xref: /linux-6.15/include/linux/netdevice.h (revision 22ac5ad4)
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 xdp state that are being passed
783  * to the netdevice through the xdp op.
784  */
785 enum xdp_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 };
801 
802 struct netlink_ext_ack;
803 
804 struct netdev_xdp {
805 	enum xdp_netdev_command command;
806 	union {
807 		/* XDP_SETUP_PROG */
808 		struct {
809 			u32 flags;
810 			struct bpf_prog *prog;
811 			struct netlink_ext_ack *extack;
812 		};
813 		/* XDP_QUERY_PROG */
814 		struct {
815 			u8 prog_attached;
816 			u32 prog_id;
817 		};
818 	};
819 };
820 
821 #ifdef CONFIG_XFRM_OFFLOAD
822 struct xfrmdev_ops {
823 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
824 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
825 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
826 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
827 				       struct xfrm_state *x);
828 };
829 #endif
830 
831 struct dev_ifalias {
832 	struct rcu_head rcuhead;
833 	char ifalias[];
834 };
835 
836 /*
837  * This structure defines the management hooks for network devices.
838  * The following hooks can be defined; unless noted otherwise, they are
839  * optional and can be filled with a null pointer.
840  *
841  * int (*ndo_init)(struct net_device *dev);
842  *     This function is called once when a network device is registered.
843  *     The network device can use this for any late stage initialization
844  *     or semantic validation. It can fail with an error code which will
845  *     be propagated back to register_netdev.
846  *
847  * void (*ndo_uninit)(struct net_device *dev);
848  *     This function is called when device is unregistered or when registration
849  *     fails. It is not called if init fails.
850  *
851  * int (*ndo_open)(struct net_device *dev);
852  *     This function is called when a network device transitions to the up
853  *     state.
854  *
855  * int (*ndo_stop)(struct net_device *dev);
856  *     This function is called when a network device transitions to the down
857  *     state.
858  *
859  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860  *                               struct net_device *dev);
861  *	Called when a packet needs to be transmitted.
862  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
863  *	the queue before that can happen; it's for obsolete devices and weird
864  *	corner cases, but the stack really does a non-trivial amount
865  *	of useless work if you return NETDEV_TX_BUSY.
866  *	Required; cannot be NULL.
867  *
868  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869  *					   struct net_device *dev
870  *					   netdev_features_t features);
871  *	Called by core transmit path to determine if device is capable of
872  *	performing offload operations on a given packet. This is to give
873  *	the device an opportunity to implement any restrictions that cannot
874  *	be otherwise expressed by feature flags. The check is called with
875  *	the set of features that the stack has calculated and it returns
876  *	those the driver believes to be appropriate.
877  *
878  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
879  *                         void *accel_priv, select_queue_fallback_t fallback);
880  *	Called to decide which queue to use when device supports multiple
881  *	transmit queues.
882  *
883  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884  *	This function is called to allow device receiver to make
885  *	changes to configuration when multicast or promiscuous is enabled.
886  *
887  * void (*ndo_set_rx_mode)(struct net_device *dev);
888  *	This function is called device changes address list filtering.
889  *	If driver handles unicast address filtering, it should set
890  *	IFF_UNICAST_FLT in its priv_flags.
891  *
892  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893  *	This function  is called when the Media Access Control address
894  *	needs to be changed. If this interface is not defined, the
895  *	MAC address can not be changed.
896  *
897  * int (*ndo_validate_addr)(struct net_device *dev);
898  *	Test if Media Access Control address is valid for the device.
899  *
900  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
901  *	Called when a user requests an ioctl which can't be handled by
902  *	the generic interface code. If not defined ioctls return
903  *	not supported error code.
904  *
905  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906  *	Used to set network devices bus interface parameters. This interface
907  *	is retained for legacy reasons; new devices should use the bus
908  *	interface (PCI) for low level management.
909  *
910  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911  *	Called when a user wants to change the Maximum Transfer Unit
912  *	of a device.
913  *
914  * void (*ndo_tx_timeout)(struct net_device *dev);
915  *	Callback used when the transmitter has not made any progress
916  *	for dev->watchdog ticks.
917  *
918  * void (*ndo_get_stats64)(struct net_device *dev,
919  *                         struct rtnl_link_stats64 *storage);
920  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
921  *	Called when a user wants to get the network device usage
922  *	statistics. Drivers must do one of the following:
923  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
924  *	   rtnl_link_stats64 structure passed by the caller.
925  *	2. Define @ndo_get_stats to update a net_device_stats structure
926  *	   (which should normally be dev->stats) and return a pointer to
927  *	   it. The structure may be changed asynchronously only if each
928  *	   field is written atomically.
929  *	3. Update dev->stats asynchronously and atomically, and define
930  *	   neither operation.
931  *
932  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
933  *	Return true if this device supports offload stats of this attr_id.
934  *
935  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
936  *	void *attr_data)
937  *	Get statistics for offload operations by attr_id. Write it into the
938  *	attr_data pointer.
939  *
940  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
941  *	If device supports VLAN filtering this function is called when a
942  *	VLAN id is registered.
943  *
944  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
945  *	If device supports VLAN filtering this function is called when a
946  *	VLAN id is unregistered.
947  *
948  * void (*ndo_poll_controller)(struct net_device *dev);
949  *
950  *	SR-IOV management functions.
951  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
952  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
953  *			  u8 qos, __be16 proto);
954  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
955  *			  int max_tx_rate);
956  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
957  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
958  * int (*ndo_get_vf_config)(struct net_device *dev,
959  *			    int vf, struct ifla_vf_info *ivf);
960  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
961  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
962  *			  struct nlattr *port[]);
963  *
964  *      Enable or disable the VF ability to query its RSS Redirection Table and
965  *      Hash Key. This is needed since on some devices VF share this information
966  *      with PF and querying it may introduce a theoretical security risk.
967  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
968  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
969  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
970  *		       void *type_data);
971  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
972  *	This is always called from the stack with the rtnl lock held and netif
973  *	tx queues stopped. This allows the netdevice to perform queue
974  *	management safely.
975  *
976  *	Fiber Channel over Ethernet (FCoE) offload functions.
977  * int (*ndo_fcoe_enable)(struct net_device *dev);
978  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
979  *	so the underlying device can perform whatever needed configuration or
980  *	initialization to support acceleration of FCoE traffic.
981  *
982  * int (*ndo_fcoe_disable)(struct net_device *dev);
983  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
984  *	so the underlying device can perform whatever needed clean-ups to
985  *	stop supporting acceleration of FCoE traffic.
986  *
987  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
988  *			     struct scatterlist *sgl, unsigned int sgc);
989  *	Called when the FCoE Initiator wants to initialize an I/O that
990  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
991  *	perform necessary setup and returns 1 to indicate the device is set up
992  *	successfully to perform DDP on this I/O, otherwise this returns 0.
993  *
994  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
995  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
996  *	indicated by the FC exchange id 'xid', so the underlying device can
997  *	clean up and reuse resources for later DDP requests.
998  *
999  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1000  *			      struct scatterlist *sgl, unsigned int sgc);
1001  *	Called when the FCoE Target wants to initialize an I/O that
1002  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1003  *	perform necessary setup and returns 1 to indicate the device is set up
1004  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1005  *
1006  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1007  *			       struct netdev_fcoe_hbainfo *hbainfo);
1008  *	Called when the FCoE Protocol stack wants information on the underlying
1009  *	device. This information is utilized by the FCoE protocol stack to
1010  *	register attributes with Fiber Channel management service as per the
1011  *	FC-GS Fabric Device Management Information(FDMI) specification.
1012  *
1013  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1014  *	Called when the underlying device wants to override default World Wide
1015  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1016  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1017  *	protocol stack to use.
1018  *
1019  *	RFS acceleration.
1020  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1021  *			    u16 rxq_index, u32 flow_id);
1022  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1023  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1024  *	Return the filter ID on success, or a negative error code.
1025  *
1026  *	Slave management functions (for bridge, bonding, etc).
1027  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1028  *	Called to make another netdev an underling.
1029  *
1030  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1031  *	Called to release previously enslaved netdev.
1032  *
1033  *      Feature/offload setting functions.
1034  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1035  *		netdev_features_t features);
1036  *	Adjusts the requested feature flags according to device-specific
1037  *	constraints, and returns the resulting flags. Must not modify
1038  *	the device state.
1039  *
1040  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1041  *	Called to update device configuration to new features. Passed
1042  *	feature set might be less than what was returned by ndo_fix_features()).
1043  *	Must return >0 or -errno if it changed dev->features itself.
1044  *
1045  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1046  *		      struct net_device *dev,
1047  *		      const unsigned char *addr, u16 vid, u16 flags)
1048  *	Adds an FDB entry to dev for addr.
1049  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1050  *		      struct net_device *dev,
1051  *		      const unsigned char *addr, u16 vid)
1052  *	Deletes the FDB entry from dev coresponding to addr.
1053  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1054  *		       struct net_device *dev, struct net_device *filter_dev,
1055  *		       int *idx)
1056  *	Used to add FDB entries to dump requests. Implementers should add
1057  *	entries to skb and update idx with the number of entries.
1058  *
1059  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1060  *			     u16 flags)
1061  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1062  *			     struct net_device *dev, u32 filter_mask,
1063  *			     int nlflags)
1064  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1065  *			     u16 flags);
1066  *
1067  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1068  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1069  *	which do not represent real hardware may define this to allow their
1070  *	userspace components to manage their virtual carrier state. Devices
1071  *	that determine carrier state from physical hardware properties (eg
1072  *	network cables) or protocol-dependent mechanisms (eg
1073  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1074  *
1075  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1076  *			       struct netdev_phys_item_id *ppid);
1077  *	Called to get ID of physical port of this device. If driver does
1078  *	not implement this, it is assumed that the hw is not able to have
1079  *	multiple net devices on single physical port.
1080  *
1081  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1082  *			      struct udp_tunnel_info *ti);
1083  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1084  *	address family that a UDP tunnel is listnening to. It is called only
1085  *	when a new port starts listening. The operation is protected by the
1086  *	RTNL.
1087  *
1088  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1089  *			      struct udp_tunnel_info *ti);
1090  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1091  *	address family that the UDP tunnel is not listening to anymore. The
1092  *	operation is protected by the RTNL.
1093  *
1094  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1095  *				 struct net_device *dev)
1096  *	Called by upper layer devices to accelerate switching or other
1097  *	station functionality into hardware. 'pdev is the lowerdev
1098  *	to use for the offload and 'dev' is the net device that will
1099  *	back the offload. Returns a pointer to the private structure
1100  *	the upper layer will maintain.
1101  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1102  *	Called by upper layer device to delete the station created
1103  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1104  *	the station and priv is the structure returned by the add
1105  *	operation.
1106  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1107  *			     int queue_index, u32 maxrate);
1108  *	Called when a user wants to set a max-rate limitation of specific
1109  *	TX queue.
1110  * int (*ndo_get_iflink)(const struct net_device *dev);
1111  *	Called to get the iflink value of this device.
1112  * void (*ndo_change_proto_down)(struct net_device *dev,
1113  *				 bool proto_down);
1114  *	This function is used to pass protocol port error state information
1115  *	to the switch driver. The switch driver can react to the proto_down
1116  *      by doing a phys down on the associated switch port.
1117  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1118  *	This function is used to get egress tunnel information for given skb.
1119  *	This is useful for retrieving outer tunnel header parameters while
1120  *	sampling packet.
1121  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1122  *	This function is used to specify the headroom that the skb must
1123  *	consider when allocation skb during packet reception. Setting
1124  *	appropriate rx headroom value allows avoiding skb head copy on
1125  *	forward. Setting a negative value resets the rx headroom to the
1126  *	default value.
1127  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1128  *	This function is used to set or query state related to XDP on the
1129  *	netdevice. See definition of enum xdp_netdev_command for details.
1130  * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1131  *	This function is used to submit a XDP packet for transmit on a
1132  *	netdevice.
1133  * void (*ndo_xdp_flush)(struct net_device *dev);
1134  *	This function is used to inform the driver to flush a particular
1135  *	xdp tx queue. Must be called on same CPU as xdp_xmit.
1136  */
1137 struct net_device_ops {
1138 	int			(*ndo_init)(struct net_device *dev);
1139 	void			(*ndo_uninit)(struct net_device *dev);
1140 	int			(*ndo_open)(struct net_device *dev);
1141 	int			(*ndo_stop)(struct net_device *dev);
1142 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1143 						  struct net_device *dev);
1144 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1145 						      struct net_device *dev,
1146 						      netdev_features_t features);
1147 	u16			(*ndo_select_queue)(struct net_device *dev,
1148 						    struct sk_buff *skb,
1149 						    void *accel_priv,
1150 						    select_queue_fallback_t fallback);
1151 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1152 						       int flags);
1153 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1154 	int			(*ndo_set_mac_address)(struct net_device *dev,
1155 						       void *addr);
1156 	int			(*ndo_validate_addr)(struct net_device *dev);
1157 	int			(*ndo_do_ioctl)(struct net_device *dev,
1158 					        struct ifreq *ifr, int cmd);
1159 	int			(*ndo_set_config)(struct net_device *dev,
1160 					          struct ifmap *map);
1161 	int			(*ndo_change_mtu)(struct net_device *dev,
1162 						  int new_mtu);
1163 	int			(*ndo_neigh_setup)(struct net_device *dev,
1164 						   struct neigh_parms *);
1165 	void			(*ndo_tx_timeout) (struct net_device *dev);
1166 
1167 	void			(*ndo_get_stats64)(struct net_device *dev,
1168 						   struct rtnl_link_stats64 *storage);
1169 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1170 	int			(*ndo_get_offload_stats)(int attr_id,
1171 							 const struct net_device *dev,
1172 							 void *attr_data);
1173 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174 
1175 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1176 						       __be16 proto, u16 vid);
1177 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1178 						        __be16 proto, u16 vid);
1179 #ifdef CONFIG_NET_POLL_CONTROLLER
1180 	void                    (*ndo_poll_controller)(struct net_device *dev);
1181 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1182 						     struct netpoll_info *info);
1183 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1184 #endif
1185 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1186 						  int queue, u8 *mac);
1187 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1188 						   int queue, u16 vlan,
1189 						   u8 qos, __be16 proto);
1190 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1191 						   int vf, int min_tx_rate,
1192 						   int max_tx_rate);
1193 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1194 						       int vf, bool setting);
1195 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1196 						    int vf, bool setting);
1197 	int			(*ndo_get_vf_config)(struct net_device *dev,
1198 						     int vf,
1199 						     struct ifla_vf_info *ivf);
1200 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1201 							 int vf, int link_state);
1202 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1203 						    int vf,
1204 						    struct ifla_vf_stats
1205 						    *vf_stats);
1206 	int			(*ndo_set_vf_port)(struct net_device *dev,
1207 						   int vf,
1208 						   struct nlattr *port[]);
1209 	int			(*ndo_get_vf_port)(struct net_device *dev,
1210 						   int vf, struct sk_buff *skb);
1211 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1212 						   int vf, u64 guid,
1213 						   int guid_type);
1214 	int			(*ndo_set_vf_rss_query_en)(
1215 						   struct net_device *dev,
1216 						   int vf, bool setting);
1217 	int			(*ndo_setup_tc)(struct net_device *dev,
1218 						enum tc_setup_type type,
1219 						void *type_data);
1220 #if IS_ENABLED(CONFIG_FCOE)
1221 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1222 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1223 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1224 						      u16 xid,
1225 						      struct scatterlist *sgl,
1226 						      unsigned int sgc);
1227 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1228 						     u16 xid);
1229 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1230 						       u16 xid,
1231 						       struct scatterlist *sgl,
1232 						       unsigned int sgc);
1233 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1234 							struct netdev_fcoe_hbainfo *hbainfo);
1235 #endif
1236 
1237 #if IS_ENABLED(CONFIG_LIBFCOE)
1238 #define NETDEV_FCOE_WWNN 0
1239 #define NETDEV_FCOE_WWPN 1
1240 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1241 						    u64 *wwn, int type);
1242 #endif
1243 
1244 #ifdef CONFIG_RFS_ACCEL
1245 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1246 						     const struct sk_buff *skb,
1247 						     u16 rxq_index,
1248 						     u32 flow_id);
1249 #endif
1250 	int			(*ndo_add_slave)(struct net_device *dev,
1251 						 struct net_device *slave_dev,
1252 						 struct netlink_ext_ack *extack);
1253 	int			(*ndo_del_slave)(struct net_device *dev,
1254 						 struct net_device *slave_dev);
1255 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1256 						    netdev_features_t features);
1257 	int			(*ndo_set_features)(struct net_device *dev,
1258 						    netdev_features_t features);
1259 	int			(*ndo_neigh_construct)(struct net_device *dev,
1260 						       struct neighbour *n);
1261 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1262 						     struct neighbour *n);
1263 
1264 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1265 					       struct nlattr *tb[],
1266 					       struct net_device *dev,
1267 					       const unsigned char *addr,
1268 					       u16 vid,
1269 					       u16 flags);
1270 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1271 					       struct nlattr *tb[],
1272 					       struct net_device *dev,
1273 					       const unsigned char *addr,
1274 					       u16 vid);
1275 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1276 						struct netlink_callback *cb,
1277 						struct net_device *dev,
1278 						struct net_device *filter_dev,
1279 						int *idx);
1280 
1281 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1282 						      struct nlmsghdr *nlh,
1283 						      u16 flags);
1284 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1285 						      u32 pid, u32 seq,
1286 						      struct net_device *dev,
1287 						      u32 filter_mask,
1288 						      int nlflags);
1289 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1290 						      struct nlmsghdr *nlh,
1291 						      u16 flags);
1292 	int			(*ndo_change_carrier)(struct net_device *dev,
1293 						      bool new_carrier);
1294 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1295 							struct netdev_phys_item_id *ppid);
1296 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1297 							  char *name, size_t len);
1298 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1299 						      struct udp_tunnel_info *ti);
1300 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1301 						      struct udp_tunnel_info *ti);
1302 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1303 							struct net_device *dev);
1304 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1305 							void *priv);
1306 
1307 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1308 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1309 						      int queue_index,
1310 						      u32 maxrate);
1311 	int			(*ndo_get_iflink)(const struct net_device *dev);
1312 	int			(*ndo_change_proto_down)(struct net_device *dev,
1313 							 bool proto_down);
1314 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1315 						       struct sk_buff *skb);
1316 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1317 						       int needed_headroom);
1318 	int			(*ndo_xdp)(struct net_device *dev,
1319 					   struct netdev_xdp *xdp);
1320 	int			(*ndo_xdp_xmit)(struct net_device *dev,
1321 						struct xdp_buff *xdp);
1322 	void			(*ndo_xdp_flush)(struct net_device *dev);
1323 };
1324 
1325 /**
1326  * enum net_device_priv_flags - &struct net_device priv_flags
1327  *
1328  * These are the &struct net_device, they are only set internally
1329  * by drivers and used in the kernel. These flags are invisible to
1330  * userspace; this means that the order of these flags can change
1331  * during any kernel release.
1332  *
1333  * You should have a pretty good reason to be extending these flags.
1334  *
1335  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1336  * @IFF_EBRIDGE: Ethernet bridging device
1337  * @IFF_BONDING: bonding master or slave
1338  * @IFF_ISATAP: ISATAP interface (RFC4214)
1339  * @IFF_WAN_HDLC: WAN HDLC device
1340  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1341  *	release skb->dst
1342  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1343  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1344  * @IFF_MACVLAN_PORT: device used as macvlan port
1345  * @IFF_BRIDGE_PORT: device used as bridge port
1346  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1347  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1348  * @IFF_UNICAST_FLT: Supports unicast filtering
1349  * @IFF_TEAM_PORT: device used as team port
1350  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1351  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1352  *	change when it's running
1353  * @IFF_MACVLAN: Macvlan device
1354  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1355  *	underlying stacked devices
1356  * @IFF_IPVLAN_MASTER: IPvlan master device
1357  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1358  * @IFF_L3MDEV_MASTER: device is an L3 master device
1359  * @IFF_NO_QUEUE: device can run without qdisc attached
1360  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1361  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1362  * @IFF_TEAM: device is a team device
1363  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1364  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1365  *	entity (i.e. the master device for bridged veth)
1366  * @IFF_MACSEC: device is a MACsec device
1367  */
1368 enum netdev_priv_flags {
1369 	IFF_802_1Q_VLAN			= 1<<0,
1370 	IFF_EBRIDGE			= 1<<1,
1371 	IFF_BONDING			= 1<<2,
1372 	IFF_ISATAP			= 1<<3,
1373 	IFF_WAN_HDLC			= 1<<4,
1374 	IFF_XMIT_DST_RELEASE		= 1<<5,
1375 	IFF_DONT_BRIDGE			= 1<<6,
1376 	IFF_DISABLE_NETPOLL		= 1<<7,
1377 	IFF_MACVLAN_PORT		= 1<<8,
1378 	IFF_BRIDGE_PORT			= 1<<9,
1379 	IFF_OVS_DATAPATH		= 1<<10,
1380 	IFF_TX_SKB_SHARING		= 1<<11,
1381 	IFF_UNICAST_FLT			= 1<<12,
1382 	IFF_TEAM_PORT			= 1<<13,
1383 	IFF_SUPP_NOFCS			= 1<<14,
1384 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1385 	IFF_MACVLAN			= 1<<16,
1386 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1387 	IFF_IPVLAN_MASTER		= 1<<18,
1388 	IFF_IPVLAN_SLAVE		= 1<<19,
1389 	IFF_L3MDEV_MASTER		= 1<<20,
1390 	IFF_NO_QUEUE			= 1<<21,
1391 	IFF_OPENVSWITCH			= 1<<22,
1392 	IFF_L3MDEV_SLAVE		= 1<<23,
1393 	IFF_TEAM			= 1<<24,
1394 	IFF_RXFH_CONFIGURED		= 1<<25,
1395 	IFF_PHONY_HEADROOM		= 1<<26,
1396 	IFF_MACSEC			= 1<<27,
1397 };
1398 
1399 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1400 #define IFF_EBRIDGE			IFF_EBRIDGE
1401 #define IFF_BONDING			IFF_BONDING
1402 #define IFF_ISATAP			IFF_ISATAP
1403 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1404 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1405 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1406 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1407 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1408 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1409 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1410 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1411 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1412 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1413 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1414 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1415 #define IFF_MACVLAN			IFF_MACVLAN
1416 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1417 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1418 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1419 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1420 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1421 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1422 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1423 #define IFF_TEAM			IFF_TEAM
1424 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1425 #define IFF_MACSEC			IFF_MACSEC
1426 
1427 /**
1428  *	struct net_device - The DEVICE structure.
1429  *
1430  *	Actually, this whole structure is a big mistake.  It mixes I/O
1431  *	data with strictly "high-level" data, and it has to know about
1432  *	almost every data structure used in the INET module.
1433  *
1434  *	@name:	This is the first field of the "visible" part of this structure
1435  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1436  *		of the interface.
1437  *
1438  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1439  *	@ifalias:	SNMP alias
1440  *	@mem_end:	Shared memory end
1441  *	@mem_start:	Shared memory start
1442  *	@base_addr:	Device I/O address
1443  *	@irq:		Device IRQ number
1444  *
1445  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1446  *
1447  *	@state:		Generic network queuing layer state, see netdev_state_t
1448  *	@dev_list:	The global list of network devices
1449  *	@napi_list:	List entry used for polling NAPI devices
1450  *	@unreg_list:	List entry  when we are unregistering the
1451  *			device; see the function unregister_netdev
1452  *	@close_list:	List entry used when we are closing the device
1453  *	@ptype_all:     Device-specific packet handlers for all protocols
1454  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1455  *
1456  *	@adj_list:	Directly linked devices, like slaves for bonding
1457  *	@features:	Currently active device features
1458  *	@hw_features:	User-changeable features
1459  *
1460  *	@wanted_features:	User-requested features
1461  *	@vlan_features:		Mask of features inheritable by VLAN devices
1462  *
1463  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1464  *				This field indicates what encapsulation
1465  *				offloads the hardware is capable of doing,
1466  *				and drivers will need to set them appropriately.
1467  *
1468  *	@mpls_features:	Mask of features inheritable by MPLS
1469  *
1470  *	@ifindex:	interface index
1471  *	@group:		The group the device belongs to
1472  *
1473  *	@stats:		Statistics struct, which was left as a legacy, use
1474  *			rtnl_link_stats64 instead
1475  *
1476  *	@rx_dropped:	Dropped packets by core network,
1477  *			do not use this in drivers
1478  *	@tx_dropped:	Dropped packets by core network,
1479  *			do not use this in drivers
1480  *	@rx_nohandler:	nohandler dropped packets by core network on
1481  *			inactive devices, do not use this in drivers
1482  *
1483  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1484  *				instead of ioctl,
1485  *				see <net/iw_handler.h> for details.
1486  *	@wireless_data:	Instance data managed by the core of wireless extensions
1487  *
1488  *	@netdev_ops:	Includes several pointers to callbacks,
1489  *			if one wants to override the ndo_*() functions
1490  *	@ethtool_ops:	Management operations
1491  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1492  *			discovery handling. Necessary for e.g. 6LoWPAN.
1493  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1494  *			of Layer 2 headers.
1495  *
1496  *	@flags:		Interface flags (a la BSD)
1497  *	@priv_flags:	Like 'flags' but invisible to userspace,
1498  *			see if.h for the definitions
1499  *	@gflags:	Global flags ( kept as legacy )
1500  *	@padded:	How much padding added by alloc_netdev()
1501  *	@operstate:	RFC2863 operstate
1502  *	@link_mode:	Mapping policy to operstate
1503  *	@if_port:	Selectable AUI, TP, ...
1504  *	@dma:		DMA channel
1505  *	@mtu:		Interface MTU value
1506  *	@min_mtu:	Interface Minimum MTU value
1507  *	@max_mtu:	Interface Maximum MTU value
1508  *	@type:		Interface hardware type
1509  *	@hard_header_len: Maximum hardware header length.
1510  *	@min_header_len:  Minimum hardware header length
1511  *
1512  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1513  *			  cases can this be guaranteed
1514  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1515  *			  cases can this be guaranteed. Some cases also use
1516  *			  LL_MAX_HEADER instead to allocate the skb
1517  *
1518  *	interface address info:
1519  *
1520  * 	@perm_addr:		Permanent hw address
1521  * 	@addr_assign_type:	Hw address assignment type
1522  * 	@addr_len:		Hardware address length
1523  *	@neigh_priv_len:	Used in neigh_alloc()
1524  * 	@dev_id:		Used to differentiate devices that share
1525  * 				the same link layer address
1526  * 	@dev_port:		Used to differentiate devices that share
1527  * 				the same function
1528  *	@addr_list_lock:	XXX: need comments on this one
1529  *	@uc_promisc:		Counter that indicates promiscuous mode
1530  *				has been enabled due to the need to listen to
1531  *				additional unicast addresses in a device that
1532  *				does not implement ndo_set_rx_mode()
1533  *	@uc:			unicast mac addresses
1534  *	@mc:			multicast mac addresses
1535  *	@dev_addrs:		list of device hw addresses
1536  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1537  *	@promiscuity:		Number of times the NIC is told to work in
1538  *				promiscuous mode; if it becomes 0 the NIC will
1539  *				exit promiscuous mode
1540  *	@allmulti:		Counter, enables or disables allmulticast mode
1541  *
1542  *	@vlan_info:	VLAN info
1543  *	@dsa_ptr:	dsa specific data
1544  *	@tipc_ptr:	TIPC specific data
1545  *	@atalk_ptr:	AppleTalk link
1546  *	@ip_ptr:	IPv4 specific data
1547  *	@dn_ptr:	DECnet specific data
1548  *	@ip6_ptr:	IPv6 specific data
1549  *	@ax25_ptr:	AX.25 specific data
1550  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1551  *
1552  *	@dev_addr:	Hw address (before bcast,
1553  *			because most packets are unicast)
1554  *
1555  *	@_rx:			Array of RX queues
1556  *	@num_rx_queues:		Number of RX queues
1557  *				allocated at register_netdev() time
1558  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1559  *
1560  *	@rx_handler:		handler for received packets
1561  *	@rx_handler_data: 	XXX: need comments on this one
1562  *	@ingress_queue:		XXX: need comments on this one
1563  *	@broadcast:		hw bcast address
1564  *
1565  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1566  *			indexed by RX queue number. Assigned by driver.
1567  *			This must only be set if the ndo_rx_flow_steer
1568  *			operation is defined
1569  *	@index_hlist:		Device index hash chain
1570  *
1571  *	@_tx:			Array of TX queues
1572  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1573  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1574  *	@qdisc:			Root qdisc from userspace point of view
1575  *	@tx_queue_len:		Max frames per queue allowed
1576  *	@tx_global_lock: 	XXX: need comments on this one
1577  *
1578  *	@xps_maps:	XXX: need comments on this one
1579  *
1580  *	@watchdog_timeo:	Represents the timeout that is used by
1581  *				the watchdog (see dev_watchdog())
1582  *	@watchdog_timer:	List of timers
1583  *
1584  *	@pcpu_refcnt:		Number of references to this device
1585  *	@todo_list:		Delayed register/unregister
1586  *	@link_watch_list:	XXX: need comments on this one
1587  *
1588  *	@reg_state:		Register/unregister state machine
1589  *	@dismantle:		Device is going to be freed
1590  *	@rtnl_link_state:	This enum represents the phases of creating
1591  *				a new link
1592  *
1593  *	@needs_free_netdev:	Should unregister perform free_netdev?
1594  *	@priv_destructor:	Called from unregister
1595  *	@npinfo:		XXX: need comments on this one
1596  * 	@nd_net:		Network namespace this network device is inside
1597  *
1598  * 	@ml_priv:	Mid-layer private
1599  * 	@lstats:	Loopback statistics
1600  * 	@tstats:	Tunnel statistics
1601  * 	@dstats:	Dummy statistics
1602  * 	@vstats:	Virtual ethernet statistics
1603  *
1604  *	@garp_port:	GARP
1605  *	@mrp_port:	MRP
1606  *
1607  *	@dev:		Class/net/name entry
1608  *	@sysfs_groups:	Space for optional device, statistics and wireless
1609  *			sysfs groups
1610  *
1611  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1612  *	@rtnl_link_ops:	Rtnl_link_ops
1613  *
1614  *	@gso_max_size:	Maximum size of generic segmentation offload
1615  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1616  *			NIC for GSO
1617  *
1618  *	@dcbnl_ops:	Data Center Bridging netlink ops
1619  *	@num_tc:	Number of traffic classes in the net device
1620  *	@tc_to_txq:	XXX: need comments on this one
1621  *	@prio_tc_map:	XXX: need comments on this one
1622  *
1623  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1624  *
1625  *	@priomap:	XXX: need comments on this one
1626  *	@phydev:	Physical device may attach itself
1627  *			for hardware timestamping
1628  *
1629  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1630  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1631  *
1632  *	@proto_down:	protocol port state information can be sent to the
1633  *			switch driver and used to set the phys state of the
1634  *			switch port.
1635  *
1636  *	FIXME: cleanup struct net_device such that network protocol info
1637  *	moves out.
1638  */
1639 
1640 struct net_device {
1641 	char			name[IFNAMSIZ];
1642 	struct hlist_node	name_hlist;
1643 	struct dev_ifalias	__rcu *ifalias;
1644 	/*
1645 	 *	I/O specific fields
1646 	 *	FIXME: Merge these and struct ifmap into one
1647 	 */
1648 	unsigned long		mem_end;
1649 	unsigned long		mem_start;
1650 	unsigned long		base_addr;
1651 	int			irq;
1652 
1653 	atomic_t		carrier_changes;
1654 
1655 	/*
1656 	 *	Some hardware also needs these fields (state,dev_list,
1657 	 *	napi_list,unreg_list,close_list) but they are not
1658 	 *	part of the usual set specified in Space.c.
1659 	 */
1660 
1661 	unsigned long		state;
1662 
1663 	struct list_head	dev_list;
1664 	struct list_head	napi_list;
1665 	struct list_head	unreg_list;
1666 	struct list_head	close_list;
1667 	struct list_head	ptype_all;
1668 	struct list_head	ptype_specific;
1669 
1670 	struct {
1671 		struct list_head upper;
1672 		struct list_head lower;
1673 	} adj_list;
1674 
1675 	netdev_features_t	features;
1676 	netdev_features_t	hw_features;
1677 	netdev_features_t	wanted_features;
1678 	netdev_features_t	vlan_features;
1679 	netdev_features_t	hw_enc_features;
1680 	netdev_features_t	mpls_features;
1681 	netdev_features_t	gso_partial_features;
1682 
1683 	int			ifindex;
1684 	int			group;
1685 
1686 	struct net_device_stats	stats;
1687 
1688 	atomic_long_t		rx_dropped;
1689 	atomic_long_t		tx_dropped;
1690 	atomic_long_t		rx_nohandler;
1691 
1692 #ifdef CONFIG_WIRELESS_EXT
1693 	const struct iw_handler_def *wireless_handlers;
1694 	struct iw_public_data	*wireless_data;
1695 #endif
1696 	const struct net_device_ops *netdev_ops;
1697 	const struct ethtool_ops *ethtool_ops;
1698 #ifdef CONFIG_NET_SWITCHDEV
1699 	const struct switchdev_ops *switchdev_ops;
1700 #endif
1701 #ifdef CONFIG_NET_L3_MASTER_DEV
1702 	const struct l3mdev_ops	*l3mdev_ops;
1703 #endif
1704 #if IS_ENABLED(CONFIG_IPV6)
1705 	const struct ndisc_ops *ndisc_ops;
1706 #endif
1707 
1708 #ifdef CONFIG_XFRM
1709 	const struct xfrmdev_ops *xfrmdev_ops;
1710 #endif
1711 
1712 	const struct header_ops *header_ops;
1713 
1714 	unsigned int		flags;
1715 	unsigned int		priv_flags;
1716 
1717 	unsigned short		gflags;
1718 	unsigned short		padded;
1719 
1720 	unsigned char		operstate;
1721 	unsigned char		link_mode;
1722 
1723 	unsigned char		if_port;
1724 	unsigned char		dma;
1725 
1726 	unsigned int		mtu;
1727 	unsigned int		min_mtu;
1728 	unsigned int		max_mtu;
1729 	unsigned short		type;
1730 	unsigned short		hard_header_len;
1731 	unsigned char		min_header_len;
1732 
1733 	unsigned short		needed_headroom;
1734 	unsigned short		needed_tailroom;
1735 
1736 	/* Interface address info. */
1737 	unsigned char		perm_addr[MAX_ADDR_LEN];
1738 	unsigned char		addr_assign_type;
1739 	unsigned char		addr_len;
1740 	unsigned short		neigh_priv_len;
1741 	unsigned short          dev_id;
1742 	unsigned short          dev_port;
1743 	spinlock_t		addr_list_lock;
1744 	unsigned char		name_assign_type;
1745 	bool			uc_promisc;
1746 	struct netdev_hw_addr_list	uc;
1747 	struct netdev_hw_addr_list	mc;
1748 	struct netdev_hw_addr_list	dev_addrs;
1749 
1750 #ifdef CONFIG_SYSFS
1751 	struct kset		*queues_kset;
1752 #endif
1753 	unsigned int		promiscuity;
1754 	unsigned int		allmulti;
1755 
1756 
1757 	/* Protocol-specific pointers */
1758 
1759 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1760 	struct vlan_info __rcu	*vlan_info;
1761 #endif
1762 #if IS_ENABLED(CONFIG_NET_DSA)
1763 	struct dsa_port		*dsa_ptr;
1764 #endif
1765 #if IS_ENABLED(CONFIG_TIPC)
1766 	struct tipc_bearer __rcu *tipc_ptr;
1767 #endif
1768 	void 			*atalk_ptr;
1769 	struct in_device __rcu	*ip_ptr;
1770 	struct dn_dev __rcu     *dn_ptr;
1771 	struct inet6_dev __rcu	*ip6_ptr;
1772 	void			*ax25_ptr;
1773 	struct wireless_dev	*ieee80211_ptr;
1774 	struct wpan_dev		*ieee802154_ptr;
1775 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1776 	struct mpls_dev __rcu	*mpls_ptr;
1777 #endif
1778 
1779 /*
1780  * Cache lines mostly used on receive path (including eth_type_trans())
1781  */
1782 	/* Interface address info used in eth_type_trans() */
1783 	unsigned char		*dev_addr;
1784 
1785 #ifdef CONFIG_SYSFS
1786 	struct netdev_rx_queue	*_rx;
1787 
1788 	unsigned int		num_rx_queues;
1789 	unsigned int		real_num_rx_queues;
1790 #endif
1791 
1792 	struct bpf_prog __rcu	*xdp_prog;
1793 	unsigned long		gro_flush_timeout;
1794 	rx_handler_func_t __rcu	*rx_handler;
1795 	void __rcu		*rx_handler_data;
1796 
1797 #ifdef CONFIG_NET_CLS_ACT
1798 	struct tcf_proto __rcu  *ingress_cl_list;
1799 #endif
1800 	struct netdev_queue __rcu *ingress_queue;
1801 #ifdef CONFIG_NETFILTER_INGRESS
1802 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1803 #endif
1804 
1805 	unsigned char		broadcast[MAX_ADDR_LEN];
1806 #ifdef CONFIG_RFS_ACCEL
1807 	struct cpu_rmap		*rx_cpu_rmap;
1808 #endif
1809 	struct hlist_node	index_hlist;
1810 
1811 /*
1812  * Cache lines mostly used on transmit path
1813  */
1814 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1815 	unsigned int		num_tx_queues;
1816 	unsigned int		real_num_tx_queues;
1817 	struct Qdisc		*qdisc;
1818 #ifdef CONFIG_NET_SCHED
1819 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1820 #endif
1821 	unsigned int		tx_queue_len;
1822 	spinlock_t		tx_global_lock;
1823 	int			watchdog_timeo;
1824 
1825 #ifdef CONFIG_XPS
1826 	struct xps_dev_maps __rcu *xps_maps;
1827 #endif
1828 #ifdef CONFIG_NET_CLS_ACT
1829 	struct tcf_proto __rcu  *egress_cl_list;
1830 #endif
1831 
1832 	/* These may be needed for future network-power-down code. */
1833 	struct timer_list	watchdog_timer;
1834 
1835 	int __percpu		*pcpu_refcnt;
1836 	struct list_head	todo_list;
1837 
1838 	struct list_head	link_watch_list;
1839 
1840 	enum { NETREG_UNINITIALIZED=0,
1841 	       NETREG_REGISTERED,	/* completed register_netdevice */
1842 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1843 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1844 	       NETREG_RELEASED,		/* called free_netdev */
1845 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1846 	} reg_state:8;
1847 
1848 	bool dismantle;
1849 
1850 	enum {
1851 		RTNL_LINK_INITIALIZED,
1852 		RTNL_LINK_INITIALIZING,
1853 	} rtnl_link_state:16;
1854 
1855 	bool needs_free_netdev;
1856 	void (*priv_destructor)(struct net_device *dev);
1857 
1858 #ifdef CONFIG_NETPOLL
1859 	struct netpoll_info __rcu	*npinfo;
1860 #endif
1861 
1862 	possible_net_t			nd_net;
1863 
1864 	/* mid-layer private */
1865 	union {
1866 		void					*ml_priv;
1867 		struct pcpu_lstats __percpu		*lstats;
1868 		struct pcpu_sw_netstats __percpu	*tstats;
1869 		struct pcpu_dstats __percpu		*dstats;
1870 		struct pcpu_vstats __percpu		*vstats;
1871 	};
1872 
1873 #if IS_ENABLED(CONFIG_GARP)
1874 	struct garp_port __rcu	*garp_port;
1875 #endif
1876 #if IS_ENABLED(CONFIG_MRP)
1877 	struct mrp_port __rcu	*mrp_port;
1878 #endif
1879 
1880 	struct device		dev;
1881 	const struct attribute_group *sysfs_groups[4];
1882 	const struct attribute_group *sysfs_rx_queue_group;
1883 
1884 	const struct rtnl_link_ops *rtnl_link_ops;
1885 
1886 	/* for setting kernel sock attribute on TCP connection setup */
1887 #define GSO_MAX_SIZE		65536
1888 	unsigned int		gso_max_size;
1889 #define GSO_MAX_SEGS		65535
1890 	u16			gso_max_segs;
1891 
1892 #ifdef CONFIG_DCB
1893 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1894 #endif
1895 	u8			num_tc;
1896 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1897 	u8			prio_tc_map[TC_BITMASK + 1];
1898 
1899 #if IS_ENABLED(CONFIG_FCOE)
1900 	unsigned int		fcoe_ddp_xid;
1901 #endif
1902 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1903 	struct netprio_map __rcu *priomap;
1904 #endif
1905 	struct phy_device	*phydev;
1906 	struct lock_class_key	*qdisc_tx_busylock;
1907 	struct lock_class_key	*qdisc_running_key;
1908 	bool			proto_down;
1909 };
1910 #define to_net_dev(d) container_of(d, struct net_device, dev)
1911 
1912 static inline bool netif_elide_gro(const struct net_device *dev)
1913 {
1914 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1915 		return true;
1916 	return false;
1917 }
1918 
1919 #define	NETDEV_ALIGN		32
1920 
1921 static inline
1922 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1923 {
1924 	return dev->prio_tc_map[prio & TC_BITMASK];
1925 }
1926 
1927 static inline
1928 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1929 {
1930 	if (tc >= dev->num_tc)
1931 		return -EINVAL;
1932 
1933 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1934 	return 0;
1935 }
1936 
1937 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1938 void netdev_reset_tc(struct net_device *dev);
1939 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1940 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1941 
1942 static inline
1943 int netdev_get_num_tc(struct net_device *dev)
1944 {
1945 	return dev->num_tc;
1946 }
1947 
1948 static inline
1949 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1950 					 unsigned int index)
1951 {
1952 	return &dev->_tx[index];
1953 }
1954 
1955 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1956 						    const struct sk_buff *skb)
1957 {
1958 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1959 }
1960 
1961 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1962 					    void (*f)(struct net_device *,
1963 						      struct netdev_queue *,
1964 						      void *),
1965 					    void *arg)
1966 {
1967 	unsigned int i;
1968 
1969 	for (i = 0; i < dev->num_tx_queues; i++)
1970 		f(dev, &dev->_tx[i], arg);
1971 }
1972 
1973 #define netdev_lockdep_set_classes(dev)				\
1974 {								\
1975 	static struct lock_class_key qdisc_tx_busylock_key;	\
1976 	static struct lock_class_key qdisc_running_key;		\
1977 	static struct lock_class_key qdisc_xmit_lock_key;	\
1978 	static struct lock_class_key dev_addr_list_lock_key;	\
1979 	unsigned int i;						\
1980 								\
1981 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1982 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1983 	lockdep_set_class(&(dev)->addr_list_lock,		\
1984 			  &dev_addr_list_lock_key); 		\
1985 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1986 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1987 				  &qdisc_xmit_lock_key);	\
1988 }
1989 
1990 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1991 				    struct sk_buff *skb,
1992 				    void *accel_priv);
1993 
1994 /* returns the headroom that the master device needs to take in account
1995  * when forwarding to this dev
1996  */
1997 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1998 {
1999 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2000 }
2001 
2002 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2003 {
2004 	if (dev->netdev_ops->ndo_set_rx_headroom)
2005 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2006 }
2007 
2008 /* set the device rx headroom to the dev's default */
2009 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2010 {
2011 	netdev_set_rx_headroom(dev, -1);
2012 }
2013 
2014 /*
2015  * Net namespace inlines
2016  */
2017 static inline
2018 struct net *dev_net(const struct net_device *dev)
2019 {
2020 	return read_pnet(&dev->nd_net);
2021 }
2022 
2023 static inline
2024 void dev_net_set(struct net_device *dev, struct net *net)
2025 {
2026 	write_pnet(&dev->nd_net, net);
2027 }
2028 
2029 /**
2030  *	netdev_priv - access network device private data
2031  *	@dev: network device
2032  *
2033  * Get network device private data
2034  */
2035 static inline void *netdev_priv(const struct net_device *dev)
2036 {
2037 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2038 }
2039 
2040 /* Set the sysfs physical device reference for the network logical device
2041  * if set prior to registration will cause a symlink during initialization.
2042  */
2043 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2044 
2045 /* Set the sysfs device type for the network logical device to allow
2046  * fine-grained identification of different network device types. For
2047  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2048  */
2049 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2050 
2051 /* Default NAPI poll() weight
2052  * Device drivers are strongly advised to not use bigger value
2053  */
2054 #define NAPI_POLL_WEIGHT 64
2055 
2056 /**
2057  *	netif_napi_add - initialize a NAPI context
2058  *	@dev:  network device
2059  *	@napi: NAPI context
2060  *	@poll: polling function
2061  *	@weight: default weight
2062  *
2063  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2064  * *any* of the other NAPI-related functions.
2065  */
2066 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2067 		    int (*poll)(struct napi_struct *, int), int weight);
2068 
2069 /**
2070  *	netif_tx_napi_add - initialize a NAPI context
2071  *	@dev:  network device
2072  *	@napi: NAPI context
2073  *	@poll: polling function
2074  *	@weight: default weight
2075  *
2076  * This variant of netif_napi_add() should be used from drivers using NAPI
2077  * to exclusively poll a TX queue.
2078  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2079  */
2080 static inline void netif_tx_napi_add(struct net_device *dev,
2081 				     struct napi_struct *napi,
2082 				     int (*poll)(struct napi_struct *, int),
2083 				     int weight)
2084 {
2085 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2086 	netif_napi_add(dev, napi, poll, weight);
2087 }
2088 
2089 /**
2090  *  netif_napi_del - remove a NAPI context
2091  *  @napi: NAPI context
2092  *
2093  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2094  */
2095 void netif_napi_del(struct napi_struct *napi);
2096 
2097 struct napi_gro_cb {
2098 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2099 	void	*frag0;
2100 
2101 	/* Length of frag0. */
2102 	unsigned int frag0_len;
2103 
2104 	/* This indicates where we are processing relative to skb->data. */
2105 	int	data_offset;
2106 
2107 	/* This is non-zero if the packet cannot be merged with the new skb. */
2108 	u16	flush;
2109 
2110 	/* Save the IP ID here and check when we get to the transport layer */
2111 	u16	flush_id;
2112 
2113 	/* Number of segments aggregated. */
2114 	u16	count;
2115 
2116 	/* Start offset for remote checksum offload */
2117 	u16	gro_remcsum_start;
2118 
2119 	/* jiffies when first packet was created/queued */
2120 	unsigned long age;
2121 
2122 	/* Used in ipv6_gro_receive() and foo-over-udp */
2123 	u16	proto;
2124 
2125 	/* This is non-zero if the packet may be of the same flow. */
2126 	u8	same_flow:1;
2127 
2128 	/* Used in tunnel GRO receive */
2129 	u8	encap_mark:1;
2130 
2131 	/* GRO checksum is valid */
2132 	u8	csum_valid:1;
2133 
2134 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2135 	u8	csum_cnt:3;
2136 
2137 	/* Free the skb? */
2138 	u8	free:2;
2139 #define NAPI_GRO_FREE		  1
2140 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2141 
2142 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2143 	u8	is_ipv6:1;
2144 
2145 	/* Used in GRE, set in fou/gue_gro_receive */
2146 	u8	is_fou:1;
2147 
2148 	/* Used to determine if flush_id can be ignored */
2149 	u8	is_atomic:1;
2150 
2151 	/* Number of gro_receive callbacks this packet already went through */
2152 	u8 recursion_counter:4;
2153 
2154 	/* 1 bit hole */
2155 
2156 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2157 	__wsum	csum;
2158 
2159 	/* used in skb_gro_receive() slow path */
2160 	struct sk_buff *last;
2161 };
2162 
2163 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2164 
2165 #define GRO_RECURSION_LIMIT 15
2166 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2167 {
2168 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2169 }
2170 
2171 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2172 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2173 						struct sk_buff **head,
2174 						struct sk_buff *skb)
2175 {
2176 	if (unlikely(gro_recursion_inc_test(skb))) {
2177 		NAPI_GRO_CB(skb)->flush |= 1;
2178 		return NULL;
2179 	}
2180 
2181 	return cb(head, skb);
2182 }
2183 
2184 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2185 					     struct sk_buff *);
2186 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2187 						   struct sock *sk,
2188 						   struct sk_buff **head,
2189 						   struct sk_buff *skb)
2190 {
2191 	if (unlikely(gro_recursion_inc_test(skb))) {
2192 		NAPI_GRO_CB(skb)->flush |= 1;
2193 		return NULL;
2194 	}
2195 
2196 	return cb(sk, head, skb);
2197 }
2198 
2199 struct packet_type {
2200 	__be16			type;	/* This is really htons(ether_type). */
2201 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2202 	int			(*func) (struct sk_buff *,
2203 					 struct net_device *,
2204 					 struct packet_type *,
2205 					 struct net_device *);
2206 	bool			(*id_match)(struct packet_type *ptype,
2207 					    struct sock *sk);
2208 	void			*af_packet_priv;
2209 	struct list_head	list;
2210 };
2211 
2212 struct offload_callbacks {
2213 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2214 						netdev_features_t features);
2215 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2216 						 struct sk_buff *skb);
2217 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2218 };
2219 
2220 struct packet_offload {
2221 	__be16			 type;	/* This is really htons(ether_type). */
2222 	u16			 priority;
2223 	struct offload_callbacks callbacks;
2224 	struct list_head	 list;
2225 };
2226 
2227 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2228 struct pcpu_sw_netstats {
2229 	u64     rx_packets;
2230 	u64     rx_bytes;
2231 	u64     tx_packets;
2232 	u64     tx_bytes;
2233 	struct u64_stats_sync   syncp;
2234 };
2235 
2236 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2237 ({									\
2238 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2239 	if (pcpu_stats)	{						\
2240 		int __cpu;						\
2241 		for_each_possible_cpu(__cpu) {				\
2242 			typeof(type) *stat;				\
2243 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2244 			u64_stats_init(&stat->syncp);			\
2245 		}							\
2246 	}								\
2247 	pcpu_stats;							\
2248 })
2249 
2250 #define netdev_alloc_pcpu_stats(type)					\
2251 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2252 
2253 enum netdev_lag_tx_type {
2254 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2255 	NETDEV_LAG_TX_TYPE_RANDOM,
2256 	NETDEV_LAG_TX_TYPE_BROADCAST,
2257 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2258 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2259 	NETDEV_LAG_TX_TYPE_HASH,
2260 };
2261 
2262 struct netdev_lag_upper_info {
2263 	enum netdev_lag_tx_type tx_type;
2264 };
2265 
2266 struct netdev_lag_lower_state_info {
2267 	u8 link_up : 1,
2268 	   tx_enabled : 1;
2269 };
2270 
2271 #include <linux/notifier.h>
2272 
2273 /* netdevice notifier chain. Please remember to update the rtnetlink
2274  * notification exclusion list in rtnetlink_event() when adding new
2275  * types.
2276  */
2277 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2278 #define NETDEV_DOWN	0x0002
2279 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2280 				   detected a hardware crash and restarted
2281 				   - we can use this eg to kick tcp sessions
2282 				   once done */
2283 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2284 #define NETDEV_REGISTER 0x0005
2285 #define NETDEV_UNREGISTER	0x0006
2286 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2287 #define NETDEV_CHANGEADDR	0x0008
2288 #define NETDEV_GOING_DOWN	0x0009
2289 #define NETDEV_CHANGENAME	0x000A
2290 #define NETDEV_FEAT_CHANGE	0x000B
2291 #define NETDEV_BONDING_FAILOVER 0x000C
2292 #define NETDEV_PRE_UP		0x000D
2293 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2294 #define NETDEV_POST_TYPE_CHANGE	0x000F
2295 #define NETDEV_POST_INIT	0x0010
2296 #define NETDEV_UNREGISTER_FINAL 0x0011
2297 #define NETDEV_RELEASE		0x0012
2298 #define NETDEV_NOTIFY_PEERS	0x0013
2299 #define NETDEV_JOIN		0x0014
2300 #define NETDEV_CHANGEUPPER	0x0015
2301 #define NETDEV_RESEND_IGMP	0x0016
2302 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2303 #define NETDEV_CHANGEINFODATA	0x0018
2304 #define NETDEV_BONDING_INFO	0x0019
2305 #define NETDEV_PRECHANGEUPPER	0x001A
2306 #define NETDEV_CHANGELOWERSTATE	0x001B
2307 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2308 #define NETDEV_UDP_TUNNEL_DROP_INFO	0x001D
2309 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2310 
2311 int register_netdevice_notifier(struct notifier_block *nb);
2312 int unregister_netdevice_notifier(struct notifier_block *nb);
2313 
2314 struct netdev_notifier_info {
2315 	struct net_device	*dev;
2316 	struct netlink_ext_ack	*extack;
2317 };
2318 
2319 struct netdev_notifier_change_info {
2320 	struct netdev_notifier_info info; /* must be first */
2321 	unsigned int flags_changed;
2322 };
2323 
2324 struct netdev_notifier_changeupper_info {
2325 	struct netdev_notifier_info info; /* must be first */
2326 	struct net_device *upper_dev; /* new upper dev */
2327 	bool master; /* is upper dev master */
2328 	bool linking; /* is the notification for link or unlink */
2329 	void *upper_info; /* upper dev info */
2330 };
2331 
2332 struct netdev_notifier_changelowerstate_info {
2333 	struct netdev_notifier_info info; /* must be first */
2334 	void *lower_state_info; /* is lower dev state */
2335 };
2336 
2337 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2338 					     struct net_device *dev)
2339 {
2340 	info->dev = dev;
2341 	info->extack = NULL;
2342 }
2343 
2344 static inline struct net_device *
2345 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2346 {
2347 	return info->dev;
2348 }
2349 
2350 static inline struct netlink_ext_ack *
2351 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2352 {
2353 	return info->extack;
2354 }
2355 
2356 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2357 
2358 
2359 extern rwlock_t				dev_base_lock;		/* Device list lock */
2360 
2361 #define for_each_netdev(net, d)		\
2362 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2363 #define for_each_netdev_reverse(net, d)	\
2364 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2365 #define for_each_netdev_rcu(net, d)		\
2366 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2367 #define for_each_netdev_safe(net, d, n)	\
2368 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2369 #define for_each_netdev_continue(net, d)		\
2370 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2371 #define for_each_netdev_continue_rcu(net, d)		\
2372 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2373 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2374 		for_each_netdev_rcu(&init_net, slave)	\
2375 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2376 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2377 
2378 static inline struct net_device *next_net_device(struct net_device *dev)
2379 {
2380 	struct list_head *lh;
2381 	struct net *net;
2382 
2383 	net = dev_net(dev);
2384 	lh = dev->dev_list.next;
2385 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2386 }
2387 
2388 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2389 {
2390 	struct list_head *lh;
2391 	struct net *net;
2392 
2393 	net = dev_net(dev);
2394 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2395 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2396 }
2397 
2398 static inline struct net_device *first_net_device(struct net *net)
2399 {
2400 	return list_empty(&net->dev_base_head) ? NULL :
2401 		net_device_entry(net->dev_base_head.next);
2402 }
2403 
2404 static inline struct net_device *first_net_device_rcu(struct net *net)
2405 {
2406 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2407 
2408 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2409 }
2410 
2411 int netdev_boot_setup_check(struct net_device *dev);
2412 unsigned long netdev_boot_base(const char *prefix, int unit);
2413 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2414 				       const char *hwaddr);
2415 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2416 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2417 void dev_add_pack(struct packet_type *pt);
2418 void dev_remove_pack(struct packet_type *pt);
2419 void __dev_remove_pack(struct packet_type *pt);
2420 void dev_add_offload(struct packet_offload *po);
2421 void dev_remove_offload(struct packet_offload *po);
2422 
2423 int dev_get_iflink(const struct net_device *dev);
2424 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2425 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2426 				      unsigned short mask);
2427 struct net_device *dev_get_by_name(struct net *net, const char *name);
2428 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2429 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2430 int dev_alloc_name(struct net_device *dev, const char *name);
2431 int dev_open(struct net_device *dev);
2432 void dev_close(struct net_device *dev);
2433 void dev_close_many(struct list_head *head, bool unlink);
2434 void dev_disable_lro(struct net_device *dev);
2435 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2436 int dev_queue_xmit(struct sk_buff *skb);
2437 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2438 int register_netdevice(struct net_device *dev);
2439 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2440 void unregister_netdevice_many(struct list_head *head);
2441 static inline void unregister_netdevice(struct net_device *dev)
2442 {
2443 	unregister_netdevice_queue(dev, NULL);
2444 }
2445 
2446 int netdev_refcnt_read(const struct net_device *dev);
2447 void free_netdev(struct net_device *dev);
2448 void netdev_freemem(struct net_device *dev);
2449 void synchronize_net(void);
2450 int init_dummy_netdev(struct net_device *dev);
2451 
2452 DECLARE_PER_CPU(int, xmit_recursion);
2453 #define XMIT_RECURSION_LIMIT	10
2454 
2455 static inline int dev_recursion_level(void)
2456 {
2457 	return this_cpu_read(xmit_recursion);
2458 }
2459 
2460 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2461 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2462 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2463 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2464 int netdev_get_name(struct net *net, char *name, int ifindex);
2465 int dev_restart(struct net_device *dev);
2466 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2467 
2468 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2469 {
2470 	return NAPI_GRO_CB(skb)->data_offset;
2471 }
2472 
2473 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2474 {
2475 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2476 }
2477 
2478 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2479 {
2480 	NAPI_GRO_CB(skb)->data_offset += len;
2481 }
2482 
2483 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2484 					unsigned int offset)
2485 {
2486 	return NAPI_GRO_CB(skb)->frag0 + offset;
2487 }
2488 
2489 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2490 {
2491 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2492 }
2493 
2494 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2495 {
2496 	NAPI_GRO_CB(skb)->frag0 = NULL;
2497 	NAPI_GRO_CB(skb)->frag0_len = 0;
2498 }
2499 
2500 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2501 					unsigned int offset)
2502 {
2503 	if (!pskb_may_pull(skb, hlen))
2504 		return NULL;
2505 
2506 	skb_gro_frag0_invalidate(skb);
2507 	return skb->data + offset;
2508 }
2509 
2510 static inline void *skb_gro_network_header(struct sk_buff *skb)
2511 {
2512 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2513 	       skb_network_offset(skb);
2514 }
2515 
2516 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2517 					const void *start, unsigned int len)
2518 {
2519 	if (NAPI_GRO_CB(skb)->csum_valid)
2520 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2521 						  csum_partial(start, len, 0));
2522 }
2523 
2524 /* GRO checksum functions. These are logical equivalents of the normal
2525  * checksum functions (in skbuff.h) except that they operate on the GRO
2526  * offsets and fields in sk_buff.
2527  */
2528 
2529 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2530 
2531 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2532 {
2533 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2534 }
2535 
2536 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2537 						      bool zero_okay,
2538 						      __sum16 check)
2539 {
2540 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2541 		skb_checksum_start_offset(skb) <
2542 		 skb_gro_offset(skb)) &&
2543 		!skb_at_gro_remcsum_start(skb) &&
2544 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2545 		(!zero_okay || check));
2546 }
2547 
2548 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2549 							   __wsum psum)
2550 {
2551 	if (NAPI_GRO_CB(skb)->csum_valid &&
2552 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2553 		return 0;
2554 
2555 	NAPI_GRO_CB(skb)->csum = psum;
2556 
2557 	return __skb_gro_checksum_complete(skb);
2558 }
2559 
2560 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2561 {
2562 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2563 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2564 		NAPI_GRO_CB(skb)->csum_cnt--;
2565 	} else {
2566 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2567 		 * verified a new top level checksum or an encapsulated one
2568 		 * during GRO. This saves work if we fallback to normal path.
2569 		 */
2570 		__skb_incr_checksum_unnecessary(skb);
2571 	}
2572 }
2573 
2574 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2575 				    compute_pseudo)			\
2576 ({									\
2577 	__sum16 __ret = 0;						\
2578 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2579 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2580 				compute_pseudo(skb, proto));		\
2581 	if (!__ret)							\
2582 		skb_gro_incr_csum_unnecessary(skb);			\
2583 	__ret;								\
2584 })
2585 
2586 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2587 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2588 
2589 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2590 					     compute_pseudo)		\
2591 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2592 
2593 #define skb_gro_checksum_simple_validate(skb)				\
2594 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2595 
2596 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2597 {
2598 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2599 		!NAPI_GRO_CB(skb)->csum_valid);
2600 }
2601 
2602 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2603 					      __sum16 check, __wsum pseudo)
2604 {
2605 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2606 	NAPI_GRO_CB(skb)->csum_valid = 1;
2607 }
2608 
2609 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2610 do {									\
2611 	if (__skb_gro_checksum_convert_check(skb))			\
2612 		__skb_gro_checksum_convert(skb, check,			\
2613 					   compute_pseudo(skb, proto));	\
2614 } while (0)
2615 
2616 struct gro_remcsum {
2617 	int offset;
2618 	__wsum delta;
2619 };
2620 
2621 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2622 {
2623 	grc->offset = 0;
2624 	grc->delta = 0;
2625 }
2626 
2627 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2628 					    unsigned int off, size_t hdrlen,
2629 					    int start, int offset,
2630 					    struct gro_remcsum *grc,
2631 					    bool nopartial)
2632 {
2633 	__wsum delta;
2634 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2635 
2636 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2637 
2638 	if (!nopartial) {
2639 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2640 		return ptr;
2641 	}
2642 
2643 	ptr = skb_gro_header_fast(skb, off);
2644 	if (skb_gro_header_hard(skb, off + plen)) {
2645 		ptr = skb_gro_header_slow(skb, off + plen, off);
2646 		if (!ptr)
2647 			return NULL;
2648 	}
2649 
2650 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2651 			       start, offset);
2652 
2653 	/* Adjust skb->csum since we changed the packet */
2654 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2655 
2656 	grc->offset = off + hdrlen + offset;
2657 	grc->delta = delta;
2658 
2659 	return ptr;
2660 }
2661 
2662 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2663 					   struct gro_remcsum *grc)
2664 {
2665 	void *ptr;
2666 	size_t plen = grc->offset + sizeof(u16);
2667 
2668 	if (!grc->delta)
2669 		return;
2670 
2671 	ptr = skb_gro_header_fast(skb, grc->offset);
2672 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2673 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2674 		if (!ptr)
2675 			return;
2676 	}
2677 
2678 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2679 }
2680 
2681 #ifdef CONFIG_XFRM_OFFLOAD
2682 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2683 {
2684 	if (PTR_ERR(pp) != -EINPROGRESS)
2685 		NAPI_GRO_CB(skb)->flush |= flush;
2686 }
2687 #else
2688 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2689 {
2690 	NAPI_GRO_CB(skb)->flush |= flush;
2691 }
2692 #endif
2693 
2694 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2695 				  unsigned short type,
2696 				  const void *daddr, const void *saddr,
2697 				  unsigned int len)
2698 {
2699 	if (!dev->header_ops || !dev->header_ops->create)
2700 		return 0;
2701 
2702 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2703 }
2704 
2705 static inline int dev_parse_header(const struct sk_buff *skb,
2706 				   unsigned char *haddr)
2707 {
2708 	const struct net_device *dev = skb->dev;
2709 
2710 	if (!dev->header_ops || !dev->header_ops->parse)
2711 		return 0;
2712 	return dev->header_ops->parse(skb, haddr);
2713 }
2714 
2715 /* ll_header must have at least hard_header_len allocated */
2716 static inline bool dev_validate_header(const struct net_device *dev,
2717 				       char *ll_header, int len)
2718 {
2719 	if (likely(len >= dev->hard_header_len))
2720 		return true;
2721 	if (len < dev->min_header_len)
2722 		return false;
2723 
2724 	if (capable(CAP_SYS_RAWIO)) {
2725 		memset(ll_header + len, 0, dev->hard_header_len - len);
2726 		return true;
2727 	}
2728 
2729 	if (dev->header_ops && dev->header_ops->validate)
2730 		return dev->header_ops->validate(ll_header, len);
2731 
2732 	return false;
2733 }
2734 
2735 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2736 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2737 static inline int unregister_gifconf(unsigned int family)
2738 {
2739 	return register_gifconf(family, NULL);
2740 }
2741 
2742 #ifdef CONFIG_NET_FLOW_LIMIT
2743 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2744 struct sd_flow_limit {
2745 	u64			count;
2746 	unsigned int		num_buckets;
2747 	unsigned int		history_head;
2748 	u16			history[FLOW_LIMIT_HISTORY];
2749 	u8			buckets[];
2750 };
2751 
2752 extern int netdev_flow_limit_table_len;
2753 #endif /* CONFIG_NET_FLOW_LIMIT */
2754 
2755 /*
2756  * Incoming packets are placed on per-CPU queues
2757  */
2758 struct softnet_data {
2759 	struct list_head	poll_list;
2760 	struct sk_buff_head	process_queue;
2761 
2762 	/* stats */
2763 	unsigned int		processed;
2764 	unsigned int		time_squeeze;
2765 	unsigned int		received_rps;
2766 #ifdef CONFIG_RPS
2767 	struct softnet_data	*rps_ipi_list;
2768 #endif
2769 #ifdef CONFIG_NET_FLOW_LIMIT
2770 	struct sd_flow_limit __rcu *flow_limit;
2771 #endif
2772 	struct Qdisc		*output_queue;
2773 	struct Qdisc		**output_queue_tailp;
2774 	struct sk_buff		*completion_queue;
2775 
2776 #ifdef CONFIG_RPS
2777 	/* input_queue_head should be written by cpu owning this struct,
2778 	 * and only read by other cpus. Worth using a cache line.
2779 	 */
2780 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2781 
2782 	/* Elements below can be accessed between CPUs for RPS/RFS */
2783 	call_single_data_t	csd ____cacheline_aligned_in_smp;
2784 	struct softnet_data	*rps_ipi_next;
2785 	unsigned int		cpu;
2786 	unsigned int		input_queue_tail;
2787 #endif
2788 	unsigned int		dropped;
2789 	struct sk_buff_head	input_pkt_queue;
2790 	struct napi_struct	backlog;
2791 
2792 };
2793 
2794 static inline void input_queue_head_incr(struct softnet_data *sd)
2795 {
2796 #ifdef CONFIG_RPS
2797 	sd->input_queue_head++;
2798 #endif
2799 }
2800 
2801 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2802 					      unsigned int *qtail)
2803 {
2804 #ifdef CONFIG_RPS
2805 	*qtail = ++sd->input_queue_tail;
2806 #endif
2807 }
2808 
2809 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2810 
2811 void __netif_schedule(struct Qdisc *q);
2812 void netif_schedule_queue(struct netdev_queue *txq);
2813 
2814 static inline void netif_tx_schedule_all(struct net_device *dev)
2815 {
2816 	unsigned int i;
2817 
2818 	for (i = 0; i < dev->num_tx_queues; i++)
2819 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2820 }
2821 
2822 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2823 {
2824 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2825 }
2826 
2827 /**
2828  *	netif_start_queue - allow transmit
2829  *	@dev: network device
2830  *
2831  *	Allow upper layers to call the device hard_start_xmit routine.
2832  */
2833 static inline void netif_start_queue(struct net_device *dev)
2834 {
2835 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2836 }
2837 
2838 static inline void netif_tx_start_all_queues(struct net_device *dev)
2839 {
2840 	unsigned int i;
2841 
2842 	for (i = 0; i < dev->num_tx_queues; i++) {
2843 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2844 		netif_tx_start_queue(txq);
2845 	}
2846 }
2847 
2848 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2849 
2850 /**
2851  *	netif_wake_queue - restart transmit
2852  *	@dev: network device
2853  *
2854  *	Allow upper layers to call the device hard_start_xmit routine.
2855  *	Used for flow control when transmit resources are available.
2856  */
2857 static inline void netif_wake_queue(struct net_device *dev)
2858 {
2859 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2860 }
2861 
2862 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2863 {
2864 	unsigned int i;
2865 
2866 	for (i = 0; i < dev->num_tx_queues; i++) {
2867 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2868 		netif_tx_wake_queue(txq);
2869 	}
2870 }
2871 
2872 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2873 {
2874 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2875 }
2876 
2877 /**
2878  *	netif_stop_queue - stop transmitted packets
2879  *	@dev: network device
2880  *
2881  *	Stop upper layers calling the device hard_start_xmit routine.
2882  *	Used for flow control when transmit resources are unavailable.
2883  */
2884 static inline void netif_stop_queue(struct net_device *dev)
2885 {
2886 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2887 }
2888 
2889 void netif_tx_stop_all_queues(struct net_device *dev);
2890 
2891 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2892 {
2893 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2894 }
2895 
2896 /**
2897  *	netif_queue_stopped - test if transmit queue is flowblocked
2898  *	@dev: network device
2899  *
2900  *	Test if transmit queue on device is currently unable to send.
2901  */
2902 static inline bool netif_queue_stopped(const struct net_device *dev)
2903 {
2904 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2905 }
2906 
2907 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2908 {
2909 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2910 }
2911 
2912 static inline bool
2913 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2914 {
2915 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2916 }
2917 
2918 static inline bool
2919 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2920 {
2921 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2922 }
2923 
2924 /**
2925  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2926  *	@dev_queue: pointer to transmit queue
2927  *
2928  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2929  * to give appropriate hint to the CPU.
2930  */
2931 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2932 {
2933 #ifdef CONFIG_BQL
2934 	prefetchw(&dev_queue->dql.num_queued);
2935 #endif
2936 }
2937 
2938 /**
2939  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2940  *	@dev_queue: pointer to transmit queue
2941  *
2942  * BQL enabled drivers might use this helper in their TX completion path,
2943  * to give appropriate hint to the CPU.
2944  */
2945 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2946 {
2947 #ifdef CONFIG_BQL
2948 	prefetchw(&dev_queue->dql.limit);
2949 #endif
2950 }
2951 
2952 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2953 					unsigned int bytes)
2954 {
2955 #ifdef CONFIG_BQL
2956 	dql_queued(&dev_queue->dql, bytes);
2957 
2958 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2959 		return;
2960 
2961 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2962 
2963 	/*
2964 	 * The XOFF flag must be set before checking the dql_avail below,
2965 	 * because in netdev_tx_completed_queue we update the dql_completed
2966 	 * before checking the XOFF flag.
2967 	 */
2968 	smp_mb();
2969 
2970 	/* check again in case another CPU has just made room avail */
2971 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2972 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2973 #endif
2974 }
2975 
2976 /**
2977  * 	netdev_sent_queue - report the number of bytes queued to hardware
2978  * 	@dev: network device
2979  * 	@bytes: number of bytes queued to the hardware device queue
2980  *
2981  * 	Report the number of bytes queued for sending/completion to the network
2982  * 	device hardware queue. @bytes should be a good approximation and should
2983  * 	exactly match netdev_completed_queue() @bytes
2984  */
2985 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2986 {
2987 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2988 }
2989 
2990 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2991 					     unsigned int pkts, unsigned int bytes)
2992 {
2993 #ifdef CONFIG_BQL
2994 	if (unlikely(!bytes))
2995 		return;
2996 
2997 	dql_completed(&dev_queue->dql, bytes);
2998 
2999 	/*
3000 	 * Without the memory barrier there is a small possiblity that
3001 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3002 	 * be stopped forever
3003 	 */
3004 	smp_mb();
3005 
3006 	if (dql_avail(&dev_queue->dql) < 0)
3007 		return;
3008 
3009 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3010 		netif_schedule_queue(dev_queue);
3011 #endif
3012 }
3013 
3014 /**
3015  * 	netdev_completed_queue - report bytes and packets completed by device
3016  * 	@dev: network device
3017  * 	@pkts: actual number of packets sent over the medium
3018  * 	@bytes: actual number of bytes sent over the medium
3019  *
3020  * 	Report the number of bytes and packets transmitted by the network device
3021  * 	hardware queue over the physical medium, @bytes must exactly match the
3022  * 	@bytes amount passed to netdev_sent_queue()
3023  */
3024 static inline void netdev_completed_queue(struct net_device *dev,
3025 					  unsigned int pkts, unsigned int bytes)
3026 {
3027 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3028 }
3029 
3030 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3031 {
3032 #ifdef CONFIG_BQL
3033 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3034 	dql_reset(&q->dql);
3035 #endif
3036 }
3037 
3038 /**
3039  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3040  * 	@dev_queue: network device
3041  *
3042  * 	Reset the bytes and packet count of a network device and clear the
3043  * 	software flow control OFF bit for this network device
3044  */
3045 static inline void netdev_reset_queue(struct net_device *dev_queue)
3046 {
3047 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3048 }
3049 
3050 /**
3051  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3052  * 	@dev: network device
3053  * 	@queue_index: given tx queue index
3054  *
3055  * 	Returns 0 if given tx queue index >= number of device tx queues,
3056  * 	otherwise returns the originally passed tx queue index.
3057  */
3058 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3059 {
3060 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3061 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3062 				     dev->name, queue_index,
3063 				     dev->real_num_tx_queues);
3064 		return 0;
3065 	}
3066 
3067 	return queue_index;
3068 }
3069 
3070 /**
3071  *	netif_running - test if up
3072  *	@dev: network device
3073  *
3074  *	Test if the device has been brought up.
3075  */
3076 static inline bool netif_running(const struct net_device *dev)
3077 {
3078 	return test_bit(__LINK_STATE_START, &dev->state);
3079 }
3080 
3081 /*
3082  * Routines to manage the subqueues on a device.  We only need start,
3083  * stop, and a check if it's stopped.  All other device management is
3084  * done at the overall netdevice level.
3085  * Also test the device if we're multiqueue.
3086  */
3087 
3088 /**
3089  *	netif_start_subqueue - allow sending packets on subqueue
3090  *	@dev: network device
3091  *	@queue_index: sub queue index
3092  *
3093  * Start individual transmit queue of a device with multiple transmit queues.
3094  */
3095 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3096 {
3097 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3098 
3099 	netif_tx_start_queue(txq);
3100 }
3101 
3102 /**
3103  *	netif_stop_subqueue - stop sending packets on subqueue
3104  *	@dev: network device
3105  *	@queue_index: sub queue index
3106  *
3107  * Stop individual transmit queue of a device with multiple transmit queues.
3108  */
3109 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3110 {
3111 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3112 	netif_tx_stop_queue(txq);
3113 }
3114 
3115 /**
3116  *	netif_subqueue_stopped - test status of subqueue
3117  *	@dev: network device
3118  *	@queue_index: sub queue index
3119  *
3120  * Check individual transmit queue of a device with multiple transmit queues.
3121  */
3122 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3123 					    u16 queue_index)
3124 {
3125 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3126 
3127 	return netif_tx_queue_stopped(txq);
3128 }
3129 
3130 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3131 					  struct sk_buff *skb)
3132 {
3133 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3134 }
3135 
3136 /**
3137  *	netif_wake_subqueue - allow sending packets on subqueue
3138  *	@dev: network device
3139  *	@queue_index: sub queue index
3140  *
3141  * Resume individual transmit queue of a device with multiple transmit queues.
3142  */
3143 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3144 {
3145 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3146 
3147 	netif_tx_wake_queue(txq);
3148 }
3149 
3150 #ifdef CONFIG_XPS
3151 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3152 			u16 index);
3153 #else
3154 static inline int netif_set_xps_queue(struct net_device *dev,
3155 				      const struct cpumask *mask,
3156 				      u16 index)
3157 {
3158 	return 0;
3159 }
3160 #endif
3161 
3162 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3163 		  unsigned int num_tx_queues);
3164 
3165 /*
3166  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3167  * as a distribution range limit for the returned value.
3168  */
3169 static inline u16 skb_tx_hash(const struct net_device *dev,
3170 			      struct sk_buff *skb)
3171 {
3172 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3173 }
3174 
3175 /**
3176  *	netif_is_multiqueue - test if device has multiple transmit queues
3177  *	@dev: network device
3178  *
3179  * Check if device has multiple transmit queues
3180  */
3181 static inline bool netif_is_multiqueue(const struct net_device *dev)
3182 {
3183 	return dev->num_tx_queues > 1;
3184 }
3185 
3186 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3187 
3188 #ifdef CONFIG_SYSFS
3189 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3190 #else
3191 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3192 						unsigned int rxq)
3193 {
3194 	return 0;
3195 }
3196 #endif
3197 
3198 #ifdef CONFIG_SYSFS
3199 static inline unsigned int get_netdev_rx_queue_index(
3200 		struct netdev_rx_queue *queue)
3201 {
3202 	struct net_device *dev = queue->dev;
3203 	int index = queue - dev->_rx;
3204 
3205 	BUG_ON(index >= dev->num_rx_queues);
3206 	return index;
3207 }
3208 #endif
3209 
3210 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3211 int netif_get_num_default_rss_queues(void);
3212 
3213 enum skb_free_reason {
3214 	SKB_REASON_CONSUMED,
3215 	SKB_REASON_DROPPED,
3216 };
3217 
3218 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3219 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3220 
3221 /*
3222  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3223  * interrupt context or with hardware interrupts being disabled.
3224  * (in_irq() || irqs_disabled())
3225  *
3226  * We provide four helpers that can be used in following contexts :
3227  *
3228  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3229  *  replacing kfree_skb(skb)
3230  *
3231  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3232  *  Typically used in place of consume_skb(skb) in TX completion path
3233  *
3234  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3235  *  replacing kfree_skb(skb)
3236  *
3237  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3238  *  and consumed a packet. Used in place of consume_skb(skb)
3239  */
3240 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3241 {
3242 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3243 }
3244 
3245 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3246 {
3247 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3248 }
3249 
3250 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3251 {
3252 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3253 }
3254 
3255 static inline void dev_consume_skb_any(struct sk_buff *skb)
3256 {
3257 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3258 }
3259 
3260 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3261 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3262 int netif_rx(struct sk_buff *skb);
3263 int netif_rx_ni(struct sk_buff *skb);
3264 int netif_receive_skb(struct sk_buff *skb);
3265 int netif_receive_skb_core(struct sk_buff *skb);
3266 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3267 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3268 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3269 gro_result_t napi_gro_frags(struct napi_struct *napi);
3270 struct packet_offload *gro_find_receive_by_type(__be16 type);
3271 struct packet_offload *gro_find_complete_by_type(__be16 type);
3272 
3273 static inline void napi_free_frags(struct napi_struct *napi)
3274 {
3275 	kfree_skb(napi->skb);
3276 	napi->skb = NULL;
3277 }
3278 
3279 bool netdev_is_rx_handler_busy(struct net_device *dev);
3280 int netdev_rx_handler_register(struct net_device *dev,
3281 			       rx_handler_func_t *rx_handler,
3282 			       void *rx_handler_data);
3283 void netdev_rx_handler_unregister(struct net_device *dev);
3284 
3285 bool dev_valid_name(const char *name);
3286 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3287 int dev_ethtool(struct net *net, struct ifreq *);
3288 unsigned int dev_get_flags(const struct net_device *);
3289 int __dev_change_flags(struct net_device *, unsigned int flags);
3290 int dev_change_flags(struct net_device *, unsigned int);
3291 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3292 			unsigned int gchanges);
3293 int dev_change_name(struct net_device *, const char *);
3294 int dev_set_alias(struct net_device *, const char *, size_t);
3295 int dev_get_alias(const struct net_device *, char *, size_t);
3296 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3297 int __dev_set_mtu(struct net_device *, int);
3298 int dev_set_mtu(struct net_device *, int);
3299 void dev_set_group(struct net_device *, int);
3300 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3301 int dev_change_carrier(struct net_device *, bool new_carrier);
3302 int dev_get_phys_port_id(struct net_device *dev,
3303 			 struct netdev_phys_item_id *ppid);
3304 int dev_get_phys_port_name(struct net_device *dev,
3305 			   char *name, size_t len);
3306 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3307 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3308 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3309 				    struct netdev_queue *txq, int *ret);
3310 
3311 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3312 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3313 		      int fd, u32 flags);
3314 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3315 
3316 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3317 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3318 bool is_skb_forwardable(const struct net_device *dev,
3319 			const struct sk_buff *skb);
3320 
3321 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3322 					       struct sk_buff *skb)
3323 {
3324 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3325 	    unlikely(!is_skb_forwardable(dev, skb))) {
3326 		atomic_long_inc(&dev->rx_dropped);
3327 		kfree_skb(skb);
3328 		return NET_RX_DROP;
3329 	}
3330 
3331 	skb_scrub_packet(skb, true);
3332 	skb->priority = 0;
3333 	return 0;
3334 }
3335 
3336 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3337 
3338 extern int		netdev_budget;
3339 extern unsigned int	netdev_budget_usecs;
3340 
3341 /* Called by rtnetlink.c:rtnl_unlock() */
3342 void netdev_run_todo(void);
3343 
3344 /**
3345  *	dev_put - release reference to device
3346  *	@dev: network device
3347  *
3348  * Release reference to device to allow it to be freed.
3349  */
3350 static inline void dev_put(struct net_device *dev)
3351 {
3352 	this_cpu_dec(*dev->pcpu_refcnt);
3353 }
3354 
3355 /**
3356  *	dev_hold - get reference to device
3357  *	@dev: network device
3358  *
3359  * Hold reference to device to keep it from being freed.
3360  */
3361 static inline void dev_hold(struct net_device *dev)
3362 {
3363 	this_cpu_inc(*dev->pcpu_refcnt);
3364 }
3365 
3366 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3367  * and _off may be called from IRQ context, but it is caller
3368  * who is responsible for serialization of these calls.
3369  *
3370  * The name carrier is inappropriate, these functions should really be
3371  * called netif_lowerlayer_*() because they represent the state of any
3372  * kind of lower layer not just hardware media.
3373  */
3374 
3375 void linkwatch_init_dev(struct net_device *dev);
3376 void linkwatch_fire_event(struct net_device *dev);
3377 void linkwatch_forget_dev(struct net_device *dev);
3378 
3379 /**
3380  *	netif_carrier_ok - test if carrier present
3381  *	@dev: network device
3382  *
3383  * Check if carrier is present on device
3384  */
3385 static inline bool netif_carrier_ok(const struct net_device *dev)
3386 {
3387 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3388 }
3389 
3390 unsigned long dev_trans_start(struct net_device *dev);
3391 
3392 void __netdev_watchdog_up(struct net_device *dev);
3393 
3394 void netif_carrier_on(struct net_device *dev);
3395 
3396 void netif_carrier_off(struct net_device *dev);
3397 
3398 /**
3399  *	netif_dormant_on - mark device as dormant.
3400  *	@dev: network device
3401  *
3402  * Mark device as dormant (as per RFC2863).
3403  *
3404  * The dormant state indicates that the relevant interface is not
3405  * actually in a condition to pass packets (i.e., it is not 'up') but is
3406  * in a "pending" state, waiting for some external event.  For "on-
3407  * demand" interfaces, this new state identifies the situation where the
3408  * interface is waiting for events to place it in the up state.
3409  */
3410 static inline void netif_dormant_on(struct net_device *dev)
3411 {
3412 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3413 		linkwatch_fire_event(dev);
3414 }
3415 
3416 /**
3417  *	netif_dormant_off - set device as not dormant.
3418  *	@dev: network device
3419  *
3420  * Device is not in dormant state.
3421  */
3422 static inline void netif_dormant_off(struct net_device *dev)
3423 {
3424 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3425 		linkwatch_fire_event(dev);
3426 }
3427 
3428 /**
3429  *	netif_dormant - test if device is dormant
3430  *	@dev: network device
3431  *
3432  * Check if device is dormant.
3433  */
3434 static inline bool netif_dormant(const struct net_device *dev)
3435 {
3436 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3437 }
3438 
3439 
3440 /**
3441  *	netif_oper_up - test if device is operational
3442  *	@dev: network device
3443  *
3444  * Check if carrier is operational
3445  */
3446 static inline bool netif_oper_up(const struct net_device *dev)
3447 {
3448 	return (dev->operstate == IF_OPER_UP ||
3449 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3450 }
3451 
3452 /**
3453  *	netif_device_present - is device available or removed
3454  *	@dev: network device
3455  *
3456  * Check if device has not been removed from system.
3457  */
3458 static inline bool netif_device_present(struct net_device *dev)
3459 {
3460 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3461 }
3462 
3463 void netif_device_detach(struct net_device *dev);
3464 
3465 void netif_device_attach(struct net_device *dev);
3466 
3467 /*
3468  * Network interface message level settings
3469  */
3470 
3471 enum {
3472 	NETIF_MSG_DRV		= 0x0001,
3473 	NETIF_MSG_PROBE		= 0x0002,
3474 	NETIF_MSG_LINK		= 0x0004,
3475 	NETIF_MSG_TIMER		= 0x0008,
3476 	NETIF_MSG_IFDOWN	= 0x0010,
3477 	NETIF_MSG_IFUP		= 0x0020,
3478 	NETIF_MSG_RX_ERR	= 0x0040,
3479 	NETIF_MSG_TX_ERR	= 0x0080,
3480 	NETIF_MSG_TX_QUEUED	= 0x0100,
3481 	NETIF_MSG_INTR		= 0x0200,
3482 	NETIF_MSG_TX_DONE	= 0x0400,
3483 	NETIF_MSG_RX_STATUS	= 0x0800,
3484 	NETIF_MSG_PKTDATA	= 0x1000,
3485 	NETIF_MSG_HW		= 0x2000,
3486 	NETIF_MSG_WOL		= 0x4000,
3487 };
3488 
3489 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3490 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3491 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3492 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3493 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3494 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3495 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3496 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3497 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3498 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3499 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3500 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3501 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3502 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3503 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3504 
3505 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3506 {
3507 	/* use default */
3508 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3509 		return default_msg_enable_bits;
3510 	if (debug_value == 0)	/* no output */
3511 		return 0;
3512 	/* set low N bits */
3513 	return (1 << debug_value) - 1;
3514 }
3515 
3516 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3517 {
3518 	spin_lock(&txq->_xmit_lock);
3519 	txq->xmit_lock_owner = cpu;
3520 }
3521 
3522 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3523 {
3524 	__acquire(&txq->_xmit_lock);
3525 	return true;
3526 }
3527 
3528 static inline void __netif_tx_release(struct netdev_queue *txq)
3529 {
3530 	__release(&txq->_xmit_lock);
3531 }
3532 
3533 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3534 {
3535 	spin_lock_bh(&txq->_xmit_lock);
3536 	txq->xmit_lock_owner = smp_processor_id();
3537 }
3538 
3539 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3540 {
3541 	bool ok = spin_trylock(&txq->_xmit_lock);
3542 	if (likely(ok))
3543 		txq->xmit_lock_owner = smp_processor_id();
3544 	return ok;
3545 }
3546 
3547 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3548 {
3549 	txq->xmit_lock_owner = -1;
3550 	spin_unlock(&txq->_xmit_lock);
3551 }
3552 
3553 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3554 {
3555 	txq->xmit_lock_owner = -1;
3556 	spin_unlock_bh(&txq->_xmit_lock);
3557 }
3558 
3559 static inline void txq_trans_update(struct netdev_queue *txq)
3560 {
3561 	if (txq->xmit_lock_owner != -1)
3562 		txq->trans_start = jiffies;
3563 }
3564 
3565 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3566 static inline void netif_trans_update(struct net_device *dev)
3567 {
3568 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3569 
3570 	if (txq->trans_start != jiffies)
3571 		txq->trans_start = jiffies;
3572 }
3573 
3574 /**
3575  *	netif_tx_lock - grab network device transmit lock
3576  *	@dev: network device
3577  *
3578  * Get network device transmit lock
3579  */
3580 static inline void netif_tx_lock(struct net_device *dev)
3581 {
3582 	unsigned int i;
3583 	int cpu;
3584 
3585 	spin_lock(&dev->tx_global_lock);
3586 	cpu = smp_processor_id();
3587 	for (i = 0; i < dev->num_tx_queues; i++) {
3588 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3589 
3590 		/* We are the only thread of execution doing a
3591 		 * freeze, but we have to grab the _xmit_lock in
3592 		 * order to synchronize with threads which are in
3593 		 * the ->hard_start_xmit() handler and already
3594 		 * checked the frozen bit.
3595 		 */
3596 		__netif_tx_lock(txq, cpu);
3597 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3598 		__netif_tx_unlock(txq);
3599 	}
3600 }
3601 
3602 static inline void netif_tx_lock_bh(struct net_device *dev)
3603 {
3604 	local_bh_disable();
3605 	netif_tx_lock(dev);
3606 }
3607 
3608 static inline void netif_tx_unlock(struct net_device *dev)
3609 {
3610 	unsigned int i;
3611 
3612 	for (i = 0; i < dev->num_tx_queues; i++) {
3613 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3614 
3615 		/* No need to grab the _xmit_lock here.  If the
3616 		 * queue is not stopped for another reason, we
3617 		 * force a schedule.
3618 		 */
3619 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3620 		netif_schedule_queue(txq);
3621 	}
3622 	spin_unlock(&dev->tx_global_lock);
3623 }
3624 
3625 static inline void netif_tx_unlock_bh(struct net_device *dev)
3626 {
3627 	netif_tx_unlock(dev);
3628 	local_bh_enable();
3629 }
3630 
3631 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3632 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3633 		__netif_tx_lock(txq, cpu);		\
3634 	} else {					\
3635 		__netif_tx_acquire(txq);		\
3636 	}						\
3637 }
3638 
3639 #define HARD_TX_TRYLOCK(dev, txq)			\
3640 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3641 		__netif_tx_trylock(txq) :		\
3642 		__netif_tx_acquire(txq))
3643 
3644 #define HARD_TX_UNLOCK(dev, txq) {			\
3645 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3646 		__netif_tx_unlock(txq);			\
3647 	} else {					\
3648 		__netif_tx_release(txq);		\
3649 	}						\
3650 }
3651 
3652 static inline void netif_tx_disable(struct net_device *dev)
3653 {
3654 	unsigned int i;
3655 	int cpu;
3656 
3657 	local_bh_disable();
3658 	cpu = smp_processor_id();
3659 	for (i = 0; i < dev->num_tx_queues; i++) {
3660 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3661 
3662 		__netif_tx_lock(txq, cpu);
3663 		netif_tx_stop_queue(txq);
3664 		__netif_tx_unlock(txq);
3665 	}
3666 	local_bh_enable();
3667 }
3668 
3669 static inline void netif_addr_lock(struct net_device *dev)
3670 {
3671 	spin_lock(&dev->addr_list_lock);
3672 }
3673 
3674 static inline void netif_addr_lock_nested(struct net_device *dev)
3675 {
3676 	int subclass = SINGLE_DEPTH_NESTING;
3677 
3678 	if (dev->netdev_ops->ndo_get_lock_subclass)
3679 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3680 
3681 	spin_lock_nested(&dev->addr_list_lock, subclass);
3682 }
3683 
3684 static inline void netif_addr_lock_bh(struct net_device *dev)
3685 {
3686 	spin_lock_bh(&dev->addr_list_lock);
3687 }
3688 
3689 static inline void netif_addr_unlock(struct net_device *dev)
3690 {
3691 	spin_unlock(&dev->addr_list_lock);
3692 }
3693 
3694 static inline void netif_addr_unlock_bh(struct net_device *dev)
3695 {
3696 	spin_unlock_bh(&dev->addr_list_lock);
3697 }
3698 
3699 /*
3700  * dev_addrs walker. Should be used only for read access. Call with
3701  * rcu_read_lock held.
3702  */
3703 #define for_each_dev_addr(dev, ha) \
3704 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3705 
3706 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3707 
3708 void ether_setup(struct net_device *dev);
3709 
3710 /* Support for loadable net-drivers */
3711 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3712 				    unsigned char name_assign_type,
3713 				    void (*setup)(struct net_device *),
3714 				    unsigned int txqs, unsigned int rxqs);
3715 int dev_get_valid_name(struct net *net, struct net_device *dev,
3716 		       const char *name);
3717 
3718 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3719 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3720 
3721 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3722 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3723 			 count)
3724 
3725 int register_netdev(struct net_device *dev);
3726 void unregister_netdev(struct net_device *dev);
3727 
3728 /* General hardware address lists handling functions */
3729 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3730 		   struct netdev_hw_addr_list *from_list, int addr_len);
3731 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3732 		      struct netdev_hw_addr_list *from_list, int addr_len);
3733 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3734 		       struct net_device *dev,
3735 		       int (*sync)(struct net_device *, const unsigned char *),
3736 		       int (*unsync)(struct net_device *,
3737 				     const unsigned char *));
3738 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3739 			  struct net_device *dev,
3740 			  int (*unsync)(struct net_device *,
3741 					const unsigned char *));
3742 void __hw_addr_init(struct netdev_hw_addr_list *list);
3743 
3744 /* Functions used for device addresses handling */
3745 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3746 		 unsigned char addr_type);
3747 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3748 		 unsigned char addr_type);
3749 void dev_addr_flush(struct net_device *dev);
3750 int dev_addr_init(struct net_device *dev);
3751 
3752 /* Functions used for unicast addresses handling */
3753 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3754 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3755 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3756 int dev_uc_sync(struct net_device *to, struct net_device *from);
3757 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3758 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3759 void dev_uc_flush(struct net_device *dev);
3760 void dev_uc_init(struct net_device *dev);
3761 
3762 /**
3763  *  __dev_uc_sync - Synchonize device's unicast list
3764  *  @dev:  device to sync
3765  *  @sync: function to call if address should be added
3766  *  @unsync: function to call if address should be removed
3767  *
3768  *  Add newly added addresses to the interface, and release
3769  *  addresses that have been deleted.
3770  */
3771 static inline int __dev_uc_sync(struct net_device *dev,
3772 				int (*sync)(struct net_device *,
3773 					    const unsigned char *),
3774 				int (*unsync)(struct net_device *,
3775 					      const unsigned char *))
3776 {
3777 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3778 }
3779 
3780 /**
3781  *  __dev_uc_unsync - Remove synchronized addresses from device
3782  *  @dev:  device to sync
3783  *  @unsync: function to call if address should be removed
3784  *
3785  *  Remove all addresses that were added to the device by dev_uc_sync().
3786  */
3787 static inline void __dev_uc_unsync(struct net_device *dev,
3788 				   int (*unsync)(struct net_device *,
3789 						 const unsigned char *))
3790 {
3791 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3792 }
3793 
3794 /* Functions used for multicast addresses handling */
3795 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3796 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3797 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3798 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3799 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3800 int dev_mc_sync(struct net_device *to, struct net_device *from);
3801 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3802 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3803 void dev_mc_flush(struct net_device *dev);
3804 void dev_mc_init(struct net_device *dev);
3805 
3806 /**
3807  *  __dev_mc_sync - Synchonize device's multicast list
3808  *  @dev:  device to sync
3809  *  @sync: function to call if address should be added
3810  *  @unsync: function to call if address should be removed
3811  *
3812  *  Add newly added addresses to the interface, and release
3813  *  addresses that have been deleted.
3814  */
3815 static inline int __dev_mc_sync(struct net_device *dev,
3816 				int (*sync)(struct net_device *,
3817 					    const unsigned char *),
3818 				int (*unsync)(struct net_device *,
3819 					      const unsigned char *))
3820 {
3821 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3822 }
3823 
3824 /**
3825  *  __dev_mc_unsync - Remove synchronized addresses from device
3826  *  @dev:  device to sync
3827  *  @unsync: function to call if address should be removed
3828  *
3829  *  Remove all addresses that were added to the device by dev_mc_sync().
3830  */
3831 static inline void __dev_mc_unsync(struct net_device *dev,
3832 				   int (*unsync)(struct net_device *,
3833 						 const unsigned char *))
3834 {
3835 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3836 }
3837 
3838 /* Functions used for secondary unicast and multicast support */
3839 void dev_set_rx_mode(struct net_device *dev);
3840 void __dev_set_rx_mode(struct net_device *dev);
3841 int dev_set_promiscuity(struct net_device *dev, int inc);
3842 int dev_set_allmulti(struct net_device *dev, int inc);
3843 void netdev_state_change(struct net_device *dev);
3844 void netdev_notify_peers(struct net_device *dev);
3845 void netdev_features_change(struct net_device *dev);
3846 /* Load a device via the kmod */
3847 void dev_load(struct net *net, const char *name);
3848 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3849 					struct rtnl_link_stats64 *storage);
3850 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3851 			     const struct net_device_stats *netdev_stats);
3852 
3853 extern int		netdev_max_backlog;
3854 extern int		netdev_tstamp_prequeue;
3855 extern int		weight_p;
3856 extern int		dev_weight_rx_bias;
3857 extern int		dev_weight_tx_bias;
3858 extern int		dev_rx_weight;
3859 extern int		dev_tx_weight;
3860 
3861 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3862 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3863 						     struct list_head **iter);
3864 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3865 						     struct list_head **iter);
3866 
3867 /* iterate through upper list, must be called under RCU read lock */
3868 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3869 	for (iter = &(dev)->adj_list.upper, \
3870 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3871 	     updev; \
3872 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3873 
3874 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3875 				  int (*fn)(struct net_device *upper_dev,
3876 					    void *data),
3877 				  void *data);
3878 
3879 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3880 				  struct net_device *upper_dev);
3881 
3882 bool netdev_has_any_upper_dev(struct net_device *dev);
3883 
3884 void *netdev_lower_get_next_private(struct net_device *dev,
3885 				    struct list_head **iter);
3886 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3887 					struct list_head **iter);
3888 
3889 #define netdev_for_each_lower_private(dev, priv, iter) \
3890 	for (iter = (dev)->adj_list.lower.next, \
3891 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3892 	     priv; \
3893 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3894 
3895 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3896 	for (iter = &(dev)->adj_list.lower, \
3897 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3898 	     priv; \
3899 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3900 
3901 void *netdev_lower_get_next(struct net_device *dev,
3902 				struct list_head **iter);
3903 
3904 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3905 	for (iter = (dev)->adj_list.lower.next, \
3906 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3907 	     ldev; \
3908 	     ldev = netdev_lower_get_next(dev, &(iter)))
3909 
3910 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3911 					     struct list_head **iter);
3912 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3913 						 struct list_head **iter);
3914 
3915 int netdev_walk_all_lower_dev(struct net_device *dev,
3916 			      int (*fn)(struct net_device *lower_dev,
3917 					void *data),
3918 			      void *data);
3919 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3920 				  int (*fn)(struct net_device *lower_dev,
3921 					    void *data),
3922 				  void *data);
3923 
3924 void *netdev_adjacent_get_private(struct list_head *adj_list);
3925 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3926 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3927 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3928 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3929 			  struct netlink_ext_ack *extack);
3930 int netdev_master_upper_dev_link(struct net_device *dev,
3931 				 struct net_device *upper_dev,
3932 				 void *upper_priv, void *upper_info,
3933 				 struct netlink_ext_ack *extack);
3934 void netdev_upper_dev_unlink(struct net_device *dev,
3935 			     struct net_device *upper_dev);
3936 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3937 void *netdev_lower_dev_get_private(struct net_device *dev,
3938 				   struct net_device *lower_dev);
3939 void netdev_lower_state_changed(struct net_device *lower_dev,
3940 				void *lower_state_info);
3941 
3942 /* RSS keys are 40 or 52 bytes long */
3943 #define NETDEV_RSS_KEY_LEN 52
3944 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3945 void netdev_rss_key_fill(void *buffer, size_t len);
3946 
3947 int dev_get_nest_level(struct net_device *dev);
3948 int skb_checksum_help(struct sk_buff *skb);
3949 int skb_crc32c_csum_help(struct sk_buff *skb);
3950 int skb_csum_hwoffload_help(struct sk_buff *skb,
3951 			    const netdev_features_t features);
3952 
3953 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3954 				  netdev_features_t features, bool tx_path);
3955 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3956 				    netdev_features_t features);
3957 
3958 struct netdev_bonding_info {
3959 	ifslave	slave;
3960 	ifbond	master;
3961 };
3962 
3963 struct netdev_notifier_bonding_info {
3964 	struct netdev_notifier_info info; /* must be first */
3965 	struct netdev_bonding_info  bonding_info;
3966 };
3967 
3968 void netdev_bonding_info_change(struct net_device *dev,
3969 				struct netdev_bonding_info *bonding_info);
3970 
3971 static inline
3972 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3973 {
3974 	return __skb_gso_segment(skb, features, true);
3975 }
3976 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3977 
3978 static inline bool can_checksum_protocol(netdev_features_t features,
3979 					 __be16 protocol)
3980 {
3981 	if (protocol == htons(ETH_P_FCOE))
3982 		return !!(features & NETIF_F_FCOE_CRC);
3983 
3984 	/* Assume this is an IP checksum (not SCTP CRC) */
3985 
3986 	if (features & NETIF_F_HW_CSUM) {
3987 		/* Can checksum everything */
3988 		return true;
3989 	}
3990 
3991 	switch (protocol) {
3992 	case htons(ETH_P_IP):
3993 		return !!(features & NETIF_F_IP_CSUM);
3994 	case htons(ETH_P_IPV6):
3995 		return !!(features & NETIF_F_IPV6_CSUM);
3996 	default:
3997 		return false;
3998 	}
3999 }
4000 
4001 #ifdef CONFIG_BUG
4002 void netdev_rx_csum_fault(struct net_device *dev);
4003 #else
4004 static inline void netdev_rx_csum_fault(struct net_device *dev)
4005 {
4006 }
4007 #endif
4008 /* rx skb timestamps */
4009 void net_enable_timestamp(void);
4010 void net_disable_timestamp(void);
4011 
4012 #ifdef CONFIG_PROC_FS
4013 int __init dev_proc_init(void);
4014 #else
4015 #define dev_proc_init() 0
4016 #endif
4017 
4018 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4019 					      struct sk_buff *skb, struct net_device *dev,
4020 					      bool more)
4021 {
4022 	skb->xmit_more = more ? 1 : 0;
4023 	return ops->ndo_start_xmit(skb, dev);
4024 }
4025 
4026 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4027 					    struct netdev_queue *txq, bool more)
4028 {
4029 	const struct net_device_ops *ops = dev->netdev_ops;
4030 	int rc;
4031 
4032 	rc = __netdev_start_xmit(ops, skb, dev, more);
4033 	if (rc == NETDEV_TX_OK)
4034 		txq_trans_update(txq);
4035 
4036 	return rc;
4037 }
4038 
4039 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4040 				const void *ns);
4041 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4042 				 const void *ns);
4043 
4044 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4045 {
4046 	return netdev_class_create_file_ns(class_attr, NULL);
4047 }
4048 
4049 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4050 {
4051 	netdev_class_remove_file_ns(class_attr, NULL);
4052 }
4053 
4054 extern const struct kobj_ns_type_operations net_ns_type_operations;
4055 
4056 const char *netdev_drivername(const struct net_device *dev);
4057 
4058 void linkwatch_run_queue(void);
4059 
4060 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4061 							  netdev_features_t f2)
4062 {
4063 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4064 		if (f1 & NETIF_F_HW_CSUM)
4065 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4066 		else
4067 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4068 	}
4069 
4070 	return f1 & f2;
4071 }
4072 
4073 static inline netdev_features_t netdev_get_wanted_features(
4074 	struct net_device *dev)
4075 {
4076 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4077 }
4078 netdev_features_t netdev_increment_features(netdev_features_t all,
4079 	netdev_features_t one, netdev_features_t mask);
4080 
4081 /* Allow TSO being used on stacked device :
4082  * Performing the GSO segmentation before last device
4083  * is a performance improvement.
4084  */
4085 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4086 							netdev_features_t mask)
4087 {
4088 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4089 }
4090 
4091 int __netdev_update_features(struct net_device *dev);
4092 void netdev_update_features(struct net_device *dev);
4093 void netdev_change_features(struct net_device *dev);
4094 
4095 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4096 					struct net_device *dev);
4097 
4098 netdev_features_t passthru_features_check(struct sk_buff *skb,
4099 					  struct net_device *dev,
4100 					  netdev_features_t features);
4101 netdev_features_t netif_skb_features(struct sk_buff *skb);
4102 
4103 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4104 {
4105 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4106 
4107 	/* check flags correspondence */
4108 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4109 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4110 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4111 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4112 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4113 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4114 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4115 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4116 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4117 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4118 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4119 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4120 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4121 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4122 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4123 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4124 
4125 	return (features & feature) == feature;
4126 }
4127 
4128 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4129 {
4130 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4131 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4132 }
4133 
4134 static inline bool netif_needs_gso(struct sk_buff *skb,
4135 				   netdev_features_t features)
4136 {
4137 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4138 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4139 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4140 }
4141 
4142 static inline void netif_set_gso_max_size(struct net_device *dev,
4143 					  unsigned int size)
4144 {
4145 	dev->gso_max_size = size;
4146 }
4147 
4148 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4149 					int pulled_hlen, u16 mac_offset,
4150 					int mac_len)
4151 {
4152 	skb->protocol = protocol;
4153 	skb->encapsulation = 1;
4154 	skb_push(skb, pulled_hlen);
4155 	skb_reset_transport_header(skb);
4156 	skb->mac_header = mac_offset;
4157 	skb->network_header = skb->mac_header + mac_len;
4158 	skb->mac_len = mac_len;
4159 }
4160 
4161 static inline bool netif_is_macsec(const struct net_device *dev)
4162 {
4163 	return dev->priv_flags & IFF_MACSEC;
4164 }
4165 
4166 static inline bool netif_is_macvlan(const struct net_device *dev)
4167 {
4168 	return dev->priv_flags & IFF_MACVLAN;
4169 }
4170 
4171 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4172 {
4173 	return dev->priv_flags & IFF_MACVLAN_PORT;
4174 }
4175 
4176 static inline bool netif_is_ipvlan(const struct net_device *dev)
4177 {
4178 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4179 }
4180 
4181 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4182 {
4183 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4184 }
4185 
4186 static inline bool netif_is_bond_master(const struct net_device *dev)
4187 {
4188 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4189 }
4190 
4191 static inline bool netif_is_bond_slave(const struct net_device *dev)
4192 {
4193 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4194 }
4195 
4196 static inline bool netif_supports_nofcs(struct net_device *dev)
4197 {
4198 	return dev->priv_flags & IFF_SUPP_NOFCS;
4199 }
4200 
4201 static inline bool netif_is_l3_master(const struct net_device *dev)
4202 {
4203 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4204 }
4205 
4206 static inline bool netif_is_l3_slave(const struct net_device *dev)
4207 {
4208 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4209 }
4210 
4211 static inline bool netif_is_bridge_master(const struct net_device *dev)
4212 {
4213 	return dev->priv_flags & IFF_EBRIDGE;
4214 }
4215 
4216 static inline bool netif_is_bridge_port(const struct net_device *dev)
4217 {
4218 	return dev->priv_flags & IFF_BRIDGE_PORT;
4219 }
4220 
4221 static inline bool netif_is_ovs_master(const struct net_device *dev)
4222 {
4223 	return dev->priv_flags & IFF_OPENVSWITCH;
4224 }
4225 
4226 static inline bool netif_is_ovs_port(const struct net_device *dev)
4227 {
4228 	return dev->priv_flags & IFF_OVS_DATAPATH;
4229 }
4230 
4231 static inline bool netif_is_team_master(const struct net_device *dev)
4232 {
4233 	return dev->priv_flags & IFF_TEAM;
4234 }
4235 
4236 static inline bool netif_is_team_port(const struct net_device *dev)
4237 {
4238 	return dev->priv_flags & IFF_TEAM_PORT;
4239 }
4240 
4241 static inline bool netif_is_lag_master(const struct net_device *dev)
4242 {
4243 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4244 }
4245 
4246 static inline bool netif_is_lag_port(const struct net_device *dev)
4247 {
4248 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4249 }
4250 
4251 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4252 {
4253 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4254 }
4255 
4256 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4257 static inline void netif_keep_dst(struct net_device *dev)
4258 {
4259 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4260 }
4261 
4262 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4263 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4264 {
4265 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4266 	return dev->priv_flags & IFF_MACSEC;
4267 }
4268 
4269 extern struct pernet_operations __net_initdata loopback_net_ops;
4270 
4271 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4272 
4273 /* netdev_printk helpers, similar to dev_printk */
4274 
4275 static inline const char *netdev_name(const struct net_device *dev)
4276 {
4277 	if (!dev->name[0] || strchr(dev->name, '%'))
4278 		return "(unnamed net_device)";
4279 	return dev->name;
4280 }
4281 
4282 static inline bool netdev_unregistering(const struct net_device *dev)
4283 {
4284 	return dev->reg_state == NETREG_UNREGISTERING;
4285 }
4286 
4287 static inline const char *netdev_reg_state(const struct net_device *dev)
4288 {
4289 	switch (dev->reg_state) {
4290 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4291 	case NETREG_REGISTERED: return "";
4292 	case NETREG_UNREGISTERING: return " (unregistering)";
4293 	case NETREG_UNREGISTERED: return " (unregistered)";
4294 	case NETREG_RELEASED: return " (released)";
4295 	case NETREG_DUMMY: return " (dummy)";
4296 	}
4297 
4298 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4299 	return " (unknown)";
4300 }
4301 
4302 __printf(3, 4)
4303 void netdev_printk(const char *level, const struct net_device *dev,
4304 		   const char *format, ...);
4305 __printf(2, 3)
4306 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4307 __printf(2, 3)
4308 void netdev_alert(const struct net_device *dev, const char *format, ...);
4309 __printf(2, 3)
4310 void netdev_crit(const struct net_device *dev, const char *format, ...);
4311 __printf(2, 3)
4312 void netdev_err(const struct net_device *dev, const char *format, ...);
4313 __printf(2, 3)
4314 void netdev_warn(const struct net_device *dev, const char *format, ...);
4315 __printf(2, 3)
4316 void netdev_notice(const struct net_device *dev, const char *format, ...);
4317 __printf(2, 3)
4318 void netdev_info(const struct net_device *dev, const char *format, ...);
4319 
4320 #define MODULE_ALIAS_NETDEV(device) \
4321 	MODULE_ALIAS("netdev-" device)
4322 
4323 #if defined(CONFIG_DYNAMIC_DEBUG)
4324 #define netdev_dbg(__dev, format, args...)			\
4325 do {								\
4326 	dynamic_netdev_dbg(__dev, format, ##args);		\
4327 } while (0)
4328 #elif defined(DEBUG)
4329 #define netdev_dbg(__dev, format, args...)			\
4330 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4331 #else
4332 #define netdev_dbg(__dev, format, args...)			\
4333 ({								\
4334 	if (0)							\
4335 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4336 })
4337 #endif
4338 
4339 #if defined(VERBOSE_DEBUG)
4340 #define netdev_vdbg	netdev_dbg
4341 #else
4342 
4343 #define netdev_vdbg(dev, format, args...)			\
4344 ({								\
4345 	if (0)							\
4346 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4347 	0;							\
4348 })
4349 #endif
4350 
4351 /*
4352  * netdev_WARN() acts like dev_printk(), but with the key difference
4353  * of using a WARN/WARN_ON to get the message out, including the
4354  * file/line information and a backtrace.
4355  */
4356 #define netdev_WARN(dev, format, args...)			\
4357 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4358 	     netdev_reg_state(dev), ##args)
4359 
4360 /* netif printk helpers, similar to netdev_printk */
4361 
4362 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4363 do {					  			\
4364 	if (netif_msg_##type(priv))				\
4365 		netdev_printk(level, (dev), fmt, ##args);	\
4366 } while (0)
4367 
4368 #define netif_level(level, priv, type, dev, fmt, args...)	\
4369 do {								\
4370 	if (netif_msg_##type(priv))				\
4371 		netdev_##level(dev, fmt, ##args);		\
4372 } while (0)
4373 
4374 #define netif_emerg(priv, type, dev, fmt, args...)		\
4375 	netif_level(emerg, priv, type, dev, fmt, ##args)
4376 #define netif_alert(priv, type, dev, fmt, args...)		\
4377 	netif_level(alert, priv, type, dev, fmt, ##args)
4378 #define netif_crit(priv, type, dev, fmt, args...)		\
4379 	netif_level(crit, priv, type, dev, fmt, ##args)
4380 #define netif_err(priv, type, dev, fmt, args...)		\
4381 	netif_level(err, priv, type, dev, fmt, ##args)
4382 #define netif_warn(priv, type, dev, fmt, args...)		\
4383 	netif_level(warn, priv, type, dev, fmt, ##args)
4384 #define netif_notice(priv, type, dev, fmt, args...)		\
4385 	netif_level(notice, priv, type, dev, fmt, ##args)
4386 #define netif_info(priv, type, dev, fmt, args...)		\
4387 	netif_level(info, priv, type, dev, fmt, ##args)
4388 
4389 #if defined(CONFIG_DYNAMIC_DEBUG)
4390 #define netif_dbg(priv, type, netdev, format, args...)		\
4391 do {								\
4392 	if (netif_msg_##type(priv))				\
4393 		dynamic_netdev_dbg(netdev, format, ##args);	\
4394 } while (0)
4395 #elif defined(DEBUG)
4396 #define netif_dbg(priv, type, dev, format, args...)		\
4397 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4398 #else
4399 #define netif_dbg(priv, type, dev, format, args...)			\
4400 ({									\
4401 	if (0)								\
4402 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4403 	0;								\
4404 })
4405 #endif
4406 
4407 /* if @cond then downgrade to debug, else print at @level */
4408 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4409 	do {                                                              \
4410 		if (cond)                                                 \
4411 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4412 		else                                                      \
4413 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4414 	} while (0)
4415 
4416 #if defined(VERBOSE_DEBUG)
4417 #define netif_vdbg	netif_dbg
4418 #else
4419 #define netif_vdbg(priv, type, dev, format, args...)		\
4420 ({								\
4421 	if (0)							\
4422 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4423 	0;							\
4424 })
4425 #endif
4426 
4427 /*
4428  *	The list of packet types we will receive (as opposed to discard)
4429  *	and the routines to invoke.
4430  *
4431  *	Why 16. Because with 16 the only overlap we get on a hash of the
4432  *	low nibble of the protocol value is RARP/SNAP/X.25.
4433  *
4434  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4435  *             sure which should go first, but I bet it won't make much
4436  *             difference if we are running VLANs.  The good news is that
4437  *             this protocol won't be in the list unless compiled in, so
4438  *             the average user (w/out VLANs) will not be adversely affected.
4439  *             --BLG
4440  *
4441  *		0800	IP
4442  *		8100    802.1Q VLAN
4443  *		0001	802.3
4444  *		0002	AX.25
4445  *		0004	802.2
4446  *		8035	RARP
4447  *		0005	SNAP
4448  *		0805	X.25
4449  *		0806	ARP
4450  *		8137	IPX
4451  *		0009	Localtalk
4452  *		86DD	IPv6
4453  */
4454 #define PTYPE_HASH_SIZE	(16)
4455 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4456 
4457 #endif	/* _LINUX_NETDEVICE_H */
4458