xref: /linux-6.15/include/linux/netdevice.h (revision f98bc10e)
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/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
40 
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
43 #ifdef CONFIG_DCB
44 #include <net/dcbnl.h>
45 #endif
46 #include <net/netprio_cgroup.h>
47 
48 #include <linux/netdev_features.h>
49 #include <linux/neighbour.h>
50 #include <uapi/linux/netdevice.h>
51 #include <uapi/linux/if_bonding.h>
52 #include <uapi/linux/pkt_cls.h>
53 #include <linux/hashtable.h>
54 
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 struct dsa_port;
59 
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
68 struct xdp_buff;
69 
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 				    const struct ethtool_ops *ops);
72 
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
75 #define NET_RX_DROP		1	/* packet dropped */
76 
77 /*
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, char *buf);
697 	ssize_t (*store)(struct netdev_rx_queue *queue,
698 			 const char *buf, size_t len);
699 };
700 
701 #ifdef CONFIG_XPS
702 /*
703  * This structure holds an XPS map which can be of variable length.  The
704  * map is an array of queues.
705  */
706 struct xps_map {
707 	unsigned int len;
708 	unsigned int alloc_len;
709 	struct rcu_head rcu;
710 	u16 queues[0];
711 };
712 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
713 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714        - sizeof(struct xps_map)) / sizeof(u16))
715 
716 /*
717  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
718  */
719 struct xps_dev_maps {
720 	struct rcu_head rcu;
721 	struct xps_map __rcu *cpu_map[0];
722 };
723 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +		\
724 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
725 #endif /* CONFIG_XPS */
726 
727 #define TC_MAX_QUEUE	16
728 #define TC_BITMASK	15
729 /* HW offloaded queuing disciplines txq count and offset maps */
730 struct netdev_tc_txq {
731 	u16 count;
732 	u16 offset;
733 };
734 
735 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736 /*
737  * This structure is to hold information about the device
738  * configured to run FCoE protocol stack.
739  */
740 struct netdev_fcoe_hbainfo {
741 	char	manufacturer[64];
742 	char	serial_number[64];
743 	char	hardware_version[64];
744 	char	driver_version[64];
745 	char	optionrom_version[64];
746 	char	firmware_version[64];
747 	char	model[256];
748 	char	model_description[256];
749 };
750 #endif
751 
752 #define MAX_PHYS_ITEM_ID_LEN 32
753 
754 /* This structure holds a unique identifier to identify some
755  * physical item (port for example) used by a netdevice.
756  */
757 struct netdev_phys_item_id {
758 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
759 	unsigned char id_len;
760 };
761 
762 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 					    struct netdev_phys_item_id *b)
764 {
765 	return a->id_len == b->id_len &&
766 	       memcmp(a->id, b->id, a->id_len) == 0;
767 }
768 
769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 				       struct sk_buff *skb);
771 
772 enum tc_setup_type {
773 	TC_SETUP_MQPRIO,
774 	TC_SETUP_CLSU32,
775 	TC_SETUP_CLSFLOWER,
776 	TC_SETUP_CLSMATCHALL,
777 	TC_SETUP_CLSBPF,
778 	TC_SETUP_BLOCK,
779 	TC_SETUP_CBS,
780 };
781 
782 /* These structures hold the attributes of bpf state that are being passed
783  * to the netdevice through the bpf op.
784  */
785 enum bpf_netdev_command {
786 	/* Set or clear a bpf program used in the earliest stages of packet
787 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
788 	 * is responsible for calling bpf_prog_put on any old progs that are
789 	 * stored. In case of error, the callee need not release the new prog
790 	 * reference, but on success it takes ownership and must bpf_prog_put
791 	 * when it is no longer used.
792 	 */
793 	XDP_SETUP_PROG,
794 	XDP_SETUP_PROG_HW,
795 	/* Check if a bpf program is set on the device.  The callee should
796 	 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
797 	 * is equivalent to XDP_ATTACHED_DRV.
798 	 */
799 	XDP_QUERY_PROG,
800 	/* BPF program for offload callbacks, invoked at program load time. */
801 	BPF_OFFLOAD_VERIFIER_PREP,
802 	BPF_OFFLOAD_TRANSLATE,
803 	BPF_OFFLOAD_DESTROY,
804 };
805 
806 struct bpf_ext_analyzer_ops;
807 struct netlink_ext_ack;
808 
809 struct netdev_bpf {
810 	enum bpf_netdev_command command;
811 	union {
812 		/* XDP_SETUP_PROG */
813 		struct {
814 			u32 flags;
815 			struct bpf_prog *prog;
816 			struct netlink_ext_ack *extack;
817 		};
818 		/* XDP_QUERY_PROG */
819 		struct {
820 			u8 prog_attached;
821 			u32 prog_id;
822 		};
823 		/* BPF_OFFLOAD_VERIFIER_PREP */
824 		struct {
825 			struct bpf_prog *prog;
826 			const struct bpf_ext_analyzer_ops *ops; /* callee set */
827 		} verifier;
828 		/* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
829 		struct {
830 			struct bpf_prog *prog;
831 		} offload;
832 	};
833 };
834 
835 #ifdef CONFIG_XFRM_OFFLOAD
836 struct xfrmdev_ops {
837 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
838 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
839 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
840 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
841 				       struct xfrm_state *x);
842 };
843 #endif
844 
845 struct dev_ifalias {
846 	struct rcu_head rcuhead;
847 	char ifalias[];
848 };
849 
850 /*
851  * This structure defines the management hooks for network devices.
852  * The following hooks can be defined; unless noted otherwise, they are
853  * optional and can be filled with a null pointer.
854  *
855  * int (*ndo_init)(struct net_device *dev);
856  *     This function is called once when a network device is registered.
857  *     The network device can use this for any late stage initialization
858  *     or semantic validation. It can fail with an error code which will
859  *     be propagated back to register_netdev.
860  *
861  * void (*ndo_uninit)(struct net_device *dev);
862  *     This function is called when device is unregistered or when registration
863  *     fails. It is not called if init fails.
864  *
865  * int (*ndo_open)(struct net_device *dev);
866  *     This function is called when a network device transitions to the up
867  *     state.
868  *
869  * int (*ndo_stop)(struct net_device *dev);
870  *     This function is called when a network device transitions to the down
871  *     state.
872  *
873  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
874  *                               struct net_device *dev);
875  *	Called when a packet needs to be transmitted.
876  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
877  *	the queue before that can happen; it's for obsolete devices and weird
878  *	corner cases, but the stack really does a non-trivial amount
879  *	of useless work if you return NETDEV_TX_BUSY.
880  *	Required; cannot be NULL.
881  *
882  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
883  *					   struct net_device *dev
884  *					   netdev_features_t features);
885  *	Called by core transmit path to determine if device is capable of
886  *	performing offload operations on a given packet. This is to give
887  *	the device an opportunity to implement any restrictions that cannot
888  *	be otherwise expressed by feature flags. The check is called with
889  *	the set of features that the stack has calculated and it returns
890  *	those the driver believes to be appropriate.
891  *
892  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
893  *                         void *accel_priv, select_queue_fallback_t fallback);
894  *	Called to decide which queue to use when device supports multiple
895  *	transmit queues.
896  *
897  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
898  *	This function is called to allow device receiver to make
899  *	changes to configuration when multicast or promiscuous is enabled.
900  *
901  * void (*ndo_set_rx_mode)(struct net_device *dev);
902  *	This function is called device changes address list filtering.
903  *	If driver handles unicast address filtering, it should set
904  *	IFF_UNICAST_FLT in its priv_flags.
905  *
906  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
907  *	This function  is called when the Media Access Control address
908  *	needs to be changed. If this interface is not defined, the
909  *	MAC address can not be changed.
910  *
911  * int (*ndo_validate_addr)(struct net_device *dev);
912  *	Test if Media Access Control address is valid for the device.
913  *
914  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
915  *	Called when a user requests an ioctl which can't be handled by
916  *	the generic interface code. If not defined ioctls return
917  *	not supported error code.
918  *
919  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
920  *	Used to set network devices bus interface parameters. This interface
921  *	is retained for legacy reasons; new devices should use the bus
922  *	interface (PCI) for low level management.
923  *
924  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
925  *	Called when a user wants to change the Maximum Transfer Unit
926  *	of a device.
927  *
928  * void (*ndo_tx_timeout)(struct net_device *dev);
929  *	Callback used when the transmitter has not made any progress
930  *	for dev->watchdog ticks.
931  *
932  * void (*ndo_get_stats64)(struct net_device *dev,
933  *                         struct rtnl_link_stats64 *storage);
934  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
935  *	Called when a user wants to get the network device usage
936  *	statistics. Drivers must do one of the following:
937  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
938  *	   rtnl_link_stats64 structure passed by the caller.
939  *	2. Define @ndo_get_stats to update a net_device_stats structure
940  *	   (which should normally be dev->stats) and return a pointer to
941  *	   it. The structure may be changed asynchronously only if each
942  *	   field is written atomically.
943  *	3. Update dev->stats asynchronously and atomically, and define
944  *	   neither operation.
945  *
946  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
947  *	Return true if this device supports offload stats of this attr_id.
948  *
949  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
950  *	void *attr_data)
951  *	Get statistics for offload operations by attr_id. Write it into the
952  *	attr_data pointer.
953  *
954  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
955  *	If device supports VLAN filtering this function is called when a
956  *	VLAN id is registered.
957  *
958  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
959  *	If device supports VLAN filtering this function is called when a
960  *	VLAN id is unregistered.
961  *
962  * void (*ndo_poll_controller)(struct net_device *dev);
963  *
964  *	SR-IOV management functions.
965  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
966  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
967  *			  u8 qos, __be16 proto);
968  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
969  *			  int max_tx_rate);
970  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
971  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
972  * int (*ndo_get_vf_config)(struct net_device *dev,
973  *			    int vf, struct ifla_vf_info *ivf);
974  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
975  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
976  *			  struct nlattr *port[]);
977  *
978  *      Enable or disable the VF ability to query its RSS Redirection Table and
979  *      Hash Key. This is needed since on some devices VF share this information
980  *      with PF and querying it may introduce a theoretical security risk.
981  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
982  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
983  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
984  *		       void *type_data);
985  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
986  *	This is always called from the stack with the rtnl lock held and netif
987  *	tx queues stopped. This allows the netdevice to perform queue
988  *	management safely.
989  *
990  *	Fiber Channel over Ethernet (FCoE) offload functions.
991  * int (*ndo_fcoe_enable)(struct net_device *dev);
992  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
993  *	so the underlying device can perform whatever needed configuration or
994  *	initialization to support acceleration of FCoE traffic.
995  *
996  * int (*ndo_fcoe_disable)(struct net_device *dev);
997  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
998  *	so the underlying device can perform whatever needed clean-ups to
999  *	stop supporting acceleration of FCoE traffic.
1000  *
1001  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1002  *			     struct scatterlist *sgl, unsigned int sgc);
1003  *	Called when the FCoE Initiator wants to initialize an I/O that
1004  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1005  *	perform necessary setup and returns 1 to indicate the device is set up
1006  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1007  *
1008  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1009  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1010  *	indicated by the FC exchange id 'xid', so the underlying device can
1011  *	clean up and reuse resources for later DDP requests.
1012  *
1013  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1014  *			      struct scatterlist *sgl, unsigned int sgc);
1015  *	Called when the FCoE Target wants to initialize an I/O that
1016  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1017  *	perform necessary setup and returns 1 to indicate the device is set up
1018  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1019  *
1020  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1021  *			       struct netdev_fcoe_hbainfo *hbainfo);
1022  *	Called when the FCoE Protocol stack wants information on the underlying
1023  *	device. This information is utilized by the FCoE protocol stack to
1024  *	register attributes with Fiber Channel management service as per the
1025  *	FC-GS Fabric Device Management Information(FDMI) specification.
1026  *
1027  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1028  *	Called when the underlying device wants to override default World Wide
1029  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1030  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1031  *	protocol stack to use.
1032  *
1033  *	RFS acceleration.
1034  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1035  *			    u16 rxq_index, u32 flow_id);
1036  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1037  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1038  *	Return the filter ID on success, or a negative error code.
1039  *
1040  *	Slave management functions (for bridge, bonding, etc).
1041  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1042  *	Called to make another netdev an underling.
1043  *
1044  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1045  *	Called to release previously enslaved netdev.
1046  *
1047  *      Feature/offload setting functions.
1048  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1049  *		netdev_features_t features);
1050  *	Adjusts the requested feature flags according to device-specific
1051  *	constraints, and returns the resulting flags. Must not modify
1052  *	the device state.
1053  *
1054  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1055  *	Called to update device configuration to new features. Passed
1056  *	feature set might be less than what was returned by ndo_fix_features()).
1057  *	Must return >0 or -errno if it changed dev->features itself.
1058  *
1059  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1060  *		      struct net_device *dev,
1061  *		      const unsigned char *addr, u16 vid, u16 flags)
1062  *	Adds an FDB entry to dev for addr.
1063  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1064  *		      struct net_device *dev,
1065  *		      const unsigned char *addr, u16 vid)
1066  *	Deletes the FDB entry from dev coresponding to addr.
1067  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1068  *		       struct net_device *dev, struct net_device *filter_dev,
1069  *		       int *idx)
1070  *	Used to add FDB entries to dump requests. Implementers should add
1071  *	entries to skb and update idx with the number of entries.
1072  *
1073  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1074  *			     u16 flags)
1075  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1076  *			     struct net_device *dev, u32 filter_mask,
1077  *			     int nlflags)
1078  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1079  *			     u16 flags);
1080  *
1081  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1082  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1083  *	which do not represent real hardware may define this to allow their
1084  *	userspace components to manage their virtual carrier state. Devices
1085  *	that determine carrier state from physical hardware properties (eg
1086  *	network cables) or protocol-dependent mechanisms (eg
1087  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1088  *
1089  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1090  *			       struct netdev_phys_item_id *ppid);
1091  *	Called to get ID of physical port of this device. If driver does
1092  *	not implement this, it is assumed that the hw is not able to have
1093  *	multiple net devices on single physical port.
1094  *
1095  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1096  *			      struct udp_tunnel_info *ti);
1097  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1098  *	address family that a UDP tunnel is listnening to. It is called only
1099  *	when a new port starts listening. The operation is protected by the
1100  *	RTNL.
1101  *
1102  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1103  *			      struct udp_tunnel_info *ti);
1104  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1105  *	address family that the UDP tunnel is not listening to anymore. The
1106  *	operation is protected by the RTNL.
1107  *
1108  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1109  *				 struct net_device *dev)
1110  *	Called by upper layer devices to accelerate switching or other
1111  *	station functionality into hardware. 'pdev is the lowerdev
1112  *	to use for the offload and 'dev' is the net device that will
1113  *	back the offload. Returns a pointer to the private structure
1114  *	the upper layer will maintain.
1115  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1116  *	Called by upper layer device to delete the station created
1117  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1118  *	the station and priv is the structure returned by the add
1119  *	operation.
1120  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1121  *			     int queue_index, u32 maxrate);
1122  *	Called when a user wants to set a max-rate limitation of specific
1123  *	TX queue.
1124  * int (*ndo_get_iflink)(const struct net_device *dev);
1125  *	Called to get the iflink value of this device.
1126  * void (*ndo_change_proto_down)(struct net_device *dev,
1127  *				 bool proto_down);
1128  *	This function is used to pass protocol port error state information
1129  *	to the switch driver. The switch driver can react to the proto_down
1130  *      by doing a phys down on the associated switch port.
1131  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1132  *	This function is used to get egress tunnel information for given skb.
1133  *	This is useful for retrieving outer tunnel header parameters while
1134  *	sampling packet.
1135  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1136  *	This function is used to specify the headroom that the skb must
1137  *	consider when allocation skb during packet reception. Setting
1138  *	appropriate rx headroom value allows avoiding skb head copy on
1139  *	forward. Setting a negative value resets the rx headroom to the
1140  *	default value.
1141  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1142  *	This function is used to set or query state related to XDP on the
1143  *	netdevice and manage BPF offload. See definition of
1144  *	enum bpf_netdev_command for details.
1145  * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1146  *	This function is used to submit a XDP packet for transmit on a
1147  *	netdevice.
1148  * void (*ndo_xdp_flush)(struct net_device *dev);
1149  *	This function is used to inform the driver to flush a particular
1150  *	xdp tx queue. Must be called on same CPU as xdp_xmit.
1151  */
1152 struct net_device_ops {
1153 	int			(*ndo_init)(struct net_device *dev);
1154 	void			(*ndo_uninit)(struct net_device *dev);
1155 	int			(*ndo_open)(struct net_device *dev);
1156 	int			(*ndo_stop)(struct net_device *dev);
1157 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1158 						  struct net_device *dev);
1159 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1160 						      struct net_device *dev,
1161 						      netdev_features_t features);
1162 	u16			(*ndo_select_queue)(struct net_device *dev,
1163 						    struct sk_buff *skb,
1164 						    void *accel_priv,
1165 						    select_queue_fallback_t fallback);
1166 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1167 						       int flags);
1168 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1169 	int			(*ndo_set_mac_address)(struct net_device *dev,
1170 						       void *addr);
1171 	int			(*ndo_validate_addr)(struct net_device *dev);
1172 	int			(*ndo_do_ioctl)(struct net_device *dev,
1173 					        struct ifreq *ifr, int cmd);
1174 	int			(*ndo_set_config)(struct net_device *dev,
1175 					          struct ifmap *map);
1176 	int			(*ndo_change_mtu)(struct net_device *dev,
1177 						  int new_mtu);
1178 	int			(*ndo_neigh_setup)(struct net_device *dev,
1179 						   struct neigh_parms *);
1180 	void			(*ndo_tx_timeout) (struct net_device *dev);
1181 
1182 	void			(*ndo_get_stats64)(struct net_device *dev,
1183 						   struct rtnl_link_stats64 *storage);
1184 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1185 	int			(*ndo_get_offload_stats)(int attr_id,
1186 							 const struct net_device *dev,
1187 							 void *attr_data);
1188 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1189 
1190 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1191 						       __be16 proto, u16 vid);
1192 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1193 						        __be16 proto, u16 vid);
1194 #ifdef CONFIG_NET_POLL_CONTROLLER
1195 	void                    (*ndo_poll_controller)(struct net_device *dev);
1196 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1197 						     struct netpoll_info *info);
1198 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1199 #endif
1200 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1201 						  int queue, u8 *mac);
1202 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1203 						   int queue, u16 vlan,
1204 						   u8 qos, __be16 proto);
1205 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1206 						   int vf, int min_tx_rate,
1207 						   int max_tx_rate);
1208 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1209 						       int vf, bool setting);
1210 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1211 						    int vf, bool setting);
1212 	int			(*ndo_get_vf_config)(struct net_device *dev,
1213 						     int vf,
1214 						     struct ifla_vf_info *ivf);
1215 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1216 							 int vf, int link_state);
1217 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1218 						    int vf,
1219 						    struct ifla_vf_stats
1220 						    *vf_stats);
1221 	int			(*ndo_set_vf_port)(struct net_device *dev,
1222 						   int vf,
1223 						   struct nlattr *port[]);
1224 	int			(*ndo_get_vf_port)(struct net_device *dev,
1225 						   int vf, struct sk_buff *skb);
1226 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1227 						   int vf, u64 guid,
1228 						   int guid_type);
1229 	int			(*ndo_set_vf_rss_query_en)(
1230 						   struct net_device *dev,
1231 						   int vf, bool setting);
1232 	int			(*ndo_setup_tc)(struct net_device *dev,
1233 						enum tc_setup_type type,
1234 						void *type_data);
1235 #if IS_ENABLED(CONFIG_FCOE)
1236 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1237 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1238 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1239 						      u16 xid,
1240 						      struct scatterlist *sgl,
1241 						      unsigned int sgc);
1242 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1243 						     u16 xid);
1244 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1245 						       u16 xid,
1246 						       struct scatterlist *sgl,
1247 						       unsigned int sgc);
1248 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1249 							struct netdev_fcoe_hbainfo *hbainfo);
1250 #endif
1251 
1252 #if IS_ENABLED(CONFIG_LIBFCOE)
1253 #define NETDEV_FCOE_WWNN 0
1254 #define NETDEV_FCOE_WWPN 1
1255 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1256 						    u64 *wwn, int type);
1257 #endif
1258 
1259 #ifdef CONFIG_RFS_ACCEL
1260 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1261 						     const struct sk_buff *skb,
1262 						     u16 rxq_index,
1263 						     u32 flow_id);
1264 #endif
1265 	int			(*ndo_add_slave)(struct net_device *dev,
1266 						 struct net_device *slave_dev,
1267 						 struct netlink_ext_ack *extack);
1268 	int			(*ndo_del_slave)(struct net_device *dev,
1269 						 struct net_device *slave_dev);
1270 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1271 						    netdev_features_t features);
1272 	int			(*ndo_set_features)(struct net_device *dev,
1273 						    netdev_features_t features);
1274 	int			(*ndo_neigh_construct)(struct net_device *dev,
1275 						       struct neighbour *n);
1276 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1277 						     struct neighbour *n);
1278 
1279 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1280 					       struct nlattr *tb[],
1281 					       struct net_device *dev,
1282 					       const unsigned char *addr,
1283 					       u16 vid,
1284 					       u16 flags);
1285 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1286 					       struct nlattr *tb[],
1287 					       struct net_device *dev,
1288 					       const unsigned char *addr,
1289 					       u16 vid);
1290 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1291 						struct netlink_callback *cb,
1292 						struct net_device *dev,
1293 						struct net_device *filter_dev,
1294 						int *idx);
1295 
1296 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1297 						      struct nlmsghdr *nlh,
1298 						      u16 flags);
1299 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1300 						      u32 pid, u32 seq,
1301 						      struct net_device *dev,
1302 						      u32 filter_mask,
1303 						      int nlflags);
1304 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1305 						      struct nlmsghdr *nlh,
1306 						      u16 flags);
1307 	int			(*ndo_change_carrier)(struct net_device *dev,
1308 						      bool new_carrier);
1309 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1310 							struct netdev_phys_item_id *ppid);
1311 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1312 							  char *name, size_t len);
1313 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1314 						      struct udp_tunnel_info *ti);
1315 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1316 						      struct udp_tunnel_info *ti);
1317 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1318 							struct net_device *dev);
1319 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1320 							void *priv);
1321 
1322 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1323 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1324 						      int queue_index,
1325 						      u32 maxrate);
1326 	int			(*ndo_get_iflink)(const struct net_device *dev);
1327 	int			(*ndo_change_proto_down)(struct net_device *dev,
1328 							 bool proto_down);
1329 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1330 						       struct sk_buff *skb);
1331 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1332 						       int needed_headroom);
1333 	int			(*ndo_bpf)(struct net_device *dev,
1334 					   struct netdev_bpf *bpf);
1335 	int			(*ndo_xdp_xmit)(struct net_device *dev,
1336 						struct xdp_buff *xdp);
1337 	void			(*ndo_xdp_flush)(struct net_device *dev);
1338 };
1339 
1340 /**
1341  * enum net_device_priv_flags - &struct net_device priv_flags
1342  *
1343  * These are the &struct net_device, they are only set internally
1344  * by drivers and used in the kernel. These flags are invisible to
1345  * userspace; this means that the order of these flags can change
1346  * during any kernel release.
1347  *
1348  * You should have a pretty good reason to be extending these flags.
1349  *
1350  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1351  * @IFF_EBRIDGE: Ethernet bridging device
1352  * @IFF_BONDING: bonding master or slave
1353  * @IFF_ISATAP: ISATAP interface (RFC4214)
1354  * @IFF_WAN_HDLC: WAN HDLC device
1355  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1356  *	release skb->dst
1357  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1358  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1359  * @IFF_MACVLAN_PORT: device used as macvlan port
1360  * @IFF_BRIDGE_PORT: device used as bridge port
1361  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1362  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1363  * @IFF_UNICAST_FLT: Supports unicast filtering
1364  * @IFF_TEAM_PORT: device used as team port
1365  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1366  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1367  *	change when it's running
1368  * @IFF_MACVLAN: Macvlan device
1369  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1370  *	underlying stacked devices
1371  * @IFF_IPVLAN_MASTER: IPvlan master device
1372  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1373  * @IFF_L3MDEV_MASTER: device is an L3 master device
1374  * @IFF_NO_QUEUE: device can run without qdisc attached
1375  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1376  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1377  * @IFF_TEAM: device is a team device
1378  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1379  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1380  *	entity (i.e. the master device for bridged veth)
1381  * @IFF_MACSEC: device is a MACsec device
1382  */
1383 enum netdev_priv_flags {
1384 	IFF_802_1Q_VLAN			= 1<<0,
1385 	IFF_EBRIDGE			= 1<<1,
1386 	IFF_BONDING			= 1<<2,
1387 	IFF_ISATAP			= 1<<3,
1388 	IFF_WAN_HDLC			= 1<<4,
1389 	IFF_XMIT_DST_RELEASE		= 1<<5,
1390 	IFF_DONT_BRIDGE			= 1<<6,
1391 	IFF_DISABLE_NETPOLL		= 1<<7,
1392 	IFF_MACVLAN_PORT		= 1<<8,
1393 	IFF_BRIDGE_PORT			= 1<<9,
1394 	IFF_OVS_DATAPATH		= 1<<10,
1395 	IFF_TX_SKB_SHARING		= 1<<11,
1396 	IFF_UNICAST_FLT			= 1<<12,
1397 	IFF_TEAM_PORT			= 1<<13,
1398 	IFF_SUPP_NOFCS			= 1<<14,
1399 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1400 	IFF_MACVLAN			= 1<<16,
1401 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1402 	IFF_IPVLAN_MASTER		= 1<<18,
1403 	IFF_IPVLAN_SLAVE		= 1<<19,
1404 	IFF_L3MDEV_MASTER		= 1<<20,
1405 	IFF_NO_QUEUE			= 1<<21,
1406 	IFF_OPENVSWITCH			= 1<<22,
1407 	IFF_L3MDEV_SLAVE		= 1<<23,
1408 	IFF_TEAM			= 1<<24,
1409 	IFF_RXFH_CONFIGURED		= 1<<25,
1410 	IFF_PHONY_HEADROOM		= 1<<26,
1411 	IFF_MACSEC			= 1<<27,
1412 };
1413 
1414 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1415 #define IFF_EBRIDGE			IFF_EBRIDGE
1416 #define IFF_BONDING			IFF_BONDING
1417 #define IFF_ISATAP			IFF_ISATAP
1418 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1419 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1420 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1421 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1422 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1423 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1424 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1425 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1426 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1427 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1428 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1429 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1430 #define IFF_MACVLAN			IFF_MACVLAN
1431 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1432 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1433 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1434 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1435 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1436 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1437 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1438 #define IFF_TEAM			IFF_TEAM
1439 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1440 #define IFF_MACSEC			IFF_MACSEC
1441 
1442 /**
1443  *	struct net_device - The DEVICE structure.
1444  *
1445  *	Actually, this whole structure is a big mistake.  It mixes I/O
1446  *	data with strictly "high-level" data, and it has to know about
1447  *	almost every data structure used in the INET module.
1448  *
1449  *	@name:	This is the first field of the "visible" part of this structure
1450  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1451  *		of the interface.
1452  *
1453  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1454  *	@ifalias:	SNMP alias
1455  *	@mem_end:	Shared memory end
1456  *	@mem_start:	Shared memory start
1457  *	@base_addr:	Device I/O address
1458  *	@irq:		Device IRQ number
1459  *
1460  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1461  *
1462  *	@state:		Generic network queuing layer state, see netdev_state_t
1463  *	@dev_list:	The global list of network devices
1464  *	@napi_list:	List entry used for polling NAPI devices
1465  *	@unreg_list:	List entry  when we are unregistering the
1466  *			device; see the function unregister_netdev
1467  *	@close_list:	List entry used when we are closing the device
1468  *	@ptype_all:     Device-specific packet handlers for all protocols
1469  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1470  *
1471  *	@adj_list:	Directly linked devices, like slaves for bonding
1472  *	@features:	Currently active device features
1473  *	@hw_features:	User-changeable features
1474  *
1475  *	@wanted_features:	User-requested features
1476  *	@vlan_features:		Mask of features inheritable by VLAN devices
1477  *
1478  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1479  *				This field indicates what encapsulation
1480  *				offloads the hardware is capable of doing,
1481  *				and drivers will need to set them appropriately.
1482  *
1483  *	@mpls_features:	Mask of features inheritable by MPLS
1484  *
1485  *	@ifindex:	interface index
1486  *	@group:		The group the device belongs to
1487  *
1488  *	@stats:		Statistics struct, which was left as a legacy, use
1489  *			rtnl_link_stats64 instead
1490  *
1491  *	@rx_dropped:	Dropped packets by core network,
1492  *			do not use this in drivers
1493  *	@tx_dropped:	Dropped packets by core network,
1494  *			do not use this in drivers
1495  *	@rx_nohandler:	nohandler dropped packets by core network on
1496  *			inactive devices, do not use this in drivers
1497  *
1498  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1499  *				instead of ioctl,
1500  *				see <net/iw_handler.h> for details.
1501  *	@wireless_data:	Instance data managed by the core of wireless extensions
1502  *
1503  *	@netdev_ops:	Includes several pointers to callbacks,
1504  *			if one wants to override the ndo_*() functions
1505  *	@ethtool_ops:	Management operations
1506  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1507  *			discovery handling. Necessary for e.g. 6LoWPAN.
1508  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1509  *			of Layer 2 headers.
1510  *
1511  *	@flags:		Interface flags (a la BSD)
1512  *	@priv_flags:	Like 'flags' but invisible to userspace,
1513  *			see if.h for the definitions
1514  *	@gflags:	Global flags ( kept as legacy )
1515  *	@padded:	How much padding added by alloc_netdev()
1516  *	@operstate:	RFC2863 operstate
1517  *	@link_mode:	Mapping policy to operstate
1518  *	@if_port:	Selectable AUI, TP, ...
1519  *	@dma:		DMA channel
1520  *	@mtu:		Interface MTU value
1521  *	@min_mtu:	Interface Minimum MTU value
1522  *	@max_mtu:	Interface Maximum MTU value
1523  *	@type:		Interface hardware type
1524  *	@hard_header_len: Maximum hardware header length.
1525  *	@min_header_len:  Minimum hardware header length
1526  *
1527  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1528  *			  cases can this be guaranteed
1529  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1530  *			  cases can this be guaranteed. Some cases also use
1531  *			  LL_MAX_HEADER instead to allocate the skb
1532  *
1533  *	interface address info:
1534  *
1535  * 	@perm_addr:		Permanent hw address
1536  * 	@addr_assign_type:	Hw address assignment type
1537  * 	@addr_len:		Hardware address length
1538  *	@neigh_priv_len:	Used in neigh_alloc()
1539  * 	@dev_id:		Used to differentiate devices that share
1540  * 				the same link layer address
1541  * 	@dev_port:		Used to differentiate devices that share
1542  * 				the same function
1543  *	@addr_list_lock:	XXX: need comments on this one
1544  *	@uc_promisc:		Counter that indicates promiscuous mode
1545  *				has been enabled due to the need to listen to
1546  *				additional unicast addresses in a device that
1547  *				does not implement ndo_set_rx_mode()
1548  *	@uc:			unicast mac addresses
1549  *	@mc:			multicast mac addresses
1550  *	@dev_addrs:		list of device hw addresses
1551  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1552  *	@promiscuity:		Number of times the NIC is told to work in
1553  *				promiscuous mode; if it becomes 0 the NIC will
1554  *				exit promiscuous mode
1555  *	@allmulti:		Counter, enables or disables allmulticast mode
1556  *
1557  *	@vlan_info:	VLAN info
1558  *	@dsa_ptr:	dsa specific data
1559  *	@tipc_ptr:	TIPC specific data
1560  *	@atalk_ptr:	AppleTalk link
1561  *	@ip_ptr:	IPv4 specific data
1562  *	@dn_ptr:	DECnet specific data
1563  *	@ip6_ptr:	IPv6 specific data
1564  *	@ax25_ptr:	AX.25 specific data
1565  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1566  *
1567  *	@dev_addr:	Hw address (before bcast,
1568  *			because most packets are unicast)
1569  *
1570  *	@_rx:			Array of RX queues
1571  *	@num_rx_queues:		Number of RX queues
1572  *				allocated at register_netdev() time
1573  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1574  *
1575  *	@rx_handler:		handler for received packets
1576  *	@rx_handler_data: 	XXX: need comments on this one
1577  *	@miniq_ingress:		ingress/clsact qdisc specific data for
1578  *				ingress processing
1579  *	@ingress_queue:		XXX: need comments on this one
1580  *	@broadcast:		hw bcast address
1581  *
1582  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1583  *			indexed by RX queue number. Assigned by driver.
1584  *			This must only be set if the ndo_rx_flow_steer
1585  *			operation is defined
1586  *	@index_hlist:		Device index hash chain
1587  *
1588  *	@_tx:			Array of TX queues
1589  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1590  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1591  *	@qdisc:			Root qdisc from userspace point of view
1592  *	@tx_queue_len:		Max frames per queue allowed
1593  *	@tx_global_lock: 	XXX: need comments on this one
1594  *
1595  *	@xps_maps:	XXX: need comments on this one
1596  *	@miniq_egress:		clsact qdisc specific data for
1597  *				egress processing
1598  *	@watchdog_timeo:	Represents the timeout that is used by
1599  *				the watchdog (see dev_watchdog())
1600  *	@watchdog_timer:	List of timers
1601  *
1602  *	@pcpu_refcnt:		Number of references to this device
1603  *	@todo_list:		Delayed register/unregister
1604  *	@link_watch_list:	XXX: need comments on this one
1605  *
1606  *	@reg_state:		Register/unregister state machine
1607  *	@dismantle:		Device is going to be freed
1608  *	@rtnl_link_state:	This enum represents the phases of creating
1609  *				a new link
1610  *
1611  *	@needs_free_netdev:	Should unregister perform free_netdev?
1612  *	@priv_destructor:	Called from unregister
1613  *	@npinfo:		XXX: need comments on this one
1614  * 	@nd_net:		Network namespace this network device is inside
1615  *
1616  * 	@ml_priv:	Mid-layer private
1617  * 	@lstats:	Loopback statistics
1618  * 	@tstats:	Tunnel statistics
1619  * 	@dstats:	Dummy statistics
1620  * 	@vstats:	Virtual ethernet statistics
1621  *
1622  *	@garp_port:	GARP
1623  *	@mrp_port:	MRP
1624  *
1625  *	@dev:		Class/net/name entry
1626  *	@sysfs_groups:	Space for optional device, statistics and wireless
1627  *			sysfs groups
1628  *
1629  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1630  *	@rtnl_link_ops:	Rtnl_link_ops
1631  *
1632  *	@gso_max_size:	Maximum size of generic segmentation offload
1633  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1634  *			NIC for GSO
1635  *
1636  *	@dcbnl_ops:	Data Center Bridging netlink ops
1637  *	@num_tc:	Number of traffic classes in the net device
1638  *	@tc_to_txq:	XXX: need comments on this one
1639  *	@prio_tc_map:	XXX: need comments on this one
1640  *
1641  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1642  *
1643  *	@priomap:	XXX: need comments on this one
1644  *	@phydev:	Physical device may attach itself
1645  *			for hardware timestamping
1646  *
1647  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1648  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1649  *
1650  *	@proto_down:	protocol port state information can be sent to the
1651  *			switch driver and used to set the phys state of the
1652  *			switch port.
1653  *
1654  *	FIXME: cleanup struct net_device such that network protocol info
1655  *	moves out.
1656  */
1657 
1658 struct net_device {
1659 	char			name[IFNAMSIZ];
1660 	struct hlist_node	name_hlist;
1661 	struct dev_ifalias	__rcu *ifalias;
1662 	/*
1663 	 *	I/O specific fields
1664 	 *	FIXME: Merge these and struct ifmap into one
1665 	 */
1666 	unsigned long		mem_end;
1667 	unsigned long		mem_start;
1668 	unsigned long		base_addr;
1669 	int			irq;
1670 
1671 	atomic_t		carrier_changes;
1672 
1673 	/*
1674 	 *	Some hardware also needs these fields (state,dev_list,
1675 	 *	napi_list,unreg_list,close_list) but they are not
1676 	 *	part of the usual set specified in Space.c.
1677 	 */
1678 
1679 	unsigned long		state;
1680 
1681 	struct list_head	dev_list;
1682 	struct list_head	napi_list;
1683 	struct list_head	unreg_list;
1684 	struct list_head	close_list;
1685 	struct list_head	ptype_all;
1686 	struct list_head	ptype_specific;
1687 
1688 	struct {
1689 		struct list_head upper;
1690 		struct list_head lower;
1691 	} adj_list;
1692 
1693 	netdev_features_t	features;
1694 	netdev_features_t	hw_features;
1695 	netdev_features_t	wanted_features;
1696 	netdev_features_t	vlan_features;
1697 	netdev_features_t	hw_enc_features;
1698 	netdev_features_t	mpls_features;
1699 	netdev_features_t	gso_partial_features;
1700 
1701 	int			ifindex;
1702 	int			group;
1703 
1704 	struct net_device_stats	stats;
1705 
1706 	atomic_long_t		rx_dropped;
1707 	atomic_long_t		tx_dropped;
1708 	atomic_long_t		rx_nohandler;
1709 
1710 #ifdef CONFIG_WIRELESS_EXT
1711 	const struct iw_handler_def *wireless_handlers;
1712 	struct iw_public_data	*wireless_data;
1713 #endif
1714 	const struct net_device_ops *netdev_ops;
1715 	const struct ethtool_ops *ethtool_ops;
1716 #ifdef CONFIG_NET_SWITCHDEV
1717 	const struct switchdev_ops *switchdev_ops;
1718 #endif
1719 #ifdef CONFIG_NET_L3_MASTER_DEV
1720 	const struct l3mdev_ops	*l3mdev_ops;
1721 #endif
1722 #if IS_ENABLED(CONFIG_IPV6)
1723 	const struct ndisc_ops *ndisc_ops;
1724 #endif
1725 
1726 #ifdef CONFIG_XFRM
1727 	const struct xfrmdev_ops *xfrmdev_ops;
1728 #endif
1729 
1730 	const struct header_ops *header_ops;
1731 
1732 	unsigned int		flags;
1733 	unsigned int		priv_flags;
1734 
1735 	unsigned short		gflags;
1736 	unsigned short		padded;
1737 
1738 	unsigned char		operstate;
1739 	unsigned char		link_mode;
1740 
1741 	unsigned char		if_port;
1742 	unsigned char		dma;
1743 
1744 	unsigned int		mtu;
1745 	unsigned int		min_mtu;
1746 	unsigned int		max_mtu;
1747 	unsigned short		type;
1748 	unsigned short		hard_header_len;
1749 	unsigned char		min_header_len;
1750 
1751 	unsigned short		needed_headroom;
1752 	unsigned short		needed_tailroom;
1753 
1754 	/* Interface address info. */
1755 	unsigned char		perm_addr[MAX_ADDR_LEN];
1756 	unsigned char		addr_assign_type;
1757 	unsigned char		addr_len;
1758 	unsigned short		neigh_priv_len;
1759 	unsigned short          dev_id;
1760 	unsigned short          dev_port;
1761 	spinlock_t		addr_list_lock;
1762 	unsigned char		name_assign_type;
1763 	bool			uc_promisc;
1764 	struct netdev_hw_addr_list	uc;
1765 	struct netdev_hw_addr_list	mc;
1766 	struct netdev_hw_addr_list	dev_addrs;
1767 
1768 #ifdef CONFIG_SYSFS
1769 	struct kset		*queues_kset;
1770 #endif
1771 	unsigned int		promiscuity;
1772 	unsigned int		allmulti;
1773 
1774 
1775 	/* Protocol-specific pointers */
1776 
1777 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1778 	struct vlan_info __rcu	*vlan_info;
1779 #endif
1780 #if IS_ENABLED(CONFIG_NET_DSA)
1781 	struct dsa_port		*dsa_ptr;
1782 #endif
1783 #if IS_ENABLED(CONFIG_TIPC)
1784 	struct tipc_bearer __rcu *tipc_ptr;
1785 #endif
1786 	void 			*atalk_ptr;
1787 	struct in_device __rcu	*ip_ptr;
1788 	struct dn_dev __rcu     *dn_ptr;
1789 	struct inet6_dev __rcu	*ip6_ptr;
1790 	void			*ax25_ptr;
1791 	struct wireless_dev	*ieee80211_ptr;
1792 	struct wpan_dev		*ieee802154_ptr;
1793 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1794 	struct mpls_dev __rcu	*mpls_ptr;
1795 #endif
1796 
1797 /*
1798  * Cache lines mostly used on receive path (including eth_type_trans())
1799  */
1800 	/* Interface address info used in eth_type_trans() */
1801 	unsigned char		*dev_addr;
1802 
1803 #ifdef CONFIG_SYSFS
1804 	struct netdev_rx_queue	*_rx;
1805 
1806 	unsigned int		num_rx_queues;
1807 	unsigned int		real_num_rx_queues;
1808 #endif
1809 
1810 	struct bpf_prog __rcu	*xdp_prog;
1811 	unsigned long		gro_flush_timeout;
1812 	rx_handler_func_t __rcu	*rx_handler;
1813 	void __rcu		*rx_handler_data;
1814 
1815 #ifdef CONFIG_NET_CLS_ACT
1816 	struct mini_Qdisc __rcu	*miniq_ingress;
1817 #endif
1818 	struct netdev_queue __rcu *ingress_queue;
1819 #ifdef CONFIG_NETFILTER_INGRESS
1820 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1821 #endif
1822 
1823 	unsigned char		broadcast[MAX_ADDR_LEN];
1824 #ifdef CONFIG_RFS_ACCEL
1825 	struct cpu_rmap		*rx_cpu_rmap;
1826 #endif
1827 	struct hlist_node	index_hlist;
1828 
1829 /*
1830  * Cache lines mostly used on transmit path
1831  */
1832 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1833 	unsigned int		num_tx_queues;
1834 	unsigned int		real_num_tx_queues;
1835 	struct Qdisc		*qdisc;
1836 #ifdef CONFIG_NET_SCHED
1837 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1838 #endif
1839 	unsigned int		tx_queue_len;
1840 	spinlock_t		tx_global_lock;
1841 	int			watchdog_timeo;
1842 
1843 #ifdef CONFIG_XPS
1844 	struct xps_dev_maps __rcu *xps_maps;
1845 #endif
1846 #ifdef CONFIG_NET_CLS_ACT
1847 	struct mini_Qdisc __rcu	*miniq_egress;
1848 #endif
1849 
1850 	/* These may be needed for future network-power-down code. */
1851 	struct timer_list	watchdog_timer;
1852 
1853 	int __percpu		*pcpu_refcnt;
1854 	struct list_head	todo_list;
1855 
1856 	struct list_head	link_watch_list;
1857 
1858 	enum { NETREG_UNINITIALIZED=0,
1859 	       NETREG_REGISTERED,	/* completed register_netdevice */
1860 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1861 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1862 	       NETREG_RELEASED,		/* called free_netdev */
1863 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1864 	} reg_state:8;
1865 
1866 	bool dismantle;
1867 
1868 	enum {
1869 		RTNL_LINK_INITIALIZED,
1870 		RTNL_LINK_INITIALIZING,
1871 	} rtnl_link_state:16;
1872 
1873 	bool needs_free_netdev;
1874 	void (*priv_destructor)(struct net_device *dev);
1875 
1876 #ifdef CONFIG_NETPOLL
1877 	struct netpoll_info __rcu	*npinfo;
1878 #endif
1879 
1880 	possible_net_t			nd_net;
1881 
1882 	/* mid-layer private */
1883 	union {
1884 		void					*ml_priv;
1885 		struct pcpu_lstats __percpu		*lstats;
1886 		struct pcpu_sw_netstats __percpu	*tstats;
1887 		struct pcpu_dstats __percpu		*dstats;
1888 		struct pcpu_vstats __percpu		*vstats;
1889 	};
1890 
1891 #if IS_ENABLED(CONFIG_GARP)
1892 	struct garp_port __rcu	*garp_port;
1893 #endif
1894 #if IS_ENABLED(CONFIG_MRP)
1895 	struct mrp_port __rcu	*mrp_port;
1896 #endif
1897 
1898 	struct device		dev;
1899 	const struct attribute_group *sysfs_groups[4];
1900 	const struct attribute_group *sysfs_rx_queue_group;
1901 
1902 	const struct rtnl_link_ops *rtnl_link_ops;
1903 
1904 	/* for setting kernel sock attribute on TCP connection setup */
1905 #define GSO_MAX_SIZE		65536
1906 	unsigned int		gso_max_size;
1907 #define GSO_MAX_SEGS		65535
1908 	u16			gso_max_segs;
1909 
1910 #ifdef CONFIG_DCB
1911 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1912 #endif
1913 	u8			num_tc;
1914 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1915 	u8			prio_tc_map[TC_BITMASK + 1];
1916 
1917 #if IS_ENABLED(CONFIG_FCOE)
1918 	unsigned int		fcoe_ddp_xid;
1919 #endif
1920 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1921 	struct netprio_map __rcu *priomap;
1922 #endif
1923 	struct phy_device	*phydev;
1924 	struct lock_class_key	*qdisc_tx_busylock;
1925 	struct lock_class_key	*qdisc_running_key;
1926 	bool			proto_down;
1927 };
1928 #define to_net_dev(d) container_of(d, struct net_device, dev)
1929 
1930 static inline bool netif_elide_gro(const struct net_device *dev)
1931 {
1932 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1933 		return true;
1934 	return false;
1935 }
1936 
1937 #define	NETDEV_ALIGN		32
1938 
1939 static inline
1940 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1941 {
1942 	return dev->prio_tc_map[prio & TC_BITMASK];
1943 }
1944 
1945 static inline
1946 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1947 {
1948 	if (tc >= dev->num_tc)
1949 		return -EINVAL;
1950 
1951 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1952 	return 0;
1953 }
1954 
1955 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1956 void netdev_reset_tc(struct net_device *dev);
1957 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1958 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1959 
1960 static inline
1961 int netdev_get_num_tc(struct net_device *dev)
1962 {
1963 	return dev->num_tc;
1964 }
1965 
1966 static inline
1967 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1968 					 unsigned int index)
1969 {
1970 	return &dev->_tx[index];
1971 }
1972 
1973 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1974 						    const struct sk_buff *skb)
1975 {
1976 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1977 }
1978 
1979 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1980 					    void (*f)(struct net_device *,
1981 						      struct netdev_queue *,
1982 						      void *),
1983 					    void *arg)
1984 {
1985 	unsigned int i;
1986 
1987 	for (i = 0; i < dev->num_tx_queues; i++)
1988 		f(dev, &dev->_tx[i], arg);
1989 }
1990 
1991 #define netdev_lockdep_set_classes(dev)				\
1992 {								\
1993 	static struct lock_class_key qdisc_tx_busylock_key;	\
1994 	static struct lock_class_key qdisc_running_key;		\
1995 	static struct lock_class_key qdisc_xmit_lock_key;	\
1996 	static struct lock_class_key dev_addr_list_lock_key;	\
1997 	unsigned int i;						\
1998 								\
1999 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2000 	(dev)->qdisc_running_key = &qdisc_running_key;		\
2001 	lockdep_set_class(&(dev)->addr_list_lock,		\
2002 			  &dev_addr_list_lock_key); 		\
2003 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2004 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2005 				  &qdisc_xmit_lock_key);	\
2006 }
2007 
2008 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2009 				    struct sk_buff *skb,
2010 				    void *accel_priv);
2011 
2012 /* returns the headroom that the master device needs to take in account
2013  * when forwarding to this dev
2014  */
2015 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2016 {
2017 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2018 }
2019 
2020 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2021 {
2022 	if (dev->netdev_ops->ndo_set_rx_headroom)
2023 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2024 }
2025 
2026 /* set the device rx headroom to the dev's default */
2027 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2028 {
2029 	netdev_set_rx_headroom(dev, -1);
2030 }
2031 
2032 /*
2033  * Net namespace inlines
2034  */
2035 static inline
2036 struct net *dev_net(const struct net_device *dev)
2037 {
2038 	return read_pnet(&dev->nd_net);
2039 }
2040 
2041 static inline
2042 void dev_net_set(struct net_device *dev, struct net *net)
2043 {
2044 	write_pnet(&dev->nd_net, net);
2045 }
2046 
2047 /**
2048  *	netdev_priv - access network device private data
2049  *	@dev: network device
2050  *
2051  * Get network device private data
2052  */
2053 static inline void *netdev_priv(const struct net_device *dev)
2054 {
2055 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2056 }
2057 
2058 /* Set the sysfs physical device reference for the network logical device
2059  * if set prior to registration will cause a symlink during initialization.
2060  */
2061 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2062 
2063 /* Set the sysfs device type for the network logical device to allow
2064  * fine-grained identification of different network device types. For
2065  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2066  */
2067 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2068 
2069 /* Default NAPI poll() weight
2070  * Device drivers are strongly advised to not use bigger value
2071  */
2072 #define NAPI_POLL_WEIGHT 64
2073 
2074 /**
2075  *	netif_napi_add - initialize a NAPI context
2076  *	@dev:  network device
2077  *	@napi: NAPI context
2078  *	@poll: polling function
2079  *	@weight: default weight
2080  *
2081  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2082  * *any* of the other NAPI-related functions.
2083  */
2084 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2085 		    int (*poll)(struct napi_struct *, int), int weight);
2086 
2087 /**
2088  *	netif_tx_napi_add - initialize a NAPI context
2089  *	@dev:  network device
2090  *	@napi: NAPI context
2091  *	@poll: polling function
2092  *	@weight: default weight
2093  *
2094  * This variant of netif_napi_add() should be used from drivers using NAPI
2095  * to exclusively poll a TX queue.
2096  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2097  */
2098 static inline void netif_tx_napi_add(struct net_device *dev,
2099 				     struct napi_struct *napi,
2100 				     int (*poll)(struct napi_struct *, int),
2101 				     int weight)
2102 {
2103 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2104 	netif_napi_add(dev, napi, poll, weight);
2105 }
2106 
2107 /**
2108  *  netif_napi_del - remove a NAPI context
2109  *  @napi: NAPI context
2110  *
2111  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2112  */
2113 void netif_napi_del(struct napi_struct *napi);
2114 
2115 struct napi_gro_cb {
2116 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2117 	void	*frag0;
2118 
2119 	/* Length of frag0. */
2120 	unsigned int frag0_len;
2121 
2122 	/* This indicates where we are processing relative to skb->data. */
2123 	int	data_offset;
2124 
2125 	/* This is non-zero if the packet cannot be merged with the new skb. */
2126 	u16	flush;
2127 
2128 	/* Save the IP ID here and check when we get to the transport layer */
2129 	u16	flush_id;
2130 
2131 	/* Number of segments aggregated. */
2132 	u16	count;
2133 
2134 	/* Start offset for remote checksum offload */
2135 	u16	gro_remcsum_start;
2136 
2137 	/* jiffies when first packet was created/queued */
2138 	unsigned long age;
2139 
2140 	/* Used in ipv6_gro_receive() and foo-over-udp */
2141 	u16	proto;
2142 
2143 	/* This is non-zero if the packet may be of the same flow. */
2144 	u8	same_flow:1;
2145 
2146 	/* Used in tunnel GRO receive */
2147 	u8	encap_mark:1;
2148 
2149 	/* GRO checksum is valid */
2150 	u8	csum_valid:1;
2151 
2152 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2153 	u8	csum_cnt:3;
2154 
2155 	/* Free the skb? */
2156 	u8	free:2;
2157 #define NAPI_GRO_FREE		  1
2158 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2159 
2160 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2161 	u8	is_ipv6:1;
2162 
2163 	/* Used in GRE, set in fou/gue_gro_receive */
2164 	u8	is_fou:1;
2165 
2166 	/* Used to determine if flush_id can be ignored */
2167 	u8	is_atomic:1;
2168 
2169 	/* Number of gro_receive callbacks this packet already went through */
2170 	u8 recursion_counter:4;
2171 
2172 	/* 1 bit hole */
2173 
2174 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2175 	__wsum	csum;
2176 
2177 	/* used in skb_gro_receive() slow path */
2178 	struct sk_buff *last;
2179 };
2180 
2181 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2182 
2183 #define GRO_RECURSION_LIMIT 15
2184 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2185 {
2186 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2187 }
2188 
2189 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2190 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2191 						struct sk_buff **head,
2192 						struct sk_buff *skb)
2193 {
2194 	if (unlikely(gro_recursion_inc_test(skb))) {
2195 		NAPI_GRO_CB(skb)->flush |= 1;
2196 		return NULL;
2197 	}
2198 
2199 	return cb(head, skb);
2200 }
2201 
2202 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2203 					     struct sk_buff *);
2204 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2205 						   struct sock *sk,
2206 						   struct sk_buff **head,
2207 						   struct sk_buff *skb)
2208 {
2209 	if (unlikely(gro_recursion_inc_test(skb))) {
2210 		NAPI_GRO_CB(skb)->flush |= 1;
2211 		return NULL;
2212 	}
2213 
2214 	return cb(sk, head, skb);
2215 }
2216 
2217 struct packet_type {
2218 	__be16			type;	/* This is really htons(ether_type). */
2219 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2220 	int			(*func) (struct sk_buff *,
2221 					 struct net_device *,
2222 					 struct packet_type *,
2223 					 struct net_device *);
2224 	bool			(*id_match)(struct packet_type *ptype,
2225 					    struct sock *sk);
2226 	void			*af_packet_priv;
2227 	struct list_head	list;
2228 };
2229 
2230 struct offload_callbacks {
2231 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2232 						netdev_features_t features);
2233 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2234 						 struct sk_buff *skb);
2235 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2236 };
2237 
2238 struct packet_offload {
2239 	__be16			 type;	/* This is really htons(ether_type). */
2240 	u16			 priority;
2241 	struct offload_callbacks callbacks;
2242 	struct list_head	 list;
2243 };
2244 
2245 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2246 struct pcpu_sw_netstats {
2247 	u64     rx_packets;
2248 	u64     rx_bytes;
2249 	u64     tx_packets;
2250 	u64     tx_bytes;
2251 	struct u64_stats_sync   syncp;
2252 };
2253 
2254 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2255 ({									\
2256 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2257 	if (pcpu_stats)	{						\
2258 		int __cpu;						\
2259 		for_each_possible_cpu(__cpu) {				\
2260 			typeof(type) *stat;				\
2261 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2262 			u64_stats_init(&stat->syncp);			\
2263 		}							\
2264 	}								\
2265 	pcpu_stats;							\
2266 })
2267 
2268 #define netdev_alloc_pcpu_stats(type)					\
2269 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2270 
2271 enum netdev_lag_tx_type {
2272 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2273 	NETDEV_LAG_TX_TYPE_RANDOM,
2274 	NETDEV_LAG_TX_TYPE_BROADCAST,
2275 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2276 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2277 	NETDEV_LAG_TX_TYPE_HASH,
2278 };
2279 
2280 struct netdev_lag_upper_info {
2281 	enum netdev_lag_tx_type tx_type;
2282 };
2283 
2284 struct netdev_lag_lower_state_info {
2285 	u8 link_up : 1,
2286 	   tx_enabled : 1;
2287 };
2288 
2289 #include <linux/notifier.h>
2290 
2291 /* netdevice notifier chain. Please remember to update the rtnetlink
2292  * notification exclusion list in rtnetlink_event() when adding new
2293  * types.
2294  */
2295 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2296 #define NETDEV_DOWN	0x0002
2297 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2298 				   detected a hardware crash and restarted
2299 				   - we can use this eg to kick tcp sessions
2300 				   once done */
2301 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2302 #define NETDEV_REGISTER 0x0005
2303 #define NETDEV_UNREGISTER	0x0006
2304 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2305 #define NETDEV_CHANGEADDR	0x0008
2306 #define NETDEV_GOING_DOWN	0x0009
2307 #define NETDEV_CHANGENAME	0x000A
2308 #define NETDEV_FEAT_CHANGE	0x000B
2309 #define NETDEV_BONDING_FAILOVER 0x000C
2310 #define NETDEV_PRE_UP		0x000D
2311 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2312 #define NETDEV_POST_TYPE_CHANGE	0x000F
2313 #define NETDEV_POST_INIT	0x0010
2314 #define NETDEV_UNREGISTER_FINAL 0x0011
2315 #define NETDEV_RELEASE		0x0012
2316 #define NETDEV_NOTIFY_PEERS	0x0013
2317 #define NETDEV_JOIN		0x0014
2318 #define NETDEV_CHANGEUPPER	0x0015
2319 #define NETDEV_RESEND_IGMP	0x0016
2320 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2321 #define NETDEV_CHANGEINFODATA	0x0018
2322 #define NETDEV_BONDING_INFO	0x0019
2323 #define NETDEV_PRECHANGEUPPER	0x001A
2324 #define NETDEV_CHANGELOWERSTATE	0x001B
2325 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2326 #define NETDEV_UDP_TUNNEL_DROP_INFO	0x001D
2327 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2328 
2329 int register_netdevice_notifier(struct notifier_block *nb);
2330 int unregister_netdevice_notifier(struct notifier_block *nb);
2331 
2332 struct netdev_notifier_info {
2333 	struct net_device	*dev;
2334 	struct netlink_ext_ack	*extack;
2335 };
2336 
2337 struct netdev_notifier_change_info {
2338 	struct netdev_notifier_info info; /* must be first */
2339 	unsigned int flags_changed;
2340 };
2341 
2342 struct netdev_notifier_changeupper_info {
2343 	struct netdev_notifier_info info; /* must be first */
2344 	struct net_device *upper_dev; /* new upper dev */
2345 	bool master; /* is upper dev master */
2346 	bool linking; /* is the notification for link or unlink */
2347 	void *upper_info; /* upper dev info */
2348 };
2349 
2350 struct netdev_notifier_changelowerstate_info {
2351 	struct netdev_notifier_info info; /* must be first */
2352 	void *lower_state_info; /* is lower dev state */
2353 };
2354 
2355 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2356 					     struct net_device *dev)
2357 {
2358 	info->dev = dev;
2359 	info->extack = NULL;
2360 }
2361 
2362 static inline struct net_device *
2363 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2364 {
2365 	return info->dev;
2366 }
2367 
2368 static inline struct netlink_ext_ack *
2369 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2370 {
2371 	return info->extack;
2372 }
2373 
2374 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2375 
2376 
2377 extern rwlock_t				dev_base_lock;		/* Device list lock */
2378 
2379 #define for_each_netdev(net, d)		\
2380 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2381 #define for_each_netdev_reverse(net, d)	\
2382 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2383 #define for_each_netdev_rcu(net, d)		\
2384 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2385 #define for_each_netdev_safe(net, d, n)	\
2386 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2387 #define for_each_netdev_continue(net, d)		\
2388 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2389 #define for_each_netdev_continue_rcu(net, d)		\
2390 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2391 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2392 		for_each_netdev_rcu(&init_net, slave)	\
2393 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2394 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2395 
2396 static inline struct net_device *next_net_device(struct net_device *dev)
2397 {
2398 	struct list_head *lh;
2399 	struct net *net;
2400 
2401 	net = dev_net(dev);
2402 	lh = dev->dev_list.next;
2403 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2404 }
2405 
2406 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2407 {
2408 	struct list_head *lh;
2409 	struct net *net;
2410 
2411 	net = dev_net(dev);
2412 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2413 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2414 }
2415 
2416 static inline struct net_device *first_net_device(struct net *net)
2417 {
2418 	return list_empty(&net->dev_base_head) ? NULL :
2419 		net_device_entry(net->dev_base_head.next);
2420 }
2421 
2422 static inline struct net_device *first_net_device_rcu(struct net *net)
2423 {
2424 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2425 
2426 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2427 }
2428 
2429 int netdev_boot_setup_check(struct net_device *dev);
2430 unsigned long netdev_boot_base(const char *prefix, int unit);
2431 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2432 				       const char *hwaddr);
2433 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2434 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2435 void dev_add_pack(struct packet_type *pt);
2436 void dev_remove_pack(struct packet_type *pt);
2437 void __dev_remove_pack(struct packet_type *pt);
2438 void dev_add_offload(struct packet_offload *po);
2439 void dev_remove_offload(struct packet_offload *po);
2440 
2441 int dev_get_iflink(const struct net_device *dev);
2442 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2443 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2444 				      unsigned short mask);
2445 struct net_device *dev_get_by_name(struct net *net, const char *name);
2446 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2447 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2448 int dev_alloc_name(struct net_device *dev, const char *name);
2449 int dev_open(struct net_device *dev);
2450 void dev_close(struct net_device *dev);
2451 void dev_close_many(struct list_head *head, bool unlink);
2452 void dev_disable_lro(struct net_device *dev);
2453 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2454 int dev_queue_xmit(struct sk_buff *skb);
2455 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2456 int register_netdevice(struct net_device *dev);
2457 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2458 void unregister_netdevice_many(struct list_head *head);
2459 static inline void unregister_netdevice(struct net_device *dev)
2460 {
2461 	unregister_netdevice_queue(dev, NULL);
2462 }
2463 
2464 int netdev_refcnt_read(const struct net_device *dev);
2465 void free_netdev(struct net_device *dev);
2466 void netdev_freemem(struct net_device *dev);
2467 void synchronize_net(void);
2468 int init_dummy_netdev(struct net_device *dev);
2469 
2470 DECLARE_PER_CPU(int, xmit_recursion);
2471 #define XMIT_RECURSION_LIMIT	10
2472 
2473 static inline int dev_recursion_level(void)
2474 {
2475 	return this_cpu_read(xmit_recursion);
2476 }
2477 
2478 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2479 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2480 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2481 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2482 int netdev_get_name(struct net *net, char *name, int ifindex);
2483 int dev_restart(struct net_device *dev);
2484 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2485 
2486 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2487 {
2488 	return NAPI_GRO_CB(skb)->data_offset;
2489 }
2490 
2491 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2492 {
2493 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2494 }
2495 
2496 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2497 {
2498 	NAPI_GRO_CB(skb)->data_offset += len;
2499 }
2500 
2501 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2502 					unsigned int offset)
2503 {
2504 	return NAPI_GRO_CB(skb)->frag0 + offset;
2505 }
2506 
2507 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2508 {
2509 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2510 }
2511 
2512 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2513 {
2514 	NAPI_GRO_CB(skb)->frag0 = NULL;
2515 	NAPI_GRO_CB(skb)->frag0_len = 0;
2516 }
2517 
2518 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2519 					unsigned int offset)
2520 {
2521 	if (!pskb_may_pull(skb, hlen))
2522 		return NULL;
2523 
2524 	skb_gro_frag0_invalidate(skb);
2525 	return skb->data + offset;
2526 }
2527 
2528 static inline void *skb_gro_network_header(struct sk_buff *skb)
2529 {
2530 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2531 	       skb_network_offset(skb);
2532 }
2533 
2534 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2535 					const void *start, unsigned int len)
2536 {
2537 	if (NAPI_GRO_CB(skb)->csum_valid)
2538 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2539 						  csum_partial(start, len, 0));
2540 }
2541 
2542 /* GRO checksum functions. These are logical equivalents of the normal
2543  * checksum functions (in skbuff.h) except that they operate on the GRO
2544  * offsets and fields in sk_buff.
2545  */
2546 
2547 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2548 
2549 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2550 {
2551 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2552 }
2553 
2554 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2555 						      bool zero_okay,
2556 						      __sum16 check)
2557 {
2558 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2559 		skb_checksum_start_offset(skb) <
2560 		 skb_gro_offset(skb)) &&
2561 		!skb_at_gro_remcsum_start(skb) &&
2562 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2563 		(!zero_okay || check));
2564 }
2565 
2566 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2567 							   __wsum psum)
2568 {
2569 	if (NAPI_GRO_CB(skb)->csum_valid &&
2570 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2571 		return 0;
2572 
2573 	NAPI_GRO_CB(skb)->csum = psum;
2574 
2575 	return __skb_gro_checksum_complete(skb);
2576 }
2577 
2578 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2579 {
2580 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2581 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2582 		NAPI_GRO_CB(skb)->csum_cnt--;
2583 	} else {
2584 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2585 		 * verified a new top level checksum or an encapsulated one
2586 		 * during GRO. This saves work if we fallback to normal path.
2587 		 */
2588 		__skb_incr_checksum_unnecessary(skb);
2589 	}
2590 }
2591 
2592 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2593 				    compute_pseudo)			\
2594 ({									\
2595 	__sum16 __ret = 0;						\
2596 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2597 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2598 				compute_pseudo(skb, proto));		\
2599 	if (!__ret)							\
2600 		skb_gro_incr_csum_unnecessary(skb);			\
2601 	__ret;								\
2602 })
2603 
2604 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2605 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2606 
2607 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2608 					     compute_pseudo)		\
2609 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2610 
2611 #define skb_gro_checksum_simple_validate(skb)				\
2612 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2613 
2614 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2615 {
2616 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2617 		!NAPI_GRO_CB(skb)->csum_valid);
2618 }
2619 
2620 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2621 					      __sum16 check, __wsum pseudo)
2622 {
2623 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2624 	NAPI_GRO_CB(skb)->csum_valid = 1;
2625 }
2626 
2627 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2628 do {									\
2629 	if (__skb_gro_checksum_convert_check(skb))			\
2630 		__skb_gro_checksum_convert(skb, check,			\
2631 					   compute_pseudo(skb, proto));	\
2632 } while (0)
2633 
2634 struct gro_remcsum {
2635 	int offset;
2636 	__wsum delta;
2637 };
2638 
2639 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2640 {
2641 	grc->offset = 0;
2642 	grc->delta = 0;
2643 }
2644 
2645 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2646 					    unsigned int off, size_t hdrlen,
2647 					    int start, int offset,
2648 					    struct gro_remcsum *grc,
2649 					    bool nopartial)
2650 {
2651 	__wsum delta;
2652 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2653 
2654 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2655 
2656 	if (!nopartial) {
2657 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2658 		return ptr;
2659 	}
2660 
2661 	ptr = skb_gro_header_fast(skb, off);
2662 	if (skb_gro_header_hard(skb, off + plen)) {
2663 		ptr = skb_gro_header_slow(skb, off + plen, off);
2664 		if (!ptr)
2665 			return NULL;
2666 	}
2667 
2668 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2669 			       start, offset);
2670 
2671 	/* Adjust skb->csum since we changed the packet */
2672 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2673 
2674 	grc->offset = off + hdrlen + offset;
2675 	grc->delta = delta;
2676 
2677 	return ptr;
2678 }
2679 
2680 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2681 					   struct gro_remcsum *grc)
2682 {
2683 	void *ptr;
2684 	size_t plen = grc->offset + sizeof(u16);
2685 
2686 	if (!grc->delta)
2687 		return;
2688 
2689 	ptr = skb_gro_header_fast(skb, grc->offset);
2690 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2691 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2692 		if (!ptr)
2693 			return;
2694 	}
2695 
2696 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2697 }
2698 
2699 #ifdef CONFIG_XFRM_OFFLOAD
2700 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2701 {
2702 	if (PTR_ERR(pp) != -EINPROGRESS)
2703 		NAPI_GRO_CB(skb)->flush |= flush;
2704 }
2705 #else
2706 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2707 {
2708 	NAPI_GRO_CB(skb)->flush |= flush;
2709 }
2710 #endif
2711 
2712 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2713 				  unsigned short type,
2714 				  const void *daddr, const void *saddr,
2715 				  unsigned int len)
2716 {
2717 	if (!dev->header_ops || !dev->header_ops->create)
2718 		return 0;
2719 
2720 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2721 }
2722 
2723 static inline int dev_parse_header(const struct sk_buff *skb,
2724 				   unsigned char *haddr)
2725 {
2726 	const struct net_device *dev = skb->dev;
2727 
2728 	if (!dev->header_ops || !dev->header_ops->parse)
2729 		return 0;
2730 	return dev->header_ops->parse(skb, haddr);
2731 }
2732 
2733 /* ll_header must have at least hard_header_len allocated */
2734 static inline bool dev_validate_header(const struct net_device *dev,
2735 				       char *ll_header, int len)
2736 {
2737 	if (likely(len >= dev->hard_header_len))
2738 		return true;
2739 	if (len < dev->min_header_len)
2740 		return false;
2741 
2742 	if (capable(CAP_SYS_RAWIO)) {
2743 		memset(ll_header + len, 0, dev->hard_header_len - len);
2744 		return true;
2745 	}
2746 
2747 	if (dev->header_ops && dev->header_ops->validate)
2748 		return dev->header_ops->validate(ll_header, len);
2749 
2750 	return false;
2751 }
2752 
2753 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2754 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2755 static inline int unregister_gifconf(unsigned int family)
2756 {
2757 	return register_gifconf(family, NULL);
2758 }
2759 
2760 #ifdef CONFIG_NET_FLOW_LIMIT
2761 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2762 struct sd_flow_limit {
2763 	u64			count;
2764 	unsigned int		num_buckets;
2765 	unsigned int		history_head;
2766 	u16			history[FLOW_LIMIT_HISTORY];
2767 	u8			buckets[];
2768 };
2769 
2770 extern int netdev_flow_limit_table_len;
2771 #endif /* CONFIG_NET_FLOW_LIMIT */
2772 
2773 /*
2774  * Incoming packets are placed on per-CPU queues
2775  */
2776 struct softnet_data {
2777 	struct list_head	poll_list;
2778 	struct sk_buff_head	process_queue;
2779 
2780 	/* stats */
2781 	unsigned int		processed;
2782 	unsigned int		time_squeeze;
2783 	unsigned int		received_rps;
2784 #ifdef CONFIG_RPS
2785 	struct softnet_data	*rps_ipi_list;
2786 #endif
2787 #ifdef CONFIG_NET_FLOW_LIMIT
2788 	struct sd_flow_limit __rcu *flow_limit;
2789 #endif
2790 	struct Qdisc		*output_queue;
2791 	struct Qdisc		**output_queue_tailp;
2792 	struct sk_buff		*completion_queue;
2793 
2794 #ifdef CONFIG_RPS
2795 	/* input_queue_head should be written by cpu owning this struct,
2796 	 * and only read by other cpus. Worth using a cache line.
2797 	 */
2798 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2799 
2800 	/* Elements below can be accessed between CPUs for RPS/RFS */
2801 	call_single_data_t	csd ____cacheline_aligned_in_smp;
2802 	struct softnet_data	*rps_ipi_next;
2803 	unsigned int		cpu;
2804 	unsigned int		input_queue_tail;
2805 #endif
2806 	unsigned int		dropped;
2807 	struct sk_buff_head	input_pkt_queue;
2808 	struct napi_struct	backlog;
2809 
2810 };
2811 
2812 static inline void input_queue_head_incr(struct softnet_data *sd)
2813 {
2814 #ifdef CONFIG_RPS
2815 	sd->input_queue_head++;
2816 #endif
2817 }
2818 
2819 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2820 					      unsigned int *qtail)
2821 {
2822 #ifdef CONFIG_RPS
2823 	*qtail = ++sd->input_queue_tail;
2824 #endif
2825 }
2826 
2827 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2828 
2829 void __netif_schedule(struct Qdisc *q);
2830 void netif_schedule_queue(struct netdev_queue *txq);
2831 
2832 static inline void netif_tx_schedule_all(struct net_device *dev)
2833 {
2834 	unsigned int i;
2835 
2836 	for (i = 0; i < dev->num_tx_queues; i++)
2837 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2838 }
2839 
2840 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2841 {
2842 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2843 }
2844 
2845 /**
2846  *	netif_start_queue - allow transmit
2847  *	@dev: network device
2848  *
2849  *	Allow upper layers to call the device hard_start_xmit routine.
2850  */
2851 static inline void netif_start_queue(struct net_device *dev)
2852 {
2853 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2854 }
2855 
2856 static inline void netif_tx_start_all_queues(struct net_device *dev)
2857 {
2858 	unsigned int i;
2859 
2860 	for (i = 0; i < dev->num_tx_queues; i++) {
2861 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2862 		netif_tx_start_queue(txq);
2863 	}
2864 }
2865 
2866 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2867 
2868 /**
2869  *	netif_wake_queue - restart transmit
2870  *	@dev: network device
2871  *
2872  *	Allow upper layers to call the device hard_start_xmit routine.
2873  *	Used for flow control when transmit resources are available.
2874  */
2875 static inline void netif_wake_queue(struct net_device *dev)
2876 {
2877 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2878 }
2879 
2880 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2881 {
2882 	unsigned int i;
2883 
2884 	for (i = 0; i < dev->num_tx_queues; i++) {
2885 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2886 		netif_tx_wake_queue(txq);
2887 	}
2888 }
2889 
2890 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2891 {
2892 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2893 }
2894 
2895 /**
2896  *	netif_stop_queue - stop transmitted packets
2897  *	@dev: network device
2898  *
2899  *	Stop upper layers calling the device hard_start_xmit routine.
2900  *	Used for flow control when transmit resources are unavailable.
2901  */
2902 static inline void netif_stop_queue(struct net_device *dev)
2903 {
2904 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2905 }
2906 
2907 void netif_tx_stop_all_queues(struct net_device *dev);
2908 
2909 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2910 {
2911 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2912 }
2913 
2914 /**
2915  *	netif_queue_stopped - test if transmit queue is flowblocked
2916  *	@dev: network device
2917  *
2918  *	Test if transmit queue on device is currently unable to send.
2919  */
2920 static inline bool netif_queue_stopped(const struct net_device *dev)
2921 {
2922 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2923 }
2924 
2925 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2926 {
2927 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2928 }
2929 
2930 static inline bool
2931 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2932 {
2933 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2934 }
2935 
2936 static inline bool
2937 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2938 {
2939 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2940 }
2941 
2942 /**
2943  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2944  *	@dev_queue: pointer to transmit queue
2945  *
2946  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2947  * to give appropriate hint to the CPU.
2948  */
2949 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2950 {
2951 #ifdef CONFIG_BQL
2952 	prefetchw(&dev_queue->dql.num_queued);
2953 #endif
2954 }
2955 
2956 /**
2957  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2958  *	@dev_queue: pointer to transmit queue
2959  *
2960  * BQL enabled drivers might use this helper in their TX completion path,
2961  * to give appropriate hint to the CPU.
2962  */
2963 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2964 {
2965 #ifdef CONFIG_BQL
2966 	prefetchw(&dev_queue->dql.limit);
2967 #endif
2968 }
2969 
2970 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2971 					unsigned int bytes)
2972 {
2973 #ifdef CONFIG_BQL
2974 	dql_queued(&dev_queue->dql, bytes);
2975 
2976 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2977 		return;
2978 
2979 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2980 
2981 	/*
2982 	 * The XOFF flag must be set before checking the dql_avail below,
2983 	 * because in netdev_tx_completed_queue we update the dql_completed
2984 	 * before checking the XOFF flag.
2985 	 */
2986 	smp_mb();
2987 
2988 	/* check again in case another CPU has just made room avail */
2989 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2990 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2991 #endif
2992 }
2993 
2994 /**
2995  * 	netdev_sent_queue - report the number of bytes queued to hardware
2996  * 	@dev: network device
2997  * 	@bytes: number of bytes queued to the hardware device queue
2998  *
2999  * 	Report the number of bytes queued for sending/completion to the network
3000  * 	device hardware queue. @bytes should be a good approximation and should
3001  * 	exactly match netdev_completed_queue() @bytes
3002  */
3003 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3004 {
3005 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3006 }
3007 
3008 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3009 					     unsigned int pkts, unsigned int bytes)
3010 {
3011 #ifdef CONFIG_BQL
3012 	if (unlikely(!bytes))
3013 		return;
3014 
3015 	dql_completed(&dev_queue->dql, bytes);
3016 
3017 	/*
3018 	 * Without the memory barrier there is a small possiblity that
3019 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3020 	 * be stopped forever
3021 	 */
3022 	smp_mb();
3023 
3024 	if (dql_avail(&dev_queue->dql) < 0)
3025 		return;
3026 
3027 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3028 		netif_schedule_queue(dev_queue);
3029 #endif
3030 }
3031 
3032 /**
3033  * 	netdev_completed_queue - report bytes and packets completed by device
3034  * 	@dev: network device
3035  * 	@pkts: actual number of packets sent over the medium
3036  * 	@bytes: actual number of bytes sent over the medium
3037  *
3038  * 	Report the number of bytes and packets transmitted by the network device
3039  * 	hardware queue over the physical medium, @bytes must exactly match the
3040  * 	@bytes amount passed to netdev_sent_queue()
3041  */
3042 static inline void netdev_completed_queue(struct net_device *dev,
3043 					  unsigned int pkts, unsigned int bytes)
3044 {
3045 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3046 }
3047 
3048 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3049 {
3050 #ifdef CONFIG_BQL
3051 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3052 	dql_reset(&q->dql);
3053 #endif
3054 }
3055 
3056 /**
3057  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3058  * 	@dev_queue: network device
3059  *
3060  * 	Reset the bytes and packet count of a network device and clear the
3061  * 	software flow control OFF bit for this network device
3062  */
3063 static inline void netdev_reset_queue(struct net_device *dev_queue)
3064 {
3065 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3066 }
3067 
3068 /**
3069  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3070  * 	@dev: network device
3071  * 	@queue_index: given tx queue index
3072  *
3073  * 	Returns 0 if given tx queue index >= number of device tx queues,
3074  * 	otherwise returns the originally passed tx queue index.
3075  */
3076 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3077 {
3078 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3079 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3080 				     dev->name, queue_index,
3081 				     dev->real_num_tx_queues);
3082 		return 0;
3083 	}
3084 
3085 	return queue_index;
3086 }
3087 
3088 /**
3089  *	netif_running - test if up
3090  *	@dev: network device
3091  *
3092  *	Test if the device has been brought up.
3093  */
3094 static inline bool netif_running(const struct net_device *dev)
3095 {
3096 	return test_bit(__LINK_STATE_START, &dev->state);
3097 }
3098 
3099 /*
3100  * Routines to manage the subqueues on a device.  We only need start,
3101  * stop, and a check if it's stopped.  All other device management is
3102  * done at the overall netdevice level.
3103  * Also test the device if we're multiqueue.
3104  */
3105 
3106 /**
3107  *	netif_start_subqueue - allow sending packets on subqueue
3108  *	@dev: network device
3109  *	@queue_index: sub queue index
3110  *
3111  * Start individual transmit queue of a device with multiple transmit queues.
3112  */
3113 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3114 {
3115 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3116 
3117 	netif_tx_start_queue(txq);
3118 }
3119 
3120 /**
3121  *	netif_stop_subqueue - stop sending packets on subqueue
3122  *	@dev: network device
3123  *	@queue_index: sub queue index
3124  *
3125  * Stop individual transmit queue of a device with multiple transmit queues.
3126  */
3127 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3128 {
3129 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3130 	netif_tx_stop_queue(txq);
3131 }
3132 
3133 /**
3134  *	netif_subqueue_stopped - test status of subqueue
3135  *	@dev: network device
3136  *	@queue_index: sub queue index
3137  *
3138  * Check individual transmit queue of a device with multiple transmit queues.
3139  */
3140 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3141 					    u16 queue_index)
3142 {
3143 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3144 
3145 	return netif_tx_queue_stopped(txq);
3146 }
3147 
3148 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3149 					  struct sk_buff *skb)
3150 {
3151 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3152 }
3153 
3154 /**
3155  *	netif_wake_subqueue - allow sending packets on subqueue
3156  *	@dev: network device
3157  *	@queue_index: sub queue index
3158  *
3159  * Resume individual transmit queue of a device with multiple transmit queues.
3160  */
3161 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3162 {
3163 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3164 
3165 	netif_tx_wake_queue(txq);
3166 }
3167 
3168 #ifdef CONFIG_XPS
3169 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3170 			u16 index);
3171 #else
3172 static inline int netif_set_xps_queue(struct net_device *dev,
3173 				      const struct cpumask *mask,
3174 				      u16 index)
3175 {
3176 	return 0;
3177 }
3178 #endif
3179 
3180 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3181 		  unsigned int num_tx_queues);
3182 
3183 /*
3184  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3185  * as a distribution range limit for the returned value.
3186  */
3187 static inline u16 skb_tx_hash(const struct net_device *dev,
3188 			      struct sk_buff *skb)
3189 {
3190 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3191 }
3192 
3193 /**
3194  *	netif_is_multiqueue - test if device has multiple transmit queues
3195  *	@dev: network device
3196  *
3197  * Check if device has multiple transmit queues
3198  */
3199 static inline bool netif_is_multiqueue(const struct net_device *dev)
3200 {
3201 	return dev->num_tx_queues > 1;
3202 }
3203 
3204 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3205 
3206 #ifdef CONFIG_SYSFS
3207 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3208 #else
3209 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3210 						unsigned int rxq)
3211 {
3212 	return 0;
3213 }
3214 #endif
3215 
3216 #ifdef CONFIG_SYSFS
3217 static inline unsigned int get_netdev_rx_queue_index(
3218 		struct netdev_rx_queue *queue)
3219 {
3220 	struct net_device *dev = queue->dev;
3221 	int index = queue - dev->_rx;
3222 
3223 	BUG_ON(index >= dev->num_rx_queues);
3224 	return index;
3225 }
3226 #endif
3227 
3228 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3229 int netif_get_num_default_rss_queues(void);
3230 
3231 enum skb_free_reason {
3232 	SKB_REASON_CONSUMED,
3233 	SKB_REASON_DROPPED,
3234 };
3235 
3236 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3237 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3238 
3239 /*
3240  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3241  * interrupt context or with hardware interrupts being disabled.
3242  * (in_irq() || irqs_disabled())
3243  *
3244  * We provide four helpers that can be used in following contexts :
3245  *
3246  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3247  *  replacing kfree_skb(skb)
3248  *
3249  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3250  *  Typically used in place of consume_skb(skb) in TX completion path
3251  *
3252  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3253  *  replacing kfree_skb(skb)
3254  *
3255  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3256  *  and consumed a packet. Used in place of consume_skb(skb)
3257  */
3258 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3259 {
3260 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3261 }
3262 
3263 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3264 {
3265 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3266 }
3267 
3268 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3269 {
3270 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3271 }
3272 
3273 static inline void dev_consume_skb_any(struct sk_buff *skb)
3274 {
3275 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3276 }
3277 
3278 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3279 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3280 int netif_rx(struct sk_buff *skb);
3281 int netif_rx_ni(struct sk_buff *skb);
3282 int netif_receive_skb(struct sk_buff *skb);
3283 int netif_receive_skb_core(struct sk_buff *skb);
3284 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3285 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3286 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3287 gro_result_t napi_gro_frags(struct napi_struct *napi);
3288 struct packet_offload *gro_find_receive_by_type(__be16 type);
3289 struct packet_offload *gro_find_complete_by_type(__be16 type);
3290 
3291 static inline void napi_free_frags(struct napi_struct *napi)
3292 {
3293 	kfree_skb(napi->skb);
3294 	napi->skb = NULL;
3295 }
3296 
3297 bool netdev_is_rx_handler_busy(struct net_device *dev);
3298 int netdev_rx_handler_register(struct net_device *dev,
3299 			       rx_handler_func_t *rx_handler,
3300 			       void *rx_handler_data);
3301 void netdev_rx_handler_unregister(struct net_device *dev);
3302 
3303 bool dev_valid_name(const char *name);
3304 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3305 int dev_ethtool(struct net *net, struct ifreq *);
3306 unsigned int dev_get_flags(const struct net_device *);
3307 int __dev_change_flags(struct net_device *, unsigned int flags);
3308 int dev_change_flags(struct net_device *, unsigned int);
3309 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3310 			unsigned int gchanges);
3311 int dev_change_name(struct net_device *, const char *);
3312 int dev_set_alias(struct net_device *, const char *, size_t);
3313 int dev_get_alias(const struct net_device *, char *, size_t);
3314 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3315 int __dev_set_mtu(struct net_device *, int);
3316 int dev_set_mtu(struct net_device *, int);
3317 void dev_set_group(struct net_device *, int);
3318 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3319 int dev_change_carrier(struct net_device *, bool new_carrier);
3320 int dev_get_phys_port_id(struct net_device *dev,
3321 			 struct netdev_phys_item_id *ppid);
3322 int dev_get_phys_port_name(struct net_device *dev,
3323 			   char *name, size_t len);
3324 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3325 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3326 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3327 				    struct netdev_queue *txq, int *ret);
3328 
3329 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3330 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3331 		      int fd, u32 flags);
3332 u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t xdp_op, u32 *prog_id);
3333 
3334 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3335 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3336 bool is_skb_forwardable(const struct net_device *dev,
3337 			const struct sk_buff *skb);
3338 
3339 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3340 					       struct sk_buff *skb)
3341 {
3342 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3343 	    unlikely(!is_skb_forwardable(dev, skb))) {
3344 		atomic_long_inc(&dev->rx_dropped);
3345 		kfree_skb(skb);
3346 		return NET_RX_DROP;
3347 	}
3348 
3349 	skb_scrub_packet(skb, true);
3350 	skb->priority = 0;
3351 	return 0;
3352 }
3353 
3354 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3355 
3356 extern int		netdev_budget;
3357 extern unsigned int	netdev_budget_usecs;
3358 
3359 /* Called by rtnetlink.c:rtnl_unlock() */
3360 void netdev_run_todo(void);
3361 
3362 /**
3363  *	dev_put - release reference to device
3364  *	@dev: network device
3365  *
3366  * Release reference to device to allow it to be freed.
3367  */
3368 static inline void dev_put(struct net_device *dev)
3369 {
3370 	this_cpu_dec(*dev->pcpu_refcnt);
3371 }
3372 
3373 /**
3374  *	dev_hold - get reference to device
3375  *	@dev: network device
3376  *
3377  * Hold reference to device to keep it from being freed.
3378  */
3379 static inline void dev_hold(struct net_device *dev)
3380 {
3381 	this_cpu_inc(*dev->pcpu_refcnt);
3382 }
3383 
3384 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3385  * and _off may be called from IRQ context, but it is caller
3386  * who is responsible for serialization of these calls.
3387  *
3388  * The name carrier is inappropriate, these functions should really be
3389  * called netif_lowerlayer_*() because they represent the state of any
3390  * kind of lower layer not just hardware media.
3391  */
3392 
3393 void linkwatch_init_dev(struct net_device *dev);
3394 void linkwatch_fire_event(struct net_device *dev);
3395 void linkwatch_forget_dev(struct net_device *dev);
3396 
3397 /**
3398  *	netif_carrier_ok - test if carrier present
3399  *	@dev: network device
3400  *
3401  * Check if carrier is present on device
3402  */
3403 static inline bool netif_carrier_ok(const struct net_device *dev)
3404 {
3405 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3406 }
3407 
3408 unsigned long dev_trans_start(struct net_device *dev);
3409 
3410 void __netdev_watchdog_up(struct net_device *dev);
3411 
3412 void netif_carrier_on(struct net_device *dev);
3413 
3414 void netif_carrier_off(struct net_device *dev);
3415 
3416 /**
3417  *	netif_dormant_on - mark device as dormant.
3418  *	@dev: network device
3419  *
3420  * Mark device as dormant (as per RFC2863).
3421  *
3422  * The dormant state indicates that the relevant interface is not
3423  * actually in a condition to pass packets (i.e., it is not 'up') but is
3424  * in a "pending" state, waiting for some external event.  For "on-
3425  * demand" interfaces, this new state identifies the situation where the
3426  * interface is waiting for events to place it in the up state.
3427  */
3428 static inline void netif_dormant_on(struct net_device *dev)
3429 {
3430 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3431 		linkwatch_fire_event(dev);
3432 }
3433 
3434 /**
3435  *	netif_dormant_off - set device as not dormant.
3436  *	@dev: network device
3437  *
3438  * Device is not in dormant state.
3439  */
3440 static inline void netif_dormant_off(struct net_device *dev)
3441 {
3442 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3443 		linkwatch_fire_event(dev);
3444 }
3445 
3446 /**
3447  *	netif_dormant - test if device is dormant
3448  *	@dev: network device
3449  *
3450  * Check if device is dormant.
3451  */
3452 static inline bool netif_dormant(const struct net_device *dev)
3453 {
3454 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3455 }
3456 
3457 
3458 /**
3459  *	netif_oper_up - test if device is operational
3460  *	@dev: network device
3461  *
3462  * Check if carrier is operational
3463  */
3464 static inline bool netif_oper_up(const struct net_device *dev)
3465 {
3466 	return (dev->operstate == IF_OPER_UP ||
3467 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3468 }
3469 
3470 /**
3471  *	netif_device_present - is device available or removed
3472  *	@dev: network device
3473  *
3474  * Check if device has not been removed from system.
3475  */
3476 static inline bool netif_device_present(struct net_device *dev)
3477 {
3478 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3479 }
3480 
3481 void netif_device_detach(struct net_device *dev);
3482 
3483 void netif_device_attach(struct net_device *dev);
3484 
3485 /*
3486  * Network interface message level settings
3487  */
3488 
3489 enum {
3490 	NETIF_MSG_DRV		= 0x0001,
3491 	NETIF_MSG_PROBE		= 0x0002,
3492 	NETIF_MSG_LINK		= 0x0004,
3493 	NETIF_MSG_TIMER		= 0x0008,
3494 	NETIF_MSG_IFDOWN	= 0x0010,
3495 	NETIF_MSG_IFUP		= 0x0020,
3496 	NETIF_MSG_RX_ERR	= 0x0040,
3497 	NETIF_MSG_TX_ERR	= 0x0080,
3498 	NETIF_MSG_TX_QUEUED	= 0x0100,
3499 	NETIF_MSG_INTR		= 0x0200,
3500 	NETIF_MSG_TX_DONE	= 0x0400,
3501 	NETIF_MSG_RX_STATUS	= 0x0800,
3502 	NETIF_MSG_PKTDATA	= 0x1000,
3503 	NETIF_MSG_HW		= 0x2000,
3504 	NETIF_MSG_WOL		= 0x4000,
3505 };
3506 
3507 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3508 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3509 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3510 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3511 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3512 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3513 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3514 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3515 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3516 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3517 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3518 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3519 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3520 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3521 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3522 
3523 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3524 {
3525 	/* use default */
3526 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3527 		return default_msg_enable_bits;
3528 	if (debug_value == 0)	/* no output */
3529 		return 0;
3530 	/* set low N bits */
3531 	return (1 << debug_value) - 1;
3532 }
3533 
3534 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3535 {
3536 	spin_lock(&txq->_xmit_lock);
3537 	txq->xmit_lock_owner = cpu;
3538 }
3539 
3540 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3541 {
3542 	__acquire(&txq->_xmit_lock);
3543 	return true;
3544 }
3545 
3546 static inline void __netif_tx_release(struct netdev_queue *txq)
3547 {
3548 	__release(&txq->_xmit_lock);
3549 }
3550 
3551 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3552 {
3553 	spin_lock_bh(&txq->_xmit_lock);
3554 	txq->xmit_lock_owner = smp_processor_id();
3555 }
3556 
3557 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3558 {
3559 	bool ok = spin_trylock(&txq->_xmit_lock);
3560 	if (likely(ok))
3561 		txq->xmit_lock_owner = smp_processor_id();
3562 	return ok;
3563 }
3564 
3565 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3566 {
3567 	txq->xmit_lock_owner = -1;
3568 	spin_unlock(&txq->_xmit_lock);
3569 }
3570 
3571 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3572 {
3573 	txq->xmit_lock_owner = -1;
3574 	spin_unlock_bh(&txq->_xmit_lock);
3575 }
3576 
3577 static inline void txq_trans_update(struct netdev_queue *txq)
3578 {
3579 	if (txq->xmit_lock_owner != -1)
3580 		txq->trans_start = jiffies;
3581 }
3582 
3583 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3584 static inline void netif_trans_update(struct net_device *dev)
3585 {
3586 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3587 
3588 	if (txq->trans_start != jiffies)
3589 		txq->trans_start = jiffies;
3590 }
3591 
3592 /**
3593  *	netif_tx_lock - grab network device transmit lock
3594  *	@dev: network device
3595  *
3596  * Get network device transmit lock
3597  */
3598 static inline void netif_tx_lock(struct net_device *dev)
3599 {
3600 	unsigned int i;
3601 	int cpu;
3602 
3603 	spin_lock(&dev->tx_global_lock);
3604 	cpu = smp_processor_id();
3605 	for (i = 0; i < dev->num_tx_queues; i++) {
3606 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3607 
3608 		/* We are the only thread of execution doing a
3609 		 * freeze, but we have to grab the _xmit_lock in
3610 		 * order to synchronize with threads which are in
3611 		 * the ->hard_start_xmit() handler and already
3612 		 * checked the frozen bit.
3613 		 */
3614 		__netif_tx_lock(txq, cpu);
3615 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3616 		__netif_tx_unlock(txq);
3617 	}
3618 }
3619 
3620 static inline void netif_tx_lock_bh(struct net_device *dev)
3621 {
3622 	local_bh_disable();
3623 	netif_tx_lock(dev);
3624 }
3625 
3626 static inline void netif_tx_unlock(struct net_device *dev)
3627 {
3628 	unsigned int i;
3629 
3630 	for (i = 0; i < dev->num_tx_queues; i++) {
3631 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3632 
3633 		/* No need to grab the _xmit_lock here.  If the
3634 		 * queue is not stopped for another reason, we
3635 		 * force a schedule.
3636 		 */
3637 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3638 		netif_schedule_queue(txq);
3639 	}
3640 	spin_unlock(&dev->tx_global_lock);
3641 }
3642 
3643 static inline void netif_tx_unlock_bh(struct net_device *dev)
3644 {
3645 	netif_tx_unlock(dev);
3646 	local_bh_enable();
3647 }
3648 
3649 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3650 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3651 		__netif_tx_lock(txq, cpu);		\
3652 	} else {					\
3653 		__netif_tx_acquire(txq);		\
3654 	}						\
3655 }
3656 
3657 #define HARD_TX_TRYLOCK(dev, txq)			\
3658 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3659 		__netif_tx_trylock(txq) :		\
3660 		__netif_tx_acquire(txq))
3661 
3662 #define HARD_TX_UNLOCK(dev, txq) {			\
3663 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3664 		__netif_tx_unlock(txq);			\
3665 	} else {					\
3666 		__netif_tx_release(txq);		\
3667 	}						\
3668 }
3669 
3670 static inline void netif_tx_disable(struct net_device *dev)
3671 {
3672 	unsigned int i;
3673 	int cpu;
3674 
3675 	local_bh_disable();
3676 	cpu = smp_processor_id();
3677 	for (i = 0; i < dev->num_tx_queues; i++) {
3678 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3679 
3680 		__netif_tx_lock(txq, cpu);
3681 		netif_tx_stop_queue(txq);
3682 		__netif_tx_unlock(txq);
3683 	}
3684 	local_bh_enable();
3685 }
3686 
3687 static inline void netif_addr_lock(struct net_device *dev)
3688 {
3689 	spin_lock(&dev->addr_list_lock);
3690 }
3691 
3692 static inline void netif_addr_lock_nested(struct net_device *dev)
3693 {
3694 	int subclass = SINGLE_DEPTH_NESTING;
3695 
3696 	if (dev->netdev_ops->ndo_get_lock_subclass)
3697 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3698 
3699 	spin_lock_nested(&dev->addr_list_lock, subclass);
3700 }
3701 
3702 static inline void netif_addr_lock_bh(struct net_device *dev)
3703 {
3704 	spin_lock_bh(&dev->addr_list_lock);
3705 }
3706 
3707 static inline void netif_addr_unlock(struct net_device *dev)
3708 {
3709 	spin_unlock(&dev->addr_list_lock);
3710 }
3711 
3712 static inline void netif_addr_unlock_bh(struct net_device *dev)
3713 {
3714 	spin_unlock_bh(&dev->addr_list_lock);
3715 }
3716 
3717 /*
3718  * dev_addrs walker. Should be used only for read access. Call with
3719  * rcu_read_lock held.
3720  */
3721 #define for_each_dev_addr(dev, ha) \
3722 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3723 
3724 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3725 
3726 void ether_setup(struct net_device *dev);
3727 
3728 /* Support for loadable net-drivers */
3729 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3730 				    unsigned char name_assign_type,
3731 				    void (*setup)(struct net_device *),
3732 				    unsigned int txqs, unsigned int rxqs);
3733 int dev_get_valid_name(struct net *net, struct net_device *dev,
3734 		       const char *name);
3735 
3736 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3737 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3738 
3739 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3740 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3741 			 count)
3742 
3743 int register_netdev(struct net_device *dev);
3744 void unregister_netdev(struct net_device *dev);
3745 
3746 /* General hardware address lists handling functions */
3747 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3748 		   struct netdev_hw_addr_list *from_list, int addr_len);
3749 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3750 		      struct netdev_hw_addr_list *from_list, int addr_len);
3751 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3752 		       struct net_device *dev,
3753 		       int (*sync)(struct net_device *, const unsigned char *),
3754 		       int (*unsync)(struct net_device *,
3755 				     const unsigned char *));
3756 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3757 			  struct net_device *dev,
3758 			  int (*unsync)(struct net_device *,
3759 					const unsigned char *));
3760 void __hw_addr_init(struct netdev_hw_addr_list *list);
3761 
3762 /* Functions used for device addresses handling */
3763 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3764 		 unsigned char addr_type);
3765 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3766 		 unsigned char addr_type);
3767 void dev_addr_flush(struct net_device *dev);
3768 int dev_addr_init(struct net_device *dev);
3769 
3770 /* Functions used for unicast addresses handling */
3771 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3772 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3773 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3774 int dev_uc_sync(struct net_device *to, struct net_device *from);
3775 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3776 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3777 void dev_uc_flush(struct net_device *dev);
3778 void dev_uc_init(struct net_device *dev);
3779 
3780 /**
3781  *  __dev_uc_sync - Synchonize device's unicast list
3782  *  @dev:  device to sync
3783  *  @sync: function to call if address should be added
3784  *  @unsync: function to call if address should be removed
3785  *
3786  *  Add newly added addresses to the interface, and release
3787  *  addresses that have been deleted.
3788  */
3789 static inline int __dev_uc_sync(struct net_device *dev,
3790 				int (*sync)(struct net_device *,
3791 					    const unsigned char *),
3792 				int (*unsync)(struct net_device *,
3793 					      const unsigned char *))
3794 {
3795 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3796 }
3797 
3798 /**
3799  *  __dev_uc_unsync - Remove synchronized addresses from device
3800  *  @dev:  device to sync
3801  *  @unsync: function to call if address should be removed
3802  *
3803  *  Remove all addresses that were added to the device by dev_uc_sync().
3804  */
3805 static inline void __dev_uc_unsync(struct net_device *dev,
3806 				   int (*unsync)(struct net_device *,
3807 						 const unsigned char *))
3808 {
3809 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3810 }
3811 
3812 /* Functions used for multicast addresses handling */
3813 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3814 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3815 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3816 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3817 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3818 int dev_mc_sync(struct net_device *to, struct net_device *from);
3819 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3820 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3821 void dev_mc_flush(struct net_device *dev);
3822 void dev_mc_init(struct net_device *dev);
3823 
3824 /**
3825  *  __dev_mc_sync - Synchonize device's multicast list
3826  *  @dev:  device to sync
3827  *  @sync: function to call if address should be added
3828  *  @unsync: function to call if address should be removed
3829  *
3830  *  Add newly added addresses to the interface, and release
3831  *  addresses that have been deleted.
3832  */
3833 static inline int __dev_mc_sync(struct net_device *dev,
3834 				int (*sync)(struct net_device *,
3835 					    const unsigned char *),
3836 				int (*unsync)(struct net_device *,
3837 					      const unsigned char *))
3838 {
3839 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3840 }
3841 
3842 /**
3843  *  __dev_mc_unsync - Remove synchronized addresses from device
3844  *  @dev:  device to sync
3845  *  @unsync: function to call if address should be removed
3846  *
3847  *  Remove all addresses that were added to the device by dev_mc_sync().
3848  */
3849 static inline void __dev_mc_unsync(struct net_device *dev,
3850 				   int (*unsync)(struct net_device *,
3851 						 const unsigned char *))
3852 {
3853 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3854 }
3855 
3856 /* Functions used for secondary unicast and multicast support */
3857 void dev_set_rx_mode(struct net_device *dev);
3858 void __dev_set_rx_mode(struct net_device *dev);
3859 int dev_set_promiscuity(struct net_device *dev, int inc);
3860 int dev_set_allmulti(struct net_device *dev, int inc);
3861 void netdev_state_change(struct net_device *dev);
3862 void netdev_notify_peers(struct net_device *dev);
3863 void netdev_features_change(struct net_device *dev);
3864 /* Load a device via the kmod */
3865 void dev_load(struct net *net, const char *name);
3866 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3867 					struct rtnl_link_stats64 *storage);
3868 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3869 			     const struct net_device_stats *netdev_stats);
3870 
3871 extern int		netdev_max_backlog;
3872 extern int		netdev_tstamp_prequeue;
3873 extern int		weight_p;
3874 extern int		dev_weight_rx_bias;
3875 extern int		dev_weight_tx_bias;
3876 extern int		dev_rx_weight;
3877 extern int		dev_tx_weight;
3878 
3879 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3880 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3881 						     struct list_head **iter);
3882 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3883 						     struct list_head **iter);
3884 
3885 /* iterate through upper list, must be called under RCU read lock */
3886 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3887 	for (iter = &(dev)->adj_list.upper, \
3888 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3889 	     updev; \
3890 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3891 
3892 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3893 				  int (*fn)(struct net_device *upper_dev,
3894 					    void *data),
3895 				  void *data);
3896 
3897 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3898 				  struct net_device *upper_dev);
3899 
3900 bool netdev_has_any_upper_dev(struct net_device *dev);
3901 
3902 void *netdev_lower_get_next_private(struct net_device *dev,
3903 				    struct list_head **iter);
3904 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3905 					struct list_head **iter);
3906 
3907 #define netdev_for_each_lower_private(dev, priv, iter) \
3908 	for (iter = (dev)->adj_list.lower.next, \
3909 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3910 	     priv; \
3911 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3912 
3913 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3914 	for (iter = &(dev)->adj_list.lower, \
3915 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3916 	     priv; \
3917 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3918 
3919 void *netdev_lower_get_next(struct net_device *dev,
3920 				struct list_head **iter);
3921 
3922 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3923 	for (iter = (dev)->adj_list.lower.next, \
3924 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3925 	     ldev; \
3926 	     ldev = netdev_lower_get_next(dev, &(iter)))
3927 
3928 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3929 					     struct list_head **iter);
3930 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3931 						 struct list_head **iter);
3932 
3933 int netdev_walk_all_lower_dev(struct net_device *dev,
3934 			      int (*fn)(struct net_device *lower_dev,
3935 					void *data),
3936 			      void *data);
3937 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3938 				  int (*fn)(struct net_device *lower_dev,
3939 					    void *data),
3940 				  void *data);
3941 
3942 void *netdev_adjacent_get_private(struct list_head *adj_list);
3943 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3944 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3945 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3946 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3947 			  struct netlink_ext_ack *extack);
3948 int netdev_master_upper_dev_link(struct net_device *dev,
3949 				 struct net_device *upper_dev,
3950 				 void *upper_priv, void *upper_info,
3951 				 struct netlink_ext_ack *extack);
3952 void netdev_upper_dev_unlink(struct net_device *dev,
3953 			     struct net_device *upper_dev);
3954 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3955 void *netdev_lower_dev_get_private(struct net_device *dev,
3956 				   struct net_device *lower_dev);
3957 void netdev_lower_state_changed(struct net_device *lower_dev,
3958 				void *lower_state_info);
3959 
3960 /* RSS keys are 40 or 52 bytes long */
3961 #define NETDEV_RSS_KEY_LEN 52
3962 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3963 void netdev_rss_key_fill(void *buffer, size_t len);
3964 
3965 int dev_get_nest_level(struct net_device *dev);
3966 int skb_checksum_help(struct sk_buff *skb);
3967 int skb_crc32c_csum_help(struct sk_buff *skb);
3968 int skb_csum_hwoffload_help(struct sk_buff *skb,
3969 			    const netdev_features_t features);
3970 
3971 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3972 				  netdev_features_t features, bool tx_path);
3973 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3974 				    netdev_features_t features);
3975 
3976 struct netdev_bonding_info {
3977 	ifslave	slave;
3978 	ifbond	master;
3979 };
3980 
3981 struct netdev_notifier_bonding_info {
3982 	struct netdev_notifier_info info; /* must be first */
3983 	struct netdev_bonding_info  bonding_info;
3984 };
3985 
3986 void netdev_bonding_info_change(struct net_device *dev,
3987 				struct netdev_bonding_info *bonding_info);
3988 
3989 static inline
3990 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3991 {
3992 	return __skb_gso_segment(skb, features, true);
3993 }
3994 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3995 
3996 static inline bool can_checksum_protocol(netdev_features_t features,
3997 					 __be16 protocol)
3998 {
3999 	if (protocol == htons(ETH_P_FCOE))
4000 		return !!(features & NETIF_F_FCOE_CRC);
4001 
4002 	/* Assume this is an IP checksum (not SCTP CRC) */
4003 
4004 	if (features & NETIF_F_HW_CSUM) {
4005 		/* Can checksum everything */
4006 		return true;
4007 	}
4008 
4009 	switch (protocol) {
4010 	case htons(ETH_P_IP):
4011 		return !!(features & NETIF_F_IP_CSUM);
4012 	case htons(ETH_P_IPV6):
4013 		return !!(features & NETIF_F_IPV6_CSUM);
4014 	default:
4015 		return false;
4016 	}
4017 }
4018 
4019 #ifdef CONFIG_BUG
4020 void netdev_rx_csum_fault(struct net_device *dev);
4021 #else
4022 static inline void netdev_rx_csum_fault(struct net_device *dev)
4023 {
4024 }
4025 #endif
4026 /* rx skb timestamps */
4027 void net_enable_timestamp(void);
4028 void net_disable_timestamp(void);
4029 
4030 #ifdef CONFIG_PROC_FS
4031 int __init dev_proc_init(void);
4032 #else
4033 #define dev_proc_init() 0
4034 #endif
4035 
4036 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4037 					      struct sk_buff *skb, struct net_device *dev,
4038 					      bool more)
4039 {
4040 	skb->xmit_more = more ? 1 : 0;
4041 	return ops->ndo_start_xmit(skb, dev);
4042 }
4043 
4044 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4045 					    struct netdev_queue *txq, bool more)
4046 {
4047 	const struct net_device_ops *ops = dev->netdev_ops;
4048 	int rc;
4049 
4050 	rc = __netdev_start_xmit(ops, skb, dev, more);
4051 	if (rc == NETDEV_TX_OK)
4052 		txq_trans_update(txq);
4053 
4054 	return rc;
4055 }
4056 
4057 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4058 				const void *ns);
4059 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4060 				 const void *ns);
4061 
4062 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4063 {
4064 	return netdev_class_create_file_ns(class_attr, NULL);
4065 }
4066 
4067 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4068 {
4069 	netdev_class_remove_file_ns(class_attr, NULL);
4070 }
4071 
4072 extern const struct kobj_ns_type_operations net_ns_type_operations;
4073 
4074 const char *netdev_drivername(const struct net_device *dev);
4075 
4076 void linkwatch_run_queue(void);
4077 
4078 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4079 							  netdev_features_t f2)
4080 {
4081 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4082 		if (f1 & NETIF_F_HW_CSUM)
4083 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4084 		else
4085 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4086 	}
4087 
4088 	return f1 & f2;
4089 }
4090 
4091 static inline netdev_features_t netdev_get_wanted_features(
4092 	struct net_device *dev)
4093 {
4094 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4095 }
4096 netdev_features_t netdev_increment_features(netdev_features_t all,
4097 	netdev_features_t one, netdev_features_t mask);
4098 
4099 /* Allow TSO being used on stacked device :
4100  * Performing the GSO segmentation before last device
4101  * is a performance improvement.
4102  */
4103 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4104 							netdev_features_t mask)
4105 {
4106 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4107 }
4108 
4109 int __netdev_update_features(struct net_device *dev);
4110 void netdev_update_features(struct net_device *dev);
4111 void netdev_change_features(struct net_device *dev);
4112 
4113 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4114 					struct net_device *dev);
4115 
4116 netdev_features_t passthru_features_check(struct sk_buff *skb,
4117 					  struct net_device *dev,
4118 					  netdev_features_t features);
4119 netdev_features_t netif_skb_features(struct sk_buff *skb);
4120 
4121 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4122 {
4123 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4124 
4125 	/* check flags correspondence */
4126 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4127 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4128 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4129 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4130 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4131 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4132 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4133 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4134 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4135 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4136 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4137 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4138 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4139 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4140 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4141 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4142 
4143 	return (features & feature) == feature;
4144 }
4145 
4146 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4147 {
4148 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4149 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4150 }
4151 
4152 static inline bool netif_needs_gso(struct sk_buff *skb,
4153 				   netdev_features_t features)
4154 {
4155 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4156 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4157 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4158 }
4159 
4160 static inline void netif_set_gso_max_size(struct net_device *dev,
4161 					  unsigned int size)
4162 {
4163 	dev->gso_max_size = size;
4164 }
4165 
4166 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4167 					int pulled_hlen, u16 mac_offset,
4168 					int mac_len)
4169 {
4170 	skb->protocol = protocol;
4171 	skb->encapsulation = 1;
4172 	skb_push(skb, pulled_hlen);
4173 	skb_reset_transport_header(skb);
4174 	skb->mac_header = mac_offset;
4175 	skb->network_header = skb->mac_header + mac_len;
4176 	skb->mac_len = mac_len;
4177 }
4178 
4179 static inline bool netif_is_macsec(const struct net_device *dev)
4180 {
4181 	return dev->priv_flags & IFF_MACSEC;
4182 }
4183 
4184 static inline bool netif_is_macvlan(const struct net_device *dev)
4185 {
4186 	return dev->priv_flags & IFF_MACVLAN;
4187 }
4188 
4189 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4190 {
4191 	return dev->priv_flags & IFF_MACVLAN_PORT;
4192 }
4193 
4194 static inline bool netif_is_ipvlan(const struct net_device *dev)
4195 {
4196 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4197 }
4198 
4199 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4200 {
4201 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4202 }
4203 
4204 static inline bool netif_is_bond_master(const struct net_device *dev)
4205 {
4206 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4207 }
4208 
4209 static inline bool netif_is_bond_slave(const struct net_device *dev)
4210 {
4211 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4212 }
4213 
4214 static inline bool netif_supports_nofcs(struct net_device *dev)
4215 {
4216 	return dev->priv_flags & IFF_SUPP_NOFCS;
4217 }
4218 
4219 static inline bool netif_is_l3_master(const struct net_device *dev)
4220 {
4221 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4222 }
4223 
4224 static inline bool netif_is_l3_slave(const struct net_device *dev)
4225 {
4226 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4227 }
4228 
4229 static inline bool netif_is_bridge_master(const struct net_device *dev)
4230 {
4231 	return dev->priv_flags & IFF_EBRIDGE;
4232 }
4233 
4234 static inline bool netif_is_bridge_port(const struct net_device *dev)
4235 {
4236 	return dev->priv_flags & IFF_BRIDGE_PORT;
4237 }
4238 
4239 static inline bool netif_is_ovs_master(const struct net_device *dev)
4240 {
4241 	return dev->priv_flags & IFF_OPENVSWITCH;
4242 }
4243 
4244 static inline bool netif_is_ovs_port(const struct net_device *dev)
4245 {
4246 	return dev->priv_flags & IFF_OVS_DATAPATH;
4247 }
4248 
4249 static inline bool netif_is_team_master(const struct net_device *dev)
4250 {
4251 	return dev->priv_flags & IFF_TEAM;
4252 }
4253 
4254 static inline bool netif_is_team_port(const struct net_device *dev)
4255 {
4256 	return dev->priv_flags & IFF_TEAM_PORT;
4257 }
4258 
4259 static inline bool netif_is_lag_master(const struct net_device *dev)
4260 {
4261 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4262 }
4263 
4264 static inline bool netif_is_lag_port(const struct net_device *dev)
4265 {
4266 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4267 }
4268 
4269 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4270 {
4271 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4272 }
4273 
4274 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4275 static inline void netif_keep_dst(struct net_device *dev)
4276 {
4277 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4278 }
4279 
4280 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4281 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4282 {
4283 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4284 	return dev->priv_flags & IFF_MACSEC;
4285 }
4286 
4287 extern struct pernet_operations __net_initdata loopback_net_ops;
4288 
4289 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4290 
4291 /* netdev_printk helpers, similar to dev_printk */
4292 
4293 static inline const char *netdev_name(const struct net_device *dev)
4294 {
4295 	if (!dev->name[0] || strchr(dev->name, '%'))
4296 		return "(unnamed net_device)";
4297 	return dev->name;
4298 }
4299 
4300 static inline bool netdev_unregistering(const struct net_device *dev)
4301 {
4302 	return dev->reg_state == NETREG_UNREGISTERING;
4303 }
4304 
4305 static inline const char *netdev_reg_state(const struct net_device *dev)
4306 {
4307 	switch (dev->reg_state) {
4308 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4309 	case NETREG_REGISTERED: return "";
4310 	case NETREG_UNREGISTERING: return " (unregistering)";
4311 	case NETREG_UNREGISTERED: return " (unregistered)";
4312 	case NETREG_RELEASED: return " (released)";
4313 	case NETREG_DUMMY: return " (dummy)";
4314 	}
4315 
4316 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4317 	return " (unknown)";
4318 }
4319 
4320 __printf(3, 4)
4321 void netdev_printk(const char *level, const struct net_device *dev,
4322 		   const char *format, ...);
4323 __printf(2, 3)
4324 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4325 __printf(2, 3)
4326 void netdev_alert(const struct net_device *dev, const char *format, ...);
4327 __printf(2, 3)
4328 void netdev_crit(const struct net_device *dev, const char *format, ...);
4329 __printf(2, 3)
4330 void netdev_err(const struct net_device *dev, const char *format, ...);
4331 __printf(2, 3)
4332 void netdev_warn(const struct net_device *dev, const char *format, ...);
4333 __printf(2, 3)
4334 void netdev_notice(const struct net_device *dev, const char *format, ...);
4335 __printf(2, 3)
4336 void netdev_info(const struct net_device *dev, const char *format, ...);
4337 
4338 #define MODULE_ALIAS_NETDEV(device) \
4339 	MODULE_ALIAS("netdev-" device)
4340 
4341 #if defined(CONFIG_DYNAMIC_DEBUG)
4342 #define netdev_dbg(__dev, format, args...)			\
4343 do {								\
4344 	dynamic_netdev_dbg(__dev, format, ##args);		\
4345 } while (0)
4346 #elif defined(DEBUG)
4347 #define netdev_dbg(__dev, format, args...)			\
4348 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4349 #else
4350 #define netdev_dbg(__dev, format, args...)			\
4351 ({								\
4352 	if (0)							\
4353 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4354 })
4355 #endif
4356 
4357 #if defined(VERBOSE_DEBUG)
4358 #define netdev_vdbg	netdev_dbg
4359 #else
4360 
4361 #define netdev_vdbg(dev, format, args...)			\
4362 ({								\
4363 	if (0)							\
4364 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4365 	0;							\
4366 })
4367 #endif
4368 
4369 /*
4370  * netdev_WARN() acts like dev_printk(), but with the key difference
4371  * of using a WARN/WARN_ON to get the message out, including the
4372  * file/line information and a backtrace.
4373  */
4374 #define netdev_WARN(dev, format, args...)			\
4375 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4376 	     netdev_reg_state(dev), ##args)
4377 
4378 /* netif printk helpers, similar to netdev_printk */
4379 
4380 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4381 do {					  			\
4382 	if (netif_msg_##type(priv))				\
4383 		netdev_printk(level, (dev), fmt, ##args);	\
4384 } while (0)
4385 
4386 #define netif_level(level, priv, type, dev, fmt, args...)	\
4387 do {								\
4388 	if (netif_msg_##type(priv))				\
4389 		netdev_##level(dev, fmt, ##args);		\
4390 } while (0)
4391 
4392 #define netif_emerg(priv, type, dev, fmt, args...)		\
4393 	netif_level(emerg, priv, type, dev, fmt, ##args)
4394 #define netif_alert(priv, type, dev, fmt, args...)		\
4395 	netif_level(alert, priv, type, dev, fmt, ##args)
4396 #define netif_crit(priv, type, dev, fmt, args...)		\
4397 	netif_level(crit, priv, type, dev, fmt, ##args)
4398 #define netif_err(priv, type, dev, fmt, args...)		\
4399 	netif_level(err, priv, type, dev, fmt, ##args)
4400 #define netif_warn(priv, type, dev, fmt, args...)		\
4401 	netif_level(warn, priv, type, dev, fmt, ##args)
4402 #define netif_notice(priv, type, dev, fmt, args...)		\
4403 	netif_level(notice, priv, type, dev, fmt, ##args)
4404 #define netif_info(priv, type, dev, fmt, args...)		\
4405 	netif_level(info, priv, type, dev, fmt, ##args)
4406 
4407 #if defined(CONFIG_DYNAMIC_DEBUG)
4408 #define netif_dbg(priv, type, netdev, format, args...)		\
4409 do {								\
4410 	if (netif_msg_##type(priv))				\
4411 		dynamic_netdev_dbg(netdev, format, ##args);	\
4412 } while (0)
4413 #elif defined(DEBUG)
4414 #define netif_dbg(priv, type, dev, format, args...)		\
4415 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4416 #else
4417 #define netif_dbg(priv, type, dev, format, args...)			\
4418 ({									\
4419 	if (0)								\
4420 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4421 	0;								\
4422 })
4423 #endif
4424 
4425 /* if @cond then downgrade to debug, else print at @level */
4426 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4427 	do {                                                              \
4428 		if (cond)                                                 \
4429 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4430 		else                                                      \
4431 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4432 	} while (0)
4433 
4434 #if defined(VERBOSE_DEBUG)
4435 #define netif_vdbg	netif_dbg
4436 #else
4437 #define netif_vdbg(priv, type, dev, format, args...)		\
4438 ({								\
4439 	if (0)							\
4440 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4441 	0;							\
4442 })
4443 #endif
4444 
4445 /*
4446  *	The list of packet types we will receive (as opposed to discard)
4447  *	and the routines to invoke.
4448  *
4449  *	Why 16. Because with 16 the only overlap we get on a hash of the
4450  *	low nibble of the protocol value is RARP/SNAP/X.25.
4451  *
4452  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4453  *             sure which should go first, but I bet it won't make much
4454  *             difference if we are running VLANs.  The good news is that
4455  *             this protocol won't be in the list unless compiled in, so
4456  *             the average user (w/out VLANs) will not be adversely affected.
4457  *             --BLG
4458  *
4459  *		0800	IP
4460  *		8100    802.1Q VLAN
4461  *		0001	802.3
4462  *		0002	AX.25
4463  *		0004	802.2
4464  *		8035	RARP
4465  *		0005	SNAP
4466  *		0805	X.25
4467  *		0806	ARP
4468  *		8137	IPX
4469  *		0009	Localtalk
4470  *		86DD	IPv6
4471  */
4472 #define PTYPE_HASH_SIZE	(16)
4473 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4474 
4475 #endif	/* _LINUX_NETDEVICE_H */
4476