xref: /linux-6.15/include/linux/netdevice.h (revision 0408c58b)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the Interfaces handler.
7  *
8  * Version:	@(#)dev.h	1.0.10	08/12/93
9  *
10  * Authors:	Ross Biro
11  *		Fred N. van Kempen, <[email protected]>
12  *		Corey Minyard <[email protected]>
13  *		Donald J. Becker, <[email protected]>
14  *		Alan Cox, <[email protected]>
15  *		Bjorn Ekwall. <[email protected]>
16  *              Pekka Riikonen <[email protected]>
17  *
18  *		This program is free software; you can redistribute it and/or
19  *		modify it under the terms of the GNU General Public License
20  *		as published by the Free Software Foundation; either version
21  *		2 of the License, or (at your option) any later version.
22  *
23  *		Moved to /usr/include/linux for NET3
24  */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27 
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35 
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41 
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #ifdef CONFIG_DCB
45 #include <net/dcbnl.h>
46 #endif
47 #include <net/netprio_cgroup.h>
48 
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 struct dsa_switch_tree;
60 
61 /* 802.11 specific */
62 struct wireless_dev;
63 /* 802.15.4 specific */
64 struct wpan_dev;
65 struct mpls_dev;
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct bpf_prog;
69 
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 				    const struct ethtool_ops *ops);
72 
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
75 #define NET_RX_DROP		1	/* packet dropped */
76 
77 /*
78  * Transmit return codes: transmit return codes originate from three different
79  * namespaces:
80  *
81  * - qdisc return codes
82  * - driver transmit return codes
83  * - errno values
84  *
85  * Drivers are allowed to return any one of those in their hard_start_xmit()
86  * function. Real network devices commonly used with qdiscs should only return
87  * the driver transmit return codes though - when qdiscs are used, the actual
88  * transmission happens asynchronously, so the value is not propagated to
89  * higher layers. Virtual network devices transmit synchronously; in this case
90  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91  * others are propagated to higher layers.
92  */
93 
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS	0x00
96 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
97 #define NET_XMIT_CN		0x02	/* congestion notification	*/
98 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
99 
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101  * indicates that the device will soon be dropping packets, or already drops
102  * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK		0xf0
108 
109 enum netdev_tx {
110 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
111 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
112 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
113 };
114 typedef enum netdev_tx netdev_tx_t;
115 
116 /*
117  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119  */
120 static inline bool dev_xmit_complete(int rc)
121 {
122 	/*
123 	 * Positive cases with an skb consumed by a driver:
124 	 * - successful transmission (rc == NETDEV_TX_OK)
125 	 * - error while transmitting (rc < 0)
126 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 	 */
128 	if (likely(rc < NET_XMIT_MASK))
129 		return true;
130 
131 	return false;
132 }
133 
134 /*
135  *	Compute the worst-case header length according to the protocols
136  *	used.
137  */
138 
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 #  define LL_MAX_HEADER 128
144 # else
145 #  define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
150 
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
157 
158 /*
159  *	Old network device statistics. Fields are native words
160  *	(unsigned long) so they can be read and written atomically.
161  */
162 
163 struct net_device_stats {
164 	unsigned long	rx_packets;
165 	unsigned long	tx_packets;
166 	unsigned long	rx_bytes;
167 	unsigned long	tx_bytes;
168 	unsigned long	rx_errors;
169 	unsigned long	tx_errors;
170 	unsigned long	rx_dropped;
171 	unsigned long	tx_dropped;
172 	unsigned long	multicast;
173 	unsigned long	collisions;
174 	unsigned long	rx_length_errors;
175 	unsigned long	rx_over_errors;
176 	unsigned long	rx_crc_errors;
177 	unsigned long	rx_frame_errors;
178 	unsigned long	rx_fifo_errors;
179 	unsigned long	rx_missed_errors;
180 	unsigned long	tx_aborted_errors;
181 	unsigned long	tx_carrier_errors;
182 	unsigned long	tx_fifo_errors;
183 	unsigned long	tx_heartbeat_errors;
184 	unsigned long	tx_window_errors;
185 	unsigned long	rx_compressed;
186 	unsigned long	tx_compressed;
187 };
188 
189 
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
192 
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 extern struct static_key rfs_needed;
197 #endif
198 
199 struct neighbour;
200 struct neigh_parms;
201 struct sk_buff;
202 
203 struct netdev_hw_addr {
204 	struct list_head	list;
205 	unsigned char		addr[MAX_ADDR_LEN];
206 	unsigned char		type;
207 #define NETDEV_HW_ADDR_T_LAN		1
208 #define NETDEV_HW_ADDR_T_SAN		2
209 #define NETDEV_HW_ADDR_T_SLAVE		3
210 #define NETDEV_HW_ADDR_T_UNICAST	4
211 #define NETDEV_HW_ADDR_T_MULTICAST	5
212 	bool			global_use;
213 	int			sync_cnt;
214 	int			refcount;
215 	int			synced;
216 	struct rcu_head		rcu_head;
217 };
218 
219 struct netdev_hw_addr_list {
220 	struct list_head	list;
221 	int			count;
222 };
223 
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 	list_for_each_entry(ha, &(l)->list, list)
228 
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238 
239 struct hh_cache {
240 	unsigned int	hh_len;
241 	seqlock_t	hh_lock;
242 
243 	/* cached hardware header; allow for machine alignment needs.        */
244 #define HH_DATA_MOD	16
245 #define HH_DATA_OFF(__len) \
246 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251 
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253  * Alternative is:
254  *   dev->hard_header_len ? (dev->hard_header_len +
255  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256  *
257  * We could use other alignment values, but we must maintain the
258  * relationship HH alignment <= LL alignment.
259  */
260 #define LL_RESERVED_SPACE(dev) \
261 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 
265 struct header_ops {
266 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
267 			   unsigned short type, const void *daddr,
268 			   const void *saddr, unsigned int len);
269 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 	void	(*cache_update)(struct hh_cache *hh,
272 				const struct net_device *dev,
273 				const unsigned char *haddr);
274 	bool	(*validate)(const char *ll_header, unsigned int len);
275 };
276 
277 /* These flag bits are private to the generic network queueing
278  * layer; they may not be explicitly referenced by any other
279  * code.
280  */
281 
282 enum netdev_state_t {
283 	__LINK_STATE_START,
284 	__LINK_STATE_PRESENT,
285 	__LINK_STATE_NOCARRIER,
286 	__LINK_STATE_LINKWATCH_PENDING,
287 	__LINK_STATE_DORMANT,
288 };
289 
290 
291 /*
292  * This structure holds boot-time configured netdevice settings. They
293  * are then used in the device probing.
294  */
295 struct netdev_boot_setup {
296 	char name[IFNAMSIZ];
297 	struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300 
301 int __init netdev_boot_setup(char *str);
302 
303 /*
304  * Structure for NAPI scheduling similar to tasklet but with weighting
305  */
306 struct napi_struct {
307 	/* The poll_list must only be managed by the entity which
308 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
309 	 * whoever atomically sets that bit can add this napi_struct
310 	 * to the per-CPU poll_list, and whoever clears that bit
311 	 * can remove from the list right before clearing the bit.
312 	 */
313 	struct list_head	poll_list;
314 
315 	unsigned long		state;
316 	int			weight;
317 	unsigned int		gro_count;
318 	int			(*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 	int			poll_owner;
321 #endif
322 	struct net_device	*dev;
323 	struct sk_buff		*gro_list;
324 	struct sk_buff		*skb;
325 	struct hrtimer		timer;
326 	struct list_head	dev_list;
327 	struct hlist_node	napi_hash_node;
328 	unsigned int		napi_id;
329 };
330 
331 enum {
332 	NAPI_STATE_SCHED,	/* Poll is scheduled */
333 	NAPI_STATE_MISSED,	/* reschedule a napi */
334 	NAPI_STATE_DISABLE,	/* Disable pending */
335 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
336 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
337 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
338 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339 };
340 
341 enum {
342 	NAPIF_STATE_SCHED	 = BIT(NAPI_STATE_SCHED),
343 	NAPIF_STATE_MISSED	 = BIT(NAPI_STATE_MISSED),
344 	NAPIF_STATE_DISABLE	 = BIT(NAPI_STATE_DISABLE),
345 	NAPIF_STATE_NPSVC	 = BIT(NAPI_STATE_NPSVC),
346 	NAPIF_STATE_HASHED	 = BIT(NAPI_STATE_HASHED),
347 	NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 	NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
349 };
350 
351 enum gro_result {
352 	GRO_MERGED,
353 	GRO_MERGED_FREE,
354 	GRO_HELD,
355 	GRO_NORMAL,
356 	GRO_DROP,
357 	GRO_CONSUMED,
358 };
359 typedef enum gro_result gro_result_t;
360 
361 /*
362  * enum rx_handler_result - Possible return values for rx_handlers.
363  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364  * further.
365  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366  * case skb->dev was changed by rx_handler.
367  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
369  *
370  * rx_handlers are functions called from inside __netif_receive_skb(), to do
371  * special processing of the skb, prior to delivery to protocol handlers.
372  *
373  * Currently, a net_device can only have a single rx_handler registered. Trying
374  * to register a second rx_handler will return -EBUSY.
375  *
376  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377  * To unregister a rx_handler on a net_device, use
378  * netdev_rx_handler_unregister().
379  *
380  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381  * do with the skb.
382  *
383  * If the rx_handler consumed the skb in some way, it should return
384  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385  * the skb to be delivered in some other way.
386  *
387  * If the rx_handler changed skb->dev, to divert the skb to another
388  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389  * new device will be called if it exists.
390  *
391  * If the rx_handler decides the skb should be ignored, it should return
392  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393  * are registered on exact device (ptype->dev == skb->dev).
394  *
395  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396  * delivered, it should return RX_HANDLER_PASS.
397  *
398  * A device without a registered rx_handler will behave as if rx_handler
399  * returned RX_HANDLER_PASS.
400  */
401 
402 enum rx_handler_result {
403 	RX_HANDLER_CONSUMED,
404 	RX_HANDLER_ANOTHER,
405 	RX_HANDLER_EXACT,
406 	RX_HANDLER_PASS,
407 };
408 typedef enum rx_handler_result rx_handler_result_t;
409 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
410 
411 void __napi_schedule(struct napi_struct *n);
412 void __napi_schedule_irqoff(struct napi_struct *n);
413 
414 static inline bool napi_disable_pending(struct napi_struct *n)
415 {
416 	return test_bit(NAPI_STATE_DISABLE, &n->state);
417 }
418 
419 bool napi_schedule_prep(struct napi_struct *n);
420 
421 /**
422  *	napi_schedule - schedule NAPI poll
423  *	@n: NAPI context
424  *
425  * Schedule NAPI poll routine to be called if it is not already
426  * running.
427  */
428 static inline void napi_schedule(struct napi_struct *n)
429 {
430 	if (napi_schedule_prep(n))
431 		__napi_schedule(n);
432 }
433 
434 /**
435  *	napi_schedule_irqoff - schedule NAPI poll
436  *	@n: NAPI context
437  *
438  * Variant of napi_schedule(), assuming hard irqs are masked.
439  */
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
441 {
442 	if (napi_schedule_prep(n))
443 		__napi_schedule_irqoff(n);
444 }
445 
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
447 static inline bool napi_reschedule(struct napi_struct *napi)
448 {
449 	if (napi_schedule_prep(napi)) {
450 		__napi_schedule(napi);
451 		return true;
452 	}
453 	return false;
454 }
455 
456 bool napi_complete_done(struct napi_struct *n, int work_done);
457 /**
458  *	napi_complete - NAPI processing complete
459  *	@n: NAPI context
460  *
461  * Mark NAPI processing as complete.
462  * Consider using napi_complete_done() instead.
463  * Return false if device should avoid rearming interrupts.
464  */
465 static inline bool napi_complete(struct napi_struct *n)
466 {
467 	return napi_complete_done(n, 0);
468 }
469 
470 /**
471  *	napi_hash_del - remove a NAPI from global table
472  *	@napi: NAPI context
473  *
474  * Warning: caller must observe RCU grace period
475  * before freeing memory containing @napi, if
476  * this function returns true.
477  * Note: core networking stack automatically calls it
478  * from netif_napi_del().
479  * Drivers might want to call this helper to combine all
480  * the needed RCU grace periods into a single one.
481  */
482 bool napi_hash_del(struct napi_struct *napi);
483 
484 /**
485  *	napi_disable - prevent NAPI from scheduling
486  *	@n: NAPI context
487  *
488  * Stop NAPI from being scheduled on this context.
489  * Waits till any outstanding processing completes.
490  */
491 void napi_disable(struct napi_struct *n);
492 
493 /**
494  *	napi_enable - enable NAPI scheduling
495  *	@n: NAPI context
496  *
497  * Resume NAPI from being scheduled on this context.
498  * Must be paired with napi_disable.
499  */
500 static inline void napi_enable(struct napi_struct *n)
501 {
502 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 	smp_mb__before_atomic();
504 	clear_bit(NAPI_STATE_SCHED, &n->state);
505 	clear_bit(NAPI_STATE_NPSVC, &n->state);
506 }
507 
508 /**
509  *	napi_synchronize - wait until NAPI is not running
510  *	@n: NAPI context
511  *
512  * Wait until NAPI is done being scheduled on this context.
513  * Waits till any outstanding processing completes but
514  * does not disable future activations.
515  */
516 static inline void napi_synchronize(const struct napi_struct *n)
517 {
518 	if (IS_ENABLED(CONFIG_SMP))
519 		while (test_bit(NAPI_STATE_SCHED, &n->state))
520 			msleep(1);
521 	else
522 		barrier();
523 }
524 
525 enum netdev_queue_state_t {
526 	__QUEUE_STATE_DRV_XOFF,
527 	__QUEUE_STATE_STACK_XOFF,
528 	__QUEUE_STATE_FROZEN,
529 };
530 
531 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
534 
535 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 					QUEUE_STATE_FROZEN)
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 					QUEUE_STATE_FROZEN)
540 
541 /*
542  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
543  * netif_tx_* functions below are used to manipulate this flag.  The
544  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545  * queue independently.  The netif_xmit_*stopped functions below are called
546  * to check if the queue has been stopped by the driver or stack (either
547  * of the XOFF bits are set in the state).  Drivers should not need to call
548  * netif_xmit*stopped functions, they should only be using netif_tx_*.
549  */
550 
551 struct netdev_queue {
552 /*
553  * read-mostly part
554  */
555 	struct net_device	*dev;
556 	struct Qdisc __rcu	*qdisc;
557 	struct Qdisc		*qdisc_sleeping;
558 #ifdef CONFIG_SYSFS
559 	struct kobject		kobj;
560 #endif
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 	int			numa_node;
563 #endif
564 	unsigned long		tx_maxrate;
565 	/*
566 	 * Number of TX timeouts for this queue
567 	 * (/sys/class/net/DEV/Q/trans_timeout)
568 	 */
569 	unsigned long		trans_timeout;
570 /*
571  * write-mostly part
572  */
573 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
574 	int			xmit_lock_owner;
575 	/*
576 	 * Time (in jiffies) of last Tx
577 	 */
578 	unsigned long		trans_start;
579 
580 	unsigned long		state;
581 
582 #ifdef CONFIG_BQL
583 	struct dql		dql;
584 #endif
585 } ____cacheline_aligned_in_smp;
586 
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588 {
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 	return q->numa_node;
591 #else
592 	return NUMA_NO_NODE;
593 #endif
594 }
595 
596 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597 {
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 	q->numa_node = node;
600 #endif
601 }
602 
603 #ifdef CONFIG_RPS
604 /*
605  * This structure holds an RPS map which can be of variable length.  The
606  * map is an array of CPUs.
607  */
608 struct rps_map {
609 	unsigned int len;
610 	struct rcu_head rcu;
611 	u16 cpus[0];
612 };
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
614 
615 /*
616  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617  * tail pointer for that CPU's input queue at the time of last enqueue, and
618  * a hardware filter index.
619  */
620 struct rps_dev_flow {
621 	u16 cpu;
622 	u16 filter;
623 	unsigned int last_qtail;
624 };
625 #define RPS_NO_FILTER 0xffff
626 
627 /*
628  * The rps_dev_flow_table structure contains a table of flow mappings.
629  */
630 struct rps_dev_flow_table {
631 	unsigned int mask;
632 	struct rcu_head rcu;
633 	struct rps_dev_flow flows[0];
634 };
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636     ((_num) * sizeof(struct rps_dev_flow)))
637 
638 /*
639  * The rps_sock_flow_table contains mappings of flows to the last CPU
640  * on which they were processed by the application (set in recvmsg).
641  * Each entry is a 32bit value. Upper part is the high-order bits
642  * of flow hash, lower part is CPU number.
643  * rps_cpu_mask is used to partition the space, depending on number of
644  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646  * meaning we use 32-6=26 bits for the hash.
647  */
648 struct rps_sock_flow_table {
649 	u32	mask;
650 
651 	u32	ents[0] ____cacheline_aligned_in_smp;
652 };
653 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
654 
655 #define RPS_NO_CPU 0xffff
656 
657 extern u32 rps_cpu_mask;
658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659 
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 					u32 hash)
662 {
663 	if (table && hash) {
664 		unsigned int index = hash & table->mask;
665 		u32 val = hash & ~rps_cpu_mask;
666 
667 		/* We only give a hint, preemption can change CPU under us */
668 		val |= raw_smp_processor_id();
669 
670 		if (table->ents[index] != val)
671 			table->ents[index] = val;
672 	}
673 }
674 
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 			 u16 filter_id);
678 #endif
679 #endif /* CONFIG_RPS */
680 
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue {
683 #ifdef CONFIG_RPS
684 	struct rps_map __rcu		*rps_map;
685 	struct rps_dev_flow_table __rcu	*rps_flow_table;
686 #endif
687 	struct kobject			kobj;
688 	struct net_device		*dev;
689 } ____cacheline_aligned_in_smp;
690 
691 /*
692  * RX queue sysfs structures and functions.
693  */
694 struct rx_queue_attribute {
695 	struct attribute attr;
696 	ssize_t (*show)(struct netdev_rx_queue *queue,
697 	    struct rx_queue_attribute *attr, char *buf);
698 	ssize_t (*store)(struct netdev_rx_queue *queue,
699 	    struct rx_queue_attribute *attr, const char *buf, size_t len);
700 };
701 
702 #ifdef CONFIG_XPS
703 /*
704  * This structure holds an XPS map which can be of variable length.  The
705  * map is an array of queues.
706  */
707 struct xps_map {
708 	unsigned int len;
709 	unsigned int alloc_len;
710 	struct rcu_head rcu;
711 	u16 queues[0];
712 };
713 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
714 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715        - sizeof(struct xps_map)) / sizeof(u16))
716 
717 /*
718  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
719  */
720 struct xps_dev_maps {
721 	struct rcu_head rcu;
722 	struct xps_map __rcu *cpu_map[0];
723 };
724 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +		\
725 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
726 #endif /* CONFIG_XPS */
727 
728 #define TC_MAX_QUEUE	16
729 #define TC_BITMASK	15
730 /* HW offloaded queuing disciplines txq count and offset maps */
731 struct netdev_tc_txq {
732 	u16 count;
733 	u16 offset;
734 };
735 
736 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
737 /*
738  * This structure is to hold information about the device
739  * configured to run FCoE protocol stack.
740  */
741 struct netdev_fcoe_hbainfo {
742 	char	manufacturer[64];
743 	char	serial_number[64];
744 	char	hardware_version[64];
745 	char	driver_version[64];
746 	char	optionrom_version[64];
747 	char	firmware_version[64];
748 	char	model[256];
749 	char	model_description[256];
750 };
751 #endif
752 
753 #define MAX_PHYS_ITEM_ID_LEN 32
754 
755 /* This structure holds a unique identifier to identify some
756  * physical item (port for example) used by a netdevice.
757  */
758 struct netdev_phys_item_id {
759 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
760 	unsigned char id_len;
761 };
762 
763 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 					    struct netdev_phys_item_id *b)
765 {
766 	return a->id_len == b->id_len &&
767 	       memcmp(a->id, b->id, a->id_len) == 0;
768 }
769 
770 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 				       struct sk_buff *skb);
772 
773 /* These structures hold the attributes of qdisc and classifiers
774  * that are being passed to the netdevice through the setup_tc op.
775  */
776 enum {
777 	TC_SETUP_MQPRIO,
778 	TC_SETUP_CLSU32,
779 	TC_SETUP_CLSFLOWER,
780 	TC_SETUP_MATCHALL,
781 	TC_SETUP_CLSBPF,
782 };
783 
784 struct tc_cls_u32_offload;
785 
786 struct tc_to_netdev {
787 	unsigned int type;
788 	union {
789 		struct tc_cls_u32_offload *cls_u32;
790 		struct tc_cls_flower_offload *cls_flower;
791 		struct tc_cls_matchall_offload *cls_mall;
792 		struct tc_cls_bpf_offload *cls_bpf;
793 		struct tc_mqprio_qopt *mqprio;
794 	};
795 	bool egress_dev;
796 };
797 
798 /* These structures hold the attributes of xdp state that are being passed
799  * to the netdevice through the xdp op.
800  */
801 enum xdp_netdev_command {
802 	/* Set or clear a bpf program used in the earliest stages of packet
803 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 	 * is responsible for calling bpf_prog_put on any old progs that are
805 	 * stored. In case of error, the callee need not release the new prog
806 	 * reference, but on success it takes ownership and must bpf_prog_put
807 	 * when it is no longer used.
808 	 */
809 	XDP_SETUP_PROG,
810 	XDP_SETUP_PROG_HW,
811 	/* Check if a bpf program is set on the device.  The callee should
812 	 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
813 	 * is equivalent to XDP_ATTACHED_DRV.
814 	 */
815 	XDP_QUERY_PROG,
816 };
817 
818 struct netlink_ext_ack;
819 
820 struct netdev_xdp {
821 	enum xdp_netdev_command command;
822 	union {
823 		/* XDP_SETUP_PROG */
824 		struct {
825 			u32 flags;
826 			struct bpf_prog *prog;
827 			struct netlink_ext_ack *extack;
828 		};
829 		/* XDP_QUERY_PROG */
830 		struct {
831 			u8 prog_attached;
832 			u32 prog_id;
833 		};
834 	};
835 };
836 
837 #ifdef CONFIG_XFRM_OFFLOAD
838 struct xfrmdev_ops {
839 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
840 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
841 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
842 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
843 				       struct xfrm_state *x);
844 };
845 #endif
846 
847 /*
848  * This structure defines the management hooks for network devices.
849  * The following hooks can be defined; unless noted otherwise, they are
850  * optional and can be filled with a null pointer.
851  *
852  * int (*ndo_init)(struct net_device *dev);
853  *     This function is called once when a network device is registered.
854  *     The network device can use this for any late stage initialization
855  *     or semantic validation. It can fail with an error code which will
856  *     be propagated back to register_netdev.
857  *
858  * void (*ndo_uninit)(struct net_device *dev);
859  *     This function is called when device is unregistered or when registration
860  *     fails. It is not called if init fails.
861  *
862  * int (*ndo_open)(struct net_device *dev);
863  *     This function is called when a network device transitions to the up
864  *     state.
865  *
866  * int (*ndo_stop)(struct net_device *dev);
867  *     This function is called when a network device transitions to the down
868  *     state.
869  *
870  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
871  *                               struct net_device *dev);
872  *	Called when a packet needs to be transmitted.
873  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
874  *	the queue before that can happen; it's for obsolete devices and weird
875  *	corner cases, but the stack really does a non-trivial amount
876  *	of useless work if you return NETDEV_TX_BUSY.
877  *	Required; cannot be NULL.
878  *
879  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
880  *					   struct net_device *dev
881  *					   netdev_features_t features);
882  *	Called by core transmit path to determine if device is capable of
883  *	performing offload operations on a given packet. This is to give
884  *	the device an opportunity to implement any restrictions that cannot
885  *	be otherwise expressed by feature flags. The check is called with
886  *	the set of features that the stack has calculated and it returns
887  *	those the driver believes to be appropriate.
888  *
889  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
890  *                         void *accel_priv, select_queue_fallback_t fallback);
891  *	Called to decide which queue to use when device supports multiple
892  *	transmit queues.
893  *
894  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
895  *	This function is called to allow device receiver to make
896  *	changes to configuration when multicast or promiscuous is enabled.
897  *
898  * void (*ndo_set_rx_mode)(struct net_device *dev);
899  *	This function is called device changes address list filtering.
900  *	If driver handles unicast address filtering, it should set
901  *	IFF_UNICAST_FLT in its priv_flags.
902  *
903  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
904  *	This function  is called when the Media Access Control address
905  *	needs to be changed. If this interface is not defined, the
906  *	MAC address can not be changed.
907  *
908  * int (*ndo_validate_addr)(struct net_device *dev);
909  *	Test if Media Access Control address is valid for the device.
910  *
911  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
912  *	Called when a user requests an ioctl which can't be handled by
913  *	the generic interface code. If not defined ioctls return
914  *	not supported error code.
915  *
916  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
917  *	Used to set network devices bus interface parameters. This interface
918  *	is retained for legacy reasons; new devices should use the bus
919  *	interface (PCI) for low level management.
920  *
921  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
922  *	Called when a user wants to change the Maximum Transfer Unit
923  *	of a device.
924  *
925  * void (*ndo_tx_timeout)(struct net_device *dev);
926  *	Callback used when the transmitter has not made any progress
927  *	for dev->watchdog ticks.
928  *
929  * void (*ndo_get_stats64)(struct net_device *dev,
930  *                         struct rtnl_link_stats64 *storage);
931  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
932  *	Called when a user wants to get the network device usage
933  *	statistics. Drivers must do one of the following:
934  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
935  *	   rtnl_link_stats64 structure passed by the caller.
936  *	2. Define @ndo_get_stats to update a net_device_stats structure
937  *	   (which should normally be dev->stats) and return a pointer to
938  *	   it. The structure may be changed asynchronously only if each
939  *	   field is written atomically.
940  *	3. Update dev->stats asynchronously and atomically, and define
941  *	   neither operation.
942  *
943  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
944  *	Return true if this device supports offload stats of this attr_id.
945  *
946  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
947  *	void *attr_data)
948  *	Get statistics for offload operations by attr_id. Write it into the
949  *	attr_data pointer.
950  *
951  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
952  *	If device supports VLAN filtering this function is called when a
953  *	VLAN id is registered.
954  *
955  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
956  *	If device supports VLAN filtering this function is called when a
957  *	VLAN id is unregistered.
958  *
959  * void (*ndo_poll_controller)(struct net_device *dev);
960  *
961  *	SR-IOV management functions.
962  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
963  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
964  *			  u8 qos, __be16 proto);
965  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
966  *			  int max_tx_rate);
967  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
968  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
969  * int (*ndo_get_vf_config)(struct net_device *dev,
970  *			    int vf, struct ifla_vf_info *ivf);
971  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
972  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
973  *			  struct nlattr *port[]);
974  *
975  *      Enable or disable the VF ability to query its RSS Redirection Table and
976  *      Hash Key. This is needed since on some devices VF share this information
977  *      with PF and querying it may introduce a theoretical security risk.
978  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
979  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
980  * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
981  *		       __be16 protocol, struct tc_to_netdev *tc);
982  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
983  *	This is always called from the stack with the rtnl lock held and netif
984  *	tx queues stopped. This allows the netdevice to perform queue
985  *	management safely.
986  *
987  *	Fiber Channel over Ethernet (FCoE) offload functions.
988  * int (*ndo_fcoe_enable)(struct net_device *dev);
989  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
990  *	so the underlying device can perform whatever needed configuration or
991  *	initialization to support acceleration of FCoE traffic.
992  *
993  * int (*ndo_fcoe_disable)(struct net_device *dev);
994  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
995  *	so the underlying device can perform whatever needed clean-ups to
996  *	stop supporting acceleration of FCoE traffic.
997  *
998  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
999  *			     struct scatterlist *sgl, unsigned int sgc);
1000  *	Called when the FCoE Initiator wants to initialize an I/O that
1001  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1002  *	perform necessary setup and returns 1 to indicate the device is set up
1003  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1004  *
1005  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1006  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1007  *	indicated by the FC exchange id 'xid', so the underlying device can
1008  *	clean up and reuse resources for later DDP requests.
1009  *
1010  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1011  *			      struct scatterlist *sgl, unsigned int sgc);
1012  *	Called when the FCoE Target wants to initialize an I/O that
1013  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1014  *	perform necessary setup and returns 1 to indicate the device is set up
1015  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1016  *
1017  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1018  *			       struct netdev_fcoe_hbainfo *hbainfo);
1019  *	Called when the FCoE Protocol stack wants information on the underlying
1020  *	device. This information is utilized by the FCoE protocol stack to
1021  *	register attributes with Fiber Channel management service as per the
1022  *	FC-GS Fabric Device Management Information(FDMI) specification.
1023  *
1024  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1025  *	Called when the underlying device wants to override default World Wide
1026  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1027  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1028  *	protocol stack to use.
1029  *
1030  *	RFS acceleration.
1031  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1032  *			    u16 rxq_index, u32 flow_id);
1033  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1034  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1035  *	Return the filter ID on success, or a negative error code.
1036  *
1037  *	Slave management functions (for bridge, bonding, etc).
1038  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1039  *	Called to make another netdev an underling.
1040  *
1041  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1042  *	Called to release previously enslaved netdev.
1043  *
1044  *      Feature/offload setting functions.
1045  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1046  *		netdev_features_t features);
1047  *	Adjusts the requested feature flags according to device-specific
1048  *	constraints, and returns the resulting flags. Must not modify
1049  *	the device state.
1050  *
1051  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1052  *	Called to update device configuration to new features. Passed
1053  *	feature set might be less than what was returned by ndo_fix_features()).
1054  *	Must return >0 or -errno if it changed dev->features itself.
1055  *
1056  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1057  *		      struct net_device *dev,
1058  *		      const unsigned char *addr, u16 vid, u16 flags)
1059  *	Adds an FDB entry to dev for addr.
1060  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1061  *		      struct net_device *dev,
1062  *		      const unsigned char *addr, u16 vid)
1063  *	Deletes the FDB entry from dev coresponding to addr.
1064  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1065  *		       struct net_device *dev, struct net_device *filter_dev,
1066  *		       int *idx)
1067  *	Used to add FDB entries to dump requests. Implementers should add
1068  *	entries to skb and update idx with the number of entries.
1069  *
1070  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1071  *			     u16 flags)
1072  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1073  *			     struct net_device *dev, u32 filter_mask,
1074  *			     int nlflags)
1075  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1076  *			     u16 flags);
1077  *
1078  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1079  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1080  *	which do not represent real hardware may define this to allow their
1081  *	userspace components to manage their virtual carrier state. Devices
1082  *	that determine carrier state from physical hardware properties (eg
1083  *	network cables) or protocol-dependent mechanisms (eg
1084  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1085  *
1086  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1087  *			       struct netdev_phys_item_id *ppid);
1088  *	Called to get ID of physical port of this device. If driver does
1089  *	not implement this, it is assumed that the hw is not able to have
1090  *	multiple net devices on single physical port.
1091  *
1092  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1093  *			      struct udp_tunnel_info *ti);
1094  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1095  *	address family that a UDP tunnel is listnening to. It is called only
1096  *	when a new port starts listening. The operation is protected by the
1097  *	RTNL.
1098  *
1099  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1100  *			      struct udp_tunnel_info *ti);
1101  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1102  *	address family that the UDP tunnel is not listening to anymore. The
1103  *	operation is protected by the RTNL.
1104  *
1105  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1106  *				 struct net_device *dev)
1107  *	Called by upper layer devices to accelerate switching or other
1108  *	station functionality into hardware. 'pdev is the lowerdev
1109  *	to use for the offload and 'dev' is the net device that will
1110  *	back the offload. Returns a pointer to the private structure
1111  *	the upper layer will maintain.
1112  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1113  *	Called by upper layer device to delete the station created
1114  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1115  *	the station and priv is the structure returned by the add
1116  *	operation.
1117  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1118  *			     int queue_index, u32 maxrate);
1119  *	Called when a user wants to set a max-rate limitation of specific
1120  *	TX queue.
1121  * int (*ndo_get_iflink)(const struct net_device *dev);
1122  *	Called to get the iflink value of this device.
1123  * void (*ndo_change_proto_down)(struct net_device *dev,
1124  *				 bool proto_down);
1125  *	This function is used to pass protocol port error state information
1126  *	to the switch driver. The switch driver can react to the proto_down
1127  *      by doing a phys down on the associated switch port.
1128  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1129  *	This function is used to get egress tunnel information for given skb.
1130  *	This is useful for retrieving outer tunnel header parameters while
1131  *	sampling packet.
1132  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1133  *	This function is used to specify the headroom that the skb must
1134  *	consider when allocation skb during packet reception. Setting
1135  *	appropriate rx headroom value allows avoiding skb head copy on
1136  *	forward. Setting a negative value resets the rx headroom to the
1137  *	default value.
1138  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1139  *	This function is used to set or query state related to XDP on the
1140  *	netdevice. See definition of enum xdp_netdev_command for details.
1141  *
1142  */
1143 struct net_device_ops {
1144 	int			(*ndo_init)(struct net_device *dev);
1145 	void			(*ndo_uninit)(struct net_device *dev);
1146 	int			(*ndo_open)(struct net_device *dev);
1147 	int			(*ndo_stop)(struct net_device *dev);
1148 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1149 						  struct net_device *dev);
1150 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1151 						      struct net_device *dev,
1152 						      netdev_features_t features);
1153 	u16			(*ndo_select_queue)(struct net_device *dev,
1154 						    struct sk_buff *skb,
1155 						    void *accel_priv,
1156 						    select_queue_fallback_t fallback);
1157 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1158 						       int flags);
1159 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1160 	int			(*ndo_set_mac_address)(struct net_device *dev,
1161 						       void *addr);
1162 	int			(*ndo_validate_addr)(struct net_device *dev);
1163 	int			(*ndo_do_ioctl)(struct net_device *dev,
1164 					        struct ifreq *ifr, int cmd);
1165 	int			(*ndo_set_config)(struct net_device *dev,
1166 					          struct ifmap *map);
1167 	int			(*ndo_change_mtu)(struct net_device *dev,
1168 						  int new_mtu);
1169 	int			(*ndo_neigh_setup)(struct net_device *dev,
1170 						   struct neigh_parms *);
1171 	void			(*ndo_tx_timeout) (struct net_device *dev);
1172 
1173 	void			(*ndo_get_stats64)(struct net_device *dev,
1174 						   struct rtnl_link_stats64 *storage);
1175 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1176 	int			(*ndo_get_offload_stats)(int attr_id,
1177 							 const struct net_device *dev,
1178 							 void *attr_data);
1179 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1180 
1181 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1182 						       __be16 proto, u16 vid);
1183 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1184 						        __be16 proto, u16 vid);
1185 #ifdef CONFIG_NET_POLL_CONTROLLER
1186 	void                    (*ndo_poll_controller)(struct net_device *dev);
1187 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1188 						     struct netpoll_info *info);
1189 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1190 #endif
1191 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1192 						  int queue, u8 *mac);
1193 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1194 						   int queue, u16 vlan,
1195 						   u8 qos, __be16 proto);
1196 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1197 						   int vf, int min_tx_rate,
1198 						   int max_tx_rate);
1199 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1200 						       int vf, bool setting);
1201 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1202 						    int vf, bool setting);
1203 	int			(*ndo_get_vf_config)(struct net_device *dev,
1204 						     int vf,
1205 						     struct ifla_vf_info *ivf);
1206 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1207 							 int vf, int link_state);
1208 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1209 						    int vf,
1210 						    struct ifla_vf_stats
1211 						    *vf_stats);
1212 	int			(*ndo_set_vf_port)(struct net_device *dev,
1213 						   int vf,
1214 						   struct nlattr *port[]);
1215 	int			(*ndo_get_vf_port)(struct net_device *dev,
1216 						   int vf, struct sk_buff *skb);
1217 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1218 						   int vf, u64 guid,
1219 						   int guid_type);
1220 	int			(*ndo_set_vf_rss_query_en)(
1221 						   struct net_device *dev,
1222 						   int vf, bool setting);
1223 	int			(*ndo_setup_tc)(struct net_device *dev,
1224 						u32 handle, u32 chain_index,
1225 						__be16 protocol,
1226 						struct tc_to_netdev *tc);
1227 #if IS_ENABLED(CONFIG_FCOE)
1228 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1229 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1230 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1231 						      u16 xid,
1232 						      struct scatterlist *sgl,
1233 						      unsigned int sgc);
1234 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1235 						     u16 xid);
1236 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1237 						       u16 xid,
1238 						       struct scatterlist *sgl,
1239 						       unsigned int sgc);
1240 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1241 							struct netdev_fcoe_hbainfo *hbainfo);
1242 #endif
1243 
1244 #if IS_ENABLED(CONFIG_LIBFCOE)
1245 #define NETDEV_FCOE_WWNN 0
1246 #define NETDEV_FCOE_WWPN 1
1247 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1248 						    u64 *wwn, int type);
1249 #endif
1250 
1251 #ifdef CONFIG_RFS_ACCEL
1252 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1253 						     const struct sk_buff *skb,
1254 						     u16 rxq_index,
1255 						     u32 flow_id);
1256 #endif
1257 	int			(*ndo_add_slave)(struct net_device *dev,
1258 						 struct net_device *slave_dev);
1259 	int			(*ndo_del_slave)(struct net_device *dev,
1260 						 struct net_device *slave_dev);
1261 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1262 						    netdev_features_t features);
1263 	int			(*ndo_set_features)(struct net_device *dev,
1264 						    netdev_features_t features);
1265 	int			(*ndo_neigh_construct)(struct net_device *dev,
1266 						       struct neighbour *n);
1267 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1268 						     struct neighbour *n);
1269 
1270 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1271 					       struct nlattr *tb[],
1272 					       struct net_device *dev,
1273 					       const unsigned char *addr,
1274 					       u16 vid,
1275 					       u16 flags);
1276 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1277 					       struct nlattr *tb[],
1278 					       struct net_device *dev,
1279 					       const unsigned char *addr,
1280 					       u16 vid);
1281 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1282 						struct netlink_callback *cb,
1283 						struct net_device *dev,
1284 						struct net_device *filter_dev,
1285 						int *idx);
1286 
1287 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1288 						      struct nlmsghdr *nlh,
1289 						      u16 flags);
1290 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1291 						      u32 pid, u32 seq,
1292 						      struct net_device *dev,
1293 						      u32 filter_mask,
1294 						      int nlflags);
1295 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1296 						      struct nlmsghdr *nlh,
1297 						      u16 flags);
1298 	int			(*ndo_change_carrier)(struct net_device *dev,
1299 						      bool new_carrier);
1300 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1301 							struct netdev_phys_item_id *ppid);
1302 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1303 							  char *name, size_t len);
1304 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1305 						      struct udp_tunnel_info *ti);
1306 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1307 						      struct udp_tunnel_info *ti);
1308 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1309 							struct net_device *dev);
1310 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1311 							void *priv);
1312 
1313 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1314 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1315 						      int queue_index,
1316 						      u32 maxrate);
1317 	int			(*ndo_get_iflink)(const struct net_device *dev);
1318 	int			(*ndo_change_proto_down)(struct net_device *dev,
1319 							 bool proto_down);
1320 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1321 						       struct sk_buff *skb);
1322 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1323 						       int needed_headroom);
1324 	int			(*ndo_xdp)(struct net_device *dev,
1325 					   struct netdev_xdp *xdp);
1326 };
1327 
1328 /**
1329  * enum net_device_priv_flags - &struct net_device priv_flags
1330  *
1331  * These are the &struct net_device, they are only set internally
1332  * by drivers and used in the kernel. These flags are invisible to
1333  * userspace; this means that the order of these flags can change
1334  * during any kernel release.
1335  *
1336  * You should have a pretty good reason to be extending these flags.
1337  *
1338  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1339  * @IFF_EBRIDGE: Ethernet bridging device
1340  * @IFF_BONDING: bonding master or slave
1341  * @IFF_ISATAP: ISATAP interface (RFC4214)
1342  * @IFF_WAN_HDLC: WAN HDLC device
1343  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1344  *	release skb->dst
1345  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1346  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1347  * @IFF_MACVLAN_PORT: device used as macvlan port
1348  * @IFF_BRIDGE_PORT: device used as bridge port
1349  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1350  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1351  * @IFF_UNICAST_FLT: Supports unicast filtering
1352  * @IFF_TEAM_PORT: device used as team port
1353  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1354  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1355  *	change when it's running
1356  * @IFF_MACVLAN: Macvlan device
1357  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1358  *	underlying stacked devices
1359  * @IFF_IPVLAN_MASTER: IPvlan master device
1360  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1361  * @IFF_L3MDEV_MASTER: device is an L3 master device
1362  * @IFF_NO_QUEUE: device can run without qdisc attached
1363  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1364  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1365  * @IFF_TEAM: device is a team device
1366  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1367  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1368  *	entity (i.e. the master device for bridged veth)
1369  * @IFF_MACSEC: device is a MACsec device
1370  */
1371 enum netdev_priv_flags {
1372 	IFF_802_1Q_VLAN			= 1<<0,
1373 	IFF_EBRIDGE			= 1<<1,
1374 	IFF_BONDING			= 1<<2,
1375 	IFF_ISATAP			= 1<<3,
1376 	IFF_WAN_HDLC			= 1<<4,
1377 	IFF_XMIT_DST_RELEASE		= 1<<5,
1378 	IFF_DONT_BRIDGE			= 1<<6,
1379 	IFF_DISABLE_NETPOLL		= 1<<7,
1380 	IFF_MACVLAN_PORT		= 1<<8,
1381 	IFF_BRIDGE_PORT			= 1<<9,
1382 	IFF_OVS_DATAPATH		= 1<<10,
1383 	IFF_TX_SKB_SHARING		= 1<<11,
1384 	IFF_UNICAST_FLT			= 1<<12,
1385 	IFF_TEAM_PORT			= 1<<13,
1386 	IFF_SUPP_NOFCS			= 1<<14,
1387 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1388 	IFF_MACVLAN			= 1<<16,
1389 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1390 	IFF_IPVLAN_MASTER		= 1<<18,
1391 	IFF_IPVLAN_SLAVE		= 1<<19,
1392 	IFF_L3MDEV_MASTER		= 1<<20,
1393 	IFF_NO_QUEUE			= 1<<21,
1394 	IFF_OPENVSWITCH			= 1<<22,
1395 	IFF_L3MDEV_SLAVE		= 1<<23,
1396 	IFF_TEAM			= 1<<24,
1397 	IFF_RXFH_CONFIGURED		= 1<<25,
1398 	IFF_PHONY_HEADROOM		= 1<<26,
1399 	IFF_MACSEC			= 1<<27,
1400 };
1401 
1402 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1403 #define IFF_EBRIDGE			IFF_EBRIDGE
1404 #define IFF_BONDING			IFF_BONDING
1405 #define IFF_ISATAP			IFF_ISATAP
1406 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1407 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1408 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1409 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1410 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1411 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1412 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1413 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1414 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1415 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1416 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1417 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1418 #define IFF_MACVLAN			IFF_MACVLAN
1419 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1420 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1421 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1422 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1423 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1424 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1425 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1426 #define IFF_TEAM			IFF_TEAM
1427 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1428 #define IFF_MACSEC			IFF_MACSEC
1429 
1430 /**
1431  *	struct net_device - The DEVICE structure.
1432  *		Actually, this whole structure is a big mistake.  It mixes I/O
1433  *		data with strictly "high-level" data, and it has to know about
1434  *		almost every data structure used in the INET module.
1435  *
1436  *	@name:	This is the first field of the "visible" part of this structure
1437  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1438  *	 	of the interface.
1439  *
1440  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1441  *	@ifalias:	SNMP alias
1442  *	@mem_end:	Shared memory end
1443  *	@mem_start:	Shared memory start
1444  *	@base_addr:	Device I/O address
1445  *	@irq:		Device IRQ number
1446  *
1447  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1448  *
1449  *	@state:		Generic network queuing layer state, see netdev_state_t
1450  *	@dev_list:	The global list of network devices
1451  *	@napi_list:	List entry used for polling NAPI devices
1452  *	@unreg_list:	List entry  when we are unregistering the
1453  *			device; see the function unregister_netdev
1454  *	@close_list:	List entry used when we are closing the device
1455  *	@ptype_all:     Device-specific packet handlers for all protocols
1456  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1457  *
1458  *	@adj_list:	Directly linked devices, like slaves for bonding
1459  *	@features:	Currently active device features
1460  *	@hw_features:	User-changeable features
1461  *
1462  *	@wanted_features:	User-requested features
1463  *	@vlan_features:		Mask of features inheritable by VLAN devices
1464  *
1465  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1466  *				This field indicates what encapsulation
1467  *				offloads the hardware is capable of doing,
1468  *				and drivers will need to set them appropriately.
1469  *
1470  *	@mpls_features:	Mask of features inheritable by MPLS
1471  *
1472  *	@ifindex:	interface index
1473  *	@group:		The group the device belongs to
1474  *
1475  *	@stats:		Statistics struct, which was left as a legacy, use
1476  *			rtnl_link_stats64 instead
1477  *
1478  *	@rx_dropped:	Dropped packets by core network,
1479  *			do not use this in drivers
1480  *	@tx_dropped:	Dropped packets by core network,
1481  *			do not use this in drivers
1482  *	@rx_nohandler:	nohandler dropped packets by core network on
1483  *			inactive devices, do not use this in drivers
1484  *
1485  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1486  *				instead of ioctl,
1487  *				see <net/iw_handler.h> for details.
1488  *	@wireless_data:	Instance data managed by the core of wireless extensions
1489  *
1490  *	@netdev_ops:	Includes several pointers to callbacks,
1491  *			if one wants to override the ndo_*() functions
1492  *	@ethtool_ops:	Management operations
1493  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1494  *			discovery handling. Necessary for e.g. 6LoWPAN.
1495  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1496  *			of Layer 2 headers.
1497  *
1498  *	@flags:		Interface flags (a la BSD)
1499  *	@priv_flags:	Like 'flags' but invisible to userspace,
1500  *			see if.h for the definitions
1501  *	@gflags:	Global flags ( kept as legacy )
1502  *	@padded:	How much padding added by alloc_netdev()
1503  *	@operstate:	RFC2863 operstate
1504  *	@link_mode:	Mapping policy to operstate
1505  *	@if_port:	Selectable AUI, TP, ...
1506  *	@dma:		DMA channel
1507  *	@mtu:		Interface MTU value
1508  *	@min_mtu:	Interface Minimum MTU value
1509  *	@max_mtu:	Interface Maximum MTU value
1510  *	@type:		Interface hardware type
1511  *	@hard_header_len: Maximum hardware header length.
1512  *	@min_header_len:  Minimum hardware header length
1513  *
1514  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1515  *			  cases can this be guaranteed
1516  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1517  *			  cases can this be guaranteed. Some cases also use
1518  *			  LL_MAX_HEADER instead to allocate the skb
1519  *
1520  *	interface address info:
1521  *
1522  * 	@perm_addr:		Permanent hw address
1523  * 	@addr_assign_type:	Hw address assignment type
1524  * 	@addr_len:		Hardware address length
1525  *	@neigh_priv_len:	Used in neigh_alloc()
1526  * 	@dev_id:		Used to differentiate devices that share
1527  * 				the same link layer address
1528  * 	@dev_port:		Used to differentiate devices that share
1529  * 				the same function
1530  *	@addr_list_lock:	XXX: need comments on this one
1531  *	@uc_promisc:		Counter that indicates promiscuous mode
1532  *				has been enabled due to the need to listen to
1533  *				additional unicast addresses in a device that
1534  *				does not implement ndo_set_rx_mode()
1535  *	@uc:			unicast mac addresses
1536  *	@mc:			multicast mac addresses
1537  *	@dev_addrs:		list of device hw addresses
1538  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1539  *	@promiscuity:		Number of times the NIC is told to work in
1540  *				promiscuous mode; if it becomes 0 the NIC will
1541  *				exit promiscuous mode
1542  *	@allmulti:		Counter, enables or disables allmulticast mode
1543  *
1544  *	@vlan_info:	VLAN info
1545  *	@dsa_ptr:	dsa specific data
1546  *	@tipc_ptr:	TIPC specific data
1547  *	@atalk_ptr:	AppleTalk link
1548  *	@ip_ptr:	IPv4 specific data
1549  *	@dn_ptr:	DECnet specific data
1550  *	@ip6_ptr:	IPv6 specific data
1551  *	@ax25_ptr:	AX.25 specific data
1552  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1553  *
1554  *	@dev_addr:	Hw address (before bcast,
1555  *			because most packets are unicast)
1556  *
1557  *	@_rx:			Array of RX queues
1558  *	@num_rx_queues:		Number of RX queues
1559  *				allocated at register_netdev() time
1560  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1561  *
1562  *	@rx_handler:		handler for received packets
1563  *	@rx_handler_data: 	XXX: need comments on this one
1564  *	@ingress_queue:		XXX: need comments on this one
1565  *	@broadcast:		hw bcast address
1566  *
1567  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1568  *			indexed by RX queue number. Assigned by driver.
1569  *			This must only be set if the ndo_rx_flow_steer
1570  *			operation is defined
1571  *	@index_hlist:		Device index hash chain
1572  *
1573  *	@_tx:			Array of TX queues
1574  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1575  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1576  *	@qdisc:			Root qdisc from userspace point of view
1577  *	@tx_queue_len:		Max frames per queue allowed
1578  *	@tx_global_lock: 	XXX: need comments on this one
1579  *
1580  *	@xps_maps:	XXX: need comments on this one
1581  *
1582  *	@watchdog_timeo:	Represents the timeout that is used by
1583  *				the watchdog (see dev_watchdog())
1584  *	@watchdog_timer:	List of timers
1585  *
1586  *	@pcpu_refcnt:		Number of references to this device
1587  *	@todo_list:		Delayed register/unregister
1588  *	@link_watch_list:	XXX: need comments on this one
1589  *
1590  *	@reg_state:		Register/unregister state machine
1591  *	@dismantle:		Device is going to be freed
1592  *	@rtnl_link_state:	This enum represents the phases of creating
1593  *				a new link
1594  *
1595  *	@needs_free_netdev:	Should unregister perform free_netdev?
1596  *	@priv_destructor:	Called from unregister
1597  *	@npinfo:		XXX: need comments on this one
1598  * 	@nd_net:		Network namespace this network device is inside
1599  *
1600  * 	@ml_priv:	Mid-layer private
1601  * 	@lstats:	Loopback statistics
1602  * 	@tstats:	Tunnel statistics
1603  * 	@dstats:	Dummy statistics
1604  * 	@vstats:	Virtual ethernet statistics
1605  *
1606  *	@garp_port:	GARP
1607  *	@mrp_port:	MRP
1608  *
1609  *	@dev:		Class/net/name entry
1610  *	@sysfs_groups:	Space for optional device, statistics and wireless
1611  *			sysfs groups
1612  *
1613  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1614  *	@rtnl_link_ops:	Rtnl_link_ops
1615  *
1616  *	@gso_max_size:	Maximum size of generic segmentation offload
1617  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1618  *			NIC for GSO
1619  *
1620  *	@dcbnl_ops:	Data Center Bridging netlink ops
1621  *	@num_tc:	Number of traffic classes in the net device
1622  *	@tc_to_txq:	XXX: need comments on this one
1623  *	@prio_tc_map:	XXX: need comments on this one
1624  *
1625  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1626  *
1627  *	@priomap:	XXX: need comments on this one
1628  *	@phydev:	Physical device may attach itself
1629  *			for hardware timestamping
1630  *
1631  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1632  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1633  *
1634  *	@proto_down:	protocol port state information can be sent to the
1635  *			switch driver and used to set the phys state of the
1636  *			switch port.
1637  *
1638  *	FIXME: cleanup struct net_device such that network protocol info
1639  *	moves out.
1640  */
1641 
1642 struct net_device {
1643 	char			name[IFNAMSIZ];
1644 	struct hlist_node	name_hlist;
1645 	char 			*ifalias;
1646 	/*
1647 	 *	I/O specific fields
1648 	 *	FIXME: Merge these and struct ifmap into one
1649 	 */
1650 	unsigned long		mem_end;
1651 	unsigned long		mem_start;
1652 	unsigned long		base_addr;
1653 	int			irq;
1654 
1655 	atomic_t		carrier_changes;
1656 
1657 	/*
1658 	 *	Some hardware also needs these fields (state,dev_list,
1659 	 *	napi_list,unreg_list,close_list) but they are not
1660 	 *	part of the usual set specified in Space.c.
1661 	 */
1662 
1663 	unsigned long		state;
1664 
1665 	struct list_head	dev_list;
1666 	struct list_head	napi_list;
1667 	struct list_head	unreg_list;
1668 	struct list_head	close_list;
1669 	struct list_head	ptype_all;
1670 	struct list_head	ptype_specific;
1671 
1672 	struct {
1673 		struct list_head upper;
1674 		struct list_head lower;
1675 	} adj_list;
1676 
1677 	netdev_features_t	features;
1678 	netdev_features_t	hw_features;
1679 	netdev_features_t	wanted_features;
1680 	netdev_features_t	vlan_features;
1681 	netdev_features_t	hw_enc_features;
1682 	netdev_features_t	mpls_features;
1683 	netdev_features_t	gso_partial_features;
1684 
1685 	int			ifindex;
1686 	int			group;
1687 
1688 	struct net_device_stats	stats;
1689 
1690 	atomic_long_t		rx_dropped;
1691 	atomic_long_t		tx_dropped;
1692 	atomic_long_t		rx_nohandler;
1693 
1694 #ifdef CONFIG_WIRELESS_EXT
1695 	const struct iw_handler_def *wireless_handlers;
1696 	struct iw_public_data	*wireless_data;
1697 #endif
1698 	const struct net_device_ops *netdev_ops;
1699 	const struct ethtool_ops *ethtool_ops;
1700 #ifdef CONFIG_NET_SWITCHDEV
1701 	const struct switchdev_ops *switchdev_ops;
1702 #endif
1703 #ifdef CONFIG_NET_L3_MASTER_DEV
1704 	const struct l3mdev_ops	*l3mdev_ops;
1705 #endif
1706 #if IS_ENABLED(CONFIG_IPV6)
1707 	const struct ndisc_ops *ndisc_ops;
1708 #endif
1709 
1710 #ifdef CONFIG_XFRM
1711 	const struct xfrmdev_ops *xfrmdev_ops;
1712 #endif
1713 
1714 	const struct header_ops *header_ops;
1715 
1716 	unsigned int		flags;
1717 	unsigned int		priv_flags;
1718 
1719 	unsigned short		gflags;
1720 	unsigned short		padded;
1721 
1722 	unsigned char		operstate;
1723 	unsigned char		link_mode;
1724 
1725 	unsigned char		if_port;
1726 	unsigned char		dma;
1727 
1728 	unsigned int		mtu;
1729 	unsigned int		min_mtu;
1730 	unsigned int		max_mtu;
1731 	unsigned short		type;
1732 	unsigned short		hard_header_len;
1733 	unsigned char		min_header_len;
1734 
1735 	unsigned short		needed_headroom;
1736 	unsigned short		needed_tailroom;
1737 
1738 	/* Interface address info. */
1739 	unsigned char		perm_addr[MAX_ADDR_LEN];
1740 	unsigned char		addr_assign_type;
1741 	unsigned char		addr_len;
1742 	unsigned short		neigh_priv_len;
1743 	unsigned short          dev_id;
1744 	unsigned short          dev_port;
1745 	spinlock_t		addr_list_lock;
1746 	unsigned char		name_assign_type;
1747 	bool			uc_promisc;
1748 	struct netdev_hw_addr_list	uc;
1749 	struct netdev_hw_addr_list	mc;
1750 	struct netdev_hw_addr_list	dev_addrs;
1751 
1752 #ifdef CONFIG_SYSFS
1753 	struct kset		*queues_kset;
1754 #endif
1755 	unsigned int		promiscuity;
1756 	unsigned int		allmulti;
1757 
1758 
1759 	/* Protocol-specific pointers */
1760 
1761 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1762 	struct vlan_info __rcu	*vlan_info;
1763 #endif
1764 #if IS_ENABLED(CONFIG_NET_DSA)
1765 	struct dsa_switch_tree	*dsa_ptr;
1766 #endif
1767 #if IS_ENABLED(CONFIG_TIPC)
1768 	struct tipc_bearer __rcu *tipc_ptr;
1769 #endif
1770 	void 			*atalk_ptr;
1771 	struct in_device __rcu	*ip_ptr;
1772 	struct dn_dev __rcu     *dn_ptr;
1773 	struct inet6_dev __rcu	*ip6_ptr;
1774 	void			*ax25_ptr;
1775 	struct wireless_dev	*ieee80211_ptr;
1776 	struct wpan_dev		*ieee802154_ptr;
1777 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1778 	struct mpls_dev __rcu	*mpls_ptr;
1779 #endif
1780 
1781 /*
1782  * Cache lines mostly used on receive path (including eth_type_trans())
1783  */
1784 	/* Interface address info used in eth_type_trans() */
1785 	unsigned char		*dev_addr;
1786 
1787 #ifdef CONFIG_SYSFS
1788 	struct netdev_rx_queue	*_rx;
1789 
1790 	unsigned int		num_rx_queues;
1791 	unsigned int		real_num_rx_queues;
1792 #endif
1793 
1794 	struct bpf_prog __rcu	*xdp_prog;
1795 	unsigned long		gro_flush_timeout;
1796 	rx_handler_func_t __rcu	*rx_handler;
1797 	void __rcu		*rx_handler_data;
1798 
1799 #ifdef CONFIG_NET_CLS_ACT
1800 	struct tcf_proto __rcu  *ingress_cl_list;
1801 #endif
1802 	struct netdev_queue __rcu *ingress_queue;
1803 #ifdef CONFIG_NETFILTER_INGRESS
1804 	struct nf_hook_entry __rcu *nf_hooks_ingress;
1805 #endif
1806 
1807 	unsigned char		broadcast[MAX_ADDR_LEN];
1808 #ifdef CONFIG_RFS_ACCEL
1809 	struct cpu_rmap		*rx_cpu_rmap;
1810 #endif
1811 	struct hlist_node	index_hlist;
1812 
1813 /*
1814  * Cache lines mostly used on transmit path
1815  */
1816 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1817 	unsigned int		num_tx_queues;
1818 	unsigned int		real_num_tx_queues;
1819 	struct Qdisc		*qdisc;
1820 #ifdef CONFIG_NET_SCHED
1821 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1822 #endif
1823 	unsigned int		tx_queue_len;
1824 	spinlock_t		tx_global_lock;
1825 	int			watchdog_timeo;
1826 
1827 #ifdef CONFIG_XPS
1828 	struct xps_dev_maps __rcu *xps_maps;
1829 #endif
1830 #ifdef CONFIG_NET_CLS_ACT
1831 	struct tcf_proto __rcu  *egress_cl_list;
1832 #endif
1833 
1834 	/* These may be needed for future network-power-down code. */
1835 	struct timer_list	watchdog_timer;
1836 
1837 	int __percpu		*pcpu_refcnt;
1838 	struct list_head	todo_list;
1839 
1840 	struct list_head	link_watch_list;
1841 
1842 	enum { NETREG_UNINITIALIZED=0,
1843 	       NETREG_REGISTERED,	/* completed register_netdevice */
1844 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1845 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1846 	       NETREG_RELEASED,		/* called free_netdev */
1847 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1848 	} reg_state:8;
1849 
1850 	bool dismantle;
1851 
1852 	enum {
1853 		RTNL_LINK_INITIALIZED,
1854 		RTNL_LINK_INITIALIZING,
1855 	} rtnl_link_state:16;
1856 
1857 	bool needs_free_netdev;
1858 	void (*priv_destructor)(struct net_device *dev);
1859 
1860 #ifdef CONFIG_NETPOLL
1861 	struct netpoll_info __rcu	*npinfo;
1862 #endif
1863 
1864 	possible_net_t			nd_net;
1865 
1866 	/* mid-layer private */
1867 	union {
1868 		void					*ml_priv;
1869 		struct pcpu_lstats __percpu		*lstats;
1870 		struct pcpu_sw_netstats __percpu	*tstats;
1871 		struct pcpu_dstats __percpu		*dstats;
1872 		struct pcpu_vstats __percpu		*vstats;
1873 	};
1874 
1875 #if IS_ENABLED(CONFIG_GARP)
1876 	struct garp_port __rcu	*garp_port;
1877 #endif
1878 #if IS_ENABLED(CONFIG_MRP)
1879 	struct mrp_port __rcu	*mrp_port;
1880 #endif
1881 
1882 	struct device		dev;
1883 	const struct attribute_group *sysfs_groups[4];
1884 	const struct attribute_group *sysfs_rx_queue_group;
1885 
1886 	const struct rtnl_link_ops *rtnl_link_ops;
1887 
1888 	/* for setting kernel sock attribute on TCP connection setup */
1889 #define GSO_MAX_SIZE		65536
1890 	unsigned int		gso_max_size;
1891 #define GSO_MAX_SEGS		65535
1892 	u16			gso_max_segs;
1893 
1894 #ifdef CONFIG_DCB
1895 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1896 #endif
1897 	u8			num_tc;
1898 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1899 	u8			prio_tc_map[TC_BITMASK + 1];
1900 
1901 #if IS_ENABLED(CONFIG_FCOE)
1902 	unsigned int		fcoe_ddp_xid;
1903 #endif
1904 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1905 	struct netprio_map __rcu *priomap;
1906 #endif
1907 	struct phy_device	*phydev;
1908 	struct lock_class_key	*qdisc_tx_busylock;
1909 	struct lock_class_key	*qdisc_running_key;
1910 	bool			proto_down;
1911 };
1912 #define to_net_dev(d) container_of(d, struct net_device, dev)
1913 
1914 static inline bool netif_elide_gro(const struct net_device *dev)
1915 {
1916 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1917 		return true;
1918 	return false;
1919 }
1920 
1921 #define	NETDEV_ALIGN		32
1922 
1923 static inline
1924 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1925 {
1926 	return dev->prio_tc_map[prio & TC_BITMASK];
1927 }
1928 
1929 static inline
1930 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1931 {
1932 	if (tc >= dev->num_tc)
1933 		return -EINVAL;
1934 
1935 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1936 	return 0;
1937 }
1938 
1939 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1940 void netdev_reset_tc(struct net_device *dev);
1941 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1942 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1943 
1944 static inline
1945 int netdev_get_num_tc(struct net_device *dev)
1946 {
1947 	return dev->num_tc;
1948 }
1949 
1950 static inline
1951 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1952 					 unsigned int index)
1953 {
1954 	return &dev->_tx[index];
1955 }
1956 
1957 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1958 						    const struct sk_buff *skb)
1959 {
1960 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1961 }
1962 
1963 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1964 					    void (*f)(struct net_device *,
1965 						      struct netdev_queue *,
1966 						      void *),
1967 					    void *arg)
1968 {
1969 	unsigned int i;
1970 
1971 	for (i = 0; i < dev->num_tx_queues; i++)
1972 		f(dev, &dev->_tx[i], arg);
1973 }
1974 
1975 #define netdev_lockdep_set_classes(dev)				\
1976 {								\
1977 	static struct lock_class_key qdisc_tx_busylock_key;	\
1978 	static struct lock_class_key qdisc_running_key;		\
1979 	static struct lock_class_key qdisc_xmit_lock_key;	\
1980 	static struct lock_class_key dev_addr_list_lock_key;	\
1981 	unsigned int i;						\
1982 								\
1983 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1984 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1985 	lockdep_set_class(&(dev)->addr_list_lock,		\
1986 			  &dev_addr_list_lock_key); 		\
1987 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1988 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1989 				  &qdisc_xmit_lock_key);	\
1990 }
1991 
1992 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1993 				    struct sk_buff *skb,
1994 				    void *accel_priv);
1995 
1996 /* returns the headroom that the master device needs to take in account
1997  * when forwarding to this dev
1998  */
1999 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2000 {
2001 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2002 }
2003 
2004 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2005 {
2006 	if (dev->netdev_ops->ndo_set_rx_headroom)
2007 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2008 }
2009 
2010 /* set the device rx headroom to the dev's default */
2011 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2012 {
2013 	netdev_set_rx_headroom(dev, -1);
2014 }
2015 
2016 /*
2017  * Net namespace inlines
2018  */
2019 static inline
2020 struct net *dev_net(const struct net_device *dev)
2021 {
2022 	return read_pnet(&dev->nd_net);
2023 }
2024 
2025 static inline
2026 void dev_net_set(struct net_device *dev, struct net *net)
2027 {
2028 	write_pnet(&dev->nd_net, net);
2029 }
2030 
2031 /**
2032  *	netdev_priv - access network device private data
2033  *	@dev: network device
2034  *
2035  * Get network device private data
2036  */
2037 static inline void *netdev_priv(const struct net_device *dev)
2038 {
2039 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2040 }
2041 
2042 /* Set the sysfs physical device reference for the network logical device
2043  * if set prior to registration will cause a symlink during initialization.
2044  */
2045 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2046 
2047 /* Set the sysfs device type for the network logical device to allow
2048  * fine-grained identification of different network device types. For
2049  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2050  */
2051 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2052 
2053 /* Default NAPI poll() weight
2054  * Device drivers are strongly advised to not use bigger value
2055  */
2056 #define NAPI_POLL_WEIGHT 64
2057 
2058 /**
2059  *	netif_napi_add - initialize a NAPI context
2060  *	@dev:  network device
2061  *	@napi: NAPI context
2062  *	@poll: polling function
2063  *	@weight: default weight
2064  *
2065  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2066  * *any* of the other NAPI-related functions.
2067  */
2068 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2069 		    int (*poll)(struct napi_struct *, int), int weight);
2070 
2071 /**
2072  *	netif_tx_napi_add - initialize a NAPI context
2073  *	@dev:  network device
2074  *	@napi: NAPI context
2075  *	@poll: polling function
2076  *	@weight: default weight
2077  *
2078  * This variant of netif_napi_add() should be used from drivers using NAPI
2079  * to exclusively poll a TX queue.
2080  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2081  */
2082 static inline void netif_tx_napi_add(struct net_device *dev,
2083 				     struct napi_struct *napi,
2084 				     int (*poll)(struct napi_struct *, int),
2085 				     int weight)
2086 {
2087 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2088 	netif_napi_add(dev, napi, poll, weight);
2089 }
2090 
2091 /**
2092  *  netif_napi_del - remove a NAPI context
2093  *  @napi: NAPI context
2094  *
2095  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2096  */
2097 void netif_napi_del(struct napi_struct *napi);
2098 
2099 struct napi_gro_cb {
2100 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2101 	void	*frag0;
2102 
2103 	/* Length of frag0. */
2104 	unsigned int frag0_len;
2105 
2106 	/* This indicates where we are processing relative to skb->data. */
2107 	int	data_offset;
2108 
2109 	/* This is non-zero if the packet cannot be merged with the new skb. */
2110 	u16	flush;
2111 
2112 	/* Save the IP ID here and check when we get to the transport layer */
2113 	u16	flush_id;
2114 
2115 	/* Number of segments aggregated. */
2116 	u16	count;
2117 
2118 	/* Start offset for remote checksum offload */
2119 	u16	gro_remcsum_start;
2120 
2121 	/* jiffies when first packet was created/queued */
2122 	unsigned long age;
2123 
2124 	/* Used in ipv6_gro_receive() and foo-over-udp */
2125 	u16	proto;
2126 
2127 	/* This is non-zero if the packet may be of the same flow. */
2128 	u8	same_flow:1;
2129 
2130 	/* Used in tunnel GRO receive */
2131 	u8	encap_mark:1;
2132 
2133 	/* GRO checksum is valid */
2134 	u8	csum_valid:1;
2135 
2136 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2137 	u8	csum_cnt:3;
2138 
2139 	/* Free the skb? */
2140 	u8	free:2;
2141 #define NAPI_GRO_FREE		  1
2142 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2143 
2144 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2145 	u8	is_ipv6:1;
2146 
2147 	/* Used in GRE, set in fou/gue_gro_receive */
2148 	u8	is_fou:1;
2149 
2150 	/* Used to determine if flush_id can be ignored */
2151 	u8	is_atomic:1;
2152 
2153 	/* Number of gro_receive callbacks this packet already went through */
2154 	u8 recursion_counter:4;
2155 
2156 	/* 1 bit hole */
2157 
2158 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2159 	__wsum	csum;
2160 
2161 	/* used in skb_gro_receive() slow path */
2162 	struct sk_buff *last;
2163 };
2164 
2165 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2166 
2167 #define GRO_RECURSION_LIMIT 15
2168 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2169 {
2170 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2171 }
2172 
2173 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2174 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2175 						struct sk_buff **head,
2176 						struct sk_buff *skb)
2177 {
2178 	if (unlikely(gro_recursion_inc_test(skb))) {
2179 		NAPI_GRO_CB(skb)->flush |= 1;
2180 		return NULL;
2181 	}
2182 
2183 	return cb(head, skb);
2184 }
2185 
2186 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2187 					     struct sk_buff *);
2188 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2189 						   struct sock *sk,
2190 						   struct sk_buff **head,
2191 						   struct sk_buff *skb)
2192 {
2193 	if (unlikely(gro_recursion_inc_test(skb))) {
2194 		NAPI_GRO_CB(skb)->flush |= 1;
2195 		return NULL;
2196 	}
2197 
2198 	return cb(sk, head, skb);
2199 }
2200 
2201 struct packet_type {
2202 	__be16			type;	/* This is really htons(ether_type). */
2203 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2204 	int			(*func) (struct sk_buff *,
2205 					 struct net_device *,
2206 					 struct packet_type *,
2207 					 struct net_device *);
2208 	bool			(*id_match)(struct packet_type *ptype,
2209 					    struct sock *sk);
2210 	void			*af_packet_priv;
2211 	struct list_head	list;
2212 };
2213 
2214 struct offload_callbacks {
2215 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2216 						netdev_features_t features);
2217 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2218 						 struct sk_buff *skb);
2219 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2220 };
2221 
2222 struct packet_offload {
2223 	__be16			 type;	/* This is really htons(ether_type). */
2224 	u16			 priority;
2225 	struct offload_callbacks callbacks;
2226 	struct list_head	 list;
2227 };
2228 
2229 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2230 struct pcpu_sw_netstats {
2231 	u64     rx_packets;
2232 	u64     rx_bytes;
2233 	u64     tx_packets;
2234 	u64     tx_bytes;
2235 	struct u64_stats_sync   syncp;
2236 };
2237 
2238 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2239 ({									\
2240 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2241 	if (pcpu_stats)	{						\
2242 		int __cpu;						\
2243 		for_each_possible_cpu(__cpu) {				\
2244 			typeof(type) *stat;				\
2245 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2246 			u64_stats_init(&stat->syncp);			\
2247 		}							\
2248 	}								\
2249 	pcpu_stats;							\
2250 })
2251 
2252 #define netdev_alloc_pcpu_stats(type)					\
2253 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2254 
2255 enum netdev_lag_tx_type {
2256 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2257 	NETDEV_LAG_TX_TYPE_RANDOM,
2258 	NETDEV_LAG_TX_TYPE_BROADCAST,
2259 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2260 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2261 	NETDEV_LAG_TX_TYPE_HASH,
2262 };
2263 
2264 struct netdev_lag_upper_info {
2265 	enum netdev_lag_tx_type tx_type;
2266 };
2267 
2268 struct netdev_lag_lower_state_info {
2269 	u8 link_up : 1,
2270 	   tx_enabled : 1;
2271 };
2272 
2273 #include <linux/notifier.h>
2274 
2275 /* netdevice notifier chain. Please remember to update the rtnetlink
2276  * notification exclusion list in rtnetlink_event() when adding new
2277  * types.
2278  */
2279 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2280 #define NETDEV_DOWN	0x0002
2281 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2282 				   detected a hardware crash and restarted
2283 				   - we can use this eg to kick tcp sessions
2284 				   once done */
2285 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2286 #define NETDEV_REGISTER 0x0005
2287 #define NETDEV_UNREGISTER	0x0006
2288 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2289 #define NETDEV_CHANGEADDR	0x0008
2290 #define NETDEV_GOING_DOWN	0x0009
2291 #define NETDEV_CHANGENAME	0x000A
2292 #define NETDEV_FEAT_CHANGE	0x000B
2293 #define NETDEV_BONDING_FAILOVER 0x000C
2294 #define NETDEV_PRE_UP		0x000D
2295 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2296 #define NETDEV_POST_TYPE_CHANGE	0x000F
2297 #define NETDEV_POST_INIT	0x0010
2298 #define NETDEV_UNREGISTER_FINAL 0x0011
2299 #define NETDEV_RELEASE		0x0012
2300 #define NETDEV_NOTIFY_PEERS	0x0013
2301 #define NETDEV_JOIN		0x0014
2302 #define NETDEV_CHANGEUPPER	0x0015
2303 #define NETDEV_RESEND_IGMP	0x0016
2304 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2305 #define NETDEV_CHANGEINFODATA	0x0018
2306 #define NETDEV_BONDING_INFO	0x0019
2307 #define NETDEV_PRECHANGEUPPER	0x001A
2308 #define NETDEV_CHANGELOWERSTATE	0x001B
2309 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2310 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2311 
2312 int register_netdevice_notifier(struct notifier_block *nb);
2313 int unregister_netdevice_notifier(struct notifier_block *nb);
2314 
2315 struct netdev_notifier_info {
2316 	struct net_device *dev;
2317 };
2318 
2319 struct netdev_notifier_change_info {
2320 	struct netdev_notifier_info info; /* must be first */
2321 	unsigned int flags_changed;
2322 };
2323 
2324 struct netdev_notifier_changeupper_info {
2325 	struct netdev_notifier_info info; /* must be first */
2326 	struct net_device *upper_dev; /* new upper dev */
2327 	bool master; /* is upper dev master */
2328 	bool linking; /* is the notification for link or unlink */
2329 	void *upper_info; /* upper dev info */
2330 };
2331 
2332 struct netdev_notifier_changelowerstate_info {
2333 	struct netdev_notifier_info info; /* must be first */
2334 	void *lower_state_info; /* is lower dev state */
2335 };
2336 
2337 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2338 					     struct net_device *dev)
2339 {
2340 	info->dev = dev;
2341 }
2342 
2343 static inline struct net_device *
2344 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2345 {
2346 	return info->dev;
2347 }
2348 
2349 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2350 
2351 
2352 extern rwlock_t				dev_base_lock;		/* Device list lock */
2353 
2354 #define for_each_netdev(net, d)		\
2355 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2356 #define for_each_netdev_reverse(net, d)	\
2357 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2358 #define for_each_netdev_rcu(net, d)		\
2359 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2360 #define for_each_netdev_safe(net, d, n)	\
2361 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2362 #define for_each_netdev_continue(net, d)		\
2363 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2364 #define for_each_netdev_continue_rcu(net, d)		\
2365 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2366 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2367 		for_each_netdev_rcu(&init_net, slave)	\
2368 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2369 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2370 
2371 static inline struct net_device *next_net_device(struct net_device *dev)
2372 {
2373 	struct list_head *lh;
2374 	struct net *net;
2375 
2376 	net = dev_net(dev);
2377 	lh = dev->dev_list.next;
2378 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2379 }
2380 
2381 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2382 {
2383 	struct list_head *lh;
2384 	struct net *net;
2385 
2386 	net = dev_net(dev);
2387 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2388 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2389 }
2390 
2391 static inline struct net_device *first_net_device(struct net *net)
2392 {
2393 	return list_empty(&net->dev_base_head) ? NULL :
2394 		net_device_entry(net->dev_base_head.next);
2395 }
2396 
2397 static inline struct net_device *first_net_device_rcu(struct net *net)
2398 {
2399 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2400 
2401 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2402 }
2403 
2404 int netdev_boot_setup_check(struct net_device *dev);
2405 unsigned long netdev_boot_base(const char *prefix, int unit);
2406 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2407 				       const char *hwaddr);
2408 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2409 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2410 void dev_add_pack(struct packet_type *pt);
2411 void dev_remove_pack(struct packet_type *pt);
2412 void __dev_remove_pack(struct packet_type *pt);
2413 void dev_add_offload(struct packet_offload *po);
2414 void dev_remove_offload(struct packet_offload *po);
2415 
2416 int dev_get_iflink(const struct net_device *dev);
2417 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2418 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2419 				      unsigned short mask);
2420 struct net_device *dev_get_by_name(struct net *net, const char *name);
2421 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2422 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2423 int dev_alloc_name(struct net_device *dev, const char *name);
2424 int dev_open(struct net_device *dev);
2425 int dev_close(struct net_device *dev);
2426 int dev_close_many(struct list_head *head, bool unlink);
2427 void dev_disable_lro(struct net_device *dev);
2428 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2429 int dev_queue_xmit(struct sk_buff *skb);
2430 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2431 int register_netdevice(struct net_device *dev);
2432 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2433 void unregister_netdevice_many(struct list_head *head);
2434 static inline void unregister_netdevice(struct net_device *dev)
2435 {
2436 	unregister_netdevice_queue(dev, NULL);
2437 }
2438 
2439 int netdev_refcnt_read(const struct net_device *dev);
2440 void free_netdev(struct net_device *dev);
2441 void netdev_freemem(struct net_device *dev);
2442 void synchronize_net(void);
2443 int init_dummy_netdev(struct net_device *dev);
2444 
2445 DECLARE_PER_CPU(int, xmit_recursion);
2446 #define XMIT_RECURSION_LIMIT	10
2447 
2448 static inline int dev_recursion_level(void)
2449 {
2450 	return this_cpu_read(xmit_recursion);
2451 }
2452 
2453 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2454 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2455 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2456 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2457 int netdev_get_name(struct net *net, char *name, int ifindex);
2458 int dev_restart(struct net_device *dev);
2459 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2460 
2461 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2462 {
2463 	return NAPI_GRO_CB(skb)->data_offset;
2464 }
2465 
2466 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2467 {
2468 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2469 }
2470 
2471 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2472 {
2473 	NAPI_GRO_CB(skb)->data_offset += len;
2474 }
2475 
2476 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2477 					unsigned int offset)
2478 {
2479 	return NAPI_GRO_CB(skb)->frag0 + offset;
2480 }
2481 
2482 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2483 {
2484 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2485 }
2486 
2487 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2488 {
2489 	NAPI_GRO_CB(skb)->frag0 = NULL;
2490 	NAPI_GRO_CB(skb)->frag0_len = 0;
2491 }
2492 
2493 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2494 					unsigned int offset)
2495 {
2496 	if (!pskb_may_pull(skb, hlen))
2497 		return NULL;
2498 
2499 	skb_gro_frag0_invalidate(skb);
2500 	return skb->data + offset;
2501 }
2502 
2503 static inline void *skb_gro_network_header(struct sk_buff *skb)
2504 {
2505 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2506 	       skb_network_offset(skb);
2507 }
2508 
2509 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2510 					const void *start, unsigned int len)
2511 {
2512 	if (NAPI_GRO_CB(skb)->csum_valid)
2513 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2514 						  csum_partial(start, len, 0));
2515 }
2516 
2517 /* GRO checksum functions. These are logical equivalents of the normal
2518  * checksum functions (in skbuff.h) except that they operate on the GRO
2519  * offsets and fields in sk_buff.
2520  */
2521 
2522 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2523 
2524 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2525 {
2526 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2527 }
2528 
2529 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2530 						      bool zero_okay,
2531 						      __sum16 check)
2532 {
2533 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2534 		skb_checksum_start_offset(skb) <
2535 		 skb_gro_offset(skb)) &&
2536 		!skb_at_gro_remcsum_start(skb) &&
2537 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2538 		(!zero_okay || check));
2539 }
2540 
2541 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2542 							   __wsum psum)
2543 {
2544 	if (NAPI_GRO_CB(skb)->csum_valid &&
2545 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2546 		return 0;
2547 
2548 	NAPI_GRO_CB(skb)->csum = psum;
2549 
2550 	return __skb_gro_checksum_complete(skb);
2551 }
2552 
2553 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2554 {
2555 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2556 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2557 		NAPI_GRO_CB(skb)->csum_cnt--;
2558 	} else {
2559 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2560 		 * verified a new top level checksum or an encapsulated one
2561 		 * during GRO. This saves work if we fallback to normal path.
2562 		 */
2563 		__skb_incr_checksum_unnecessary(skb);
2564 	}
2565 }
2566 
2567 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2568 				    compute_pseudo)			\
2569 ({									\
2570 	__sum16 __ret = 0;						\
2571 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2572 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2573 				compute_pseudo(skb, proto));		\
2574 	if (!__ret)							\
2575 		skb_gro_incr_csum_unnecessary(skb);			\
2576 	__ret;								\
2577 })
2578 
2579 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2580 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2581 
2582 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2583 					     compute_pseudo)		\
2584 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2585 
2586 #define skb_gro_checksum_simple_validate(skb)				\
2587 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2588 
2589 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2590 {
2591 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2592 		!NAPI_GRO_CB(skb)->csum_valid);
2593 }
2594 
2595 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2596 					      __sum16 check, __wsum pseudo)
2597 {
2598 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2599 	NAPI_GRO_CB(skb)->csum_valid = 1;
2600 }
2601 
2602 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2603 do {									\
2604 	if (__skb_gro_checksum_convert_check(skb))			\
2605 		__skb_gro_checksum_convert(skb, check,			\
2606 					   compute_pseudo(skb, proto));	\
2607 } while (0)
2608 
2609 struct gro_remcsum {
2610 	int offset;
2611 	__wsum delta;
2612 };
2613 
2614 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2615 {
2616 	grc->offset = 0;
2617 	grc->delta = 0;
2618 }
2619 
2620 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2621 					    unsigned int off, size_t hdrlen,
2622 					    int start, int offset,
2623 					    struct gro_remcsum *grc,
2624 					    bool nopartial)
2625 {
2626 	__wsum delta;
2627 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2628 
2629 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2630 
2631 	if (!nopartial) {
2632 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2633 		return ptr;
2634 	}
2635 
2636 	ptr = skb_gro_header_fast(skb, off);
2637 	if (skb_gro_header_hard(skb, off + plen)) {
2638 		ptr = skb_gro_header_slow(skb, off + plen, off);
2639 		if (!ptr)
2640 			return NULL;
2641 	}
2642 
2643 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2644 			       start, offset);
2645 
2646 	/* Adjust skb->csum since we changed the packet */
2647 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2648 
2649 	grc->offset = off + hdrlen + offset;
2650 	grc->delta = delta;
2651 
2652 	return ptr;
2653 }
2654 
2655 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2656 					   struct gro_remcsum *grc)
2657 {
2658 	void *ptr;
2659 	size_t plen = grc->offset + sizeof(u16);
2660 
2661 	if (!grc->delta)
2662 		return;
2663 
2664 	ptr = skb_gro_header_fast(skb, grc->offset);
2665 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2666 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2667 		if (!ptr)
2668 			return;
2669 	}
2670 
2671 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2672 }
2673 
2674 #ifdef CONFIG_XFRM_OFFLOAD
2675 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2676 {
2677 	if (PTR_ERR(pp) != -EINPROGRESS)
2678 		NAPI_GRO_CB(skb)->flush |= flush;
2679 }
2680 #else
2681 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2682 {
2683 	NAPI_GRO_CB(skb)->flush |= flush;
2684 }
2685 #endif
2686 
2687 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2688 				  unsigned short type,
2689 				  const void *daddr, const void *saddr,
2690 				  unsigned int len)
2691 {
2692 	if (!dev->header_ops || !dev->header_ops->create)
2693 		return 0;
2694 
2695 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2696 }
2697 
2698 static inline int dev_parse_header(const struct sk_buff *skb,
2699 				   unsigned char *haddr)
2700 {
2701 	const struct net_device *dev = skb->dev;
2702 
2703 	if (!dev->header_ops || !dev->header_ops->parse)
2704 		return 0;
2705 	return dev->header_ops->parse(skb, haddr);
2706 }
2707 
2708 /* ll_header must have at least hard_header_len allocated */
2709 static inline bool dev_validate_header(const struct net_device *dev,
2710 				       char *ll_header, int len)
2711 {
2712 	if (likely(len >= dev->hard_header_len))
2713 		return true;
2714 	if (len < dev->min_header_len)
2715 		return false;
2716 
2717 	if (capable(CAP_SYS_RAWIO)) {
2718 		memset(ll_header + len, 0, dev->hard_header_len - len);
2719 		return true;
2720 	}
2721 
2722 	if (dev->header_ops && dev->header_ops->validate)
2723 		return dev->header_ops->validate(ll_header, len);
2724 
2725 	return false;
2726 }
2727 
2728 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2729 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2730 static inline int unregister_gifconf(unsigned int family)
2731 {
2732 	return register_gifconf(family, NULL);
2733 }
2734 
2735 #ifdef CONFIG_NET_FLOW_LIMIT
2736 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2737 struct sd_flow_limit {
2738 	u64			count;
2739 	unsigned int		num_buckets;
2740 	unsigned int		history_head;
2741 	u16			history[FLOW_LIMIT_HISTORY];
2742 	u8			buckets[];
2743 };
2744 
2745 extern int netdev_flow_limit_table_len;
2746 #endif /* CONFIG_NET_FLOW_LIMIT */
2747 
2748 /*
2749  * Incoming packets are placed on per-CPU queues
2750  */
2751 struct softnet_data {
2752 	struct list_head	poll_list;
2753 	struct sk_buff_head	process_queue;
2754 
2755 	/* stats */
2756 	unsigned int		processed;
2757 	unsigned int		time_squeeze;
2758 	unsigned int		received_rps;
2759 #ifdef CONFIG_RPS
2760 	struct softnet_data	*rps_ipi_list;
2761 #endif
2762 #ifdef CONFIG_NET_FLOW_LIMIT
2763 	struct sd_flow_limit __rcu *flow_limit;
2764 #endif
2765 	struct Qdisc		*output_queue;
2766 	struct Qdisc		**output_queue_tailp;
2767 	struct sk_buff		*completion_queue;
2768 
2769 #ifdef CONFIG_RPS
2770 	/* input_queue_head should be written by cpu owning this struct,
2771 	 * and only read by other cpus. Worth using a cache line.
2772 	 */
2773 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2774 
2775 	/* Elements below can be accessed between CPUs for RPS/RFS */
2776 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2777 	struct softnet_data	*rps_ipi_next;
2778 	unsigned int		cpu;
2779 	unsigned int		input_queue_tail;
2780 #endif
2781 	unsigned int		dropped;
2782 	struct sk_buff_head	input_pkt_queue;
2783 	struct napi_struct	backlog;
2784 
2785 };
2786 
2787 static inline void input_queue_head_incr(struct softnet_data *sd)
2788 {
2789 #ifdef CONFIG_RPS
2790 	sd->input_queue_head++;
2791 #endif
2792 }
2793 
2794 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2795 					      unsigned int *qtail)
2796 {
2797 #ifdef CONFIG_RPS
2798 	*qtail = ++sd->input_queue_tail;
2799 #endif
2800 }
2801 
2802 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2803 
2804 void __netif_schedule(struct Qdisc *q);
2805 void netif_schedule_queue(struct netdev_queue *txq);
2806 
2807 static inline void netif_tx_schedule_all(struct net_device *dev)
2808 {
2809 	unsigned int i;
2810 
2811 	for (i = 0; i < dev->num_tx_queues; i++)
2812 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2813 }
2814 
2815 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2816 {
2817 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2818 }
2819 
2820 /**
2821  *	netif_start_queue - allow transmit
2822  *	@dev: network device
2823  *
2824  *	Allow upper layers to call the device hard_start_xmit routine.
2825  */
2826 static inline void netif_start_queue(struct net_device *dev)
2827 {
2828 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2829 }
2830 
2831 static inline void netif_tx_start_all_queues(struct net_device *dev)
2832 {
2833 	unsigned int i;
2834 
2835 	for (i = 0; i < dev->num_tx_queues; i++) {
2836 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2837 		netif_tx_start_queue(txq);
2838 	}
2839 }
2840 
2841 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2842 
2843 /**
2844  *	netif_wake_queue - restart transmit
2845  *	@dev: network device
2846  *
2847  *	Allow upper layers to call the device hard_start_xmit routine.
2848  *	Used for flow control when transmit resources are available.
2849  */
2850 static inline void netif_wake_queue(struct net_device *dev)
2851 {
2852 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2853 }
2854 
2855 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2856 {
2857 	unsigned int i;
2858 
2859 	for (i = 0; i < dev->num_tx_queues; i++) {
2860 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2861 		netif_tx_wake_queue(txq);
2862 	}
2863 }
2864 
2865 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2866 {
2867 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2868 }
2869 
2870 /**
2871  *	netif_stop_queue - stop transmitted packets
2872  *	@dev: network device
2873  *
2874  *	Stop upper layers calling the device hard_start_xmit routine.
2875  *	Used for flow control when transmit resources are unavailable.
2876  */
2877 static inline void netif_stop_queue(struct net_device *dev)
2878 {
2879 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2880 }
2881 
2882 void netif_tx_stop_all_queues(struct net_device *dev);
2883 
2884 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2885 {
2886 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2887 }
2888 
2889 /**
2890  *	netif_queue_stopped - test if transmit queue is flowblocked
2891  *	@dev: network device
2892  *
2893  *	Test if transmit queue on device is currently unable to send.
2894  */
2895 static inline bool netif_queue_stopped(const struct net_device *dev)
2896 {
2897 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2898 }
2899 
2900 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2901 {
2902 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2903 }
2904 
2905 static inline bool
2906 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2907 {
2908 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2909 }
2910 
2911 static inline bool
2912 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2913 {
2914 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2915 }
2916 
2917 /**
2918  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2919  *	@dev_queue: pointer to transmit queue
2920  *
2921  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2922  * to give appropriate hint to the CPU.
2923  */
2924 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2925 {
2926 #ifdef CONFIG_BQL
2927 	prefetchw(&dev_queue->dql.num_queued);
2928 #endif
2929 }
2930 
2931 /**
2932  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2933  *	@dev_queue: pointer to transmit queue
2934  *
2935  * BQL enabled drivers might use this helper in their TX completion path,
2936  * to give appropriate hint to the CPU.
2937  */
2938 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2939 {
2940 #ifdef CONFIG_BQL
2941 	prefetchw(&dev_queue->dql.limit);
2942 #endif
2943 }
2944 
2945 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2946 					unsigned int bytes)
2947 {
2948 #ifdef CONFIG_BQL
2949 	dql_queued(&dev_queue->dql, bytes);
2950 
2951 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2952 		return;
2953 
2954 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2955 
2956 	/*
2957 	 * The XOFF flag must be set before checking the dql_avail below,
2958 	 * because in netdev_tx_completed_queue we update the dql_completed
2959 	 * before checking the XOFF flag.
2960 	 */
2961 	smp_mb();
2962 
2963 	/* check again in case another CPU has just made room avail */
2964 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2965 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2966 #endif
2967 }
2968 
2969 /**
2970  * 	netdev_sent_queue - report the number of bytes queued to hardware
2971  * 	@dev: network device
2972  * 	@bytes: number of bytes queued to the hardware device queue
2973  *
2974  * 	Report the number of bytes queued for sending/completion to the network
2975  * 	device hardware queue. @bytes should be a good approximation and should
2976  * 	exactly match netdev_completed_queue() @bytes
2977  */
2978 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2979 {
2980 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2981 }
2982 
2983 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2984 					     unsigned int pkts, unsigned int bytes)
2985 {
2986 #ifdef CONFIG_BQL
2987 	if (unlikely(!bytes))
2988 		return;
2989 
2990 	dql_completed(&dev_queue->dql, bytes);
2991 
2992 	/*
2993 	 * Without the memory barrier there is a small possiblity that
2994 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2995 	 * be stopped forever
2996 	 */
2997 	smp_mb();
2998 
2999 	if (dql_avail(&dev_queue->dql) < 0)
3000 		return;
3001 
3002 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3003 		netif_schedule_queue(dev_queue);
3004 #endif
3005 }
3006 
3007 /**
3008  * 	netdev_completed_queue - report bytes and packets completed by device
3009  * 	@dev: network device
3010  * 	@pkts: actual number of packets sent over the medium
3011  * 	@bytes: actual number of bytes sent over the medium
3012  *
3013  * 	Report the number of bytes and packets transmitted by the network device
3014  * 	hardware queue over the physical medium, @bytes must exactly match the
3015  * 	@bytes amount passed to netdev_sent_queue()
3016  */
3017 static inline void netdev_completed_queue(struct net_device *dev,
3018 					  unsigned int pkts, unsigned int bytes)
3019 {
3020 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3021 }
3022 
3023 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3024 {
3025 #ifdef CONFIG_BQL
3026 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3027 	dql_reset(&q->dql);
3028 #endif
3029 }
3030 
3031 /**
3032  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3033  * 	@dev_queue: network device
3034  *
3035  * 	Reset the bytes and packet count of a network device and clear the
3036  * 	software flow control OFF bit for this network device
3037  */
3038 static inline void netdev_reset_queue(struct net_device *dev_queue)
3039 {
3040 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3041 }
3042 
3043 /**
3044  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3045  * 	@dev: network device
3046  * 	@queue_index: given tx queue index
3047  *
3048  * 	Returns 0 if given tx queue index >= number of device tx queues,
3049  * 	otherwise returns the originally passed tx queue index.
3050  */
3051 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3052 {
3053 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3054 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3055 				     dev->name, queue_index,
3056 				     dev->real_num_tx_queues);
3057 		return 0;
3058 	}
3059 
3060 	return queue_index;
3061 }
3062 
3063 /**
3064  *	netif_running - test if up
3065  *	@dev: network device
3066  *
3067  *	Test if the device has been brought up.
3068  */
3069 static inline bool netif_running(const struct net_device *dev)
3070 {
3071 	return test_bit(__LINK_STATE_START, &dev->state);
3072 }
3073 
3074 /*
3075  * Routines to manage the subqueues on a device.  We only need start,
3076  * stop, and a check if it's stopped.  All other device management is
3077  * done at the overall netdevice level.
3078  * Also test the device if we're multiqueue.
3079  */
3080 
3081 /**
3082  *	netif_start_subqueue - allow sending packets on subqueue
3083  *	@dev: network device
3084  *	@queue_index: sub queue index
3085  *
3086  * Start individual transmit queue of a device with multiple transmit queues.
3087  */
3088 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3089 {
3090 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3091 
3092 	netif_tx_start_queue(txq);
3093 }
3094 
3095 /**
3096  *	netif_stop_subqueue - stop sending packets on subqueue
3097  *	@dev: network device
3098  *	@queue_index: sub queue index
3099  *
3100  * Stop individual transmit queue of a device with multiple transmit queues.
3101  */
3102 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3103 {
3104 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3105 	netif_tx_stop_queue(txq);
3106 }
3107 
3108 /**
3109  *	netif_subqueue_stopped - test status of subqueue
3110  *	@dev: network device
3111  *	@queue_index: sub queue index
3112  *
3113  * Check individual transmit queue of a device with multiple transmit queues.
3114  */
3115 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3116 					    u16 queue_index)
3117 {
3118 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3119 
3120 	return netif_tx_queue_stopped(txq);
3121 }
3122 
3123 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3124 					  struct sk_buff *skb)
3125 {
3126 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3127 }
3128 
3129 /**
3130  *	netif_wake_subqueue - allow sending packets on subqueue
3131  *	@dev: network device
3132  *	@queue_index: sub queue index
3133  *
3134  * Resume individual transmit queue of a device with multiple transmit queues.
3135  */
3136 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3137 {
3138 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3139 
3140 	netif_tx_wake_queue(txq);
3141 }
3142 
3143 #ifdef CONFIG_XPS
3144 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3145 			u16 index);
3146 #else
3147 static inline int netif_set_xps_queue(struct net_device *dev,
3148 				      const struct cpumask *mask,
3149 				      u16 index)
3150 {
3151 	return 0;
3152 }
3153 #endif
3154 
3155 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3156 		  unsigned int num_tx_queues);
3157 
3158 /*
3159  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3160  * as a distribution range limit for the returned value.
3161  */
3162 static inline u16 skb_tx_hash(const struct net_device *dev,
3163 			      struct sk_buff *skb)
3164 {
3165 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3166 }
3167 
3168 /**
3169  *	netif_is_multiqueue - test if device has multiple transmit queues
3170  *	@dev: network device
3171  *
3172  * Check if device has multiple transmit queues
3173  */
3174 static inline bool netif_is_multiqueue(const struct net_device *dev)
3175 {
3176 	return dev->num_tx_queues > 1;
3177 }
3178 
3179 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3180 
3181 #ifdef CONFIG_SYSFS
3182 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3183 #else
3184 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3185 						unsigned int rxq)
3186 {
3187 	return 0;
3188 }
3189 #endif
3190 
3191 #ifdef CONFIG_SYSFS
3192 static inline unsigned int get_netdev_rx_queue_index(
3193 		struct netdev_rx_queue *queue)
3194 {
3195 	struct net_device *dev = queue->dev;
3196 	int index = queue - dev->_rx;
3197 
3198 	BUG_ON(index >= dev->num_rx_queues);
3199 	return index;
3200 }
3201 #endif
3202 
3203 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3204 int netif_get_num_default_rss_queues(void);
3205 
3206 enum skb_free_reason {
3207 	SKB_REASON_CONSUMED,
3208 	SKB_REASON_DROPPED,
3209 };
3210 
3211 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3212 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3213 
3214 /*
3215  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3216  * interrupt context or with hardware interrupts being disabled.
3217  * (in_irq() || irqs_disabled())
3218  *
3219  * We provide four helpers that can be used in following contexts :
3220  *
3221  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3222  *  replacing kfree_skb(skb)
3223  *
3224  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3225  *  Typically used in place of consume_skb(skb) in TX completion path
3226  *
3227  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3228  *  replacing kfree_skb(skb)
3229  *
3230  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3231  *  and consumed a packet. Used in place of consume_skb(skb)
3232  */
3233 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3234 {
3235 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3236 }
3237 
3238 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3239 {
3240 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3241 }
3242 
3243 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3244 {
3245 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3246 }
3247 
3248 static inline void dev_consume_skb_any(struct sk_buff *skb)
3249 {
3250 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3251 }
3252 
3253 int netif_rx(struct sk_buff *skb);
3254 int netif_rx_ni(struct sk_buff *skb);
3255 int netif_receive_skb(struct sk_buff *skb);
3256 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3257 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3258 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3259 gro_result_t napi_gro_frags(struct napi_struct *napi);
3260 struct packet_offload *gro_find_receive_by_type(__be16 type);
3261 struct packet_offload *gro_find_complete_by_type(__be16 type);
3262 
3263 static inline void napi_free_frags(struct napi_struct *napi)
3264 {
3265 	kfree_skb(napi->skb);
3266 	napi->skb = NULL;
3267 }
3268 
3269 bool netdev_is_rx_handler_busy(struct net_device *dev);
3270 int netdev_rx_handler_register(struct net_device *dev,
3271 			       rx_handler_func_t *rx_handler,
3272 			       void *rx_handler_data);
3273 void netdev_rx_handler_unregister(struct net_device *dev);
3274 
3275 bool dev_valid_name(const char *name);
3276 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3277 int dev_ethtool(struct net *net, struct ifreq *);
3278 unsigned int dev_get_flags(const struct net_device *);
3279 int __dev_change_flags(struct net_device *, unsigned int flags);
3280 int dev_change_flags(struct net_device *, unsigned int);
3281 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3282 			unsigned int gchanges);
3283 int dev_change_name(struct net_device *, const char *);
3284 int dev_set_alias(struct net_device *, const char *, size_t);
3285 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3286 int dev_set_mtu(struct net_device *, int);
3287 void dev_set_group(struct net_device *, int);
3288 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3289 int dev_change_carrier(struct net_device *, bool new_carrier);
3290 int dev_get_phys_port_id(struct net_device *dev,
3291 			 struct netdev_phys_item_id *ppid);
3292 int dev_get_phys_port_name(struct net_device *dev,
3293 			   char *name, size_t len);
3294 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3295 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3296 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3297 				    struct netdev_queue *txq, int *ret);
3298 
3299 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3300 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3301 		      int fd, u32 flags);
3302 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3303 
3304 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3305 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3306 bool is_skb_forwardable(const struct net_device *dev,
3307 			const struct sk_buff *skb);
3308 
3309 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3310 					       struct sk_buff *skb)
3311 {
3312 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3313 	    unlikely(!is_skb_forwardable(dev, skb))) {
3314 		atomic_long_inc(&dev->rx_dropped);
3315 		kfree_skb(skb);
3316 		return NET_RX_DROP;
3317 	}
3318 
3319 	skb_scrub_packet(skb, true);
3320 	skb->priority = 0;
3321 	return 0;
3322 }
3323 
3324 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3325 
3326 extern int		netdev_budget;
3327 extern unsigned int	netdev_budget_usecs;
3328 
3329 /* Called by rtnetlink.c:rtnl_unlock() */
3330 void netdev_run_todo(void);
3331 
3332 /**
3333  *	dev_put - release reference to device
3334  *	@dev: network device
3335  *
3336  * Release reference to device to allow it to be freed.
3337  */
3338 static inline void dev_put(struct net_device *dev)
3339 {
3340 	this_cpu_dec(*dev->pcpu_refcnt);
3341 }
3342 
3343 /**
3344  *	dev_hold - get reference to device
3345  *	@dev: network device
3346  *
3347  * Hold reference to device to keep it from being freed.
3348  */
3349 static inline void dev_hold(struct net_device *dev)
3350 {
3351 	this_cpu_inc(*dev->pcpu_refcnt);
3352 }
3353 
3354 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3355  * and _off may be called from IRQ context, but it is caller
3356  * who is responsible for serialization of these calls.
3357  *
3358  * The name carrier is inappropriate, these functions should really be
3359  * called netif_lowerlayer_*() because they represent the state of any
3360  * kind of lower layer not just hardware media.
3361  */
3362 
3363 void linkwatch_init_dev(struct net_device *dev);
3364 void linkwatch_fire_event(struct net_device *dev);
3365 void linkwatch_forget_dev(struct net_device *dev);
3366 
3367 /**
3368  *	netif_carrier_ok - test if carrier present
3369  *	@dev: network device
3370  *
3371  * Check if carrier is present on device
3372  */
3373 static inline bool netif_carrier_ok(const struct net_device *dev)
3374 {
3375 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3376 }
3377 
3378 unsigned long dev_trans_start(struct net_device *dev);
3379 
3380 void __netdev_watchdog_up(struct net_device *dev);
3381 
3382 void netif_carrier_on(struct net_device *dev);
3383 
3384 void netif_carrier_off(struct net_device *dev);
3385 
3386 /**
3387  *	netif_dormant_on - mark device as dormant.
3388  *	@dev: network device
3389  *
3390  * Mark device as dormant (as per RFC2863).
3391  *
3392  * The dormant state indicates that the relevant interface is not
3393  * actually in a condition to pass packets (i.e., it is not 'up') but is
3394  * in a "pending" state, waiting for some external event.  For "on-
3395  * demand" interfaces, this new state identifies the situation where the
3396  * interface is waiting for events to place it in the up state.
3397  */
3398 static inline void netif_dormant_on(struct net_device *dev)
3399 {
3400 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3401 		linkwatch_fire_event(dev);
3402 }
3403 
3404 /**
3405  *	netif_dormant_off - set device as not dormant.
3406  *	@dev: network device
3407  *
3408  * Device is not in dormant state.
3409  */
3410 static inline void netif_dormant_off(struct net_device *dev)
3411 {
3412 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3413 		linkwatch_fire_event(dev);
3414 }
3415 
3416 /**
3417  *	netif_dormant - test if device is dormant
3418  *	@dev: network device
3419  *
3420  * Check if device is dormant.
3421  */
3422 static inline bool netif_dormant(const struct net_device *dev)
3423 {
3424 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3425 }
3426 
3427 
3428 /**
3429  *	netif_oper_up - test if device is operational
3430  *	@dev: network device
3431  *
3432  * Check if carrier is operational
3433  */
3434 static inline bool netif_oper_up(const struct net_device *dev)
3435 {
3436 	return (dev->operstate == IF_OPER_UP ||
3437 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3438 }
3439 
3440 /**
3441  *	netif_device_present - is device available or removed
3442  *	@dev: network device
3443  *
3444  * Check if device has not been removed from system.
3445  */
3446 static inline bool netif_device_present(struct net_device *dev)
3447 {
3448 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3449 }
3450 
3451 void netif_device_detach(struct net_device *dev);
3452 
3453 void netif_device_attach(struct net_device *dev);
3454 
3455 /*
3456  * Network interface message level settings
3457  */
3458 
3459 enum {
3460 	NETIF_MSG_DRV		= 0x0001,
3461 	NETIF_MSG_PROBE		= 0x0002,
3462 	NETIF_MSG_LINK		= 0x0004,
3463 	NETIF_MSG_TIMER		= 0x0008,
3464 	NETIF_MSG_IFDOWN	= 0x0010,
3465 	NETIF_MSG_IFUP		= 0x0020,
3466 	NETIF_MSG_RX_ERR	= 0x0040,
3467 	NETIF_MSG_TX_ERR	= 0x0080,
3468 	NETIF_MSG_TX_QUEUED	= 0x0100,
3469 	NETIF_MSG_INTR		= 0x0200,
3470 	NETIF_MSG_TX_DONE	= 0x0400,
3471 	NETIF_MSG_RX_STATUS	= 0x0800,
3472 	NETIF_MSG_PKTDATA	= 0x1000,
3473 	NETIF_MSG_HW		= 0x2000,
3474 	NETIF_MSG_WOL		= 0x4000,
3475 };
3476 
3477 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3478 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3479 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3480 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3481 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3482 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3483 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3484 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3485 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3486 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3487 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3488 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3489 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3490 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3491 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3492 
3493 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3494 {
3495 	/* use default */
3496 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3497 		return default_msg_enable_bits;
3498 	if (debug_value == 0)	/* no output */
3499 		return 0;
3500 	/* set low N bits */
3501 	return (1 << debug_value) - 1;
3502 }
3503 
3504 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3505 {
3506 	spin_lock(&txq->_xmit_lock);
3507 	txq->xmit_lock_owner = cpu;
3508 }
3509 
3510 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3511 {
3512 	__acquire(&txq->_xmit_lock);
3513 	return true;
3514 }
3515 
3516 static inline void __netif_tx_release(struct netdev_queue *txq)
3517 {
3518 	__release(&txq->_xmit_lock);
3519 }
3520 
3521 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3522 {
3523 	spin_lock_bh(&txq->_xmit_lock);
3524 	txq->xmit_lock_owner = smp_processor_id();
3525 }
3526 
3527 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3528 {
3529 	bool ok = spin_trylock(&txq->_xmit_lock);
3530 	if (likely(ok))
3531 		txq->xmit_lock_owner = smp_processor_id();
3532 	return ok;
3533 }
3534 
3535 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3536 {
3537 	txq->xmit_lock_owner = -1;
3538 	spin_unlock(&txq->_xmit_lock);
3539 }
3540 
3541 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3542 {
3543 	txq->xmit_lock_owner = -1;
3544 	spin_unlock_bh(&txq->_xmit_lock);
3545 }
3546 
3547 static inline void txq_trans_update(struct netdev_queue *txq)
3548 {
3549 	if (txq->xmit_lock_owner != -1)
3550 		txq->trans_start = jiffies;
3551 }
3552 
3553 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3554 static inline void netif_trans_update(struct net_device *dev)
3555 {
3556 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3557 
3558 	if (txq->trans_start != jiffies)
3559 		txq->trans_start = jiffies;
3560 }
3561 
3562 /**
3563  *	netif_tx_lock - grab network device transmit lock
3564  *	@dev: network device
3565  *
3566  * Get network device transmit lock
3567  */
3568 static inline void netif_tx_lock(struct net_device *dev)
3569 {
3570 	unsigned int i;
3571 	int cpu;
3572 
3573 	spin_lock(&dev->tx_global_lock);
3574 	cpu = smp_processor_id();
3575 	for (i = 0; i < dev->num_tx_queues; i++) {
3576 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3577 
3578 		/* We are the only thread of execution doing a
3579 		 * freeze, but we have to grab the _xmit_lock in
3580 		 * order to synchronize with threads which are in
3581 		 * the ->hard_start_xmit() handler and already
3582 		 * checked the frozen bit.
3583 		 */
3584 		__netif_tx_lock(txq, cpu);
3585 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3586 		__netif_tx_unlock(txq);
3587 	}
3588 }
3589 
3590 static inline void netif_tx_lock_bh(struct net_device *dev)
3591 {
3592 	local_bh_disable();
3593 	netif_tx_lock(dev);
3594 }
3595 
3596 static inline void netif_tx_unlock(struct net_device *dev)
3597 {
3598 	unsigned int i;
3599 
3600 	for (i = 0; i < dev->num_tx_queues; i++) {
3601 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3602 
3603 		/* No need to grab the _xmit_lock here.  If the
3604 		 * queue is not stopped for another reason, we
3605 		 * force a schedule.
3606 		 */
3607 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3608 		netif_schedule_queue(txq);
3609 	}
3610 	spin_unlock(&dev->tx_global_lock);
3611 }
3612 
3613 static inline void netif_tx_unlock_bh(struct net_device *dev)
3614 {
3615 	netif_tx_unlock(dev);
3616 	local_bh_enable();
3617 }
3618 
3619 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3620 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3621 		__netif_tx_lock(txq, cpu);		\
3622 	} else {					\
3623 		__netif_tx_acquire(txq);		\
3624 	}						\
3625 }
3626 
3627 #define HARD_TX_TRYLOCK(dev, txq)			\
3628 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3629 		__netif_tx_trylock(txq) :		\
3630 		__netif_tx_acquire(txq))
3631 
3632 #define HARD_TX_UNLOCK(dev, txq) {			\
3633 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3634 		__netif_tx_unlock(txq);			\
3635 	} else {					\
3636 		__netif_tx_release(txq);		\
3637 	}						\
3638 }
3639 
3640 static inline void netif_tx_disable(struct net_device *dev)
3641 {
3642 	unsigned int i;
3643 	int cpu;
3644 
3645 	local_bh_disable();
3646 	cpu = smp_processor_id();
3647 	for (i = 0; i < dev->num_tx_queues; i++) {
3648 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3649 
3650 		__netif_tx_lock(txq, cpu);
3651 		netif_tx_stop_queue(txq);
3652 		__netif_tx_unlock(txq);
3653 	}
3654 	local_bh_enable();
3655 }
3656 
3657 static inline void netif_addr_lock(struct net_device *dev)
3658 {
3659 	spin_lock(&dev->addr_list_lock);
3660 }
3661 
3662 static inline void netif_addr_lock_nested(struct net_device *dev)
3663 {
3664 	int subclass = SINGLE_DEPTH_NESTING;
3665 
3666 	if (dev->netdev_ops->ndo_get_lock_subclass)
3667 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3668 
3669 	spin_lock_nested(&dev->addr_list_lock, subclass);
3670 }
3671 
3672 static inline void netif_addr_lock_bh(struct net_device *dev)
3673 {
3674 	spin_lock_bh(&dev->addr_list_lock);
3675 }
3676 
3677 static inline void netif_addr_unlock(struct net_device *dev)
3678 {
3679 	spin_unlock(&dev->addr_list_lock);
3680 }
3681 
3682 static inline void netif_addr_unlock_bh(struct net_device *dev)
3683 {
3684 	spin_unlock_bh(&dev->addr_list_lock);
3685 }
3686 
3687 /*
3688  * dev_addrs walker. Should be used only for read access. Call with
3689  * rcu_read_lock held.
3690  */
3691 #define for_each_dev_addr(dev, ha) \
3692 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3693 
3694 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3695 
3696 void ether_setup(struct net_device *dev);
3697 
3698 /* Support for loadable net-drivers */
3699 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3700 				    unsigned char name_assign_type,
3701 				    void (*setup)(struct net_device *),
3702 				    unsigned int txqs, unsigned int rxqs);
3703 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3704 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3705 
3706 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3707 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3708 			 count)
3709 
3710 int register_netdev(struct net_device *dev);
3711 void unregister_netdev(struct net_device *dev);
3712 
3713 /* General hardware address lists handling functions */
3714 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3715 		   struct netdev_hw_addr_list *from_list, int addr_len);
3716 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3717 		      struct netdev_hw_addr_list *from_list, int addr_len);
3718 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3719 		       struct net_device *dev,
3720 		       int (*sync)(struct net_device *, const unsigned char *),
3721 		       int (*unsync)(struct net_device *,
3722 				     const unsigned char *));
3723 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3724 			  struct net_device *dev,
3725 			  int (*unsync)(struct net_device *,
3726 					const unsigned char *));
3727 void __hw_addr_init(struct netdev_hw_addr_list *list);
3728 
3729 /* Functions used for device addresses handling */
3730 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3731 		 unsigned char addr_type);
3732 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3733 		 unsigned char addr_type);
3734 void dev_addr_flush(struct net_device *dev);
3735 int dev_addr_init(struct net_device *dev);
3736 
3737 /* Functions used for unicast addresses handling */
3738 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3739 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3740 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3741 int dev_uc_sync(struct net_device *to, struct net_device *from);
3742 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3743 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3744 void dev_uc_flush(struct net_device *dev);
3745 void dev_uc_init(struct net_device *dev);
3746 
3747 /**
3748  *  __dev_uc_sync - Synchonize device's unicast list
3749  *  @dev:  device to sync
3750  *  @sync: function to call if address should be added
3751  *  @unsync: function to call if address should be removed
3752  *
3753  *  Add newly added addresses to the interface, and release
3754  *  addresses that have been deleted.
3755  */
3756 static inline int __dev_uc_sync(struct net_device *dev,
3757 				int (*sync)(struct net_device *,
3758 					    const unsigned char *),
3759 				int (*unsync)(struct net_device *,
3760 					      const unsigned char *))
3761 {
3762 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3763 }
3764 
3765 /**
3766  *  __dev_uc_unsync - Remove synchronized addresses from device
3767  *  @dev:  device to sync
3768  *  @unsync: function to call if address should be removed
3769  *
3770  *  Remove all addresses that were added to the device by dev_uc_sync().
3771  */
3772 static inline void __dev_uc_unsync(struct net_device *dev,
3773 				   int (*unsync)(struct net_device *,
3774 						 const unsigned char *))
3775 {
3776 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3777 }
3778 
3779 /* Functions used for multicast addresses handling */
3780 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3781 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3782 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3783 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3784 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3785 int dev_mc_sync(struct net_device *to, struct net_device *from);
3786 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3787 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3788 void dev_mc_flush(struct net_device *dev);
3789 void dev_mc_init(struct net_device *dev);
3790 
3791 /**
3792  *  __dev_mc_sync - Synchonize device's multicast list
3793  *  @dev:  device to sync
3794  *  @sync: function to call if address should be added
3795  *  @unsync: function to call if address should be removed
3796  *
3797  *  Add newly added addresses to the interface, and release
3798  *  addresses that have been deleted.
3799  */
3800 static inline int __dev_mc_sync(struct net_device *dev,
3801 				int (*sync)(struct net_device *,
3802 					    const unsigned char *),
3803 				int (*unsync)(struct net_device *,
3804 					      const unsigned char *))
3805 {
3806 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3807 }
3808 
3809 /**
3810  *  __dev_mc_unsync - Remove synchronized addresses from device
3811  *  @dev:  device to sync
3812  *  @unsync: function to call if address should be removed
3813  *
3814  *  Remove all addresses that were added to the device by dev_mc_sync().
3815  */
3816 static inline void __dev_mc_unsync(struct net_device *dev,
3817 				   int (*unsync)(struct net_device *,
3818 						 const unsigned char *))
3819 {
3820 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3821 }
3822 
3823 /* Functions used for secondary unicast and multicast support */
3824 void dev_set_rx_mode(struct net_device *dev);
3825 void __dev_set_rx_mode(struct net_device *dev);
3826 int dev_set_promiscuity(struct net_device *dev, int inc);
3827 int dev_set_allmulti(struct net_device *dev, int inc);
3828 void netdev_state_change(struct net_device *dev);
3829 void netdev_notify_peers(struct net_device *dev);
3830 void netdev_features_change(struct net_device *dev);
3831 /* Load a device via the kmod */
3832 void dev_load(struct net *net, const char *name);
3833 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3834 					struct rtnl_link_stats64 *storage);
3835 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3836 			     const struct net_device_stats *netdev_stats);
3837 
3838 extern int		netdev_max_backlog;
3839 extern int		netdev_tstamp_prequeue;
3840 extern int		weight_p;
3841 extern int		dev_weight_rx_bias;
3842 extern int		dev_weight_tx_bias;
3843 extern int		dev_rx_weight;
3844 extern int		dev_tx_weight;
3845 
3846 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3847 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3848 						     struct list_head **iter);
3849 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3850 						     struct list_head **iter);
3851 
3852 /* iterate through upper list, must be called under RCU read lock */
3853 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3854 	for (iter = &(dev)->adj_list.upper, \
3855 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3856 	     updev; \
3857 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3858 
3859 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3860 				  int (*fn)(struct net_device *upper_dev,
3861 					    void *data),
3862 				  void *data);
3863 
3864 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3865 				  struct net_device *upper_dev);
3866 
3867 void *netdev_lower_get_next_private(struct net_device *dev,
3868 				    struct list_head **iter);
3869 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3870 					struct list_head **iter);
3871 
3872 #define netdev_for_each_lower_private(dev, priv, iter) \
3873 	for (iter = (dev)->adj_list.lower.next, \
3874 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3875 	     priv; \
3876 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3877 
3878 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3879 	for (iter = &(dev)->adj_list.lower, \
3880 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3881 	     priv; \
3882 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3883 
3884 void *netdev_lower_get_next(struct net_device *dev,
3885 				struct list_head **iter);
3886 
3887 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3888 	for (iter = (dev)->adj_list.lower.next, \
3889 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3890 	     ldev; \
3891 	     ldev = netdev_lower_get_next(dev, &(iter)))
3892 
3893 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3894 					     struct list_head **iter);
3895 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3896 						 struct list_head **iter);
3897 
3898 int netdev_walk_all_lower_dev(struct net_device *dev,
3899 			      int (*fn)(struct net_device *lower_dev,
3900 					void *data),
3901 			      void *data);
3902 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3903 				  int (*fn)(struct net_device *lower_dev,
3904 					    void *data),
3905 				  void *data);
3906 
3907 void *netdev_adjacent_get_private(struct list_head *adj_list);
3908 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3909 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3910 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3911 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3912 int netdev_master_upper_dev_link(struct net_device *dev,
3913 				 struct net_device *upper_dev,
3914 				 void *upper_priv, void *upper_info);
3915 void netdev_upper_dev_unlink(struct net_device *dev,
3916 			     struct net_device *upper_dev);
3917 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3918 void *netdev_lower_dev_get_private(struct net_device *dev,
3919 				   struct net_device *lower_dev);
3920 void netdev_lower_state_changed(struct net_device *lower_dev,
3921 				void *lower_state_info);
3922 
3923 /* RSS keys are 40 or 52 bytes long */
3924 #define NETDEV_RSS_KEY_LEN 52
3925 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3926 void netdev_rss_key_fill(void *buffer, size_t len);
3927 
3928 int dev_get_nest_level(struct net_device *dev);
3929 int skb_checksum_help(struct sk_buff *skb);
3930 int skb_crc32c_csum_help(struct sk_buff *skb);
3931 int skb_csum_hwoffload_help(struct sk_buff *skb,
3932 			    const netdev_features_t features);
3933 
3934 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3935 				  netdev_features_t features, bool tx_path);
3936 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3937 				    netdev_features_t features);
3938 
3939 struct netdev_bonding_info {
3940 	ifslave	slave;
3941 	ifbond	master;
3942 };
3943 
3944 struct netdev_notifier_bonding_info {
3945 	struct netdev_notifier_info info; /* must be first */
3946 	struct netdev_bonding_info  bonding_info;
3947 };
3948 
3949 void netdev_bonding_info_change(struct net_device *dev,
3950 				struct netdev_bonding_info *bonding_info);
3951 
3952 static inline
3953 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3954 {
3955 	return __skb_gso_segment(skb, features, true);
3956 }
3957 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3958 
3959 static inline bool can_checksum_protocol(netdev_features_t features,
3960 					 __be16 protocol)
3961 {
3962 	if (protocol == htons(ETH_P_FCOE))
3963 		return !!(features & NETIF_F_FCOE_CRC);
3964 
3965 	/* Assume this is an IP checksum (not SCTP CRC) */
3966 
3967 	if (features & NETIF_F_HW_CSUM) {
3968 		/* Can checksum everything */
3969 		return true;
3970 	}
3971 
3972 	switch (protocol) {
3973 	case htons(ETH_P_IP):
3974 		return !!(features & NETIF_F_IP_CSUM);
3975 	case htons(ETH_P_IPV6):
3976 		return !!(features & NETIF_F_IPV6_CSUM);
3977 	default:
3978 		return false;
3979 	}
3980 }
3981 
3982 #ifdef CONFIG_BUG
3983 void netdev_rx_csum_fault(struct net_device *dev);
3984 #else
3985 static inline void netdev_rx_csum_fault(struct net_device *dev)
3986 {
3987 }
3988 #endif
3989 /* rx skb timestamps */
3990 void net_enable_timestamp(void);
3991 void net_disable_timestamp(void);
3992 
3993 #ifdef CONFIG_PROC_FS
3994 int __init dev_proc_init(void);
3995 #else
3996 #define dev_proc_init() 0
3997 #endif
3998 
3999 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4000 					      struct sk_buff *skb, struct net_device *dev,
4001 					      bool more)
4002 {
4003 	skb->xmit_more = more ? 1 : 0;
4004 	return ops->ndo_start_xmit(skb, dev);
4005 }
4006 
4007 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4008 					    struct netdev_queue *txq, bool more)
4009 {
4010 	const struct net_device_ops *ops = dev->netdev_ops;
4011 	int rc;
4012 
4013 	rc = __netdev_start_xmit(ops, skb, dev, more);
4014 	if (rc == NETDEV_TX_OK)
4015 		txq_trans_update(txq);
4016 
4017 	return rc;
4018 }
4019 
4020 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4021 				const void *ns);
4022 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4023 				 const void *ns);
4024 
4025 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4026 {
4027 	return netdev_class_create_file_ns(class_attr, NULL);
4028 }
4029 
4030 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4031 {
4032 	netdev_class_remove_file_ns(class_attr, NULL);
4033 }
4034 
4035 extern struct kobj_ns_type_operations net_ns_type_operations;
4036 
4037 const char *netdev_drivername(const struct net_device *dev);
4038 
4039 void linkwatch_run_queue(void);
4040 
4041 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4042 							  netdev_features_t f2)
4043 {
4044 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4045 		if (f1 & NETIF_F_HW_CSUM)
4046 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4047 		else
4048 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4049 	}
4050 
4051 	return f1 & f2;
4052 }
4053 
4054 static inline netdev_features_t netdev_get_wanted_features(
4055 	struct net_device *dev)
4056 {
4057 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4058 }
4059 netdev_features_t netdev_increment_features(netdev_features_t all,
4060 	netdev_features_t one, netdev_features_t mask);
4061 
4062 /* Allow TSO being used on stacked device :
4063  * Performing the GSO segmentation before last device
4064  * is a performance improvement.
4065  */
4066 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4067 							netdev_features_t mask)
4068 {
4069 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4070 }
4071 
4072 int __netdev_update_features(struct net_device *dev);
4073 void netdev_update_features(struct net_device *dev);
4074 void netdev_change_features(struct net_device *dev);
4075 
4076 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4077 					struct net_device *dev);
4078 
4079 netdev_features_t passthru_features_check(struct sk_buff *skb,
4080 					  struct net_device *dev,
4081 					  netdev_features_t features);
4082 netdev_features_t netif_skb_features(struct sk_buff *skb);
4083 
4084 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4085 {
4086 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4087 
4088 	/* check flags correspondence */
4089 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4090 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4091 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4092 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4093 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4094 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4095 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4096 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4097 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4098 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4099 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4100 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4101 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4102 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4103 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4104 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4105 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4106 
4107 	return (features & feature) == feature;
4108 }
4109 
4110 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4111 {
4112 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4113 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4114 }
4115 
4116 static inline bool netif_needs_gso(struct sk_buff *skb,
4117 				   netdev_features_t features)
4118 {
4119 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4120 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4121 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4122 }
4123 
4124 static inline void netif_set_gso_max_size(struct net_device *dev,
4125 					  unsigned int size)
4126 {
4127 	dev->gso_max_size = size;
4128 }
4129 
4130 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4131 					int pulled_hlen, u16 mac_offset,
4132 					int mac_len)
4133 {
4134 	skb->protocol = protocol;
4135 	skb->encapsulation = 1;
4136 	skb_push(skb, pulled_hlen);
4137 	skb_reset_transport_header(skb);
4138 	skb->mac_header = mac_offset;
4139 	skb->network_header = skb->mac_header + mac_len;
4140 	skb->mac_len = mac_len;
4141 }
4142 
4143 static inline bool netif_is_macsec(const struct net_device *dev)
4144 {
4145 	return dev->priv_flags & IFF_MACSEC;
4146 }
4147 
4148 static inline bool netif_is_macvlan(const struct net_device *dev)
4149 {
4150 	return dev->priv_flags & IFF_MACVLAN;
4151 }
4152 
4153 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4154 {
4155 	return dev->priv_flags & IFF_MACVLAN_PORT;
4156 }
4157 
4158 static inline bool netif_is_ipvlan(const struct net_device *dev)
4159 {
4160 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4161 }
4162 
4163 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4164 {
4165 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4166 }
4167 
4168 static inline bool netif_is_bond_master(const struct net_device *dev)
4169 {
4170 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4171 }
4172 
4173 static inline bool netif_is_bond_slave(const struct net_device *dev)
4174 {
4175 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4176 }
4177 
4178 static inline bool netif_supports_nofcs(struct net_device *dev)
4179 {
4180 	return dev->priv_flags & IFF_SUPP_NOFCS;
4181 }
4182 
4183 static inline bool netif_is_l3_master(const struct net_device *dev)
4184 {
4185 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4186 }
4187 
4188 static inline bool netif_is_l3_slave(const struct net_device *dev)
4189 {
4190 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4191 }
4192 
4193 static inline bool netif_is_bridge_master(const struct net_device *dev)
4194 {
4195 	return dev->priv_flags & IFF_EBRIDGE;
4196 }
4197 
4198 static inline bool netif_is_bridge_port(const struct net_device *dev)
4199 {
4200 	return dev->priv_flags & IFF_BRIDGE_PORT;
4201 }
4202 
4203 static inline bool netif_is_ovs_master(const struct net_device *dev)
4204 {
4205 	return dev->priv_flags & IFF_OPENVSWITCH;
4206 }
4207 
4208 static inline bool netif_is_ovs_port(const struct net_device *dev)
4209 {
4210 	return dev->priv_flags & IFF_OVS_DATAPATH;
4211 }
4212 
4213 static inline bool netif_is_team_master(const struct net_device *dev)
4214 {
4215 	return dev->priv_flags & IFF_TEAM;
4216 }
4217 
4218 static inline bool netif_is_team_port(const struct net_device *dev)
4219 {
4220 	return dev->priv_flags & IFF_TEAM_PORT;
4221 }
4222 
4223 static inline bool netif_is_lag_master(const struct net_device *dev)
4224 {
4225 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4226 }
4227 
4228 static inline bool netif_is_lag_port(const struct net_device *dev)
4229 {
4230 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4231 }
4232 
4233 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4234 {
4235 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4236 }
4237 
4238 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4239 static inline void netif_keep_dst(struct net_device *dev)
4240 {
4241 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4242 }
4243 
4244 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4245 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4246 {
4247 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4248 	return dev->priv_flags & IFF_MACSEC;
4249 }
4250 
4251 extern struct pernet_operations __net_initdata loopback_net_ops;
4252 
4253 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4254 
4255 /* netdev_printk helpers, similar to dev_printk */
4256 
4257 static inline const char *netdev_name(const struct net_device *dev)
4258 {
4259 	if (!dev->name[0] || strchr(dev->name, '%'))
4260 		return "(unnamed net_device)";
4261 	return dev->name;
4262 }
4263 
4264 static inline bool netdev_unregistering(const struct net_device *dev)
4265 {
4266 	return dev->reg_state == NETREG_UNREGISTERING;
4267 }
4268 
4269 static inline const char *netdev_reg_state(const struct net_device *dev)
4270 {
4271 	switch (dev->reg_state) {
4272 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4273 	case NETREG_REGISTERED: return "";
4274 	case NETREG_UNREGISTERING: return " (unregistering)";
4275 	case NETREG_UNREGISTERED: return " (unregistered)";
4276 	case NETREG_RELEASED: return " (released)";
4277 	case NETREG_DUMMY: return " (dummy)";
4278 	}
4279 
4280 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4281 	return " (unknown)";
4282 }
4283 
4284 __printf(3, 4)
4285 void netdev_printk(const char *level, const struct net_device *dev,
4286 		   const char *format, ...);
4287 __printf(2, 3)
4288 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4289 __printf(2, 3)
4290 void netdev_alert(const struct net_device *dev, const char *format, ...);
4291 __printf(2, 3)
4292 void netdev_crit(const struct net_device *dev, const char *format, ...);
4293 __printf(2, 3)
4294 void netdev_err(const struct net_device *dev, const char *format, ...);
4295 __printf(2, 3)
4296 void netdev_warn(const struct net_device *dev, const char *format, ...);
4297 __printf(2, 3)
4298 void netdev_notice(const struct net_device *dev, const char *format, ...);
4299 __printf(2, 3)
4300 void netdev_info(const struct net_device *dev, const char *format, ...);
4301 
4302 #define MODULE_ALIAS_NETDEV(device) \
4303 	MODULE_ALIAS("netdev-" device)
4304 
4305 #if defined(CONFIG_DYNAMIC_DEBUG)
4306 #define netdev_dbg(__dev, format, args...)			\
4307 do {								\
4308 	dynamic_netdev_dbg(__dev, format, ##args);		\
4309 } while (0)
4310 #elif defined(DEBUG)
4311 #define netdev_dbg(__dev, format, args...)			\
4312 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4313 #else
4314 #define netdev_dbg(__dev, format, args...)			\
4315 ({								\
4316 	if (0)							\
4317 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4318 })
4319 #endif
4320 
4321 #if defined(VERBOSE_DEBUG)
4322 #define netdev_vdbg	netdev_dbg
4323 #else
4324 
4325 #define netdev_vdbg(dev, format, args...)			\
4326 ({								\
4327 	if (0)							\
4328 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4329 	0;							\
4330 })
4331 #endif
4332 
4333 /*
4334  * netdev_WARN() acts like dev_printk(), but with the key difference
4335  * of using a WARN/WARN_ON to get the message out, including the
4336  * file/line information and a backtrace.
4337  */
4338 #define netdev_WARN(dev, format, args...)			\
4339 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4340 	     netdev_reg_state(dev), ##args)
4341 
4342 /* netif printk helpers, similar to netdev_printk */
4343 
4344 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4345 do {					  			\
4346 	if (netif_msg_##type(priv))				\
4347 		netdev_printk(level, (dev), fmt, ##args);	\
4348 } while (0)
4349 
4350 #define netif_level(level, priv, type, dev, fmt, args...)	\
4351 do {								\
4352 	if (netif_msg_##type(priv))				\
4353 		netdev_##level(dev, fmt, ##args);		\
4354 } while (0)
4355 
4356 #define netif_emerg(priv, type, dev, fmt, args...)		\
4357 	netif_level(emerg, priv, type, dev, fmt, ##args)
4358 #define netif_alert(priv, type, dev, fmt, args...)		\
4359 	netif_level(alert, priv, type, dev, fmt, ##args)
4360 #define netif_crit(priv, type, dev, fmt, args...)		\
4361 	netif_level(crit, priv, type, dev, fmt, ##args)
4362 #define netif_err(priv, type, dev, fmt, args...)		\
4363 	netif_level(err, priv, type, dev, fmt, ##args)
4364 #define netif_warn(priv, type, dev, fmt, args...)		\
4365 	netif_level(warn, priv, type, dev, fmt, ##args)
4366 #define netif_notice(priv, type, dev, fmt, args...)		\
4367 	netif_level(notice, priv, type, dev, fmt, ##args)
4368 #define netif_info(priv, type, dev, fmt, args...)		\
4369 	netif_level(info, priv, type, dev, fmt, ##args)
4370 
4371 #if defined(CONFIG_DYNAMIC_DEBUG)
4372 #define netif_dbg(priv, type, netdev, format, args...)		\
4373 do {								\
4374 	if (netif_msg_##type(priv))				\
4375 		dynamic_netdev_dbg(netdev, format, ##args);	\
4376 } while (0)
4377 #elif defined(DEBUG)
4378 #define netif_dbg(priv, type, dev, format, args...)		\
4379 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4380 #else
4381 #define netif_dbg(priv, type, dev, format, args...)			\
4382 ({									\
4383 	if (0)								\
4384 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4385 	0;								\
4386 })
4387 #endif
4388 
4389 /* if @cond then downgrade to debug, else print at @level */
4390 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4391 	do {                                                              \
4392 		if (cond)                                                 \
4393 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4394 		else                                                      \
4395 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4396 	} while (0)
4397 
4398 #if defined(VERBOSE_DEBUG)
4399 #define netif_vdbg	netif_dbg
4400 #else
4401 #define netif_vdbg(priv, type, dev, format, args...)		\
4402 ({								\
4403 	if (0)							\
4404 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4405 	0;							\
4406 })
4407 #endif
4408 
4409 /*
4410  *	The list of packet types we will receive (as opposed to discard)
4411  *	and the routines to invoke.
4412  *
4413  *	Why 16. Because with 16 the only overlap we get on a hash of the
4414  *	low nibble of the protocol value is RARP/SNAP/X.25.
4415  *
4416  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4417  *             sure which should go first, but I bet it won't make much
4418  *             difference if we are running VLANs.  The good news is that
4419  *             this protocol won't be in the list unless compiled in, so
4420  *             the average user (w/out VLANs) will not be adversely affected.
4421  *             --BLG
4422  *
4423  *		0800	IP
4424  *		8100    802.1Q VLAN
4425  *		0001	802.3
4426  *		0002	AX.25
4427  *		0004	802.2
4428  *		8035	RARP
4429  *		0005	SNAP
4430  *		0805	X.25
4431  *		0806	ARP
4432  *		8137	IPX
4433  *		0009	Localtalk
4434  *		86DD	IPv6
4435  */
4436 #define PTYPE_HASH_SIZE	(16)
4437 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4438 
4439 #endif	/* _LINUX_NETDEVICE_H */
4440