1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
36
37 #include <linux/ethtool.h>
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 #include <net/xdp.h>
44
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
51
52 struct netpoll_info;
53 struct device;
54 struct phy_device;
55 struct dsa_port;
56
57 struct sfp_bus;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63 /* UDP Tunnel offloads */
64 struct udp_tunnel_info;
65 struct bpf_prog;
66 struct xdp_buff;
67
68 void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
70
71 /* Backlog congestion levels */
72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73 #define NET_RX_DROP 1 /* packet dropped */
74
75 /*
76 * Transmit return codes: transmit return codes originate from three different
77 * namespaces:
78 *
79 * - qdisc return codes
80 * - driver transmit return codes
81 * - errno values
82 *
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
87 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
89 * others are propagated to higher layers.
90 */
91
92 /* qdisc ->enqueue() return codes. */
93 #define NET_XMIT_SUCCESS 0x00
94 #define NET_XMIT_DROP 0x01 /* skb dropped */
95 #define NET_XMIT_CN 0x02 /* congestion notification */
96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
97
98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103
104 /* Driver transmit return codes */
105 #define NETDEV_TX_MASK 0xf0
106
107 enum netdev_tx {
108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
111 };
112 typedef enum netdev_tx netdev_tx_t;
113
114 /*
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 */
118 static inline bool dev_xmit_complete(int rc)
119 {
120 /*
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 */
126 if (likely(rc < NET_XMIT_MASK))
127 return true;
128
129 return false;
130 }
131
132 /*
133 * Compute the worst-case header length according to the protocols
134 * used.
135 */
136
137 #if defined(CONFIG_HYPERV_NET)
138 # define LL_MAX_HEADER 128
139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140 # if defined(CONFIG_MAC80211_MESH)
141 # define LL_MAX_HEADER 128
142 # else
143 # define LL_MAX_HEADER 96
144 # endif
145 #else
146 # define LL_MAX_HEADER 32
147 #endif
148
149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
151 #define MAX_HEADER LL_MAX_HEADER
152 #else
153 #define MAX_HEADER (LL_MAX_HEADER + 48)
154 #endif
155
156 /*
157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
159 */
160
161 struct net_device_stats {
162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
171 unsigned long collisions;
172 unsigned long rx_length_errors;
173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
185 };
186
187
188 #include <linux/cache.h>
189 #include <linux/skbuff.h>
190
191 #ifdef CONFIG_RPS
192 #include <linux/static_key.h>
193 extern struct static_key_false rps_needed;
194 extern struct static_key_false rfs_needed;
195 #endif
196
197 struct neighbour;
198 struct neigh_parms;
199 struct sk_buff;
200
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
210 bool global_use;
211 int sync_cnt;
212 int refcount;
213 int synced;
214 struct rcu_head rcu_head;
215 };
216
217 struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220 };
221
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236
237 struct hh_cache {
238 unsigned int hh_len;
239 seqlock_t hh_lock;
240
241 /* cached hardware header; allow for machine alignment needs. */
242 #define HH_DATA_MOD 16
243 #define HH_DATA_OFF(__len) \
244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
245 #define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248 };
249
250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
251 * Alternative is:
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 *
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
257 */
258 #define LL_RESERVED_SPACE(dev) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262
263 struct header_ops {
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
266 const void *saddr, unsigned int len);
267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
272 bool (*validate)(const char *ll_header, unsigned int len);
273 __be16 (*parse_protocol)(const struct sk_buff *skb);
274 };
275
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
278 * code.
279 */
280
281 enum netdev_state_t {
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
287 };
288
289
290 /*
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
293 */
294 struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297 };
298 #define NETDEV_BOOT_SETUP_MAX 8
299
300 int __init netdev_boot_setup(char *str);
301
302 struct gro_list {
303 struct list_head list;
304 int count;
305 };
306
307 /*
308 * size of gro hash buckets, must less than bit number of
309 * napi_struct::gro_bitmask
310 */
311 #define GRO_HASH_BUCKETS 8
312
313 /*
314 * Structure for NAPI scheduling similar to tasklet but with weighting
315 */
316 struct napi_struct {
317 /* The poll_list must only be managed by the entity which
318 * changes the state of the NAPI_STATE_SCHED bit. This means
319 * whoever atomically sets that bit can add this napi_struct
320 * to the per-CPU poll_list, and whoever clears that bit
321 * can remove from the list right before clearing the bit.
322 */
323 struct list_head poll_list;
324
325 unsigned long state;
326 int weight;
327 unsigned long gro_bitmask;
328 int (*poll)(struct napi_struct *, int);
329 #ifdef CONFIG_NETPOLL
330 int poll_owner;
331 #endif
332 struct net_device *dev;
333 struct gro_list gro_hash[GRO_HASH_BUCKETS];
334 struct sk_buff *skb;
335 struct hrtimer timer;
336 struct list_head dev_list;
337 struct hlist_node napi_hash_node;
338 unsigned int napi_id;
339 };
340
341 enum {
342 NAPI_STATE_SCHED, /* Poll is scheduled */
343 NAPI_STATE_MISSED, /* reschedule a napi */
344 NAPI_STATE_DISABLE, /* Disable pending */
345 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
346 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
347 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
348 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
349 };
350
351 enum {
352 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
353 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
354 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
355 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
356 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
357 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
358 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
359 };
360
361 enum gro_result {
362 GRO_MERGED,
363 GRO_MERGED_FREE,
364 GRO_HELD,
365 GRO_NORMAL,
366 GRO_DROP,
367 GRO_CONSUMED,
368 };
369 typedef enum gro_result gro_result_t;
370
371 /*
372 * enum rx_handler_result - Possible return values for rx_handlers.
373 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
374 * further.
375 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
376 * case skb->dev was changed by rx_handler.
377 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
378 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
379 *
380 * rx_handlers are functions called from inside __netif_receive_skb(), to do
381 * special processing of the skb, prior to delivery to protocol handlers.
382 *
383 * Currently, a net_device can only have a single rx_handler registered. Trying
384 * to register a second rx_handler will return -EBUSY.
385 *
386 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
387 * To unregister a rx_handler on a net_device, use
388 * netdev_rx_handler_unregister().
389 *
390 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
391 * do with the skb.
392 *
393 * If the rx_handler consumed the skb in some way, it should return
394 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
395 * the skb to be delivered in some other way.
396 *
397 * If the rx_handler changed skb->dev, to divert the skb to another
398 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
399 * new device will be called if it exists.
400 *
401 * If the rx_handler decides the skb should be ignored, it should return
402 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
403 * are registered on exact device (ptype->dev == skb->dev).
404 *
405 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
406 * delivered, it should return RX_HANDLER_PASS.
407 *
408 * A device without a registered rx_handler will behave as if rx_handler
409 * returned RX_HANDLER_PASS.
410 */
411
412 enum rx_handler_result {
413 RX_HANDLER_CONSUMED,
414 RX_HANDLER_ANOTHER,
415 RX_HANDLER_EXACT,
416 RX_HANDLER_PASS,
417 };
418 typedef enum rx_handler_result rx_handler_result_t;
419 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
420
421 void __napi_schedule(struct napi_struct *n);
422 void __napi_schedule_irqoff(struct napi_struct *n);
423
424 static inline bool napi_disable_pending(struct napi_struct *n)
425 {
426 return test_bit(NAPI_STATE_DISABLE, &n->state);
427 }
428
429 bool napi_schedule_prep(struct napi_struct *n);
430
431 /**
432 * napi_schedule - schedule NAPI poll
433 * @n: NAPI context
434 *
435 * Schedule NAPI poll routine to be called if it is not already
436 * running.
437 */
438 static inline void napi_schedule(struct napi_struct *n)
439 {
440 if (napi_schedule_prep(n))
441 __napi_schedule(n);
442 }
443
444 /**
445 * napi_schedule_irqoff - schedule NAPI poll
446 * @n: NAPI context
447 *
448 * Variant of napi_schedule(), assuming hard irqs are masked.
449 */
450 static inline void napi_schedule_irqoff(struct napi_struct *n)
451 {
452 if (napi_schedule_prep(n))
453 __napi_schedule_irqoff(n);
454 }
455
456 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
457 static inline bool napi_reschedule(struct napi_struct *napi)
458 {
459 if (napi_schedule_prep(napi)) {
460 __napi_schedule(napi);
461 return true;
462 }
463 return false;
464 }
465
466 bool napi_complete_done(struct napi_struct *n, int work_done);
467 /**
468 * napi_complete - NAPI processing complete
469 * @n: NAPI context
470 *
471 * Mark NAPI processing as complete.
472 * Consider using napi_complete_done() instead.
473 * Return false if device should avoid rearming interrupts.
474 */
475 static inline bool napi_complete(struct napi_struct *n)
476 {
477 return napi_complete_done(n, 0);
478 }
479
480 /**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: NAPI context
483 *
484 * Warning: caller must observe RCU grace period
485 * before freeing memory containing @napi, if
486 * this function returns true.
487 * Note: core networking stack automatically calls it
488 * from netif_napi_del().
489 * Drivers might want to call this helper to combine all
490 * the needed RCU grace periods into a single one.
491 */
492 bool napi_hash_del(struct napi_struct *napi);
493
494 /**
495 * napi_disable - prevent NAPI from scheduling
496 * @n: NAPI context
497 *
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
500 */
501 void napi_disable(struct napi_struct *n);
502
503 /**
504 * napi_enable - enable NAPI scheduling
505 * @n: NAPI context
506 *
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
509 */
510 static inline void napi_enable(struct napi_struct *n)
511 {
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
513 smp_mb__before_atomic();
514 clear_bit(NAPI_STATE_SCHED, &n->state);
515 clear_bit(NAPI_STATE_NPSVC, &n->state);
516 }
517
518 /**
519 * napi_synchronize - wait until NAPI is not running
520 * @n: NAPI context
521 *
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
525 */
526 static inline void napi_synchronize(const struct napi_struct *n)
527 {
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
530 msleep(1);
531 else
532 barrier();
533 }
534
535 /**
536 * napi_if_scheduled_mark_missed - if napi is running, set the
537 * NAPIF_STATE_MISSED
538 * @n: NAPI context
539 *
540 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
541 * NAPI is scheduled.
542 **/
543 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
544 {
545 unsigned long val, new;
546
547 do {
548 val = READ_ONCE(n->state);
549 if (val & NAPIF_STATE_DISABLE)
550 return true;
551
552 if (!(val & NAPIF_STATE_SCHED))
553 return false;
554
555 new = val | NAPIF_STATE_MISSED;
556 } while (cmpxchg(&n->state, val, new) != val);
557
558 return true;
559 }
560
561 enum netdev_queue_state_t {
562 __QUEUE_STATE_DRV_XOFF,
563 __QUEUE_STATE_STACK_XOFF,
564 __QUEUE_STATE_FROZEN,
565 };
566
567 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
568 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
569 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
570
571 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
572 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
573 QUEUE_STATE_FROZEN)
574 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
575 QUEUE_STATE_FROZEN)
576
577 /*
578 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
579 * netif_tx_* functions below are used to manipulate this flag. The
580 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
581 * queue independently. The netif_xmit_*stopped functions below are called
582 * to check if the queue has been stopped by the driver or stack (either
583 * of the XOFF bits are set in the state). Drivers should not need to call
584 * netif_xmit*stopped functions, they should only be using netif_tx_*.
585 */
586
587 struct netdev_queue {
588 /*
589 * read-mostly part
590 */
591 struct net_device *dev;
592 struct Qdisc __rcu *qdisc;
593 struct Qdisc *qdisc_sleeping;
594 #ifdef CONFIG_SYSFS
595 struct kobject kobj;
596 #endif
597 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 int numa_node;
599 #endif
600 unsigned long tx_maxrate;
601 /*
602 * Number of TX timeouts for this queue
603 * (/sys/class/net/DEV/Q/trans_timeout)
604 */
605 unsigned long trans_timeout;
606
607 /* Subordinate device that the queue has been assigned to */
608 struct net_device *sb_dev;
609 #ifdef CONFIG_XDP_SOCKETS
610 struct xdp_umem *umem;
611 #endif
612 /*
613 * write-mostly part
614 */
615 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
616 int xmit_lock_owner;
617 /*
618 * Time (in jiffies) of last Tx
619 */
620 unsigned long trans_start;
621
622 unsigned long state;
623
624 #ifdef CONFIG_BQL
625 struct dql dql;
626 #endif
627 } ____cacheline_aligned_in_smp;
628
629 extern int sysctl_fb_tunnels_only_for_init_net;
630 extern int sysctl_devconf_inherit_init_net;
631
632 static inline bool net_has_fallback_tunnels(const struct net *net)
633 {
634 return net == &init_net ||
635 !IS_ENABLED(CONFIG_SYSCTL) ||
636 !sysctl_fb_tunnels_only_for_init_net;
637 }
638
639 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
640 {
641 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
642 return q->numa_node;
643 #else
644 return NUMA_NO_NODE;
645 #endif
646 }
647
648 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
649 {
650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
651 q->numa_node = node;
652 #endif
653 }
654
655 #ifdef CONFIG_RPS
656 /*
657 * This structure holds an RPS map which can be of variable length. The
658 * map is an array of CPUs.
659 */
660 struct rps_map {
661 unsigned int len;
662 struct rcu_head rcu;
663 u16 cpus[0];
664 };
665 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
666
667 /*
668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
669 * tail pointer for that CPU's input queue at the time of last enqueue, and
670 * a hardware filter index.
671 */
672 struct rps_dev_flow {
673 u16 cpu;
674 u16 filter;
675 unsigned int last_qtail;
676 };
677 #define RPS_NO_FILTER 0xffff
678
679 /*
680 * The rps_dev_flow_table structure contains a table of flow mappings.
681 */
682 struct rps_dev_flow_table {
683 unsigned int mask;
684 struct rcu_head rcu;
685 struct rps_dev_flow flows[0];
686 };
687 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
688 ((_num) * sizeof(struct rps_dev_flow)))
689
690 /*
691 * The rps_sock_flow_table contains mappings of flows to the last CPU
692 * on which they were processed by the application (set in recvmsg).
693 * Each entry is a 32bit value. Upper part is the high-order bits
694 * of flow hash, lower part is CPU number.
695 * rps_cpu_mask is used to partition the space, depending on number of
696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
698 * meaning we use 32-6=26 bits for the hash.
699 */
700 struct rps_sock_flow_table {
701 u32 mask;
702
703 u32 ents[0] ____cacheline_aligned_in_smp;
704 };
705 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
706
707 #define RPS_NO_CPU 0xffff
708
709 extern u32 rps_cpu_mask;
710 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
711
712 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
713 u32 hash)
714 {
715 if (table && hash) {
716 unsigned int index = hash & table->mask;
717 u32 val = hash & ~rps_cpu_mask;
718
719 /* We only give a hint, preemption can change CPU under us */
720 val |= raw_smp_processor_id();
721
722 if (table->ents[index] != val)
723 table->ents[index] = val;
724 }
725 }
726
727 #ifdef CONFIG_RFS_ACCEL
728 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
729 u16 filter_id);
730 #endif
731 #endif /* CONFIG_RPS */
732
733 /* This structure contains an instance of an RX queue. */
734 struct netdev_rx_queue {
735 #ifdef CONFIG_RPS
736 struct rps_map __rcu *rps_map;
737 struct rps_dev_flow_table __rcu *rps_flow_table;
738 #endif
739 struct kobject kobj;
740 struct net_device *dev;
741 struct xdp_rxq_info xdp_rxq;
742 #ifdef CONFIG_XDP_SOCKETS
743 struct xdp_umem *umem;
744 #endif
745 } ____cacheline_aligned_in_smp;
746
747 /*
748 * RX queue sysfs structures and functions.
749 */
750 struct rx_queue_attribute {
751 struct attribute attr;
752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
753 ssize_t (*store)(struct netdev_rx_queue *queue,
754 const char *buf, size_t len);
755 };
756
757 #ifdef CONFIG_XPS
758 /*
759 * This structure holds an XPS map which can be of variable length. The
760 * map is an array of queues.
761 */
762 struct xps_map {
763 unsigned int len;
764 unsigned int alloc_len;
765 struct rcu_head rcu;
766 u16 queues[0];
767 };
768 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
769 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
770 - sizeof(struct xps_map)) / sizeof(u16))
771
772 /*
773 * This structure holds all XPS maps for device. Maps are indexed by CPU.
774 */
775 struct xps_dev_maps {
776 struct rcu_head rcu;
777 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
778 };
779
780 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
781 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
782
783 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
784 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
785
786 #endif /* CONFIG_XPS */
787
788 #define TC_MAX_QUEUE 16
789 #define TC_BITMASK 15
790 /* HW offloaded queuing disciplines txq count and offset maps */
791 struct netdev_tc_txq {
792 u16 count;
793 u16 offset;
794 };
795
796 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
797 /*
798 * This structure is to hold information about the device
799 * configured to run FCoE protocol stack.
800 */
801 struct netdev_fcoe_hbainfo {
802 char manufacturer[64];
803 char serial_number[64];
804 char hardware_version[64];
805 char driver_version[64];
806 char optionrom_version[64];
807 char firmware_version[64];
808 char model[256];
809 char model_description[256];
810 };
811 #endif
812
813 #define MAX_PHYS_ITEM_ID_LEN 32
814
815 /* This structure holds a unique identifier to identify some
816 * physical item (port for example) used by a netdevice.
817 */
818 struct netdev_phys_item_id {
819 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
820 unsigned char id_len;
821 };
822
823 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
824 struct netdev_phys_item_id *b)
825 {
826 return a->id_len == b->id_len &&
827 memcmp(a->id, b->id, a->id_len) == 0;
828 }
829
830 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
831 struct sk_buff *skb,
832 struct net_device *sb_dev);
833
834 enum tc_setup_type {
835 TC_SETUP_QDISC_MQPRIO,
836 TC_SETUP_CLSU32,
837 TC_SETUP_CLSFLOWER,
838 TC_SETUP_CLSMATCHALL,
839 TC_SETUP_CLSBPF,
840 TC_SETUP_BLOCK,
841 TC_SETUP_QDISC_CBS,
842 TC_SETUP_QDISC_RED,
843 TC_SETUP_QDISC_PRIO,
844 TC_SETUP_QDISC_MQ,
845 TC_SETUP_QDISC_ETF,
846 TC_SETUP_ROOT_QDISC,
847 TC_SETUP_QDISC_GRED,
848 };
849
850 /* These structures hold the attributes of bpf state that are being passed
851 * to the netdevice through the bpf op.
852 */
853 enum bpf_netdev_command {
854 /* Set or clear a bpf program used in the earliest stages of packet
855 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
856 * is responsible for calling bpf_prog_put on any old progs that are
857 * stored. In case of error, the callee need not release the new prog
858 * reference, but on success it takes ownership and must bpf_prog_put
859 * when it is no longer used.
860 */
861 XDP_SETUP_PROG,
862 XDP_SETUP_PROG_HW,
863 XDP_QUERY_PROG,
864 XDP_QUERY_PROG_HW,
865 /* BPF program for offload callbacks, invoked at program load time. */
866 BPF_OFFLOAD_MAP_ALLOC,
867 BPF_OFFLOAD_MAP_FREE,
868 XDP_SETUP_XSK_UMEM,
869 };
870
871 struct bpf_prog_offload_ops;
872 struct netlink_ext_ack;
873 struct xdp_umem;
874
875 struct netdev_bpf {
876 enum bpf_netdev_command command;
877 union {
878 /* XDP_SETUP_PROG */
879 struct {
880 u32 flags;
881 struct bpf_prog *prog;
882 struct netlink_ext_ack *extack;
883 };
884 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
885 struct {
886 u32 prog_id;
887 /* flags with which program was installed */
888 u32 prog_flags;
889 };
890 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
891 struct {
892 struct bpf_offloaded_map *offmap;
893 };
894 /* XDP_SETUP_XSK_UMEM */
895 struct {
896 struct xdp_umem *umem;
897 u16 queue_id;
898 } xsk;
899 };
900 };
901
902 #ifdef CONFIG_XFRM_OFFLOAD
903 struct xfrmdev_ops {
904 int (*xdo_dev_state_add) (struct xfrm_state *x);
905 void (*xdo_dev_state_delete) (struct xfrm_state *x);
906 void (*xdo_dev_state_free) (struct xfrm_state *x);
907 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
908 struct xfrm_state *x);
909 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
910 };
911 #endif
912
913 struct dev_ifalias {
914 struct rcu_head rcuhead;
915 char ifalias[];
916 };
917
918 struct devlink;
919 struct tlsdev_ops;
920
921 /*
922 * This structure defines the management hooks for network devices.
923 * The following hooks can be defined; unless noted otherwise, they are
924 * optional and can be filled with a null pointer.
925 *
926 * int (*ndo_init)(struct net_device *dev);
927 * This function is called once when a network device is registered.
928 * The network device can use this for any late stage initialization
929 * or semantic validation. It can fail with an error code which will
930 * be propagated back to register_netdev.
931 *
932 * void (*ndo_uninit)(struct net_device *dev);
933 * This function is called when device is unregistered or when registration
934 * fails. It is not called if init fails.
935 *
936 * int (*ndo_open)(struct net_device *dev);
937 * This function is called when a network device transitions to the up
938 * state.
939 *
940 * int (*ndo_stop)(struct net_device *dev);
941 * This function is called when a network device transitions to the down
942 * state.
943 *
944 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
945 * struct net_device *dev);
946 * Called when a packet needs to be transmitted.
947 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
948 * the queue before that can happen; it's for obsolete devices and weird
949 * corner cases, but the stack really does a non-trivial amount
950 * of useless work if you return NETDEV_TX_BUSY.
951 * Required; cannot be NULL.
952 *
953 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
954 * struct net_device *dev
955 * netdev_features_t features);
956 * Called by core transmit path to determine if device is capable of
957 * performing offload operations on a given packet. This is to give
958 * the device an opportunity to implement any restrictions that cannot
959 * be otherwise expressed by feature flags. The check is called with
960 * the set of features that the stack has calculated and it returns
961 * those the driver believes to be appropriate.
962 *
963 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
964 * struct net_device *sb_dev);
965 * Called to decide which queue to use when device supports multiple
966 * transmit queues.
967 *
968 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
969 * This function is called to allow device receiver to make
970 * changes to configuration when multicast or promiscuous is enabled.
971 *
972 * void (*ndo_set_rx_mode)(struct net_device *dev);
973 * This function is called device changes address list filtering.
974 * If driver handles unicast address filtering, it should set
975 * IFF_UNICAST_FLT in its priv_flags.
976 *
977 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
978 * This function is called when the Media Access Control address
979 * needs to be changed. If this interface is not defined, the
980 * MAC address can not be changed.
981 *
982 * int (*ndo_validate_addr)(struct net_device *dev);
983 * Test if Media Access Control address is valid for the device.
984 *
985 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
986 * Called when a user requests an ioctl which can't be handled by
987 * the generic interface code. If not defined ioctls return
988 * not supported error code.
989 *
990 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
991 * Used to set network devices bus interface parameters. This interface
992 * is retained for legacy reasons; new devices should use the bus
993 * interface (PCI) for low level management.
994 *
995 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
996 * Called when a user wants to change the Maximum Transfer Unit
997 * of a device.
998 *
999 * void (*ndo_tx_timeout)(struct net_device *dev);
1000 * Callback used when the transmitter has not made any progress
1001 * for dev->watchdog ticks.
1002 *
1003 * void (*ndo_get_stats64)(struct net_device *dev,
1004 * struct rtnl_link_stats64 *storage);
1005 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1006 * Called when a user wants to get the network device usage
1007 * statistics. Drivers must do one of the following:
1008 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1009 * rtnl_link_stats64 structure passed by the caller.
1010 * 2. Define @ndo_get_stats to update a net_device_stats structure
1011 * (which should normally be dev->stats) and return a pointer to
1012 * it. The structure may be changed asynchronously only if each
1013 * field is written atomically.
1014 * 3. Update dev->stats asynchronously and atomically, and define
1015 * neither operation.
1016 *
1017 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1018 * Return true if this device supports offload stats of this attr_id.
1019 *
1020 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1021 * void *attr_data)
1022 * Get statistics for offload operations by attr_id. Write it into the
1023 * attr_data pointer.
1024 *
1025 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1026 * If device supports VLAN filtering this function is called when a
1027 * VLAN id is registered.
1028 *
1029 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1030 * If device supports VLAN filtering this function is called when a
1031 * VLAN id is unregistered.
1032 *
1033 * void (*ndo_poll_controller)(struct net_device *dev);
1034 *
1035 * SR-IOV management functions.
1036 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1037 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1038 * u8 qos, __be16 proto);
1039 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1040 * int max_tx_rate);
1041 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1042 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1043 * int (*ndo_get_vf_config)(struct net_device *dev,
1044 * int vf, struct ifla_vf_info *ivf);
1045 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1046 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1047 * struct nlattr *port[]);
1048 *
1049 * Enable or disable the VF ability to query its RSS Redirection Table and
1050 * Hash Key. This is needed since on some devices VF share this information
1051 * with PF and querying it may introduce a theoretical security risk.
1052 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1053 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1054 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1055 * void *type_data);
1056 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1057 * This is always called from the stack with the rtnl lock held and netif
1058 * tx queues stopped. This allows the netdevice to perform queue
1059 * management safely.
1060 *
1061 * Fiber Channel over Ethernet (FCoE) offload functions.
1062 * int (*ndo_fcoe_enable)(struct net_device *dev);
1063 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1064 * so the underlying device can perform whatever needed configuration or
1065 * initialization to support acceleration of FCoE traffic.
1066 *
1067 * int (*ndo_fcoe_disable)(struct net_device *dev);
1068 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1069 * so the underlying device can perform whatever needed clean-ups to
1070 * stop supporting acceleration of FCoE traffic.
1071 *
1072 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1073 * struct scatterlist *sgl, unsigned int sgc);
1074 * Called when the FCoE Initiator wants to initialize an I/O that
1075 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1076 * perform necessary setup and returns 1 to indicate the device is set up
1077 * successfully to perform DDP on this I/O, otherwise this returns 0.
1078 *
1079 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1080 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1081 * indicated by the FC exchange id 'xid', so the underlying device can
1082 * clean up and reuse resources for later DDP requests.
1083 *
1084 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1085 * struct scatterlist *sgl, unsigned int sgc);
1086 * Called when the FCoE Target wants to initialize an I/O that
1087 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1088 * perform necessary setup and returns 1 to indicate the device is set up
1089 * successfully to perform DDP on this I/O, otherwise this returns 0.
1090 *
1091 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1092 * struct netdev_fcoe_hbainfo *hbainfo);
1093 * Called when the FCoE Protocol stack wants information on the underlying
1094 * device. This information is utilized by the FCoE protocol stack to
1095 * register attributes with Fiber Channel management service as per the
1096 * FC-GS Fabric Device Management Information(FDMI) specification.
1097 *
1098 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1099 * Called when the underlying device wants to override default World Wide
1100 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1101 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1102 * protocol stack to use.
1103 *
1104 * RFS acceleration.
1105 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1106 * u16 rxq_index, u32 flow_id);
1107 * Set hardware filter for RFS. rxq_index is the target queue index;
1108 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1109 * Return the filter ID on success, or a negative error code.
1110 *
1111 * Slave management functions (for bridge, bonding, etc).
1112 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1113 * Called to make another netdev an underling.
1114 *
1115 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1116 * Called to release previously enslaved netdev.
1117 *
1118 * Feature/offload setting functions.
1119 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1120 * netdev_features_t features);
1121 * Adjusts the requested feature flags according to device-specific
1122 * constraints, and returns the resulting flags. Must not modify
1123 * the device state.
1124 *
1125 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1126 * Called to update device configuration to new features. Passed
1127 * feature set might be less than what was returned by ndo_fix_features()).
1128 * Must return >0 or -errno if it changed dev->features itself.
1129 *
1130 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1131 * struct net_device *dev,
1132 * const unsigned char *addr, u16 vid, u16 flags,
1133 * struct netlink_ext_ack *extack);
1134 * Adds an FDB entry to dev for addr.
1135 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1136 * struct net_device *dev,
1137 * const unsigned char *addr, u16 vid)
1138 * Deletes the FDB entry from dev coresponding to addr.
1139 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1140 * struct net_device *dev, struct net_device *filter_dev,
1141 * int *idx)
1142 * Used to add FDB entries to dump requests. Implementers should add
1143 * entries to skb and update idx with the number of entries.
1144 *
1145 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1146 * u16 flags, struct netlink_ext_ack *extack)
1147 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1148 * struct net_device *dev, u32 filter_mask,
1149 * int nlflags)
1150 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1151 * u16 flags);
1152 *
1153 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1154 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1155 * which do not represent real hardware may define this to allow their
1156 * userspace components to manage their virtual carrier state. Devices
1157 * that determine carrier state from physical hardware properties (eg
1158 * network cables) or protocol-dependent mechanisms (eg
1159 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1160 *
1161 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1162 * struct netdev_phys_item_id *ppid);
1163 * Called to get ID of physical port of this device. If driver does
1164 * not implement this, it is assumed that the hw is not able to have
1165 * multiple net devices on single physical port.
1166 *
1167 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1168 * struct netdev_phys_item_id *ppid)
1169 * Called to get the parent ID of the physical port of this device.
1170 *
1171 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1172 * struct udp_tunnel_info *ti);
1173 * Called by UDP tunnel to notify a driver about the UDP port and socket
1174 * address family that a UDP tunnel is listnening to. It is called only
1175 * when a new port starts listening. The operation is protected by the
1176 * RTNL.
1177 *
1178 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1179 * struct udp_tunnel_info *ti);
1180 * Called by UDP tunnel to notify the driver about a UDP port and socket
1181 * address family that the UDP tunnel is not listening to anymore. The
1182 * operation is protected by the RTNL.
1183 *
1184 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1185 * struct net_device *dev)
1186 * Called by upper layer devices to accelerate switching or other
1187 * station functionality into hardware. 'pdev is the lowerdev
1188 * to use for the offload and 'dev' is the net device that will
1189 * back the offload. Returns a pointer to the private structure
1190 * the upper layer will maintain.
1191 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1192 * Called by upper layer device to delete the station created
1193 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1194 * the station and priv is the structure returned by the add
1195 * operation.
1196 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1197 * int queue_index, u32 maxrate);
1198 * Called when a user wants to set a max-rate limitation of specific
1199 * TX queue.
1200 * int (*ndo_get_iflink)(const struct net_device *dev);
1201 * Called to get the iflink value of this device.
1202 * void (*ndo_change_proto_down)(struct net_device *dev,
1203 * bool proto_down);
1204 * This function is used to pass protocol port error state information
1205 * to the switch driver. The switch driver can react to the proto_down
1206 * by doing a phys down on the associated switch port.
1207 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1208 * This function is used to get egress tunnel information for given skb.
1209 * This is useful for retrieving outer tunnel header parameters while
1210 * sampling packet.
1211 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1212 * This function is used to specify the headroom that the skb must
1213 * consider when allocation skb during packet reception. Setting
1214 * appropriate rx headroom value allows avoiding skb head copy on
1215 * forward. Setting a negative value resets the rx headroom to the
1216 * default value.
1217 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1218 * This function is used to set or query state related to XDP on the
1219 * netdevice and manage BPF offload. See definition of
1220 * enum bpf_netdev_command for details.
1221 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1222 * u32 flags);
1223 * This function is used to submit @n XDP packets for transmit on a
1224 * netdevice. Returns number of frames successfully transmitted, frames
1225 * that got dropped are freed/returned via xdp_return_frame().
1226 * Returns negative number, means general error invoking ndo, meaning
1227 * no frames were xmit'ed and core-caller will free all frames.
1228 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1229 * Get devlink port instance associated with a given netdev.
1230 * Called with a reference on the netdevice and devlink locks only,
1231 * rtnl_lock is not held.
1232 */
1233 struct net_device_ops {
1234 int (*ndo_init)(struct net_device *dev);
1235 void (*ndo_uninit)(struct net_device *dev);
1236 int (*ndo_open)(struct net_device *dev);
1237 int (*ndo_stop)(struct net_device *dev);
1238 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1239 struct net_device *dev);
1240 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1241 struct net_device *dev,
1242 netdev_features_t features);
1243 u16 (*ndo_select_queue)(struct net_device *dev,
1244 struct sk_buff *skb,
1245 struct net_device *sb_dev);
1246 void (*ndo_change_rx_flags)(struct net_device *dev,
1247 int flags);
1248 void (*ndo_set_rx_mode)(struct net_device *dev);
1249 int (*ndo_set_mac_address)(struct net_device *dev,
1250 void *addr);
1251 int (*ndo_validate_addr)(struct net_device *dev);
1252 int (*ndo_do_ioctl)(struct net_device *dev,
1253 struct ifreq *ifr, int cmd);
1254 int (*ndo_set_config)(struct net_device *dev,
1255 struct ifmap *map);
1256 int (*ndo_change_mtu)(struct net_device *dev,
1257 int new_mtu);
1258 int (*ndo_neigh_setup)(struct net_device *dev,
1259 struct neigh_parms *);
1260 void (*ndo_tx_timeout) (struct net_device *dev);
1261
1262 void (*ndo_get_stats64)(struct net_device *dev,
1263 struct rtnl_link_stats64 *storage);
1264 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1265 int (*ndo_get_offload_stats)(int attr_id,
1266 const struct net_device *dev,
1267 void *attr_data);
1268 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1269
1270 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1271 __be16 proto, u16 vid);
1272 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1273 __be16 proto, u16 vid);
1274 #ifdef CONFIG_NET_POLL_CONTROLLER
1275 void (*ndo_poll_controller)(struct net_device *dev);
1276 int (*ndo_netpoll_setup)(struct net_device *dev,
1277 struct netpoll_info *info);
1278 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1279 #endif
1280 int (*ndo_set_vf_mac)(struct net_device *dev,
1281 int queue, u8 *mac);
1282 int (*ndo_set_vf_vlan)(struct net_device *dev,
1283 int queue, u16 vlan,
1284 u8 qos, __be16 proto);
1285 int (*ndo_set_vf_rate)(struct net_device *dev,
1286 int vf, int min_tx_rate,
1287 int max_tx_rate);
1288 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1289 int vf, bool setting);
1290 int (*ndo_set_vf_trust)(struct net_device *dev,
1291 int vf, bool setting);
1292 int (*ndo_get_vf_config)(struct net_device *dev,
1293 int vf,
1294 struct ifla_vf_info *ivf);
1295 int (*ndo_set_vf_link_state)(struct net_device *dev,
1296 int vf, int link_state);
1297 int (*ndo_get_vf_stats)(struct net_device *dev,
1298 int vf,
1299 struct ifla_vf_stats
1300 *vf_stats);
1301 int (*ndo_set_vf_port)(struct net_device *dev,
1302 int vf,
1303 struct nlattr *port[]);
1304 int (*ndo_get_vf_port)(struct net_device *dev,
1305 int vf, struct sk_buff *skb);
1306 int (*ndo_set_vf_guid)(struct net_device *dev,
1307 int vf, u64 guid,
1308 int guid_type);
1309 int (*ndo_set_vf_rss_query_en)(
1310 struct net_device *dev,
1311 int vf, bool setting);
1312 int (*ndo_setup_tc)(struct net_device *dev,
1313 enum tc_setup_type type,
1314 void *type_data);
1315 #if IS_ENABLED(CONFIG_FCOE)
1316 int (*ndo_fcoe_enable)(struct net_device *dev);
1317 int (*ndo_fcoe_disable)(struct net_device *dev);
1318 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1319 u16 xid,
1320 struct scatterlist *sgl,
1321 unsigned int sgc);
1322 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1323 u16 xid);
1324 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1325 u16 xid,
1326 struct scatterlist *sgl,
1327 unsigned int sgc);
1328 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1329 struct netdev_fcoe_hbainfo *hbainfo);
1330 #endif
1331
1332 #if IS_ENABLED(CONFIG_LIBFCOE)
1333 #define NETDEV_FCOE_WWNN 0
1334 #define NETDEV_FCOE_WWPN 1
1335 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1336 u64 *wwn, int type);
1337 #endif
1338
1339 #ifdef CONFIG_RFS_ACCEL
1340 int (*ndo_rx_flow_steer)(struct net_device *dev,
1341 const struct sk_buff *skb,
1342 u16 rxq_index,
1343 u32 flow_id);
1344 #endif
1345 int (*ndo_add_slave)(struct net_device *dev,
1346 struct net_device *slave_dev,
1347 struct netlink_ext_ack *extack);
1348 int (*ndo_del_slave)(struct net_device *dev,
1349 struct net_device *slave_dev);
1350 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1351 netdev_features_t features);
1352 int (*ndo_set_features)(struct net_device *dev,
1353 netdev_features_t features);
1354 int (*ndo_neigh_construct)(struct net_device *dev,
1355 struct neighbour *n);
1356 void (*ndo_neigh_destroy)(struct net_device *dev,
1357 struct neighbour *n);
1358
1359 int (*ndo_fdb_add)(struct ndmsg *ndm,
1360 struct nlattr *tb[],
1361 struct net_device *dev,
1362 const unsigned char *addr,
1363 u16 vid,
1364 u16 flags,
1365 struct netlink_ext_ack *extack);
1366 int (*ndo_fdb_del)(struct ndmsg *ndm,
1367 struct nlattr *tb[],
1368 struct net_device *dev,
1369 const unsigned char *addr,
1370 u16 vid);
1371 int (*ndo_fdb_dump)(struct sk_buff *skb,
1372 struct netlink_callback *cb,
1373 struct net_device *dev,
1374 struct net_device *filter_dev,
1375 int *idx);
1376 int (*ndo_fdb_get)(struct sk_buff *skb,
1377 struct nlattr *tb[],
1378 struct net_device *dev,
1379 const unsigned char *addr,
1380 u16 vid, u32 portid, u32 seq,
1381 struct netlink_ext_ack *extack);
1382 int (*ndo_bridge_setlink)(struct net_device *dev,
1383 struct nlmsghdr *nlh,
1384 u16 flags,
1385 struct netlink_ext_ack *extack);
1386 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1387 u32 pid, u32 seq,
1388 struct net_device *dev,
1389 u32 filter_mask,
1390 int nlflags);
1391 int (*ndo_bridge_dellink)(struct net_device *dev,
1392 struct nlmsghdr *nlh,
1393 u16 flags);
1394 int (*ndo_change_carrier)(struct net_device *dev,
1395 bool new_carrier);
1396 int (*ndo_get_phys_port_id)(struct net_device *dev,
1397 struct netdev_phys_item_id *ppid);
1398 int (*ndo_get_port_parent_id)(struct net_device *dev,
1399 struct netdev_phys_item_id *ppid);
1400 int (*ndo_get_phys_port_name)(struct net_device *dev,
1401 char *name, size_t len);
1402 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1403 struct udp_tunnel_info *ti);
1404 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1405 struct udp_tunnel_info *ti);
1406 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1407 struct net_device *dev);
1408 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1409 void *priv);
1410
1411 int (*ndo_get_lock_subclass)(struct net_device *dev);
1412 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1413 int queue_index,
1414 u32 maxrate);
1415 int (*ndo_get_iflink)(const struct net_device *dev);
1416 int (*ndo_change_proto_down)(struct net_device *dev,
1417 bool proto_down);
1418 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1419 struct sk_buff *skb);
1420 void (*ndo_set_rx_headroom)(struct net_device *dev,
1421 int needed_headroom);
1422 int (*ndo_bpf)(struct net_device *dev,
1423 struct netdev_bpf *bpf);
1424 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1425 struct xdp_frame **xdp,
1426 u32 flags);
1427 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1428 u32 queue_id);
1429 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1430 };
1431
1432 /**
1433 * enum net_device_priv_flags - &struct net_device priv_flags
1434 *
1435 * These are the &struct net_device, they are only set internally
1436 * by drivers and used in the kernel. These flags are invisible to
1437 * userspace; this means that the order of these flags can change
1438 * during any kernel release.
1439 *
1440 * You should have a pretty good reason to be extending these flags.
1441 *
1442 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1443 * @IFF_EBRIDGE: Ethernet bridging device
1444 * @IFF_BONDING: bonding master or slave
1445 * @IFF_ISATAP: ISATAP interface (RFC4214)
1446 * @IFF_WAN_HDLC: WAN HDLC device
1447 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1448 * release skb->dst
1449 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1450 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1451 * @IFF_MACVLAN_PORT: device used as macvlan port
1452 * @IFF_BRIDGE_PORT: device used as bridge port
1453 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1454 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1455 * @IFF_UNICAST_FLT: Supports unicast filtering
1456 * @IFF_TEAM_PORT: device used as team port
1457 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1458 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1459 * change when it's running
1460 * @IFF_MACVLAN: Macvlan device
1461 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1462 * underlying stacked devices
1463 * @IFF_L3MDEV_MASTER: device is an L3 master device
1464 * @IFF_NO_QUEUE: device can run without qdisc attached
1465 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1466 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1467 * @IFF_TEAM: device is a team device
1468 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1469 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1470 * entity (i.e. the master device for bridged veth)
1471 * @IFF_MACSEC: device is a MACsec device
1472 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1473 * @IFF_FAILOVER: device is a failover master device
1474 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1475 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1476 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1477 */
1478 enum netdev_priv_flags {
1479 IFF_802_1Q_VLAN = 1<<0,
1480 IFF_EBRIDGE = 1<<1,
1481 IFF_BONDING = 1<<2,
1482 IFF_ISATAP = 1<<3,
1483 IFF_WAN_HDLC = 1<<4,
1484 IFF_XMIT_DST_RELEASE = 1<<5,
1485 IFF_DONT_BRIDGE = 1<<6,
1486 IFF_DISABLE_NETPOLL = 1<<7,
1487 IFF_MACVLAN_PORT = 1<<8,
1488 IFF_BRIDGE_PORT = 1<<9,
1489 IFF_OVS_DATAPATH = 1<<10,
1490 IFF_TX_SKB_SHARING = 1<<11,
1491 IFF_UNICAST_FLT = 1<<12,
1492 IFF_TEAM_PORT = 1<<13,
1493 IFF_SUPP_NOFCS = 1<<14,
1494 IFF_LIVE_ADDR_CHANGE = 1<<15,
1495 IFF_MACVLAN = 1<<16,
1496 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1497 IFF_L3MDEV_MASTER = 1<<18,
1498 IFF_NO_QUEUE = 1<<19,
1499 IFF_OPENVSWITCH = 1<<20,
1500 IFF_L3MDEV_SLAVE = 1<<21,
1501 IFF_TEAM = 1<<22,
1502 IFF_RXFH_CONFIGURED = 1<<23,
1503 IFF_PHONY_HEADROOM = 1<<24,
1504 IFF_MACSEC = 1<<25,
1505 IFF_NO_RX_HANDLER = 1<<26,
1506 IFF_FAILOVER = 1<<27,
1507 IFF_FAILOVER_SLAVE = 1<<28,
1508 IFF_L3MDEV_RX_HANDLER = 1<<29,
1509 IFF_LIVE_RENAME_OK = 1<<30,
1510 };
1511
1512 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1513 #define IFF_EBRIDGE IFF_EBRIDGE
1514 #define IFF_BONDING IFF_BONDING
1515 #define IFF_ISATAP IFF_ISATAP
1516 #define IFF_WAN_HDLC IFF_WAN_HDLC
1517 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1518 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1519 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1520 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1521 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1522 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1523 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1524 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1525 #define IFF_TEAM_PORT IFF_TEAM_PORT
1526 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1527 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1528 #define IFF_MACVLAN IFF_MACVLAN
1529 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1530 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1531 #define IFF_NO_QUEUE IFF_NO_QUEUE
1532 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1533 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1534 #define IFF_TEAM IFF_TEAM
1535 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1536 #define IFF_MACSEC IFF_MACSEC
1537 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1538 #define IFF_FAILOVER IFF_FAILOVER
1539 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1540 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1541 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1542
1543 /**
1544 * struct net_device - The DEVICE structure.
1545 *
1546 * Actually, this whole structure is a big mistake. It mixes I/O
1547 * data with strictly "high-level" data, and it has to know about
1548 * almost every data structure used in the INET module.
1549 *
1550 * @name: This is the first field of the "visible" part of this structure
1551 * (i.e. as seen by users in the "Space.c" file). It is the name
1552 * of the interface.
1553 *
1554 * @name_hlist: Device name hash chain, please keep it close to name[]
1555 * @ifalias: SNMP alias
1556 * @mem_end: Shared memory end
1557 * @mem_start: Shared memory start
1558 * @base_addr: Device I/O address
1559 * @irq: Device IRQ number
1560 *
1561 * @state: Generic network queuing layer state, see netdev_state_t
1562 * @dev_list: The global list of network devices
1563 * @napi_list: List entry used for polling NAPI devices
1564 * @unreg_list: List entry when we are unregistering the
1565 * device; see the function unregister_netdev
1566 * @close_list: List entry used when we are closing the device
1567 * @ptype_all: Device-specific packet handlers for all protocols
1568 * @ptype_specific: Device-specific, protocol-specific packet handlers
1569 *
1570 * @adj_list: Directly linked devices, like slaves for bonding
1571 * @features: Currently active device features
1572 * @hw_features: User-changeable features
1573 *
1574 * @wanted_features: User-requested features
1575 * @vlan_features: Mask of features inheritable by VLAN devices
1576 *
1577 * @hw_enc_features: Mask of features inherited by encapsulating devices
1578 * This field indicates what encapsulation
1579 * offloads the hardware is capable of doing,
1580 * and drivers will need to set them appropriately.
1581 *
1582 * @mpls_features: Mask of features inheritable by MPLS
1583 *
1584 * @ifindex: interface index
1585 * @group: The group the device belongs to
1586 *
1587 * @stats: Statistics struct, which was left as a legacy, use
1588 * rtnl_link_stats64 instead
1589 *
1590 * @rx_dropped: Dropped packets by core network,
1591 * do not use this in drivers
1592 * @tx_dropped: Dropped packets by core network,
1593 * do not use this in drivers
1594 * @rx_nohandler: nohandler dropped packets by core network on
1595 * inactive devices, do not use this in drivers
1596 * @carrier_up_count: Number of times the carrier has been up
1597 * @carrier_down_count: Number of times the carrier has been down
1598 *
1599 * @wireless_handlers: List of functions to handle Wireless Extensions,
1600 * instead of ioctl,
1601 * see <net/iw_handler.h> for details.
1602 * @wireless_data: Instance data managed by the core of wireless extensions
1603 *
1604 * @netdev_ops: Includes several pointers to callbacks,
1605 * if one wants to override the ndo_*() functions
1606 * @ethtool_ops: Management operations
1607 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1608 * discovery handling. Necessary for e.g. 6LoWPAN.
1609 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1610 * of Layer 2 headers.
1611 *
1612 * @flags: Interface flags (a la BSD)
1613 * @priv_flags: Like 'flags' but invisible to userspace,
1614 * see if.h for the definitions
1615 * @gflags: Global flags ( kept as legacy )
1616 * @padded: How much padding added by alloc_netdev()
1617 * @operstate: RFC2863 operstate
1618 * @link_mode: Mapping policy to operstate
1619 * @if_port: Selectable AUI, TP, ...
1620 * @dma: DMA channel
1621 * @mtu: Interface MTU value
1622 * @min_mtu: Interface Minimum MTU value
1623 * @max_mtu: Interface Maximum MTU value
1624 * @type: Interface hardware type
1625 * @hard_header_len: Maximum hardware header length.
1626 * @min_header_len: Minimum hardware header length
1627 *
1628 * @needed_headroom: Extra headroom the hardware may need, but not in all
1629 * cases can this be guaranteed
1630 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1631 * cases can this be guaranteed. Some cases also use
1632 * LL_MAX_HEADER instead to allocate the skb
1633 *
1634 * interface address info:
1635 *
1636 * @perm_addr: Permanent hw address
1637 * @addr_assign_type: Hw address assignment type
1638 * @addr_len: Hardware address length
1639 * @neigh_priv_len: Used in neigh_alloc()
1640 * @dev_id: Used to differentiate devices that share
1641 * the same link layer address
1642 * @dev_port: Used to differentiate devices that share
1643 * the same function
1644 * @addr_list_lock: XXX: need comments on this one
1645 * @uc_promisc: Counter that indicates promiscuous mode
1646 * has been enabled due to the need to listen to
1647 * additional unicast addresses in a device that
1648 * does not implement ndo_set_rx_mode()
1649 * @uc: unicast mac addresses
1650 * @mc: multicast mac addresses
1651 * @dev_addrs: list of device hw addresses
1652 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1653 * @promiscuity: Number of times the NIC is told to work in
1654 * promiscuous mode; if it becomes 0 the NIC will
1655 * exit promiscuous mode
1656 * @allmulti: Counter, enables or disables allmulticast mode
1657 *
1658 * @vlan_info: VLAN info
1659 * @dsa_ptr: dsa specific data
1660 * @tipc_ptr: TIPC specific data
1661 * @atalk_ptr: AppleTalk link
1662 * @ip_ptr: IPv4 specific data
1663 * @dn_ptr: DECnet specific data
1664 * @ip6_ptr: IPv6 specific data
1665 * @ax25_ptr: AX.25 specific data
1666 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1667 *
1668 * @dev_addr: Hw address (before bcast,
1669 * because most packets are unicast)
1670 *
1671 * @_rx: Array of RX queues
1672 * @num_rx_queues: Number of RX queues
1673 * allocated at register_netdev() time
1674 * @real_num_rx_queues: Number of RX queues currently active in device
1675 *
1676 * @rx_handler: handler for received packets
1677 * @rx_handler_data: XXX: need comments on this one
1678 * @miniq_ingress: ingress/clsact qdisc specific data for
1679 * ingress processing
1680 * @ingress_queue: XXX: need comments on this one
1681 * @broadcast: hw bcast address
1682 *
1683 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1684 * indexed by RX queue number. Assigned by driver.
1685 * This must only be set if the ndo_rx_flow_steer
1686 * operation is defined
1687 * @index_hlist: Device index hash chain
1688 *
1689 * @_tx: Array of TX queues
1690 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1691 * @real_num_tx_queues: Number of TX queues currently active in device
1692 * @qdisc: Root qdisc from userspace point of view
1693 * @tx_queue_len: Max frames per queue allowed
1694 * @tx_global_lock: XXX: need comments on this one
1695 *
1696 * @xps_maps: XXX: need comments on this one
1697 * @miniq_egress: clsact qdisc specific data for
1698 * egress processing
1699 * @watchdog_timeo: Represents the timeout that is used by
1700 * the watchdog (see dev_watchdog())
1701 * @watchdog_timer: List of timers
1702 *
1703 * @pcpu_refcnt: Number of references to this device
1704 * @todo_list: Delayed register/unregister
1705 * @link_watch_list: XXX: need comments on this one
1706 *
1707 * @reg_state: Register/unregister state machine
1708 * @dismantle: Device is going to be freed
1709 * @rtnl_link_state: This enum represents the phases of creating
1710 * a new link
1711 *
1712 * @needs_free_netdev: Should unregister perform free_netdev?
1713 * @priv_destructor: Called from unregister
1714 * @npinfo: XXX: need comments on this one
1715 * @nd_net: Network namespace this network device is inside
1716 *
1717 * @ml_priv: Mid-layer private
1718 * @lstats: Loopback statistics
1719 * @tstats: Tunnel statistics
1720 * @dstats: Dummy statistics
1721 * @vstats: Virtual ethernet statistics
1722 *
1723 * @garp_port: GARP
1724 * @mrp_port: MRP
1725 *
1726 * @dev: Class/net/name entry
1727 * @sysfs_groups: Space for optional device, statistics and wireless
1728 * sysfs groups
1729 *
1730 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1731 * @rtnl_link_ops: Rtnl_link_ops
1732 *
1733 * @gso_max_size: Maximum size of generic segmentation offload
1734 * @gso_max_segs: Maximum number of segments that can be passed to the
1735 * NIC for GSO
1736 *
1737 * @dcbnl_ops: Data Center Bridging netlink ops
1738 * @num_tc: Number of traffic classes in the net device
1739 * @tc_to_txq: XXX: need comments on this one
1740 * @prio_tc_map: XXX: need comments on this one
1741 *
1742 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1743 *
1744 * @priomap: XXX: need comments on this one
1745 * @phydev: Physical device may attach itself
1746 * for hardware timestamping
1747 * @sfp_bus: attached &struct sfp_bus structure.
1748 *
1749 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1750 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1751 *
1752 * @proto_down: protocol port state information can be sent to the
1753 * switch driver and used to set the phys state of the
1754 * switch port.
1755 *
1756 * @wol_enabled: Wake-on-LAN is enabled
1757 *
1758 * FIXME: cleanup struct net_device such that network protocol info
1759 * moves out.
1760 */
1761
1762 struct net_device {
1763 char name[IFNAMSIZ];
1764 struct hlist_node name_hlist;
1765 struct dev_ifalias __rcu *ifalias;
1766 /*
1767 * I/O specific fields
1768 * FIXME: Merge these and struct ifmap into one
1769 */
1770 unsigned long mem_end;
1771 unsigned long mem_start;
1772 unsigned long base_addr;
1773 int irq;
1774
1775 /*
1776 * Some hardware also needs these fields (state,dev_list,
1777 * napi_list,unreg_list,close_list) but they are not
1778 * part of the usual set specified in Space.c.
1779 */
1780
1781 unsigned long state;
1782
1783 struct list_head dev_list;
1784 struct list_head napi_list;
1785 struct list_head unreg_list;
1786 struct list_head close_list;
1787 struct list_head ptype_all;
1788 struct list_head ptype_specific;
1789
1790 struct {
1791 struct list_head upper;
1792 struct list_head lower;
1793 } adj_list;
1794
1795 netdev_features_t features;
1796 netdev_features_t hw_features;
1797 netdev_features_t wanted_features;
1798 netdev_features_t vlan_features;
1799 netdev_features_t hw_enc_features;
1800 netdev_features_t mpls_features;
1801 netdev_features_t gso_partial_features;
1802
1803 int ifindex;
1804 int group;
1805
1806 struct net_device_stats stats;
1807
1808 atomic_long_t rx_dropped;
1809 atomic_long_t tx_dropped;
1810 atomic_long_t rx_nohandler;
1811
1812 /* Stats to monitor link on/off, flapping */
1813 atomic_t carrier_up_count;
1814 atomic_t carrier_down_count;
1815
1816 #ifdef CONFIG_WIRELESS_EXT
1817 const struct iw_handler_def *wireless_handlers;
1818 struct iw_public_data *wireless_data;
1819 #endif
1820 const struct net_device_ops *netdev_ops;
1821 const struct ethtool_ops *ethtool_ops;
1822 #ifdef CONFIG_NET_L3_MASTER_DEV
1823 const struct l3mdev_ops *l3mdev_ops;
1824 #endif
1825 #if IS_ENABLED(CONFIG_IPV6)
1826 const struct ndisc_ops *ndisc_ops;
1827 #endif
1828
1829 #ifdef CONFIG_XFRM_OFFLOAD
1830 const struct xfrmdev_ops *xfrmdev_ops;
1831 #endif
1832
1833 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1834 const struct tlsdev_ops *tlsdev_ops;
1835 #endif
1836
1837 const struct header_ops *header_ops;
1838
1839 unsigned int flags;
1840 unsigned int priv_flags;
1841
1842 unsigned short gflags;
1843 unsigned short padded;
1844
1845 unsigned char operstate;
1846 unsigned char link_mode;
1847
1848 unsigned char if_port;
1849 unsigned char dma;
1850
1851 /* Note : dev->mtu is often read without holding a lock.
1852 * Writers usually hold RTNL.
1853 * It is recommended to use READ_ONCE() to annotate the reads,
1854 * and to use WRITE_ONCE() to annotate the writes.
1855 */
1856 unsigned int mtu;
1857 unsigned int min_mtu;
1858 unsigned int max_mtu;
1859 unsigned short type;
1860 unsigned short hard_header_len;
1861 unsigned char min_header_len;
1862
1863 unsigned short needed_headroom;
1864 unsigned short needed_tailroom;
1865
1866 /* Interface address info. */
1867 unsigned char perm_addr[MAX_ADDR_LEN];
1868 unsigned char addr_assign_type;
1869 unsigned char addr_len;
1870 unsigned short neigh_priv_len;
1871 unsigned short dev_id;
1872 unsigned short dev_port;
1873 spinlock_t addr_list_lock;
1874 unsigned char name_assign_type;
1875 bool uc_promisc;
1876 struct netdev_hw_addr_list uc;
1877 struct netdev_hw_addr_list mc;
1878 struct netdev_hw_addr_list dev_addrs;
1879
1880 #ifdef CONFIG_SYSFS
1881 struct kset *queues_kset;
1882 #endif
1883 unsigned int promiscuity;
1884 unsigned int allmulti;
1885
1886
1887 /* Protocol-specific pointers */
1888
1889 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1890 struct vlan_info __rcu *vlan_info;
1891 #endif
1892 #if IS_ENABLED(CONFIG_NET_DSA)
1893 struct dsa_port *dsa_ptr;
1894 #endif
1895 #if IS_ENABLED(CONFIG_TIPC)
1896 struct tipc_bearer __rcu *tipc_ptr;
1897 #endif
1898 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1899 void *atalk_ptr;
1900 #endif
1901 struct in_device __rcu *ip_ptr;
1902 #if IS_ENABLED(CONFIG_DECNET)
1903 struct dn_dev __rcu *dn_ptr;
1904 #endif
1905 struct inet6_dev __rcu *ip6_ptr;
1906 #if IS_ENABLED(CONFIG_AX25)
1907 void *ax25_ptr;
1908 #endif
1909 struct wireless_dev *ieee80211_ptr;
1910 struct wpan_dev *ieee802154_ptr;
1911 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1912 struct mpls_dev __rcu *mpls_ptr;
1913 #endif
1914
1915 /*
1916 * Cache lines mostly used on receive path (including eth_type_trans())
1917 */
1918 /* Interface address info used in eth_type_trans() */
1919 unsigned char *dev_addr;
1920
1921 struct netdev_rx_queue *_rx;
1922 unsigned int num_rx_queues;
1923 unsigned int real_num_rx_queues;
1924
1925 struct bpf_prog __rcu *xdp_prog;
1926 unsigned long gro_flush_timeout;
1927 rx_handler_func_t __rcu *rx_handler;
1928 void __rcu *rx_handler_data;
1929
1930 #ifdef CONFIG_NET_CLS_ACT
1931 struct mini_Qdisc __rcu *miniq_ingress;
1932 #endif
1933 struct netdev_queue __rcu *ingress_queue;
1934 #ifdef CONFIG_NETFILTER_INGRESS
1935 struct nf_hook_entries __rcu *nf_hooks_ingress;
1936 #endif
1937
1938 unsigned char broadcast[MAX_ADDR_LEN];
1939 #ifdef CONFIG_RFS_ACCEL
1940 struct cpu_rmap *rx_cpu_rmap;
1941 #endif
1942 struct hlist_node index_hlist;
1943
1944 /*
1945 * Cache lines mostly used on transmit path
1946 */
1947 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1948 unsigned int num_tx_queues;
1949 unsigned int real_num_tx_queues;
1950 struct Qdisc *qdisc;
1951 #ifdef CONFIG_NET_SCHED
1952 DECLARE_HASHTABLE (qdisc_hash, 4);
1953 #endif
1954 unsigned int tx_queue_len;
1955 spinlock_t tx_global_lock;
1956 int watchdog_timeo;
1957
1958 #ifdef CONFIG_XPS
1959 struct xps_dev_maps __rcu *xps_cpus_map;
1960 struct xps_dev_maps __rcu *xps_rxqs_map;
1961 #endif
1962 #ifdef CONFIG_NET_CLS_ACT
1963 struct mini_Qdisc __rcu *miniq_egress;
1964 #endif
1965
1966 /* These may be needed for future network-power-down code. */
1967 struct timer_list watchdog_timer;
1968
1969 int __percpu *pcpu_refcnt;
1970 struct list_head todo_list;
1971
1972 struct list_head link_watch_list;
1973
1974 enum { NETREG_UNINITIALIZED=0,
1975 NETREG_REGISTERED, /* completed register_netdevice */
1976 NETREG_UNREGISTERING, /* called unregister_netdevice */
1977 NETREG_UNREGISTERED, /* completed unregister todo */
1978 NETREG_RELEASED, /* called free_netdev */
1979 NETREG_DUMMY, /* dummy device for NAPI poll */
1980 } reg_state:8;
1981
1982 bool dismantle;
1983
1984 enum {
1985 RTNL_LINK_INITIALIZED,
1986 RTNL_LINK_INITIALIZING,
1987 } rtnl_link_state:16;
1988
1989 bool needs_free_netdev;
1990 void (*priv_destructor)(struct net_device *dev);
1991
1992 #ifdef CONFIG_NETPOLL
1993 struct netpoll_info __rcu *npinfo;
1994 #endif
1995
1996 possible_net_t nd_net;
1997
1998 /* mid-layer private */
1999 union {
2000 void *ml_priv;
2001 struct pcpu_lstats __percpu *lstats;
2002 struct pcpu_sw_netstats __percpu *tstats;
2003 struct pcpu_dstats __percpu *dstats;
2004 };
2005
2006 #if IS_ENABLED(CONFIG_GARP)
2007 struct garp_port __rcu *garp_port;
2008 #endif
2009 #if IS_ENABLED(CONFIG_MRP)
2010 struct mrp_port __rcu *mrp_port;
2011 #endif
2012
2013 struct device dev;
2014 const struct attribute_group *sysfs_groups[4];
2015 const struct attribute_group *sysfs_rx_queue_group;
2016
2017 const struct rtnl_link_ops *rtnl_link_ops;
2018
2019 /* for setting kernel sock attribute on TCP connection setup */
2020 #define GSO_MAX_SIZE 65536
2021 unsigned int gso_max_size;
2022 #define GSO_MAX_SEGS 65535
2023 u16 gso_max_segs;
2024
2025 #ifdef CONFIG_DCB
2026 const struct dcbnl_rtnl_ops *dcbnl_ops;
2027 #endif
2028 s16 num_tc;
2029 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2030 u8 prio_tc_map[TC_BITMASK + 1];
2031
2032 #if IS_ENABLED(CONFIG_FCOE)
2033 unsigned int fcoe_ddp_xid;
2034 #endif
2035 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2036 struct netprio_map __rcu *priomap;
2037 #endif
2038 struct phy_device *phydev;
2039 struct sfp_bus *sfp_bus;
2040 struct lock_class_key *qdisc_tx_busylock;
2041 struct lock_class_key *qdisc_running_key;
2042 bool proto_down;
2043 unsigned wol_enabled:1;
2044 };
2045 #define to_net_dev(d) container_of(d, struct net_device, dev)
2046
2047 static inline bool netif_elide_gro(const struct net_device *dev)
2048 {
2049 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2050 return true;
2051 return false;
2052 }
2053
2054 #define NETDEV_ALIGN 32
2055
2056 static inline
2057 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2058 {
2059 return dev->prio_tc_map[prio & TC_BITMASK];
2060 }
2061
2062 static inline
2063 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2064 {
2065 if (tc >= dev->num_tc)
2066 return -EINVAL;
2067
2068 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2069 return 0;
2070 }
2071
2072 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2073 void netdev_reset_tc(struct net_device *dev);
2074 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2075 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2076
2077 static inline
2078 int netdev_get_num_tc(struct net_device *dev)
2079 {
2080 return dev->num_tc;
2081 }
2082
2083 void netdev_unbind_sb_channel(struct net_device *dev,
2084 struct net_device *sb_dev);
2085 int netdev_bind_sb_channel_queue(struct net_device *dev,
2086 struct net_device *sb_dev,
2087 u8 tc, u16 count, u16 offset);
2088 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2089 static inline int netdev_get_sb_channel(struct net_device *dev)
2090 {
2091 return max_t(int, -dev->num_tc, 0);
2092 }
2093
2094 static inline
2095 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2096 unsigned int index)
2097 {
2098 return &dev->_tx[index];
2099 }
2100
2101 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2102 const struct sk_buff *skb)
2103 {
2104 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2105 }
2106
2107 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2108 void (*f)(struct net_device *,
2109 struct netdev_queue *,
2110 void *),
2111 void *arg)
2112 {
2113 unsigned int i;
2114
2115 for (i = 0; i < dev->num_tx_queues; i++)
2116 f(dev, &dev->_tx[i], arg);
2117 }
2118
2119 #define netdev_lockdep_set_classes(dev) \
2120 { \
2121 static struct lock_class_key qdisc_tx_busylock_key; \
2122 static struct lock_class_key qdisc_running_key; \
2123 static struct lock_class_key qdisc_xmit_lock_key; \
2124 static struct lock_class_key dev_addr_list_lock_key; \
2125 unsigned int i; \
2126 \
2127 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2128 (dev)->qdisc_running_key = &qdisc_running_key; \
2129 lockdep_set_class(&(dev)->addr_list_lock, \
2130 &dev_addr_list_lock_key); \
2131 for (i = 0; i < (dev)->num_tx_queues; i++) \
2132 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2133 &qdisc_xmit_lock_key); \
2134 }
2135
2136 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2137 struct net_device *sb_dev);
2138 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2139 struct sk_buff *skb,
2140 struct net_device *sb_dev);
2141
2142 /* returns the headroom that the master device needs to take in account
2143 * when forwarding to this dev
2144 */
2145 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2146 {
2147 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2148 }
2149
2150 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2151 {
2152 if (dev->netdev_ops->ndo_set_rx_headroom)
2153 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2154 }
2155
2156 /* set the device rx headroom to the dev's default */
2157 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2158 {
2159 netdev_set_rx_headroom(dev, -1);
2160 }
2161
2162 /*
2163 * Net namespace inlines
2164 */
2165 static inline
2166 struct net *dev_net(const struct net_device *dev)
2167 {
2168 return read_pnet(&dev->nd_net);
2169 }
2170
2171 static inline
2172 void dev_net_set(struct net_device *dev, struct net *net)
2173 {
2174 write_pnet(&dev->nd_net, net);
2175 }
2176
2177 /**
2178 * netdev_priv - access network device private data
2179 * @dev: network device
2180 *
2181 * Get network device private data
2182 */
2183 static inline void *netdev_priv(const struct net_device *dev)
2184 {
2185 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2186 }
2187
2188 /* Set the sysfs physical device reference for the network logical device
2189 * if set prior to registration will cause a symlink during initialization.
2190 */
2191 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2192
2193 /* Set the sysfs device type for the network logical device to allow
2194 * fine-grained identification of different network device types. For
2195 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2196 */
2197 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2198
2199 /* Default NAPI poll() weight
2200 * Device drivers are strongly advised to not use bigger value
2201 */
2202 #define NAPI_POLL_WEIGHT 64
2203
2204 /**
2205 * netif_napi_add - initialize a NAPI context
2206 * @dev: network device
2207 * @napi: NAPI context
2208 * @poll: polling function
2209 * @weight: default weight
2210 *
2211 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2212 * *any* of the other NAPI-related functions.
2213 */
2214 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2215 int (*poll)(struct napi_struct *, int), int weight);
2216
2217 /**
2218 * netif_tx_napi_add - initialize a NAPI context
2219 * @dev: network device
2220 * @napi: NAPI context
2221 * @poll: polling function
2222 * @weight: default weight
2223 *
2224 * This variant of netif_napi_add() should be used from drivers using NAPI
2225 * to exclusively poll a TX queue.
2226 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2227 */
2228 static inline void netif_tx_napi_add(struct net_device *dev,
2229 struct napi_struct *napi,
2230 int (*poll)(struct napi_struct *, int),
2231 int weight)
2232 {
2233 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2234 netif_napi_add(dev, napi, poll, weight);
2235 }
2236
2237 /**
2238 * netif_napi_del - remove a NAPI context
2239 * @napi: NAPI context
2240 *
2241 * netif_napi_del() removes a NAPI context from the network device NAPI list
2242 */
2243 void netif_napi_del(struct napi_struct *napi);
2244
2245 struct napi_gro_cb {
2246 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2247 void *frag0;
2248
2249 /* Length of frag0. */
2250 unsigned int frag0_len;
2251
2252 /* This indicates where we are processing relative to skb->data. */
2253 int data_offset;
2254
2255 /* This is non-zero if the packet cannot be merged with the new skb. */
2256 u16 flush;
2257
2258 /* Save the IP ID here and check when we get to the transport layer */
2259 u16 flush_id;
2260
2261 /* Number of segments aggregated. */
2262 u16 count;
2263
2264 /* Start offset for remote checksum offload */
2265 u16 gro_remcsum_start;
2266
2267 /* jiffies when first packet was created/queued */
2268 unsigned long age;
2269
2270 /* Used in ipv6_gro_receive() and foo-over-udp */
2271 u16 proto;
2272
2273 /* This is non-zero if the packet may be of the same flow. */
2274 u8 same_flow:1;
2275
2276 /* Used in tunnel GRO receive */
2277 u8 encap_mark:1;
2278
2279 /* GRO checksum is valid */
2280 u8 csum_valid:1;
2281
2282 /* Number of checksums via CHECKSUM_UNNECESSARY */
2283 u8 csum_cnt:3;
2284
2285 /* Free the skb? */
2286 u8 free:2;
2287 #define NAPI_GRO_FREE 1
2288 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2289
2290 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2291 u8 is_ipv6:1;
2292
2293 /* Used in GRE, set in fou/gue_gro_receive */
2294 u8 is_fou:1;
2295
2296 /* Used to determine if flush_id can be ignored */
2297 u8 is_atomic:1;
2298
2299 /* Number of gro_receive callbacks this packet already went through */
2300 u8 recursion_counter:4;
2301
2302 /* 1 bit hole */
2303
2304 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2305 __wsum csum;
2306
2307 /* used in skb_gro_receive() slow path */
2308 struct sk_buff *last;
2309 };
2310
2311 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2312
2313 #define GRO_RECURSION_LIMIT 15
2314 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2315 {
2316 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2317 }
2318
2319 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2320 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2321 struct list_head *head,
2322 struct sk_buff *skb)
2323 {
2324 if (unlikely(gro_recursion_inc_test(skb))) {
2325 NAPI_GRO_CB(skb)->flush |= 1;
2326 return NULL;
2327 }
2328
2329 return cb(head, skb);
2330 }
2331
2332 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2333 struct sk_buff *);
2334 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2335 struct sock *sk,
2336 struct list_head *head,
2337 struct sk_buff *skb)
2338 {
2339 if (unlikely(gro_recursion_inc_test(skb))) {
2340 NAPI_GRO_CB(skb)->flush |= 1;
2341 return NULL;
2342 }
2343
2344 return cb(sk, head, skb);
2345 }
2346
2347 struct packet_type {
2348 __be16 type; /* This is really htons(ether_type). */
2349 bool ignore_outgoing;
2350 struct net_device *dev; /* NULL is wildcarded here */
2351 int (*func) (struct sk_buff *,
2352 struct net_device *,
2353 struct packet_type *,
2354 struct net_device *);
2355 void (*list_func) (struct list_head *,
2356 struct packet_type *,
2357 struct net_device *);
2358 bool (*id_match)(struct packet_type *ptype,
2359 struct sock *sk);
2360 void *af_packet_priv;
2361 struct list_head list;
2362 };
2363
2364 struct offload_callbacks {
2365 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2366 netdev_features_t features);
2367 struct sk_buff *(*gro_receive)(struct list_head *head,
2368 struct sk_buff *skb);
2369 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2370 };
2371
2372 struct packet_offload {
2373 __be16 type; /* This is really htons(ether_type). */
2374 u16 priority;
2375 struct offload_callbacks callbacks;
2376 struct list_head list;
2377 };
2378
2379 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2380 struct pcpu_sw_netstats {
2381 u64 rx_packets;
2382 u64 rx_bytes;
2383 u64 tx_packets;
2384 u64 tx_bytes;
2385 struct u64_stats_sync syncp;
2386 } __aligned(4 * sizeof(u64));
2387
2388 struct pcpu_lstats {
2389 u64 packets;
2390 u64 bytes;
2391 struct u64_stats_sync syncp;
2392 } __aligned(2 * sizeof(u64));
2393
2394 #define __netdev_alloc_pcpu_stats(type, gfp) \
2395 ({ \
2396 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2397 if (pcpu_stats) { \
2398 int __cpu; \
2399 for_each_possible_cpu(__cpu) { \
2400 typeof(type) *stat; \
2401 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2402 u64_stats_init(&stat->syncp); \
2403 } \
2404 } \
2405 pcpu_stats; \
2406 })
2407
2408 #define netdev_alloc_pcpu_stats(type) \
2409 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2410
2411 enum netdev_lag_tx_type {
2412 NETDEV_LAG_TX_TYPE_UNKNOWN,
2413 NETDEV_LAG_TX_TYPE_RANDOM,
2414 NETDEV_LAG_TX_TYPE_BROADCAST,
2415 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2416 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2417 NETDEV_LAG_TX_TYPE_HASH,
2418 };
2419
2420 enum netdev_lag_hash {
2421 NETDEV_LAG_HASH_NONE,
2422 NETDEV_LAG_HASH_L2,
2423 NETDEV_LAG_HASH_L34,
2424 NETDEV_LAG_HASH_L23,
2425 NETDEV_LAG_HASH_E23,
2426 NETDEV_LAG_HASH_E34,
2427 NETDEV_LAG_HASH_UNKNOWN,
2428 };
2429
2430 struct netdev_lag_upper_info {
2431 enum netdev_lag_tx_type tx_type;
2432 enum netdev_lag_hash hash_type;
2433 };
2434
2435 struct netdev_lag_lower_state_info {
2436 u8 link_up : 1,
2437 tx_enabled : 1;
2438 };
2439
2440 #include <linux/notifier.h>
2441
2442 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2443 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2444 * adding new types.
2445 */
2446 enum netdev_cmd {
2447 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2448 NETDEV_DOWN,
2449 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2450 detected a hardware crash and restarted
2451 - we can use this eg to kick tcp sessions
2452 once done */
2453 NETDEV_CHANGE, /* Notify device state change */
2454 NETDEV_REGISTER,
2455 NETDEV_UNREGISTER,
2456 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2457 NETDEV_CHANGEADDR, /* notify after the address change */
2458 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2459 NETDEV_GOING_DOWN,
2460 NETDEV_CHANGENAME,
2461 NETDEV_FEAT_CHANGE,
2462 NETDEV_BONDING_FAILOVER,
2463 NETDEV_PRE_UP,
2464 NETDEV_PRE_TYPE_CHANGE,
2465 NETDEV_POST_TYPE_CHANGE,
2466 NETDEV_POST_INIT,
2467 NETDEV_RELEASE,
2468 NETDEV_NOTIFY_PEERS,
2469 NETDEV_JOIN,
2470 NETDEV_CHANGEUPPER,
2471 NETDEV_RESEND_IGMP,
2472 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2473 NETDEV_CHANGEINFODATA,
2474 NETDEV_BONDING_INFO,
2475 NETDEV_PRECHANGEUPPER,
2476 NETDEV_CHANGELOWERSTATE,
2477 NETDEV_UDP_TUNNEL_PUSH_INFO,
2478 NETDEV_UDP_TUNNEL_DROP_INFO,
2479 NETDEV_CHANGE_TX_QUEUE_LEN,
2480 NETDEV_CVLAN_FILTER_PUSH_INFO,
2481 NETDEV_CVLAN_FILTER_DROP_INFO,
2482 NETDEV_SVLAN_FILTER_PUSH_INFO,
2483 NETDEV_SVLAN_FILTER_DROP_INFO,
2484 };
2485 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2486
2487 int register_netdevice_notifier(struct notifier_block *nb);
2488 int unregister_netdevice_notifier(struct notifier_block *nb);
2489
2490 struct netdev_notifier_info {
2491 struct net_device *dev;
2492 struct netlink_ext_ack *extack;
2493 };
2494
2495 struct netdev_notifier_info_ext {
2496 struct netdev_notifier_info info; /* must be first */
2497 union {
2498 u32 mtu;
2499 } ext;
2500 };
2501
2502 struct netdev_notifier_change_info {
2503 struct netdev_notifier_info info; /* must be first */
2504 unsigned int flags_changed;
2505 };
2506
2507 struct netdev_notifier_changeupper_info {
2508 struct netdev_notifier_info info; /* must be first */
2509 struct net_device *upper_dev; /* new upper dev */
2510 bool master; /* is upper dev master */
2511 bool linking; /* is the notification for link or unlink */
2512 void *upper_info; /* upper dev info */
2513 };
2514
2515 struct netdev_notifier_changelowerstate_info {
2516 struct netdev_notifier_info info; /* must be first */
2517 void *lower_state_info; /* is lower dev state */
2518 };
2519
2520 struct netdev_notifier_pre_changeaddr_info {
2521 struct netdev_notifier_info info; /* must be first */
2522 const unsigned char *dev_addr;
2523 };
2524
2525 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2526 struct net_device *dev)
2527 {
2528 info->dev = dev;
2529 info->extack = NULL;
2530 }
2531
2532 static inline struct net_device *
2533 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2534 {
2535 return info->dev;
2536 }
2537
2538 static inline struct netlink_ext_ack *
2539 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2540 {
2541 return info->extack;
2542 }
2543
2544 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2545
2546
2547 extern rwlock_t dev_base_lock; /* Device list lock */
2548
2549 #define for_each_netdev(net, d) \
2550 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2551 #define for_each_netdev_reverse(net, d) \
2552 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2553 #define for_each_netdev_rcu(net, d) \
2554 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2555 #define for_each_netdev_safe(net, d, n) \
2556 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2557 #define for_each_netdev_continue(net, d) \
2558 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2559 #define for_each_netdev_continue_rcu(net, d) \
2560 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2561 #define for_each_netdev_in_bond_rcu(bond, slave) \
2562 for_each_netdev_rcu(&init_net, slave) \
2563 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2564 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2565
2566 static inline struct net_device *next_net_device(struct net_device *dev)
2567 {
2568 struct list_head *lh;
2569 struct net *net;
2570
2571 net = dev_net(dev);
2572 lh = dev->dev_list.next;
2573 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2574 }
2575
2576 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2577 {
2578 struct list_head *lh;
2579 struct net *net;
2580
2581 net = dev_net(dev);
2582 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2583 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2584 }
2585
2586 static inline struct net_device *first_net_device(struct net *net)
2587 {
2588 return list_empty(&net->dev_base_head) ? NULL :
2589 net_device_entry(net->dev_base_head.next);
2590 }
2591
2592 static inline struct net_device *first_net_device_rcu(struct net *net)
2593 {
2594 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2595
2596 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2597 }
2598
2599 int netdev_boot_setup_check(struct net_device *dev);
2600 unsigned long netdev_boot_base(const char *prefix, int unit);
2601 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2602 const char *hwaddr);
2603 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2604 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2605 void dev_add_pack(struct packet_type *pt);
2606 void dev_remove_pack(struct packet_type *pt);
2607 void __dev_remove_pack(struct packet_type *pt);
2608 void dev_add_offload(struct packet_offload *po);
2609 void dev_remove_offload(struct packet_offload *po);
2610
2611 int dev_get_iflink(const struct net_device *dev);
2612 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2613 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2614 unsigned short mask);
2615 struct net_device *dev_get_by_name(struct net *net, const char *name);
2616 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2617 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2618 int dev_alloc_name(struct net_device *dev, const char *name);
2619 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2620 void dev_close(struct net_device *dev);
2621 void dev_close_many(struct list_head *head, bool unlink);
2622 void dev_disable_lro(struct net_device *dev);
2623 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2624 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2625 struct net_device *sb_dev);
2626 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2627 struct net_device *sb_dev);
2628 int dev_queue_xmit(struct sk_buff *skb);
2629 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2630 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2631 int register_netdevice(struct net_device *dev);
2632 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2633 void unregister_netdevice_many(struct list_head *head);
2634 static inline void unregister_netdevice(struct net_device *dev)
2635 {
2636 unregister_netdevice_queue(dev, NULL);
2637 }
2638
2639 int netdev_refcnt_read(const struct net_device *dev);
2640 void free_netdev(struct net_device *dev);
2641 void netdev_freemem(struct net_device *dev);
2642 void synchronize_net(void);
2643 int init_dummy_netdev(struct net_device *dev);
2644
2645 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2646 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2647 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2648 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2649 int netdev_get_name(struct net *net, char *name, int ifindex);
2650 int dev_restart(struct net_device *dev);
2651 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2652
2653 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2654 {
2655 return NAPI_GRO_CB(skb)->data_offset;
2656 }
2657
2658 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2659 {
2660 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2661 }
2662
2663 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2664 {
2665 NAPI_GRO_CB(skb)->data_offset += len;
2666 }
2667
2668 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2669 unsigned int offset)
2670 {
2671 return NAPI_GRO_CB(skb)->frag0 + offset;
2672 }
2673
2674 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2675 {
2676 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2677 }
2678
2679 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2680 {
2681 NAPI_GRO_CB(skb)->frag0 = NULL;
2682 NAPI_GRO_CB(skb)->frag0_len = 0;
2683 }
2684
2685 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2686 unsigned int offset)
2687 {
2688 if (!pskb_may_pull(skb, hlen))
2689 return NULL;
2690
2691 skb_gro_frag0_invalidate(skb);
2692 return skb->data + offset;
2693 }
2694
2695 static inline void *skb_gro_network_header(struct sk_buff *skb)
2696 {
2697 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2698 skb_network_offset(skb);
2699 }
2700
2701 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2702 const void *start, unsigned int len)
2703 {
2704 if (NAPI_GRO_CB(skb)->csum_valid)
2705 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2706 csum_partial(start, len, 0));
2707 }
2708
2709 /* GRO checksum functions. These are logical equivalents of the normal
2710 * checksum functions (in skbuff.h) except that they operate on the GRO
2711 * offsets and fields in sk_buff.
2712 */
2713
2714 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2715
2716 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2717 {
2718 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2719 }
2720
2721 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2722 bool zero_okay,
2723 __sum16 check)
2724 {
2725 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2726 skb_checksum_start_offset(skb) <
2727 skb_gro_offset(skb)) &&
2728 !skb_at_gro_remcsum_start(skb) &&
2729 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2730 (!zero_okay || check));
2731 }
2732
2733 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2734 __wsum psum)
2735 {
2736 if (NAPI_GRO_CB(skb)->csum_valid &&
2737 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2738 return 0;
2739
2740 NAPI_GRO_CB(skb)->csum = psum;
2741
2742 return __skb_gro_checksum_complete(skb);
2743 }
2744
2745 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2746 {
2747 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2748 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2749 NAPI_GRO_CB(skb)->csum_cnt--;
2750 } else {
2751 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2752 * verified a new top level checksum or an encapsulated one
2753 * during GRO. This saves work if we fallback to normal path.
2754 */
2755 __skb_incr_checksum_unnecessary(skb);
2756 }
2757 }
2758
2759 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2760 compute_pseudo) \
2761 ({ \
2762 __sum16 __ret = 0; \
2763 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2764 __ret = __skb_gro_checksum_validate_complete(skb, \
2765 compute_pseudo(skb, proto)); \
2766 if (!__ret) \
2767 skb_gro_incr_csum_unnecessary(skb); \
2768 __ret; \
2769 })
2770
2771 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2772 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2773
2774 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2775 compute_pseudo) \
2776 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2777
2778 #define skb_gro_checksum_simple_validate(skb) \
2779 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2780
2781 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2782 {
2783 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2784 !NAPI_GRO_CB(skb)->csum_valid);
2785 }
2786
2787 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2788 __sum16 check, __wsum pseudo)
2789 {
2790 NAPI_GRO_CB(skb)->csum = ~pseudo;
2791 NAPI_GRO_CB(skb)->csum_valid = 1;
2792 }
2793
2794 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2795 do { \
2796 if (__skb_gro_checksum_convert_check(skb)) \
2797 __skb_gro_checksum_convert(skb, check, \
2798 compute_pseudo(skb, proto)); \
2799 } while (0)
2800
2801 struct gro_remcsum {
2802 int offset;
2803 __wsum delta;
2804 };
2805
2806 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2807 {
2808 grc->offset = 0;
2809 grc->delta = 0;
2810 }
2811
2812 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2813 unsigned int off, size_t hdrlen,
2814 int start, int offset,
2815 struct gro_remcsum *grc,
2816 bool nopartial)
2817 {
2818 __wsum delta;
2819 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2820
2821 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2822
2823 if (!nopartial) {
2824 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2825 return ptr;
2826 }
2827
2828 ptr = skb_gro_header_fast(skb, off);
2829 if (skb_gro_header_hard(skb, off + plen)) {
2830 ptr = skb_gro_header_slow(skb, off + plen, off);
2831 if (!ptr)
2832 return NULL;
2833 }
2834
2835 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2836 start, offset);
2837
2838 /* Adjust skb->csum since we changed the packet */
2839 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2840
2841 grc->offset = off + hdrlen + offset;
2842 grc->delta = delta;
2843
2844 return ptr;
2845 }
2846
2847 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2848 struct gro_remcsum *grc)
2849 {
2850 void *ptr;
2851 size_t plen = grc->offset + sizeof(u16);
2852
2853 if (!grc->delta)
2854 return;
2855
2856 ptr = skb_gro_header_fast(skb, grc->offset);
2857 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2858 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2859 if (!ptr)
2860 return;
2861 }
2862
2863 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2864 }
2865
2866 #ifdef CONFIG_XFRM_OFFLOAD
2867 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2868 {
2869 if (PTR_ERR(pp) != -EINPROGRESS)
2870 NAPI_GRO_CB(skb)->flush |= flush;
2871 }
2872 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2873 struct sk_buff *pp,
2874 int flush,
2875 struct gro_remcsum *grc)
2876 {
2877 if (PTR_ERR(pp) != -EINPROGRESS) {
2878 NAPI_GRO_CB(skb)->flush |= flush;
2879 skb_gro_remcsum_cleanup(skb, grc);
2880 skb->remcsum_offload = 0;
2881 }
2882 }
2883 #else
2884 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2885 {
2886 NAPI_GRO_CB(skb)->flush |= flush;
2887 }
2888 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2889 struct sk_buff *pp,
2890 int flush,
2891 struct gro_remcsum *grc)
2892 {
2893 NAPI_GRO_CB(skb)->flush |= flush;
2894 skb_gro_remcsum_cleanup(skb, grc);
2895 skb->remcsum_offload = 0;
2896 }
2897 #endif
2898
2899 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2900 unsigned short type,
2901 const void *daddr, const void *saddr,
2902 unsigned int len)
2903 {
2904 if (!dev->header_ops || !dev->header_ops->create)
2905 return 0;
2906
2907 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2908 }
2909
2910 static inline int dev_parse_header(const struct sk_buff *skb,
2911 unsigned char *haddr)
2912 {
2913 const struct net_device *dev = skb->dev;
2914
2915 if (!dev->header_ops || !dev->header_ops->parse)
2916 return 0;
2917 return dev->header_ops->parse(skb, haddr);
2918 }
2919
2920 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2921 {
2922 const struct net_device *dev = skb->dev;
2923
2924 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2925 return 0;
2926 return dev->header_ops->parse_protocol(skb);
2927 }
2928
2929 /* ll_header must have at least hard_header_len allocated */
2930 static inline bool dev_validate_header(const struct net_device *dev,
2931 char *ll_header, int len)
2932 {
2933 if (likely(len >= dev->hard_header_len))
2934 return true;
2935 if (len < dev->min_header_len)
2936 return false;
2937
2938 if (capable(CAP_SYS_RAWIO)) {
2939 memset(ll_header + len, 0, dev->hard_header_len - len);
2940 return true;
2941 }
2942
2943 if (dev->header_ops && dev->header_ops->validate)
2944 return dev->header_ops->validate(ll_header, len);
2945
2946 return false;
2947 }
2948
2949 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2950 int len, int size);
2951 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2952 static inline int unregister_gifconf(unsigned int family)
2953 {
2954 return register_gifconf(family, NULL);
2955 }
2956
2957 #ifdef CONFIG_NET_FLOW_LIMIT
2958 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2959 struct sd_flow_limit {
2960 u64 count;
2961 unsigned int num_buckets;
2962 unsigned int history_head;
2963 u16 history[FLOW_LIMIT_HISTORY];
2964 u8 buckets[];
2965 };
2966
2967 extern int netdev_flow_limit_table_len;
2968 #endif /* CONFIG_NET_FLOW_LIMIT */
2969
2970 /*
2971 * Incoming packets are placed on per-CPU queues
2972 */
2973 struct softnet_data {
2974 struct list_head poll_list;
2975 struct sk_buff_head process_queue;
2976
2977 /* stats */
2978 unsigned int processed;
2979 unsigned int time_squeeze;
2980 unsigned int received_rps;
2981 #ifdef CONFIG_RPS
2982 struct softnet_data *rps_ipi_list;
2983 #endif
2984 #ifdef CONFIG_NET_FLOW_LIMIT
2985 struct sd_flow_limit __rcu *flow_limit;
2986 #endif
2987 struct Qdisc *output_queue;
2988 struct Qdisc **output_queue_tailp;
2989 struct sk_buff *completion_queue;
2990 #ifdef CONFIG_XFRM_OFFLOAD
2991 struct sk_buff_head xfrm_backlog;
2992 #endif
2993 /* written and read only by owning cpu: */
2994 struct {
2995 u16 recursion;
2996 u8 more;
2997 } xmit;
2998 #ifdef CONFIG_RPS
2999 /* input_queue_head should be written by cpu owning this struct,
3000 * and only read by other cpus. Worth using a cache line.
3001 */
3002 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3003
3004 /* Elements below can be accessed between CPUs for RPS/RFS */
3005 call_single_data_t csd ____cacheline_aligned_in_smp;
3006 struct softnet_data *rps_ipi_next;
3007 unsigned int cpu;
3008 unsigned int input_queue_tail;
3009 #endif
3010 unsigned int dropped;
3011 struct sk_buff_head input_pkt_queue;
3012 struct napi_struct backlog;
3013
3014 };
3015
3016 static inline void input_queue_head_incr(struct softnet_data *sd)
3017 {
3018 #ifdef CONFIG_RPS
3019 sd->input_queue_head++;
3020 #endif
3021 }
3022
3023 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3024 unsigned int *qtail)
3025 {
3026 #ifdef CONFIG_RPS
3027 *qtail = ++sd->input_queue_tail;
3028 #endif
3029 }
3030
3031 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3032
3033 static inline int dev_recursion_level(void)
3034 {
3035 return this_cpu_read(softnet_data.xmit.recursion);
3036 }
3037
3038 #define XMIT_RECURSION_LIMIT 10
3039 static inline bool dev_xmit_recursion(void)
3040 {
3041 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3042 XMIT_RECURSION_LIMIT);
3043 }
3044
3045 static inline void dev_xmit_recursion_inc(void)
3046 {
3047 __this_cpu_inc(softnet_data.xmit.recursion);
3048 }
3049
3050 static inline void dev_xmit_recursion_dec(void)
3051 {
3052 __this_cpu_dec(softnet_data.xmit.recursion);
3053 }
3054
3055 void __netif_schedule(struct Qdisc *q);
3056 void netif_schedule_queue(struct netdev_queue *txq);
3057
3058 static inline void netif_tx_schedule_all(struct net_device *dev)
3059 {
3060 unsigned int i;
3061
3062 for (i = 0; i < dev->num_tx_queues; i++)
3063 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3064 }
3065
3066 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3067 {
3068 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3069 }
3070
3071 /**
3072 * netif_start_queue - allow transmit
3073 * @dev: network device
3074 *
3075 * Allow upper layers to call the device hard_start_xmit routine.
3076 */
3077 static inline void netif_start_queue(struct net_device *dev)
3078 {
3079 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3080 }
3081
3082 static inline void netif_tx_start_all_queues(struct net_device *dev)
3083 {
3084 unsigned int i;
3085
3086 for (i = 0; i < dev->num_tx_queues; i++) {
3087 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3088 netif_tx_start_queue(txq);
3089 }
3090 }
3091
3092 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3093
3094 /**
3095 * netif_wake_queue - restart transmit
3096 * @dev: network device
3097 *
3098 * Allow upper layers to call the device hard_start_xmit routine.
3099 * Used for flow control when transmit resources are available.
3100 */
3101 static inline void netif_wake_queue(struct net_device *dev)
3102 {
3103 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3104 }
3105
3106 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3107 {
3108 unsigned int i;
3109
3110 for (i = 0; i < dev->num_tx_queues; i++) {
3111 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3112 netif_tx_wake_queue(txq);
3113 }
3114 }
3115
3116 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3117 {
3118 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3119 }
3120
3121 /**
3122 * netif_stop_queue - stop transmitted packets
3123 * @dev: network device
3124 *
3125 * Stop upper layers calling the device hard_start_xmit routine.
3126 * Used for flow control when transmit resources are unavailable.
3127 */
3128 static inline void netif_stop_queue(struct net_device *dev)
3129 {
3130 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3131 }
3132
3133 void netif_tx_stop_all_queues(struct net_device *dev);
3134
3135 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3136 {
3137 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3138 }
3139
3140 /**
3141 * netif_queue_stopped - test if transmit queue is flowblocked
3142 * @dev: network device
3143 *
3144 * Test if transmit queue on device is currently unable to send.
3145 */
3146 static inline bool netif_queue_stopped(const struct net_device *dev)
3147 {
3148 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3149 }
3150
3151 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3152 {
3153 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3154 }
3155
3156 static inline bool
3157 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3158 {
3159 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3160 }
3161
3162 static inline bool
3163 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3164 {
3165 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3166 }
3167
3168 /**
3169 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3170 * @dev_queue: pointer to transmit queue
3171 *
3172 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3173 * to give appropriate hint to the CPU.
3174 */
3175 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3176 {
3177 #ifdef CONFIG_BQL
3178 prefetchw(&dev_queue->dql.num_queued);
3179 #endif
3180 }
3181
3182 /**
3183 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3184 * @dev_queue: pointer to transmit queue
3185 *
3186 * BQL enabled drivers might use this helper in their TX completion path,
3187 * to give appropriate hint to the CPU.
3188 */
3189 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3190 {
3191 #ifdef CONFIG_BQL
3192 prefetchw(&dev_queue->dql.limit);
3193 #endif
3194 }
3195
3196 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3197 unsigned int bytes)
3198 {
3199 #ifdef CONFIG_BQL
3200 dql_queued(&dev_queue->dql, bytes);
3201
3202 if (likely(dql_avail(&dev_queue->dql) >= 0))
3203 return;
3204
3205 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3206
3207 /*
3208 * The XOFF flag must be set before checking the dql_avail below,
3209 * because in netdev_tx_completed_queue we update the dql_completed
3210 * before checking the XOFF flag.
3211 */
3212 smp_mb();
3213
3214 /* check again in case another CPU has just made room avail */
3215 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3216 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3217 #endif
3218 }
3219
3220 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3221 * that they should not test BQL status themselves.
3222 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3223 * skb of a batch.
3224 * Returns true if the doorbell must be used to kick the NIC.
3225 */
3226 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3227 unsigned int bytes,
3228 bool xmit_more)
3229 {
3230 if (xmit_more) {
3231 #ifdef CONFIG_BQL
3232 dql_queued(&dev_queue->dql, bytes);
3233 #endif
3234 return netif_tx_queue_stopped(dev_queue);
3235 }
3236 netdev_tx_sent_queue(dev_queue, bytes);
3237 return true;
3238 }
3239
3240 /**
3241 * netdev_sent_queue - report the number of bytes queued to hardware
3242 * @dev: network device
3243 * @bytes: number of bytes queued to the hardware device queue
3244 *
3245 * Report the number of bytes queued for sending/completion to the network
3246 * device hardware queue. @bytes should be a good approximation and should
3247 * exactly match netdev_completed_queue() @bytes
3248 */
3249 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3250 {
3251 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3252 }
3253
3254 static inline bool __netdev_sent_queue(struct net_device *dev,
3255 unsigned int bytes,
3256 bool xmit_more)
3257 {
3258 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3259 xmit_more);
3260 }
3261
3262 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3263 unsigned int pkts, unsigned int bytes)
3264 {
3265 #ifdef CONFIG_BQL
3266 if (unlikely(!bytes))
3267 return;
3268
3269 dql_completed(&dev_queue->dql, bytes);
3270
3271 /*
3272 * Without the memory barrier there is a small possiblity that
3273 * netdev_tx_sent_queue will miss the update and cause the queue to
3274 * be stopped forever
3275 */
3276 smp_mb();
3277
3278 if (dql_avail(&dev_queue->dql) < 0)
3279 return;
3280
3281 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3282 netif_schedule_queue(dev_queue);
3283 #endif
3284 }
3285
3286 /**
3287 * netdev_completed_queue - report bytes and packets completed by device
3288 * @dev: network device
3289 * @pkts: actual number of packets sent over the medium
3290 * @bytes: actual number of bytes sent over the medium
3291 *
3292 * Report the number of bytes and packets transmitted by the network device
3293 * hardware queue over the physical medium, @bytes must exactly match the
3294 * @bytes amount passed to netdev_sent_queue()
3295 */
3296 static inline void netdev_completed_queue(struct net_device *dev,
3297 unsigned int pkts, unsigned int bytes)
3298 {
3299 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3300 }
3301
3302 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3303 {
3304 #ifdef CONFIG_BQL
3305 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3306 dql_reset(&q->dql);
3307 #endif
3308 }
3309
3310 /**
3311 * netdev_reset_queue - reset the packets and bytes count of a network device
3312 * @dev_queue: network device
3313 *
3314 * Reset the bytes and packet count of a network device and clear the
3315 * software flow control OFF bit for this network device
3316 */
3317 static inline void netdev_reset_queue(struct net_device *dev_queue)
3318 {
3319 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3320 }
3321
3322 /**
3323 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3324 * @dev: network device
3325 * @queue_index: given tx queue index
3326 *
3327 * Returns 0 if given tx queue index >= number of device tx queues,
3328 * otherwise returns the originally passed tx queue index.
3329 */
3330 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3331 {
3332 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3333 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3334 dev->name, queue_index,
3335 dev->real_num_tx_queues);
3336 return 0;
3337 }
3338
3339 return queue_index;
3340 }
3341
3342 /**
3343 * netif_running - test if up
3344 * @dev: network device
3345 *
3346 * Test if the device has been brought up.
3347 */
3348 static inline bool netif_running(const struct net_device *dev)
3349 {
3350 return test_bit(__LINK_STATE_START, &dev->state);
3351 }
3352
3353 /*
3354 * Routines to manage the subqueues on a device. We only need start,
3355 * stop, and a check if it's stopped. All other device management is
3356 * done at the overall netdevice level.
3357 * Also test the device if we're multiqueue.
3358 */
3359
3360 /**
3361 * netif_start_subqueue - allow sending packets on subqueue
3362 * @dev: network device
3363 * @queue_index: sub queue index
3364 *
3365 * Start individual transmit queue of a device with multiple transmit queues.
3366 */
3367 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3368 {
3369 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3370
3371 netif_tx_start_queue(txq);
3372 }
3373
3374 /**
3375 * netif_stop_subqueue - stop sending packets on subqueue
3376 * @dev: network device
3377 * @queue_index: sub queue index
3378 *
3379 * Stop individual transmit queue of a device with multiple transmit queues.
3380 */
3381 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3382 {
3383 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3384 netif_tx_stop_queue(txq);
3385 }
3386
3387 /**
3388 * netif_subqueue_stopped - test status of subqueue
3389 * @dev: network device
3390 * @queue_index: sub queue index
3391 *
3392 * Check individual transmit queue of a device with multiple transmit queues.
3393 */
3394 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3395 u16 queue_index)
3396 {
3397 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3398
3399 return netif_tx_queue_stopped(txq);
3400 }
3401
3402 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3403 struct sk_buff *skb)
3404 {
3405 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3406 }
3407
3408 /**
3409 * netif_wake_subqueue - allow sending packets on subqueue
3410 * @dev: network device
3411 * @queue_index: sub queue index
3412 *
3413 * Resume individual transmit queue of a device with multiple transmit queues.
3414 */
3415 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3416 {
3417 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3418
3419 netif_tx_wake_queue(txq);
3420 }
3421
3422 #ifdef CONFIG_XPS
3423 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3424 u16 index);
3425 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3426 u16 index, bool is_rxqs_map);
3427
3428 /**
3429 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3430 * @j: CPU/Rx queue index
3431 * @mask: bitmask of all cpus/rx queues
3432 * @nr_bits: number of bits in the bitmask
3433 *
3434 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3435 */
3436 static inline bool netif_attr_test_mask(unsigned long j,
3437 const unsigned long *mask,
3438 unsigned int nr_bits)
3439 {
3440 cpu_max_bits_warn(j, nr_bits);
3441 return test_bit(j, mask);
3442 }
3443
3444 /**
3445 * netif_attr_test_online - Test for online CPU/Rx queue
3446 * @j: CPU/Rx queue index
3447 * @online_mask: bitmask for CPUs/Rx queues that are online
3448 * @nr_bits: number of bits in the bitmask
3449 *
3450 * Returns true if a CPU/Rx queue is online.
3451 */
3452 static inline bool netif_attr_test_online(unsigned long j,
3453 const unsigned long *online_mask,
3454 unsigned int nr_bits)
3455 {
3456 cpu_max_bits_warn(j, nr_bits);
3457
3458 if (online_mask)
3459 return test_bit(j, online_mask);
3460
3461 return (j < nr_bits);
3462 }
3463
3464 /**
3465 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3466 * @n: CPU/Rx queue index
3467 * @srcp: the cpumask/Rx queue mask pointer
3468 * @nr_bits: number of bits in the bitmask
3469 *
3470 * Returns >= nr_bits if no further CPUs/Rx queues set.
3471 */
3472 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3473 unsigned int nr_bits)
3474 {
3475 /* -1 is a legal arg here. */
3476 if (n != -1)
3477 cpu_max_bits_warn(n, nr_bits);
3478
3479 if (srcp)
3480 return find_next_bit(srcp, nr_bits, n + 1);
3481
3482 return n + 1;
3483 }
3484
3485 /**
3486 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3487 * @n: CPU/Rx queue index
3488 * @src1p: the first CPUs/Rx queues mask pointer
3489 * @src2p: the second CPUs/Rx queues mask pointer
3490 * @nr_bits: number of bits in the bitmask
3491 *
3492 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3493 */
3494 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3495 const unsigned long *src2p,
3496 unsigned int nr_bits)
3497 {
3498 /* -1 is a legal arg here. */
3499 if (n != -1)
3500 cpu_max_bits_warn(n, nr_bits);
3501
3502 if (src1p && src2p)
3503 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3504 else if (src1p)
3505 return find_next_bit(src1p, nr_bits, n + 1);
3506 else if (src2p)
3507 return find_next_bit(src2p, nr_bits, n + 1);
3508
3509 return n + 1;
3510 }
3511 #else
3512 static inline int netif_set_xps_queue(struct net_device *dev,
3513 const struct cpumask *mask,
3514 u16 index)
3515 {
3516 return 0;
3517 }
3518
3519 static inline int __netif_set_xps_queue(struct net_device *dev,
3520 const unsigned long *mask,
3521 u16 index, bool is_rxqs_map)
3522 {
3523 return 0;
3524 }
3525 #endif
3526
3527 /**
3528 * netif_is_multiqueue - test if device has multiple transmit queues
3529 * @dev: network device
3530 *
3531 * Check if device has multiple transmit queues
3532 */
3533 static inline bool netif_is_multiqueue(const struct net_device *dev)
3534 {
3535 return dev->num_tx_queues > 1;
3536 }
3537
3538 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3539
3540 #ifdef CONFIG_SYSFS
3541 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3542 #else
3543 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3544 unsigned int rxqs)
3545 {
3546 dev->real_num_rx_queues = rxqs;
3547 return 0;
3548 }
3549 #endif
3550
3551 static inline struct netdev_rx_queue *
3552 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3553 {
3554 return dev->_rx + rxq;
3555 }
3556
3557 #ifdef CONFIG_SYSFS
3558 static inline unsigned int get_netdev_rx_queue_index(
3559 struct netdev_rx_queue *queue)
3560 {
3561 struct net_device *dev = queue->dev;
3562 int index = queue - dev->_rx;
3563
3564 BUG_ON(index >= dev->num_rx_queues);
3565 return index;
3566 }
3567 #endif
3568
3569 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3570 int netif_get_num_default_rss_queues(void);
3571
3572 enum skb_free_reason {
3573 SKB_REASON_CONSUMED,
3574 SKB_REASON_DROPPED,
3575 };
3576
3577 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3578 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3579
3580 /*
3581 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3582 * interrupt context or with hardware interrupts being disabled.
3583 * (in_irq() || irqs_disabled())
3584 *
3585 * We provide four helpers that can be used in following contexts :
3586 *
3587 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3588 * replacing kfree_skb(skb)
3589 *
3590 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3591 * Typically used in place of consume_skb(skb) in TX completion path
3592 *
3593 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3594 * replacing kfree_skb(skb)
3595 *
3596 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3597 * and consumed a packet. Used in place of consume_skb(skb)
3598 */
3599 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3600 {
3601 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3602 }
3603
3604 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3605 {
3606 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3607 }
3608
3609 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3610 {
3611 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3612 }
3613
3614 static inline void dev_consume_skb_any(struct sk_buff *skb)
3615 {
3616 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3617 }
3618
3619 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3620 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3621 int netif_rx(struct sk_buff *skb);
3622 int netif_rx_ni(struct sk_buff *skb);
3623 int netif_receive_skb(struct sk_buff *skb);
3624 int netif_receive_skb_core(struct sk_buff *skb);
3625 void netif_receive_skb_list(struct list_head *head);
3626 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3627 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3628 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3629 gro_result_t napi_gro_frags(struct napi_struct *napi);
3630 struct packet_offload *gro_find_receive_by_type(__be16 type);
3631 struct packet_offload *gro_find_complete_by_type(__be16 type);
3632
3633 static inline void napi_free_frags(struct napi_struct *napi)
3634 {
3635 kfree_skb(napi->skb);
3636 napi->skb = NULL;
3637 }
3638
3639 bool netdev_is_rx_handler_busy(struct net_device *dev);
3640 int netdev_rx_handler_register(struct net_device *dev,
3641 rx_handler_func_t *rx_handler,
3642 void *rx_handler_data);
3643 void netdev_rx_handler_unregister(struct net_device *dev);
3644
3645 bool dev_valid_name(const char *name);
3646 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3647 bool *need_copyout);
3648 int dev_ifconf(struct net *net, struct ifconf *, int);
3649 int dev_ethtool(struct net *net, struct ifreq *);
3650 unsigned int dev_get_flags(const struct net_device *);
3651 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3652 struct netlink_ext_ack *extack);
3653 int dev_change_flags(struct net_device *dev, unsigned int flags,
3654 struct netlink_ext_ack *extack);
3655 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3656 unsigned int gchanges);
3657 int dev_change_name(struct net_device *, const char *);
3658 int dev_set_alias(struct net_device *, const char *, size_t);
3659 int dev_get_alias(const struct net_device *, char *, size_t);
3660 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3661 int __dev_set_mtu(struct net_device *, int);
3662 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3663 struct netlink_ext_ack *extack);
3664 int dev_set_mtu(struct net_device *, int);
3665 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3666 void dev_set_group(struct net_device *, int);
3667 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3668 struct netlink_ext_ack *extack);
3669 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3670 struct netlink_ext_ack *extack);
3671 int dev_change_carrier(struct net_device *, bool new_carrier);
3672 int dev_get_phys_port_id(struct net_device *dev,
3673 struct netdev_phys_item_id *ppid);
3674 int dev_get_phys_port_name(struct net_device *dev,
3675 char *name, size_t len);
3676 int dev_get_port_parent_id(struct net_device *dev,
3677 struct netdev_phys_item_id *ppid, bool recurse);
3678 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3679 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3680 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3681 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3682 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3683 struct netdev_queue *txq, int *ret);
3684
3685 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3686 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3687 int fd, u32 flags);
3688 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3689 enum bpf_netdev_command cmd);
3690 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3691
3692 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3693 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3694 bool is_skb_forwardable(const struct net_device *dev,
3695 const struct sk_buff *skb);
3696
3697 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3698 struct sk_buff *skb)
3699 {
3700 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3701 unlikely(!is_skb_forwardable(dev, skb))) {
3702 atomic_long_inc(&dev->rx_dropped);
3703 kfree_skb(skb);
3704 return NET_RX_DROP;
3705 }
3706
3707 skb_scrub_packet(skb, true);
3708 skb->priority = 0;
3709 return 0;
3710 }
3711
3712 bool dev_nit_active(struct net_device *dev);
3713 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3714
3715 extern int netdev_budget;
3716 extern unsigned int netdev_budget_usecs;
3717
3718 /* Called by rtnetlink.c:rtnl_unlock() */
3719 void netdev_run_todo(void);
3720
3721 /**
3722 * dev_put - release reference to device
3723 * @dev: network device
3724 *
3725 * Release reference to device to allow it to be freed.
3726 */
3727 static inline void dev_put(struct net_device *dev)
3728 {
3729 this_cpu_dec(*dev->pcpu_refcnt);
3730 }
3731
3732 /**
3733 * dev_hold - get reference to device
3734 * @dev: network device
3735 *
3736 * Hold reference to device to keep it from being freed.
3737 */
3738 static inline void dev_hold(struct net_device *dev)
3739 {
3740 this_cpu_inc(*dev->pcpu_refcnt);
3741 }
3742
3743 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3744 * and _off may be called from IRQ context, but it is caller
3745 * who is responsible for serialization of these calls.
3746 *
3747 * The name carrier is inappropriate, these functions should really be
3748 * called netif_lowerlayer_*() because they represent the state of any
3749 * kind of lower layer not just hardware media.
3750 */
3751
3752 void linkwatch_init_dev(struct net_device *dev);
3753 void linkwatch_fire_event(struct net_device *dev);
3754 void linkwatch_forget_dev(struct net_device *dev);
3755
3756 /**
3757 * netif_carrier_ok - test if carrier present
3758 * @dev: network device
3759 *
3760 * Check if carrier is present on device
3761 */
3762 static inline bool netif_carrier_ok(const struct net_device *dev)
3763 {
3764 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3765 }
3766
3767 unsigned long dev_trans_start(struct net_device *dev);
3768
3769 void __netdev_watchdog_up(struct net_device *dev);
3770
3771 void netif_carrier_on(struct net_device *dev);
3772
3773 void netif_carrier_off(struct net_device *dev);
3774
3775 /**
3776 * netif_dormant_on - mark device as dormant.
3777 * @dev: network device
3778 *
3779 * Mark device as dormant (as per RFC2863).
3780 *
3781 * The dormant state indicates that the relevant interface is not
3782 * actually in a condition to pass packets (i.e., it is not 'up') but is
3783 * in a "pending" state, waiting for some external event. For "on-
3784 * demand" interfaces, this new state identifies the situation where the
3785 * interface is waiting for events to place it in the up state.
3786 */
3787 static inline void netif_dormant_on(struct net_device *dev)
3788 {
3789 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3790 linkwatch_fire_event(dev);
3791 }
3792
3793 /**
3794 * netif_dormant_off - set device as not dormant.
3795 * @dev: network device
3796 *
3797 * Device is not in dormant state.
3798 */
3799 static inline void netif_dormant_off(struct net_device *dev)
3800 {
3801 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3802 linkwatch_fire_event(dev);
3803 }
3804
3805 /**
3806 * netif_dormant - test if device is dormant
3807 * @dev: network device
3808 *
3809 * Check if device is dormant.
3810 */
3811 static inline bool netif_dormant(const struct net_device *dev)
3812 {
3813 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3814 }
3815
3816
3817 /**
3818 * netif_oper_up - test if device is operational
3819 * @dev: network device
3820 *
3821 * Check if carrier is operational
3822 */
3823 static inline bool netif_oper_up(const struct net_device *dev)
3824 {
3825 return (dev->operstate == IF_OPER_UP ||
3826 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3827 }
3828
3829 /**
3830 * netif_device_present - is device available or removed
3831 * @dev: network device
3832 *
3833 * Check if device has not been removed from system.
3834 */
3835 static inline bool netif_device_present(struct net_device *dev)
3836 {
3837 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3838 }
3839
3840 void netif_device_detach(struct net_device *dev);
3841
3842 void netif_device_attach(struct net_device *dev);
3843
3844 /*
3845 * Network interface message level settings
3846 */
3847
3848 enum {
3849 NETIF_MSG_DRV = 0x0001,
3850 NETIF_MSG_PROBE = 0x0002,
3851 NETIF_MSG_LINK = 0x0004,
3852 NETIF_MSG_TIMER = 0x0008,
3853 NETIF_MSG_IFDOWN = 0x0010,
3854 NETIF_MSG_IFUP = 0x0020,
3855 NETIF_MSG_RX_ERR = 0x0040,
3856 NETIF_MSG_TX_ERR = 0x0080,
3857 NETIF_MSG_TX_QUEUED = 0x0100,
3858 NETIF_MSG_INTR = 0x0200,
3859 NETIF_MSG_TX_DONE = 0x0400,
3860 NETIF_MSG_RX_STATUS = 0x0800,
3861 NETIF_MSG_PKTDATA = 0x1000,
3862 NETIF_MSG_HW = 0x2000,
3863 NETIF_MSG_WOL = 0x4000,
3864 };
3865
3866 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3867 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3868 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3869 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3870 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3871 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3872 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3873 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3874 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3875 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3876 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3877 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3878 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3879 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3880 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3881
3882 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3883 {
3884 /* use default */
3885 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3886 return default_msg_enable_bits;
3887 if (debug_value == 0) /* no output */
3888 return 0;
3889 /* set low N bits */
3890 return (1U << debug_value) - 1;
3891 }
3892
3893 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3894 {
3895 spin_lock(&txq->_xmit_lock);
3896 txq->xmit_lock_owner = cpu;
3897 }
3898
3899 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3900 {
3901 __acquire(&txq->_xmit_lock);
3902 return true;
3903 }
3904
3905 static inline void __netif_tx_release(struct netdev_queue *txq)
3906 {
3907 __release(&txq->_xmit_lock);
3908 }
3909
3910 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3911 {
3912 spin_lock_bh(&txq->_xmit_lock);
3913 txq->xmit_lock_owner = smp_processor_id();
3914 }
3915
3916 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3917 {
3918 bool ok = spin_trylock(&txq->_xmit_lock);
3919 if (likely(ok))
3920 txq->xmit_lock_owner = smp_processor_id();
3921 return ok;
3922 }
3923
3924 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3925 {
3926 txq->xmit_lock_owner = -1;
3927 spin_unlock(&txq->_xmit_lock);
3928 }
3929
3930 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3931 {
3932 txq->xmit_lock_owner = -1;
3933 spin_unlock_bh(&txq->_xmit_lock);
3934 }
3935
3936 static inline void txq_trans_update(struct netdev_queue *txq)
3937 {
3938 if (txq->xmit_lock_owner != -1)
3939 txq->trans_start = jiffies;
3940 }
3941
3942 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3943 static inline void netif_trans_update(struct net_device *dev)
3944 {
3945 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3946
3947 if (txq->trans_start != jiffies)
3948 txq->trans_start = jiffies;
3949 }
3950
3951 /**
3952 * netif_tx_lock - grab network device transmit lock
3953 * @dev: network device
3954 *
3955 * Get network device transmit lock
3956 */
3957 static inline void netif_tx_lock(struct net_device *dev)
3958 {
3959 unsigned int i;
3960 int cpu;
3961
3962 spin_lock(&dev->tx_global_lock);
3963 cpu = smp_processor_id();
3964 for (i = 0; i < dev->num_tx_queues; i++) {
3965 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3966
3967 /* We are the only thread of execution doing a
3968 * freeze, but we have to grab the _xmit_lock in
3969 * order to synchronize with threads which are in
3970 * the ->hard_start_xmit() handler and already
3971 * checked the frozen bit.
3972 */
3973 __netif_tx_lock(txq, cpu);
3974 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3975 __netif_tx_unlock(txq);
3976 }
3977 }
3978
3979 static inline void netif_tx_lock_bh(struct net_device *dev)
3980 {
3981 local_bh_disable();
3982 netif_tx_lock(dev);
3983 }
3984
3985 static inline void netif_tx_unlock(struct net_device *dev)
3986 {
3987 unsigned int i;
3988
3989 for (i = 0; i < dev->num_tx_queues; i++) {
3990 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3991
3992 /* No need to grab the _xmit_lock here. If the
3993 * queue is not stopped for another reason, we
3994 * force a schedule.
3995 */
3996 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3997 netif_schedule_queue(txq);
3998 }
3999 spin_unlock(&dev->tx_global_lock);
4000 }
4001
4002 static inline void netif_tx_unlock_bh(struct net_device *dev)
4003 {
4004 netif_tx_unlock(dev);
4005 local_bh_enable();
4006 }
4007
4008 #define HARD_TX_LOCK(dev, txq, cpu) { \
4009 if ((dev->features & NETIF_F_LLTX) == 0) { \
4010 __netif_tx_lock(txq, cpu); \
4011 } else { \
4012 __netif_tx_acquire(txq); \
4013 } \
4014 }
4015
4016 #define HARD_TX_TRYLOCK(dev, txq) \
4017 (((dev->features & NETIF_F_LLTX) == 0) ? \
4018 __netif_tx_trylock(txq) : \
4019 __netif_tx_acquire(txq))
4020
4021 #define HARD_TX_UNLOCK(dev, txq) { \
4022 if ((dev->features & NETIF_F_LLTX) == 0) { \
4023 __netif_tx_unlock(txq); \
4024 } else { \
4025 __netif_tx_release(txq); \
4026 } \
4027 }
4028
4029 static inline void netif_tx_disable(struct net_device *dev)
4030 {
4031 unsigned int i;
4032 int cpu;
4033
4034 local_bh_disable();
4035 cpu = smp_processor_id();
4036 for (i = 0; i < dev->num_tx_queues; i++) {
4037 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4038
4039 __netif_tx_lock(txq, cpu);
4040 netif_tx_stop_queue(txq);
4041 __netif_tx_unlock(txq);
4042 }
4043 local_bh_enable();
4044 }
4045
4046 static inline void netif_addr_lock(struct net_device *dev)
4047 {
4048 spin_lock(&dev->addr_list_lock);
4049 }
4050
4051 static inline void netif_addr_lock_nested(struct net_device *dev)
4052 {
4053 int subclass = SINGLE_DEPTH_NESTING;
4054
4055 if (dev->netdev_ops->ndo_get_lock_subclass)
4056 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4057
4058 spin_lock_nested(&dev->addr_list_lock, subclass);
4059 }
4060
4061 static inline void netif_addr_lock_bh(struct net_device *dev)
4062 {
4063 spin_lock_bh(&dev->addr_list_lock);
4064 }
4065
4066 static inline void netif_addr_unlock(struct net_device *dev)
4067 {
4068 spin_unlock(&dev->addr_list_lock);
4069 }
4070
4071 static inline void netif_addr_unlock_bh(struct net_device *dev)
4072 {
4073 spin_unlock_bh(&dev->addr_list_lock);
4074 }
4075
4076 /*
4077 * dev_addrs walker. Should be used only for read access. Call with
4078 * rcu_read_lock held.
4079 */
4080 #define for_each_dev_addr(dev, ha) \
4081 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4082
4083 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4084
4085 void ether_setup(struct net_device *dev);
4086
4087 /* Support for loadable net-drivers */
4088 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4089 unsigned char name_assign_type,
4090 void (*setup)(struct net_device *),
4091 unsigned int txqs, unsigned int rxqs);
4092 int dev_get_valid_name(struct net *net, struct net_device *dev,
4093 const char *name);
4094
4095 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4096 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4097
4098 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4099 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4100 count)
4101
4102 int register_netdev(struct net_device *dev);
4103 void unregister_netdev(struct net_device *dev);
4104
4105 /* General hardware address lists handling functions */
4106 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4107 struct netdev_hw_addr_list *from_list, int addr_len);
4108 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4109 struct netdev_hw_addr_list *from_list, int addr_len);
4110 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4111 struct net_device *dev,
4112 int (*sync)(struct net_device *, const unsigned char *),
4113 int (*unsync)(struct net_device *,
4114 const unsigned char *));
4115 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4116 struct net_device *dev,
4117 int (*sync)(struct net_device *,
4118 const unsigned char *, int),
4119 int (*unsync)(struct net_device *,
4120 const unsigned char *, int));
4121 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4122 struct net_device *dev,
4123 int (*unsync)(struct net_device *,
4124 const unsigned char *, int));
4125 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4126 struct net_device *dev,
4127 int (*unsync)(struct net_device *,
4128 const unsigned char *));
4129 void __hw_addr_init(struct netdev_hw_addr_list *list);
4130
4131 /* Functions used for device addresses handling */
4132 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4133 unsigned char addr_type);
4134 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4135 unsigned char addr_type);
4136 void dev_addr_flush(struct net_device *dev);
4137 int dev_addr_init(struct net_device *dev);
4138
4139 /* Functions used for unicast addresses handling */
4140 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4141 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4142 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4143 int dev_uc_sync(struct net_device *to, struct net_device *from);
4144 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4145 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4146 void dev_uc_flush(struct net_device *dev);
4147 void dev_uc_init(struct net_device *dev);
4148
4149 /**
4150 * __dev_uc_sync - Synchonize device's unicast list
4151 * @dev: device to sync
4152 * @sync: function to call if address should be added
4153 * @unsync: function to call if address should be removed
4154 *
4155 * Add newly added addresses to the interface, and release
4156 * addresses that have been deleted.
4157 */
4158 static inline int __dev_uc_sync(struct net_device *dev,
4159 int (*sync)(struct net_device *,
4160 const unsigned char *),
4161 int (*unsync)(struct net_device *,
4162 const unsigned char *))
4163 {
4164 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4165 }
4166
4167 /**
4168 * __dev_uc_unsync - Remove synchronized addresses from device
4169 * @dev: device to sync
4170 * @unsync: function to call if address should be removed
4171 *
4172 * Remove all addresses that were added to the device by dev_uc_sync().
4173 */
4174 static inline void __dev_uc_unsync(struct net_device *dev,
4175 int (*unsync)(struct net_device *,
4176 const unsigned char *))
4177 {
4178 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4179 }
4180
4181 /* Functions used for multicast addresses handling */
4182 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4183 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4184 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4185 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4186 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4187 int dev_mc_sync(struct net_device *to, struct net_device *from);
4188 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4189 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4190 void dev_mc_flush(struct net_device *dev);
4191 void dev_mc_init(struct net_device *dev);
4192
4193 /**
4194 * __dev_mc_sync - Synchonize device's multicast list
4195 * @dev: device to sync
4196 * @sync: function to call if address should be added
4197 * @unsync: function to call if address should be removed
4198 *
4199 * Add newly added addresses to the interface, and release
4200 * addresses that have been deleted.
4201 */
4202 static inline int __dev_mc_sync(struct net_device *dev,
4203 int (*sync)(struct net_device *,
4204 const unsigned char *),
4205 int (*unsync)(struct net_device *,
4206 const unsigned char *))
4207 {
4208 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4209 }
4210
4211 /**
4212 * __dev_mc_unsync - Remove synchronized addresses from device
4213 * @dev: device to sync
4214 * @unsync: function to call if address should be removed
4215 *
4216 * Remove all addresses that were added to the device by dev_mc_sync().
4217 */
4218 static inline void __dev_mc_unsync(struct net_device *dev,
4219 int (*unsync)(struct net_device *,
4220 const unsigned char *))
4221 {
4222 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4223 }
4224
4225 /* Functions used for secondary unicast and multicast support */
4226 void dev_set_rx_mode(struct net_device *dev);
4227 void __dev_set_rx_mode(struct net_device *dev);
4228 int dev_set_promiscuity(struct net_device *dev, int inc);
4229 int dev_set_allmulti(struct net_device *dev, int inc);
4230 void netdev_state_change(struct net_device *dev);
4231 void netdev_notify_peers(struct net_device *dev);
4232 void netdev_features_change(struct net_device *dev);
4233 /* Load a device via the kmod */
4234 void dev_load(struct net *net, const char *name);
4235 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4236 struct rtnl_link_stats64 *storage);
4237 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4238 const struct net_device_stats *netdev_stats);
4239
4240 extern int netdev_max_backlog;
4241 extern int netdev_tstamp_prequeue;
4242 extern int weight_p;
4243 extern int dev_weight_rx_bias;
4244 extern int dev_weight_tx_bias;
4245 extern int dev_rx_weight;
4246 extern int dev_tx_weight;
4247
4248 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4249 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4250 struct list_head **iter);
4251 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4252 struct list_head **iter);
4253
4254 /* iterate through upper list, must be called under RCU read lock */
4255 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4256 for (iter = &(dev)->adj_list.upper, \
4257 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4258 updev; \
4259 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4260
4261 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4262 int (*fn)(struct net_device *upper_dev,
4263 void *data),
4264 void *data);
4265
4266 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4267 struct net_device *upper_dev);
4268
4269 bool netdev_has_any_upper_dev(struct net_device *dev);
4270
4271 void *netdev_lower_get_next_private(struct net_device *dev,
4272 struct list_head **iter);
4273 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4274 struct list_head **iter);
4275
4276 #define netdev_for_each_lower_private(dev, priv, iter) \
4277 for (iter = (dev)->adj_list.lower.next, \
4278 priv = netdev_lower_get_next_private(dev, &(iter)); \
4279 priv; \
4280 priv = netdev_lower_get_next_private(dev, &(iter)))
4281
4282 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4283 for (iter = &(dev)->adj_list.lower, \
4284 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4285 priv; \
4286 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4287
4288 void *netdev_lower_get_next(struct net_device *dev,
4289 struct list_head **iter);
4290
4291 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4292 for (iter = (dev)->adj_list.lower.next, \
4293 ldev = netdev_lower_get_next(dev, &(iter)); \
4294 ldev; \
4295 ldev = netdev_lower_get_next(dev, &(iter)))
4296
4297 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4298 struct list_head **iter);
4299 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4300 struct list_head **iter);
4301
4302 int netdev_walk_all_lower_dev(struct net_device *dev,
4303 int (*fn)(struct net_device *lower_dev,
4304 void *data),
4305 void *data);
4306 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4307 int (*fn)(struct net_device *lower_dev,
4308 void *data),
4309 void *data);
4310
4311 void *netdev_adjacent_get_private(struct list_head *adj_list);
4312 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4313 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4314 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4315 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4316 struct netlink_ext_ack *extack);
4317 int netdev_master_upper_dev_link(struct net_device *dev,
4318 struct net_device *upper_dev,
4319 void *upper_priv, void *upper_info,
4320 struct netlink_ext_ack *extack);
4321 void netdev_upper_dev_unlink(struct net_device *dev,
4322 struct net_device *upper_dev);
4323 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4324 void *netdev_lower_dev_get_private(struct net_device *dev,
4325 struct net_device *lower_dev);
4326 void netdev_lower_state_changed(struct net_device *lower_dev,
4327 void *lower_state_info);
4328
4329 /* RSS keys are 40 or 52 bytes long */
4330 #define NETDEV_RSS_KEY_LEN 52
4331 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4332 void netdev_rss_key_fill(void *buffer, size_t len);
4333
4334 int dev_get_nest_level(struct net_device *dev);
4335 int skb_checksum_help(struct sk_buff *skb);
4336 int skb_crc32c_csum_help(struct sk_buff *skb);
4337 int skb_csum_hwoffload_help(struct sk_buff *skb,
4338 const netdev_features_t features);
4339
4340 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4341 netdev_features_t features, bool tx_path);
4342 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4343 netdev_features_t features);
4344
4345 struct netdev_bonding_info {
4346 ifslave slave;
4347 ifbond master;
4348 };
4349
4350 struct netdev_notifier_bonding_info {
4351 struct netdev_notifier_info info; /* must be first */
4352 struct netdev_bonding_info bonding_info;
4353 };
4354
4355 void netdev_bonding_info_change(struct net_device *dev,
4356 struct netdev_bonding_info *bonding_info);
4357
4358 static inline
4359 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4360 {
4361 return __skb_gso_segment(skb, features, true);
4362 }
4363 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4364
4365 static inline bool can_checksum_protocol(netdev_features_t features,
4366 __be16 protocol)
4367 {
4368 if (protocol == htons(ETH_P_FCOE))
4369 return !!(features & NETIF_F_FCOE_CRC);
4370
4371 /* Assume this is an IP checksum (not SCTP CRC) */
4372
4373 if (features & NETIF_F_HW_CSUM) {
4374 /* Can checksum everything */
4375 return true;
4376 }
4377
4378 switch (protocol) {
4379 case htons(ETH_P_IP):
4380 return !!(features & NETIF_F_IP_CSUM);
4381 case htons(ETH_P_IPV6):
4382 return !!(features & NETIF_F_IPV6_CSUM);
4383 default:
4384 return false;
4385 }
4386 }
4387
4388 #ifdef CONFIG_BUG
4389 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4390 #else
4391 static inline void netdev_rx_csum_fault(struct net_device *dev,
4392 struct sk_buff *skb)
4393 {
4394 }
4395 #endif
4396 /* rx skb timestamps */
4397 void net_enable_timestamp(void);
4398 void net_disable_timestamp(void);
4399
4400 #ifdef CONFIG_PROC_FS
4401 int __init dev_proc_init(void);
4402 #else
4403 #define dev_proc_init() 0
4404 #endif
4405
4406 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4407 struct sk_buff *skb, struct net_device *dev,
4408 bool more)
4409 {
4410 __this_cpu_write(softnet_data.xmit.more, more);
4411 return ops->ndo_start_xmit(skb, dev);
4412 }
4413
4414 static inline bool netdev_xmit_more(void)
4415 {
4416 return __this_cpu_read(softnet_data.xmit.more);
4417 }
4418
4419 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4420 struct netdev_queue *txq, bool more)
4421 {
4422 const struct net_device_ops *ops = dev->netdev_ops;
4423 netdev_tx_t rc;
4424
4425 rc = __netdev_start_xmit(ops, skb, dev, more);
4426 if (rc == NETDEV_TX_OK)
4427 txq_trans_update(txq);
4428
4429 return rc;
4430 }
4431
4432 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4433 const void *ns);
4434 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4435 const void *ns);
4436
4437 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4438 {
4439 return netdev_class_create_file_ns(class_attr, NULL);
4440 }
4441
4442 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4443 {
4444 netdev_class_remove_file_ns(class_attr, NULL);
4445 }
4446
4447 extern const struct kobj_ns_type_operations net_ns_type_operations;
4448
4449 const char *netdev_drivername(const struct net_device *dev);
4450
4451 void linkwatch_run_queue(void);
4452
4453 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4454 netdev_features_t f2)
4455 {
4456 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4457 if (f1 & NETIF_F_HW_CSUM)
4458 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4459 else
4460 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4461 }
4462
4463 return f1 & f2;
4464 }
4465
4466 static inline netdev_features_t netdev_get_wanted_features(
4467 struct net_device *dev)
4468 {
4469 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4470 }
4471 netdev_features_t netdev_increment_features(netdev_features_t all,
4472 netdev_features_t one, netdev_features_t mask);
4473
4474 /* Allow TSO being used on stacked device :
4475 * Performing the GSO segmentation before last device
4476 * is a performance improvement.
4477 */
4478 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4479 netdev_features_t mask)
4480 {
4481 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4482 }
4483
4484 int __netdev_update_features(struct net_device *dev);
4485 void netdev_update_features(struct net_device *dev);
4486 void netdev_change_features(struct net_device *dev);
4487
4488 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4489 struct net_device *dev);
4490
4491 netdev_features_t passthru_features_check(struct sk_buff *skb,
4492 struct net_device *dev,
4493 netdev_features_t features);
4494 netdev_features_t netif_skb_features(struct sk_buff *skb);
4495
4496 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4497 {
4498 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4499
4500 /* check flags correspondence */
4501 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4502 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4503 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4504 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4505 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4506 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4507 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4508 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4509 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4510 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4511 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4512 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4513 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4514 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4515 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4516 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4517 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4518 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4519
4520 return (features & feature) == feature;
4521 }
4522
4523 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4524 {
4525 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4526 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4527 }
4528
4529 static inline bool netif_needs_gso(struct sk_buff *skb,
4530 netdev_features_t features)
4531 {
4532 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4533 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4534 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4535 }
4536
4537 static inline void netif_set_gso_max_size(struct net_device *dev,
4538 unsigned int size)
4539 {
4540 dev->gso_max_size = size;
4541 }
4542
4543 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4544 int pulled_hlen, u16 mac_offset,
4545 int mac_len)
4546 {
4547 skb->protocol = protocol;
4548 skb->encapsulation = 1;
4549 skb_push(skb, pulled_hlen);
4550 skb_reset_transport_header(skb);
4551 skb->mac_header = mac_offset;
4552 skb->network_header = skb->mac_header + mac_len;
4553 skb->mac_len = mac_len;
4554 }
4555
4556 static inline bool netif_is_macsec(const struct net_device *dev)
4557 {
4558 return dev->priv_flags & IFF_MACSEC;
4559 }
4560
4561 static inline bool netif_is_macvlan(const struct net_device *dev)
4562 {
4563 return dev->priv_flags & IFF_MACVLAN;
4564 }
4565
4566 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4567 {
4568 return dev->priv_flags & IFF_MACVLAN_PORT;
4569 }
4570
4571 static inline bool netif_is_bond_master(const struct net_device *dev)
4572 {
4573 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4574 }
4575
4576 static inline bool netif_is_bond_slave(const struct net_device *dev)
4577 {
4578 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4579 }
4580
4581 static inline bool netif_supports_nofcs(struct net_device *dev)
4582 {
4583 return dev->priv_flags & IFF_SUPP_NOFCS;
4584 }
4585
4586 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4587 {
4588 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4589 }
4590
4591 static inline bool netif_is_l3_master(const struct net_device *dev)
4592 {
4593 return dev->priv_flags & IFF_L3MDEV_MASTER;
4594 }
4595
4596 static inline bool netif_is_l3_slave(const struct net_device *dev)
4597 {
4598 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4599 }
4600
4601 static inline bool netif_is_bridge_master(const struct net_device *dev)
4602 {
4603 return dev->priv_flags & IFF_EBRIDGE;
4604 }
4605
4606 static inline bool netif_is_bridge_port(const struct net_device *dev)
4607 {
4608 return dev->priv_flags & IFF_BRIDGE_PORT;
4609 }
4610
4611 static inline bool netif_is_ovs_master(const struct net_device *dev)
4612 {
4613 return dev->priv_flags & IFF_OPENVSWITCH;
4614 }
4615
4616 static inline bool netif_is_ovs_port(const struct net_device *dev)
4617 {
4618 return dev->priv_flags & IFF_OVS_DATAPATH;
4619 }
4620
4621 static inline bool netif_is_team_master(const struct net_device *dev)
4622 {
4623 return dev->priv_flags & IFF_TEAM;
4624 }
4625
4626 static inline bool netif_is_team_port(const struct net_device *dev)
4627 {
4628 return dev->priv_flags & IFF_TEAM_PORT;
4629 }
4630
4631 static inline bool netif_is_lag_master(const struct net_device *dev)
4632 {
4633 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4634 }
4635
4636 static inline bool netif_is_lag_port(const struct net_device *dev)
4637 {
4638 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4639 }
4640
4641 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4642 {
4643 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4644 }
4645
4646 static inline bool netif_is_failover(const struct net_device *dev)
4647 {
4648 return dev->priv_flags & IFF_FAILOVER;
4649 }
4650
4651 static inline bool netif_is_failover_slave(const struct net_device *dev)
4652 {
4653 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4654 }
4655
4656 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4657 static inline void netif_keep_dst(struct net_device *dev)
4658 {
4659 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4660 }
4661
4662 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4663 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4664 {
4665 /* TODO: reserve and use an additional IFF bit, if we get more users */
4666 return dev->priv_flags & IFF_MACSEC;
4667 }
4668
4669 extern struct pernet_operations __net_initdata loopback_net_ops;
4670
4671 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4672
4673 /* netdev_printk helpers, similar to dev_printk */
4674
4675 static inline const char *netdev_name(const struct net_device *dev)
4676 {
4677 if (!dev->name[0] || strchr(dev->name, '%'))
4678 return "(unnamed net_device)";
4679 return dev->name;
4680 }
4681
4682 static inline bool netdev_unregistering(const struct net_device *dev)
4683 {
4684 return dev->reg_state == NETREG_UNREGISTERING;
4685 }
4686
4687 static inline const char *netdev_reg_state(const struct net_device *dev)
4688 {
4689 switch (dev->reg_state) {
4690 case NETREG_UNINITIALIZED: return " (uninitialized)";
4691 case NETREG_REGISTERED: return "";
4692 case NETREG_UNREGISTERING: return " (unregistering)";
4693 case NETREG_UNREGISTERED: return " (unregistered)";
4694 case NETREG_RELEASED: return " (released)";
4695 case NETREG_DUMMY: return " (dummy)";
4696 }
4697
4698 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4699 return " (unknown)";
4700 }
4701
4702 __printf(3, 4) __cold
4703 void netdev_printk(const char *level, const struct net_device *dev,
4704 const char *format, ...);
4705 __printf(2, 3) __cold
4706 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4707 __printf(2, 3) __cold
4708 void netdev_alert(const struct net_device *dev, const char *format, ...);
4709 __printf(2, 3) __cold
4710 void netdev_crit(const struct net_device *dev, const char *format, ...);
4711 __printf(2, 3) __cold
4712 void netdev_err(const struct net_device *dev, const char *format, ...);
4713 __printf(2, 3) __cold
4714 void netdev_warn(const struct net_device *dev, const char *format, ...);
4715 __printf(2, 3) __cold
4716 void netdev_notice(const struct net_device *dev, const char *format, ...);
4717 __printf(2, 3) __cold
4718 void netdev_info(const struct net_device *dev, const char *format, ...);
4719
4720 #define netdev_level_once(level, dev, fmt, ...) \
4721 do { \
4722 static bool __print_once __read_mostly; \
4723 \
4724 if (!__print_once) { \
4725 __print_once = true; \
4726 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4727 } \
4728 } while (0)
4729
4730 #define netdev_emerg_once(dev, fmt, ...) \
4731 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4732 #define netdev_alert_once(dev, fmt, ...) \
4733 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4734 #define netdev_crit_once(dev, fmt, ...) \
4735 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4736 #define netdev_err_once(dev, fmt, ...) \
4737 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4738 #define netdev_warn_once(dev, fmt, ...) \
4739 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4740 #define netdev_notice_once(dev, fmt, ...) \
4741 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4742 #define netdev_info_once(dev, fmt, ...) \
4743 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4744
4745 #define MODULE_ALIAS_NETDEV(device) \
4746 MODULE_ALIAS("netdev-" device)
4747
4748 #if defined(CONFIG_DYNAMIC_DEBUG)
4749 #define netdev_dbg(__dev, format, args...) \
4750 do { \
4751 dynamic_netdev_dbg(__dev, format, ##args); \
4752 } while (0)
4753 #elif defined(DEBUG)
4754 #define netdev_dbg(__dev, format, args...) \
4755 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4756 #else
4757 #define netdev_dbg(__dev, format, args...) \
4758 ({ \
4759 if (0) \
4760 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4761 })
4762 #endif
4763
4764 #if defined(VERBOSE_DEBUG)
4765 #define netdev_vdbg netdev_dbg
4766 #else
4767
4768 #define netdev_vdbg(dev, format, args...) \
4769 ({ \
4770 if (0) \
4771 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4772 0; \
4773 })
4774 #endif
4775
4776 /*
4777 * netdev_WARN() acts like dev_printk(), but with the key difference
4778 * of using a WARN/WARN_ON to get the message out, including the
4779 * file/line information and a backtrace.
4780 */
4781 #define netdev_WARN(dev, format, args...) \
4782 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4783 netdev_reg_state(dev), ##args)
4784
4785 #define netdev_WARN_ONCE(dev, format, args...) \
4786 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4787 netdev_reg_state(dev), ##args)
4788
4789 /* netif printk helpers, similar to netdev_printk */
4790
4791 #define netif_printk(priv, type, level, dev, fmt, args...) \
4792 do { \
4793 if (netif_msg_##type(priv)) \
4794 netdev_printk(level, (dev), fmt, ##args); \
4795 } while (0)
4796
4797 #define netif_level(level, priv, type, dev, fmt, args...) \
4798 do { \
4799 if (netif_msg_##type(priv)) \
4800 netdev_##level(dev, fmt, ##args); \
4801 } while (0)
4802
4803 #define netif_emerg(priv, type, dev, fmt, args...) \
4804 netif_level(emerg, priv, type, dev, fmt, ##args)
4805 #define netif_alert(priv, type, dev, fmt, args...) \
4806 netif_level(alert, priv, type, dev, fmt, ##args)
4807 #define netif_crit(priv, type, dev, fmt, args...) \
4808 netif_level(crit, priv, type, dev, fmt, ##args)
4809 #define netif_err(priv, type, dev, fmt, args...) \
4810 netif_level(err, priv, type, dev, fmt, ##args)
4811 #define netif_warn(priv, type, dev, fmt, args...) \
4812 netif_level(warn, priv, type, dev, fmt, ##args)
4813 #define netif_notice(priv, type, dev, fmt, args...) \
4814 netif_level(notice, priv, type, dev, fmt, ##args)
4815 #define netif_info(priv, type, dev, fmt, args...) \
4816 netif_level(info, priv, type, dev, fmt, ##args)
4817
4818 #if defined(CONFIG_DYNAMIC_DEBUG)
4819 #define netif_dbg(priv, type, netdev, format, args...) \
4820 do { \
4821 if (netif_msg_##type(priv)) \
4822 dynamic_netdev_dbg(netdev, format, ##args); \
4823 } while (0)
4824 #elif defined(DEBUG)
4825 #define netif_dbg(priv, type, dev, format, args...) \
4826 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4827 #else
4828 #define netif_dbg(priv, type, dev, format, args...) \
4829 ({ \
4830 if (0) \
4831 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4832 0; \
4833 })
4834 #endif
4835
4836 /* if @cond then downgrade to debug, else print at @level */
4837 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4838 do { \
4839 if (cond) \
4840 netif_dbg(priv, type, netdev, fmt, ##args); \
4841 else \
4842 netif_ ## level(priv, type, netdev, fmt, ##args); \
4843 } while (0)
4844
4845 #if defined(VERBOSE_DEBUG)
4846 #define netif_vdbg netif_dbg
4847 #else
4848 #define netif_vdbg(priv, type, dev, format, args...) \
4849 ({ \
4850 if (0) \
4851 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4852 0; \
4853 })
4854 #endif
4855
4856 /*
4857 * The list of packet types we will receive (as opposed to discard)
4858 * and the routines to invoke.
4859 *
4860 * Why 16. Because with 16 the only overlap we get on a hash of the
4861 * low nibble of the protocol value is RARP/SNAP/X.25.
4862 *
4863 * 0800 IP
4864 * 0001 802.3
4865 * 0002 AX.25
4866 * 0004 802.2
4867 * 8035 RARP
4868 * 0005 SNAP
4869 * 0805 X.25
4870 * 0806 ARP
4871 * 8137 IPX
4872 * 0009 Localtalk
4873 * 86DD IPv6
4874 */
4875 #define PTYPE_HASH_SIZE (16)
4876 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4877
4878 extern struct net_device *blackhole_netdev;
4879
4880 #endif /* _LINUX_NETDEVICE_H */
4881
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.