]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/netdevice.h
net: Add asynchronous callbacks for xfrm on layer 2.
[mirror_ubuntu-jammy-kernel.git] / include / linux / netdevice.h
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, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
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_QDISC_MQPRIO,
774 TC_SETUP_CLSU32,
775 TC_SETUP_CLSFLOWER,
776 TC_SETUP_CLSMATCHALL,
777 TC_SETUP_CLSBPF,
778 TC_SETUP_BLOCK,
779 TC_SETUP_QDISC_CBS,
780 TC_SETUP_QDISC_RED,
781 };
782
783 /* These structures hold the attributes of bpf state that are being passed
784 * to the netdevice through the bpf op.
785 */
786 enum bpf_netdev_command {
787 /* Set or clear a bpf program used in the earliest stages of packet
788 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
789 * is responsible for calling bpf_prog_put on any old progs that are
790 * stored. In case of error, the callee need not release the new prog
791 * reference, but on success it takes ownership and must bpf_prog_put
792 * when it is no longer used.
793 */
794 XDP_SETUP_PROG,
795 XDP_SETUP_PROG_HW,
796 /* Check if a bpf program is set on the device. The callee should
797 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
798 * is equivalent to XDP_ATTACHED_DRV.
799 */
800 XDP_QUERY_PROG,
801 /* BPF program for offload callbacks, invoked at program load time. */
802 BPF_OFFLOAD_VERIFIER_PREP,
803 BPF_OFFLOAD_TRANSLATE,
804 BPF_OFFLOAD_DESTROY,
805 };
806
807 struct bpf_ext_analyzer_ops;
808 struct netlink_ext_ack;
809
810 struct netdev_bpf {
811 enum bpf_netdev_command command;
812 union {
813 /* XDP_SETUP_PROG */
814 struct {
815 u32 flags;
816 struct bpf_prog *prog;
817 struct netlink_ext_ack *extack;
818 };
819 /* XDP_QUERY_PROG */
820 struct {
821 u8 prog_attached;
822 u32 prog_id;
823 /* flags with which program was installed */
824 u32 prog_flags;
825 };
826 /* BPF_OFFLOAD_VERIFIER_PREP */
827 struct {
828 struct bpf_prog *prog;
829 const struct bpf_ext_analyzer_ops *ops; /* callee set */
830 } verifier;
831 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
832 struct {
833 struct bpf_prog *prog;
834 } offload;
835 };
836 };
837
838 #ifdef CONFIG_XFRM_OFFLOAD
839 struct xfrmdev_ops {
840 int (*xdo_dev_state_add) (struct xfrm_state *x);
841 void (*xdo_dev_state_delete) (struct xfrm_state *x);
842 void (*xdo_dev_state_free) (struct xfrm_state *x);
843 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
844 struct xfrm_state *x);
845 };
846 #endif
847
848 struct dev_ifalias {
849 struct rcu_head rcuhead;
850 char ifalias[];
851 };
852
853 /*
854 * This structure defines the management hooks for network devices.
855 * The following hooks can be defined; unless noted otherwise, they are
856 * optional and can be filled with a null pointer.
857 *
858 * int (*ndo_init)(struct net_device *dev);
859 * This function is called once when a network device is registered.
860 * The network device can use this for any late stage initialization
861 * or semantic validation. It can fail with an error code which will
862 * be propagated back to register_netdev.
863 *
864 * void (*ndo_uninit)(struct net_device *dev);
865 * This function is called when device is unregistered or when registration
866 * fails. It is not called if init fails.
867 *
868 * int (*ndo_open)(struct net_device *dev);
869 * This function is called when a network device transitions to the up
870 * state.
871 *
872 * int (*ndo_stop)(struct net_device *dev);
873 * This function is called when a network device transitions to the down
874 * state.
875 *
876 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
877 * struct net_device *dev);
878 * Called when a packet needs to be transmitted.
879 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
880 * the queue before that can happen; it's for obsolete devices and weird
881 * corner cases, but the stack really does a non-trivial amount
882 * of useless work if you return NETDEV_TX_BUSY.
883 * Required; cannot be NULL.
884 *
885 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
886 * struct net_device *dev
887 * netdev_features_t features);
888 * Called by core transmit path to determine if device is capable of
889 * performing offload operations on a given packet. This is to give
890 * the device an opportunity to implement any restrictions that cannot
891 * be otherwise expressed by feature flags. The check is called with
892 * the set of features that the stack has calculated and it returns
893 * those the driver believes to be appropriate.
894 *
895 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
896 * void *accel_priv, select_queue_fallback_t fallback);
897 * Called to decide which queue to use when device supports multiple
898 * transmit queues.
899 *
900 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
901 * This function is called to allow device receiver to make
902 * changes to configuration when multicast or promiscuous is enabled.
903 *
904 * void (*ndo_set_rx_mode)(struct net_device *dev);
905 * This function is called device changes address list filtering.
906 * If driver handles unicast address filtering, it should set
907 * IFF_UNICAST_FLT in its priv_flags.
908 *
909 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
910 * This function is called when the Media Access Control address
911 * needs to be changed. If this interface is not defined, the
912 * MAC address can not be changed.
913 *
914 * int (*ndo_validate_addr)(struct net_device *dev);
915 * Test if Media Access Control address is valid for the device.
916 *
917 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
918 * Called when a user requests an ioctl which can't be handled by
919 * the generic interface code. If not defined ioctls return
920 * not supported error code.
921 *
922 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
923 * Used to set network devices bus interface parameters. This interface
924 * is retained for legacy reasons; new devices should use the bus
925 * interface (PCI) for low level management.
926 *
927 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
928 * Called when a user wants to change the Maximum Transfer Unit
929 * of a device.
930 *
931 * void (*ndo_tx_timeout)(struct net_device *dev);
932 * Callback used when the transmitter has not made any progress
933 * for dev->watchdog ticks.
934 *
935 * void (*ndo_get_stats64)(struct net_device *dev,
936 * struct rtnl_link_stats64 *storage);
937 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
938 * Called when a user wants to get the network device usage
939 * statistics. Drivers must do one of the following:
940 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
941 * rtnl_link_stats64 structure passed by the caller.
942 * 2. Define @ndo_get_stats to update a net_device_stats structure
943 * (which should normally be dev->stats) and return a pointer to
944 * it. The structure may be changed asynchronously only if each
945 * field is written atomically.
946 * 3. Update dev->stats asynchronously and atomically, and define
947 * neither operation.
948 *
949 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
950 * Return true if this device supports offload stats of this attr_id.
951 *
952 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
953 * void *attr_data)
954 * Get statistics for offload operations by attr_id. Write it into the
955 * attr_data pointer.
956 *
957 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
958 * If device supports VLAN filtering this function is called when a
959 * VLAN id is registered.
960 *
961 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
962 * If device supports VLAN filtering this function is called when a
963 * VLAN id is unregistered.
964 *
965 * void (*ndo_poll_controller)(struct net_device *dev);
966 *
967 * SR-IOV management functions.
968 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
969 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
970 * u8 qos, __be16 proto);
971 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
972 * int max_tx_rate);
973 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
974 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
975 * int (*ndo_get_vf_config)(struct net_device *dev,
976 * int vf, struct ifla_vf_info *ivf);
977 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
978 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
979 * struct nlattr *port[]);
980 *
981 * Enable or disable the VF ability to query its RSS Redirection Table and
982 * Hash Key. This is needed since on some devices VF share this information
983 * with PF and querying it may introduce a theoretical security risk.
984 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
985 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
986 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
987 * void *type_data);
988 * Called to setup any 'tc' scheduler, classifier or action on @dev.
989 * This is always called from the stack with the rtnl lock held and netif
990 * tx queues stopped. This allows the netdevice to perform queue
991 * management safely.
992 *
993 * Fiber Channel over Ethernet (FCoE) offload functions.
994 * int (*ndo_fcoe_enable)(struct net_device *dev);
995 * Called when the FCoE protocol stack wants to start using LLD for FCoE
996 * so the underlying device can perform whatever needed configuration or
997 * initialization to support acceleration of FCoE traffic.
998 *
999 * int (*ndo_fcoe_disable)(struct net_device *dev);
1000 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1001 * so the underlying device can perform whatever needed clean-ups to
1002 * stop supporting acceleration of FCoE traffic.
1003 *
1004 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1005 * struct scatterlist *sgl, unsigned int sgc);
1006 * Called when the FCoE Initiator wants to initialize an I/O that
1007 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1008 * perform necessary setup and returns 1 to indicate the device is set up
1009 * successfully to perform DDP on this I/O, otherwise this returns 0.
1010 *
1011 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1012 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1013 * indicated by the FC exchange id 'xid', so the underlying device can
1014 * clean up and reuse resources for later DDP requests.
1015 *
1016 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1017 * struct scatterlist *sgl, unsigned int sgc);
1018 * Called when the FCoE Target wants to initialize an I/O that
1019 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1020 * perform necessary setup and returns 1 to indicate the device is set up
1021 * successfully to perform DDP on this I/O, otherwise this returns 0.
1022 *
1023 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1024 * struct netdev_fcoe_hbainfo *hbainfo);
1025 * Called when the FCoE Protocol stack wants information on the underlying
1026 * device. This information is utilized by the FCoE protocol stack to
1027 * register attributes with Fiber Channel management service as per the
1028 * FC-GS Fabric Device Management Information(FDMI) specification.
1029 *
1030 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1031 * Called when the underlying device wants to override default World Wide
1032 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1033 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1034 * protocol stack to use.
1035 *
1036 * RFS acceleration.
1037 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1038 * u16 rxq_index, u32 flow_id);
1039 * Set hardware filter for RFS. rxq_index is the target queue index;
1040 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1041 * Return the filter ID on success, or a negative error code.
1042 *
1043 * Slave management functions (for bridge, bonding, etc).
1044 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1045 * Called to make another netdev an underling.
1046 *
1047 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1048 * Called to release previously enslaved netdev.
1049 *
1050 * Feature/offload setting functions.
1051 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1052 * netdev_features_t features);
1053 * Adjusts the requested feature flags according to device-specific
1054 * constraints, and returns the resulting flags. Must not modify
1055 * the device state.
1056 *
1057 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1058 * Called to update device configuration to new features. Passed
1059 * feature set might be less than what was returned by ndo_fix_features()).
1060 * Must return >0 or -errno if it changed dev->features itself.
1061 *
1062 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1063 * struct net_device *dev,
1064 * const unsigned char *addr, u16 vid, u16 flags)
1065 * Adds an FDB entry to dev for addr.
1066 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1067 * struct net_device *dev,
1068 * const unsigned char *addr, u16 vid)
1069 * Deletes the FDB entry from dev coresponding to addr.
1070 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1071 * struct net_device *dev, struct net_device *filter_dev,
1072 * int *idx)
1073 * Used to add FDB entries to dump requests. Implementers should add
1074 * entries to skb and update idx with the number of entries.
1075 *
1076 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1077 * u16 flags)
1078 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1079 * struct net_device *dev, u32 filter_mask,
1080 * int nlflags)
1081 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1082 * u16 flags);
1083 *
1084 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1085 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1086 * which do not represent real hardware may define this to allow their
1087 * userspace components to manage their virtual carrier state. Devices
1088 * that determine carrier state from physical hardware properties (eg
1089 * network cables) or protocol-dependent mechanisms (eg
1090 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1091 *
1092 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1093 * struct netdev_phys_item_id *ppid);
1094 * Called to get ID of physical port of this device. If driver does
1095 * not implement this, it is assumed that the hw is not able to have
1096 * multiple net devices on single physical port.
1097 *
1098 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1099 * struct udp_tunnel_info *ti);
1100 * Called by UDP tunnel to notify a driver about the UDP port and socket
1101 * address family that a UDP tunnel is listnening to. It is called only
1102 * when a new port starts listening. The operation is protected by the
1103 * RTNL.
1104 *
1105 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1106 * struct udp_tunnel_info *ti);
1107 * Called by UDP tunnel to notify the driver about a UDP port and socket
1108 * address family that the UDP tunnel is not listening to anymore. The
1109 * operation is protected by the RTNL.
1110 *
1111 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1112 * struct net_device *dev)
1113 * Called by upper layer devices to accelerate switching or other
1114 * station functionality into hardware. 'pdev is the lowerdev
1115 * to use for the offload and 'dev' is the net device that will
1116 * back the offload. Returns a pointer to the private structure
1117 * the upper layer will maintain.
1118 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1119 * Called by upper layer device to delete the station created
1120 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1121 * the station and priv is the structure returned by the add
1122 * operation.
1123 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1124 * int queue_index, u32 maxrate);
1125 * Called when a user wants to set a max-rate limitation of specific
1126 * TX queue.
1127 * int (*ndo_get_iflink)(const struct net_device *dev);
1128 * Called to get the iflink value of this device.
1129 * void (*ndo_change_proto_down)(struct net_device *dev,
1130 * bool proto_down);
1131 * This function is used to pass protocol port error state information
1132 * to the switch driver. The switch driver can react to the proto_down
1133 * by doing a phys down on the associated switch port.
1134 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1135 * This function is used to get egress tunnel information for given skb.
1136 * This is useful for retrieving outer tunnel header parameters while
1137 * sampling packet.
1138 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1139 * This function is used to specify the headroom that the skb must
1140 * consider when allocation skb during packet reception. Setting
1141 * appropriate rx headroom value allows avoiding skb head copy on
1142 * forward. Setting a negative value resets the rx headroom to the
1143 * default value.
1144 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1145 * This function is used to set or query state related to XDP on the
1146 * netdevice and manage BPF offload. See definition of
1147 * enum bpf_netdev_command for details.
1148 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1149 * This function is used to submit a XDP packet for transmit on a
1150 * netdevice.
1151 * void (*ndo_xdp_flush)(struct net_device *dev);
1152 * This function is used to inform the driver to flush a particular
1153 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1154 */
1155 struct net_device_ops {
1156 int (*ndo_init)(struct net_device *dev);
1157 void (*ndo_uninit)(struct net_device *dev);
1158 int (*ndo_open)(struct net_device *dev);
1159 int (*ndo_stop)(struct net_device *dev);
1160 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1161 struct net_device *dev);
1162 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1163 struct net_device *dev,
1164 netdev_features_t features);
1165 u16 (*ndo_select_queue)(struct net_device *dev,
1166 struct sk_buff *skb,
1167 void *accel_priv,
1168 select_queue_fallback_t fallback);
1169 void (*ndo_change_rx_flags)(struct net_device *dev,
1170 int flags);
1171 void (*ndo_set_rx_mode)(struct net_device *dev);
1172 int (*ndo_set_mac_address)(struct net_device *dev,
1173 void *addr);
1174 int (*ndo_validate_addr)(struct net_device *dev);
1175 int (*ndo_do_ioctl)(struct net_device *dev,
1176 struct ifreq *ifr, int cmd);
1177 int (*ndo_set_config)(struct net_device *dev,
1178 struct ifmap *map);
1179 int (*ndo_change_mtu)(struct net_device *dev,
1180 int new_mtu);
1181 int (*ndo_neigh_setup)(struct net_device *dev,
1182 struct neigh_parms *);
1183 void (*ndo_tx_timeout) (struct net_device *dev);
1184
1185 void (*ndo_get_stats64)(struct net_device *dev,
1186 struct rtnl_link_stats64 *storage);
1187 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1188 int (*ndo_get_offload_stats)(int attr_id,
1189 const struct net_device *dev,
1190 void *attr_data);
1191 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1192
1193 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1194 __be16 proto, u16 vid);
1195 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1196 __be16 proto, u16 vid);
1197 #ifdef CONFIG_NET_POLL_CONTROLLER
1198 void (*ndo_poll_controller)(struct net_device *dev);
1199 int (*ndo_netpoll_setup)(struct net_device *dev,
1200 struct netpoll_info *info);
1201 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1202 #endif
1203 int (*ndo_set_vf_mac)(struct net_device *dev,
1204 int queue, u8 *mac);
1205 int (*ndo_set_vf_vlan)(struct net_device *dev,
1206 int queue, u16 vlan,
1207 u8 qos, __be16 proto);
1208 int (*ndo_set_vf_rate)(struct net_device *dev,
1209 int vf, int min_tx_rate,
1210 int max_tx_rate);
1211 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1212 int vf, bool setting);
1213 int (*ndo_set_vf_trust)(struct net_device *dev,
1214 int vf, bool setting);
1215 int (*ndo_get_vf_config)(struct net_device *dev,
1216 int vf,
1217 struct ifla_vf_info *ivf);
1218 int (*ndo_set_vf_link_state)(struct net_device *dev,
1219 int vf, int link_state);
1220 int (*ndo_get_vf_stats)(struct net_device *dev,
1221 int vf,
1222 struct ifla_vf_stats
1223 *vf_stats);
1224 int (*ndo_set_vf_port)(struct net_device *dev,
1225 int vf,
1226 struct nlattr *port[]);
1227 int (*ndo_get_vf_port)(struct net_device *dev,
1228 int vf, struct sk_buff *skb);
1229 int (*ndo_set_vf_guid)(struct net_device *dev,
1230 int vf, u64 guid,
1231 int guid_type);
1232 int (*ndo_set_vf_rss_query_en)(
1233 struct net_device *dev,
1234 int vf, bool setting);
1235 int (*ndo_setup_tc)(struct net_device *dev,
1236 enum tc_setup_type type,
1237 void *type_data);
1238 #if IS_ENABLED(CONFIG_FCOE)
1239 int (*ndo_fcoe_enable)(struct net_device *dev);
1240 int (*ndo_fcoe_disable)(struct net_device *dev);
1241 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1242 u16 xid,
1243 struct scatterlist *sgl,
1244 unsigned int sgc);
1245 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1246 u16 xid);
1247 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1248 u16 xid,
1249 struct scatterlist *sgl,
1250 unsigned int sgc);
1251 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1252 struct netdev_fcoe_hbainfo *hbainfo);
1253 #endif
1254
1255 #if IS_ENABLED(CONFIG_LIBFCOE)
1256 #define NETDEV_FCOE_WWNN 0
1257 #define NETDEV_FCOE_WWPN 1
1258 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1259 u64 *wwn, int type);
1260 #endif
1261
1262 #ifdef CONFIG_RFS_ACCEL
1263 int (*ndo_rx_flow_steer)(struct net_device *dev,
1264 const struct sk_buff *skb,
1265 u16 rxq_index,
1266 u32 flow_id);
1267 #endif
1268 int (*ndo_add_slave)(struct net_device *dev,
1269 struct net_device *slave_dev,
1270 struct netlink_ext_ack *extack);
1271 int (*ndo_del_slave)(struct net_device *dev,
1272 struct net_device *slave_dev);
1273 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1274 netdev_features_t features);
1275 int (*ndo_set_features)(struct net_device *dev,
1276 netdev_features_t features);
1277 int (*ndo_neigh_construct)(struct net_device *dev,
1278 struct neighbour *n);
1279 void (*ndo_neigh_destroy)(struct net_device *dev,
1280 struct neighbour *n);
1281
1282 int (*ndo_fdb_add)(struct ndmsg *ndm,
1283 struct nlattr *tb[],
1284 struct net_device *dev,
1285 const unsigned char *addr,
1286 u16 vid,
1287 u16 flags);
1288 int (*ndo_fdb_del)(struct ndmsg *ndm,
1289 struct nlattr *tb[],
1290 struct net_device *dev,
1291 const unsigned char *addr,
1292 u16 vid);
1293 int (*ndo_fdb_dump)(struct sk_buff *skb,
1294 struct netlink_callback *cb,
1295 struct net_device *dev,
1296 struct net_device *filter_dev,
1297 int *idx);
1298
1299 int (*ndo_bridge_setlink)(struct net_device *dev,
1300 struct nlmsghdr *nlh,
1301 u16 flags);
1302 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1303 u32 pid, u32 seq,
1304 struct net_device *dev,
1305 u32 filter_mask,
1306 int nlflags);
1307 int (*ndo_bridge_dellink)(struct net_device *dev,
1308 struct nlmsghdr *nlh,
1309 u16 flags);
1310 int (*ndo_change_carrier)(struct net_device *dev,
1311 bool new_carrier);
1312 int (*ndo_get_phys_port_id)(struct net_device *dev,
1313 struct netdev_phys_item_id *ppid);
1314 int (*ndo_get_phys_port_name)(struct net_device *dev,
1315 char *name, size_t len);
1316 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1317 struct udp_tunnel_info *ti);
1318 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1319 struct udp_tunnel_info *ti);
1320 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1321 struct net_device *dev);
1322 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1323 void *priv);
1324
1325 int (*ndo_get_lock_subclass)(struct net_device *dev);
1326 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1327 int queue_index,
1328 u32 maxrate);
1329 int (*ndo_get_iflink)(const struct net_device *dev);
1330 int (*ndo_change_proto_down)(struct net_device *dev,
1331 bool proto_down);
1332 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1333 struct sk_buff *skb);
1334 void (*ndo_set_rx_headroom)(struct net_device *dev,
1335 int needed_headroom);
1336 int (*ndo_bpf)(struct net_device *dev,
1337 struct netdev_bpf *bpf);
1338 int (*ndo_xdp_xmit)(struct net_device *dev,
1339 struct xdp_buff *xdp);
1340 void (*ndo_xdp_flush)(struct net_device *dev);
1341 };
1342
1343 /**
1344 * enum net_device_priv_flags - &struct net_device priv_flags
1345 *
1346 * These are the &struct net_device, they are only set internally
1347 * by drivers and used in the kernel. These flags are invisible to
1348 * userspace; this means that the order of these flags can change
1349 * during any kernel release.
1350 *
1351 * You should have a pretty good reason to be extending these flags.
1352 *
1353 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1354 * @IFF_EBRIDGE: Ethernet bridging device
1355 * @IFF_BONDING: bonding master or slave
1356 * @IFF_ISATAP: ISATAP interface (RFC4214)
1357 * @IFF_WAN_HDLC: WAN HDLC device
1358 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1359 * release skb->dst
1360 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1361 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1362 * @IFF_MACVLAN_PORT: device used as macvlan port
1363 * @IFF_BRIDGE_PORT: device used as bridge port
1364 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1365 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1366 * @IFF_UNICAST_FLT: Supports unicast filtering
1367 * @IFF_TEAM_PORT: device used as team port
1368 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1369 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1370 * change when it's running
1371 * @IFF_MACVLAN: Macvlan device
1372 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1373 * underlying stacked devices
1374 * @IFF_IPVLAN_MASTER: IPvlan master device
1375 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1376 * @IFF_L3MDEV_MASTER: device is an L3 master device
1377 * @IFF_NO_QUEUE: device can run without qdisc attached
1378 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1379 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1380 * @IFF_TEAM: device is a team device
1381 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1382 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1383 * entity (i.e. the master device for bridged veth)
1384 * @IFF_MACSEC: device is a MACsec device
1385 */
1386 enum netdev_priv_flags {
1387 IFF_802_1Q_VLAN = 1<<0,
1388 IFF_EBRIDGE = 1<<1,
1389 IFF_BONDING = 1<<2,
1390 IFF_ISATAP = 1<<3,
1391 IFF_WAN_HDLC = 1<<4,
1392 IFF_XMIT_DST_RELEASE = 1<<5,
1393 IFF_DONT_BRIDGE = 1<<6,
1394 IFF_DISABLE_NETPOLL = 1<<7,
1395 IFF_MACVLAN_PORT = 1<<8,
1396 IFF_BRIDGE_PORT = 1<<9,
1397 IFF_OVS_DATAPATH = 1<<10,
1398 IFF_TX_SKB_SHARING = 1<<11,
1399 IFF_UNICAST_FLT = 1<<12,
1400 IFF_TEAM_PORT = 1<<13,
1401 IFF_SUPP_NOFCS = 1<<14,
1402 IFF_LIVE_ADDR_CHANGE = 1<<15,
1403 IFF_MACVLAN = 1<<16,
1404 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1405 IFF_IPVLAN_MASTER = 1<<18,
1406 IFF_IPVLAN_SLAVE = 1<<19,
1407 IFF_L3MDEV_MASTER = 1<<20,
1408 IFF_NO_QUEUE = 1<<21,
1409 IFF_OPENVSWITCH = 1<<22,
1410 IFF_L3MDEV_SLAVE = 1<<23,
1411 IFF_TEAM = 1<<24,
1412 IFF_RXFH_CONFIGURED = 1<<25,
1413 IFF_PHONY_HEADROOM = 1<<26,
1414 IFF_MACSEC = 1<<27,
1415 };
1416
1417 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1418 #define IFF_EBRIDGE IFF_EBRIDGE
1419 #define IFF_BONDING IFF_BONDING
1420 #define IFF_ISATAP IFF_ISATAP
1421 #define IFF_WAN_HDLC IFF_WAN_HDLC
1422 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1423 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1424 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1425 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1426 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1427 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1428 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1429 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1430 #define IFF_TEAM_PORT IFF_TEAM_PORT
1431 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1432 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1433 #define IFF_MACVLAN IFF_MACVLAN
1434 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1435 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1436 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1437 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1438 #define IFF_NO_QUEUE IFF_NO_QUEUE
1439 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1440 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1441 #define IFF_TEAM IFF_TEAM
1442 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1443 #define IFF_MACSEC IFF_MACSEC
1444
1445 /**
1446 * struct net_device - The DEVICE structure.
1447 *
1448 * Actually, this whole structure is a big mistake. It mixes I/O
1449 * data with strictly "high-level" data, and it has to know about
1450 * almost every data structure used in the INET module.
1451 *
1452 * @name: This is the first field of the "visible" part of this structure
1453 * (i.e. as seen by users in the "Space.c" file). It is the name
1454 * of the interface.
1455 *
1456 * @name_hlist: Device name hash chain, please keep it close to name[]
1457 * @ifalias: SNMP alias
1458 * @mem_end: Shared memory end
1459 * @mem_start: Shared memory start
1460 * @base_addr: Device I/O address
1461 * @irq: Device IRQ number
1462 *
1463 * @carrier_changes: Stats to monitor carrier on<->off transitions
1464 *
1465 * @state: Generic network queuing layer state, see netdev_state_t
1466 * @dev_list: The global list of network devices
1467 * @napi_list: List entry used for polling NAPI devices
1468 * @unreg_list: List entry when we are unregistering the
1469 * device; see the function unregister_netdev
1470 * @close_list: List entry used when we are closing the device
1471 * @ptype_all: Device-specific packet handlers for all protocols
1472 * @ptype_specific: Device-specific, protocol-specific packet handlers
1473 *
1474 * @adj_list: Directly linked devices, like slaves for bonding
1475 * @features: Currently active device features
1476 * @hw_features: User-changeable features
1477 *
1478 * @wanted_features: User-requested features
1479 * @vlan_features: Mask of features inheritable by VLAN devices
1480 *
1481 * @hw_enc_features: Mask of features inherited by encapsulating devices
1482 * This field indicates what encapsulation
1483 * offloads the hardware is capable of doing,
1484 * and drivers will need to set them appropriately.
1485 *
1486 * @mpls_features: Mask of features inheritable by MPLS
1487 *
1488 * @ifindex: interface index
1489 * @group: The group the device belongs to
1490 *
1491 * @stats: Statistics struct, which was left as a legacy, use
1492 * rtnl_link_stats64 instead
1493 *
1494 * @rx_dropped: Dropped packets by core network,
1495 * do not use this in drivers
1496 * @tx_dropped: Dropped packets by core network,
1497 * do not use this in drivers
1498 * @rx_nohandler: nohandler dropped packets by core network on
1499 * inactive devices, do not use this in drivers
1500 *
1501 * @wireless_handlers: List of functions to handle Wireless Extensions,
1502 * instead of ioctl,
1503 * see <net/iw_handler.h> for details.
1504 * @wireless_data: Instance data managed by the core of wireless extensions
1505 *
1506 * @netdev_ops: Includes several pointers to callbacks,
1507 * if one wants to override the ndo_*() functions
1508 * @ethtool_ops: Management operations
1509 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1510 * discovery handling. Necessary for e.g. 6LoWPAN.
1511 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1512 * of Layer 2 headers.
1513 *
1514 * @flags: Interface flags (a la BSD)
1515 * @priv_flags: Like 'flags' but invisible to userspace,
1516 * see if.h for the definitions
1517 * @gflags: Global flags ( kept as legacy )
1518 * @padded: How much padding added by alloc_netdev()
1519 * @operstate: RFC2863 operstate
1520 * @link_mode: Mapping policy to operstate
1521 * @if_port: Selectable AUI, TP, ...
1522 * @dma: DMA channel
1523 * @mtu: Interface MTU value
1524 * @min_mtu: Interface Minimum MTU value
1525 * @max_mtu: Interface Maximum MTU value
1526 * @type: Interface hardware type
1527 * @hard_header_len: Maximum hardware header length.
1528 * @min_header_len: Minimum hardware header length
1529 *
1530 * @needed_headroom: Extra headroom the hardware may need, but not in all
1531 * cases can this be guaranteed
1532 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1533 * cases can this be guaranteed. Some cases also use
1534 * LL_MAX_HEADER instead to allocate the skb
1535 *
1536 * interface address info:
1537 *
1538 * @perm_addr: Permanent hw address
1539 * @addr_assign_type: Hw address assignment type
1540 * @addr_len: Hardware address length
1541 * @neigh_priv_len: Used in neigh_alloc()
1542 * @dev_id: Used to differentiate devices that share
1543 * the same link layer address
1544 * @dev_port: Used to differentiate devices that share
1545 * the same function
1546 * @addr_list_lock: XXX: need comments on this one
1547 * @uc_promisc: Counter that indicates promiscuous mode
1548 * has been enabled due to the need to listen to
1549 * additional unicast addresses in a device that
1550 * does not implement ndo_set_rx_mode()
1551 * @uc: unicast mac addresses
1552 * @mc: multicast mac addresses
1553 * @dev_addrs: list of device hw addresses
1554 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1555 * @promiscuity: Number of times the NIC is told to work in
1556 * promiscuous mode; if it becomes 0 the NIC will
1557 * exit promiscuous mode
1558 * @allmulti: Counter, enables or disables allmulticast mode
1559 *
1560 * @vlan_info: VLAN info
1561 * @dsa_ptr: dsa specific data
1562 * @tipc_ptr: TIPC specific data
1563 * @atalk_ptr: AppleTalk link
1564 * @ip_ptr: IPv4 specific data
1565 * @dn_ptr: DECnet specific data
1566 * @ip6_ptr: IPv6 specific data
1567 * @ax25_ptr: AX.25 specific data
1568 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1569 *
1570 * @dev_addr: Hw address (before bcast,
1571 * because most packets are unicast)
1572 *
1573 * @_rx: Array of RX queues
1574 * @num_rx_queues: Number of RX queues
1575 * allocated at register_netdev() time
1576 * @real_num_rx_queues: Number of RX queues currently active in device
1577 *
1578 * @rx_handler: handler for received packets
1579 * @rx_handler_data: XXX: need comments on this one
1580 * @miniq_ingress: ingress/clsact qdisc specific data for
1581 * ingress processing
1582 * @ingress_queue: XXX: need comments on this one
1583 * @broadcast: hw bcast address
1584 *
1585 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1586 * indexed by RX queue number. Assigned by driver.
1587 * This must only be set if the ndo_rx_flow_steer
1588 * operation is defined
1589 * @index_hlist: Device index hash chain
1590 *
1591 * @_tx: Array of TX queues
1592 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1593 * @real_num_tx_queues: Number of TX queues currently active in device
1594 * @qdisc: Root qdisc from userspace point of view
1595 * @tx_queue_len: Max frames per queue allowed
1596 * @tx_global_lock: XXX: need comments on this one
1597 *
1598 * @xps_maps: XXX: need comments on this one
1599 * @miniq_egress: clsact qdisc specific data for
1600 * egress processing
1601 * @watchdog_timeo: Represents the timeout that is used by
1602 * the watchdog (see dev_watchdog())
1603 * @watchdog_timer: List of timers
1604 *
1605 * @pcpu_refcnt: Number of references to this device
1606 * @todo_list: Delayed register/unregister
1607 * @link_watch_list: XXX: need comments on this one
1608 *
1609 * @reg_state: Register/unregister state machine
1610 * @dismantle: Device is going to be freed
1611 * @rtnl_link_state: This enum represents the phases of creating
1612 * a new link
1613 *
1614 * @needs_free_netdev: Should unregister perform free_netdev?
1615 * @priv_destructor: Called from unregister
1616 * @npinfo: XXX: need comments on this one
1617 * @nd_net: Network namespace this network device is inside
1618 *
1619 * @ml_priv: Mid-layer private
1620 * @lstats: Loopback statistics
1621 * @tstats: Tunnel statistics
1622 * @dstats: Dummy statistics
1623 * @vstats: Virtual ethernet statistics
1624 *
1625 * @garp_port: GARP
1626 * @mrp_port: MRP
1627 *
1628 * @dev: Class/net/name entry
1629 * @sysfs_groups: Space for optional device, statistics and wireless
1630 * sysfs groups
1631 *
1632 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1633 * @rtnl_link_ops: Rtnl_link_ops
1634 *
1635 * @gso_max_size: Maximum size of generic segmentation offload
1636 * @gso_max_segs: Maximum number of segments that can be passed to the
1637 * NIC for GSO
1638 *
1639 * @dcbnl_ops: Data Center Bridging netlink ops
1640 * @num_tc: Number of traffic classes in the net device
1641 * @tc_to_txq: XXX: need comments on this one
1642 * @prio_tc_map: XXX: need comments on this one
1643 *
1644 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1645 *
1646 * @priomap: XXX: need comments on this one
1647 * @phydev: Physical device may attach itself
1648 * for hardware timestamping
1649 *
1650 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1651 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1652 *
1653 * @proto_down: protocol port state information can be sent to the
1654 * switch driver and used to set the phys state of the
1655 * switch port.
1656 *
1657 * FIXME: cleanup struct net_device such that network protocol info
1658 * moves out.
1659 */
1660
1661 struct net_device {
1662 char name[IFNAMSIZ];
1663 struct hlist_node name_hlist;
1664 struct dev_ifalias __rcu *ifalias;
1665 /*
1666 * I/O specific fields
1667 * FIXME: Merge these and struct ifmap into one
1668 */
1669 unsigned long mem_end;
1670 unsigned long mem_start;
1671 unsigned long base_addr;
1672 int irq;
1673
1674 atomic_t carrier_changes;
1675
1676 /*
1677 * Some hardware also needs these fields (state,dev_list,
1678 * napi_list,unreg_list,close_list) but they are not
1679 * part of the usual set specified in Space.c.
1680 */
1681
1682 unsigned long state;
1683
1684 struct list_head dev_list;
1685 struct list_head napi_list;
1686 struct list_head unreg_list;
1687 struct list_head close_list;
1688 struct list_head ptype_all;
1689 struct list_head ptype_specific;
1690
1691 struct {
1692 struct list_head upper;
1693 struct list_head lower;
1694 } adj_list;
1695
1696 netdev_features_t features;
1697 netdev_features_t hw_features;
1698 netdev_features_t wanted_features;
1699 netdev_features_t vlan_features;
1700 netdev_features_t hw_enc_features;
1701 netdev_features_t mpls_features;
1702 netdev_features_t gso_partial_features;
1703
1704 int ifindex;
1705 int group;
1706
1707 struct net_device_stats stats;
1708
1709 atomic_long_t rx_dropped;
1710 atomic_long_t tx_dropped;
1711 atomic_long_t rx_nohandler;
1712
1713 #ifdef CONFIG_WIRELESS_EXT
1714 const struct iw_handler_def *wireless_handlers;
1715 struct iw_public_data *wireless_data;
1716 #endif
1717 const struct net_device_ops *netdev_ops;
1718 const struct ethtool_ops *ethtool_ops;
1719 #ifdef CONFIG_NET_SWITCHDEV
1720 const struct switchdev_ops *switchdev_ops;
1721 #endif
1722 #ifdef CONFIG_NET_L3_MASTER_DEV
1723 const struct l3mdev_ops *l3mdev_ops;
1724 #endif
1725 #if IS_ENABLED(CONFIG_IPV6)
1726 const struct ndisc_ops *ndisc_ops;
1727 #endif
1728
1729 #ifdef CONFIG_XFRM
1730 const struct xfrmdev_ops *xfrmdev_ops;
1731 #endif
1732
1733 const struct header_ops *header_ops;
1734
1735 unsigned int flags;
1736 unsigned int priv_flags;
1737
1738 unsigned short gflags;
1739 unsigned short padded;
1740
1741 unsigned char operstate;
1742 unsigned char link_mode;
1743
1744 unsigned char if_port;
1745 unsigned char dma;
1746
1747 unsigned int mtu;
1748 unsigned int min_mtu;
1749 unsigned int max_mtu;
1750 unsigned short type;
1751 unsigned short hard_header_len;
1752 unsigned char min_header_len;
1753
1754 unsigned short needed_headroom;
1755 unsigned short needed_tailroom;
1756
1757 /* Interface address info. */
1758 unsigned char perm_addr[MAX_ADDR_LEN];
1759 unsigned char addr_assign_type;
1760 unsigned char addr_len;
1761 unsigned short neigh_priv_len;
1762 unsigned short dev_id;
1763 unsigned short dev_port;
1764 spinlock_t addr_list_lock;
1765 unsigned char name_assign_type;
1766 bool uc_promisc;
1767 struct netdev_hw_addr_list uc;
1768 struct netdev_hw_addr_list mc;
1769 struct netdev_hw_addr_list dev_addrs;
1770
1771 #ifdef CONFIG_SYSFS
1772 struct kset *queues_kset;
1773 #endif
1774 unsigned int promiscuity;
1775 unsigned int allmulti;
1776
1777
1778 /* Protocol-specific pointers */
1779
1780 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1781 struct vlan_info __rcu *vlan_info;
1782 #endif
1783 #if IS_ENABLED(CONFIG_NET_DSA)
1784 struct dsa_port *dsa_ptr;
1785 #endif
1786 #if IS_ENABLED(CONFIG_TIPC)
1787 struct tipc_bearer __rcu *tipc_ptr;
1788 #endif
1789 void *atalk_ptr;
1790 struct in_device __rcu *ip_ptr;
1791 struct dn_dev __rcu *dn_ptr;
1792 struct inet6_dev __rcu *ip6_ptr;
1793 void *ax25_ptr;
1794 struct wireless_dev *ieee80211_ptr;
1795 struct wpan_dev *ieee802154_ptr;
1796 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1797 struct mpls_dev __rcu *mpls_ptr;
1798 #endif
1799
1800 /*
1801 * Cache lines mostly used on receive path (including eth_type_trans())
1802 */
1803 /* Interface address info used in eth_type_trans() */
1804 unsigned char *dev_addr;
1805
1806 #ifdef CONFIG_SYSFS
1807 struct netdev_rx_queue *_rx;
1808
1809 unsigned int num_rx_queues;
1810 unsigned int real_num_rx_queues;
1811 #endif
1812
1813 struct bpf_prog __rcu *xdp_prog;
1814 unsigned long gro_flush_timeout;
1815 rx_handler_func_t __rcu *rx_handler;
1816 void __rcu *rx_handler_data;
1817
1818 #ifdef CONFIG_NET_CLS_ACT
1819 struct mini_Qdisc __rcu *miniq_ingress;
1820 #endif
1821 struct netdev_queue __rcu *ingress_queue;
1822 #ifdef CONFIG_NETFILTER_INGRESS
1823 struct nf_hook_entries __rcu *nf_hooks_ingress;
1824 #endif
1825
1826 unsigned char broadcast[MAX_ADDR_LEN];
1827 #ifdef CONFIG_RFS_ACCEL
1828 struct cpu_rmap *rx_cpu_rmap;
1829 #endif
1830 struct hlist_node index_hlist;
1831
1832 /*
1833 * Cache lines mostly used on transmit path
1834 */
1835 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1836 unsigned int num_tx_queues;
1837 unsigned int real_num_tx_queues;
1838 struct Qdisc *qdisc;
1839 #ifdef CONFIG_NET_SCHED
1840 DECLARE_HASHTABLE (qdisc_hash, 4);
1841 #endif
1842 unsigned int tx_queue_len;
1843 spinlock_t tx_global_lock;
1844 int watchdog_timeo;
1845
1846 #ifdef CONFIG_XPS
1847 struct xps_dev_maps __rcu *xps_maps;
1848 #endif
1849 #ifdef CONFIG_NET_CLS_ACT
1850 struct mini_Qdisc __rcu *miniq_egress;
1851 #endif
1852
1853 /* These may be needed for future network-power-down code. */
1854 struct timer_list watchdog_timer;
1855
1856 int __percpu *pcpu_refcnt;
1857 struct list_head todo_list;
1858
1859 struct list_head link_watch_list;
1860
1861 enum { NETREG_UNINITIALIZED=0,
1862 NETREG_REGISTERED, /* completed register_netdevice */
1863 NETREG_UNREGISTERING, /* called unregister_netdevice */
1864 NETREG_UNREGISTERED, /* completed unregister todo */
1865 NETREG_RELEASED, /* called free_netdev */
1866 NETREG_DUMMY, /* dummy device for NAPI poll */
1867 } reg_state:8;
1868
1869 bool dismantle;
1870
1871 enum {
1872 RTNL_LINK_INITIALIZED,
1873 RTNL_LINK_INITIALIZING,
1874 } rtnl_link_state:16;
1875
1876 bool needs_free_netdev;
1877 void (*priv_destructor)(struct net_device *dev);
1878
1879 #ifdef CONFIG_NETPOLL
1880 struct netpoll_info __rcu *npinfo;
1881 #endif
1882
1883 possible_net_t nd_net;
1884
1885 /* mid-layer private */
1886 union {
1887 void *ml_priv;
1888 struct pcpu_lstats __percpu *lstats;
1889 struct pcpu_sw_netstats __percpu *tstats;
1890 struct pcpu_dstats __percpu *dstats;
1891 struct pcpu_vstats __percpu *vstats;
1892 };
1893
1894 #if IS_ENABLED(CONFIG_GARP)
1895 struct garp_port __rcu *garp_port;
1896 #endif
1897 #if IS_ENABLED(CONFIG_MRP)
1898 struct mrp_port __rcu *mrp_port;
1899 #endif
1900
1901 struct device dev;
1902 const struct attribute_group *sysfs_groups[4];
1903 const struct attribute_group *sysfs_rx_queue_group;
1904
1905 const struct rtnl_link_ops *rtnl_link_ops;
1906
1907 /* for setting kernel sock attribute on TCP connection setup */
1908 #define GSO_MAX_SIZE 65536
1909 unsigned int gso_max_size;
1910 #define GSO_MAX_SEGS 65535
1911 u16 gso_max_segs;
1912
1913 #ifdef CONFIG_DCB
1914 const struct dcbnl_rtnl_ops *dcbnl_ops;
1915 #endif
1916 u8 num_tc;
1917 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1918 u8 prio_tc_map[TC_BITMASK + 1];
1919
1920 #if IS_ENABLED(CONFIG_FCOE)
1921 unsigned int fcoe_ddp_xid;
1922 #endif
1923 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1924 struct netprio_map __rcu *priomap;
1925 #endif
1926 struct phy_device *phydev;
1927 struct lock_class_key *qdisc_tx_busylock;
1928 struct lock_class_key *qdisc_running_key;
1929 bool proto_down;
1930 };
1931 #define to_net_dev(d) container_of(d, struct net_device, dev)
1932
1933 static inline bool netif_elide_gro(const struct net_device *dev)
1934 {
1935 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1936 return true;
1937 return false;
1938 }
1939
1940 #define NETDEV_ALIGN 32
1941
1942 static inline
1943 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1944 {
1945 return dev->prio_tc_map[prio & TC_BITMASK];
1946 }
1947
1948 static inline
1949 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1950 {
1951 if (tc >= dev->num_tc)
1952 return -EINVAL;
1953
1954 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1955 return 0;
1956 }
1957
1958 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1959 void netdev_reset_tc(struct net_device *dev);
1960 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1961 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1962
1963 static inline
1964 int netdev_get_num_tc(struct net_device *dev)
1965 {
1966 return dev->num_tc;
1967 }
1968
1969 static inline
1970 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1971 unsigned int index)
1972 {
1973 return &dev->_tx[index];
1974 }
1975
1976 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1977 const struct sk_buff *skb)
1978 {
1979 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1980 }
1981
1982 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1983 void (*f)(struct net_device *,
1984 struct netdev_queue *,
1985 void *),
1986 void *arg)
1987 {
1988 unsigned int i;
1989
1990 for (i = 0; i < dev->num_tx_queues; i++)
1991 f(dev, &dev->_tx[i], arg);
1992 }
1993
1994 #define netdev_lockdep_set_classes(dev) \
1995 { \
1996 static struct lock_class_key qdisc_tx_busylock_key; \
1997 static struct lock_class_key qdisc_running_key; \
1998 static struct lock_class_key qdisc_xmit_lock_key; \
1999 static struct lock_class_key dev_addr_list_lock_key; \
2000 unsigned int i; \
2001 \
2002 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2003 (dev)->qdisc_running_key = &qdisc_running_key; \
2004 lockdep_set_class(&(dev)->addr_list_lock, \
2005 &dev_addr_list_lock_key); \
2006 for (i = 0; i < (dev)->num_tx_queues; i++) \
2007 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2008 &qdisc_xmit_lock_key); \
2009 }
2010
2011 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2012 struct sk_buff *skb,
2013 void *accel_priv);
2014
2015 /* returns the headroom that the master device needs to take in account
2016 * when forwarding to this dev
2017 */
2018 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2019 {
2020 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2021 }
2022
2023 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2024 {
2025 if (dev->netdev_ops->ndo_set_rx_headroom)
2026 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2027 }
2028
2029 /* set the device rx headroom to the dev's default */
2030 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2031 {
2032 netdev_set_rx_headroom(dev, -1);
2033 }
2034
2035 /*
2036 * Net namespace inlines
2037 */
2038 static inline
2039 struct net *dev_net(const struct net_device *dev)
2040 {
2041 return read_pnet(&dev->nd_net);
2042 }
2043
2044 static inline
2045 void dev_net_set(struct net_device *dev, struct net *net)
2046 {
2047 write_pnet(&dev->nd_net, net);
2048 }
2049
2050 /**
2051 * netdev_priv - access network device private data
2052 * @dev: network device
2053 *
2054 * Get network device private data
2055 */
2056 static inline void *netdev_priv(const struct net_device *dev)
2057 {
2058 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2059 }
2060
2061 /* Set the sysfs physical device reference for the network logical device
2062 * if set prior to registration will cause a symlink during initialization.
2063 */
2064 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2065
2066 /* Set the sysfs device type for the network logical device to allow
2067 * fine-grained identification of different network device types. For
2068 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2069 */
2070 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2071
2072 /* Default NAPI poll() weight
2073 * Device drivers are strongly advised to not use bigger value
2074 */
2075 #define NAPI_POLL_WEIGHT 64
2076
2077 /**
2078 * netif_napi_add - initialize a NAPI context
2079 * @dev: network device
2080 * @napi: NAPI context
2081 * @poll: polling function
2082 * @weight: default weight
2083 *
2084 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2085 * *any* of the other NAPI-related functions.
2086 */
2087 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2088 int (*poll)(struct napi_struct *, int), int weight);
2089
2090 /**
2091 * netif_tx_napi_add - initialize a NAPI context
2092 * @dev: network device
2093 * @napi: NAPI context
2094 * @poll: polling function
2095 * @weight: default weight
2096 *
2097 * This variant of netif_napi_add() should be used from drivers using NAPI
2098 * to exclusively poll a TX queue.
2099 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2100 */
2101 static inline void netif_tx_napi_add(struct net_device *dev,
2102 struct napi_struct *napi,
2103 int (*poll)(struct napi_struct *, int),
2104 int weight)
2105 {
2106 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2107 netif_napi_add(dev, napi, poll, weight);
2108 }
2109
2110 /**
2111 * netif_napi_del - remove a NAPI context
2112 * @napi: NAPI context
2113 *
2114 * netif_napi_del() removes a NAPI context from the network device NAPI list
2115 */
2116 void netif_napi_del(struct napi_struct *napi);
2117
2118 struct napi_gro_cb {
2119 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2120 void *frag0;
2121
2122 /* Length of frag0. */
2123 unsigned int frag0_len;
2124
2125 /* This indicates where we are processing relative to skb->data. */
2126 int data_offset;
2127
2128 /* This is non-zero if the packet cannot be merged with the new skb. */
2129 u16 flush;
2130
2131 /* Save the IP ID here and check when we get to the transport layer */
2132 u16 flush_id;
2133
2134 /* Number of segments aggregated. */
2135 u16 count;
2136
2137 /* Start offset for remote checksum offload */
2138 u16 gro_remcsum_start;
2139
2140 /* jiffies when first packet was created/queued */
2141 unsigned long age;
2142
2143 /* Used in ipv6_gro_receive() and foo-over-udp */
2144 u16 proto;
2145
2146 /* This is non-zero if the packet may be of the same flow. */
2147 u8 same_flow:1;
2148
2149 /* Used in tunnel GRO receive */
2150 u8 encap_mark:1;
2151
2152 /* GRO checksum is valid */
2153 u8 csum_valid:1;
2154
2155 /* Number of checksums via CHECKSUM_UNNECESSARY */
2156 u8 csum_cnt:3;
2157
2158 /* Free the skb? */
2159 u8 free:2;
2160 #define NAPI_GRO_FREE 1
2161 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2162
2163 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2164 u8 is_ipv6:1;
2165
2166 /* Used in GRE, set in fou/gue_gro_receive */
2167 u8 is_fou:1;
2168
2169 /* Used to determine if flush_id can be ignored */
2170 u8 is_atomic:1;
2171
2172 /* Number of gro_receive callbacks this packet already went through */
2173 u8 recursion_counter:4;
2174
2175 /* 1 bit hole */
2176
2177 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2178 __wsum csum;
2179
2180 /* used in skb_gro_receive() slow path */
2181 struct sk_buff *last;
2182 };
2183
2184 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2185
2186 #define GRO_RECURSION_LIMIT 15
2187 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2188 {
2189 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2190 }
2191
2192 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2193 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2194 struct sk_buff **head,
2195 struct sk_buff *skb)
2196 {
2197 if (unlikely(gro_recursion_inc_test(skb))) {
2198 NAPI_GRO_CB(skb)->flush |= 1;
2199 return NULL;
2200 }
2201
2202 return cb(head, skb);
2203 }
2204
2205 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2206 struct sk_buff *);
2207 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2208 struct sock *sk,
2209 struct sk_buff **head,
2210 struct sk_buff *skb)
2211 {
2212 if (unlikely(gro_recursion_inc_test(skb))) {
2213 NAPI_GRO_CB(skb)->flush |= 1;
2214 return NULL;
2215 }
2216
2217 return cb(sk, head, skb);
2218 }
2219
2220 struct packet_type {
2221 __be16 type; /* This is really htons(ether_type). */
2222 struct net_device *dev; /* NULL is wildcarded here */
2223 int (*func) (struct sk_buff *,
2224 struct net_device *,
2225 struct packet_type *,
2226 struct net_device *);
2227 bool (*id_match)(struct packet_type *ptype,
2228 struct sock *sk);
2229 void *af_packet_priv;
2230 struct list_head list;
2231 };
2232
2233 struct offload_callbacks {
2234 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2235 netdev_features_t features);
2236 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2237 struct sk_buff *skb);
2238 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2239 };
2240
2241 struct packet_offload {
2242 __be16 type; /* This is really htons(ether_type). */
2243 u16 priority;
2244 struct offload_callbacks callbacks;
2245 struct list_head list;
2246 };
2247
2248 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2249 struct pcpu_sw_netstats {
2250 u64 rx_packets;
2251 u64 rx_bytes;
2252 u64 tx_packets;
2253 u64 tx_bytes;
2254 struct u64_stats_sync syncp;
2255 };
2256
2257 #define __netdev_alloc_pcpu_stats(type, gfp) \
2258 ({ \
2259 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2260 if (pcpu_stats) { \
2261 int __cpu; \
2262 for_each_possible_cpu(__cpu) { \
2263 typeof(type) *stat; \
2264 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2265 u64_stats_init(&stat->syncp); \
2266 } \
2267 } \
2268 pcpu_stats; \
2269 })
2270
2271 #define netdev_alloc_pcpu_stats(type) \
2272 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2273
2274 enum netdev_lag_tx_type {
2275 NETDEV_LAG_TX_TYPE_UNKNOWN,
2276 NETDEV_LAG_TX_TYPE_RANDOM,
2277 NETDEV_LAG_TX_TYPE_BROADCAST,
2278 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2279 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2280 NETDEV_LAG_TX_TYPE_HASH,
2281 };
2282
2283 struct netdev_lag_upper_info {
2284 enum netdev_lag_tx_type tx_type;
2285 };
2286
2287 struct netdev_lag_lower_state_info {
2288 u8 link_up : 1,
2289 tx_enabled : 1;
2290 };
2291
2292 #include <linux/notifier.h>
2293
2294 /* netdevice notifier chain. Please remember to update the rtnetlink
2295 * notification exclusion list in rtnetlink_event() when adding new
2296 * types.
2297 */
2298 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2299 #define NETDEV_DOWN 0x0002
2300 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2301 detected a hardware crash and restarted
2302 - we can use this eg to kick tcp sessions
2303 once done */
2304 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2305 #define NETDEV_REGISTER 0x0005
2306 #define NETDEV_UNREGISTER 0x0006
2307 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2308 #define NETDEV_CHANGEADDR 0x0008
2309 #define NETDEV_GOING_DOWN 0x0009
2310 #define NETDEV_CHANGENAME 0x000A
2311 #define NETDEV_FEAT_CHANGE 0x000B
2312 #define NETDEV_BONDING_FAILOVER 0x000C
2313 #define NETDEV_PRE_UP 0x000D
2314 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2315 #define NETDEV_POST_TYPE_CHANGE 0x000F
2316 #define NETDEV_POST_INIT 0x0010
2317 #define NETDEV_UNREGISTER_FINAL 0x0011
2318 #define NETDEV_RELEASE 0x0012
2319 #define NETDEV_NOTIFY_PEERS 0x0013
2320 #define NETDEV_JOIN 0x0014
2321 #define NETDEV_CHANGEUPPER 0x0015
2322 #define NETDEV_RESEND_IGMP 0x0016
2323 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2324 #define NETDEV_CHANGEINFODATA 0x0018
2325 #define NETDEV_BONDING_INFO 0x0019
2326 #define NETDEV_PRECHANGEUPPER 0x001A
2327 #define NETDEV_CHANGELOWERSTATE 0x001B
2328 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2329 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2330 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2331
2332 int register_netdevice_notifier(struct notifier_block *nb);
2333 int unregister_netdevice_notifier(struct notifier_block *nb);
2334
2335 struct netdev_notifier_info {
2336 struct net_device *dev;
2337 struct netlink_ext_ack *extack;
2338 };
2339
2340 struct netdev_notifier_change_info {
2341 struct netdev_notifier_info info; /* must be first */
2342 unsigned int flags_changed;
2343 };
2344
2345 struct netdev_notifier_changeupper_info {
2346 struct netdev_notifier_info info; /* must be first */
2347 struct net_device *upper_dev; /* new upper dev */
2348 bool master; /* is upper dev master */
2349 bool linking; /* is the notification for link or unlink */
2350 void *upper_info; /* upper dev info */
2351 };
2352
2353 struct netdev_notifier_changelowerstate_info {
2354 struct netdev_notifier_info info; /* must be first */
2355 void *lower_state_info; /* is lower dev state */
2356 };
2357
2358 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2359 struct net_device *dev)
2360 {
2361 info->dev = dev;
2362 info->extack = NULL;
2363 }
2364
2365 static inline struct net_device *
2366 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2367 {
2368 return info->dev;
2369 }
2370
2371 static inline struct netlink_ext_ack *
2372 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2373 {
2374 return info->extack;
2375 }
2376
2377 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2378
2379
2380 extern rwlock_t dev_base_lock; /* Device list lock */
2381
2382 #define for_each_netdev(net, d) \
2383 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2384 #define for_each_netdev_reverse(net, d) \
2385 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2386 #define for_each_netdev_rcu(net, d) \
2387 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2388 #define for_each_netdev_safe(net, d, n) \
2389 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2390 #define for_each_netdev_continue(net, d) \
2391 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2392 #define for_each_netdev_continue_rcu(net, d) \
2393 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2394 #define for_each_netdev_in_bond_rcu(bond, slave) \
2395 for_each_netdev_rcu(&init_net, slave) \
2396 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2397 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2398
2399 static inline struct net_device *next_net_device(struct net_device *dev)
2400 {
2401 struct list_head *lh;
2402 struct net *net;
2403
2404 net = dev_net(dev);
2405 lh = dev->dev_list.next;
2406 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2407 }
2408
2409 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2410 {
2411 struct list_head *lh;
2412 struct net *net;
2413
2414 net = dev_net(dev);
2415 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2416 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2417 }
2418
2419 static inline struct net_device *first_net_device(struct net *net)
2420 {
2421 return list_empty(&net->dev_base_head) ? NULL :
2422 net_device_entry(net->dev_base_head.next);
2423 }
2424
2425 static inline struct net_device *first_net_device_rcu(struct net *net)
2426 {
2427 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2428
2429 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2430 }
2431
2432 int netdev_boot_setup_check(struct net_device *dev);
2433 unsigned long netdev_boot_base(const char *prefix, int unit);
2434 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2435 const char *hwaddr);
2436 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2437 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2438 void dev_add_pack(struct packet_type *pt);
2439 void dev_remove_pack(struct packet_type *pt);
2440 void __dev_remove_pack(struct packet_type *pt);
2441 void dev_add_offload(struct packet_offload *po);
2442 void dev_remove_offload(struct packet_offload *po);
2443
2444 int dev_get_iflink(const struct net_device *dev);
2445 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2446 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2447 unsigned short mask);
2448 struct net_device *dev_get_by_name(struct net *net, const char *name);
2449 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2450 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2451 int dev_alloc_name(struct net_device *dev, const char *name);
2452 int dev_open(struct net_device *dev);
2453 void dev_close(struct net_device *dev);
2454 void dev_close_many(struct list_head *head, bool unlink);
2455 void dev_disable_lro(struct net_device *dev);
2456 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2457 int dev_queue_xmit(struct sk_buff *skb);
2458 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2459 int register_netdevice(struct net_device *dev);
2460 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2461 void unregister_netdevice_many(struct list_head *head);
2462 static inline void unregister_netdevice(struct net_device *dev)
2463 {
2464 unregister_netdevice_queue(dev, NULL);
2465 }
2466
2467 int netdev_refcnt_read(const struct net_device *dev);
2468 void free_netdev(struct net_device *dev);
2469 void netdev_freemem(struct net_device *dev);
2470 void synchronize_net(void);
2471 int init_dummy_netdev(struct net_device *dev);
2472
2473 DECLARE_PER_CPU(int, xmit_recursion);
2474 #define XMIT_RECURSION_LIMIT 10
2475
2476 static inline int dev_recursion_level(void)
2477 {
2478 return this_cpu_read(xmit_recursion);
2479 }
2480
2481 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2482 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2483 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2484 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2485 int netdev_get_name(struct net *net, char *name, int ifindex);
2486 int dev_restart(struct net_device *dev);
2487 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2488
2489 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2490 {
2491 return NAPI_GRO_CB(skb)->data_offset;
2492 }
2493
2494 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2495 {
2496 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2497 }
2498
2499 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2500 {
2501 NAPI_GRO_CB(skb)->data_offset += len;
2502 }
2503
2504 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2505 unsigned int offset)
2506 {
2507 return NAPI_GRO_CB(skb)->frag0 + offset;
2508 }
2509
2510 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2511 {
2512 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2513 }
2514
2515 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2516 {
2517 NAPI_GRO_CB(skb)->frag0 = NULL;
2518 NAPI_GRO_CB(skb)->frag0_len = 0;
2519 }
2520
2521 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2522 unsigned int offset)
2523 {
2524 if (!pskb_may_pull(skb, hlen))
2525 return NULL;
2526
2527 skb_gro_frag0_invalidate(skb);
2528 return skb->data + offset;
2529 }
2530
2531 static inline void *skb_gro_network_header(struct sk_buff *skb)
2532 {
2533 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2534 skb_network_offset(skb);
2535 }
2536
2537 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2538 const void *start, unsigned int len)
2539 {
2540 if (NAPI_GRO_CB(skb)->csum_valid)
2541 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2542 csum_partial(start, len, 0));
2543 }
2544
2545 /* GRO checksum functions. These are logical equivalents of the normal
2546 * checksum functions (in skbuff.h) except that they operate on the GRO
2547 * offsets and fields in sk_buff.
2548 */
2549
2550 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2551
2552 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2553 {
2554 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2555 }
2556
2557 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2558 bool zero_okay,
2559 __sum16 check)
2560 {
2561 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2562 skb_checksum_start_offset(skb) <
2563 skb_gro_offset(skb)) &&
2564 !skb_at_gro_remcsum_start(skb) &&
2565 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2566 (!zero_okay || check));
2567 }
2568
2569 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2570 __wsum psum)
2571 {
2572 if (NAPI_GRO_CB(skb)->csum_valid &&
2573 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2574 return 0;
2575
2576 NAPI_GRO_CB(skb)->csum = psum;
2577
2578 return __skb_gro_checksum_complete(skb);
2579 }
2580
2581 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2582 {
2583 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2584 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2585 NAPI_GRO_CB(skb)->csum_cnt--;
2586 } else {
2587 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2588 * verified a new top level checksum or an encapsulated one
2589 * during GRO. This saves work if we fallback to normal path.
2590 */
2591 __skb_incr_checksum_unnecessary(skb);
2592 }
2593 }
2594
2595 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2596 compute_pseudo) \
2597 ({ \
2598 __sum16 __ret = 0; \
2599 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2600 __ret = __skb_gro_checksum_validate_complete(skb, \
2601 compute_pseudo(skb, proto)); \
2602 if (!__ret) \
2603 skb_gro_incr_csum_unnecessary(skb); \
2604 __ret; \
2605 })
2606
2607 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2608 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2609
2610 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2611 compute_pseudo) \
2612 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2613
2614 #define skb_gro_checksum_simple_validate(skb) \
2615 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2616
2617 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2618 {
2619 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2620 !NAPI_GRO_CB(skb)->csum_valid);
2621 }
2622
2623 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2624 __sum16 check, __wsum pseudo)
2625 {
2626 NAPI_GRO_CB(skb)->csum = ~pseudo;
2627 NAPI_GRO_CB(skb)->csum_valid = 1;
2628 }
2629
2630 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2631 do { \
2632 if (__skb_gro_checksum_convert_check(skb)) \
2633 __skb_gro_checksum_convert(skb, check, \
2634 compute_pseudo(skb, proto)); \
2635 } while (0)
2636
2637 struct gro_remcsum {
2638 int offset;
2639 __wsum delta;
2640 };
2641
2642 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2643 {
2644 grc->offset = 0;
2645 grc->delta = 0;
2646 }
2647
2648 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2649 unsigned int off, size_t hdrlen,
2650 int start, int offset,
2651 struct gro_remcsum *grc,
2652 bool nopartial)
2653 {
2654 __wsum delta;
2655 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2656
2657 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2658
2659 if (!nopartial) {
2660 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2661 return ptr;
2662 }
2663
2664 ptr = skb_gro_header_fast(skb, off);
2665 if (skb_gro_header_hard(skb, off + plen)) {
2666 ptr = skb_gro_header_slow(skb, off + plen, off);
2667 if (!ptr)
2668 return NULL;
2669 }
2670
2671 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2672 start, offset);
2673
2674 /* Adjust skb->csum since we changed the packet */
2675 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2676
2677 grc->offset = off + hdrlen + offset;
2678 grc->delta = delta;
2679
2680 return ptr;
2681 }
2682
2683 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2684 struct gro_remcsum *grc)
2685 {
2686 void *ptr;
2687 size_t plen = grc->offset + sizeof(u16);
2688
2689 if (!grc->delta)
2690 return;
2691
2692 ptr = skb_gro_header_fast(skb, grc->offset);
2693 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2694 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2695 if (!ptr)
2696 return;
2697 }
2698
2699 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2700 }
2701
2702 #ifdef CONFIG_XFRM_OFFLOAD
2703 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2704 {
2705 if (PTR_ERR(pp) != -EINPROGRESS)
2706 NAPI_GRO_CB(skb)->flush |= flush;
2707 }
2708 #else
2709 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2710 {
2711 NAPI_GRO_CB(skb)->flush |= flush;
2712 }
2713 #endif
2714
2715 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2716 unsigned short type,
2717 const void *daddr, const void *saddr,
2718 unsigned int len)
2719 {
2720 if (!dev->header_ops || !dev->header_ops->create)
2721 return 0;
2722
2723 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2724 }
2725
2726 static inline int dev_parse_header(const struct sk_buff *skb,
2727 unsigned char *haddr)
2728 {
2729 const struct net_device *dev = skb->dev;
2730
2731 if (!dev->header_ops || !dev->header_ops->parse)
2732 return 0;
2733 return dev->header_ops->parse(skb, haddr);
2734 }
2735
2736 /* ll_header must have at least hard_header_len allocated */
2737 static inline bool dev_validate_header(const struct net_device *dev,
2738 char *ll_header, int len)
2739 {
2740 if (likely(len >= dev->hard_header_len))
2741 return true;
2742 if (len < dev->min_header_len)
2743 return false;
2744
2745 if (capable(CAP_SYS_RAWIO)) {
2746 memset(ll_header + len, 0, dev->hard_header_len - len);
2747 return true;
2748 }
2749
2750 if (dev->header_ops && dev->header_ops->validate)
2751 return dev->header_ops->validate(ll_header, len);
2752
2753 return false;
2754 }
2755
2756 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2757 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2758 static inline int unregister_gifconf(unsigned int family)
2759 {
2760 return register_gifconf(family, NULL);
2761 }
2762
2763 #ifdef CONFIG_NET_FLOW_LIMIT
2764 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2765 struct sd_flow_limit {
2766 u64 count;
2767 unsigned int num_buckets;
2768 unsigned int history_head;
2769 u16 history[FLOW_LIMIT_HISTORY];
2770 u8 buckets[];
2771 };
2772
2773 extern int netdev_flow_limit_table_len;
2774 #endif /* CONFIG_NET_FLOW_LIMIT */
2775
2776 /*
2777 * Incoming packets are placed on per-CPU queues
2778 */
2779 struct softnet_data {
2780 struct list_head poll_list;
2781 struct sk_buff_head process_queue;
2782
2783 /* stats */
2784 unsigned int processed;
2785 unsigned int time_squeeze;
2786 unsigned int received_rps;
2787 #ifdef CONFIG_RPS
2788 struct softnet_data *rps_ipi_list;
2789 #endif
2790 #ifdef CONFIG_NET_FLOW_LIMIT
2791 struct sd_flow_limit __rcu *flow_limit;
2792 #endif
2793 struct Qdisc *output_queue;
2794 struct Qdisc **output_queue_tailp;
2795 struct sk_buff *completion_queue;
2796 #ifdef CONFIG_XFRM_OFFLOAD
2797 struct sk_buff_head xfrm_backlog;
2798 #endif
2799 #ifdef CONFIG_RPS
2800 /* input_queue_head should be written by cpu owning this struct,
2801 * and only read by other cpus. Worth using a cache line.
2802 */
2803 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2804
2805 /* Elements below can be accessed between CPUs for RPS/RFS */
2806 call_single_data_t csd ____cacheline_aligned_in_smp;
2807 struct softnet_data *rps_ipi_next;
2808 unsigned int cpu;
2809 unsigned int input_queue_tail;
2810 #endif
2811 unsigned int dropped;
2812 struct sk_buff_head input_pkt_queue;
2813 struct napi_struct backlog;
2814
2815 };
2816
2817 static inline void input_queue_head_incr(struct softnet_data *sd)
2818 {
2819 #ifdef CONFIG_RPS
2820 sd->input_queue_head++;
2821 #endif
2822 }
2823
2824 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2825 unsigned int *qtail)
2826 {
2827 #ifdef CONFIG_RPS
2828 *qtail = ++sd->input_queue_tail;
2829 #endif
2830 }
2831
2832 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2833
2834 void __netif_schedule(struct Qdisc *q);
2835 void netif_schedule_queue(struct netdev_queue *txq);
2836
2837 static inline void netif_tx_schedule_all(struct net_device *dev)
2838 {
2839 unsigned int i;
2840
2841 for (i = 0; i < dev->num_tx_queues; i++)
2842 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2843 }
2844
2845 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2846 {
2847 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2848 }
2849
2850 /**
2851 * netif_start_queue - allow transmit
2852 * @dev: network device
2853 *
2854 * Allow upper layers to call the device hard_start_xmit routine.
2855 */
2856 static inline void netif_start_queue(struct net_device *dev)
2857 {
2858 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2859 }
2860
2861 static inline void netif_tx_start_all_queues(struct net_device *dev)
2862 {
2863 unsigned int i;
2864
2865 for (i = 0; i < dev->num_tx_queues; i++) {
2866 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2867 netif_tx_start_queue(txq);
2868 }
2869 }
2870
2871 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2872
2873 /**
2874 * netif_wake_queue - restart transmit
2875 * @dev: network device
2876 *
2877 * Allow upper layers to call the device hard_start_xmit routine.
2878 * Used for flow control when transmit resources are available.
2879 */
2880 static inline void netif_wake_queue(struct net_device *dev)
2881 {
2882 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2883 }
2884
2885 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2886 {
2887 unsigned int i;
2888
2889 for (i = 0; i < dev->num_tx_queues; i++) {
2890 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2891 netif_tx_wake_queue(txq);
2892 }
2893 }
2894
2895 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2896 {
2897 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2898 }
2899
2900 /**
2901 * netif_stop_queue - stop transmitted packets
2902 * @dev: network device
2903 *
2904 * Stop upper layers calling the device hard_start_xmit routine.
2905 * Used for flow control when transmit resources are unavailable.
2906 */
2907 static inline void netif_stop_queue(struct net_device *dev)
2908 {
2909 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2910 }
2911
2912 void netif_tx_stop_all_queues(struct net_device *dev);
2913
2914 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2915 {
2916 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2917 }
2918
2919 /**
2920 * netif_queue_stopped - test if transmit queue is flowblocked
2921 * @dev: network device
2922 *
2923 * Test if transmit queue on device is currently unable to send.
2924 */
2925 static inline bool netif_queue_stopped(const struct net_device *dev)
2926 {
2927 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2928 }
2929
2930 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2931 {
2932 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2933 }
2934
2935 static inline bool
2936 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2937 {
2938 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2939 }
2940
2941 static inline bool
2942 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2943 {
2944 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2945 }
2946
2947 /**
2948 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2949 * @dev_queue: pointer to transmit queue
2950 *
2951 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2952 * to give appropriate hint to the CPU.
2953 */
2954 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2955 {
2956 #ifdef CONFIG_BQL
2957 prefetchw(&dev_queue->dql.num_queued);
2958 #endif
2959 }
2960
2961 /**
2962 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2963 * @dev_queue: pointer to transmit queue
2964 *
2965 * BQL enabled drivers might use this helper in their TX completion path,
2966 * to give appropriate hint to the CPU.
2967 */
2968 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2969 {
2970 #ifdef CONFIG_BQL
2971 prefetchw(&dev_queue->dql.limit);
2972 #endif
2973 }
2974
2975 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2976 unsigned int bytes)
2977 {
2978 #ifdef CONFIG_BQL
2979 dql_queued(&dev_queue->dql, bytes);
2980
2981 if (likely(dql_avail(&dev_queue->dql) >= 0))
2982 return;
2983
2984 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2985
2986 /*
2987 * The XOFF flag must be set before checking the dql_avail below,
2988 * because in netdev_tx_completed_queue we update the dql_completed
2989 * before checking the XOFF flag.
2990 */
2991 smp_mb();
2992
2993 /* check again in case another CPU has just made room avail */
2994 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2995 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2996 #endif
2997 }
2998
2999 /**
3000 * netdev_sent_queue - report the number of bytes queued to hardware
3001 * @dev: network device
3002 * @bytes: number of bytes queued to the hardware device queue
3003 *
3004 * Report the number of bytes queued for sending/completion to the network
3005 * device hardware queue. @bytes should be a good approximation and should
3006 * exactly match netdev_completed_queue() @bytes
3007 */
3008 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3009 {
3010 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3011 }
3012
3013 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3014 unsigned int pkts, unsigned int bytes)
3015 {
3016 #ifdef CONFIG_BQL
3017 if (unlikely(!bytes))
3018 return;
3019
3020 dql_completed(&dev_queue->dql, bytes);
3021
3022 /*
3023 * Without the memory barrier there is a small possiblity that
3024 * netdev_tx_sent_queue will miss the update and cause the queue to
3025 * be stopped forever
3026 */
3027 smp_mb();
3028
3029 if (dql_avail(&dev_queue->dql) < 0)
3030 return;
3031
3032 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3033 netif_schedule_queue(dev_queue);
3034 #endif
3035 }
3036
3037 /**
3038 * netdev_completed_queue - report bytes and packets completed by device
3039 * @dev: network device
3040 * @pkts: actual number of packets sent over the medium
3041 * @bytes: actual number of bytes sent over the medium
3042 *
3043 * Report the number of bytes and packets transmitted by the network device
3044 * hardware queue over the physical medium, @bytes must exactly match the
3045 * @bytes amount passed to netdev_sent_queue()
3046 */
3047 static inline void netdev_completed_queue(struct net_device *dev,
3048 unsigned int pkts, unsigned int bytes)
3049 {
3050 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3051 }
3052
3053 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3054 {
3055 #ifdef CONFIG_BQL
3056 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3057 dql_reset(&q->dql);
3058 #endif
3059 }
3060
3061 /**
3062 * netdev_reset_queue - reset the packets and bytes count of a network device
3063 * @dev_queue: network device
3064 *
3065 * Reset the bytes and packet count of a network device and clear the
3066 * software flow control OFF bit for this network device
3067 */
3068 static inline void netdev_reset_queue(struct net_device *dev_queue)
3069 {
3070 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3071 }
3072
3073 /**
3074 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3075 * @dev: network device
3076 * @queue_index: given tx queue index
3077 *
3078 * Returns 0 if given tx queue index >= number of device tx queues,
3079 * otherwise returns the originally passed tx queue index.
3080 */
3081 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3082 {
3083 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3084 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3085 dev->name, queue_index,
3086 dev->real_num_tx_queues);
3087 return 0;
3088 }
3089
3090 return queue_index;
3091 }
3092
3093 /**
3094 * netif_running - test if up
3095 * @dev: network device
3096 *
3097 * Test if the device has been brought up.
3098 */
3099 static inline bool netif_running(const struct net_device *dev)
3100 {
3101 return test_bit(__LINK_STATE_START, &dev->state);
3102 }
3103
3104 /*
3105 * Routines to manage the subqueues on a device. We only need start,
3106 * stop, and a check if it's stopped. All other device management is
3107 * done at the overall netdevice level.
3108 * Also test the device if we're multiqueue.
3109 */
3110
3111 /**
3112 * netif_start_subqueue - allow sending packets on subqueue
3113 * @dev: network device
3114 * @queue_index: sub queue index
3115 *
3116 * Start individual transmit queue of a device with multiple transmit queues.
3117 */
3118 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3119 {
3120 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3121
3122 netif_tx_start_queue(txq);
3123 }
3124
3125 /**
3126 * netif_stop_subqueue - stop sending packets on subqueue
3127 * @dev: network device
3128 * @queue_index: sub queue index
3129 *
3130 * Stop individual transmit queue of a device with multiple transmit queues.
3131 */
3132 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3133 {
3134 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3135 netif_tx_stop_queue(txq);
3136 }
3137
3138 /**
3139 * netif_subqueue_stopped - test status of subqueue
3140 * @dev: network device
3141 * @queue_index: sub queue index
3142 *
3143 * Check individual transmit queue of a device with multiple transmit queues.
3144 */
3145 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3146 u16 queue_index)
3147 {
3148 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3149
3150 return netif_tx_queue_stopped(txq);
3151 }
3152
3153 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3154 struct sk_buff *skb)
3155 {
3156 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3157 }
3158
3159 /**
3160 * netif_wake_subqueue - allow sending packets on subqueue
3161 * @dev: network device
3162 * @queue_index: sub queue index
3163 *
3164 * Resume individual transmit queue of a device with multiple transmit queues.
3165 */
3166 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3167 {
3168 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3169
3170 netif_tx_wake_queue(txq);
3171 }
3172
3173 #ifdef CONFIG_XPS
3174 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3175 u16 index);
3176 #else
3177 static inline int netif_set_xps_queue(struct net_device *dev,
3178 const struct cpumask *mask,
3179 u16 index)
3180 {
3181 return 0;
3182 }
3183 #endif
3184
3185 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3186 unsigned int num_tx_queues);
3187
3188 /*
3189 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3190 * as a distribution range limit for the returned value.
3191 */
3192 static inline u16 skb_tx_hash(const struct net_device *dev,
3193 struct sk_buff *skb)
3194 {
3195 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3196 }
3197
3198 /**
3199 * netif_is_multiqueue - test if device has multiple transmit queues
3200 * @dev: network device
3201 *
3202 * Check if device has multiple transmit queues
3203 */
3204 static inline bool netif_is_multiqueue(const struct net_device *dev)
3205 {
3206 return dev->num_tx_queues > 1;
3207 }
3208
3209 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3210
3211 #ifdef CONFIG_SYSFS
3212 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3213 #else
3214 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3215 unsigned int rxq)
3216 {
3217 return 0;
3218 }
3219 #endif
3220
3221 #ifdef CONFIG_SYSFS
3222 static inline unsigned int get_netdev_rx_queue_index(
3223 struct netdev_rx_queue *queue)
3224 {
3225 struct net_device *dev = queue->dev;
3226 int index = queue - dev->_rx;
3227
3228 BUG_ON(index >= dev->num_rx_queues);
3229 return index;
3230 }
3231 #endif
3232
3233 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3234 int netif_get_num_default_rss_queues(void);
3235
3236 enum skb_free_reason {
3237 SKB_REASON_CONSUMED,
3238 SKB_REASON_DROPPED,
3239 };
3240
3241 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3242 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3243
3244 /*
3245 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3246 * interrupt context or with hardware interrupts being disabled.
3247 * (in_irq() || irqs_disabled())
3248 *
3249 * We provide four helpers that can be used in following contexts :
3250 *
3251 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3252 * replacing kfree_skb(skb)
3253 *
3254 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3255 * Typically used in place of consume_skb(skb) in TX completion path
3256 *
3257 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3258 * replacing kfree_skb(skb)
3259 *
3260 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3261 * and consumed a packet. Used in place of consume_skb(skb)
3262 */
3263 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3264 {
3265 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3266 }
3267
3268 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3269 {
3270 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3271 }
3272
3273 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3274 {
3275 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3276 }
3277
3278 static inline void dev_consume_skb_any(struct sk_buff *skb)
3279 {
3280 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3281 }
3282
3283 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3284 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3285 int netif_rx(struct sk_buff *skb);
3286 int netif_rx_ni(struct sk_buff *skb);
3287 int netif_receive_skb(struct sk_buff *skb);
3288 int netif_receive_skb_core(struct sk_buff *skb);
3289 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3290 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3291 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3292 gro_result_t napi_gro_frags(struct napi_struct *napi);
3293 struct packet_offload *gro_find_receive_by_type(__be16 type);
3294 struct packet_offload *gro_find_complete_by_type(__be16 type);
3295
3296 static inline void napi_free_frags(struct napi_struct *napi)
3297 {
3298 kfree_skb(napi->skb);
3299 napi->skb = NULL;
3300 }
3301
3302 bool netdev_is_rx_handler_busy(struct net_device *dev);
3303 int netdev_rx_handler_register(struct net_device *dev,
3304 rx_handler_func_t *rx_handler,
3305 void *rx_handler_data);
3306 void netdev_rx_handler_unregister(struct net_device *dev);
3307
3308 bool dev_valid_name(const char *name);
3309 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3310 int dev_ethtool(struct net *net, struct ifreq *);
3311 unsigned int dev_get_flags(const struct net_device *);
3312 int __dev_change_flags(struct net_device *, unsigned int flags);
3313 int dev_change_flags(struct net_device *, unsigned int);
3314 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3315 unsigned int gchanges);
3316 int dev_change_name(struct net_device *, const char *);
3317 int dev_set_alias(struct net_device *, const char *, size_t);
3318 int dev_get_alias(const struct net_device *, char *, size_t);
3319 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3320 int __dev_set_mtu(struct net_device *, int);
3321 int dev_set_mtu(struct net_device *, int);
3322 void dev_set_group(struct net_device *, int);
3323 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3324 int dev_change_carrier(struct net_device *, bool new_carrier);
3325 int dev_get_phys_port_id(struct net_device *dev,
3326 struct netdev_phys_item_id *ppid);
3327 int dev_get_phys_port_name(struct net_device *dev,
3328 char *name, size_t len);
3329 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3330 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3331 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3332 struct netdev_queue *txq, int *ret);
3333
3334 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3335 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3336 int fd, u32 flags);
3337 void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3338 struct netdev_bpf *xdp);
3339
3340 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3341 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3342 bool is_skb_forwardable(const struct net_device *dev,
3343 const struct sk_buff *skb);
3344
3345 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3346 struct sk_buff *skb)
3347 {
3348 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3349 unlikely(!is_skb_forwardable(dev, skb))) {
3350 atomic_long_inc(&dev->rx_dropped);
3351 kfree_skb(skb);
3352 return NET_RX_DROP;
3353 }
3354
3355 skb_scrub_packet(skb, true);
3356 skb->priority = 0;
3357 return 0;
3358 }
3359
3360 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3361
3362 extern int netdev_budget;
3363 extern unsigned int netdev_budget_usecs;
3364
3365 /* Called by rtnetlink.c:rtnl_unlock() */
3366 void netdev_run_todo(void);
3367
3368 /**
3369 * dev_put - release reference to device
3370 * @dev: network device
3371 *
3372 * Release reference to device to allow it to be freed.
3373 */
3374 static inline void dev_put(struct net_device *dev)
3375 {
3376 this_cpu_dec(*dev->pcpu_refcnt);
3377 }
3378
3379 /**
3380 * dev_hold - get reference to device
3381 * @dev: network device
3382 *
3383 * Hold reference to device to keep it from being freed.
3384 */
3385 static inline void dev_hold(struct net_device *dev)
3386 {
3387 this_cpu_inc(*dev->pcpu_refcnt);
3388 }
3389
3390 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3391 * and _off may be called from IRQ context, but it is caller
3392 * who is responsible for serialization of these calls.
3393 *
3394 * The name carrier is inappropriate, these functions should really be
3395 * called netif_lowerlayer_*() because they represent the state of any
3396 * kind of lower layer not just hardware media.
3397 */
3398
3399 void linkwatch_init_dev(struct net_device *dev);
3400 void linkwatch_fire_event(struct net_device *dev);
3401 void linkwatch_forget_dev(struct net_device *dev);
3402
3403 /**
3404 * netif_carrier_ok - test if carrier present
3405 * @dev: network device
3406 *
3407 * Check if carrier is present on device
3408 */
3409 static inline bool netif_carrier_ok(const struct net_device *dev)
3410 {
3411 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3412 }
3413
3414 unsigned long dev_trans_start(struct net_device *dev);
3415
3416 void __netdev_watchdog_up(struct net_device *dev);
3417
3418 void netif_carrier_on(struct net_device *dev);
3419
3420 void netif_carrier_off(struct net_device *dev);
3421
3422 /**
3423 * netif_dormant_on - mark device as dormant.
3424 * @dev: network device
3425 *
3426 * Mark device as dormant (as per RFC2863).
3427 *
3428 * The dormant state indicates that the relevant interface is not
3429 * actually in a condition to pass packets (i.e., it is not 'up') but is
3430 * in a "pending" state, waiting for some external event. For "on-
3431 * demand" interfaces, this new state identifies the situation where the
3432 * interface is waiting for events to place it in the up state.
3433 */
3434 static inline void netif_dormant_on(struct net_device *dev)
3435 {
3436 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3437 linkwatch_fire_event(dev);
3438 }
3439
3440 /**
3441 * netif_dormant_off - set device as not dormant.
3442 * @dev: network device
3443 *
3444 * Device is not in dormant state.
3445 */
3446 static inline void netif_dormant_off(struct net_device *dev)
3447 {
3448 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3449 linkwatch_fire_event(dev);
3450 }
3451
3452 /**
3453 * netif_dormant - test if device is dormant
3454 * @dev: network device
3455 *
3456 * Check if device is dormant.
3457 */
3458 static inline bool netif_dormant(const struct net_device *dev)
3459 {
3460 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3461 }
3462
3463
3464 /**
3465 * netif_oper_up - test if device is operational
3466 * @dev: network device
3467 *
3468 * Check if carrier is operational
3469 */
3470 static inline bool netif_oper_up(const struct net_device *dev)
3471 {
3472 return (dev->operstate == IF_OPER_UP ||
3473 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3474 }
3475
3476 /**
3477 * netif_device_present - is device available or removed
3478 * @dev: network device
3479 *
3480 * Check if device has not been removed from system.
3481 */
3482 static inline bool netif_device_present(struct net_device *dev)
3483 {
3484 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3485 }
3486
3487 void netif_device_detach(struct net_device *dev);
3488
3489 void netif_device_attach(struct net_device *dev);
3490
3491 /*
3492 * Network interface message level settings
3493 */
3494
3495 enum {
3496 NETIF_MSG_DRV = 0x0001,
3497 NETIF_MSG_PROBE = 0x0002,
3498 NETIF_MSG_LINK = 0x0004,
3499 NETIF_MSG_TIMER = 0x0008,
3500 NETIF_MSG_IFDOWN = 0x0010,
3501 NETIF_MSG_IFUP = 0x0020,
3502 NETIF_MSG_RX_ERR = 0x0040,
3503 NETIF_MSG_TX_ERR = 0x0080,
3504 NETIF_MSG_TX_QUEUED = 0x0100,
3505 NETIF_MSG_INTR = 0x0200,
3506 NETIF_MSG_TX_DONE = 0x0400,
3507 NETIF_MSG_RX_STATUS = 0x0800,
3508 NETIF_MSG_PKTDATA = 0x1000,
3509 NETIF_MSG_HW = 0x2000,
3510 NETIF_MSG_WOL = 0x4000,
3511 };
3512
3513 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3514 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3515 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3516 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3517 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3518 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3519 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3520 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3521 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3522 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3523 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3524 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3525 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3526 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3527 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3528
3529 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3530 {
3531 /* use default */
3532 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3533 return default_msg_enable_bits;
3534 if (debug_value == 0) /* no output */
3535 return 0;
3536 /* set low N bits */
3537 return (1 << debug_value) - 1;
3538 }
3539
3540 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3541 {
3542 spin_lock(&txq->_xmit_lock);
3543 txq->xmit_lock_owner = cpu;
3544 }
3545
3546 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3547 {
3548 __acquire(&txq->_xmit_lock);
3549 return true;
3550 }
3551
3552 static inline void __netif_tx_release(struct netdev_queue *txq)
3553 {
3554 __release(&txq->_xmit_lock);
3555 }
3556
3557 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3558 {
3559 spin_lock_bh(&txq->_xmit_lock);
3560 txq->xmit_lock_owner = smp_processor_id();
3561 }
3562
3563 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3564 {
3565 bool ok = spin_trylock(&txq->_xmit_lock);
3566 if (likely(ok))
3567 txq->xmit_lock_owner = smp_processor_id();
3568 return ok;
3569 }
3570
3571 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3572 {
3573 txq->xmit_lock_owner = -1;
3574 spin_unlock(&txq->_xmit_lock);
3575 }
3576
3577 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3578 {
3579 txq->xmit_lock_owner = -1;
3580 spin_unlock_bh(&txq->_xmit_lock);
3581 }
3582
3583 static inline void txq_trans_update(struct netdev_queue *txq)
3584 {
3585 if (txq->xmit_lock_owner != -1)
3586 txq->trans_start = jiffies;
3587 }
3588
3589 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3590 static inline void netif_trans_update(struct net_device *dev)
3591 {
3592 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3593
3594 if (txq->trans_start != jiffies)
3595 txq->trans_start = jiffies;
3596 }
3597
3598 /**
3599 * netif_tx_lock - grab network device transmit lock
3600 * @dev: network device
3601 *
3602 * Get network device transmit lock
3603 */
3604 static inline void netif_tx_lock(struct net_device *dev)
3605 {
3606 unsigned int i;
3607 int cpu;
3608
3609 spin_lock(&dev->tx_global_lock);
3610 cpu = smp_processor_id();
3611 for (i = 0; i < dev->num_tx_queues; i++) {
3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3613
3614 /* We are the only thread of execution doing a
3615 * freeze, but we have to grab the _xmit_lock in
3616 * order to synchronize with threads which are in
3617 * the ->hard_start_xmit() handler and already
3618 * checked the frozen bit.
3619 */
3620 __netif_tx_lock(txq, cpu);
3621 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3622 __netif_tx_unlock(txq);
3623 }
3624 }
3625
3626 static inline void netif_tx_lock_bh(struct net_device *dev)
3627 {
3628 local_bh_disable();
3629 netif_tx_lock(dev);
3630 }
3631
3632 static inline void netif_tx_unlock(struct net_device *dev)
3633 {
3634 unsigned int i;
3635
3636 for (i = 0; i < dev->num_tx_queues; i++) {
3637 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3638
3639 /* No need to grab the _xmit_lock here. If the
3640 * queue is not stopped for another reason, we
3641 * force a schedule.
3642 */
3643 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3644 netif_schedule_queue(txq);
3645 }
3646 spin_unlock(&dev->tx_global_lock);
3647 }
3648
3649 static inline void netif_tx_unlock_bh(struct net_device *dev)
3650 {
3651 netif_tx_unlock(dev);
3652 local_bh_enable();
3653 }
3654
3655 #define HARD_TX_LOCK(dev, txq, cpu) { \
3656 if ((dev->features & NETIF_F_LLTX) == 0) { \
3657 __netif_tx_lock(txq, cpu); \
3658 } else { \
3659 __netif_tx_acquire(txq); \
3660 } \
3661 }
3662
3663 #define HARD_TX_TRYLOCK(dev, txq) \
3664 (((dev->features & NETIF_F_LLTX) == 0) ? \
3665 __netif_tx_trylock(txq) : \
3666 __netif_tx_acquire(txq))
3667
3668 #define HARD_TX_UNLOCK(dev, txq) { \
3669 if ((dev->features & NETIF_F_LLTX) == 0) { \
3670 __netif_tx_unlock(txq); \
3671 } else { \
3672 __netif_tx_release(txq); \
3673 } \
3674 }
3675
3676 static inline void netif_tx_disable(struct net_device *dev)
3677 {
3678 unsigned int i;
3679 int cpu;
3680
3681 local_bh_disable();
3682 cpu = smp_processor_id();
3683 for (i = 0; i < dev->num_tx_queues; i++) {
3684 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3685
3686 __netif_tx_lock(txq, cpu);
3687 netif_tx_stop_queue(txq);
3688 __netif_tx_unlock(txq);
3689 }
3690 local_bh_enable();
3691 }
3692
3693 static inline void netif_addr_lock(struct net_device *dev)
3694 {
3695 spin_lock(&dev->addr_list_lock);
3696 }
3697
3698 static inline void netif_addr_lock_nested(struct net_device *dev)
3699 {
3700 int subclass = SINGLE_DEPTH_NESTING;
3701
3702 if (dev->netdev_ops->ndo_get_lock_subclass)
3703 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3704
3705 spin_lock_nested(&dev->addr_list_lock, subclass);
3706 }
3707
3708 static inline void netif_addr_lock_bh(struct net_device *dev)
3709 {
3710 spin_lock_bh(&dev->addr_list_lock);
3711 }
3712
3713 static inline void netif_addr_unlock(struct net_device *dev)
3714 {
3715 spin_unlock(&dev->addr_list_lock);
3716 }
3717
3718 static inline void netif_addr_unlock_bh(struct net_device *dev)
3719 {
3720 spin_unlock_bh(&dev->addr_list_lock);
3721 }
3722
3723 /*
3724 * dev_addrs walker. Should be used only for read access. Call with
3725 * rcu_read_lock held.
3726 */
3727 #define for_each_dev_addr(dev, ha) \
3728 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3729
3730 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3731
3732 void ether_setup(struct net_device *dev);
3733
3734 /* Support for loadable net-drivers */
3735 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3736 unsigned char name_assign_type,
3737 void (*setup)(struct net_device *),
3738 unsigned int txqs, unsigned int rxqs);
3739 int dev_get_valid_name(struct net *net, struct net_device *dev,
3740 const char *name);
3741
3742 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3743 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3744
3745 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3746 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3747 count)
3748
3749 int register_netdev(struct net_device *dev);
3750 void unregister_netdev(struct net_device *dev);
3751
3752 /* General hardware address lists handling functions */
3753 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3754 struct netdev_hw_addr_list *from_list, int addr_len);
3755 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3756 struct netdev_hw_addr_list *from_list, int addr_len);
3757 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3758 struct net_device *dev,
3759 int (*sync)(struct net_device *, const unsigned char *),
3760 int (*unsync)(struct net_device *,
3761 const unsigned char *));
3762 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3763 struct net_device *dev,
3764 int (*unsync)(struct net_device *,
3765 const unsigned char *));
3766 void __hw_addr_init(struct netdev_hw_addr_list *list);
3767
3768 /* Functions used for device addresses handling */
3769 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3770 unsigned char addr_type);
3771 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3772 unsigned char addr_type);
3773 void dev_addr_flush(struct net_device *dev);
3774 int dev_addr_init(struct net_device *dev);
3775
3776 /* Functions used for unicast addresses handling */
3777 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3778 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3779 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3780 int dev_uc_sync(struct net_device *to, struct net_device *from);
3781 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3782 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3783 void dev_uc_flush(struct net_device *dev);
3784 void dev_uc_init(struct net_device *dev);
3785
3786 /**
3787 * __dev_uc_sync - Synchonize device's unicast list
3788 * @dev: device to sync
3789 * @sync: function to call if address should be added
3790 * @unsync: function to call if address should be removed
3791 *
3792 * Add newly added addresses to the interface, and release
3793 * addresses that have been deleted.
3794 */
3795 static inline int __dev_uc_sync(struct net_device *dev,
3796 int (*sync)(struct net_device *,
3797 const unsigned char *),
3798 int (*unsync)(struct net_device *,
3799 const unsigned char *))
3800 {
3801 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3802 }
3803
3804 /**
3805 * __dev_uc_unsync - Remove synchronized addresses from device
3806 * @dev: device to sync
3807 * @unsync: function to call if address should be removed
3808 *
3809 * Remove all addresses that were added to the device by dev_uc_sync().
3810 */
3811 static inline void __dev_uc_unsync(struct net_device *dev,
3812 int (*unsync)(struct net_device *,
3813 const unsigned char *))
3814 {
3815 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3816 }
3817
3818 /* Functions used for multicast addresses handling */
3819 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3820 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3821 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3822 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3823 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3824 int dev_mc_sync(struct net_device *to, struct net_device *from);
3825 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3826 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3827 void dev_mc_flush(struct net_device *dev);
3828 void dev_mc_init(struct net_device *dev);
3829
3830 /**
3831 * __dev_mc_sync - Synchonize device's multicast list
3832 * @dev: device to sync
3833 * @sync: function to call if address should be added
3834 * @unsync: function to call if address should be removed
3835 *
3836 * Add newly added addresses to the interface, and release
3837 * addresses that have been deleted.
3838 */
3839 static inline int __dev_mc_sync(struct net_device *dev,
3840 int (*sync)(struct net_device *,
3841 const unsigned char *),
3842 int (*unsync)(struct net_device *,
3843 const unsigned char *))
3844 {
3845 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3846 }
3847
3848 /**
3849 * __dev_mc_unsync - Remove synchronized addresses from device
3850 * @dev: device to sync
3851 * @unsync: function to call if address should be removed
3852 *
3853 * Remove all addresses that were added to the device by dev_mc_sync().
3854 */
3855 static inline void __dev_mc_unsync(struct net_device *dev,
3856 int (*unsync)(struct net_device *,
3857 const unsigned char *))
3858 {
3859 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3860 }
3861
3862 /* Functions used for secondary unicast and multicast support */
3863 void dev_set_rx_mode(struct net_device *dev);
3864 void __dev_set_rx_mode(struct net_device *dev);
3865 int dev_set_promiscuity(struct net_device *dev, int inc);
3866 int dev_set_allmulti(struct net_device *dev, int inc);
3867 void netdev_state_change(struct net_device *dev);
3868 void netdev_notify_peers(struct net_device *dev);
3869 void netdev_features_change(struct net_device *dev);
3870 /* Load a device via the kmod */
3871 void dev_load(struct net *net, const char *name);
3872 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3873 struct rtnl_link_stats64 *storage);
3874 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3875 const struct net_device_stats *netdev_stats);
3876
3877 extern int netdev_max_backlog;
3878 extern int netdev_tstamp_prequeue;
3879 extern int weight_p;
3880 extern int dev_weight_rx_bias;
3881 extern int dev_weight_tx_bias;
3882 extern int dev_rx_weight;
3883 extern int dev_tx_weight;
3884
3885 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3886 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3887 struct list_head **iter);
3888 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3889 struct list_head **iter);
3890
3891 /* iterate through upper list, must be called under RCU read lock */
3892 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3893 for (iter = &(dev)->adj_list.upper, \
3894 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3895 updev; \
3896 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3897
3898 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3899 int (*fn)(struct net_device *upper_dev,
3900 void *data),
3901 void *data);
3902
3903 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3904 struct net_device *upper_dev);
3905
3906 bool netdev_has_any_upper_dev(struct net_device *dev);
3907
3908 void *netdev_lower_get_next_private(struct net_device *dev,
3909 struct list_head **iter);
3910 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3911 struct list_head **iter);
3912
3913 #define netdev_for_each_lower_private(dev, priv, iter) \
3914 for (iter = (dev)->adj_list.lower.next, \
3915 priv = netdev_lower_get_next_private(dev, &(iter)); \
3916 priv; \
3917 priv = netdev_lower_get_next_private(dev, &(iter)))
3918
3919 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3920 for (iter = &(dev)->adj_list.lower, \
3921 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3922 priv; \
3923 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3924
3925 void *netdev_lower_get_next(struct net_device *dev,
3926 struct list_head **iter);
3927
3928 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3929 for (iter = (dev)->adj_list.lower.next, \
3930 ldev = netdev_lower_get_next(dev, &(iter)); \
3931 ldev; \
3932 ldev = netdev_lower_get_next(dev, &(iter)))
3933
3934 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3935 struct list_head **iter);
3936 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3937 struct list_head **iter);
3938
3939 int netdev_walk_all_lower_dev(struct net_device *dev,
3940 int (*fn)(struct net_device *lower_dev,
3941 void *data),
3942 void *data);
3943 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3944 int (*fn)(struct net_device *lower_dev,
3945 void *data),
3946 void *data);
3947
3948 void *netdev_adjacent_get_private(struct list_head *adj_list);
3949 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3950 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3951 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3952 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3953 struct netlink_ext_ack *extack);
3954 int netdev_master_upper_dev_link(struct net_device *dev,
3955 struct net_device *upper_dev,
3956 void *upper_priv, void *upper_info,
3957 struct netlink_ext_ack *extack);
3958 void netdev_upper_dev_unlink(struct net_device *dev,
3959 struct net_device *upper_dev);
3960 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3961 void *netdev_lower_dev_get_private(struct net_device *dev,
3962 struct net_device *lower_dev);
3963 void netdev_lower_state_changed(struct net_device *lower_dev,
3964 void *lower_state_info);
3965
3966 /* RSS keys are 40 or 52 bytes long */
3967 #define NETDEV_RSS_KEY_LEN 52
3968 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3969 void netdev_rss_key_fill(void *buffer, size_t len);
3970
3971 int dev_get_nest_level(struct net_device *dev);
3972 int skb_checksum_help(struct sk_buff *skb);
3973 int skb_crc32c_csum_help(struct sk_buff *skb);
3974 int skb_csum_hwoffload_help(struct sk_buff *skb,
3975 const netdev_features_t features);
3976
3977 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3978 netdev_features_t features, bool tx_path);
3979 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3980 netdev_features_t features);
3981
3982 struct netdev_bonding_info {
3983 ifslave slave;
3984 ifbond master;
3985 };
3986
3987 struct netdev_notifier_bonding_info {
3988 struct netdev_notifier_info info; /* must be first */
3989 struct netdev_bonding_info bonding_info;
3990 };
3991
3992 void netdev_bonding_info_change(struct net_device *dev,
3993 struct netdev_bonding_info *bonding_info);
3994
3995 static inline
3996 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3997 {
3998 return __skb_gso_segment(skb, features, true);
3999 }
4000 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4001
4002 static inline bool can_checksum_protocol(netdev_features_t features,
4003 __be16 protocol)
4004 {
4005 if (protocol == htons(ETH_P_FCOE))
4006 return !!(features & NETIF_F_FCOE_CRC);
4007
4008 /* Assume this is an IP checksum (not SCTP CRC) */
4009
4010 if (features & NETIF_F_HW_CSUM) {
4011 /* Can checksum everything */
4012 return true;
4013 }
4014
4015 switch (protocol) {
4016 case htons(ETH_P_IP):
4017 return !!(features & NETIF_F_IP_CSUM);
4018 case htons(ETH_P_IPV6):
4019 return !!(features & NETIF_F_IPV6_CSUM);
4020 default:
4021 return false;
4022 }
4023 }
4024
4025 #ifdef CONFIG_BUG
4026 void netdev_rx_csum_fault(struct net_device *dev);
4027 #else
4028 static inline void netdev_rx_csum_fault(struct net_device *dev)
4029 {
4030 }
4031 #endif
4032 /* rx skb timestamps */
4033 void net_enable_timestamp(void);
4034 void net_disable_timestamp(void);
4035
4036 #ifdef CONFIG_PROC_FS
4037 int __init dev_proc_init(void);
4038 #else
4039 #define dev_proc_init() 0
4040 #endif
4041
4042 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4043 struct sk_buff *skb, struct net_device *dev,
4044 bool more)
4045 {
4046 skb->xmit_more = more ? 1 : 0;
4047 return ops->ndo_start_xmit(skb, dev);
4048 }
4049
4050 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4051 struct netdev_queue *txq, bool more)
4052 {
4053 const struct net_device_ops *ops = dev->netdev_ops;
4054 int rc;
4055
4056 rc = __netdev_start_xmit(ops, skb, dev, more);
4057 if (rc == NETDEV_TX_OK)
4058 txq_trans_update(txq);
4059
4060 return rc;
4061 }
4062
4063 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4064 const void *ns);
4065 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4066 const void *ns);
4067
4068 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4069 {
4070 return netdev_class_create_file_ns(class_attr, NULL);
4071 }
4072
4073 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4074 {
4075 netdev_class_remove_file_ns(class_attr, NULL);
4076 }
4077
4078 extern const struct kobj_ns_type_operations net_ns_type_operations;
4079
4080 const char *netdev_drivername(const struct net_device *dev);
4081
4082 void linkwatch_run_queue(void);
4083
4084 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4085 netdev_features_t f2)
4086 {
4087 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4088 if (f1 & NETIF_F_HW_CSUM)
4089 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4090 else
4091 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4092 }
4093
4094 return f1 & f2;
4095 }
4096
4097 static inline netdev_features_t netdev_get_wanted_features(
4098 struct net_device *dev)
4099 {
4100 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4101 }
4102 netdev_features_t netdev_increment_features(netdev_features_t all,
4103 netdev_features_t one, netdev_features_t mask);
4104
4105 /* Allow TSO being used on stacked device :
4106 * Performing the GSO segmentation before last device
4107 * is a performance improvement.
4108 */
4109 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4110 netdev_features_t mask)
4111 {
4112 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4113 }
4114
4115 int __netdev_update_features(struct net_device *dev);
4116 void netdev_update_features(struct net_device *dev);
4117 void netdev_change_features(struct net_device *dev);
4118
4119 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4120 struct net_device *dev);
4121
4122 netdev_features_t passthru_features_check(struct sk_buff *skb,
4123 struct net_device *dev,
4124 netdev_features_t features);
4125 netdev_features_t netif_skb_features(struct sk_buff *skb);
4126
4127 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4128 {
4129 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4130
4131 /* check flags correspondence */
4132 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4133 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4134 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4135 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4136 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4137 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4138 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4139 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4140 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4141 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4142 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4143 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4144 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4145 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4146 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4147 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4148 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4149
4150 return (features & feature) == feature;
4151 }
4152
4153 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4154 {
4155 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4156 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4157 }
4158
4159 static inline bool netif_needs_gso(struct sk_buff *skb,
4160 netdev_features_t features)
4161 {
4162 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4163 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4164 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4165 }
4166
4167 static inline void netif_set_gso_max_size(struct net_device *dev,
4168 unsigned int size)
4169 {
4170 dev->gso_max_size = size;
4171 }
4172
4173 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4174 int pulled_hlen, u16 mac_offset,
4175 int mac_len)
4176 {
4177 skb->protocol = protocol;
4178 skb->encapsulation = 1;
4179 skb_push(skb, pulled_hlen);
4180 skb_reset_transport_header(skb);
4181 skb->mac_header = mac_offset;
4182 skb->network_header = skb->mac_header + mac_len;
4183 skb->mac_len = mac_len;
4184 }
4185
4186 static inline bool netif_is_macsec(const struct net_device *dev)
4187 {
4188 return dev->priv_flags & IFF_MACSEC;
4189 }
4190
4191 static inline bool netif_is_macvlan(const struct net_device *dev)
4192 {
4193 return dev->priv_flags & IFF_MACVLAN;
4194 }
4195
4196 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4197 {
4198 return dev->priv_flags & IFF_MACVLAN_PORT;
4199 }
4200
4201 static inline bool netif_is_ipvlan(const struct net_device *dev)
4202 {
4203 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4204 }
4205
4206 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4207 {
4208 return dev->priv_flags & IFF_IPVLAN_MASTER;
4209 }
4210
4211 static inline bool netif_is_bond_master(const struct net_device *dev)
4212 {
4213 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4214 }
4215
4216 static inline bool netif_is_bond_slave(const struct net_device *dev)
4217 {
4218 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4219 }
4220
4221 static inline bool netif_supports_nofcs(struct net_device *dev)
4222 {
4223 return dev->priv_flags & IFF_SUPP_NOFCS;
4224 }
4225
4226 static inline bool netif_is_l3_master(const struct net_device *dev)
4227 {
4228 return dev->priv_flags & IFF_L3MDEV_MASTER;
4229 }
4230
4231 static inline bool netif_is_l3_slave(const struct net_device *dev)
4232 {
4233 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4234 }
4235
4236 static inline bool netif_is_bridge_master(const struct net_device *dev)
4237 {
4238 return dev->priv_flags & IFF_EBRIDGE;
4239 }
4240
4241 static inline bool netif_is_bridge_port(const struct net_device *dev)
4242 {
4243 return dev->priv_flags & IFF_BRIDGE_PORT;
4244 }
4245
4246 static inline bool netif_is_ovs_master(const struct net_device *dev)
4247 {
4248 return dev->priv_flags & IFF_OPENVSWITCH;
4249 }
4250
4251 static inline bool netif_is_ovs_port(const struct net_device *dev)
4252 {
4253 return dev->priv_flags & IFF_OVS_DATAPATH;
4254 }
4255
4256 static inline bool netif_is_team_master(const struct net_device *dev)
4257 {
4258 return dev->priv_flags & IFF_TEAM;
4259 }
4260
4261 static inline bool netif_is_team_port(const struct net_device *dev)
4262 {
4263 return dev->priv_flags & IFF_TEAM_PORT;
4264 }
4265
4266 static inline bool netif_is_lag_master(const struct net_device *dev)
4267 {
4268 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4269 }
4270
4271 static inline bool netif_is_lag_port(const struct net_device *dev)
4272 {
4273 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4274 }
4275
4276 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4277 {
4278 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4279 }
4280
4281 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4282 static inline void netif_keep_dst(struct net_device *dev)
4283 {
4284 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4285 }
4286
4287 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4288 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4289 {
4290 /* TODO: reserve and use an additional IFF bit, if we get more users */
4291 return dev->priv_flags & IFF_MACSEC;
4292 }
4293
4294 extern struct pernet_operations __net_initdata loopback_net_ops;
4295
4296 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4297
4298 /* netdev_printk helpers, similar to dev_printk */
4299
4300 static inline const char *netdev_name(const struct net_device *dev)
4301 {
4302 if (!dev->name[0] || strchr(dev->name, '%'))
4303 return "(unnamed net_device)";
4304 return dev->name;
4305 }
4306
4307 static inline bool netdev_unregistering(const struct net_device *dev)
4308 {
4309 return dev->reg_state == NETREG_UNREGISTERING;
4310 }
4311
4312 static inline const char *netdev_reg_state(const struct net_device *dev)
4313 {
4314 switch (dev->reg_state) {
4315 case NETREG_UNINITIALIZED: return " (uninitialized)";
4316 case NETREG_REGISTERED: return "";
4317 case NETREG_UNREGISTERING: return " (unregistering)";
4318 case NETREG_UNREGISTERED: return " (unregistered)";
4319 case NETREG_RELEASED: return " (released)";
4320 case NETREG_DUMMY: return " (dummy)";
4321 }
4322
4323 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4324 return " (unknown)";
4325 }
4326
4327 __printf(3, 4)
4328 void netdev_printk(const char *level, const struct net_device *dev,
4329 const char *format, ...);
4330 __printf(2, 3)
4331 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4332 __printf(2, 3)
4333 void netdev_alert(const struct net_device *dev, const char *format, ...);
4334 __printf(2, 3)
4335 void netdev_crit(const struct net_device *dev, const char *format, ...);
4336 __printf(2, 3)
4337 void netdev_err(const struct net_device *dev, const char *format, ...);
4338 __printf(2, 3)
4339 void netdev_warn(const struct net_device *dev, const char *format, ...);
4340 __printf(2, 3)
4341 void netdev_notice(const struct net_device *dev, const char *format, ...);
4342 __printf(2, 3)
4343 void netdev_info(const struct net_device *dev, const char *format, ...);
4344
4345 #define netdev_level_once(level, dev, fmt, ...) \
4346 do { \
4347 static bool __print_once __read_mostly; \
4348 \
4349 if (!__print_once) { \
4350 __print_once = true; \
4351 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4352 } \
4353 } while (0)
4354
4355 #define netdev_emerg_once(dev, fmt, ...) \
4356 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4357 #define netdev_alert_once(dev, fmt, ...) \
4358 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4359 #define netdev_crit_once(dev, fmt, ...) \
4360 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4361 #define netdev_err_once(dev, fmt, ...) \
4362 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4363 #define netdev_warn_once(dev, fmt, ...) \
4364 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4365 #define netdev_notice_once(dev, fmt, ...) \
4366 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4367 #define netdev_info_once(dev, fmt, ...) \
4368 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4369
4370 #define MODULE_ALIAS_NETDEV(device) \
4371 MODULE_ALIAS("netdev-" device)
4372
4373 #if defined(CONFIG_DYNAMIC_DEBUG)
4374 #define netdev_dbg(__dev, format, args...) \
4375 do { \
4376 dynamic_netdev_dbg(__dev, format, ##args); \
4377 } while (0)
4378 #elif defined(DEBUG)
4379 #define netdev_dbg(__dev, format, args...) \
4380 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4381 #else
4382 #define netdev_dbg(__dev, format, args...) \
4383 ({ \
4384 if (0) \
4385 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4386 })
4387 #endif
4388
4389 #if defined(VERBOSE_DEBUG)
4390 #define netdev_vdbg netdev_dbg
4391 #else
4392
4393 #define netdev_vdbg(dev, format, args...) \
4394 ({ \
4395 if (0) \
4396 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4397 0; \
4398 })
4399 #endif
4400
4401 /*
4402 * netdev_WARN() acts like dev_printk(), but with the key difference
4403 * of using a WARN/WARN_ON to get the message out, including the
4404 * file/line information and a backtrace.
4405 */
4406 #define netdev_WARN(dev, format, args...) \
4407 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4408 netdev_reg_state(dev), ##args)
4409
4410 #define netdev_WARN_ONCE(dev, condition, format, arg...) \
4411 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev) \
4412 netdev_reg_state(dev), ##args)
4413
4414 /* netif printk helpers, similar to netdev_printk */
4415
4416 #define netif_printk(priv, type, level, dev, fmt, args...) \
4417 do { \
4418 if (netif_msg_##type(priv)) \
4419 netdev_printk(level, (dev), fmt, ##args); \
4420 } while (0)
4421
4422 #define netif_level(level, priv, type, dev, fmt, args...) \
4423 do { \
4424 if (netif_msg_##type(priv)) \
4425 netdev_##level(dev, fmt, ##args); \
4426 } while (0)
4427
4428 #define netif_emerg(priv, type, dev, fmt, args...) \
4429 netif_level(emerg, priv, type, dev, fmt, ##args)
4430 #define netif_alert(priv, type, dev, fmt, args...) \
4431 netif_level(alert, priv, type, dev, fmt, ##args)
4432 #define netif_crit(priv, type, dev, fmt, args...) \
4433 netif_level(crit, priv, type, dev, fmt, ##args)
4434 #define netif_err(priv, type, dev, fmt, args...) \
4435 netif_level(err, priv, type, dev, fmt, ##args)
4436 #define netif_warn(priv, type, dev, fmt, args...) \
4437 netif_level(warn, priv, type, dev, fmt, ##args)
4438 #define netif_notice(priv, type, dev, fmt, args...) \
4439 netif_level(notice, priv, type, dev, fmt, ##args)
4440 #define netif_info(priv, type, dev, fmt, args...) \
4441 netif_level(info, priv, type, dev, fmt, ##args)
4442
4443 #if defined(CONFIG_DYNAMIC_DEBUG)
4444 #define netif_dbg(priv, type, netdev, format, args...) \
4445 do { \
4446 if (netif_msg_##type(priv)) \
4447 dynamic_netdev_dbg(netdev, format, ##args); \
4448 } while (0)
4449 #elif defined(DEBUG)
4450 #define netif_dbg(priv, type, dev, format, args...) \
4451 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4452 #else
4453 #define netif_dbg(priv, type, dev, format, args...) \
4454 ({ \
4455 if (0) \
4456 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4457 0; \
4458 })
4459 #endif
4460
4461 /* if @cond then downgrade to debug, else print at @level */
4462 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4463 do { \
4464 if (cond) \
4465 netif_dbg(priv, type, netdev, fmt, ##args); \
4466 else \
4467 netif_ ## level(priv, type, netdev, fmt, ##args); \
4468 } while (0)
4469
4470 #if defined(VERBOSE_DEBUG)
4471 #define netif_vdbg netif_dbg
4472 #else
4473 #define netif_vdbg(priv, type, dev, format, args...) \
4474 ({ \
4475 if (0) \
4476 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4477 0; \
4478 })
4479 #endif
4480
4481 /*
4482 * The list of packet types we will receive (as opposed to discard)
4483 * and the routines to invoke.
4484 *
4485 * Why 16. Because with 16 the only overlap we get on a hash of the
4486 * low nibble of the protocol value is RARP/SNAP/X.25.
4487 *
4488 * 0800 IP
4489 * 0001 802.3
4490 * 0002 AX.25
4491 * 0004 802.2
4492 * 8035 RARP
4493 * 0005 SNAP
4494 * 0805 X.25
4495 * 0806 ARP
4496 * 8137 IPX
4497 * 0009 Localtalk
4498 * 86DD IPv6
4499 */
4500 #define PTYPE_HASH_SIZE (16)
4501 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4502
4503 #endif /* _LINUX_NETDEVICE_H */