features. Passed 1165 * feature set might be less than what was returned by ndo_fix_features()). 1166 * Must return >0 or -errno if it changed dev->features itself. 1167 * 1168 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1169 * struct net_device *dev, 1170 * const unsigned char *addr, u16 vid, u16 flags, 1171 * struct netlink_ext_ack *extack); 1172 * Adds an FDB entry to dev for addr. 1173 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1174 * struct net_device *dev, 1175 * const unsigned char *addr, u16 vid) 1176 * Deletes the FDB entry from dev coresponding to addr. 1177 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1178 * struct net_device *dev, struct net_device *filter_dev, 1179 * int *idx) 1180 * Used to add FDB entries to dump requests. Implementers should add 1181 * entries to skb and update idx with the number of entries. 1182 * 1183 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1184 * u16 flags, struct netlink_ext_ack *extack) 1185 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1186 * struct net_device *dev, u32 filter_mask, 1187 * int nlflags) 1188 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1189 * u16 flags); 1190 * 1191 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1192 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1193 * which do not represent real hardware may define this to allow their 1194 * userspace components to manage their virtual carrier state. Devices 1195 * that determine carrier state from physical hardware properties (eg 1196 * network cables) or protocol-dependent mechanisms (eg 1197 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1198 * 1199 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1200 * struct netdev_phys_item_id *ppid); 1201 * Called to get ID of physical port of this device. If driver does 1202 * not implement this, it is assumed that the hw is not able to have 1203 * multiple net devices on single physical port. 1204 * 1205 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1206 * struct netdev_phys_item_id *ppid) 1207 * Called to get the parent ID of the physical port of this device. 1208 * 1209 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1210 * struct udp_tunnel_info *ti); 1211 * Called by UDP tunnel to notify a driver about the UDP port and socket 1212 * address family that a UDP tunnel is listnening to. It is called only 1213 * when a new port starts listening. The operation is protected by the 1214 * RTNL. 1215 * 1216 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1217 * struct udp_tunnel_info *ti); 1218 * Called by UDP tunnel to notify the driver about a UDP port and socket 1219 * address family that the UDP tunnel is not listening to anymore. The 1220 * operation is protected by the RTNL. 1221 * 1222 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1223 * struct net_device *dev) 1224 * Called by upper layer devices to accelerate switching or other 1225 * station functionality into hardware. 'pdev is the lowerdev 1226 * to use for the offload and 'dev' is the net device that will 1227 * back the offload. Returns a pointer to the private structure 1228 * the upper layer will maintain. 1229 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1230 * Called by upper layer device to delete the station created 1231 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1232 * the station and priv is the structure returned by the add 1233 * operation. 1234 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1235 * int queue_index, u32 maxrate); 1236 * Called when a user wants to set a max-rate limitation of specific 1237 * TX queue. 1238 * int (*ndo_get_iflink)(const struct net_device *dev); 1239 * Called to get the iflink value of this device. 1240 * void (*ndo_change_proto_down)(struct net_device *dev, 1241 * bool proto_down); 1242 * This function is used to pass protocol port error state information 1243 * to the switch driver. The switch driver can react to the proto_down 1244 * by doing a phys down on the associated switch port. 1245 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1246 * This function is used to get egress tunnel information for given skb. 1247 * This is useful for retrieving outer tunnel header parameters while 1248 * sampling packet. 1249 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1250 * This function is used to specify the headroom that the skb must 1251 * consider when allocation skb during packet reception. Setting 1252 * appropriate rx headroom value allows avoiding skb head copy on 1253 * forward. Setting a negative value resets the rx headroom to the 1254 * default value. 1255 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1256 * This function is used to set or query state related to XDP on the 1257 * netdevice and manage BPF offload. See definition of 1258 * enum bpf_netdev_command for details. 1259 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1260 * u32 flags); 1261 * This function is used to submit @n XDP packets for transmit on a 1262 * netdevice. Returns number of frames successfully transmitted, frames 1263 * that got dropped are freed/returned via xdp_return_frame(). 1264 * Returns negative number, means general error invoking ndo, meaning 1265 * no frames were xmit'ed and core-caller will free all frames. 1266 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1267 * This function is used to wake up the softirq, ksoftirqd or kthread 1268 * responsible for sending and/or receiving packets on a specific 1269 * queue id bound to an AF_XDP socket. The flags field specifies if 1270 * only RX, only Tx, or both should be woken up using the flags 1271 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1272 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1273 * Get devlink port instance associated with a given netdev. 1274 * Called with a reference on the netdevice and devlink locks only, 1275 * rtnl_lock is not held. 1276 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1277 * int cmd); 1278 * Add, change, delete or get information on an IPv4 tunnel. 1279 */ 1280 struct net_device_ops { 1281 int (*ndo_init)(struct net_device *dev); 1282 void (*ndo_uninit)(struct net_device *dev); 1283 int (*ndo_open)(struct net_device *dev); 1284 int (*ndo_stop)(struct net_device *dev); 1285 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1286 struct net_device *dev); 1287 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1288 struct net_device *dev, 1289 netdev_features_t features); 1290 u16 (*ndo_select_queue)(struct net_device *dev, 1291 struct sk_buff *skb, 1292 struct net_device *sb_dev); 1293 void (*ndo_change_rx_flags)(struct net_device *dev, 1294 int flags); 1295 void (*ndo_set_rx_mode)(struct net_device *dev); 1296 int (*ndo_set_mac_address)(struct net_device *dev, 1297 void *addr); 1298 int (*ndo_validate_addr)(struct net_device *dev); 1299 int (*ndo_do_ioctl)(struct net_device *dev, 1300 struct ifreq *ifr, int cmd); 1301 int (*ndo_set_config)(struct net_device *dev, 1302 struct ifmap *map); 1303 int (*ndo_change_mtu)(struct net_device *dev, 1304 int new_mtu); 1305 int (*ndo_neigh_setup)(struct net_device *dev, 1306 struct neigh_parms *); 1307 void (*ndo_tx_timeout) (struct net_device *dev, 1308 unsigned int txqueue); 1309 1310 void (*ndo_get_stats64)(struct net_device *dev, 1311 struct rtnl_link_stats64 *storage); 1312 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1313 int (*ndo_get_offload_stats)(int attr_id, 1314 const struct net_device *dev, 1315 void *attr_data); 1316 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1317 1318 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1319 __be16 proto, u16 vid); 1320 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1321 __be16 proto, u16 vid); 1322 #ifdef CONFIG_NET_POLL_CONTROLLER 1323 void (*ndo_poll_controller)(struct net_device *dev); 1324 int (*ndo_netpoll_setup)(struct net_device *dev, 1325 struct netpoll_info *info); 1326 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1327 #endif 1328 int (*ndo_set_vf_mac)(struct net_device *dev, 1329 int queue, u8 *mac); 1330 int (*ndo_set_vf_vlan)(struct net_device *dev, 1331 int queue, u16 vlan, 1332 u8 qos, __be16 proto); 1333 int (*ndo_set_vf_rate)(struct net_device *dev, 1334 int vf, int min_tx_rate, 1335 int max_tx_rate); 1336 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1337 int vf, bool setting); 1338 int (*ndo_set_vf_trust)(struct net_device *dev, 1339 int vf, bool setting); 1340 int (*ndo_get_vf_config)(struct net_device *dev, 1341 int vf, 1342 struct ifla_vf_info *ivf); 1343 int (*ndo_set_vf_link_state)(struct net_device *dev, 1344 int vf, int link_state); 1345 int (*ndo_get_vf_stats)(struct net_device *dev, 1346 int vf, 1347 struct ifla_vf_stats 1348 *vf_stats); 1349 int (*ndo_set_vf_port)(struct net_device *dev, 1350 int vf, 1351 struct nlattr *port[]); 1352 int (*ndo_get_vf_port)(struct net_device *dev, 1353 int vf, struct sk_buff *skb); 1354 int (*ndo_get_vf_guid)(struct net_device *dev, 1355 int vf, 1356 struct ifla_vf_guid *node_guid, 1357 struct ifla_vf_guid *port_guid); 1358 int (*ndo_set_vf_guid)(struct net_device *dev, 1359 int vf, u64 guid, 1360 int guid_type); 1361 int (*ndo_set_vf_rss_query_en)( 1362 struct net_device *dev, 1363 int vf, bool setting); 1364 int (*ndo_setup_tc)(struct net_device *dev, 1365 enum tc_setup_type type, 1366 void *type_data); 1367 #if IS_ENABLED(CONFIG_FCOE) 1368 int (*ndo_fcoe_enable)(struct net_device *dev); 1369 int (*ndo_fcoe_disable)(struct net_device *dev); 1370 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1371 u16 xid, 1372 struct scatterlist *sgl, 1373 unsigned int sgc); 1374 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1375 u16 xid); 1376 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1377 u16 xid, 1378 struct scatterlist *sgl, 1379 unsigned int sgc); 1380 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1381 struct netdev_fcoe_hbainfo *hbainfo); 1382 #endif 1383 1384 #if IS_ENABLED(CONFIG_LIBFCOE) 1385 #define NETDEV_FCOE_WWNN 0 1386 #define NETDEV_FCOE_WWPN 1 1387 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1388 u64 *wwn, int type); 1389 #endif 1390 1391 #ifdef CONFIG_RFS_ACCEL 1392 int (*ndo_rx_flow_steer)(struct net_device *dev, 1393 const struct sk_buff *skb, 1394 u16 rxq_index, 1395 u32 flow_id); 1396 #endif 1397 int (*ndo_add_slave)(struct net_device *dev, 1398 struct net_device *slave_dev, 1399 struct netlink_ext_ack *extack); 1400 int (*ndo_del_slave)(struct net_device *dev, 1401 struct net_device *slave_dev); 1402 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1403 struct sk_buff *skb, 1404 bool all_slaves); 1405 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1406 netdev_features_t features); 1407 int (*ndo_set_features)(struct net_device *dev, 1408 netdev_features_t features); 1409 int (*ndo_neigh_construct)(struct net_device *dev, 1410 struct neighbour *n); 1411 void (*ndo_neigh_destroy)(struct net_device *dev, 1412 struct neighbour *n); 1413 1414 int (*ndo_fdb_add)(struct ndmsg *ndm, 1415 struct nlattr *tb[], 1416 struct net_device *dev, 1417 const unsigned char *addr, 1418 u16 vid, 1419 u16 flags, 1420 struct netlink_ext_ack *extack); 1421 int (*ndo_fdb_del)(struct ndmsg *ndm, 1422 struct nlattr *tb[], 1423 struct net_device *dev, 1424 const unsigned char *addr, 1425 u16 vid); 1426 int (*ndo_fdb_dump)(struct sk_buff *skb, 1427 struct netlink_callback *cb, 1428 struct net_device *dev, 1429 struct net_device *filter_dev, 1430 int *idx); 1431 int (*ndo_fdb_get)(struct sk_buff *skb, 1432 struct nlattr *tb[], 1433 struct net_device *dev, 1434 const unsigned char *addr, 1435 u16 vid, u32 portid, u32 seq, 1436 struct netlink_ext_ack *extack); 1437 int (*ndo_bridge_setlink)(struct net_device *dev, 1438 struct nlmsghdr *nlh, 1439 u16 flags, 1440 struct netlink_ext_ack *extack); 1441 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1442 u32 pid, u32 seq, 1443 struct net_device *dev, 1444 u32 filter_mask, 1445 int nlflags); 1446 int (*ndo_bridge_dellink)(struct net_device *dev, 1447 struct nlmsghdr *nlh, 1448 u16 flags); 1449 int (*ndo_change_carrier)(struct net_device *dev, 1450 bool new_carrier); 1451 int (*ndo_get_phys_port_id)(struct net_device *dev, 1452 struct netdev_phys_item_id *ppid); 1453 int (*ndo_get_port_parent_id)(struct net_device *dev, 1454 struct netdev_phys_item_id *ppid); 1455 int (*ndo_get_phys_port_name)(struct net_device *dev, 1456 char *name, size_t len); 1457 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1458 struct udp_tunnel_info *ti); 1459 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1460 struct udp_tunnel_info *ti); 1461 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1462 struct net_device *dev); 1463 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1464 void *priv); 1465 1466 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1467 int queue_index, 1468 u32 maxrate); 1469 int (*ndo_get_iflink)(const struct net_device *dev); 1470 int (*ndo_change_proto_down)(struct net_device *dev, 1471 bool proto_down); 1472 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1473 struct sk_buff *skb); 1474 void (*ndo_set_rx_headroom)(struct net_device *dev, 1475 int needed_headroom); 1476 int (*ndo_bpf)(struct net_device *dev, 1477 struct netdev_bpf *bpf); 1478 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1479 struct xdp_frame **xdp, 1480 u32 flags); 1481 int (*ndo_xsk_wakeup)(struct net_device *dev, 1482 u32 queue_id, u32 flags); 1483 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1484 int (*ndo_tunnel_ctl)(struct net_device *dev, 1485 struct ip_tunnel_parm *p, int cmd); 1486 }; 1487 1488 /** 1489 * enum net_device_priv_flags - &struct net_device priv_flags 1490 * 1491 * These are the &struct net_device, they are only set internally 1492 * by drivers and used in the kernel. These flags are invisible to 1493 * userspace; this means that the order of these flags can change 1494 * during any kernel release. 1495 * 1496 * You should have a pretty good reason to be extending these flags. 1497 * 1498 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1499 * @IFF_EBRIDGE: Ethernet bridging device 1500 * @IFF_BONDING: bonding master or slave 1501 * @IFF_ISATAP: ISATAP interface (RFC4214) 1502 * @IFF_WAN_HDLC: WAN HDLC device 1503 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1504 * release skb->dst 1505 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1506 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1507 * @IFF_MACVLAN_PORT: device used as macvlan port 1508 * @IFF_BRIDGE_PORT: device used as bridge port 1509 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1510 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1511 * @IFF_UNICAST_FLT: Supports unicast filtering 1512 * @IFF_TEAM_PORT: device used as team port 1513 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1514 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1515 * change when it's running 1516 * @IFF_MACVLAN: Macvlan device 1517 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1518 * underlying stacked devices 1519 * @IFF_L3MDEV_MASTER: device is an L3 master device 1520 * @IFF_NO_QUEUE: device can run without qdisc attached 1521 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1522 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1523 * @IFF_TEAM: device is a team device 1524 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1525 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1526 * entity (i.e. the master device for bridged veth) 1527 * @IFF_MACSEC: device is a MACsec device 1528 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1529 * @IFF_FAILOVER: device is a failover master device 1530 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1531 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1532 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1533 */ 1534 enum netdev_priv_flags { 1535 IFF_802_1Q_VLAN = 1<<0, 1536 IFF_EBRIDGE = 1<<1, 1537 IFF_BONDING = 1<<2, 1538 IFF_ISATAP = 1<<3, 1539 IFF_WAN_HDLC = 1<<4, 1540 IFF_XMIT_DST_RELEASE = 1<<5, 1541 IFF_DONT_BRIDGE = 1<<6, 1542 IFF_DISABLE_NETPOLL = 1<<7, 1543 IFF_MACVLAN_PORT = 1<<8, 1544 IFF_BRIDGE_PORT = 1<<9, 1545 IFF_OVS_DATAPATH = 1<<10, 1546 IFF_TX_SKB_SHARING = 1<<11, 1547 IFF_UNICAST_FLT = 1<<12, 1548 IFF_TEAM_PORT = 1<<13, 1549 IFF_SUPP_NOFCS = 1<<14, 1550 IFF_LIVE_ADDR_CHANGE = 1<<15, 1551 IFF_MACVLAN = 1<<16, 1552 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1553 IFF_L3MDEV_MASTER = 1<<18, 1554 IFF_NO_QUEUE = 1<<19, 1555 IFF_OPENVSWITCH = 1<<20, 1556 IFF_L3MDEV_SLAVE = 1<<21, 1557 IFF_TEAM = 1<<22, 1558 IFF_RXFH_CONFIGURED = 1<<23, 1559 IFF_PHONY_HEADROOM = 1<<24, 1560 IFF_MACSEC = 1<<25, 1561 IFF_NO_RX_HANDLER = 1<<26, 1562 IFF_FAILOVER = 1<<27, 1563 IFF_FAILOVER_SLAVE = 1<<28, 1564 IFF_L3MDEV_RX_HANDLER = 1<<29, 1565 IFF_LIVE_RENAME_OK = 1<<30, 1566 }; 1567 1568 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1569 #define IFF_EBRIDGE IFF_EBRIDGE 1570 #define IFF_BONDING IFF_BONDING 1571 #define IFF_ISATAP IFF_ISATAP 1572 #define IFF_WAN_HDLC IFF_WAN_HDLC 1573 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1574 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1575 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1576 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1577 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1578 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1579 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1580 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1581 #define IFF_TEAM_PORT IFF_TEAM_PORT 1582 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1583 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1584 #define IFF_MACVLAN IFF_MACVLAN 1585 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1586 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1587 #define IFF_NO_QUEUE IFF_NO_QUEUE 1588 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1589 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1590 #define IFF_TEAM IFF_TEAM 1591 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1592 #define IFF_MACSEC IFF_MACSEC 1593 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1594 #define IFF_FAILOVER IFF_FAILOVER 1595 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1596 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1597 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1598 1599 /** 1600 * struct net_device - The DEVICE structure. 1601 * 1602 * Actually, this whole structure is a big mistake. It mixes I/O 1603 * data with strictly "high-level" data, and it has to know about 1604 * almost every data structure used in the INET module. 1605 * 1606 * @name: This is the first field of the "visible" part of this structure 1607 * (i.e. as seen by users in the "Space.c" file). It is the name 1608 * of the interface. 1609 * 1610 * @name_node: Name hashlist node 1611 * @ifalias: SNMP alias 1612 * @mem_end: Shared memory end 1613 * @mem_start: Shared memory start 1614 * @base_addr: Device I/O address 1615 * @irq: Device IRQ number 1616 * 1617 * @state: Generic network queuing layer state, see netdev_state_t 1618 * @dev_list: The global list of network devices 1619 * @napi_list: List entry used for polling NAPI devices 1620 * @unreg_list: List entry when we are unregistering the 1621 * device; see the function unregister_netdev 1622 * @close_list: List entry used when we are closing the device 1623 * @ptype_all: Device-specific packet handlers for all protocols 1624 * @ptype_specific: Device-specific, protocol-specific packet handlers 1625 * 1626 * @adj_list: Directly linked devices, like slaves for bonding 1627 * @features: Currently active device features 1628 * @hw_features: User-changeable features 1629 * 1630 * @wanted_features: User-requested features 1631 * @vlan_features: Mask of features inheritable by VLAN devices 1632 * 1633 * @hw_enc_features: Mask of features inherited by encapsulating devices 1634 * This field indicates what encapsulation 1635 * offloads the hardware is capable of doing, 1636 * and drivers will need to set them appropriately. 1637 * 1638 * @mpls_features: Mask of features inheritable by MPLS 1639 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1640 * 1641 * @ifindex: interface index 1642 * @group: The group the device belongs to 1643 * 1644 * @stats: Statistics struct, which was left as a legacy, use 1645 * rtnl_link_stats64 instead 1646 * 1647 * @rx_dropped: Dropped packets by core network, 1648 * do not use this in drivers 1649 * @tx_dropped: Dropped packets by core network, 1650 * do not use this in drivers 1651 * @rx_nohandler: nohandler dropped packets by core network on 1652 * inactive devices, do not use this in drivers 1653 * @carrier_up_count: Number of times the carrier has been up 1654 * @carrier_down_count: Number of times the carrier has been down 1655 * 1656 * @wireless_handlers: List of functions to handle Wireless Extensions, 1657 * instead of ioctl, 1658 * see <net/iw_handler.h> for details. 1659 * @wireless_data: Instance data managed by the core of wireless extensions 1660 * 1661 * @netdev_ops: Includes several pointers to callbacks, 1662 * if one wants to override the ndo_*() functions 1663 * @ethtool_ops: Management operations 1664 * @l3mdev_ops: Layer 3 master device operations 1665 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1666 * discovery handling. Necessary for e.g. 6LoWPAN. 1667 * @xfrmdev_ops: Transformation offload operations 1668 * @tlsdev_ops: Transport Layer Security offload operations 1669 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1670 * of Layer 2 headers. 1671 * 1672 * @flags: Interface flags (a la BSD) 1673 * @priv_flags: Like 'flags' but invisible to userspace, 1674 * see if.h for the definitions 1675 * @gflags: Global flags ( kept as legacy ) 1676 * @padded: How much padding added by alloc_netdev() 1677 * @operstate: RFC2863 operstate 1678 * @link_mode: Mapping policy to operstate 1679 * @if_port: Selectable AUI, TP, ... 1680 * @dma: DMA channel 1681 * @mtu: Interface MTU value 1682 * @min_mtu: Interface Minimum MTU value 1683 * @max_mtu: Interface Maximum MTU value 1684 * @type: Interface hardware type 1685 * @hard_header_len: Maximum hardware header length. 1686 * @min_header_len: Minimum hardware header length 1687 * 1688 * @needed_headroom: Extra headroom the hardware may need, but not in all 1689 * cases can this be guaranteed 1690 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1691 * cases can this be guaranteed. Some cases also use 1692 * LL_MAX_HEADER instead to allocate the skb 1693 * 1694 * interface address info: 1695 * 1696 * @perm_addr: Permanent hw address 1697 * @addr_assign_type: Hw address assignment type 1698 * @addr_len: Hardware address length 1699 * @upper_level: Maximum depth level of upper devices. 1700 * @lower_level: Maximum depth level of lower devices. 1701 * @neigh_priv_len: Used in neigh_alloc() 1702 * @dev_id: Used to differentiate devices that share 1703 * the same link layer address 1704 * @dev_port: Used to differentiate devices that share 1705 * the same function 1706 * @addr_list_lock: XXX: need comments on this one 1707 * @name_assign_type: network interface name assignment type 1708 * @uc_promisc: Counter that indicates promiscuous mode 1709 * has been enabled due to the need to listen to 1710 * additional unicast addresses in a device that 1711 * does not implement ndo_set_rx_mode() 1712 * @uc: unicast mac addresses 1713 * @mc: multicast mac addresses 1714 * @dev_addrs: list of device hw addresses 1715 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1716 * @promiscuity: Number of times the NIC is told to work in 1717 * promiscuous mode; if it becomes 0 the NIC will 1718 * exit promiscuous mode 1719 * @allmulti: Counter, enables or disables allmulticast mode 1720 * 1721 * @vlan_info: VLAN info 1722 * @dsa_ptr: dsa specific data 1723 * @tipc_ptr: TIPC specific data 1724 * @atalk_ptr: AppleTalk link 1725 * @ip_ptr: IPv4 specific data 1726 * @dn_ptr: DECnet specific data 1727 * @ip6_ptr: IPv6 specific data 1728 * @ax25_ptr: AX.25 specific data 1729 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1730 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1731 * device struct 1732 * @mpls_ptr: mpls_dev struct pointer 1733 * 1734 * @dev_addr: Hw address (before bcast, 1735 * because most packets are unicast) 1736 * 1737 * @_rx: Array of RX queues 1738 * @num_rx_queues: Number of RX queues 1739 * allocated at register_netdev() time 1740 * @real_num_rx_queues: Number of RX queues currently active in device 1741 * @xdp_prog: XDP sockets filter program pointer 1742 * @gro_flush_timeout: timeout for GRO layer in NAPI 1743 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1744 * allow to avoid NIC hard IRQ, on busy queues. 1745 * 1746 * @rx_handler: handler for received packets 1747 * @rx_handler_data: XXX: need comments on this one 1748 * @miniq_ingress: ingress/clsact qdisc specific data for 1749 * ingress processing 1750 * @ingress_queue: XXX: need comments on this one 1751 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1752 * @broadcast: hw bcast address 1753 * 1754 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1755 * indexed by RX queue number. Assigned by driver. 1756 * This must only be set if the ndo_rx_flow_steer 1757 * operation is defined 1758 * @index_hlist: Device index hash chain 1759 * 1760 * @_tx: Array of TX queues 1761 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1762 * @real_num_tx_queues: Number of TX queues currently active in device 1763 * @qdisc: Root qdisc from userspace point of view 1764 * @tx_queue_len: Max frames per queue allowed 1765 * @tx_global_lock: XXX: need comments on this one 1766 * @xdp_bulkq: XDP device bulk queue 1767 * @xps_cpus_map: all CPUs map for XPS device 1768 * @xps_rxqs_map: all RXQs map for XPS device 1769 * 1770 * @xps_maps: XXX: need comments on this one 1771 * @miniq_egress: clsact qdisc specific data for 1772 * egress processing 1773 * @qdisc_hash: qdisc hash table 1774 * @watchdog_timeo: Represents the timeout that is used by 1775 * the watchdog (see dev_watchdog()) 1776 * @watchdog_timer: List of timers 1777 * 1778 * @proto_down_reason: reason a netdev interface is held down 1779 * @pcpu_refcnt: Number of references to this device 1780 * @todo_list: Delayed register/unregister 1781 * @link_watch_list: XXX: need comments on this one 1782 * 1783 * @reg_state: Register/unregister state machine 1784 * @dismantle: Device is going to be freed 1785 * @rtnl_link_state: This enum represents the phases of creating 1786 * a new link 1787 * 1788 * @needs_free_netdev: Should unregister perform free_netdev? 1789 * @priv_destructor: Called from unregister 1790 * @npinfo: XXX: need comments on this one 1791 * @nd_net: Network namespace this network device is inside 1792 * 1793 * @ml_priv: Mid-layer private 1794 * @lstats: Loopback statistics 1795 * @tstats: Tunnel statistics 1796 * @dstats: Dummy statistics 1797 * @vstats: Virtual ethernet statistics 1798 * 1799 * @garp_port: GARP 1800 * @mrp_port: MRP 1801 * 1802 * @dev: Class/net/name entry 1803 * @sysfs_groups: Space for optional device, statistics and wireless 1804 * sysfs groups 1805 * 1806 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1807 * @rtnl_link_ops: Rtnl_link_ops 1808 * 1809 * @gso_max_size: Maximum size of generic segmentation offload 1810 * @gso_max_segs: Maximum number of segments that can be passed to the 1811 * NIC for GSO 1812 * 1813 * @dcbnl_ops: Data Center Bridging netlink ops 1814 * @num_tc: Number of traffic classes in the net device 1815 * @tc_to_txq: XXX: need comments on this one 1816 * @prio_tc_map: XXX: need comments on this one 1817 * 1818 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1819 * 1820 * @priomap: XXX: need comments on this one 1821 * @phydev: Physical device may attach itself 1822 * for hardware timestamping 1823 * @sfp_bus: attached &struct sfp_bus structure. 1824 * 1825 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1826 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1827 * 1828 * @proto_down: protocol port state information can be sent to the 1829 * switch driver and used to set the phys state of the 1830 * switch port. 1831 * 1832 * @wol_enabled: Wake-on-LAN is enabled 1833 * 1834 * @net_notifier_list: List of per-net netdev notifier block 1835 * that follow this device when it is moved 1836 * to another network namespace. 1837 * 1838 * @macsec_ops: MACsec offloading ops 1839 * 1840 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1841 * offload capabilities of the device 1842 * @udp_tunnel_nic: UDP tunnel offload state 1843 * @xdp_state: stores info on attached XDP BPF programs 1844 * 1845 * FIXME: cleanup struct net_device such that network protocol info 1846 * moves out. 1847 */ 1848 1849 struct net_device { 1850 char name[IFNAMSIZ]; 1851 struct netdev_name_node *name_node; 1852 struct dev_ifalias __rcu *ifalias; 1853 /* 1854 * I/O specific fields 1855 * FIXME: Merge these and struct ifmap into one 1856 */ 1857 unsigned long mem_end; 1858 unsigned long mem_start; 1859 unsigned long base_addr; 1860 int irq; 1861 1862 /* 1863 * Some hardware also needs these fields (state,dev_list, 1864 * napi_list,unreg_list,close_list) but they are not 1865 * part of the usual set specified in Space.c. 1866 */ 1867 1868 unsigned long state; 1869 1870 struct list_head dev_list; 1871 struct list_head napi_list; 1872 struct list_head unreg_list; 1873 struct list_head close_list; 1874 struct list_head ptype_all; 1875 struct list_head ptype_specific; 1876 1877 struct { 1878 struct list_head upper; 1879 struct list_head lower; 1880 } adj_list; 1881 1882 netdev_features_t features; 1883 netdev_features_t hw_features; 1884 netdev_features_t wanted_features; 1885 netdev_features_t vlan_features; 1886 netdev_features_t hw_enc_features; 1887 netdev_features_t mpls_features; 1888 netdev_features_t gso_partial_features; 1889 1890 int ifindex; 1891 int group; 1892 1893 struct net_device_stats stats; 1894 1895 atomic_long_t rx_dropped; 1896 atomic_long_t tx_dropped; 1897 atomic_long_t rx_nohandler; 1898 1899 /* Stats to monitor link on/off, flapping */ 1900 atomic_t carrier_up_count; 1901 atomic_t carrier_down_count; 1902 1903 #ifdef CONFIG_WIRELESS_EXT 1904 const struct iw_handler_def *wireless_handlers; 1905 struct iw_public_data *wireless_data; 1906 #endif 1907 const struct net_device_ops *netdev_ops; 1908 const struct ethtool_ops *ethtool_ops; 1909 #ifdef CONFIG_NET_L3_MASTER_DEV 1910 const struct l3mdev_ops *l3mdev_ops; 1911 #endif 1912 #if IS_ENABLED(CONFIG_IPV6) 1913 const struct ndisc_ops *ndisc_ops; 1914 #endif 1915 1916 #ifdef CONFIG_XFRM_OFFLOAD 1917 const struct xfrmdev_ops *xfrmdev_ops; 1918 #endif 1919 1920 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1921 const struct tlsdev_ops *tlsdev_ops; 1922 #endif 1923 1924 const struct header_ops *header_ops; 1925 1926 unsigned int flags; 1927 unsigned int priv_flags; 1928 1929 unsigned short gflags; 1930 unsigned short padded; 1931 1932 unsigned char operstate; 1933 unsigned char link_mode; 1934 1935 unsigned char if_port; 1936 unsigned char dma; 1937 1938 /* Note : dev->mtu is often read without holding a lock. 1939 * Writers usually hold RTNL. 1940 * It is recommended to use READ_ONCE() to annotate the reads, 1941 * and to use WRITE_ONCE() to annotate the writes. 1942 */ 1943 unsigned int mtu; 1944 unsigned int min_mtu; 1945 unsigned int max_mtu; 1946 unsigned short type; 1947 unsigned short hard_header_len; 1948 unsigned char min_header_len; 1949 1950 unsigned short needed_headroom; 1951 unsigned short needed_tailroom; 1952 1953 /* Interface address info. */ 1954 unsigned char perm_addr[MAX_ADDR_LEN]; 1955 unsigned char addr_assign_type; 1956 unsigned char addr_len; 1957 unsigned char upper_level; 1958 unsigned char lower_level; 1959 unsigned short neigh_priv_len; 1960 unsigned short dev_id; 1961 unsigned short dev_port; 1962 spinlock_t addr_list_lock; 1963 unsigned char name_assign_type; 1964 bool uc_promisc; 1965 struct netdev_hw_addr_list uc; 1966 struct netdev_hw_addr_list mc; 1967 struct netdev_hw_addr_list dev_addrs; 1968 1969 #ifdef CONFIG_SYSFS 1970 struct kset *queues_kset; 1971 #endif 1972 unsigned int promiscuity; 1973 unsigned int allmulti; 1974 1975 1976 /* Protocol-specific pointers */ 1977 1978 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1979 struct vlan_info __rcu *vlan_info; 1980 #endif 1981 #if IS_ENABLED(CONFIG_NET_DSA) 1982 struct dsa_port *dsa_ptr; 1983 #endif 1984 #if IS_ENABLED(CONFIG_TIPC) 1985 struct tipc_bearer __rcu *tipc_ptr; 1986 #endif 1987 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1988 void *atalk_ptr; 1989 #endif 1990 struct in_device __rcu *ip_ptr; 1991 #if IS_ENABLED(CONFIG_DECNET) 1992 struct dn_dev __rcu *dn_ptr; 1993 #endif 1994 struct inet6_dev __rcu *ip6_ptr; 1995 #if IS_ENABLED(CONFIG_AX25) 1996 void *ax25_ptr; 1997 #endif 1998 struct wireless_dev *ieee80211_ptr; 1999 struct wpan_dev *ieee802154_ptr; 2000 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2001 struct mpls_dev __rcu *mpls_ptr; 2002 #endif 2003 2004 /* 2005 * Cache lines mostly used on receive path (including eth_type_trans()) 2006 */ 2007 /* Interface address info used in eth_type_trans() */ 2008 unsigned char *dev_addr; 2009 2010 struct netdev_rx_queue *_rx; 2011 unsigned int num_rx_queues; 2012 unsigned int real_num_rx_queues; 2013 2014 struct bpf_prog __rcu *xdp_prog; 2015 unsigned long gro_flush_timeout; 2016 int napi_defer_hard_irqs; 2017 rx_handler_func_t __rcu *rx_handler; 2018 void __rcu *rx_handler_data; 2019 2020 #ifdef CONFIG_NET_CLS_ACT 2021 struct mini_Qdisc __rcu *miniq_ingress; 2022 #endif 2023 struct netdev_queue __rcu *ingress_queue; 2024 #ifdef CONFIG_NETFILTER_INGRESS 2025 struct nf_hook_entries __rcu *nf_hooks_ingress; 2026 #endif 2027 2028 unsigned char broadcast[MAX_ADDR_LEN]; 2029 #ifdef CONFIG_RFS_ACCEL 2030 struct cpu_rmap *rx_cpu_rmap; 2031 #endif 2032 struct hlist_node index_hlist; 2033 2034 /* 2035 * Cache lines mostly used on transmit path 2036 */ 2037 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2038 unsigned int num_tx_queues; 2039 unsigned int real_num_tx_queues; 2040 struct Qdisc *qdisc; 2041 unsigned int tx_queue_len; 2042 spinlock_t tx_global_lock; 2043 2044 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2045 2046 #ifdef CONFIG_XPS 2047 struct xps_dev_maps __rcu *xps_cpus_map; 2048 struct xps_dev_maps __rcu *xps_rxqs_map; 2049 #endif 2050 #ifdef CONFIG_NET_CLS_ACT 2051 struct mini_Qdisc __rcu *miniq_egress; 2052 #endif 2053 2054 #ifdef CONFIG_NET_SCHED 2055 DECLARE_HASHTABLE (qdisc_hash, 4); 2056 #endif 2057 /* These may be needed for future network-power-down code. */ 2058 struct timer_list watchdog_timer; 2059 int watchdog_timeo; 2060 2061 u32 proto_down_reason; 2062 2063 struct list_head todo_list; 2064 int __percpu *pcpu_refcnt; 2065 2066 struct list_head link_watch_list; 2067 2068 enum { NETREG_UNINITIALIZED=0, 2069 NETREG_REGISTERED, /* completed register_netdevice */ 2070 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2071 NETREG_UNREGISTERED, /* completed unregister todo */ 2072 NETREG_RELEASED, /* called free_netdev */ 2073 NETREG_DUMMY, /* dummy device for NAPI poll */ 2074 } reg_state:8; 2075 2076 bool dismantle; 2077 2078 enum { 2079 RTNL_LINK_INITIALIZED, 2080 RTNL_LINK_INITIALIZING, 2081 } rtnl_link_state:16; 2082 2083 bool needs_free_netdev; 2084 void (*priv_destructor)(struct net_device *dev); 2085 2086 #ifdef CONFIG_NETPOLL 2087 struct netpoll_info __rcu *npinfo; 2088 #endif 2089 2090 possible_net_t nd_net; 2091 2092 /* mid-layer private */ 2093 union { 2094 void *ml_priv; 2095 struct pcpu_lstats __percpu *lstats; 2096 struct pcpu_sw_netstats __percpu *tstats; 2097 struct pcpu_dstats __percpu *dstats; 2098 }; 2099 2100 #if IS_ENABLED(CONFIG_GARP) 2101 struct garp_port __rcu *garp_port; 2102 #endif 2103 #if IS_ENABLED(CONFIG_MRP) 2104 struct mrp_port __rcu *mrp_port; 2105 #endif 2106 2107 struct device dev; 2108 const struct attribute_group *sysfs_groups[4]; 2109 const struct attribute_group *sysfs_rx_queue_group; 2110 2111 const struct rtnl_link_ops *rtnl_link_ops; 2112 2113 /* for setting kernel sock attribute on TCP connection setup */ 2114 #define GSO_MAX_SIZE 65536 2115 unsigned int gso_max_size; 2116 #define GSO_MAX_SEGS 65535 2117 u16 gso_max_segs; 2118 2119 #ifdef CONFIG_DCB 2120 const struct dcbnl_rtnl_ops *dcbnl_ops; 2121 #endif 2122 s16 num_tc; 2123 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2124 u8 prio_tc_map[TC_BITMASK + 1]; 2125 2126 #if IS_ENABLED(CONFIG_FCOE) 2127 unsigned int fcoe_ddp_xid; 2128 #endif 2129 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2130 struct netprio_map __rcu *priomap; 2131 #endif 2132 struct phy_device *phydev; 2133 struct sfp_bus *sfp_bus; 2134 struct lock_class_key *qdisc_tx_busylock; 2135 struct lock_class_key *qdisc_running_key; 2136 bool proto_down; 2137 unsigned wol_enabled:1; 2138 2139 struct list_head net_notifier_list; 2140 2141 #if IS_ENABLED(CONFIG_MACSEC) 2142 /* MACsec management functions */ 2143 const struct macsec_ops *macsec_ops; 2144 #endif 2145 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2146 struct udp_tunnel_nic *udp_tunnel_nic; 2147 2148 /* protected by rtnl_lock */ 2149 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2150 }; 2151 #define to_net_dev(d) container_of(d, struct net_device, dev) 2152 2153 static inline bool netif_elide_gro(const struct net_device *dev) 2154 { 2155 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2156 return true; 2157 return false; 2158 } 2159 2160 #define NETDEV_ALIGN 32 2161 2162 static inline 2163 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2164 { 2165 return dev->prio_tc_map[prio & TC_BITMASK]; 2166 } 2167 2168 static inline 2169 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2170 { 2171 if (tc >= dev->num_tc) 2172 return -EINVAL; 2173 2174 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2175 return 0; 2176 } 2177 2178 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2179 void netdev_reset_tc(struct net_device *dev); 2180 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2181 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2182 2183 static inline 2184 int netdev_get_num_tc(struct net_device *dev) 2185 { 2186 return dev->num_tc; 2187 } 2188 2189 static inline void net_prefetch(void *p) 2190 { 2191 prefetch(p); 2192 #if L1_CACHE_BYTES < 128 2193 prefetch((u8 *)p + L1_CACHE_BYTES); 2194 #endif 2195 } 2196 2197 static inline void net_prefetchw(void *p) 2198 { 2199 prefetchw(p); 2200 #if L1_CACHE_BYTES < 128 2201 prefetchw((u8 *)p + L1_CACHE_BYTES); 2202 #endif 2203 } 2204 2205 void netdev_unbind_sb_channel(struct net_device *dev, 2206 struct net_device *sb_dev); 2207 int netdev_bind_sb_channel_queue(struct net_device *dev, 2208 struct net_device *sb_dev, 2209 u8 tc, u16 count, u16 offset); 2210 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2211 static inline int netdev_get_sb_channel(struct net_device *dev) 2212 { 2213 return max_t(int, -dev->num_tc, 0); 2214 } 2215 2216 static inline 2217 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2218 unsigned int index) 2219 { 2220 return &dev->_tx[index]; 2221 } 2222 2223 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2224 const struct sk_buff *skb) 2225 { 2226 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2227 } 2228 2229 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2230 void (*f)(struct net_device *, 2231 struct netdev_queue *, 2232 void *), 2233 void *arg) 2234 { 2235 unsigned int i; 2236 2237 for (i = 0; i < dev->num_tx_queues; i++) 2238 f(dev, &dev->_tx[i], arg); 2239 } 2240 2241 #define netdev_lockdep_set_classes(dev) \ 2242 { \ 2243 static struct lock_class_key qdisc_tx_busylock_key; \ 2244 static struct lock_class_key qdisc_running_key; \ 2245 static struct lock_class_key qdisc_xmit_lock_key; \ 2246 static struct lock_class_key dev_addr_list_lock_key; \ 2247 unsigned int i; \ 2248 \ 2249 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2250 (dev)->qdisc_running_key = &qdisc_running_key; \ 2251 lockdep_set_class(&(dev)->addr_list_lock, \ 2252 &dev_addr_list_lock_key); \ 2253 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2254 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2255 &qdisc_xmit_lock_key); \ 2256 } 2257 2258 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2259 struct net_device *sb_dev); 2260 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2261 struct sk_buff *skb, 2262 struct net_device *sb_dev); 2263 2264 /* returns the headroom that the master device needs to take in account 2265 * when forwarding to this dev 2266 */ 2267 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2268 { 2269 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2270 } 2271 2272 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2273 { 2274 if (dev->netdev_ops->ndo_set_rx_headroom) 2275 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2276 } 2277 2278 /* set the device rx headroom to the dev's default */ 2279 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2280 { 2281 netdev_set_rx_headroom(dev, -1); 2282 } 2283 2284 /* 2285 * Net namespace inlines 2286 */ 2287 static inline 2288 struct net *dev_net(const struct net_device *dev) 2289 { 2290 return read_pnet(&dev->nd_net); 2291 } 2292 2293 static inline 2294 void dev_net_set(struct net_device *dev, struct net *net) 2295 { 2296 write_pnet(&dev->nd_net, net); 2297 } 2298 2299 /** 2300 * netdev_priv - access network device private data 2301 * @dev: network device 2302 * 2303 * Get network device private data 2304 */ 2305 static inline void *netdev_priv(const struct net_device *dev) 2306 { 2307 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2308 } 2309 2310 /* Set the sysfs physical device reference for the network logical device 2311 * if set prior to registration will cause a symlink during initialization. 2312 */ 2313 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2314 2315 /* Set the sysfs device type for the network logical device to allow 2316 * fine-grained identification of different network device types. For 2317 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2318 */ 2319 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2320 2321 /* Default NAPI poll() weight 2322 * Device drivers are strongly advised to not use bigger value 2323 */ 2324 #define NAPI_POLL_WEIGHT 64 2325 2326 /** 2327 * netif_napi_add - initialize a NAPI context 2328 * @dev: network device 2329 * @napi: NAPI context 2330 * @poll: polling function 2331 * @weight: default weight 2332 * 2333 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2334 * *any* of the other NAPI-related functions. 2335 */ 2336 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2337 int (*poll)(struct napi_struct *, int), int weight); 2338 2339 /** 2340 * netif_tx_napi_add - initialize a NAPI context 2341 * @dev: network device 2342 * @napi: NAPI context 2343 * @poll: polling function 2344 * @weight: default weight 2345 * 2346 * This variant of netif_napi_add() should be used from drivers using NAPI 2347 * to exclusively poll a TX queue. 2348 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2349 */ 2350 static inline void netif_tx_napi_add(struct net_device *dev, 2351 struct napi_struct *napi, 2352 int (*poll)(struct napi_struct *, int), 2353 int weight) 2354 { 2355 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2356 netif_napi_add(dev, napi, poll, weight); 2357 } 2358 2359 /** 2360 * __netif_napi_del - remove a NAPI context 2361 * @napi: NAPI context 2362 * 2363 * Warning: caller must observe RCU grace period before freeing memory 2364 * containing @napi. Drivers might want to call this helper to combine 2365 * all the needed RCU grace periods into a single one. 2366 */ 2367 void __netif_napi_del(struct napi_struct *napi); 2368 2369 /** 2370 * netif_napi_del - remove a NAPI context 2371 * @napi: NAPI context 2372 * 2373 * netif_napi_del() removes a NAPI context from the network device NAPI list 2374 */ 2375 static inline void netif_napi_del(struct napi_struct *napi) 2376 { 2377 __netif_napi_del(napi); 2378 synchronize_net(); 2379 } 2380 2381 struct napi_gro_cb { 2382 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2383 void *frag0; 2384 2385 /* Length of frag0. */ 2386 unsigned int frag0_len; 2387 2388 /* This indicates where we are processing relative to skb->data. */ 2389 int data_offset; 2390 2391 /* This is non-zero if the packet cannot be merged with the new skb. */ 2392 u16 flush; 2393 2394 /* Save the IP ID here and check when we get to the transport layer */ 2395 u16 flush_id; 2396 2397 /* Number of segments aggregated. */ 2398 u16 count; 2399 2400 /* Start offset for remote checksum offload */ 2401 u16 gro_remcsum_start; 2402 2403 /* jiffies when first packet was created/queued */ 2404 unsigned long age; 2405 2406 /* Used in ipv6_gro_receive() and foo-over-udp */ 2407 u16 proto; 2408 2409 /* This is non-zero if the packet may be of the same flow. */ 2410 u8 same_flow:1; 2411 2412 /* Used in tunnel GRO receive */ 2413 u8 encap_mark:1; 2414 2415 /* GRO checksum is valid */ 2416 u8 csum_valid:1; 2417 2418 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2419 u8 csum_cnt:3; 2420 2421 /* Free the skb? */ 2422 u8 free:2; 2423 #define NAPI_GRO_FREE 1 2424 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2425 2426 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2427 u8 is_ipv6:1; 2428 2429 /* Used in GRE, set in fou/gue_gro_receive */ 2430 u8 is_fou:1; 2431 2432 /* Used to determine if flush_id can be ignored */ 2433 u8 is_atomic:1; 2434 2435 /* Number of gro_receive callbacks this packet already went through */ 2436 u8 recursion_counter:4; 2437 2438 /* GRO is done by frag_list pointer chaining. */ 2439 u8 is_flist:1; 2440 2441 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2442 __wsum csum; 2443 2444 /* used in skb_gro_receive() slow path */ 2445 struct sk_buff *last; 2446 }; 2447 2448 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2449 2450 #define GRO_RECURSION_LIMIT 15 2451 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2452 { 2453 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2454 } 2455 2456 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2457 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2458 struct list_head *head, 2459 struct sk_buff *skb) 2460 { 2461 if (unlikely(gro_recursion_inc_test(skb))) { 2462 NAPI_GRO_CB(skb)->flush |= 1; 2463 return NULL; 2464 } 2465 2466 return cb(head, skb); 2467 } 2468 2469 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2470 struct sk_buff *); 2471 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2472 struct sock *sk, 2473 struct list_head *head, 2474 struct sk_buff *skb) 2475 { 2476 if (unlikely(gro_recursion_inc_test(skb))) { 2477 NAPI_GRO_CB(skb)->flush |= 1; 2478 return NULL; 2479 } 2480 2481 return cb(sk, head, skb); 2482 } 2483 2484 struct packet_type { 2485 __be16 type; /* This is really htons(ether_type). */ 2486 bool ignore_outgoing; 2487 struct net_device *dev; /* NULL is wildcarded here */ 2488 int (*func) (struct sk_buff *, 2489 struct net_device *, 2490 struct packet_type *, 2491 struct net_device *); 2492 void (*list_func) (struct list_head *, 2493 struct packet_type *, 2494 struct net_device *); 2495 bool (*id_match)(struct packet_type *ptype, 2496 struct sock *sk); 2497 void *af_packet_priv; 2498 struct list_head list; 2499 }; 2500 2501 struct offload_callbacks { 2502 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2503 netdev_features_t features); 2504 struct sk_buff *(*gro_receive)(struct list_head *head, 2505 struct sk_buff *skb); 2506 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2507 }; 2508 2509 struct packet_offload { 2510 __be16 type; /* This is really htons(ether_type). */ 2511 u16 priority; 2512 struct offload_callbacks callbacks; 2513 struct list_head list; 2514 }; 2515 2516 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2517 struct pcpu_sw_netstats { 2518 u64 rx_packets; 2519 u64 rx_bytes; 2520 u64 tx_packets; 2521 u64 tx_bytes; 2522 struct u64_stats_sync syncp; 2523 } __aligned(4 * sizeof(u64)); 2524 2525 struct pcpu_lstats { 2526 u64_stats_t packets; 2527 u64_stats_t bytes; 2528 struct u64_stats_sync syncp; 2529 } __aligned(2 * sizeof(u64)); 2530 2531 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2532 2533 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2534 { 2535 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2536 2537 u64_stats_update_begin(&lstats->syncp); 2538 u64_stats_add(&lstats->bytes, len); 2539 u64_stats_inc(&lstats->packets); 2540 u64_stats_update_end(&lstats->syncp); 2541 } 2542 2543 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2544 ({ \ 2545 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2546 if (pcpu_stats) { \ 2547 int __cpu; \ 2548 for_each_possible_cpu(__cpu) { \ 2549 typeof(type) *stat; \ 2550 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2551 u64_stats_init(&stat->syncp); \ 2552 } \ 2553 } \ 2554 pcpu_stats; \ 2555 }) 2556 2557 #define netdev_alloc_pcpu_stats(type) \ 2558 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2559 2560 enum netdev_lag_tx_type { 2561 NETDEV_LAG_TX_TYPE_UNKNOWN, 2562 NETDEV_LAG_TX_TYPE_RANDOM, 2563 NETDEV_LAG_TX_TYPE_BROADCAST, 2564 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2565 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2566 NETDEV_LAG_TX_TYPE_HASH, 2567 }; 2568 2569 enum netdev_lag_hash { 2570 NETDEV_LAG_HASH_NONE, 2571 NETDEV_LAG_HASH_L2, 2572 NETDEV_LAG_HASH_L34, 2573 NETDEV_LAG_HASH_L23, 2574 NETDEV_LAG_HASH_E23, 2575 NETDEV_LAG_HASH_E34, 2576 NETDEV_LAG_HASH_UNKNOWN, 2577 }; 2578 2579 struct netdev_lag_upper_info { 2580 enum netdev_lag_tx_type tx_type; 2581 enum netdev_lag_hash hash_type; 2582 }; 2583 2584 struct netdev_lag_lower_state_info { 2585 u8 link_up : 1, 2586 tx_enabled : 1; 2587 }; 2588 2589 #include <linux/notifier.h> 2590 2591 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2592 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2593 * adding new types. 2594 */ 2595 enum netdev_cmd { 2596 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2597 NETDEV_DOWN, 2598 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2599 detected a hardware crash and restarted 2600 - we can use this eg to kick tcp sessions 2601 once done */ 2602 NETDEV_CHANGE, /* Notify device state change */ 2603 NETDEV_REGISTER, 2604 NETDEV_UNREGISTER, 2605 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2606 NETDEV_CHANGEADDR, /* notify after the address change */ 2607 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2608 NETDEV_GOING_DOWN, 2609 NETDEV_CHANGENAME, 2610 NETDEV_FEAT_CHANGE, 2611 NETDEV_BONDING_FAILOVER, 2612 NETDEV_PRE_UP, 2613 NETDEV_PRE_TYPE_CHANGE, 2614 NETDEV_POST_TYPE_CHANGE, 2615 NETDEV_POST_INIT, 2616 NETDEV_RELEASE, 2617 NETDEV_NOTIFY_PEERS, 2618 NETDEV_JOIN, 2619 NETDEV_CHANGEUPPER, 2620 NETDEV_RESEND_IGMP, 2621 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2622 NETDEV_CHANGEINFODATA, 2623 NETDEV_BONDING_INFO, 2624 NETDEV_PRECHANGEUPPER, 2625 NETDEV_CHANGELOWERSTATE, 2626 NETDEV_UDP_TUNNEL_PUSH_INFO, 2627 NETDEV_UDP_TUNNEL_DROP_INFO, 2628 NETDEV_CHANGE_TX_QUEUE_LEN, 2629 NETDEV_CVLAN_FILTER_PUSH_INFO, 2630 NETDEV_CVLAN_FILTER_DROP_INFO, 2631 NETDEV_SVLAN_FILTER_PUSH_INFO, 2632 NETDEV_SVLAN_FILTER_DROP_INFO, 2633 }; 2634 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2635 2636 int register_netdevice_notifier(struct notifier_block *nb); 2637 int unregister_netdevice_notifier(struct notifier_block *nb); 2638 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2639 int unregister_netdevice_notifier_net(struct net *net, 2640 struct notifier_block *nb); 2641 int register_netdevice_notifier_dev_net(struct net_device *dev, 2642 struct notifier_block *nb, 2643 struct netdev_net_notifier *nn); 2644 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2645 struct notifier_block *nb, 2646 struct netdev_net_notifier *nn); 2647 2648 struct netdev_notifier_info { 2649 struct net_device *dev; 2650 struct netlink_ext_ack *extack; 2651 }; 2652 2653 struct netdev_notifier_info_ext { 2654 struct netdev_notifier_info info; /* must be first */ 2655 union { 2656 u32 mtu; 2657 } ext; 2658 }; 2659 2660 struct netdev_notifier_change_info { 2661 struct netdev_notifier_info info; /* must be first */ 2662 unsigned int flags_changed; 2663 }; 2664 2665 struct netdev_notifier_changeupper_info { 2666 struct netdev_notifier_info info; /* must be first */ 2667 struct net_device *upper_dev; /* new upper dev */ 2668 bool master; /* is upper dev master */ 2669 bool linking; /* is the notification for link or unlink */ 2670 void *upper_info; /* upper dev info */ 2671 }; 2672 2673 struct netdev_notifier_changelowerstate_info { 2674 struct netdev_notifier_info info; /* must be first */ 2675 void *lower_state_info; /* is lower dev state */ 2676 }; 2677 2678 struct netdev_notifier_pre_changeaddr_info { 2679 struct netdev_notifier_info info; /* must be first */ 2680 const unsigned char *dev_addr; 2681 }; 2682 2683 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2684 struct net_device *dev) 2685 { 2686 info->dev = dev; 2687 info->extack = NULL; 2688 } 2689 2690 static inline struct net_device * 2691 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2692 { 2693 return info->dev; 2694 } 2695 2696 static inline struct netlink_ext_ack * 2697 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2698 { 2699 return info->extack; 2700 } 2701 2702 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2703 2704 2705 extern rwlock_t dev_base_lock; /* Device list lock */ 2706 2707 #define for_each_netdev(net, d) \ 2708 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2709 #define for_each_netdev_reverse(net, d) \ 2710 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2711 #define for_each_netdev_rcu(net, d) \ 2712 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2713 #define for_each_netdev_safe(net, d, n) \ 2714 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2715 #define for_each_netdev_continue(net, d) \ 2716 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2717 #define for_each_netdev_continue_reverse(net, d) \ 2718 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2719 dev_list) 2720 #define for_each_netdev_continue_rcu(net, d) \ 2721 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2722 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2723 for_each_netdev_rcu(&init_net, slave) \ 2724 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2725 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2726 2727 static inline struct net_device *next_net_device(struct net_device *dev) 2728 { 2729 struct list_head *lh; 2730 struct net *net; 2731 2732 net = dev_net(dev); 2733 lh = dev->dev_list.next; 2734 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2735 } 2736 2737 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2738 { 2739 struct list_head *lh; 2740 struct net *net; 2741 2742 net = dev_net(dev); 2743 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2744 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2745 } 2746 2747 static inline struct net_device *first_net_device(struct net *net) 2748 { 2749 return list_empty(&net->dev_base_head) ? NULL : 2750 net_device_entry(net->dev_base_head.next); 2751 } 2752 2753 static inline struct net_device *first_net_device_rcu(struct net *net) 2754 { 2755 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2756 2757 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2758 } 2759 2760 int netdev_boot_setup_check(struct net_device *dev); 2761 unsigned long netdev_boot_base(const char *prefix, int unit); 2762 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2763 const char *hwaddr); 2764 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2765 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2766 void dev_add_pack(struct packet_type *pt); 2767 void dev_remove_pack(struct packet_type *pt); 2768 void __dev_remove_pack(struct packet_type *pt); 2769 void dev_add_offload(struct packet_offload *po); 2770 void dev_remove_offload(struct packet_offload *po); 2771 2772 int dev_get_iflink(const struct net_device *dev); 2773 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2774 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2775 unsigned short mask); 2776 struct net_device *dev_get_by_name(struct net *net, const char *name); 2777 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2778 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2779 int dev_alloc_name(struct net_device *dev, const char *name); 2780 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2781 void dev_close(struct net_device *dev); 2782 void dev_close_many(struct list_head *head, bool unlink); 2783 void dev_disable_lro(struct net_device *dev); 2784 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2785 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2786 struct net_device *sb_dev); 2787 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2788 struct net_device *sb_dev); 2789 int dev_queue_xmit(struct sk_buff *skb); 2790 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2791 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2792 int register_netdevice(struct net_device *dev); 2793 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2794 void unregister_netdevice_many(struct list_head *head); 2795 static inline void unregister_netdevice(struct net_device *dev) 2796 { 2797 unregister_netdevice_queue(dev, NULL); 2798 } 2799 2800 int netdev_refcnt_read(const struct net_device *dev); 2801 void free_netdev(struct net_device *dev); 2802 void netdev_freemem(struct net_device *dev); 2803 int init_dummy_netdev(struct net_device *dev); 2804 2805 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2806 struct sk_buff *skb, 2807 bool all_slaves); 2808 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2809 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2810 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2811 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2812 int netdev_get_name(struct net *net, char *name, int ifindex); 2813 int dev_restart(struct net_device *dev); 2814 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2815 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2816 2817 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2818 { 2819 return NAPI_GRO_CB(skb)->data_offset; 2820 } 2821 2822 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2823 { 2824 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2825 } 2826 2827 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2828 { 2829 NAPI_GRO_CB(skb)->data_offset += len; 2830 } 2831 2832 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2833 unsigned int offset) 2834 { 2835 return NAPI_GRO_CB(skb)->frag0 + offset; 2836 } 2837 2838 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2839 { 2840 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2841 } 2842 2843 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2844 { 2845 NAPI_GRO_CB(skb)->frag0 = NULL; 2846 NAPI_GRO_CB(skb)->frag0_len = 0; 2847 } 2848 2849 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2850 unsigned int offset) 2851 { 2852 if (!pskb_may_pull(skb, hlen)) 2853 return NULL; 2854 2855 skb_gro_frag0_invalidate(skb); 2856 return skb->data + offset; 2857 } 2858 2859 static inline void *skb_gro_network_header(struct sk_buff *skb) 2860 { 2861 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2862 skb_network_offset(skb); 2863 } 2864 2865 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2866 const void *start, unsigned int len) 2867 { 2868 if (NAPI_GRO_CB(skb)->csum_valid) 2869 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2870 csum_partial(start, len, 0)); 2871 } 2872 2873 /* GRO checksum functions. These are logical equivalents of the normal 2874 * checksum functions (in skbuff.h) except that they operate on the GRO 2875 * offsets and fields in sk_buff. 2876 */ 2877 2878 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2879 2880 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2881 { 2882 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2883 } 2884 2885 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2886 bool zero_okay, 2887 __sum16 check) 2888 { 2889 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2890 skb_checksum_start_offset(skb) < 2891 skb_gro_offset(skb)) && 2892 !skb_at_gro_remcsum_start(skb) && 2893 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2894 (!zero_okay || check)); 2895 } 2896 2897 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2898 __wsum psum) 2899 { 2900 if (NAPI_GRO_CB(skb)->csum_valid && 2901 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2902 return 0; 2903 2904 NAPI_GRO_CB(skb)->csum = psum; 2905 2906 return __skb_gro_checksum_complete(skb); 2907 } 2908 2909 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2910 { 2911 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2912 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2913 NAPI_GRO_CB(skb)->csum_cnt--; 2914 } else { 2915 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2916 * verified a new top level checksum or an encapsulated one 2917 * during GRO. This saves work if we fallback to normal path. 2918 */ 2919 __skb_incr_checksum_unnecessary(skb); 2920 } 2921 } 2922 2923 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2924 compute_pseudo) \ 2925 ({ \ 2926 __sum16 __ret = 0; \ 2927 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2928 __ret = __skb_gro_checksum_validate_complete(skb, \ 2929 compute_pseudo(skb, proto)); \ 2930 if (!__ret) \ 2931 skb_gro_incr_csum_unnecessary(skb); \ 2932 __ret; \ 2933 }) 2934 2935 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2936 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2937 2938 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2939 compute_pseudo) \ 2940 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2941 2942 #define skb_gro_checksum_simple_validate(skb) \ 2943 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2944 2945 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2946 { 2947 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2948 !NAPI_GRO_CB(skb)->csum_valid); 2949 } 2950 2951 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2952 __wsum pseudo) 2953 { 2954 NAPI_GRO_CB(skb)->csum = ~pseudo; 2955 NAPI_GRO_CB(skb)->csum_valid = 1; 2956 } 2957 2958 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2959 do { \ 2960 if (__skb_gro_checksum_convert_check(skb)) \ 2961 __skb_gro_checksum_convert(skb, \ 2962 compute_pseudo(skb, proto)); \ 2963 } while (0) 2964 2965 struct gro_remcsum { 2966 int offset; 2967 __wsum delta; 2968 }; 2969 2970 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2971 { 2972 grc->offset = 0; 2973 grc->delta = 0; 2974 } 2975 2976 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2977 unsigned int off, size_t hdrlen, 2978 int start, int offset, 2979 struct gro_remcsum *grc, 2980 bool nopartial) 2981 { 2982 __wsum delta; 2983 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2984 2985 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2986 2987 if (!nopartial) { 2988 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2989 return ptr; 2990 } 2991 2992 ptr = skb_gro_header_fast(skb, off); 2993 if (skb_gro_header_hard(skb, off + plen)) { 2994 ptr = skb_gro_header_slow(skb, off + plen, off); 2995 if (!ptr) 2996 return NULL; 2997 } 2998 2999 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3000 start, offset); 3001 3002 /* Adjust skb->csum since we changed the packet */ 3003 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3004 3005 grc->offset = off + hdrlen + offset; 3006 grc->delta = delta; 3007 3008 return ptr; 3009 } 3010 3011 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3012 struct gro_remcsum *grc) 3013 { 3014 void *ptr; 3015 size_t plen = grc->offset + sizeof(u16); 3016 3017 if (!grc->delta) 3018 return; 3019 3020 ptr = skb_gro_header_fast(skb, grc->offset); 3021 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3022 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3023 if (!ptr) 3024 return; 3025 } 3026 3027 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3028 } 3029 3030 #ifdef CONFIG_XFRM_OFFLOAD 3031 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3032 { 3033 if (PTR_ERR(pp) != -EINPROGRESS) 3034 NAPI_GRO_CB(skb)->flush |= flush; 3035 } 3036 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3037 struct sk_buff *pp, 3038 int flush, 3039 struct gro_remcsum *grc) 3040 { 3041 if (PTR_ERR(pp) != -EINPROGRESS) { 3042 NAPI_GRO_CB(skb)->flush |= flush; 3043 skb_gro_remcsum_cleanup(skb, grc); 3044 skb->remcsum_offload = 0; 3045 } 3046 } 3047 #else 3048 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3049 { 3050 NAPI_GRO_CB(skb)->flush |= flush; 3051 } 3052 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3053 struct sk_buff *pp, 3054 int flush, 3055 struct gro_remcsum *grc) 3056 { 3057 NAPI_GRO_CB(skb)->flush |= flush; 3058 skb_gro_remcsum_cleanup(skb, grc); 3059 skb->remcsum_offload = 0; 3060 } 3061 #endif 3062 3063 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3064 unsigned short type, 3065 const void *daddr, const void *saddr, 3066 unsigned int len) 3067 { 3068 if (!dev->header_ops || !dev->header_ops->create) 3069 return 0; 3070 3071 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3072 } 3073 3074 static inline int dev_parse_header(const struct sk_buff *skb, 3075 unsigned char *haddr) 3076 { 3077 const struct net_device *dev = skb->dev; 3078 3079 if (!dev->header_ops || !dev->header_ops->parse) 3080 return 0; 3081 return dev->header_ops->parse(skb, haddr); 3082 } 3083 3084 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3085 { 3086 const struct net_device *dev = skb->dev; 3087 3088 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3089 return 0; 3090 return dev->header_ops->parse_protocol(skb); 3091 } 3092 3093 /* ll_header must have at least hard_header_len allocated */ 3094 static inline bool dev_validate_header(const struct net_device *dev, 3095 char *ll_header, int len) 3096 { 3097 if (likely(len >= dev->hard_header_len)) 3098 return true; 3099 if (len < dev->min_header_len) 3100 return false; 3101 3102 if (capable(CAP_SYS_RAWIO)) { 3103 memset(ll_header + len, 0, dev->hard_header_len - len); 3104 return true; 3105 } 3106 3107 if (dev->header_ops && dev->header_ops->validate) 3108 return dev->header_ops->validate(ll_header, len); 3109 3110 return false; 3111 } 3112 3113 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3114 int len, int size); 3115 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3116 static inline int unregister_gifconf(unsigned int family) 3117 { 3118 return register_gifconf(family, NULL); 3119 } 3120 3121 #ifdef CONFIG_NET_FLOW_LIMIT 3122 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3123 struct sd_flow_limit { 3124 u64 count; 3125 unsigned int num_buckets; 3126 unsigned int history_head; 3127 u16 history[FLOW_LIMIT_HISTORY]; 3128 u8 buckets[]; 3129 }; 3130 3131 extern int netdev_flow_limit_table_len; 3132 #endif /* CONFIG_NET_FLOW_LIMIT */ 3133 3134 /* 3135 * Incoming packets are placed on per-CPU queues 3136 */ 3137 struct softnet_data { 3138 struct list_head poll_list; 3139 struct sk_buff_head process_queue; 3140 3141 /* stats */ 3142 unsigned int processed; 3143 unsigned int time_squeeze; 3144 unsigned int received_rps; 3145 #ifdef CONFIG_RPS 3146 struct softnet_data *rps_ipi_list; 3147 #endif 3148 #ifdef CONFIG_NET_FLOW_LIMIT 3149 struct sd_flow_limit __rcu *flow_limit; 3150 #endif 3151 struct Qdisc *output_queue; 3152 struct Qdisc **output_queue_tailp; 3153 struct sk_buff *completion_queue; 3154 #ifdef CONFIG_XFRM_OFFLOAD 3155 struct sk_buff_head xfrm_backlog; 3156 #endif 3157 /* written and read only by owning cpu: */ 3158 struct { 3159 u16 recursion; 3160 u8 more; 3161 } xmit; 3162 #ifdef CONFIG_RPS 3163 /* input_queue_head should be written by cpu owning this struct, 3164 * and only read by other cpus. Worth using a cache line. 3165 */ 3166 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3167 3168 /* Elements below can be accessed between CPUs for RPS/RFS */ 3169 call_single_data_t csd ____cacheline_aligned_in_smp; 3170 struct softnet_data *rps_ipi_next; 3171 unsigned int cpu; 3172 unsigned int input_queue_tail; 3173 #endif 3174 unsigned int dropped; 3175 struct sk_buff_head input_pkt_queue; 3176 struct napi_struct backlog; 3177 3178 }; 3179 3180 static inline void input_queue_head_incr(struct softnet_data *sd) 3181 { 3182 #ifdef CONFIG_RPS 3183 sd->input_queue_head++; 3184 #endif 3185 } 3186 3187 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3188 unsigned int *qtail) 3189 { 3190 #ifdef CONFIG_RPS 3191 *qtail = ++sd->input_queue_tail; 3192 #endif 3193 } 3194 3195 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3196 3197 static inline int dev_recursion_level(void) 3198 { 3199 return this_cpu_read(softnet_data.xmit.recursion); 3200 } 3201 3202 #define XMIT_RECURSION_LIMIT 8 3203 static inline bool dev_xmit_recursion(void) 3204 { 3205 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3206 XMIT_RECURSION_LIMIT); 3207 } 3208 3209 static inline void dev_xmit_recursion_inc(void) 3210 { 3211 __this_cpu_inc(softnet_data.xmit.recursion); 3212 } 3213 3214 static inline void dev_xmit_recursion_dec(void) 3215 { 3216 __this_cpu_dec(softnet_data.xmit.recursion); 3217 } 3218 3219 void __netif_schedule(struct Qdisc *q); 3220 void netif_schedule_queue(struct netdev_queue *txq); 3221 3222 static inline void netif_tx_schedule_all(struct net_device *dev) 3223 { 3224 unsigned int i; 3225 3226 for (i = 0; i < dev->num_tx_queues; i++) 3227 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3228 } 3229 3230 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3231 { 3232 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3233 } 3234 3235 /** 3236 * netif_start_queue - allow transmit 3237 * @dev: network device 3238 * 3239 * Allow upper layers to call the device hard_start_xmit routine. 3240 */ 3241 static inline void netif_start_queue(struct net_device *dev) 3242 { 3243 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3244 } 3245 3246 static inline void netif_tx_start_all_queues(struct net_device *dev) 3247 { 3248 unsigned int i; 3249 3250 for (i = 0; i < dev->num_tx_queues; i++) { 3251 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3252 netif_tx_start_queue(txq); 3253 } 3254 } 3255 3256 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3257 3258 /** 3259 * netif_wake_queue - restart transmit 3260 * @dev: network device 3261 * 3262 * Allow upper layers to call the device hard_start_xmit routine. 3263 * Used for flow control when transmit resources are available. 3264 */ 3265 static inline void netif_wake_queue(struct net_device *dev) 3266 { 3267 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3268 } 3269 3270 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3271 { 3272 unsigned int i; 3273 3274 for (i = 0; i < dev->num_tx_queues; i++) { 3275 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3276 netif_tx_wake_queue(txq); 3277 } 3278 } 3279 3280 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3281 { 3282 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3283 } 3284 3285 /** 3286 * netif_stop_queue - stop transmitted packets 3287 * @dev: network device 3288 * 3289 * Stop upper layers calling the device hard_start_xmit routine. 3290 * Used for flow control when transmit resources are unavailable. 3291 */ 3292 static inline void netif_stop_queue(struct net_device *dev) 3293 { 3294 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3295 } 3296 3297 void netif_tx_stop_all_queues(struct net_device *dev); 3298 3299 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3300 { 3301 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3302 } 3303 3304 /** 3305 * netif_queue_stopped - test if transmit queue is flowblocked 3306 * @dev: network device 3307 * 3308 * Test if transmit queue on device is currently unable to send. 3309 */ 3310 static inline bool netif_queue_stopped(const struct net_device *dev) 3311 { 3312 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3313 } 3314 3315 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3316 { 3317 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3318 } 3319 3320 static inline bool 3321 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3322 { 3323 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3324 } 3325 3326 static inline bool 3327 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3328 { 3329 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3330 } 3331 3332 /** 3333 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3334 * @dev_queue: pointer to transmit queue 3335 * 3336 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3337 * to give appropriate hint to the CPU. 3338 */ 3339 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3340 { 3341 #ifdef CONFIG_BQL 3342 prefetchw(&dev_queue->dql.num_queued); 3343 #endif 3344 } 3345 3346 /** 3347 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3348 * @dev_queue: pointer to transmit queue 3349 * 3350 * BQL enabled drivers might use this helper in their TX completion path, 3351 * to give appropriate hint to the CPU. 3352 */ 3353 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3354 { 3355 #ifdef CONFIG_BQL 3356 prefetchw(&dev_queue->dql.limit); 3357 #endif 3358 } 3359 3360 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3361 unsigned int bytes) 3362 { 3363 #ifdef CONFIG_BQL 3364 dql_queued(&dev_queue->dql, bytes); 3365 3366 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3367 return; 3368 3369 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3370 3371 /* 3372 * The XOFF flag must be set before checking the dql_avail below, 3373 * because in netdev_tx_completed_queue we update the dql_completed 3374 * before checking the XOFF flag. 3375 */ 3376 smp_mb(); 3377 3378 /* check again in case another CPU has just made room avail */ 3379 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3380 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3381 #endif 3382 } 3383 3384 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3385 * that they should not test BQL status themselves. 3386 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3387 * skb of a batch. 3388 * Returns true if the doorbell must be used to kick the NIC. 3389 */ 3390 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3391 unsigned int bytes, 3392 bool xmit_more) 3393 { 3394 if (xmit_more) { 3395 #ifdef CONFIG_BQL 3396 dql_queued(&dev_queue->dql, bytes); 3397 #endif 3398 return netif_tx_queue_stopped(dev_queue); 3399 } 3400 netdev_tx_sent_queue(dev_queue, bytes); 3401 return true; 3402 } 3403 3404 /** 3405 * netdev_sent_queue - report the number of bytes queued to hardware 3406 * @dev: network device 3407 * @bytes: number of bytes queued to the hardware device queue 3408 * 3409 * Report the number of bytes queued for sending/completion to the network 3410 * device hardware queue. @bytes should be a good approximation and should 3411 * exactly match netdev_completed_queue() @bytes 3412 */ 3413 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3414 { 3415 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3416 } 3417 3418 static inline bool __netdev_sent_queue(struct net_device *dev, 3419 unsigned int bytes, 3420 bool xmit_more) 3421 { 3422 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3423 xmit_more); 3424 } 3425 3426 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3427 unsigned int pkts, unsigned int bytes) 3428 { 3429 #ifdef CONFIG_BQL 3430 if (unlikely(!bytes)) 3431 return; 3432 3433 dql_completed(&dev_queue->dql, bytes); 3434 3435 /* 3436 * Without the memory barrier there is a small possiblity that 3437 * netdev_tx_sent_queue will miss the update and cause the queue to 3438 * be stopped forever 3439 */ 3440 smp_mb(); 3441 3442 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3443 return; 3444 3445 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3446 netif_schedule_queue(dev_queue); 3447 #endif 3448 } 3449 3450 /** 3451 * netdev_completed_queue - report bytes and packets completed by device 3452 * @dev: network device 3453 * @pkts: actual number of packets sent over the medium 3454 * @bytes: actual number of bytes sent over the medium 3455 * 3456 * Report the number of bytes and packets transmitted by the network device 3457 * hardware queue over the physical medium, @bytes must exactly match the 3458 * @bytes amount passed to netdev_sent_queue() 3459 */ 3460 static inline void netdev_completed_queue(struct net_device *dev, 3461 unsigned int pkts, unsigned int bytes) 3462 { 3463 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3464 } 3465 3466 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3467 { 3468 #ifdef CONFIG_BQL 3469 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3470 dql_reset(&q->dql); 3471 #endif 3472 } 3473 3474 /** 3475 * netdev_reset_queue - reset the packets and bytes count of a network device 3476 * @dev_queue: network device 3477 * 3478 * Reset the bytes and packet count of a network device and clear the 3479 * software flow control OFF bit for this network device 3480 */ 3481 static inline void netdev_reset_queue(struct net_device *dev_queue) 3482 { 3483 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3484 } 3485 3486 /** 3487 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3488 * @dev: network device 3489 * @queue_index: given tx queue index 3490 * 3491 * Returns 0 if given tx queue index >= number of device tx queues, 3492 * otherwise returns the originally passed tx queue index. 3493 */ 3494 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3495 { 3496 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3497 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3498 dev->name, queue_index, 3499 dev->real_num_tx_queues); 3500 return 0; 3501 } 3502 3503 return queue_index; 3504 } 3505 3506 /** 3507 * netif_running - test if up 3508 * @dev: network device 3509 * 3510 * Test if the device has been brought up. 3511 */ 3512 static inline bool netif_running(const struct net_device *dev) 3513 { 3514 return test_bit(__LINK_STATE_START, &dev->state); 3515 } 3516 3517 /* 3518 * Routines to manage the subqueues on a device. We only need start, 3519 * stop, and a check if it's stopped. All other device management is 3520 * done at the overall netdevice level. 3521 * Also test the device if we're multiqueue. 3522 */ 3523 3524 /** 3525 * netif_start_subqueue - allow sending packets on subqueue 3526 * @dev: network device 3527 * @queue_index: sub queue index 3528 * 3529 * Start individual transmit queue of a device with multiple transmit queues. 3530 */ 3531 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3532 { 3533 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3534 3535 netif_tx_start_queue(txq); 3536 } 3537 3538 /** 3539 * netif_stop_subqueue - stop sending packets on subqueue 3540 * @dev: network device 3541 * @queue_index: sub queue index 3542 * 3543 * Stop individual transmit queue of a device with multiple transmit queues. 3544 */ 3545 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3546 { 3547 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3548 netif_tx_stop_queue(txq); 3549 } 3550 3551 /** 3552 * netif_subqueue_stopped - test status of subqueue 3553 * @dev: network device 3554 * @queue_index: sub queue index 3555 * 3556 * Check individual transmit queue of a device with multiple transmit queues. 3557 */ 3558 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3559 u16 queue_index) 3560 { 3561 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3562 3563 return netif_tx_queue_stopped(txq); 3564 } 3565 3566 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3567 struct sk_buff *skb) 3568 { 3569 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3570 } 3571 3572 /** 3573 * netif_wake_subqueue - allow sending packets on subqueue 3574 * @dev: network device 3575 * @queue_index: sub queue index 3576 * 3577 * Resume individual transmit queue of a device with multiple transmit queues. 3578 */ 3579 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3580 { 3581 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3582 3583 netif_tx_wake_queue(txq); 3584 } 3585 3586 #ifdef CONFIG_XPS 3587 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3588 u16 index); 3589 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3590 u16 index, bool is_rxqs_map); 3591 3592 /** 3593 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3594 * @j: CPU/Rx queue index 3595 * @mask: bitmask of all cpus/rx queues 3596 * @nr_bits: number of bits in the bitmask 3597 * 3598 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3599 */ 3600 static inline bool netif_attr_test_mask(unsigned long j, 3601 const unsigned long *mask, 3602 unsigned int nr_bits) 3603 { 3604 cpu_max_bits_warn(j, nr_bits); 3605 return test_bit(j, mask); 3606 } 3607 3608 /** 3609 * netif_attr_test_online - Test for online CPU/Rx queue 3610 * @j: CPU/Rx queue index 3611 * @online_mask: bitmask for CPUs/Rx queues that are online 3612 * @nr_bits: number of bits in the bitmask 3613 * 3614 * Returns true if a CPU/Rx queue is online. 3615 */ 3616 static inline bool netif_attr_test_online(unsigned long j, 3617 const unsigned long *online_mask, 3618 unsigned int nr_bits) 3619 { 3620 cpu_max_bits_warn(j, nr_bits); 3621 3622 if (online_mask) 3623 return test_bit(j, online_mask); 3624 3625 return (j < nr_bits); 3626 } 3627 3628 /** 3629 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3630 * @n: CPU/Rx queue index 3631 * @srcp: the cpumask/Rx queue mask pointer 3632 * @nr_bits: number of bits in the bitmask 3633 * 3634 * Returns >= nr_bits if no further CPUs/Rx queues set. 3635 */ 3636 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3637 unsigned int nr_bits) 3638 { 3639 /* -1 is a legal arg here. */ 3640 if (n != -1) 3641 cpu_max_bits_warn(n, nr_bits); 3642 3643 if (srcp) 3644 return find_next_bit(srcp, nr_bits, n + 1); 3645 3646 return n + 1; 3647 } 3648 3649 /** 3650 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3651 * @n: CPU/Rx queue index 3652 * @src1p: the first CPUs/Rx queues mask pointer 3653 * @src2p: the second CPUs/Rx queues mask pointer 3654 * @nr_bits: number of bits in the bitmask 3655 * 3656 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3657 */ 3658 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3659 const unsigned long *src2p, 3660 unsigned int nr_bits) 3661 { 3662 /* -1 is a legal arg here. */ 3663 if (n != -1) 3664 cpu_max_bits_warn(n, nr_bits); 3665 3666 if (src1p && src2p) 3667 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3668 else if (src1p) 3669 return find_next_bit(src1p, nr_bits, n + 1); 3670 else if (src2p) 3671 return find_next_bit(src2p, nr_bits, n + 1); 3672 3673 return n + 1; 3674 } 3675 #else 3676 static inline int netif_set_xps_queue(struct net_device *dev, 3677 const struct cpumask *mask, 3678 u16 index) 3679 { 3680 return 0; 3681 } 3682 3683 static inline int __netif_set_xps_queue(struct net_device *dev, 3684 const unsigned long *mask, 3685 u16 index, bool is_rxqs_map) 3686 { 3687 return 0; 3688 } 3689 #endif 3690 3691 /** 3692 * netif_is_multiqueue - test if device has multiple transmit queues 3693 * @dev: network device 3694 * 3695 * Check if device has multiple transmit queues 3696 */ 3697 static inline bool netif_is_multiqueue(const struct net_device *dev) 3698 { 3699 return dev->num_tx_queues > 1; 3700 } 3701 3702 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3703 3704 #ifdef CONFIG_SYSFS 3705 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3706 #else 3707 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3708 unsigned int rxqs) 3709 { 3710 dev->real_num_rx_queues = rxqs; 3711 return 0; 3712 } 3713 #endif 3714 3715 static inline struct netdev_rx_queue * 3716 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3717 { 3718 return dev->_rx + rxq; 3719 } 3720 3721 #ifdef CONFIG_SYSFS 3722 static inline unsigned int get_netdev_rx_queue_index( 3723 struct netdev_rx_queue *queue) 3724 { 3725 struct net_device *dev = queue->dev; 3726 int index = queue - dev->_rx; 3727 3728 BUG_ON(index >= dev->num_rx_queues); 3729 return index; 3730 } 3731 #endif 3732 3733 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3734 int netif_get_num_default_rss_queues(void); 3735 3736 enum skb_free_reason { 3737 SKB_REASON_CONSUMED, 3738 SKB_REASON_DROPPED, 3739 }; 3740 3741 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3742 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3743 3744 /* 3745 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3746 * interrupt context or with hardware interrupts being disabled. 3747 * (in_irq() || irqs_disabled()) 3748 * 3749 * We provide four helpers that can be used in following contexts : 3750 * 3751 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3752 * replacing kfree_skb(skb) 3753 * 3754 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3755 * Typically used in place of consume_skb(skb) in TX completion path 3756 * 3757 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3758 * replacing kfree_skb(skb) 3759 * 3760 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3761 * and consumed a packet. Used in place of consume_skb(skb) 3762 */ 3763 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3764 { 3765 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3766 } 3767 3768 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3769 { 3770 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3771 } 3772 3773 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3774 { 3775 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3776 } 3777 3778 static inline void dev_consume_skb_any(struct sk_buff *skb) 3779 { 3780 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3781 } 3782 3783 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3784 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3785 int netif_rx(struct sk_buff *skb); 3786 int netif_rx_ni(struct sk_buff *skb); 3787 int netif_rx_any_context(struct sk_buff *skb); 3788 int netif_receive_skb(struct sk_buff *skb); 3789 int netif_receive_skb_core(struct sk_buff *skb); 3790 void netif_receive_skb_list(struct list_head *head); 3791 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3792 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3793 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3794 gro_result_t napi_gro_frags(struct napi_struct *napi); 3795 struct packet_offload *gro_find_receive_by_type(__be16 type); 3796 struct packet_offload *gro_find_complete_by_type(__be16 type); 3797 3798 static inline void napi_free_frags(struct napi_struct *napi) 3799 { 3800 kfree_skb(napi->skb); 3801 napi->skb = NULL; 3802 } 3803 3804 bool netdev_is_rx_handler_busy(struct net_device *dev); 3805 int netdev_rx_handler_register(struct net_device *dev, 3806 rx_handler_func_t *rx_handler, 3807 void *rx_handler_data); 3808 void netdev_rx_handler_unregister(struct net_device *dev); 3809 3810 bool dev_valid_name(const char *name); 3811 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3812 bool *need_copyout); 3813 int dev_ifconf(struct net *net, struct ifconf *, int); 3814 int dev_ethtool(struct net *net, struct ifreq *); 3815 unsigned int dev_get_flags(const struct net_device *); 3816 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3817 struct netlink_ext_ack *extack); 3818 int dev_change_flags(struct net_device *dev, unsigned int flags, 3819 struct netlink_ext_ack *extack); 3820 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3821 unsigned int gchanges); 3822 int dev_change_name(struct net_device *, const char *); 3823 int dev_set_alias(struct net_device *, const char *, size_t); 3824 int dev_get_alias(const struct net_device *, char *, size_t); 3825 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3826 int __dev_set_mtu(struct net_device *, int); 3827 int dev_validate_mtu(struct net_device *dev, int mtu, 3828 struct netlink_ext_ack *extack); 3829 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3830 struct netlink_ext_ack *extack); 3831 int dev_set_mtu(struct net_device *, int); 3832 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3833 void dev_set_group(struct net_device *, int); 3834 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3835 struct netlink_ext_ack *extack); 3836 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3837 struct netlink_ext_ack *extack); 3838 int dev_change_carrier(struct net_device *, bool new_carrier); 3839 int dev_get_phys_port_id(struct net_device *dev, 3840 struct netdev_phys_item_id *ppid); 3841 int dev_get_phys_port_name(struct net_device *dev, 3842 char *name, size_t len); 3843 int dev_get_port_parent_id(struct net_device *dev, 3844 struct netdev_phys_item_id *ppid, bool recurse); 3845 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3846 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3847 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3848 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3849 u32 value); 3850 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3851 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3852 struct netdev_queue *txq, int *ret); 3853 3854 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3855 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3856 int fd, int expected_fd, u32 flags); 3857 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3858 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3859 3860 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3861 3862 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3863 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3864 bool is_skb_forwardable(const struct net_device *dev, 3865 const struct sk_buff *skb); 3866 3867 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3868 struct sk_buff *skb) 3869 { 3870 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3871 unlikely(!is_skb_forwardable(dev, skb))) { 3872 atomic_long_inc(&dev->rx_dropped); 3873 kfree_skb(skb); 3874 return NET_RX_DROP; 3875 } 3876 3877 skb_scrub_packet(skb, true); 3878 skb->priority = 0; 3879 return 0; 3880 } 3881 3882 bool dev_nit_active(struct net_device *dev); 3883 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3884 3885 extern int netdev_budget; 3886 extern unsigned int netdev_budget_usecs; 3887 3888 /* Called by rtnetlink.c:rtnl_unlock() */ 3889 void netdev_run_todo(void); 3890 3891 /** 3892 * dev_put - release reference to device 3893 * @dev: network device 3894 * 3895 * Release reference to device to allow it to be freed. 3896 */ 3897 static inline void dev_put(struct net_device *dev) 3898 { 3899 this_cpu_dec(*dev->pcpu_refcnt); 3900 } 3901 3902 /** 3903 * dev_hold - get reference to device 3904 * @dev: network device 3905 * 3906 * Hold reference to device to keep it from being freed. 3907 */ 3908 static inline void dev_hold(struct net_device *dev) 3909 { 3910 this_cpu_inc(*dev->pcpu_refcnt); 3911 } 3912 3913 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3914 * and _off may be called from IRQ context, but it is caller 3915 * who is responsible for serialization of these calls. 3916 * 3917 * The name carrier is inappropriate, these functions should really be 3918 * called netif_lowerlayer_*() because they represent the state of any 3919 * kind of lower layer not just hardware media. 3920 */ 3921 3922 void linkwatch_init_dev(struct net_device *dev); 3923 void linkwatch_fire_event(struct net_device *dev); 3924 void linkwatch_forget_dev(struct net_device *dev); 3925 3926 /** 3927 * netif_carrier_ok - test if carrier present 3928 * @dev: network device 3929 * 3930 * Check if carrier is present on device 3931 */ 3932 static inline bool netif_carrier_ok(const struct net_device *dev) 3933 { 3934 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3935 } 3936 3937 unsigned long dev_trans_start(struct net_device *dev); 3938 3939 void __netdev_watchdog_up(struct net_device *dev); 3940 3941 void netif_carrier_on(struct net_device *dev); 3942 3943 void netif_carrier_off(struct net_device *dev); 3944 3945 /** 3946 * netif_dormant_on - mark device as dormant. 3947 * @dev: network device 3948 * 3949 * Mark device as dormant (as per RFC2863). 3950 * 3951 * The dormant state indicates that the relevant interface is not 3952 * actually in a condition to pass packets (i.e., it is not 'up') but is 3953 * in a "pending" state, waiting for some external event. For "on- 3954 * demand" interfaces, this new state identifies the situation where the 3955 * interface is waiting for events to place it in the up state. 3956 */ 3957 static inline void netif_dormant_on(struct net_device *dev) 3958 { 3959 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3960 linkwatch_fire_event(dev); 3961 } 3962 3963 /** 3964 * netif_dormant_off - set device as not dormant. 3965 * @dev: network device 3966 * 3967 * Device is not in dormant state. 3968 */ 3969 static inline void netif_dormant_off(struct net_device *dev) 3970 { 3971 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3972 linkwatch_fire_event(dev); 3973 } 3974 3975 /** 3976 * netif_dormant - test if device is dormant 3977 * @dev: network device 3978 * 3979 * Check if device is dormant. 3980 */ 3981 static inline bool netif_dormant(const struct net_device *dev) 3982 { 3983 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3984 } 3985 3986 3987 /** 3988 * netif_testing_on - mark device as under test. 3989 * @dev: network device 3990 * 3991 * Mark device as under test (as per RFC2863). 3992 * 3993 * The testing state indicates that some test(s) must be performed on 3994 * the interface. After completion, of the test, the interface state 3995 * will change to up, dormant, or down, as appropriate. 3996 */ 3997 static inline void netif_testing_on(struct net_device *dev) 3998 { 3999 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4000 linkwatch_fire_event(dev); 4001 } 4002 4003 /** 4004 * netif_testing_off - set device as not under test. 4005 * @dev: network device 4006 * 4007 * Device is not in testing state. 4008 */ 4009 static inline void netif_testing_off(struct net_device *dev) 4010 { 4011 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4012 linkwatch_fire_event(dev); 4013 } 4014 4015 /** 4016 * netif_testing - test if device is under test 4017 * @dev: network device 4018 * 4019 * Check if device is under test 4020 */ 4021 static inline bool netif_testing(const struct net_device *dev) 4022 { 4023 return test_bit(__LINK_STATE_TESTING, &dev->state); 4024 } 4025 4026 4027 /** 4028 * netif_oper_up - test if device is operational 4029 * @dev: network device 4030 * 4031 * Check if carrier is operational 4032 */ 4033 static inline bool netif_oper_up(const struct net_device *dev) 4034 { 4035 return (dev->operstate == IF_OPER_UP || 4036 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4037 } 4038 4039 /** 4040 * netif_device_present - is device available or removed 4041 * @dev: network device 4042 * 4043 * Check if device has not been removed from system. 4044 */ 4045 static inline bool netif_device_present(struct net_device *dev) 4046 { 4047 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4048 } 4049 4050 void netif_device_detach(struct net_device *dev); 4051 4052 void netif_device_attach(struct net_device *dev); 4053 4054 /* 4055 * Network interface message level settings 4056 */ 4057 4058 enum { 4059 NETIF_MSG_DRV_BIT, 4060 NETIF_MSG_PROBE_BIT, 4061 NETIF_MSG_LINK_BIT, 4062 NETIF_MSG_TIMER_BIT, 4063 NETIF_MSG_IFDOWN_BIT, 4064 NETIF_MSG_IFUP_BIT, 4065 NETIF_MSG_RX_ERR_BIT, 4066 NETIF_MSG_TX_ERR_BIT, 4067 NETIF_MSG_TX_QUEUED_BIT, 4068 NETIF_MSG_INTR_BIT, 4069 NETIF_MSG_TX_DONE_BIT, 4070 NETIF_MSG_RX_STATUS_BIT, 4071 NETIF_MSG_PKTDATA_BIT, 4072 NETIF_MSG_HW_BIT, 4073 NETIF_MSG_WOL_BIT, 4074 4075 /* When you add a new bit above, update netif_msg_class_names array 4076 * in net/ethtool/common.c 4077 */ 4078 NETIF_MSG_CLASS_COUNT, 4079 }; 4080 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4081 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4082 4083 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4084 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4085 4086 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4087 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4088 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4089 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4090 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4091 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4092 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4093 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4094 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4095 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4096 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4097 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4098 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4099 #define NETIF_MSG_HW __NETIF_MSG(HW) 4100 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4101 4102 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4103 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4104 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4105 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4106 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4107 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4108 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4109 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4110 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4111 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4112 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4113 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4114 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4115 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4116 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4117 4118 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4119 { 4120 /* use default */ 4121 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4122 return default_msg_enable_bits; 4123 if (debug_value == 0) /* no output */ 4124 return 0; 4125 /* set low N bits */ 4126 return (1U << debug_value) - 1; 4127 } 4128 4129 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4130 { 4131 spin_lock(&txq->_xmit_lock); 4132 txq->xmit_lock_owner = cpu; 4133 } 4134 4135 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4136 { 4137 __acquire(&txq->_xmit_lock); 4138 return true; 4139 } 4140 4141 static inline void __netif_tx_release(struct netdev_queue *txq) 4142 { 4143 __release(&txq->_xmit_lock); 4144 } 4145 4146 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4147 { 4148 spin_lock_bh(&txq->_xmit_lock); 4149 txq->xmit_lock_owner = smp_processor_id(); 4150 } 4151 4152 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4153 { 4154 bool ok = spin_trylock(&txq->_xmit_lock); 4155 if (likely(ok)) 4156 txq->xmit_lock_owner = smp_processor_id(); 4157 return ok; 4158 } 4159 4160 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4161 { 4162 txq->xmit_lock_owner = -1; 4163 spin_unlock(&txq->_xmit_lock); 4164 } 4165 4166 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4167 { 4168 txq->xmit_lock_owner = -1; 4169 spin_unlock_bh(&txq->_xmit_lock); 4170 } 4171 4172 static inline void txq_trans_update(struct netdev_queue *txq) 4173 { 4174 if (txq->xmit_lock_owner != -1) 4175 txq->trans_start = jiffies; 4176 } 4177 4178 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4179 static inline void netif_trans_update(struct net_device *dev) 4180 { 4181 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4182 4183 if (txq->trans_start != jiffies) 4184 txq->trans_start = jiffies; 4185 } 4186 4187 /** 4188 * netif_tx_lock - grab network device transmit lock 4189 * @dev: network device 4190 * 4191 * Get network device transmit lock 4192 */ 4193 static inline void netif_tx_lock(struct net_device *dev) 4194 { 4195 unsigned int i; 4196 int cpu; 4197 4198 spin_lock(&dev->tx_global_lock); 4199 cpu = smp_processor_id(); 4200 for (i = 0; i < dev->num_tx_queues; i++) { 4201 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4202 4203 /* We are the only thread of execution doing a 4204 * freeze, but we have to grab the _xmit_lock in 4205 * order to synchronize with threads which are in 4206 * the ->hard_start_xmit() handler and already 4207 * checked the frozen bit. 4208 */ 4209 __netif_tx_lock(txq, cpu); 4210 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4211 __netif_tx_unlock(txq); 4212 } 4213 } 4214 4215 static inline void netif_tx_lock_bh(struct net_device *dev) 4216 { 4217 local_bh_disable(); 4218 netif_tx_lock(dev); 4219 } 4220 4221 static inline void netif_tx_unlock(struct net_device *dev) 4222 { 4223 unsigned int i; 4224 4225 for (i = 0; i < dev->num_tx_queues; i++) { 4226 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4227 4228 /* No need to grab the _xmit_lock here. If the 4229 * queue is not stopped for another reason, we 4230 * force a schedule. 4231 */ 4232 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4233 netif_schedule_queue(txq); 4234 } 4235 spin_unlock(&dev->tx_global_lock); 4236 } 4237 4238 static inline void netif_tx_unlock_bh(struct net_device *dev) 4239 { 4240 netif_tx_unlock(dev); 4241 local_bh_enable(); 4242 } 4243 4244 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4245 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4246 __netif_tx_lock(txq, cpu); \ 4247 } else { \ 4248 __netif_tx_acquire(txq); \ 4249 } \ 4250 } 4251 4252 #define HARD_TX_TRYLOCK(dev, txq) \ 4253 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4254 __netif_tx_trylock(txq) : \ 4255 __netif_tx_acquire(txq)) 4256 4257 #define HARD_TX_UNLOCK(dev, txq) { \ 4258 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4259 __netif_tx_unlock(txq); \ 4260 } else { \ 4261 __netif_tx_release(txq); \ 4262 } \ 4263 } 4264 4265 static inline void netif_tx_disable(struct net_device *dev) 4266 { 4267 unsigned int i; 4268 int cpu; 4269 4270 local_bh_disable(); 4271 cpu = smp_processor_id(); 4272 for (i = 0; i < dev->num_tx_queues; i++) { 4273 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4274 4275 __netif_tx_lock(txq, cpu); 4276 netif_tx_stop_queue(txq); 4277 __netif_tx_unlock(txq); 4278 } 4279 local_bh_enable(); 4280 } 4281 4282 static inline void netif_addr_lock(struct net_device *dev) 4283 { 4284 spin_lock(&dev->addr_list_lock); 4285 } 4286 4287 static inline void netif_addr_lock_nested(struct net_device *dev) 4288 { 4289 spin_lock_nested(&dev->addr_list_lock, dev->lower_level); 4290 } 4291 4292 static inline void netif_addr_lock_bh(struct net_device *dev) 4293 { 4294 spin_lock_bh(&dev->addr_list_lock); 4295 } 4296 4297 static inline void netif_addr_unlock(struct net_device *dev) 4298 { 4299 spin_unlock(&dev->addr_list_lock); 4300 } 4301 4302 static inline void netif_addr_unlock_bh(struct net_device *dev) 4303 { 4304 spin_unlock_bh(&dev->addr_list_lock); 4305 } 4306 4307 /* 4308 * dev_addrs walker. Should be used only for read access. Call with 4309 * rcu_read_lock held. 4310 */ 4311 #define for_each_dev_addr(dev, ha) \ 4312 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4313 4314 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4315 4316 void ether_setup(struct net_device *dev); 4317 4318 /* Support for loadable net-drivers */ 4319 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4320 unsigned char name_assign_type, 4321 void (*setup)(struct net_device *), 4322 unsigned int txqs, unsigned int rxqs); 4323 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4324 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4325 4326 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4327 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4328 count) 4329 4330 int register_netdev(struct net_device *dev); 4331 void unregister_netdev(struct net_device *dev); 4332 4333 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4334 4335 /* General hardware address lists handling functions */ 4336 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4337 struct netdev_hw_addr_list *from_list, int addr_len); 4338 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4339 struct netdev_hw_addr_list *from_list, int addr_len); 4340 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4341 struct net_device *dev, 4342 int (*sync)(struct net_device *, const unsigned char *), 4343 int (*unsync)(struct net_device *, 4344 const unsigned char *)); 4345 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4346 struct net_device *dev, 4347 int (*sync)(struct net_device *, 4348 const unsigned char *, int), 4349 int (*unsync)(struct net_device *, 4350 const unsigned char *, int)); 4351 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4352 struct net_device *dev, 4353 int (*unsync)(struct net_device *, 4354 const unsigned char *, int)); 4355 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4356 struct net_device *dev, 4357 int (*unsync)(struct net_device *, 4358 const unsigned char *)); 4359 void __hw_addr_init(struct netdev_hw_addr_list *list); 4360 4361 /* Functions used for device addresses handling */ 4362 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4363 unsigned char addr_type); 4364 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4365 unsigned char addr_type); 4366 void dev_addr_flush(struct net_device *dev); 4367 int dev_addr_init(struct net_device *dev); 4368 4369 /* Functions used for unicast addresses handling */ 4370 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4371 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4372 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4373 int dev_uc_sync(struct net_device *to, struct net_device *from); 4374 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4375 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4376 void dev_uc_flush(struct net_device *dev); 4377 void dev_uc_init(struct net_device *dev); 4378 4379 /** 4380 * __dev_uc_sync - Synchonize device's unicast list 4381 * @dev: device to sync 4382 * @sync: function to call if address should be added 4383 * @unsync: function to call if address should be removed 4384 * 4385 * Add newly added addresses to the interface, and release 4386 * addresses that have been deleted. 4387 */ 4388 static inline int __dev_uc_sync(struct net_device *dev, 4389 int (*sync)(struct net_device *, 4390 const unsigned char *), 4391 int (*unsync)(struct net_device *, 4392 const unsigned char *)) 4393 { 4394 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4395 } 4396 4397 /** 4398 * __dev_uc_unsync - Remove synchronized addresses from device 4399 * @dev: device to sync 4400 * @unsync: function to call if address should be removed 4401 * 4402 * Remove all addresses that were added to the device by dev_uc_sync(). 4403 */ 4404 static inline void __dev_uc_unsync(struct net_device *dev, 4405 int (*unsync)(struct net_device *, 4406 const unsigned char *)) 4407 { 4408 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4409 } 4410 4411 /* Functions used for multicast addresses handling */ 4412 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4413 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4414 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4415 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4416 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4417 int dev_mc_sync(struct net_device *to, struct net_device *from); 4418 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4419 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4420 void dev_mc_flush(struct net_device *dev); 4421 void dev_mc_init(struct net_device *dev); 4422 4423 /** 4424 * __dev_mc_sync - Synchonize device's multicast list 4425 * @dev: device to sync 4426 * @sync: function to call if address should be added 4427 * @unsync: function to call if address should be removed 4428 * 4429 * Add newly added addresses to the interface, and release 4430 * addresses that have been deleted. 4431 */ 4432 static inline int __dev_mc_sync(struct net_device *dev, 4433 int (*sync)(struct net_device *, 4434 const unsigned char *), 4435 int (*unsync)(struct net_device *, 4436 const unsigned char *)) 4437 { 4438 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4439 } 4440 4441 /** 4442 * __dev_mc_unsync - Remove synchronized addresses from device 4443 * @dev: device to sync 4444 * @unsync: function to call if address should be removed 4445 * 4446 * Remove all addresses that were added to the device by dev_mc_sync(). 4447 */ 4448 static inline void __dev_mc_unsync(struct net_device *dev, 4449 int (*unsync)(struct net_device *, 4450 const unsigned char *)) 4451 { 4452 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4453 } 4454 4455 /* Functions used for secondary unicast and multicast support */ 4456 void dev_set_rx_mode(struct net_device *dev); 4457 void __dev_set_rx_mode(struct net_device *dev); 4458 int dev_set_promiscuity(struct net_device *dev, int inc); 4459 int dev_set_allmulti(struct net_device *dev, int inc); 4460 void netdev_state_change(struct net_device *dev); 4461 void netdev_notify_peers(struct net_device *dev); 4462 void netdev_features_change(struct net_device *dev); 4463 /* Load a device via the kmod */ 4464 void dev_load(struct net *net, const char *name); 4465 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4466 struct rtnl_link_stats64 *storage); 4467 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4468 const struct net_device_stats *netdev_stats); 4469 4470 extern int netdev_max_backlog; 4471 extern int netdev_tstamp_prequeue; 4472 extern int weight_p; 4473 extern int dev_weight_rx_bias; 4474 extern int dev_weight_tx_bias; 4475 extern int dev_rx_weight; 4476 extern int dev_tx_weight; 4477 extern int gro_normal_batch; 4478 4479 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4480 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4481 struct list_head **iter); 4482 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4483 struct list_head **iter); 4484 4485 /* iterate through upper list, must be called under RCU read lock */ 4486 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4487 for (iter = &(dev)->adj_list.upper, \ 4488 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4489 updev; \ 4490 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4491 4492 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4493 int (*fn)(struct net_device *upper_dev, 4494 void *data), 4495 void *data); 4496 4497 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4498 struct net_device *upper_dev); 4499 4500 bool netdev_has_any_upper_dev(struct net_device *dev); 4501 4502 void *netdev_lower_get_next_private(struct net_device *dev, 4503 struct list_head **iter); 4504 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4505 struct list_head **iter); 4506 4507 #define netdev_for_each_lower_private(dev, priv, iter) \ 4508 for (iter = (dev)->adj_list.lower.next, \ 4509 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4510 priv; \ 4511 priv = netdev_lower_get_next_private(dev, &(iter))) 4512 4513 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4514 for (iter = &(dev)->adj_list.lower, \ 4515 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4516 priv; \ 4517 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4518 4519 void *netdev_lower_get_next(struct net_device *dev, 4520 struct list_head **iter); 4521 4522 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4523 for (iter = (dev)->adj_list.lower.next, \ 4524 ldev = netdev_lower_get_next(dev, &(iter)); \ 4525 ldev; \ 4526 ldev = netdev_lower_get_next(dev, &(iter))) 4527 4528 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4529 struct list_head **iter); 4530 int netdev_walk_all_lower_dev(struct net_device *dev, 4531 int (*fn)(struct net_device *lower_dev, 4532 void *data), 4533 void *data); 4534 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4535 int (*fn)(struct net_device *lower_dev, 4536 void *data), 4537 void *data); 4538 4539 void *netdev_adjacent_get_private(struct list_head *adj_list); 4540 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4541 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4542 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4543 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4544 struct netlink_ext_ack *extack); 4545 int netdev_master_upper_dev_link(struct net_device *dev, 4546 struct net_device *upper_dev, 4547 void *upper_priv, void *upper_info, 4548 struct netlink_ext_ack *extack); 4549 void netdev_upper_dev_unlink(struct net_device *dev, 4550 struct net_device *upper_dev); 4551 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4552 struct net_device *new_dev, 4553 struct net_device *dev, 4554 struct netlink_ext_ack *extack); 4555 void netdev_adjacent_change_commit(struct net_device *old_dev, 4556 struct net_device *new_dev, 4557 struct net_device *dev); 4558 void netdev_adjacent_change_abort(struct net_device *old_dev, 4559 struct net_device *new_dev, 4560 struct net_device *dev); 4561 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4562 void *netdev_lower_dev_get_private(struct net_device *dev, 4563 struct net_device *lower_dev); 4564 void netdev_lower_state_changed(struct net_device *lower_dev, 4565 void *lower_state_info); 4566 4567 /* RSS keys are 40 or 52 bytes long */ 4568 #define NETDEV_RSS_KEY_LEN 52 4569 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4570 void netdev_rss_key_fill(void *buffer, size_t len); 4571 4572 int skb_checksum_help(struct sk_buff *skb); 4573 int skb_crc32c_csum_help(struct sk_buff *skb); 4574 int skb_csum_hwoffload_help(struct sk_buff *skb, 4575 const netdev_features_t features); 4576 4577 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4578 netdev_features_t features, bool tx_path); 4579 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4580 netdev_features_t features); 4581 4582 struct netdev_bonding_info { 4583 ifslave slave; 4584 ifbond master; 4585 }; 4586 4587 struct netdev_notifier_bonding_info { 4588 struct netdev_notifier_info info; /* must be first */ 4589 struct netdev_bonding_info bonding_info; 4590 }; 4591 4592 void netdev_bonding_info_change(struct net_device *dev, 4593 struct netdev_bonding_info *bonding_info); 4594 4595 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4596 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4597 #else 4598 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4599 const void *data) 4600 { 4601 } 4602 #endif 4603 4604 static inline 4605 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4606 { 4607 return __skb_gso_segment(skb, features, true); 4608 } 4609 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4610 4611 static inline bool can_checksum_protocol(netdev_features_t features, 4612 __be16 protocol) 4613 { 4614 if (protocol == htons(ETH_P_FCOE)) 4615 return !!(features & NETIF_F_FCOE_CRC); 4616 4617 /* Assume this is an IP checksum (not SCTP CRC) */ 4618 4619 if (features & NETIF_F_HW_CSUM) { 4620 /* Can checksum everything */ 4621 return true; 4622 } 4623 4624 switch (protocol) { 4625 case htons(ETH_P_IP): 4626 return !!(features & NETIF_F_IP_CSUM); 4627 case htons(ETH_P_IPV6): 4628 return !!(features & NETIF_F_IPV6_CSUM); 4629 default: 4630 return false; 4631 } 4632 } 4633 4634 #ifdef CONFIG_BUG 4635 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4636 #else 4637 static inline void netdev_rx_csum_fault(struct net_device *dev, 4638 struct sk_buff *skb) 4639 { 4640 } 4641 #endif 4642 /* rx skb timestamps */ 4643 void net_enable_timestamp(void); 4644 void net_disable_timestamp(void); 4645 4646 #ifdef CONFIG_PROC_FS 4647 int __init dev_proc_init(void); 4648 #else 4649 #define dev_proc_init() 0 4650 #endif 4651 4652 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4653 struct sk_buff *skb, struct net_device *dev, 4654 bool more) 4655 { 4656 __this_cpu_write(softnet_data.xmit.more, more); 4657 return ops->ndo_start_xmit(skb, dev); 4658 } 4659 4660 static inline bool netdev_xmit_more(void) 4661 { 4662 return __this_cpu_read(softnet_data.xmit.more); 4663 } 4664 4665 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4666 struct netdev_queue *txq, bool more) 4667 { 4668 const struct net_device_ops *ops = dev->netdev_ops; 4669 netdev_tx_t rc; 4670 4671 rc = __netdev_start_xmit(ops, skb, dev, more); 4672 if (rc == NETDEV_TX_OK) 4673 txq_trans_update(txq); 4674 4675 return rc; 4676 } 4677 4678 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4679 const void *ns); 4680 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4681 const void *ns); 4682 4683 extern const struct kobj_ns_type_operations net_ns_type_operations; 4684 4685 const char *netdev_drivername(const struct net_device *dev); 4686 4687 void linkwatch_run_queue(void); 4688 4689 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4690 netdev_features_t f2) 4691 { 4692 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4693 if (f1 & NETIF_F_HW_CSUM) 4694 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4695 else 4696 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4697 } 4698 4699 return f1 & f2; 4700 } 4701 4702 static inline netdev_features_t netdev_get_wanted_features( 4703 struct net_device *dev) 4704 { 4705 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4706 } 4707 netdev_features_t netdev_increment_features(netdev_features_t all, 4708 netdev_features_t one, netdev_features_t mask); 4709 4710 /* Allow TSO being used on stacked device : 4711 * Performing the GSO segmentation before last device 4712 * is a performance improvement. 4713 */ 4714 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4715 netdev_features_t mask) 4716 { 4717 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4718 } 4719 4720 int __netdev_update_features(struct net_device *dev); 4721 void netdev_update_features(struct net_device *dev); 4722 void netdev_change_features(struct net_device *dev); 4723 4724 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4725 struct net_device *dev); 4726 4727 netdev_features_t passthru_features_check(struct sk_buff *skb, 4728 struct net_device *dev, 4729 netdev_features_t features); 4730 netdev_features_t netif_skb_features(struct sk_buff *skb); 4731 4732 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4733 { 4734 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4735 4736 /* check flags correspondence */ 4737 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4738 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4739 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4740 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4741 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4742 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4743 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4744 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4745 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4746 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4747 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4748 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4749 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4750 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4751 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4752 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4753 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4754 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4755 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4756 4757 return (features & feature) == feature; 4758 } 4759 4760 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4761 { 4762 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4763 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4764 } 4765 4766 static inline bool netif_needs_gso(struct sk_buff *skb, 4767 netdev_features_t features) 4768 { 4769 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4770 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4771 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4772 } 4773 4774 static inline void netif_set_gso_max_size(struct net_device *dev, 4775 unsigned int size) 4776 { 4777 dev->gso_max_size = size; 4778 } 4779 4780 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4781 int pulled_hlen, u16 mac_offset, 4782 int mac_len) 4783 { 4784 skb->protocol = protocol; 4785 skb->encapsulation = 1; 4786 skb_push(skb, pulled_hlen); 4787 skb_reset_transport_header(skb); 4788 skb->mac_header = mac_offset; 4789 skb->network_header = skb->mac_header + mac_len; 4790 skb->mac_len = mac_len; 4791 } 4792 4793 static inline bool netif_is_macsec(const struct net_device *dev) 4794 { 4795 return dev->priv_flags & IFF_MACSEC; 4796 } 4797 4798 static inline bool netif_is_macvlan(const struct net_device *dev) 4799 { 4800 return dev->priv_flags & IFF_MACVLAN; 4801 } 4802 4803 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4804 { 4805 return dev->priv_flags & IFF_MACVLAN_PORT; 4806 } 4807 4808 static inline bool netif_is_bond_master(const struct net_device *dev) 4809 { 4810 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4811 } 4812 4813 static inline bool netif_is_bond_slave(const struct net_device *dev) 4814 { 4815 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4816 } 4817 4818 static inline bool netif_supports_nofcs(struct net_device *dev) 4819 { 4820 return dev->priv_flags & IFF_SUPP_NOFCS; 4821 } 4822 4823 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4824 { 4825 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4826 } 4827 4828 static inline bool netif_is_l3_master(const struct net_device *dev) 4829 { 4830 return dev->priv_flags & IFF_L3MDEV_MASTER; 4831 } 4832 4833 static inline bool netif_is_l3_slave(const struct net_device *dev) 4834 { 4835 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4836 } 4837 4838 static inline bool netif_is_bridge_master(const struct net_device *dev) 4839 { 4840 return dev->priv_flags & IFF_EBRIDGE; 4841 } 4842 4843 static inline bool netif_is_bridge_port(const struct net_device *dev) 4844 { 4845 return dev->priv_flags & IFF_BRIDGE_PORT; 4846 } 4847 4848 static inline bool netif_is_ovs_master(const struct net_device *dev) 4849 { 4850 return dev->priv_flags & IFF_OPENVSWITCH; 4851 } 4852 4853 static inline bool netif_is_ovs_port(const struct net_device *dev) 4854 { 4855 return dev->priv_flags & IFF_OVS_DATAPATH; 4856 } 4857 4858 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4859 { 4860 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4861 } 4862 4863 static inline bool netif_is_team_master(const struct net_device *dev) 4864 { 4865 return dev->priv_flags & IFF_TEAM; 4866 } 4867 4868 static inline bool netif_is_team_port(const struct net_device *dev) 4869 { 4870 return dev->priv_flags & IFF_TEAM_PORT; 4871 } 4872 4873 static inline bool netif_is_lag_master(const struct net_device *dev) 4874 { 4875 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4876 } 4877 4878 static inline bool netif_is_lag_port(const struct net_device *dev) 4879 { 4880 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4881 } 4882 4883 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4884 { 4885 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4886 } 4887 4888 static inline bool netif_is_failover(const struct net_device *dev) 4889 { 4890 return dev->priv_flags & IFF_FAILOVER; 4891 } 4892 4893 static inline bool netif_is_failover_slave(const struct net_device *dev) 4894 { 4895 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4896 } 4897 4898 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4899 static inline void netif_keep_dst(struct net_device *dev) 4900 { 4901 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4902 } 4903 4904 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4905 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4906 { 4907 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4908 return dev->priv_flags & IFF_MACSEC; 4909 } 4910 4911 extern struct pernet_operations __net_initdata loopback_net_ops; 4912 4913 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4914 4915 /* netdev_printk helpers, similar to dev_printk */ 4916 4917 static inline const char *netdev_name(const struct net_device *dev) 4918 { 4919 if (!dev->name[0] || strchr(dev->name, '%')) 4920 return "(unnamed net_device)"; 4921 return dev->name; 4922 } 4923 4924 static inline bool netdev_unregistering(const struct net_device *dev) 4925 { 4926 return dev->reg_state == NETREG_UNREGISTERING; 4927 } 4928 4929 static inline const char *netdev_reg_state(const struct net_device *dev) 4930 { 4931 switch (dev->reg_state) { 4932 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4933 case NETREG_REGISTERED: return ""; 4934 case NETREG_UNREGISTERING: return " (unregistering)"; 4935 case NETREG_UNREGISTERED: return " (unregistered)"; 4936 case NETREG_RELEASED: return " (released)"; 4937 case NETREG_DUMMY: return " (dummy)"; 4938 } 4939 4940 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4941 return " (unknown)"; 4942 } 4943 4944 __printf(3, 4) __cold 4945 void netdev_printk(const char *level, const struct net_device *dev, 4946 const char *format, ...); 4947 __printf(2, 3) __cold 4948 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4949 __printf(2, 3) __cold 4950 void netdev_alert(const struct net_device *dev, const char *format, ...); 4951 __printf(2, 3) __cold 4952 void netdev_crit(const struct net_device *dev, const char *format, ...); 4953 __printf(2, 3) __cold 4954 void netdev_err(const struct net_device *dev, const char *format, ...); 4955 __printf(2, 3) __cold 4956 void netdev_warn(const struct net_device *dev, const char *format, ...); 4957 __printf(2, 3) __cold 4958 void netdev_notice(const struct net_device *dev, const char *format, ...); 4959 __printf(2, 3) __cold 4960 void netdev_info(const struct net_device *dev, const char *format, ...); 4961 4962 #define netdev_level_once(level, dev, fmt, ...) \ 4963 do { \ 4964 static bool __print_once __read_mostly; \ 4965 \ 4966 if (!__print_once) { \ 4967 __print_once = true; \ 4968 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4969 } \ 4970 } while (0) 4971 4972 #define netdev_emerg_once(dev, fmt, ...) \ 4973 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4974 #define netdev_alert_once(dev, fmt, ...) \ 4975 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4976 #define netdev_crit_once(dev, fmt, ...) \ 4977 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4978 #define netdev_err_once(dev, fmt, ...) \ 4979 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4980 #define netdev_warn_once(dev, fmt, ...) \ 4981 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4982 #define netdev_notice_once(dev, fmt, ...) \ 4983 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4984 #define netdev_info_once(dev, fmt, ...) \ 4985 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4986 4987 #define MODULE_ALIAS_NETDEV(device) \ 4988 MODULE_ALIAS("netdev-" device) 4989 4990 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 4991 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 4992 #define netdev_dbg(__dev, format, args...) \ 4993 do { \ 4994 dynamic_netdev_dbg(__dev, format, ##args); \ 4995 } while (0) 4996 #elif defined(DEBUG) 4997 #define netdev_dbg(__dev, format, args...) \ 4998 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4999 #else 5000 #define netdev_dbg(__dev, format, args...) \ 5001 ({ \ 5002 if (0) \ 5003 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5004 }) 5005 #endif 5006 5007 #if defined(VERBOSE_DEBUG) 5008 #define netdev_vdbg netdev_dbg 5009 #else 5010 5011 #define netdev_vdbg(dev, format, args...) \ 5012 ({ \ 5013 if (0) \ 5014 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5015 0; \ 5016 }) 5017 #endif 5018 5019 /* 5020 * netdev_WARN() acts like dev_printk(), but with the key difference 5021 * of using a WARN/WARN_ON to get the message out, including the 5022 * file/line information and a backtrace. 5023 */ 5024 #define netdev_WARN(dev, format, args...) \ 5025 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5026 netdev_reg_state(dev), ##args) 5027 5028 #define netdev_WARN_ONCE(dev, format, args...) \ 5029 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5030 netdev_reg_state(dev), ##args) 5031 5032 /* netif printk helpers, similar to netdev_printk */ 5033 5034 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5035 do { \ 5036 if (netif_msg_##type(priv)) \ 5037 netdev_printk(level, (dev), fmt, ##args); \ 5038 } while (0) 5039 5040 #define netif_level(level, priv, type, dev, fmt, args...) \ 5041 do { \ 5042 if (netif_msg_##type(priv)) \ 5043 netdev_##level(dev, fmt, ##args); \ 5044 } while (0) 5045 5046 #define netif_emerg(priv, type, dev, fmt, args...) \ 5047 netif_level(emerg, priv, type, dev, fmt, ##args) 5048 #define netif_alert(priv, type, dev, fmt, args...) \ 5049 netif_level(alert, priv, type, dev, fmt, ##args) 5050 #define netif_crit(priv, type, dev, fmt, args...) \ 5051 netif_level(crit, priv, type, dev, fmt, ##args) 5052 #define netif_err(priv, type, dev, fmt, args...) \ 5053 netif_level(err, priv, type, dev, fmt, ##args) 5054 #define netif_warn(priv, type, dev, fmt, args...) \ 5055 netif_level(warn, priv, type, dev, fmt, ##args) 5056 #define netif_notice(priv, type, dev, fmt, args...) \ 5057 netif_level(notice, priv, type, dev, fmt, ##args) 5058 #define netif_info(priv, type, dev, fmt, args...) \ 5059 netif_level(info, priv, type, dev, fmt, ##args) 5060 5061 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5062 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5063 #define netif_dbg(priv, type, netdev, format, args...) \ 5064 do { \ 5065 if (netif_msg_##type(priv)) \ 5066 dynamic_netdev_dbg(netdev, format, ##args); \ 5067 } while (0) 5068 #elif defined(DEBUG) 5069 #define netif_dbg(priv, type, dev, format, args...) \ 5070 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5071 #else 5072 #define netif_dbg(priv, type, dev, format, args...) \ 5073 ({ \ 5074 if (0) \ 5075 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5076 0; \ 5077 }) 5078 #endif 5079 5080 /* if @cond then downgrade to debug, else print at @level */ 5081 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5082 do { \ 5083 if (cond) \ 5084 netif_dbg(priv, type, netdev, fmt, ##args); \ 5085 else \ 5086 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5087 } while (0) 5088 5089 #if defined(VERBOSE_DEBUG) 5090 #define netif_vdbg netif_dbg 5091 #else 5092 #define netif_vdbg(priv, type, dev, format, args...) \ 5093 ({ \ 5094 if (0) \ 5095 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5096 0; \ 5097 }) 5098 #endif 5099 5100 /* 5101 * The list of packet types we will receive (as opposed to discard) 5102 * and the routines to invoke. 5103 * 5104 * Why 16. Because with 16 the only overlap we get on a hash of the 5105 * low nibble of the protocol value is RARP/SNAP/X.25. 5106 * 5107 * 0800 IP 5108 * 0001 802.3 5109 * 0002 AX.25 5110 * 0004 802.2 5111 * 8035 RARP 5112 * 0005 SNAP 5113 * 0805 X.25 5114 * 0806 ARP 5115 * 8137 IPX 5116 * 0009 Localtalk 5117 * 86DD IPv6 5118 */ 5119 #define PTYPE_HASH_SIZE (16) 5120 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5121 5122 extern struct net_device *blackhole_netdev; 5123 5124 #endif /* _LINUX_NETDEVICE_H */ 5125