]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/netdevice.h
1d238d54c4840da5bdc0e5aa0ec960959b578f9a
[mirror_ubuntu-bionic-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/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #ifdef CONFIG_DCB
45 #include <net/dcbnl.h>
46 #endif
47 #include <net/netprio_cgroup.h>
48
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
55
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 struct dsa_switch_tree;
60
61 /* 802.11 specific */
62 struct wireless_dev;
63 /* 802.15.4 specific */
64 struct wpan_dev;
65 struct mpls_dev;
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct bpf_prog;
69 struct xdp_buff;
70
71 void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
73
74 /* Backlog congestion levels */
75 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76 #define NET_RX_DROP 1 /* packet dropped */
77
78 /*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
92 * others are propagated to higher layers.
93 */
94
95 /* qdisc ->enqueue() return codes. */
96 #define NET_XMIT_SUCCESS 0x00
97 #define NET_XMIT_DROP 0x01 /* skb dropped */
98 #define NET_XMIT_CN 0x02 /* congestion notification */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
109
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 };
115 typedef enum netdev_tx netdev_tx_t;
116
117 /*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121 static inline bool dev_xmit_complete(int rc)
122 {
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133 }
134
135 /*
136 * Compute the worst-case header length according to the protocols
137 * used.
138 */
139
140 #if defined(CONFIG_HYPERV_NET)
141 # define LL_MAX_HEADER 128
142 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
143 # if defined(CONFIG_MAC80211_MESH)
144 # define LL_MAX_HEADER 128
145 # else
146 # define LL_MAX_HEADER 96
147 # endif
148 #else
149 # define LL_MAX_HEADER 32
150 #endif
151
152 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
154 #define MAX_HEADER LL_MAX_HEADER
155 #else
156 #define MAX_HEADER (LL_MAX_HEADER + 48)
157 #endif
158
159 /*
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
162 */
163
164 struct net_device_stats {
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
174 unsigned long collisions;
175 unsigned long rx_length_errors;
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188 };
189
190
191 #include <linux/cache.h>
192 #include <linux/skbuff.h>
193
194 #ifdef CONFIG_RPS
195 #include <linux/static_key.h>
196 extern struct static_key rps_needed;
197 extern struct static_key rfs_needed;
198 #endif
199
200 struct neighbour;
201 struct neigh_parms;
202 struct sk_buff;
203
204 struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
208 #define NETDEV_HW_ADDR_T_LAN 1
209 #define NETDEV_HW_ADDR_T_SAN 2
210 #define NETDEV_HW_ADDR_T_SLAVE 3
211 #define NETDEV_HW_ADDR_T_UNICAST 4
212 #define NETDEV_HW_ADDR_T_MULTICAST 5
213 bool global_use;
214 int sync_cnt;
215 int refcount;
216 int synced;
217 struct rcu_head rcu_head;
218 };
219
220 struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223 };
224
225 #define netdev_hw_addr_list_count(l) ((l)->count)
226 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227 #define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
229
230 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232 #define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234
235 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
237 #define netdev_for_each_mc_addr(ha, dev) \
238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
239
240 struct hh_cache {
241 unsigned int hh_len;
242 seqlock_t hh_lock;
243
244 /* cached hardware header; allow for machine alignment needs. */
245 #define HH_DATA_MOD 16
246 #define HH_DATA_OFF(__len) \
247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
248 #define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 };
252
253 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261 #define LL_RESERVED_SPACE(dev) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265
266 struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
269 const void *saddr, unsigned int len);
270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275 bool (*validate)(const char *ll_header, unsigned int len);
276 };
277
278 /* These flag bits are private to the generic network queueing
279 * layer; they may not be explicitly referenced by any other
280 * code.
281 */
282
283 enum netdev_state_t {
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
286 __LINK_STATE_NOCARRIER,
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
289 };
290
291
292 /*
293 * This structure holds boot-time configured netdevice settings. They
294 * are then used in the device probing.
295 */
296 struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299 };
300 #define NETDEV_BOOT_SETUP_MAX 8
301
302 int __init netdev_boot_setup(char *str);
303
304 /*
305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307 struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
311 * to the per-CPU poll_list, and whoever clears that bit
312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
318 unsigned int gro_count;
319 int (*poll)(struct napi_struct *, int);
320 #ifdef CONFIG_NETPOLL
321 int poll_owner;
322 #endif
323 struct net_device *dev;
324 struct sk_buff *gro_list;
325 struct sk_buff *skb;
326 struct hrtimer timer;
327 struct list_head dev_list;
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
330 };
331
332 enum {
333 NAPI_STATE_SCHED, /* Poll is scheduled */
334 NAPI_STATE_MISSED, /* reschedule a napi */
335 NAPI_STATE_DISABLE, /* Disable pending */
336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340 };
341
342 enum {
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
350 };
351
352 enum gro_result {
353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
358 GRO_CONSUMED,
359 };
360 typedef enum gro_result gro_result_t;
361
362 /*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
384 * If the rx_handler consumed the skb in some way, it should return
385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
386 * the skb to be delivered in some other way.
387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
392 * If the rx_handler decides the skb should be ignored, it should return
393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
394 * are registered on exact device (ptype->dev == skb->dev).
395 *
396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403 enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408 };
409 typedef enum rx_handler_result rx_handler_result_t;
410 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
411
412 void __napi_schedule(struct napi_struct *n);
413 void __napi_schedule_irqoff(struct napi_struct *n);
414
415 static inline bool napi_disable_pending(struct napi_struct *n)
416 {
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418 }
419
420 bool napi_schedule_prep(struct napi_struct *n);
421
422 /**
423 * napi_schedule - schedule NAPI poll
424 * @n: NAPI context
425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429 static inline void napi_schedule(struct napi_struct *n)
430 {
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433 }
434
435 /**
436 * napi_schedule_irqoff - schedule NAPI poll
437 * @n: NAPI context
438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441 static inline void napi_schedule_irqoff(struct napi_struct *n)
442 {
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445 }
446
447 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
448 static inline bool napi_reschedule(struct napi_struct *napi)
449 {
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
452 return true;
453 }
454 return false;
455 }
456
457 bool napi_complete_done(struct napi_struct *n, int work_done);
458 /**
459 * napi_complete - NAPI processing complete
460 * @n: NAPI context
461 *
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
464 * Return false if device should avoid rearming interrupts.
465 */
466 static inline bool napi_complete(struct napi_struct *n)
467 {
468 return napi_complete_done(n, 0);
469 }
470
471 /**
472 * napi_hash_del - remove a NAPI from global table
473 * @napi: NAPI context
474 *
475 * Warning: caller must observe RCU grace period
476 * before freeing memory containing @napi, if
477 * this function returns true.
478 * Note: core networking stack automatically calls it
479 * from netif_napi_del().
480 * Drivers might want to call this helper to combine all
481 * the needed RCU grace periods into a single one.
482 */
483 bool napi_hash_del(struct napi_struct *napi);
484
485 /**
486 * napi_disable - prevent NAPI from scheduling
487 * @n: NAPI context
488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
492 void napi_disable(struct napi_struct *n);
493
494 /**
495 * napi_enable - enable NAPI scheduling
496 * @n: NAPI context
497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501 static inline void napi_enable(struct napi_struct *n)
502 {
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
504 smp_mb__before_atomic();
505 clear_bit(NAPI_STATE_SCHED, &n->state);
506 clear_bit(NAPI_STATE_NPSVC, &n->state);
507 }
508
509 /**
510 * napi_synchronize - wait until NAPI is not running
511 * @n: NAPI context
512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517 static inline void napi_synchronize(const struct napi_struct *n)
518 {
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
524 }
525
526 enum netdev_queue_state_t {
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
529 __QUEUE_STATE_FROZEN,
530 };
531
532 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
542 /*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
551
552 struct netdev_queue {
553 /*
554 * read-mostly part
555 */
556 struct net_device *dev;
557 struct Qdisc __rcu *qdisc;
558 struct Qdisc *qdisc_sleeping;
559 #ifdef CONFIG_SYSFS
560 struct kobject kobj;
561 #endif
562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564 #endif
565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
571 /*
572 * write-mostly part
573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
576 /*
577 * Time (in jiffies) of last Tx
578 */
579 unsigned long trans_start;
580
581 unsigned long state;
582
583 #ifdef CONFIG_BQL
584 struct dql dql;
585 #endif
586 } ____cacheline_aligned_in_smp;
587
588 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589 {
590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 return q->numa_node;
592 #else
593 return NUMA_NO_NODE;
594 #endif
595 }
596
597 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598 {
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 q->numa_node = node;
601 #endif
602 }
603
604 #ifdef CONFIG_RPS
605 /*
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
608 */
609 struct rps_map {
610 unsigned int len;
611 struct rcu_head rcu;
612 u16 cpus[0];
613 };
614 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
615
616 /*
617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
620 */
621 struct rps_dev_flow {
622 u16 cpu;
623 u16 filter;
624 unsigned int last_qtail;
625 };
626 #define RPS_NO_FILTER 0xffff
627
628 /*
629 * The rps_dev_flow_table structure contains a table of flow mappings.
630 */
631 struct rps_dev_flow_table {
632 unsigned int mask;
633 struct rcu_head rcu;
634 struct rps_dev_flow flows[0];
635 };
636 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
637 ((_num) * sizeof(struct rps_dev_flow)))
638
639 /*
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
644 * rps_cpu_mask is used to partition the space, depending on number of
645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
647 * meaning we use 32-6=26 bits for the hash.
648 */
649 struct rps_sock_flow_table {
650 u32 mask;
651
652 u32 ents[0] ____cacheline_aligned_in_smp;
653 };
654 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655
656 #define RPS_NO_CPU 0xffff
657
658 extern u32 rps_cpu_mask;
659 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660
661 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 u32 hash)
663 {
664 if (table && hash) {
665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
667
668 /* We only give a hint, preemption can change CPU under us */
669 val |= raw_smp_processor_id();
670
671 if (table->ents[index] != val)
672 table->ents[index] = val;
673 }
674 }
675
676 #ifdef CONFIG_RFS_ACCEL
677 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 u16 filter_id);
679 #endif
680 #endif /* CONFIG_RPS */
681
682 /* This structure contains an instance of an RX queue. */
683 struct netdev_rx_queue {
684 #ifdef CONFIG_RPS
685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
687 #endif
688 struct kobject kobj;
689 struct net_device *dev;
690 } ____cacheline_aligned_in_smp;
691
692 /*
693 * RX queue sysfs structures and functions.
694 */
695 struct rx_queue_attribute {
696 struct attribute attr;
697 ssize_t (*show)(struct netdev_rx_queue *queue,
698 struct rx_queue_attribute *attr, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 struct rx_queue_attribute *attr, const char *buf, size_t len);
701 };
702
703 #ifdef CONFIG_XPS
704 /*
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
707 */
708 struct xps_map {
709 unsigned int len;
710 unsigned int alloc_len;
711 struct rcu_head rcu;
712 u16 queues[0];
713 };
714 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
715 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
717
718 /*
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 */
721 struct xps_dev_maps {
722 struct rcu_head rcu;
723 struct xps_map __rcu *cpu_map[0];
724 };
725 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
727 #endif /* CONFIG_XPS */
728
729 #define TC_MAX_QUEUE 16
730 #define TC_BITMASK 15
731 /* HW offloaded queuing disciplines txq count and offset maps */
732 struct netdev_tc_txq {
733 u16 count;
734 u16 offset;
735 };
736
737 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738 /*
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
741 */
742 struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
749 char model[256];
750 char model_description[256];
751 };
752 #endif
753
754 #define MAX_PHYS_ITEM_ID_LEN 32
755
756 /* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
758 */
759 struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
761 unsigned char id_len;
762 };
763
764 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
766 {
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
769 }
770
771 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
773
774 enum tc_setup_type {
775 TC_SETUP_MQPRIO,
776 TC_SETUP_CLSU32,
777 TC_SETUP_CLSFLOWER,
778 TC_SETUP_CLSMATCHALL,
779 TC_SETUP_CLSBPF,
780 };
781
782 /* These structures hold the attributes of xdp state that are being passed
783 * to the netdevice through the xdp op.
784 */
785 enum xdp_netdev_command {
786 /* Set or clear a bpf program used in the earliest stages of packet
787 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
788 * is responsible for calling bpf_prog_put on any old progs that are
789 * stored. In case of error, the callee need not release the new prog
790 * reference, but on success it takes ownership and must bpf_prog_put
791 * when it is no longer used.
792 */
793 XDP_SETUP_PROG,
794 XDP_SETUP_PROG_HW,
795 /* Check if a bpf program is set on the device. The callee should
796 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
797 * is equivalent to XDP_ATTACHED_DRV.
798 */
799 XDP_QUERY_PROG,
800 };
801
802 struct netlink_ext_ack;
803
804 struct netdev_xdp {
805 enum xdp_netdev_command command;
806 union {
807 /* XDP_SETUP_PROG */
808 struct {
809 u32 flags;
810 struct bpf_prog *prog;
811 struct netlink_ext_ack *extack;
812 };
813 /* XDP_QUERY_PROG */
814 struct {
815 u8 prog_attached;
816 u32 prog_id;
817 };
818 };
819 };
820
821 #ifdef CONFIG_XFRM_OFFLOAD
822 struct xfrmdev_ops {
823 int (*xdo_dev_state_add) (struct xfrm_state *x);
824 void (*xdo_dev_state_delete) (struct xfrm_state *x);
825 void (*xdo_dev_state_free) (struct xfrm_state *x);
826 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
827 struct xfrm_state *x);
828 };
829 #endif
830
831 /*
832 * This structure defines the management hooks for network devices.
833 * The following hooks can be defined; unless noted otherwise, they are
834 * optional and can be filled with a null pointer.
835 *
836 * int (*ndo_init)(struct net_device *dev);
837 * This function is called once when a network device is registered.
838 * The network device can use this for any late stage initialization
839 * or semantic validation. It can fail with an error code which will
840 * be propagated back to register_netdev.
841 *
842 * void (*ndo_uninit)(struct net_device *dev);
843 * This function is called when device is unregistered or when registration
844 * fails. It is not called if init fails.
845 *
846 * int (*ndo_open)(struct net_device *dev);
847 * This function is called when a network device transitions to the up
848 * state.
849 *
850 * int (*ndo_stop)(struct net_device *dev);
851 * This function is called when a network device transitions to the down
852 * state.
853 *
854 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
855 * struct net_device *dev);
856 * Called when a packet needs to be transmitted.
857 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
858 * the queue before that can happen; it's for obsolete devices and weird
859 * corner cases, but the stack really does a non-trivial amount
860 * of useless work if you return NETDEV_TX_BUSY.
861 * Required; cannot be NULL.
862 *
863 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
864 * struct net_device *dev
865 * netdev_features_t features);
866 * Called by core transmit path to determine if device is capable of
867 * performing offload operations on a given packet. This is to give
868 * the device an opportunity to implement any restrictions that cannot
869 * be otherwise expressed by feature flags. The check is called with
870 * the set of features that the stack has calculated and it returns
871 * those the driver believes to be appropriate.
872 *
873 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
874 * void *accel_priv, select_queue_fallback_t fallback);
875 * Called to decide which queue to use when device supports multiple
876 * transmit queues.
877 *
878 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
879 * This function is called to allow device receiver to make
880 * changes to configuration when multicast or promiscuous is enabled.
881 *
882 * void (*ndo_set_rx_mode)(struct net_device *dev);
883 * This function is called device changes address list filtering.
884 * If driver handles unicast address filtering, it should set
885 * IFF_UNICAST_FLT in its priv_flags.
886 *
887 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
888 * This function is called when the Media Access Control address
889 * needs to be changed. If this interface is not defined, the
890 * MAC address can not be changed.
891 *
892 * int (*ndo_validate_addr)(struct net_device *dev);
893 * Test if Media Access Control address is valid for the device.
894 *
895 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
896 * Called when a user requests an ioctl which can't be handled by
897 * the generic interface code. If not defined ioctls return
898 * not supported error code.
899 *
900 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
901 * Used to set network devices bus interface parameters. This interface
902 * is retained for legacy reasons; new devices should use the bus
903 * interface (PCI) for low level management.
904 *
905 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
906 * Called when a user wants to change the Maximum Transfer Unit
907 * of a device.
908 *
909 * void (*ndo_tx_timeout)(struct net_device *dev);
910 * Callback used when the transmitter has not made any progress
911 * for dev->watchdog ticks.
912 *
913 * void (*ndo_get_stats64)(struct net_device *dev,
914 * struct rtnl_link_stats64 *storage);
915 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
916 * Called when a user wants to get the network device usage
917 * statistics. Drivers must do one of the following:
918 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
919 * rtnl_link_stats64 structure passed by the caller.
920 * 2. Define @ndo_get_stats to update a net_device_stats structure
921 * (which should normally be dev->stats) and return a pointer to
922 * it. The structure may be changed asynchronously only if each
923 * field is written atomically.
924 * 3. Update dev->stats asynchronously and atomically, and define
925 * neither operation.
926 *
927 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
928 * Return true if this device supports offload stats of this attr_id.
929 *
930 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
931 * void *attr_data)
932 * Get statistics for offload operations by attr_id. Write it into the
933 * attr_data pointer.
934 *
935 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
936 * If device supports VLAN filtering this function is called when a
937 * VLAN id is registered.
938 *
939 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
940 * If device supports VLAN filtering this function is called when a
941 * VLAN id is unregistered.
942 *
943 * void (*ndo_poll_controller)(struct net_device *dev);
944 *
945 * SR-IOV management functions.
946 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
947 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
948 * u8 qos, __be16 proto);
949 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
950 * int max_tx_rate);
951 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
952 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
953 * int (*ndo_get_vf_config)(struct net_device *dev,
954 * int vf, struct ifla_vf_info *ivf);
955 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
956 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
957 * struct nlattr *port[]);
958 *
959 * Enable or disable the VF ability to query its RSS Redirection Table and
960 * Hash Key. This is needed since on some devices VF share this information
961 * with PF and querying it may introduce a theoretical security risk.
962 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
963 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
964 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
965 * void *type_data);
966 * Called to setup any 'tc' scheduler, classifier or action on @dev.
967 * This is always called from the stack with the rtnl lock held and netif
968 * tx queues stopped. This allows the netdevice to perform queue
969 * management safely.
970 *
971 * Fiber Channel over Ethernet (FCoE) offload functions.
972 * int (*ndo_fcoe_enable)(struct net_device *dev);
973 * Called when the FCoE protocol stack wants to start using LLD for FCoE
974 * so the underlying device can perform whatever needed configuration or
975 * initialization to support acceleration of FCoE traffic.
976 *
977 * int (*ndo_fcoe_disable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
979 * so the underlying device can perform whatever needed clean-ups to
980 * stop supporting acceleration of FCoE traffic.
981 *
982 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
983 * struct scatterlist *sgl, unsigned int sgc);
984 * Called when the FCoE Initiator wants to initialize an I/O that
985 * is a possible candidate for Direct Data Placement (DDP). The LLD can
986 * perform necessary setup and returns 1 to indicate the device is set up
987 * successfully to perform DDP on this I/O, otherwise this returns 0.
988 *
989 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
990 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
991 * indicated by the FC exchange id 'xid', so the underlying device can
992 * clean up and reuse resources for later DDP requests.
993 *
994 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
995 * struct scatterlist *sgl, unsigned int sgc);
996 * Called when the FCoE Target wants to initialize an I/O that
997 * is a possible candidate for Direct Data Placement (DDP). The LLD can
998 * perform necessary setup and returns 1 to indicate the device is set up
999 * successfully to perform DDP on this I/O, otherwise this returns 0.
1000 *
1001 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1002 * struct netdev_fcoe_hbainfo *hbainfo);
1003 * Called when the FCoE Protocol stack wants information on the underlying
1004 * device. This information is utilized by the FCoE protocol stack to
1005 * register attributes with Fiber Channel management service as per the
1006 * FC-GS Fabric Device Management Information(FDMI) specification.
1007 *
1008 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1009 * Called when the underlying device wants to override default World Wide
1010 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1011 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1012 * protocol stack to use.
1013 *
1014 * RFS acceleration.
1015 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1016 * u16 rxq_index, u32 flow_id);
1017 * Set hardware filter for RFS. rxq_index is the target queue index;
1018 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1019 * Return the filter ID on success, or a negative error code.
1020 *
1021 * Slave management functions (for bridge, bonding, etc).
1022 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to make another netdev an underling.
1024 *
1025 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to release previously enslaved netdev.
1027 *
1028 * Feature/offload setting functions.
1029 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1030 * netdev_features_t features);
1031 * Adjusts the requested feature flags according to device-specific
1032 * constraints, and returns the resulting flags. Must not modify
1033 * the device state.
1034 *
1035 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1036 * Called to update device configuration to new features. Passed
1037 * feature set might be less than what was returned by ndo_fix_features()).
1038 * Must return >0 or -errno if it changed dev->features itself.
1039 *
1040 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1041 * struct net_device *dev,
1042 * const unsigned char *addr, u16 vid, u16 flags)
1043 * Adds an FDB entry to dev for addr.
1044 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1045 * struct net_device *dev,
1046 * const unsigned char *addr, u16 vid)
1047 * Deletes the FDB entry from dev coresponding to addr.
1048 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1049 * struct net_device *dev, struct net_device *filter_dev,
1050 * int *idx)
1051 * Used to add FDB entries to dump requests. Implementers should add
1052 * entries to skb and update idx with the number of entries.
1053 *
1054 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1055 * u16 flags)
1056 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1057 * struct net_device *dev, u32 filter_mask,
1058 * int nlflags)
1059 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1060 * u16 flags);
1061 *
1062 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1063 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1064 * which do not represent real hardware may define this to allow their
1065 * userspace components to manage their virtual carrier state. Devices
1066 * that determine carrier state from physical hardware properties (eg
1067 * network cables) or protocol-dependent mechanisms (eg
1068 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1069 *
1070 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1071 * struct netdev_phys_item_id *ppid);
1072 * Called to get ID of physical port of this device. If driver does
1073 * not implement this, it is assumed that the hw is not able to have
1074 * multiple net devices on single physical port.
1075 *
1076 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1077 * struct udp_tunnel_info *ti);
1078 * Called by UDP tunnel to notify a driver about the UDP port and socket
1079 * address family that a UDP tunnel is listnening to. It is called only
1080 * when a new port starts listening. The operation is protected by the
1081 * RTNL.
1082 *
1083 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1084 * struct udp_tunnel_info *ti);
1085 * Called by UDP tunnel to notify the driver about a UDP port and socket
1086 * address family that the UDP tunnel is not listening to anymore. The
1087 * operation is protected by the RTNL.
1088 *
1089 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1090 * struct net_device *dev)
1091 * Called by upper layer devices to accelerate switching or other
1092 * station functionality into hardware. 'pdev is the lowerdev
1093 * to use for the offload and 'dev' is the net device that will
1094 * back the offload. Returns a pointer to the private structure
1095 * the upper layer will maintain.
1096 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1097 * Called by upper layer device to delete the station created
1098 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1099 * the station and priv is the structure returned by the add
1100 * operation.
1101 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1102 * int queue_index, u32 maxrate);
1103 * Called when a user wants to set a max-rate limitation of specific
1104 * TX queue.
1105 * int (*ndo_get_iflink)(const struct net_device *dev);
1106 * Called to get the iflink value of this device.
1107 * void (*ndo_change_proto_down)(struct net_device *dev,
1108 * bool proto_down);
1109 * This function is used to pass protocol port error state information
1110 * to the switch driver. The switch driver can react to the proto_down
1111 * by doing a phys down on the associated switch port.
1112 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1113 * This function is used to get egress tunnel information for given skb.
1114 * This is useful for retrieving outer tunnel header parameters while
1115 * sampling packet.
1116 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1117 * This function is used to specify the headroom that the skb must
1118 * consider when allocation skb during packet reception. Setting
1119 * appropriate rx headroom value allows avoiding skb head copy on
1120 * forward. Setting a negative value resets the rx headroom to the
1121 * default value.
1122 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1123 * This function is used to set or query state related to XDP on the
1124 * netdevice. See definition of enum xdp_netdev_command for details.
1125 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1126 * This function is used to submit a XDP packet for transmit on a
1127 * netdevice.
1128 * void (*ndo_xdp_flush)(struct net_device *dev);
1129 * This function is used to inform the driver to flush a paticular
1130 * xpd tx queue. Must be called on same CPU as xdp_xmit.
1131 */
1132 struct net_device_ops {
1133 int (*ndo_init)(struct net_device *dev);
1134 void (*ndo_uninit)(struct net_device *dev);
1135 int (*ndo_open)(struct net_device *dev);
1136 int (*ndo_stop)(struct net_device *dev);
1137 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1138 struct net_device *dev);
1139 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1140 struct net_device *dev,
1141 netdev_features_t features);
1142 u16 (*ndo_select_queue)(struct net_device *dev,
1143 struct sk_buff *skb,
1144 void *accel_priv,
1145 select_queue_fallback_t fallback);
1146 void (*ndo_change_rx_flags)(struct net_device *dev,
1147 int flags);
1148 void (*ndo_set_rx_mode)(struct net_device *dev);
1149 int (*ndo_set_mac_address)(struct net_device *dev,
1150 void *addr);
1151 int (*ndo_validate_addr)(struct net_device *dev);
1152 int (*ndo_do_ioctl)(struct net_device *dev,
1153 struct ifreq *ifr, int cmd);
1154 int (*ndo_set_config)(struct net_device *dev,
1155 struct ifmap *map);
1156 int (*ndo_change_mtu)(struct net_device *dev,
1157 int new_mtu);
1158 int (*ndo_neigh_setup)(struct net_device *dev,
1159 struct neigh_parms *);
1160 void (*ndo_tx_timeout) (struct net_device *dev);
1161
1162 void (*ndo_get_stats64)(struct net_device *dev,
1163 struct rtnl_link_stats64 *storage);
1164 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1165 int (*ndo_get_offload_stats)(int attr_id,
1166 const struct net_device *dev,
1167 void *attr_data);
1168 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1169
1170 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1171 __be16 proto, u16 vid);
1172 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1173 __be16 proto, u16 vid);
1174 #ifdef CONFIG_NET_POLL_CONTROLLER
1175 void (*ndo_poll_controller)(struct net_device *dev);
1176 int (*ndo_netpoll_setup)(struct net_device *dev,
1177 struct netpoll_info *info);
1178 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1179 #endif
1180 int (*ndo_set_vf_mac)(struct net_device *dev,
1181 int queue, u8 *mac);
1182 int (*ndo_set_vf_vlan)(struct net_device *dev,
1183 int queue, u16 vlan,
1184 u8 qos, __be16 proto);
1185 int (*ndo_set_vf_rate)(struct net_device *dev,
1186 int vf, int min_tx_rate,
1187 int max_tx_rate);
1188 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1189 int vf, bool setting);
1190 int (*ndo_set_vf_trust)(struct net_device *dev,
1191 int vf, bool setting);
1192 int (*ndo_get_vf_config)(struct net_device *dev,
1193 int vf,
1194 struct ifla_vf_info *ivf);
1195 int (*ndo_set_vf_link_state)(struct net_device *dev,
1196 int vf, int link_state);
1197 int (*ndo_get_vf_stats)(struct net_device *dev,
1198 int vf,
1199 struct ifla_vf_stats
1200 *vf_stats);
1201 int (*ndo_set_vf_port)(struct net_device *dev,
1202 int vf,
1203 struct nlattr *port[]);
1204 int (*ndo_get_vf_port)(struct net_device *dev,
1205 int vf, struct sk_buff *skb);
1206 int (*ndo_set_vf_guid)(struct net_device *dev,
1207 int vf, u64 guid,
1208 int guid_type);
1209 int (*ndo_set_vf_rss_query_en)(
1210 struct net_device *dev,
1211 int vf, bool setting);
1212 int (*ndo_setup_tc)(struct net_device *dev,
1213 enum tc_setup_type type,
1214 void *type_data);
1215 #if IS_ENABLED(CONFIG_FCOE)
1216 int (*ndo_fcoe_enable)(struct net_device *dev);
1217 int (*ndo_fcoe_disable)(struct net_device *dev);
1218 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1219 u16 xid,
1220 struct scatterlist *sgl,
1221 unsigned int sgc);
1222 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1223 u16 xid);
1224 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1225 u16 xid,
1226 struct scatterlist *sgl,
1227 unsigned int sgc);
1228 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1229 struct netdev_fcoe_hbainfo *hbainfo);
1230 #endif
1231
1232 #if IS_ENABLED(CONFIG_LIBFCOE)
1233 #define NETDEV_FCOE_WWNN 0
1234 #define NETDEV_FCOE_WWPN 1
1235 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1236 u64 *wwn, int type);
1237 #endif
1238
1239 #ifdef CONFIG_RFS_ACCEL
1240 int (*ndo_rx_flow_steer)(struct net_device *dev,
1241 const struct sk_buff *skb,
1242 u16 rxq_index,
1243 u32 flow_id);
1244 #endif
1245 int (*ndo_add_slave)(struct net_device *dev,
1246 struct net_device *slave_dev);
1247 int (*ndo_del_slave)(struct net_device *dev,
1248 struct net_device *slave_dev);
1249 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1250 netdev_features_t features);
1251 int (*ndo_set_features)(struct net_device *dev,
1252 netdev_features_t features);
1253 int (*ndo_neigh_construct)(struct net_device *dev,
1254 struct neighbour *n);
1255 void (*ndo_neigh_destroy)(struct net_device *dev,
1256 struct neighbour *n);
1257
1258 int (*ndo_fdb_add)(struct ndmsg *ndm,
1259 struct nlattr *tb[],
1260 struct net_device *dev,
1261 const unsigned char *addr,
1262 u16 vid,
1263 u16 flags);
1264 int (*ndo_fdb_del)(struct ndmsg *ndm,
1265 struct nlattr *tb[],
1266 struct net_device *dev,
1267 const unsigned char *addr,
1268 u16 vid);
1269 int (*ndo_fdb_dump)(struct sk_buff *skb,
1270 struct netlink_callback *cb,
1271 struct net_device *dev,
1272 struct net_device *filter_dev,
1273 int *idx);
1274
1275 int (*ndo_bridge_setlink)(struct net_device *dev,
1276 struct nlmsghdr *nlh,
1277 u16 flags);
1278 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1279 u32 pid, u32 seq,
1280 struct net_device *dev,
1281 u32 filter_mask,
1282 int nlflags);
1283 int (*ndo_bridge_dellink)(struct net_device *dev,
1284 struct nlmsghdr *nlh,
1285 u16 flags);
1286 int (*ndo_change_carrier)(struct net_device *dev,
1287 bool new_carrier);
1288 int (*ndo_get_phys_port_id)(struct net_device *dev,
1289 struct netdev_phys_item_id *ppid);
1290 int (*ndo_get_phys_port_name)(struct net_device *dev,
1291 char *name, size_t len);
1292 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1293 struct udp_tunnel_info *ti);
1294 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1295 struct udp_tunnel_info *ti);
1296 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1297 struct net_device *dev);
1298 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1299 void *priv);
1300
1301 int (*ndo_get_lock_subclass)(struct net_device *dev);
1302 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1303 int queue_index,
1304 u32 maxrate);
1305 int (*ndo_get_iflink)(const struct net_device *dev);
1306 int (*ndo_change_proto_down)(struct net_device *dev,
1307 bool proto_down);
1308 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1309 struct sk_buff *skb);
1310 void (*ndo_set_rx_headroom)(struct net_device *dev,
1311 int needed_headroom);
1312 int (*ndo_xdp)(struct net_device *dev,
1313 struct netdev_xdp *xdp);
1314 int (*ndo_xdp_xmit)(struct net_device *dev,
1315 struct xdp_buff *xdp);
1316 void (*ndo_xdp_flush)(struct net_device *dev);
1317 };
1318
1319 /**
1320 * enum net_device_priv_flags - &struct net_device priv_flags
1321 *
1322 * These are the &struct net_device, they are only set internally
1323 * by drivers and used in the kernel. These flags are invisible to
1324 * userspace; this means that the order of these flags can change
1325 * during any kernel release.
1326 *
1327 * You should have a pretty good reason to be extending these flags.
1328 *
1329 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1330 * @IFF_EBRIDGE: Ethernet bridging device
1331 * @IFF_BONDING: bonding master or slave
1332 * @IFF_ISATAP: ISATAP interface (RFC4214)
1333 * @IFF_WAN_HDLC: WAN HDLC device
1334 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1335 * release skb->dst
1336 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1337 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1338 * @IFF_MACVLAN_PORT: device used as macvlan port
1339 * @IFF_BRIDGE_PORT: device used as bridge port
1340 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1341 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1342 * @IFF_UNICAST_FLT: Supports unicast filtering
1343 * @IFF_TEAM_PORT: device used as team port
1344 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1345 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1346 * change when it's running
1347 * @IFF_MACVLAN: Macvlan device
1348 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1349 * underlying stacked devices
1350 * @IFF_IPVLAN_MASTER: IPvlan master device
1351 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1352 * @IFF_L3MDEV_MASTER: device is an L3 master device
1353 * @IFF_NO_QUEUE: device can run without qdisc attached
1354 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1355 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1356 * @IFF_TEAM: device is a team device
1357 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1358 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1359 * entity (i.e. the master device for bridged veth)
1360 * @IFF_MACSEC: device is a MACsec device
1361 */
1362 enum netdev_priv_flags {
1363 IFF_802_1Q_VLAN = 1<<0,
1364 IFF_EBRIDGE = 1<<1,
1365 IFF_BONDING = 1<<2,
1366 IFF_ISATAP = 1<<3,
1367 IFF_WAN_HDLC = 1<<4,
1368 IFF_XMIT_DST_RELEASE = 1<<5,
1369 IFF_DONT_BRIDGE = 1<<6,
1370 IFF_DISABLE_NETPOLL = 1<<7,
1371 IFF_MACVLAN_PORT = 1<<8,
1372 IFF_BRIDGE_PORT = 1<<9,
1373 IFF_OVS_DATAPATH = 1<<10,
1374 IFF_TX_SKB_SHARING = 1<<11,
1375 IFF_UNICAST_FLT = 1<<12,
1376 IFF_TEAM_PORT = 1<<13,
1377 IFF_SUPP_NOFCS = 1<<14,
1378 IFF_LIVE_ADDR_CHANGE = 1<<15,
1379 IFF_MACVLAN = 1<<16,
1380 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1381 IFF_IPVLAN_MASTER = 1<<18,
1382 IFF_IPVLAN_SLAVE = 1<<19,
1383 IFF_L3MDEV_MASTER = 1<<20,
1384 IFF_NO_QUEUE = 1<<21,
1385 IFF_OPENVSWITCH = 1<<22,
1386 IFF_L3MDEV_SLAVE = 1<<23,
1387 IFF_TEAM = 1<<24,
1388 IFF_RXFH_CONFIGURED = 1<<25,
1389 IFF_PHONY_HEADROOM = 1<<26,
1390 IFF_MACSEC = 1<<27,
1391 };
1392
1393 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1394 #define IFF_EBRIDGE IFF_EBRIDGE
1395 #define IFF_BONDING IFF_BONDING
1396 #define IFF_ISATAP IFF_ISATAP
1397 #define IFF_WAN_HDLC IFF_WAN_HDLC
1398 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1399 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1400 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1401 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1402 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1403 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1404 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1405 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1406 #define IFF_TEAM_PORT IFF_TEAM_PORT
1407 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1408 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1409 #define IFF_MACVLAN IFF_MACVLAN
1410 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1411 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1412 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1413 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1414 #define IFF_NO_QUEUE IFF_NO_QUEUE
1415 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1416 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1417 #define IFF_TEAM IFF_TEAM
1418 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1419 #define IFF_MACSEC IFF_MACSEC
1420
1421 /**
1422 * struct net_device - The DEVICE structure.
1423 *
1424 * Actually, this whole structure is a big mistake. It mixes I/O
1425 * data with strictly "high-level" data, and it has to know about
1426 * almost every data structure used in the INET module.
1427 *
1428 * @name: This is the first field of the "visible" part of this structure
1429 * (i.e. as seen by users in the "Space.c" file). It is the name
1430 * of the interface.
1431 *
1432 * @name_hlist: Device name hash chain, please keep it close to name[]
1433 * @ifalias: SNMP alias
1434 * @mem_end: Shared memory end
1435 * @mem_start: Shared memory start
1436 * @base_addr: Device I/O address
1437 * @irq: Device IRQ number
1438 *
1439 * @carrier_changes: Stats to monitor carrier on<->off transitions
1440 *
1441 * @state: Generic network queuing layer state, see netdev_state_t
1442 * @dev_list: The global list of network devices
1443 * @napi_list: List entry used for polling NAPI devices
1444 * @unreg_list: List entry when we are unregistering the
1445 * device; see the function unregister_netdev
1446 * @close_list: List entry used when we are closing the device
1447 * @ptype_all: Device-specific packet handlers for all protocols
1448 * @ptype_specific: Device-specific, protocol-specific packet handlers
1449 *
1450 * @adj_list: Directly linked devices, like slaves for bonding
1451 * @features: Currently active device features
1452 * @hw_features: User-changeable features
1453 *
1454 * @wanted_features: User-requested features
1455 * @vlan_features: Mask of features inheritable by VLAN devices
1456 *
1457 * @hw_enc_features: Mask of features inherited by encapsulating devices
1458 * This field indicates what encapsulation
1459 * offloads the hardware is capable of doing,
1460 * and drivers will need to set them appropriately.
1461 *
1462 * @mpls_features: Mask of features inheritable by MPLS
1463 *
1464 * @ifindex: interface index
1465 * @group: The group the device belongs to
1466 *
1467 * @stats: Statistics struct, which was left as a legacy, use
1468 * rtnl_link_stats64 instead
1469 *
1470 * @rx_dropped: Dropped packets by core network,
1471 * do not use this in drivers
1472 * @tx_dropped: Dropped packets by core network,
1473 * do not use this in drivers
1474 * @rx_nohandler: nohandler dropped packets by core network on
1475 * inactive devices, do not use this in drivers
1476 *
1477 * @wireless_handlers: List of functions to handle Wireless Extensions,
1478 * instead of ioctl,
1479 * see <net/iw_handler.h> for details.
1480 * @wireless_data: Instance data managed by the core of wireless extensions
1481 *
1482 * @netdev_ops: Includes several pointers to callbacks,
1483 * if one wants to override the ndo_*() functions
1484 * @ethtool_ops: Management operations
1485 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1486 * discovery handling. Necessary for e.g. 6LoWPAN.
1487 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1488 * of Layer 2 headers.
1489 *
1490 * @flags: Interface flags (a la BSD)
1491 * @priv_flags: Like 'flags' but invisible to userspace,
1492 * see if.h for the definitions
1493 * @gflags: Global flags ( kept as legacy )
1494 * @padded: How much padding added by alloc_netdev()
1495 * @operstate: RFC2863 operstate
1496 * @link_mode: Mapping policy to operstate
1497 * @if_port: Selectable AUI, TP, ...
1498 * @dma: DMA channel
1499 * @mtu: Interface MTU value
1500 * @min_mtu: Interface Minimum MTU value
1501 * @max_mtu: Interface Maximum MTU value
1502 * @type: Interface hardware type
1503 * @hard_header_len: Maximum hardware header length.
1504 * @min_header_len: Minimum hardware header length
1505 *
1506 * @needed_headroom: Extra headroom the hardware may need, but not in all
1507 * cases can this be guaranteed
1508 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1509 * cases can this be guaranteed. Some cases also use
1510 * LL_MAX_HEADER instead to allocate the skb
1511 *
1512 * interface address info:
1513 *
1514 * @perm_addr: Permanent hw address
1515 * @addr_assign_type: Hw address assignment type
1516 * @addr_len: Hardware address length
1517 * @neigh_priv_len: Used in neigh_alloc()
1518 * @dev_id: Used to differentiate devices that share
1519 * the same link layer address
1520 * @dev_port: Used to differentiate devices that share
1521 * the same function
1522 * @addr_list_lock: XXX: need comments on this one
1523 * @uc_promisc: Counter that indicates promiscuous mode
1524 * has been enabled due to the need to listen to
1525 * additional unicast addresses in a device that
1526 * does not implement ndo_set_rx_mode()
1527 * @uc: unicast mac addresses
1528 * @mc: multicast mac addresses
1529 * @dev_addrs: list of device hw addresses
1530 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1531 * @promiscuity: Number of times the NIC is told to work in
1532 * promiscuous mode; if it becomes 0 the NIC will
1533 * exit promiscuous mode
1534 * @allmulti: Counter, enables or disables allmulticast mode
1535 *
1536 * @vlan_info: VLAN info
1537 * @dsa_ptr: dsa specific data
1538 * @tipc_ptr: TIPC specific data
1539 * @atalk_ptr: AppleTalk link
1540 * @ip_ptr: IPv4 specific data
1541 * @dn_ptr: DECnet specific data
1542 * @ip6_ptr: IPv6 specific data
1543 * @ax25_ptr: AX.25 specific data
1544 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1545 *
1546 * @dev_addr: Hw address (before bcast,
1547 * because most packets are unicast)
1548 *
1549 * @_rx: Array of RX queues
1550 * @num_rx_queues: Number of RX queues
1551 * allocated at register_netdev() time
1552 * @real_num_rx_queues: Number of RX queues currently active in device
1553 *
1554 * @rx_handler: handler for received packets
1555 * @rx_handler_data: XXX: need comments on this one
1556 * @ingress_queue: XXX: need comments on this one
1557 * @broadcast: hw bcast address
1558 *
1559 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1560 * indexed by RX queue number. Assigned by driver.
1561 * This must only be set if the ndo_rx_flow_steer
1562 * operation is defined
1563 * @index_hlist: Device index hash chain
1564 *
1565 * @_tx: Array of TX queues
1566 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1567 * @real_num_tx_queues: Number of TX queues currently active in device
1568 * @qdisc: Root qdisc from userspace point of view
1569 * @tx_queue_len: Max frames per queue allowed
1570 * @tx_global_lock: XXX: need comments on this one
1571 *
1572 * @xps_maps: XXX: need comments on this one
1573 *
1574 * @watchdog_timeo: Represents the timeout that is used by
1575 * the watchdog (see dev_watchdog())
1576 * @watchdog_timer: List of timers
1577 *
1578 * @pcpu_refcnt: Number of references to this device
1579 * @todo_list: Delayed register/unregister
1580 * @link_watch_list: XXX: need comments on this one
1581 *
1582 * @reg_state: Register/unregister state machine
1583 * @dismantle: Device is going to be freed
1584 * @rtnl_link_state: This enum represents the phases of creating
1585 * a new link
1586 *
1587 * @needs_free_netdev: Should unregister perform free_netdev?
1588 * @priv_destructor: Called from unregister
1589 * @npinfo: XXX: need comments on this one
1590 * @nd_net: Network namespace this network device is inside
1591 *
1592 * @ml_priv: Mid-layer private
1593 * @lstats: Loopback statistics
1594 * @tstats: Tunnel statistics
1595 * @dstats: Dummy statistics
1596 * @vstats: Virtual ethernet statistics
1597 *
1598 * @garp_port: GARP
1599 * @mrp_port: MRP
1600 *
1601 * @dev: Class/net/name entry
1602 * @sysfs_groups: Space for optional device, statistics and wireless
1603 * sysfs groups
1604 *
1605 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1606 * @rtnl_link_ops: Rtnl_link_ops
1607 *
1608 * @gso_max_size: Maximum size of generic segmentation offload
1609 * @gso_max_segs: Maximum number of segments that can be passed to the
1610 * NIC for GSO
1611 *
1612 * @dcbnl_ops: Data Center Bridging netlink ops
1613 * @num_tc: Number of traffic classes in the net device
1614 * @tc_to_txq: XXX: need comments on this one
1615 * @prio_tc_map: XXX: need comments on this one
1616 *
1617 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1618 *
1619 * @priomap: XXX: need comments on this one
1620 * @phydev: Physical device may attach itself
1621 * for hardware timestamping
1622 *
1623 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1624 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1625 *
1626 * @proto_down: protocol port state information can be sent to the
1627 * switch driver and used to set the phys state of the
1628 * switch port.
1629 *
1630 * FIXME: cleanup struct net_device such that network protocol info
1631 * moves out.
1632 */
1633
1634 struct net_device {
1635 char name[IFNAMSIZ];
1636 struct hlist_node name_hlist;
1637 char *ifalias;
1638 /*
1639 * I/O specific fields
1640 * FIXME: Merge these and struct ifmap into one
1641 */
1642 unsigned long mem_end;
1643 unsigned long mem_start;
1644 unsigned long base_addr;
1645 int irq;
1646
1647 atomic_t carrier_changes;
1648
1649 /*
1650 * Some hardware also needs these fields (state,dev_list,
1651 * napi_list,unreg_list,close_list) but they are not
1652 * part of the usual set specified in Space.c.
1653 */
1654
1655 unsigned long state;
1656
1657 struct list_head dev_list;
1658 struct list_head napi_list;
1659 struct list_head unreg_list;
1660 struct list_head close_list;
1661 struct list_head ptype_all;
1662 struct list_head ptype_specific;
1663
1664 struct {
1665 struct list_head upper;
1666 struct list_head lower;
1667 } adj_list;
1668
1669 netdev_features_t features;
1670 netdev_features_t hw_features;
1671 netdev_features_t wanted_features;
1672 netdev_features_t vlan_features;
1673 netdev_features_t hw_enc_features;
1674 netdev_features_t mpls_features;
1675 netdev_features_t gso_partial_features;
1676
1677 int ifindex;
1678 int group;
1679
1680 struct net_device_stats stats;
1681
1682 atomic_long_t rx_dropped;
1683 atomic_long_t tx_dropped;
1684 atomic_long_t rx_nohandler;
1685
1686 #ifdef CONFIG_WIRELESS_EXT
1687 const struct iw_handler_def *wireless_handlers;
1688 struct iw_public_data *wireless_data;
1689 #endif
1690 const struct net_device_ops *netdev_ops;
1691 const struct ethtool_ops *ethtool_ops;
1692 #ifdef CONFIG_NET_SWITCHDEV
1693 const struct switchdev_ops *switchdev_ops;
1694 #endif
1695 #ifdef CONFIG_NET_L3_MASTER_DEV
1696 const struct l3mdev_ops *l3mdev_ops;
1697 #endif
1698 #if IS_ENABLED(CONFIG_IPV6)
1699 const struct ndisc_ops *ndisc_ops;
1700 #endif
1701
1702 #ifdef CONFIG_XFRM
1703 const struct xfrmdev_ops *xfrmdev_ops;
1704 #endif
1705
1706 const struct header_ops *header_ops;
1707
1708 unsigned int flags;
1709 unsigned int priv_flags;
1710
1711 unsigned short gflags;
1712 unsigned short padded;
1713
1714 unsigned char operstate;
1715 unsigned char link_mode;
1716
1717 unsigned char if_port;
1718 unsigned char dma;
1719
1720 unsigned int mtu;
1721 unsigned int min_mtu;
1722 unsigned int max_mtu;
1723 unsigned short type;
1724 unsigned short hard_header_len;
1725 unsigned char min_header_len;
1726
1727 unsigned short needed_headroom;
1728 unsigned short needed_tailroom;
1729
1730 /* Interface address info. */
1731 unsigned char perm_addr[MAX_ADDR_LEN];
1732 unsigned char addr_assign_type;
1733 unsigned char addr_len;
1734 unsigned short neigh_priv_len;
1735 unsigned short dev_id;
1736 unsigned short dev_port;
1737 spinlock_t addr_list_lock;
1738 unsigned char name_assign_type;
1739 bool uc_promisc;
1740 struct netdev_hw_addr_list uc;
1741 struct netdev_hw_addr_list mc;
1742 struct netdev_hw_addr_list dev_addrs;
1743
1744 #ifdef CONFIG_SYSFS
1745 struct kset *queues_kset;
1746 #endif
1747 unsigned int promiscuity;
1748 unsigned int allmulti;
1749
1750
1751 /* Protocol-specific pointers */
1752
1753 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1754 struct vlan_info __rcu *vlan_info;
1755 #endif
1756 #if IS_ENABLED(CONFIG_NET_DSA)
1757 struct dsa_switch_tree *dsa_ptr;
1758 #endif
1759 #if IS_ENABLED(CONFIG_TIPC)
1760 struct tipc_bearer __rcu *tipc_ptr;
1761 #endif
1762 void *atalk_ptr;
1763 struct in_device __rcu *ip_ptr;
1764 struct dn_dev __rcu *dn_ptr;
1765 struct inet6_dev __rcu *ip6_ptr;
1766 void *ax25_ptr;
1767 struct wireless_dev *ieee80211_ptr;
1768 struct wpan_dev *ieee802154_ptr;
1769 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1770 struct mpls_dev __rcu *mpls_ptr;
1771 #endif
1772
1773 /*
1774 * Cache lines mostly used on receive path (including eth_type_trans())
1775 */
1776 /* Interface address info used in eth_type_trans() */
1777 unsigned char *dev_addr;
1778
1779 #ifdef CONFIG_SYSFS
1780 struct netdev_rx_queue *_rx;
1781
1782 unsigned int num_rx_queues;
1783 unsigned int real_num_rx_queues;
1784 #endif
1785
1786 struct bpf_prog __rcu *xdp_prog;
1787 unsigned long gro_flush_timeout;
1788 rx_handler_func_t __rcu *rx_handler;
1789 void __rcu *rx_handler_data;
1790
1791 #ifdef CONFIG_NET_CLS_ACT
1792 struct tcf_proto __rcu *ingress_cl_list;
1793 #endif
1794 struct netdev_queue __rcu *ingress_queue;
1795 #ifdef CONFIG_NETFILTER_INGRESS
1796 struct nf_hook_entry __rcu *nf_hooks_ingress;
1797 #endif
1798
1799 unsigned char broadcast[MAX_ADDR_LEN];
1800 #ifdef CONFIG_RFS_ACCEL
1801 struct cpu_rmap *rx_cpu_rmap;
1802 #endif
1803 struct hlist_node index_hlist;
1804
1805 /*
1806 * Cache lines mostly used on transmit path
1807 */
1808 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1809 unsigned int num_tx_queues;
1810 unsigned int real_num_tx_queues;
1811 struct Qdisc *qdisc;
1812 #ifdef CONFIG_NET_SCHED
1813 DECLARE_HASHTABLE (qdisc_hash, 4);
1814 #endif
1815 unsigned int tx_queue_len;
1816 spinlock_t tx_global_lock;
1817 int watchdog_timeo;
1818
1819 #ifdef CONFIG_XPS
1820 struct xps_dev_maps __rcu *xps_maps;
1821 #endif
1822 #ifdef CONFIG_NET_CLS_ACT
1823 struct tcf_proto __rcu *egress_cl_list;
1824 #endif
1825
1826 /* These may be needed for future network-power-down code. */
1827 struct timer_list watchdog_timer;
1828
1829 int __percpu *pcpu_refcnt;
1830 struct list_head todo_list;
1831
1832 struct list_head link_watch_list;
1833
1834 enum { NETREG_UNINITIALIZED=0,
1835 NETREG_REGISTERED, /* completed register_netdevice */
1836 NETREG_UNREGISTERING, /* called unregister_netdevice */
1837 NETREG_UNREGISTERED, /* completed unregister todo */
1838 NETREG_RELEASED, /* called free_netdev */
1839 NETREG_DUMMY, /* dummy device for NAPI poll */
1840 } reg_state:8;
1841
1842 bool dismantle;
1843
1844 enum {
1845 RTNL_LINK_INITIALIZED,
1846 RTNL_LINK_INITIALIZING,
1847 } rtnl_link_state:16;
1848
1849 bool needs_free_netdev;
1850 void (*priv_destructor)(struct net_device *dev);
1851
1852 #ifdef CONFIG_NETPOLL
1853 struct netpoll_info __rcu *npinfo;
1854 #endif
1855
1856 possible_net_t nd_net;
1857
1858 /* mid-layer private */
1859 union {
1860 void *ml_priv;
1861 struct pcpu_lstats __percpu *lstats;
1862 struct pcpu_sw_netstats __percpu *tstats;
1863 struct pcpu_dstats __percpu *dstats;
1864 struct pcpu_vstats __percpu *vstats;
1865 };
1866
1867 #if IS_ENABLED(CONFIG_GARP)
1868 struct garp_port __rcu *garp_port;
1869 #endif
1870 #if IS_ENABLED(CONFIG_MRP)
1871 struct mrp_port __rcu *mrp_port;
1872 #endif
1873
1874 struct device dev;
1875 const struct attribute_group *sysfs_groups[4];
1876 const struct attribute_group *sysfs_rx_queue_group;
1877
1878 const struct rtnl_link_ops *rtnl_link_ops;
1879
1880 /* for setting kernel sock attribute on TCP connection setup */
1881 #define GSO_MAX_SIZE 65536
1882 unsigned int gso_max_size;
1883 #define GSO_MAX_SEGS 65535
1884 u16 gso_max_segs;
1885
1886 #ifdef CONFIG_DCB
1887 const struct dcbnl_rtnl_ops *dcbnl_ops;
1888 #endif
1889 u8 num_tc;
1890 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1891 u8 prio_tc_map[TC_BITMASK + 1];
1892
1893 #if IS_ENABLED(CONFIG_FCOE)
1894 unsigned int fcoe_ddp_xid;
1895 #endif
1896 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1897 struct netprio_map __rcu *priomap;
1898 #endif
1899 struct phy_device *phydev;
1900 struct lock_class_key *qdisc_tx_busylock;
1901 struct lock_class_key *qdisc_running_key;
1902 bool proto_down;
1903 };
1904 #define to_net_dev(d) container_of(d, struct net_device, dev)
1905
1906 static inline bool netif_elide_gro(const struct net_device *dev)
1907 {
1908 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1909 return true;
1910 return false;
1911 }
1912
1913 #define NETDEV_ALIGN 32
1914
1915 static inline
1916 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1917 {
1918 return dev->prio_tc_map[prio & TC_BITMASK];
1919 }
1920
1921 static inline
1922 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1923 {
1924 if (tc >= dev->num_tc)
1925 return -EINVAL;
1926
1927 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1928 return 0;
1929 }
1930
1931 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1932 void netdev_reset_tc(struct net_device *dev);
1933 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1934 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1935
1936 static inline
1937 int netdev_get_num_tc(struct net_device *dev)
1938 {
1939 return dev->num_tc;
1940 }
1941
1942 static inline
1943 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1944 unsigned int index)
1945 {
1946 return &dev->_tx[index];
1947 }
1948
1949 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1950 const struct sk_buff *skb)
1951 {
1952 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1953 }
1954
1955 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1956 void (*f)(struct net_device *,
1957 struct netdev_queue *,
1958 void *),
1959 void *arg)
1960 {
1961 unsigned int i;
1962
1963 for (i = 0; i < dev->num_tx_queues; i++)
1964 f(dev, &dev->_tx[i], arg);
1965 }
1966
1967 #define netdev_lockdep_set_classes(dev) \
1968 { \
1969 static struct lock_class_key qdisc_tx_busylock_key; \
1970 static struct lock_class_key qdisc_running_key; \
1971 static struct lock_class_key qdisc_xmit_lock_key; \
1972 static struct lock_class_key dev_addr_list_lock_key; \
1973 unsigned int i; \
1974 \
1975 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1976 (dev)->qdisc_running_key = &qdisc_running_key; \
1977 lockdep_set_class(&(dev)->addr_list_lock, \
1978 &dev_addr_list_lock_key); \
1979 for (i = 0; i < (dev)->num_tx_queues; i++) \
1980 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1981 &qdisc_xmit_lock_key); \
1982 }
1983
1984 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1985 struct sk_buff *skb,
1986 void *accel_priv);
1987
1988 /* returns the headroom that the master device needs to take in account
1989 * when forwarding to this dev
1990 */
1991 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1992 {
1993 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1994 }
1995
1996 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1997 {
1998 if (dev->netdev_ops->ndo_set_rx_headroom)
1999 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2000 }
2001
2002 /* set the device rx headroom to the dev's default */
2003 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2004 {
2005 netdev_set_rx_headroom(dev, -1);
2006 }
2007
2008 /*
2009 * Net namespace inlines
2010 */
2011 static inline
2012 struct net *dev_net(const struct net_device *dev)
2013 {
2014 return read_pnet(&dev->nd_net);
2015 }
2016
2017 static inline
2018 void dev_net_set(struct net_device *dev, struct net *net)
2019 {
2020 write_pnet(&dev->nd_net, net);
2021 }
2022
2023 /**
2024 * netdev_priv - access network device private data
2025 * @dev: network device
2026 *
2027 * Get network device private data
2028 */
2029 static inline void *netdev_priv(const struct net_device *dev)
2030 {
2031 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2032 }
2033
2034 /* Set the sysfs physical device reference for the network logical device
2035 * if set prior to registration will cause a symlink during initialization.
2036 */
2037 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2038
2039 /* Set the sysfs device type for the network logical device to allow
2040 * fine-grained identification of different network device types. For
2041 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2042 */
2043 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2044
2045 /* Default NAPI poll() weight
2046 * Device drivers are strongly advised to not use bigger value
2047 */
2048 #define NAPI_POLL_WEIGHT 64
2049
2050 /**
2051 * netif_napi_add - initialize a NAPI context
2052 * @dev: network device
2053 * @napi: NAPI context
2054 * @poll: polling function
2055 * @weight: default weight
2056 *
2057 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2058 * *any* of the other NAPI-related functions.
2059 */
2060 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2061 int (*poll)(struct napi_struct *, int), int weight);
2062
2063 /**
2064 * netif_tx_napi_add - initialize a NAPI context
2065 * @dev: network device
2066 * @napi: NAPI context
2067 * @poll: polling function
2068 * @weight: default weight
2069 *
2070 * This variant of netif_napi_add() should be used from drivers using NAPI
2071 * to exclusively poll a TX queue.
2072 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2073 */
2074 static inline void netif_tx_napi_add(struct net_device *dev,
2075 struct napi_struct *napi,
2076 int (*poll)(struct napi_struct *, int),
2077 int weight)
2078 {
2079 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2080 netif_napi_add(dev, napi, poll, weight);
2081 }
2082
2083 /**
2084 * netif_napi_del - remove a NAPI context
2085 * @napi: NAPI context
2086 *
2087 * netif_napi_del() removes a NAPI context from the network device NAPI list
2088 */
2089 void netif_napi_del(struct napi_struct *napi);
2090
2091 struct napi_gro_cb {
2092 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2093 void *frag0;
2094
2095 /* Length of frag0. */
2096 unsigned int frag0_len;
2097
2098 /* This indicates where we are processing relative to skb->data. */
2099 int data_offset;
2100
2101 /* This is non-zero if the packet cannot be merged with the new skb. */
2102 u16 flush;
2103
2104 /* Save the IP ID here and check when we get to the transport layer */
2105 u16 flush_id;
2106
2107 /* Number of segments aggregated. */
2108 u16 count;
2109
2110 /* Start offset for remote checksum offload */
2111 u16 gro_remcsum_start;
2112
2113 /* jiffies when first packet was created/queued */
2114 unsigned long age;
2115
2116 /* Used in ipv6_gro_receive() and foo-over-udp */
2117 u16 proto;
2118
2119 /* This is non-zero if the packet may be of the same flow. */
2120 u8 same_flow:1;
2121
2122 /* Used in tunnel GRO receive */
2123 u8 encap_mark:1;
2124
2125 /* GRO checksum is valid */
2126 u8 csum_valid:1;
2127
2128 /* Number of checksums via CHECKSUM_UNNECESSARY */
2129 u8 csum_cnt:3;
2130
2131 /* Free the skb? */
2132 u8 free:2;
2133 #define NAPI_GRO_FREE 1
2134 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2135
2136 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2137 u8 is_ipv6:1;
2138
2139 /* Used in GRE, set in fou/gue_gro_receive */
2140 u8 is_fou:1;
2141
2142 /* Used to determine if flush_id can be ignored */
2143 u8 is_atomic:1;
2144
2145 /* Number of gro_receive callbacks this packet already went through */
2146 u8 recursion_counter:4;
2147
2148 /* 1 bit hole */
2149
2150 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2151 __wsum csum;
2152
2153 /* used in skb_gro_receive() slow path */
2154 struct sk_buff *last;
2155 };
2156
2157 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2158
2159 #define GRO_RECURSION_LIMIT 15
2160 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2161 {
2162 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2163 }
2164
2165 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2166 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2167 struct sk_buff **head,
2168 struct sk_buff *skb)
2169 {
2170 if (unlikely(gro_recursion_inc_test(skb))) {
2171 NAPI_GRO_CB(skb)->flush |= 1;
2172 return NULL;
2173 }
2174
2175 return cb(head, skb);
2176 }
2177
2178 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2179 struct sk_buff *);
2180 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2181 struct sock *sk,
2182 struct sk_buff **head,
2183 struct sk_buff *skb)
2184 {
2185 if (unlikely(gro_recursion_inc_test(skb))) {
2186 NAPI_GRO_CB(skb)->flush |= 1;
2187 return NULL;
2188 }
2189
2190 return cb(sk, head, skb);
2191 }
2192
2193 struct packet_type {
2194 __be16 type; /* This is really htons(ether_type). */
2195 struct net_device *dev; /* NULL is wildcarded here */
2196 int (*func) (struct sk_buff *,
2197 struct net_device *,
2198 struct packet_type *,
2199 struct net_device *);
2200 bool (*id_match)(struct packet_type *ptype,
2201 struct sock *sk);
2202 void *af_packet_priv;
2203 struct list_head list;
2204 };
2205
2206 struct offload_callbacks {
2207 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2208 netdev_features_t features);
2209 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2210 struct sk_buff *skb);
2211 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2212 };
2213
2214 struct packet_offload {
2215 __be16 type; /* This is really htons(ether_type). */
2216 u16 priority;
2217 struct offload_callbacks callbacks;
2218 struct list_head list;
2219 };
2220
2221 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2222 struct pcpu_sw_netstats {
2223 u64 rx_packets;
2224 u64 rx_bytes;
2225 u64 tx_packets;
2226 u64 tx_bytes;
2227 struct u64_stats_sync syncp;
2228 };
2229
2230 #define __netdev_alloc_pcpu_stats(type, gfp) \
2231 ({ \
2232 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2233 if (pcpu_stats) { \
2234 int __cpu; \
2235 for_each_possible_cpu(__cpu) { \
2236 typeof(type) *stat; \
2237 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2238 u64_stats_init(&stat->syncp); \
2239 } \
2240 } \
2241 pcpu_stats; \
2242 })
2243
2244 #define netdev_alloc_pcpu_stats(type) \
2245 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2246
2247 enum netdev_lag_tx_type {
2248 NETDEV_LAG_TX_TYPE_UNKNOWN,
2249 NETDEV_LAG_TX_TYPE_RANDOM,
2250 NETDEV_LAG_TX_TYPE_BROADCAST,
2251 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2252 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2253 NETDEV_LAG_TX_TYPE_HASH,
2254 };
2255
2256 struct netdev_lag_upper_info {
2257 enum netdev_lag_tx_type tx_type;
2258 };
2259
2260 struct netdev_lag_lower_state_info {
2261 u8 link_up : 1,
2262 tx_enabled : 1;
2263 };
2264
2265 #include <linux/notifier.h>
2266
2267 /* netdevice notifier chain. Please remember to update the rtnetlink
2268 * notification exclusion list in rtnetlink_event() when adding new
2269 * types.
2270 */
2271 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2272 #define NETDEV_DOWN 0x0002
2273 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2274 detected a hardware crash and restarted
2275 - we can use this eg to kick tcp sessions
2276 once done */
2277 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2278 #define NETDEV_REGISTER 0x0005
2279 #define NETDEV_UNREGISTER 0x0006
2280 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2281 #define NETDEV_CHANGEADDR 0x0008
2282 #define NETDEV_GOING_DOWN 0x0009
2283 #define NETDEV_CHANGENAME 0x000A
2284 #define NETDEV_FEAT_CHANGE 0x000B
2285 #define NETDEV_BONDING_FAILOVER 0x000C
2286 #define NETDEV_PRE_UP 0x000D
2287 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2288 #define NETDEV_POST_TYPE_CHANGE 0x000F
2289 #define NETDEV_POST_INIT 0x0010
2290 #define NETDEV_UNREGISTER_FINAL 0x0011
2291 #define NETDEV_RELEASE 0x0012
2292 #define NETDEV_NOTIFY_PEERS 0x0013
2293 #define NETDEV_JOIN 0x0014
2294 #define NETDEV_CHANGEUPPER 0x0015
2295 #define NETDEV_RESEND_IGMP 0x0016
2296 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2297 #define NETDEV_CHANGEINFODATA 0x0018
2298 #define NETDEV_BONDING_INFO 0x0019
2299 #define NETDEV_PRECHANGEUPPER 0x001A
2300 #define NETDEV_CHANGELOWERSTATE 0x001B
2301 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2302 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2303 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2304
2305 int register_netdevice_notifier(struct notifier_block *nb);
2306 int unregister_netdevice_notifier(struct notifier_block *nb);
2307
2308 struct netdev_notifier_info {
2309 struct net_device *dev;
2310 };
2311
2312 struct netdev_notifier_change_info {
2313 struct netdev_notifier_info info; /* must be first */
2314 unsigned int flags_changed;
2315 };
2316
2317 struct netdev_notifier_changeupper_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 struct net_device *upper_dev; /* new upper dev */
2320 bool master; /* is upper dev master */
2321 bool linking; /* is the notification for link or unlink */
2322 void *upper_info; /* upper dev info */
2323 };
2324
2325 struct netdev_notifier_changelowerstate_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 void *lower_state_info; /* is lower dev state */
2328 };
2329
2330 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2331 struct net_device *dev)
2332 {
2333 info->dev = dev;
2334 }
2335
2336 static inline struct net_device *
2337 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2338 {
2339 return info->dev;
2340 }
2341
2342 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2343
2344
2345 extern rwlock_t dev_base_lock; /* Device list lock */
2346
2347 #define for_each_netdev(net, d) \
2348 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2349 #define for_each_netdev_reverse(net, d) \
2350 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2351 #define for_each_netdev_rcu(net, d) \
2352 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2353 #define for_each_netdev_safe(net, d, n) \
2354 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2355 #define for_each_netdev_continue(net, d) \
2356 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2357 #define for_each_netdev_continue_rcu(net, d) \
2358 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2359 #define for_each_netdev_in_bond_rcu(bond, slave) \
2360 for_each_netdev_rcu(&init_net, slave) \
2361 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2362 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2363
2364 static inline struct net_device *next_net_device(struct net_device *dev)
2365 {
2366 struct list_head *lh;
2367 struct net *net;
2368
2369 net = dev_net(dev);
2370 lh = dev->dev_list.next;
2371 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2372 }
2373
2374 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2375 {
2376 struct list_head *lh;
2377 struct net *net;
2378
2379 net = dev_net(dev);
2380 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382 }
2383
2384 static inline struct net_device *first_net_device(struct net *net)
2385 {
2386 return list_empty(&net->dev_base_head) ? NULL :
2387 net_device_entry(net->dev_base_head.next);
2388 }
2389
2390 static inline struct net_device *first_net_device_rcu(struct net *net)
2391 {
2392 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2393
2394 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2395 }
2396
2397 int netdev_boot_setup_check(struct net_device *dev);
2398 unsigned long netdev_boot_base(const char *prefix, int unit);
2399 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2400 const char *hwaddr);
2401 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403 void dev_add_pack(struct packet_type *pt);
2404 void dev_remove_pack(struct packet_type *pt);
2405 void __dev_remove_pack(struct packet_type *pt);
2406 void dev_add_offload(struct packet_offload *po);
2407 void dev_remove_offload(struct packet_offload *po);
2408
2409 int dev_get_iflink(const struct net_device *dev);
2410 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2411 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2412 unsigned short mask);
2413 struct net_device *dev_get_by_name(struct net *net, const char *name);
2414 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2415 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2416 int dev_alloc_name(struct net_device *dev, const char *name);
2417 int dev_open(struct net_device *dev);
2418 void dev_close(struct net_device *dev);
2419 void dev_close_many(struct list_head *head, bool unlink);
2420 void dev_disable_lro(struct net_device *dev);
2421 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2422 int dev_queue_xmit(struct sk_buff *skb);
2423 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2424 int register_netdevice(struct net_device *dev);
2425 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2426 void unregister_netdevice_many(struct list_head *head);
2427 static inline void unregister_netdevice(struct net_device *dev)
2428 {
2429 unregister_netdevice_queue(dev, NULL);
2430 }
2431
2432 int netdev_refcnt_read(const struct net_device *dev);
2433 void free_netdev(struct net_device *dev);
2434 void netdev_freemem(struct net_device *dev);
2435 void synchronize_net(void);
2436 int init_dummy_netdev(struct net_device *dev);
2437
2438 DECLARE_PER_CPU(int, xmit_recursion);
2439 #define XMIT_RECURSION_LIMIT 10
2440
2441 static inline int dev_recursion_level(void)
2442 {
2443 return this_cpu_read(xmit_recursion);
2444 }
2445
2446 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2447 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2448 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2449 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2450 int netdev_get_name(struct net *net, char *name, int ifindex);
2451 int dev_restart(struct net_device *dev);
2452 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2453
2454 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2455 {
2456 return NAPI_GRO_CB(skb)->data_offset;
2457 }
2458
2459 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2460 {
2461 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2462 }
2463
2464 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2465 {
2466 NAPI_GRO_CB(skb)->data_offset += len;
2467 }
2468
2469 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2470 unsigned int offset)
2471 {
2472 return NAPI_GRO_CB(skb)->frag0 + offset;
2473 }
2474
2475 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2476 {
2477 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2478 }
2479
2480 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2481 {
2482 NAPI_GRO_CB(skb)->frag0 = NULL;
2483 NAPI_GRO_CB(skb)->frag0_len = 0;
2484 }
2485
2486 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2487 unsigned int offset)
2488 {
2489 if (!pskb_may_pull(skb, hlen))
2490 return NULL;
2491
2492 skb_gro_frag0_invalidate(skb);
2493 return skb->data + offset;
2494 }
2495
2496 static inline void *skb_gro_network_header(struct sk_buff *skb)
2497 {
2498 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2499 skb_network_offset(skb);
2500 }
2501
2502 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2503 const void *start, unsigned int len)
2504 {
2505 if (NAPI_GRO_CB(skb)->csum_valid)
2506 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2507 csum_partial(start, len, 0));
2508 }
2509
2510 /* GRO checksum functions. These are logical equivalents of the normal
2511 * checksum functions (in skbuff.h) except that they operate on the GRO
2512 * offsets and fields in sk_buff.
2513 */
2514
2515 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2516
2517 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2518 {
2519 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2520 }
2521
2522 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2523 bool zero_okay,
2524 __sum16 check)
2525 {
2526 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2527 skb_checksum_start_offset(skb) <
2528 skb_gro_offset(skb)) &&
2529 !skb_at_gro_remcsum_start(skb) &&
2530 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2531 (!zero_okay || check));
2532 }
2533
2534 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2535 __wsum psum)
2536 {
2537 if (NAPI_GRO_CB(skb)->csum_valid &&
2538 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2539 return 0;
2540
2541 NAPI_GRO_CB(skb)->csum = psum;
2542
2543 return __skb_gro_checksum_complete(skb);
2544 }
2545
2546 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2547 {
2548 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2549 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2550 NAPI_GRO_CB(skb)->csum_cnt--;
2551 } else {
2552 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2553 * verified a new top level checksum or an encapsulated one
2554 * during GRO. This saves work if we fallback to normal path.
2555 */
2556 __skb_incr_checksum_unnecessary(skb);
2557 }
2558 }
2559
2560 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2561 compute_pseudo) \
2562 ({ \
2563 __sum16 __ret = 0; \
2564 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2565 __ret = __skb_gro_checksum_validate_complete(skb, \
2566 compute_pseudo(skb, proto)); \
2567 if (!__ret) \
2568 skb_gro_incr_csum_unnecessary(skb); \
2569 __ret; \
2570 })
2571
2572 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2573 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2574
2575 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2576 compute_pseudo) \
2577 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2578
2579 #define skb_gro_checksum_simple_validate(skb) \
2580 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2581
2582 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2583 {
2584 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2585 !NAPI_GRO_CB(skb)->csum_valid);
2586 }
2587
2588 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2589 __sum16 check, __wsum pseudo)
2590 {
2591 NAPI_GRO_CB(skb)->csum = ~pseudo;
2592 NAPI_GRO_CB(skb)->csum_valid = 1;
2593 }
2594
2595 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2596 do { \
2597 if (__skb_gro_checksum_convert_check(skb)) \
2598 __skb_gro_checksum_convert(skb, check, \
2599 compute_pseudo(skb, proto)); \
2600 } while (0)
2601
2602 struct gro_remcsum {
2603 int offset;
2604 __wsum delta;
2605 };
2606
2607 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2608 {
2609 grc->offset = 0;
2610 grc->delta = 0;
2611 }
2612
2613 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2614 unsigned int off, size_t hdrlen,
2615 int start, int offset,
2616 struct gro_remcsum *grc,
2617 bool nopartial)
2618 {
2619 __wsum delta;
2620 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2621
2622 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2623
2624 if (!nopartial) {
2625 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2626 return ptr;
2627 }
2628
2629 ptr = skb_gro_header_fast(skb, off);
2630 if (skb_gro_header_hard(skb, off + plen)) {
2631 ptr = skb_gro_header_slow(skb, off + plen, off);
2632 if (!ptr)
2633 return NULL;
2634 }
2635
2636 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2637 start, offset);
2638
2639 /* Adjust skb->csum since we changed the packet */
2640 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2641
2642 grc->offset = off + hdrlen + offset;
2643 grc->delta = delta;
2644
2645 return ptr;
2646 }
2647
2648 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2649 struct gro_remcsum *grc)
2650 {
2651 void *ptr;
2652 size_t plen = grc->offset + sizeof(u16);
2653
2654 if (!grc->delta)
2655 return;
2656
2657 ptr = skb_gro_header_fast(skb, grc->offset);
2658 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2659 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2660 if (!ptr)
2661 return;
2662 }
2663
2664 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2665 }
2666
2667 #ifdef CONFIG_XFRM_OFFLOAD
2668 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2669 {
2670 if (PTR_ERR(pp) != -EINPROGRESS)
2671 NAPI_GRO_CB(skb)->flush |= flush;
2672 }
2673 #else
2674 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2675 {
2676 NAPI_GRO_CB(skb)->flush |= flush;
2677 }
2678 #endif
2679
2680 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2681 unsigned short type,
2682 const void *daddr, const void *saddr,
2683 unsigned int len)
2684 {
2685 if (!dev->header_ops || !dev->header_ops->create)
2686 return 0;
2687
2688 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2689 }
2690
2691 static inline int dev_parse_header(const struct sk_buff *skb,
2692 unsigned char *haddr)
2693 {
2694 const struct net_device *dev = skb->dev;
2695
2696 if (!dev->header_ops || !dev->header_ops->parse)
2697 return 0;
2698 return dev->header_ops->parse(skb, haddr);
2699 }
2700
2701 /* ll_header must have at least hard_header_len allocated */
2702 static inline bool dev_validate_header(const struct net_device *dev,
2703 char *ll_header, int len)
2704 {
2705 if (likely(len >= dev->hard_header_len))
2706 return true;
2707 if (len < dev->min_header_len)
2708 return false;
2709
2710 if (capable(CAP_SYS_RAWIO)) {
2711 memset(ll_header + len, 0, dev->hard_header_len - len);
2712 return true;
2713 }
2714
2715 if (dev->header_ops && dev->header_ops->validate)
2716 return dev->header_ops->validate(ll_header, len);
2717
2718 return false;
2719 }
2720
2721 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2722 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2723 static inline int unregister_gifconf(unsigned int family)
2724 {
2725 return register_gifconf(family, NULL);
2726 }
2727
2728 #ifdef CONFIG_NET_FLOW_LIMIT
2729 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2730 struct sd_flow_limit {
2731 u64 count;
2732 unsigned int num_buckets;
2733 unsigned int history_head;
2734 u16 history[FLOW_LIMIT_HISTORY];
2735 u8 buckets[];
2736 };
2737
2738 extern int netdev_flow_limit_table_len;
2739 #endif /* CONFIG_NET_FLOW_LIMIT */
2740
2741 /*
2742 * Incoming packets are placed on per-CPU queues
2743 */
2744 struct softnet_data {
2745 struct list_head poll_list;
2746 struct sk_buff_head process_queue;
2747
2748 /* stats */
2749 unsigned int processed;
2750 unsigned int time_squeeze;
2751 unsigned int received_rps;
2752 #ifdef CONFIG_RPS
2753 struct softnet_data *rps_ipi_list;
2754 #endif
2755 #ifdef CONFIG_NET_FLOW_LIMIT
2756 struct sd_flow_limit __rcu *flow_limit;
2757 #endif
2758 struct Qdisc *output_queue;
2759 struct Qdisc **output_queue_tailp;
2760 struct sk_buff *completion_queue;
2761
2762 #ifdef CONFIG_RPS
2763 /* input_queue_head should be written by cpu owning this struct,
2764 * and only read by other cpus. Worth using a cache line.
2765 */
2766 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2767
2768 /* Elements below can be accessed between CPUs for RPS/RFS */
2769 struct call_single_data csd ____cacheline_aligned_in_smp;
2770 struct softnet_data *rps_ipi_next;
2771 unsigned int cpu;
2772 unsigned int input_queue_tail;
2773 #endif
2774 unsigned int dropped;
2775 struct sk_buff_head input_pkt_queue;
2776 struct napi_struct backlog;
2777
2778 };
2779
2780 static inline void input_queue_head_incr(struct softnet_data *sd)
2781 {
2782 #ifdef CONFIG_RPS
2783 sd->input_queue_head++;
2784 #endif
2785 }
2786
2787 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2788 unsigned int *qtail)
2789 {
2790 #ifdef CONFIG_RPS
2791 *qtail = ++sd->input_queue_tail;
2792 #endif
2793 }
2794
2795 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2796
2797 void __netif_schedule(struct Qdisc *q);
2798 void netif_schedule_queue(struct netdev_queue *txq);
2799
2800 static inline void netif_tx_schedule_all(struct net_device *dev)
2801 {
2802 unsigned int i;
2803
2804 for (i = 0; i < dev->num_tx_queues; i++)
2805 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2806 }
2807
2808 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2809 {
2810 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2811 }
2812
2813 /**
2814 * netif_start_queue - allow transmit
2815 * @dev: network device
2816 *
2817 * Allow upper layers to call the device hard_start_xmit routine.
2818 */
2819 static inline void netif_start_queue(struct net_device *dev)
2820 {
2821 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2822 }
2823
2824 static inline void netif_tx_start_all_queues(struct net_device *dev)
2825 {
2826 unsigned int i;
2827
2828 for (i = 0; i < dev->num_tx_queues; i++) {
2829 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2830 netif_tx_start_queue(txq);
2831 }
2832 }
2833
2834 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2835
2836 /**
2837 * netif_wake_queue - restart transmit
2838 * @dev: network device
2839 *
2840 * Allow upper layers to call the device hard_start_xmit routine.
2841 * Used for flow control when transmit resources are available.
2842 */
2843 static inline void netif_wake_queue(struct net_device *dev)
2844 {
2845 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2846 }
2847
2848 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2849 {
2850 unsigned int i;
2851
2852 for (i = 0; i < dev->num_tx_queues; i++) {
2853 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2854 netif_tx_wake_queue(txq);
2855 }
2856 }
2857
2858 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2859 {
2860 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2861 }
2862
2863 /**
2864 * netif_stop_queue - stop transmitted packets
2865 * @dev: network device
2866 *
2867 * Stop upper layers calling the device hard_start_xmit routine.
2868 * Used for flow control when transmit resources are unavailable.
2869 */
2870 static inline void netif_stop_queue(struct net_device *dev)
2871 {
2872 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2873 }
2874
2875 void netif_tx_stop_all_queues(struct net_device *dev);
2876
2877 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2878 {
2879 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2880 }
2881
2882 /**
2883 * netif_queue_stopped - test if transmit queue is flowblocked
2884 * @dev: network device
2885 *
2886 * Test if transmit queue on device is currently unable to send.
2887 */
2888 static inline bool netif_queue_stopped(const struct net_device *dev)
2889 {
2890 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2891 }
2892
2893 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2894 {
2895 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2896 }
2897
2898 static inline bool
2899 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2900 {
2901 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2902 }
2903
2904 static inline bool
2905 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2906 {
2907 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2908 }
2909
2910 /**
2911 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2912 * @dev_queue: pointer to transmit queue
2913 *
2914 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2915 * to give appropriate hint to the CPU.
2916 */
2917 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2918 {
2919 #ifdef CONFIG_BQL
2920 prefetchw(&dev_queue->dql.num_queued);
2921 #endif
2922 }
2923
2924 /**
2925 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2926 * @dev_queue: pointer to transmit queue
2927 *
2928 * BQL enabled drivers might use this helper in their TX completion path,
2929 * to give appropriate hint to the CPU.
2930 */
2931 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2932 {
2933 #ifdef CONFIG_BQL
2934 prefetchw(&dev_queue->dql.limit);
2935 #endif
2936 }
2937
2938 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2939 unsigned int bytes)
2940 {
2941 #ifdef CONFIG_BQL
2942 dql_queued(&dev_queue->dql, bytes);
2943
2944 if (likely(dql_avail(&dev_queue->dql) >= 0))
2945 return;
2946
2947 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2948
2949 /*
2950 * The XOFF flag must be set before checking the dql_avail below,
2951 * because in netdev_tx_completed_queue we update the dql_completed
2952 * before checking the XOFF flag.
2953 */
2954 smp_mb();
2955
2956 /* check again in case another CPU has just made room avail */
2957 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2958 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2959 #endif
2960 }
2961
2962 /**
2963 * netdev_sent_queue - report the number of bytes queued to hardware
2964 * @dev: network device
2965 * @bytes: number of bytes queued to the hardware device queue
2966 *
2967 * Report the number of bytes queued for sending/completion to the network
2968 * device hardware queue. @bytes should be a good approximation and should
2969 * exactly match netdev_completed_queue() @bytes
2970 */
2971 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2972 {
2973 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2974 }
2975
2976 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2977 unsigned int pkts, unsigned int bytes)
2978 {
2979 #ifdef CONFIG_BQL
2980 if (unlikely(!bytes))
2981 return;
2982
2983 dql_completed(&dev_queue->dql, bytes);
2984
2985 /*
2986 * Without the memory barrier there is a small possiblity that
2987 * netdev_tx_sent_queue will miss the update and cause the queue to
2988 * be stopped forever
2989 */
2990 smp_mb();
2991
2992 if (dql_avail(&dev_queue->dql) < 0)
2993 return;
2994
2995 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2996 netif_schedule_queue(dev_queue);
2997 #endif
2998 }
2999
3000 /**
3001 * netdev_completed_queue - report bytes and packets completed by device
3002 * @dev: network device
3003 * @pkts: actual number of packets sent over the medium
3004 * @bytes: actual number of bytes sent over the medium
3005 *
3006 * Report the number of bytes and packets transmitted by the network device
3007 * hardware queue over the physical medium, @bytes must exactly match the
3008 * @bytes amount passed to netdev_sent_queue()
3009 */
3010 static inline void netdev_completed_queue(struct net_device *dev,
3011 unsigned int pkts, unsigned int bytes)
3012 {
3013 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3014 }
3015
3016 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3017 {
3018 #ifdef CONFIG_BQL
3019 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3020 dql_reset(&q->dql);
3021 #endif
3022 }
3023
3024 /**
3025 * netdev_reset_queue - reset the packets and bytes count of a network device
3026 * @dev_queue: network device
3027 *
3028 * Reset the bytes and packet count of a network device and clear the
3029 * software flow control OFF bit for this network device
3030 */
3031 static inline void netdev_reset_queue(struct net_device *dev_queue)
3032 {
3033 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3034 }
3035
3036 /**
3037 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3038 * @dev: network device
3039 * @queue_index: given tx queue index
3040 *
3041 * Returns 0 if given tx queue index >= number of device tx queues,
3042 * otherwise returns the originally passed tx queue index.
3043 */
3044 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3045 {
3046 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3047 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3048 dev->name, queue_index,
3049 dev->real_num_tx_queues);
3050 return 0;
3051 }
3052
3053 return queue_index;
3054 }
3055
3056 /**
3057 * netif_running - test if up
3058 * @dev: network device
3059 *
3060 * Test if the device has been brought up.
3061 */
3062 static inline bool netif_running(const struct net_device *dev)
3063 {
3064 return test_bit(__LINK_STATE_START, &dev->state);
3065 }
3066
3067 /*
3068 * Routines to manage the subqueues on a device. We only need start,
3069 * stop, and a check if it's stopped. All other device management is
3070 * done at the overall netdevice level.
3071 * Also test the device if we're multiqueue.
3072 */
3073
3074 /**
3075 * netif_start_subqueue - allow sending packets on subqueue
3076 * @dev: network device
3077 * @queue_index: sub queue index
3078 *
3079 * Start individual transmit queue of a device with multiple transmit queues.
3080 */
3081 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3082 {
3083 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3084
3085 netif_tx_start_queue(txq);
3086 }
3087
3088 /**
3089 * netif_stop_subqueue - stop sending packets on subqueue
3090 * @dev: network device
3091 * @queue_index: sub queue index
3092 *
3093 * Stop individual transmit queue of a device with multiple transmit queues.
3094 */
3095 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3096 {
3097 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3098 netif_tx_stop_queue(txq);
3099 }
3100
3101 /**
3102 * netif_subqueue_stopped - test status of subqueue
3103 * @dev: network device
3104 * @queue_index: sub queue index
3105 *
3106 * Check individual transmit queue of a device with multiple transmit queues.
3107 */
3108 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3109 u16 queue_index)
3110 {
3111 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3112
3113 return netif_tx_queue_stopped(txq);
3114 }
3115
3116 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3117 struct sk_buff *skb)
3118 {
3119 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3120 }
3121
3122 /**
3123 * netif_wake_subqueue - allow sending packets on subqueue
3124 * @dev: network device
3125 * @queue_index: sub queue index
3126 *
3127 * Resume individual transmit queue of a device with multiple transmit queues.
3128 */
3129 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3130 {
3131 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3132
3133 netif_tx_wake_queue(txq);
3134 }
3135
3136 #ifdef CONFIG_XPS
3137 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3138 u16 index);
3139 #else
3140 static inline int netif_set_xps_queue(struct net_device *dev,
3141 const struct cpumask *mask,
3142 u16 index)
3143 {
3144 return 0;
3145 }
3146 #endif
3147
3148 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3149 unsigned int num_tx_queues);
3150
3151 /*
3152 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3153 * as a distribution range limit for the returned value.
3154 */
3155 static inline u16 skb_tx_hash(const struct net_device *dev,
3156 struct sk_buff *skb)
3157 {
3158 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3159 }
3160
3161 /**
3162 * netif_is_multiqueue - test if device has multiple transmit queues
3163 * @dev: network device
3164 *
3165 * Check if device has multiple transmit queues
3166 */
3167 static inline bool netif_is_multiqueue(const struct net_device *dev)
3168 {
3169 return dev->num_tx_queues > 1;
3170 }
3171
3172 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3173
3174 #ifdef CONFIG_SYSFS
3175 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3176 #else
3177 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3178 unsigned int rxq)
3179 {
3180 return 0;
3181 }
3182 #endif
3183
3184 #ifdef CONFIG_SYSFS
3185 static inline unsigned int get_netdev_rx_queue_index(
3186 struct netdev_rx_queue *queue)
3187 {
3188 struct net_device *dev = queue->dev;
3189 int index = queue - dev->_rx;
3190
3191 BUG_ON(index >= dev->num_rx_queues);
3192 return index;
3193 }
3194 #endif
3195
3196 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3197 int netif_get_num_default_rss_queues(void);
3198
3199 enum skb_free_reason {
3200 SKB_REASON_CONSUMED,
3201 SKB_REASON_DROPPED,
3202 };
3203
3204 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3205 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3206
3207 /*
3208 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3209 * interrupt context or with hardware interrupts being disabled.
3210 * (in_irq() || irqs_disabled())
3211 *
3212 * We provide four helpers that can be used in following contexts :
3213 *
3214 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3215 * replacing kfree_skb(skb)
3216 *
3217 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3218 * Typically used in place of consume_skb(skb) in TX completion path
3219 *
3220 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3221 * replacing kfree_skb(skb)
3222 *
3223 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3224 * and consumed a packet. Used in place of consume_skb(skb)
3225 */
3226 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3227 {
3228 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3229 }
3230
3231 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3232 {
3233 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3234 }
3235
3236 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3237 {
3238 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3239 }
3240
3241 static inline void dev_consume_skb_any(struct sk_buff *skb)
3242 {
3243 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3244 }
3245
3246 int netif_rx(struct sk_buff *skb);
3247 int netif_rx_ni(struct sk_buff *skb);
3248 int netif_receive_skb(struct sk_buff *skb);
3249 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3250 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3251 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3252 gro_result_t napi_gro_frags(struct napi_struct *napi);
3253 struct packet_offload *gro_find_receive_by_type(__be16 type);
3254 struct packet_offload *gro_find_complete_by_type(__be16 type);
3255
3256 static inline void napi_free_frags(struct napi_struct *napi)
3257 {
3258 kfree_skb(napi->skb);
3259 napi->skb = NULL;
3260 }
3261
3262 bool netdev_is_rx_handler_busy(struct net_device *dev);
3263 int netdev_rx_handler_register(struct net_device *dev,
3264 rx_handler_func_t *rx_handler,
3265 void *rx_handler_data);
3266 void netdev_rx_handler_unregister(struct net_device *dev);
3267
3268 bool dev_valid_name(const char *name);
3269 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3270 int dev_ethtool(struct net *net, struct ifreq *);
3271 unsigned int dev_get_flags(const struct net_device *);
3272 int __dev_change_flags(struct net_device *, unsigned int flags);
3273 int dev_change_flags(struct net_device *, unsigned int);
3274 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3275 unsigned int gchanges);
3276 int dev_change_name(struct net_device *, const char *);
3277 int dev_set_alias(struct net_device *, const char *, size_t);
3278 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3279 int __dev_set_mtu(struct net_device *, int);
3280 int dev_set_mtu(struct net_device *, int);
3281 void dev_set_group(struct net_device *, int);
3282 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3283 int dev_change_carrier(struct net_device *, bool new_carrier);
3284 int dev_get_phys_port_id(struct net_device *dev,
3285 struct netdev_phys_item_id *ppid);
3286 int dev_get_phys_port_name(struct net_device *dev,
3287 char *name, size_t len);
3288 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3289 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3290 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3291 struct netdev_queue *txq, int *ret);
3292
3293 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3294 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3295 int fd, u32 flags);
3296 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3297
3298 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3299 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3300 bool is_skb_forwardable(const struct net_device *dev,
3301 const struct sk_buff *skb);
3302
3303 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3304 struct sk_buff *skb)
3305 {
3306 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3307 unlikely(!is_skb_forwardable(dev, skb))) {
3308 atomic_long_inc(&dev->rx_dropped);
3309 kfree_skb(skb);
3310 return NET_RX_DROP;
3311 }
3312
3313 skb_scrub_packet(skb, true);
3314 skb->priority = 0;
3315 return 0;
3316 }
3317
3318 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3319
3320 extern int netdev_budget;
3321 extern unsigned int netdev_budget_usecs;
3322
3323 /* Called by rtnetlink.c:rtnl_unlock() */
3324 void netdev_run_todo(void);
3325
3326 /**
3327 * dev_put - release reference to device
3328 * @dev: network device
3329 *
3330 * Release reference to device to allow it to be freed.
3331 */
3332 static inline void dev_put(struct net_device *dev)
3333 {
3334 this_cpu_dec(*dev->pcpu_refcnt);
3335 }
3336
3337 /**
3338 * dev_hold - get reference to device
3339 * @dev: network device
3340 *
3341 * Hold reference to device to keep it from being freed.
3342 */
3343 static inline void dev_hold(struct net_device *dev)
3344 {
3345 this_cpu_inc(*dev->pcpu_refcnt);
3346 }
3347
3348 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3349 * and _off may be called from IRQ context, but it is caller
3350 * who is responsible for serialization of these calls.
3351 *
3352 * The name carrier is inappropriate, these functions should really be
3353 * called netif_lowerlayer_*() because they represent the state of any
3354 * kind of lower layer not just hardware media.
3355 */
3356
3357 void linkwatch_init_dev(struct net_device *dev);
3358 void linkwatch_fire_event(struct net_device *dev);
3359 void linkwatch_forget_dev(struct net_device *dev);
3360
3361 /**
3362 * netif_carrier_ok - test if carrier present
3363 * @dev: network device
3364 *
3365 * Check if carrier is present on device
3366 */
3367 static inline bool netif_carrier_ok(const struct net_device *dev)
3368 {
3369 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3370 }
3371
3372 unsigned long dev_trans_start(struct net_device *dev);
3373
3374 void __netdev_watchdog_up(struct net_device *dev);
3375
3376 void netif_carrier_on(struct net_device *dev);
3377
3378 void netif_carrier_off(struct net_device *dev);
3379
3380 /**
3381 * netif_dormant_on - mark device as dormant.
3382 * @dev: network device
3383 *
3384 * Mark device as dormant (as per RFC2863).
3385 *
3386 * The dormant state indicates that the relevant interface is not
3387 * actually in a condition to pass packets (i.e., it is not 'up') but is
3388 * in a "pending" state, waiting for some external event. For "on-
3389 * demand" interfaces, this new state identifies the situation where the
3390 * interface is waiting for events to place it in the up state.
3391 */
3392 static inline void netif_dormant_on(struct net_device *dev)
3393 {
3394 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3395 linkwatch_fire_event(dev);
3396 }
3397
3398 /**
3399 * netif_dormant_off - set device as not dormant.
3400 * @dev: network device
3401 *
3402 * Device is not in dormant state.
3403 */
3404 static inline void netif_dormant_off(struct net_device *dev)
3405 {
3406 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3408 }
3409
3410 /**
3411 * netif_dormant - test if device is dormant
3412 * @dev: network device
3413 *
3414 * Check if device is dormant.
3415 */
3416 static inline bool netif_dormant(const struct net_device *dev)
3417 {
3418 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3419 }
3420
3421
3422 /**
3423 * netif_oper_up - test if device is operational
3424 * @dev: network device
3425 *
3426 * Check if carrier is operational
3427 */
3428 static inline bool netif_oper_up(const struct net_device *dev)
3429 {
3430 return (dev->operstate == IF_OPER_UP ||
3431 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3432 }
3433
3434 /**
3435 * netif_device_present - is device available or removed
3436 * @dev: network device
3437 *
3438 * Check if device has not been removed from system.
3439 */
3440 static inline bool netif_device_present(struct net_device *dev)
3441 {
3442 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3443 }
3444
3445 void netif_device_detach(struct net_device *dev);
3446
3447 void netif_device_attach(struct net_device *dev);
3448
3449 /*
3450 * Network interface message level settings
3451 */
3452
3453 enum {
3454 NETIF_MSG_DRV = 0x0001,
3455 NETIF_MSG_PROBE = 0x0002,
3456 NETIF_MSG_LINK = 0x0004,
3457 NETIF_MSG_TIMER = 0x0008,
3458 NETIF_MSG_IFDOWN = 0x0010,
3459 NETIF_MSG_IFUP = 0x0020,
3460 NETIF_MSG_RX_ERR = 0x0040,
3461 NETIF_MSG_TX_ERR = 0x0080,
3462 NETIF_MSG_TX_QUEUED = 0x0100,
3463 NETIF_MSG_INTR = 0x0200,
3464 NETIF_MSG_TX_DONE = 0x0400,
3465 NETIF_MSG_RX_STATUS = 0x0800,
3466 NETIF_MSG_PKTDATA = 0x1000,
3467 NETIF_MSG_HW = 0x2000,
3468 NETIF_MSG_WOL = 0x4000,
3469 };
3470
3471 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3472 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3473 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3474 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3475 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3476 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3477 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3478 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3479 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3480 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3481 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3482 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3483 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3484 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3485 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3486
3487 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3488 {
3489 /* use default */
3490 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3491 return default_msg_enable_bits;
3492 if (debug_value == 0) /* no output */
3493 return 0;
3494 /* set low N bits */
3495 return (1 << debug_value) - 1;
3496 }
3497
3498 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3499 {
3500 spin_lock(&txq->_xmit_lock);
3501 txq->xmit_lock_owner = cpu;
3502 }
3503
3504 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3505 {
3506 __acquire(&txq->_xmit_lock);
3507 return true;
3508 }
3509
3510 static inline void __netif_tx_release(struct netdev_queue *txq)
3511 {
3512 __release(&txq->_xmit_lock);
3513 }
3514
3515 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3516 {
3517 spin_lock_bh(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = smp_processor_id();
3519 }
3520
3521 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3522 {
3523 bool ok = spin_trylock(&txq->_xmit_lock);
3524 if (likely(ok))
3525 txq->xmit_lock_owner = smp_processor_id();
3526 return ok;
3527 }
3528
3529 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3530 {
3531 txq->xmit_lock_owner = -1;
3532 spin_unlock(&txq->_xmit_lock);
3533 }
3534
3535 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3536 {
3537 txq->xmit_lock_owner = -1;
3538 spin_unlock_bh(&txq->_xmit_lock);
3539 }
3540
3541 static inline void txq_trans_update(struct netdev_queue *txq)
3542 {
3543 if (txq->xmit_lock_owner != -1)
3544 txq->trans_start = jiffies;
3545 }
3546
3547 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3548 static inline void netif_trans_update(struct net_device *dev)
3549 {
3550 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3551
3552 if (txq->trans_start != jiffies)
3553 txq->trans_start = jiffies;
3554 }
3555
3556 /**
3557 * netif_tx_lock - grab network device transmit lock
3558 * @dev: network device
3559 *
3560 * Get network device transmit lock
3561 */
3562 static inline void netif_tx_lock(struct net_device *dev)
3563 {
3564 unsigned int i;
3565 int cpu;
3566
3567 spin_lock(&dev->tx_global_lock);
3568 cpu = smp_processor_id();
3569 for (i = 0; i < dev->num_tx_queues; i++) {
3570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3571
3572 /* We are the only thread of execution doing a
3573 * freeze, but we have to grab the _xmit_lock in
3574 * order to synchronize with threads which are in
3575 * the ->hard_start_xmit() handler and already
3576 * checked the frozen bit.
3577 */
3578 __netif_tx_lock(txq, cpu);
3579 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 __netif_tx_unlock(txq);
3581 }
3582 }
3583
3584 static inline void netif_tx_lock_bh(struct net_device *dev)
3585 {
3586 local_bh_disable();
3587 netif_tx_lock(dev);
3588 }
3589
3590 static inline void netif_tx_unlock(struct net_device *dev)
3591 {
3592 unsigned int i;
3593
3594 for (i = 0; i < dev->num_tx_queues; i++) {
3595 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3596
3597 /* No need to grab the _xmit_lock here. If the
3598 * queue is not stopped for another reason, we
3599 * force a schedule.
3600 */
3601 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3602 netif_schedule_queue(txq);
3603 }
3604 spin_unlock(&dev->tx_global_lock);
3605 }
3606
3607 static inline void netif_tx_unlock_bh(struct net_device *dev)
3608 {
3609 netif_tx_unlock(dev);
3610 local_bh_enable();
3611 }
3612
3613 #define HARD_TX_LOCK(dev, txq, cpu) { \
3614 if ((dev->features & NETIF_F_LLTX) == 0) { \
3615 __netif_tx_lock(txq, cpu); \
3616 } else { \
3617 __netif_tx_acquire(txq); \
3618 } \
3619 }
3620
3621 #define HARD_TX_TRYLOCK(dev, txq) \
3622 (((dev->features & NETIF_F_LLTX) == 0) ? \
3623 __netif_tx_trylock(txq) : \
3624 __netif_tx_acquire(txq))
3625
3626 #define HARD_TX_UNLOCK(dev, txq) { \
3627 if ((dev->features & NETIF_F_LLTX) == 0) { \
3628 __netif_tx_unlock(txq); \
3629 } else { \
3630 __netif_tx_release(txq); \
3631 } \
3632 }
3633
3634 static inline void netif_tx_disable(struct net_device *dev)
3635 {
3636 unsigned int i;
3637 int cpu;
3638
3639 local_bh_disable();
3640 cpu = smp_processor_id();
3641 for (i = 0; i < dev->num_tx_queues; i++) {
3642 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3643
3644 __netif_tx_lock(txq, cpu);
3645 netif_tx_stop_queue(txq);
3646 __netif_tx_unlock(txq);
3647 }
3648 local_bh_enable();
3649 }
3650
3651 static inline void netif_addr_lock(struct net_device *dev)
3652 {
3653 spin_lock(&dev->addr_list_lock);
3654 }
3655
3656 static inline void netif_addr_lock_nested(struct net_device *dev)
3657 {
3658 int subclass = SINGLE_DEPTH_NESTING;
3659
3660 if (dev->netdev_ops->ndo_get_lock_subclass)
3661 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3662
3663 spin_lock_nested(&dev->addr_list_lock, subclass);
3664 }
3665
3666 static inline void netif_addr_lock_bh(struct net_device *dev)
3667 {
3668 spin_lock_bh(&dev->addr_list_lock);
3669 }
3670
3671 static inline void netif_addr_unlock(struct net_device *dev)
3672 {
3673 spin_unlock(&dev->addr_list_lock);
3674 }
3675
3676 static inline void netif_addr_unlock_bh(struct net_device *dev)
3677 {
3678 spin_unlock_bh(&dev->addr_list_lock);
3679 }
3680
3681 /*
3682 * dev_addrs walker. Should be used only for read access. Call with
3683 * rcu_read_lock held.
3684 */
3685 #define for_each_dev_addr(dev, ha) \
3686 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3687
3688 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3689
3690 void ether_setup(struct net_device *dev);
3691
3692 /* Support for loadable net-drivers */
3693 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3694 unsigned char name_assign_type,
3695 void (*setup)(struct net_device *),
3696 unsigned int txqs, unsigned int rxqs);
3697 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3698 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3699
3700 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3701 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3702 count)
3703
3704 int register_netdev(struct net_device *dev);
3705 void unregister_netdev(struct net_device *dev);
3706
3707 /* General hardware address lists handling functions */
3708 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3709 struct netdev_hw_addr_list *from_list, int addr_len);
3710 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3711 struct netdev_hw_addr_list *from_list, int addr_len);
3712 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3713 struct net_device *dev,
3714 int (*sync)(struct net_device *, const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *));
3717 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3718 struct net_device *dev,
3719 int (*unsync)(struct net_device *,
3720 const unsigned char *));
3721 void __hw_addr_init(struct netdev_hw_addr_list *list);
3722
3723 /* Functions used for device addresses handling */
3724 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3725 unsigned char addr_type);
3726 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3727 unsigned char addr_type);
3728 void dev_addr_flush(struct net_device *dev);
3729 int dev_addr_init(struct net_device *dev);
3730
3731 /* Functions used for unicast addresses handling */
3732 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3733 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3734 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3735 int dev_uc_sync(struct net_device *to, struct net_device *from);
3736 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3737 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3738 void dev_uc_flush(struct net_device *dev);
3739 void dev_uc_init(struct net_device *dev);
3740
3741 /**
3742 * __dev_uc_sync - Synchonize device's unicast list
3743 * @dev: device to sync
3744 * @sync: function to call if address should be added
3745 * @unsync: function to call if address should be removed
3746 *
3747 * Add newly added addresses to the interface, and release
3748 * addresses that have been deleted.
3749 */
3750 static inline int __dev_uc_sync(struct net_device *dev,
3751 int (*sync)(struct net_device *,
3752 const unsigned char *),
3753 int (*unsync)(struct net_device *,
3754 const unsigned char *))
3755 {
3756 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3757 }
3758
3759 /**
3760 * __dev_uc_unsync - Remove synchronized addresses from device
3761 * @dev: device to sync
3762 * @unsync: function to call if address should be removed
3763 *
3764 * Remove all addresses that were added to the device by dev_uc_sync().
3765 */
3766 static inline void __dev_uc_unsync(struct net_device *dev,
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3769 {
3770 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3771 }
3772
3773 /* Functions used for multicast addresses handling */
3774 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3775 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3776 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3777 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3778 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3779 int dev_mc_sync(struct net_device *to, struct net_device *from);
3780 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3781 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3782 void dev_mc_flush(struct net_device *dev);
3783 void dev_mc_init(struct net_device *dev);
3784
3785 /**
3786 * __dev_mc_sync - Synchonize device's multicast list
3787 * @dev: device to sync
3788 * @sync: function to call if address should be added
3789 * @unsync: function to call if address should be removed
3790 *
3791 * Add newly added addresses to the interface, and release
3792 * addresses that have been deleted.
3793 */
3794 static inline int __dev_mc_sync(struct net_device *dev,
3795 int (*sync)(struct net_device *,
3796 const unsigned char *),
3797 int (*unsync)(struct net_device *,
3798 const unsigned char *))
3799 {
3800 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3801 }
3802
3803 /**
3804 * __dev_mc_unsync - Remove synchronized addresses from device
3805 * @dev: device to sync
3806 * @unsync: function to call if address should be removed
3807 *
3808 * Remove all addresses that were added to the device by dev_mc_sync().
3809 */
3810 static inline void __dev_mc_unsync(struct net_device *dev,
3811 int (*unsync)(struct net_device *,
3812 const unsigned char *))
3813 {
3814 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3815 }
3816
3817 /* Functions used for secondary unicast and multicast support */
3818 void dev_set_rx_mode(struct net_device *dev);
3819 void __dev_set_rx_mode(struct net_device *dev);
3820 int dev_set_promiscuity(struct net_device *dev, int inc);
3821 int dev_set_allmulti(struct net_device *dev, int inc);
3822 void netdev_state_change(struct net_device *dev);
3823 void netdev_notify_peers(struct net_device *dev);
3824 void netdev_features_change(struct net_device *dev);
3825 /* Load a device via the kmod */
3826 void dev_load(struct net *net, const char *name);
3827 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3828 struct rtnl_link_stats64 *storage);
3829 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3830 const struct net_device_stats *netdev_stats);
3831
3832 extern int netdev_max_backlog;
3833 extern int netdev_tstamp_prequeue;
3834 extern int weight_p;
3835 extern int dev_weight_rx_bias;
3836 extern int dev_weight_tx_bias;
3837 extern int dev_rx_weight;
3838 extern int dev_tx_weight;
3839
3840 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3841 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3842 struct list_head **iter);
3843 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3844 struct list_head **iter);
3845
3846 /* iterate through upper list, must be called under RCU read lock */
3847 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3848 for (iter = &(dev)->adj_list.upper, \
3849 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3850 updev; \
3851 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3852
3853 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3854 int (*fn)(struct net_device *upper_dev,
3855 void *data),
3856 void *data);
3857
3858 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3859 struct net_device *upper_dev);
3860
3861 void *netdev_lower_get_next_private(struct net_device *dev,
3862 struct list_head **iter);
3863 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3864 struct list_head **iter);
3865
3866 #define netdev_for_each_lower_private(dev, priv, iter) \
3867 for (iter = (dev)->adj_list.lower.next, \
3868 priv = netdev_lower_get_next_private(dev, &(iter)); \
3869 priv; \
3870 priv = netdev_lower_get_next_private(dev, &(iter)))
3871
3872 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3873 for (iter = &(dev)->adj_list.lower, \
3874 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3875 priv; \
3876 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3877
3878 void *netdev_lower_get_next(struct net_device *dev,
3879 struct list_head **iter);
3880
3881 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3882 for (iter = (dev)->adj_list.lower.next, \
3883 ldev = netdev_lower_get_next(dev, &(iter)); \
3884 ldev; \
3885 ldev = netdev_lower_get_next(dev, &(iter)))
3886
3887 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3888 struct list_head **iter);
3889 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3890 struct list_head **iter);
3891
3892 int netdev_walk_all_lower_dev(struct net_device *dev,
3893 int (*fn)(struct net_device *lower_dev,
3894 void *data),
3895 void *data);
3896 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3897 int (*fn)(struct net_device *lower_dev,
3898 void *data),
3899 void *data);
3900
3901 void *netdev_adjacent_get_private(struct list_head *adj_list);
3902 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3903 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3904 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3905 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3906 int netdev_master_upper_dev_link(struct net_device *dev,
3907 struct net_device *upper_dev,
3908 void *upper_priv, void *upper_info);
3909 void netdev_upper_dev_unlink(struct net_device *dev,
3910 struct net_device *upper_dev);
3911 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3912 void *netdev_lower_dev_get_private(struct net_device *dev,
3913 struct net_device *lower_dev);
3914 void netdev_lower_state_changed(struct net_device *lower_dev,
3915 void *lower_state_info);
3916
3917 /* RSS keys are 40 or 52 bytes long */
3918 #define NETDEV_RSS_KEY_LEN 52
3919 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3920 void netdev_rss_key_fill(void *buffer, size_t len);
3921
3922 int dev_get_nest_level(struct net_device *dev);
3923 int skb_checksum_help(struct sk_buff *skb);
3924 int skb_crc32c_csum_help(struct sk_buff *skb);
3925 int skb_csum_hwoffload_help(struct sk_buff *skb,
3926 const netdev_features_t features);
3927
3928 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3929 netdev_features_t features, bool tx_path);
3930 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3931 netdev_features_t features);
3932
3933 struct netdev_bonding_info {
3934 ifslave slave;
3935 ifbond master;
3936 };
3937
3938 struct netdev_notifier_bonding_info {
3939 struct netdev_notifier_info info; /* must be first */
3940 struct netdev_bonding_info bonding_info;
3941 };
3942
3943 void netdev_bonding_info_change(struct net_device *dev,
3944 struct netdev_bonding_info *bonding_info);
3945
3946 static inline
3947 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3948 {
3949 return __skb_gso_segment(skb, features, true);
3950 }
3951 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3952
3953 static inline bool can_checksum_protocol(netdev_features_t features,
3954 __be16 protocol)
3955 {
3956 if (protocol == htons(ETH_P_FCOE))
3957 return !!(features & NETIF_F_FCOE_CRC);
3958
3959 /* Assume this is an IP checksum (not SCTP CRC) */
3960
3961 if (features & NETIF_F_HW_CSUM) {
3962 /* Can checksum everything */
3963 return true;
3964 }
3965
3966 switch (protocol) {
3967 case htons(ETH_P_IP):
3968 return !!(features & NETIF_F_IP_CSUM);
3969 case htons(ETH_P_IPV6):
3970 return !!(features & NETIF_F_IPV6_CSUM);
3971 default:
3972 return false;
3973 }
3974 }
3975
3976 #ifdef CONFIG_BUG
3977 void netdev_rx_csum_fault(struct net_device *dev);
3978 #else
3979 static inline void netdev_rx_csum_fault(struct net_device *dev)
3980 {
3981 }
3982 #endif
3983 /* rx skb timestamps */
3984 void net_enable_timestamp(void);
3985 void net_disable_timestamp(void);
3986
3987 #ifdef CONFIG_PROC_FS
3988 int __init dev_proc_init(void);
3989 #else
3990 #define dev_proc_init() 0
3991 #endif
3992
3993 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3994 struct sk_buff *skb, struct net_device *dev,
3995 bool more)
3996 {
3997 skb->xmit_more = more ? 1 : 0;
3998 return ops->ndo_start_xmit(skb, dev);
3999 }
4000
4001 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4002 struct netdev_queue *txq, bool more)
4003 {
4004 const struct net_device_ops *ops = dev->netdev_ops;
4005 int rc;
4006
4007 rc = __netdev_start_xmit(ops, skb, dev, more);
4008 if (rc == NETDEV_TX_OK)
4009 txq_trans_update(txq);
4010
4011 return rc;
4012 }
4013
4014 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4015 const void *ns);
4016 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4017 const void *ns);
4018
4019 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4020 {
4021 return netdev_class_create_file_ns(class_attr, NULL);
4022 }
4023
4024 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4025 {
4026 netdev_class_remove_file_ns(class_attr, NULL);
4027 }
4028
4029 extern struct kobj_ns_type_operations net_ns_type_operations;
4030
4031 const char *netdev_drivername(const struct net_device *dev);
4032
4033 void linkwatch_run_queue(void);
4034
4035 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4036 netdev_features_t f2)
4037 {
4038 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4039 if (f1 & NETIF_F_HW_CSUM)
4040 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4041 else
4042 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4043 }
4044
4045 return f1 & f2;
4046 }
4047
4048 static inline netdev_features_t netdev_get_wanted_features(
4049 struct net_device *dev)
4050 {
4051 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4052 }
4053 netdev_features_t netdev_increment_features(netdev_features_t all,
4054 netdev_features_t one, netdev_features_t mask);
4055
4056 /* Allow TSO being used on stacked device :
4057 * Performing the GSO segmentation before last device
4058 * is a performance improvement.
4059 */
4060 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4061 netdev_features_t mask)
4062 {
4063 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4064 }
4065
4066 int __netdev_update_features(struct net_device *dev);
4067 void netdev_update_features(struct net_device *dev);
4068 void netdev_change_features(struct net_device *dev);
4069
4070 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4071 struct net_device *dev);
4072
4073 netdev_features_t passthru_features_check(struct sk_buff *skb,
4074 struct net_device *dev,
4075 netdev_features_t features);
4076 netdev_features_t netif_skb_features(struct sk_buff *skb);
4077
4078 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4079 {
4080 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4081
4082 /* check flags correspondence */
4083 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4084 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4085 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4086 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4087 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4088 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4089 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4090 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4091 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4092 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4093 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4094 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4095 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4096 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4097 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4098 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4099
4100 return (features & feature) == feature;
4101 }
4102
4103 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4104 {
4105 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4106 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4107 }
4108
4109 static inline bool netif_needs_gso(struct sk_buff *skb,
4110 netdev_features_t features)
4111 {
4112 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4113 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4114 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4115 }
4116
4117 static inline void netif_set_gso_max_size(struct net_device *dev,
4118 unsigned int size)
4119 {
4120 dev->gso_max_size = size;
4121 }
4122
4123 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4124 int pulled_hlen, u16 mac_offset,
4125 int mac_len)
4126 {
4127 skb->protocol = protocol;
4128 skb->encapsulation = 1;
4129 skb_push(skb, pulled_hlen);
4130 skb_reset_transport_header(skb);
4131 skb->mac_header = mac_offset;
4132 skb->network_header = skb->mac_header + mac_len;
4133 skb->mac_len = mac_len;
4134 }
4135
4136 static inline bool netif_is_macsec(const struct net_device *dev)
4137 {
4138 return dev->priv_flags & IFF_MACSEC;
4139 }
4140
4141 static inline bool netif_is_macvlan(const struct net_device *dev)
4142 {
4143 return dev->priv_flags & IFF_MACVLAN;
4144 }
4145
4146 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4147 {
4148 return dev->priv_flags & IFF_MACVLAN_PORT;
4149 }
4150
4151 static inline bool netif_is_ipvlan(const struct net_device *dev)
4152 {
4153 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4154 }
4155
4156 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4157 {
4158 return dev->priv_flags & IFF_IPVLAN_MASTER;
4159 }
4160
4161 static inline bool netif_is_bond_master(const struct net_device *dev)
4162 {
4163 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4164 }
4165
4166 static inline bool netif_is_bond_slave(const struct net_device *dev)
4167 {
4168 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4169 }
4170
4171 static inline bool netif_supports_nofcs(struct net_device *dev)
4172 {
4173 return dev->priv_flags & IFF_SUPP_NOFCS;
4174 }
4175
4176 static inline bool netif_is_l3_master(const struct net_device *dev)
4177 {
4178 return dev->priv_flags & IFF_L3MDEV_MASTER;
4179 }
4180
4181 static inline bool netif_is_l3_slave(const struct net_device *dev)
4182 {
4183 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4184 }
4185
4186 static inline bool netif_is_bridge_master(const struct net_device *dev)
4187 {
4188 return dev->priv_flags & IFF_EBRIDGE;
4189 }
4190
4191 static inline bool netif_is_bridge_port(const struct net_device *dev)
4192 {
4193 return dev->priv_flags & IFF_BRIDGE_PORT;
4194 }
4195
4196 static inline bool netif_is_ovs_master(const struct net_device *dev)
4197 {
4198 return dev->priv_flags & IFF_OPENVSWITCH;
4199 }
4200
4201 static inline bool netif_is_ovs_port(const struct net_device *dev)
4202 {
4203 return dev->priv_flags & IFF_OVS_DATAPATH;
4204 }
4205
4206 static inline bool netif_is_team_master(const struct net_device *dev)
4207 {
4208 return dev->priv_flags & IFF_TEAM;
4209 }
4210
4211 static inline bool netif_is_team_port(const struct net_device *dev)
4212 {
4213 return dev->priv_flags & IFF_TEAM_PORT;
4214 }
4215
4216 static inline bool netif_is_lag_master(const struct net_device *dev)
4217 {
4218 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4219 }
4220
4221 static inline bool netif_is_lag_port(const struct net_device *dev)
4222 {
4223 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4224 }
4225
4226 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4227 {
4228 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4229 }
4230
4231 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4232 static inline void netif_keep_dst(struct net_device *dev)
4233 {
4234 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4235 }
4236
4237 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4238 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4239 {
4240 /* TODO: reserve and use an additional IFF bit, if we get more users */
4241 return dev->priv_flags & IFF_MACSEC;
4242 }
4243
4244 extern struct pernet_operations __net_initdata loopback_net_ops;
4245
4246 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4247
4248 /* netdev_printk helpers, similar to dev_printk */
4249
4250 static inline const char *netdev_name(const struct net_device *dev)
4251 {
4252 if (!dev->name[0] || strchr(dev->name, '%'))
4253 return "(unnamed net_device)";
4254 return dev->name;
4255 }
4256
4257 static inline bool netdev_unregistering(const struct net_device *dev)
4258 {
4259 return dev->reg_state == NETREG_UNREGISTERING;
4260 }
4261
4262 static inline const char *netdev_reg_state(const struct net_device *dev)
4263 {
4264 switch (dev->reg_state) {
4265 case NETREG_UNINITIALIZED: return " (uninitialized)";
4266 case NETREG_REGISTERED: return "";
4267 case NETREG_UNREGISTERING: return " (unregistering)";
4268 case NETREG_UNREGISTERED: return " (unregistered)";
4269 case NETREG_RELEASED: return " (released)";
4270 case NETREG_DUMMY: return " (dummy)";
4271 }
4272
4273 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4274 return " (unknown)";
4275 }
4276
4277 __printf(3, 4)
4278 void netdev_printk(const char *level, const struct net_device *dev,
4279 const char *format, ...);
4280 __printf(2, 3)
4281 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4282 __printf(2, 3)
4283 void netdev_alert(const struct net_device *dev, const char *format, ...);
4284 __printf(2, 3)
4285 void netdev_crit(const struct net_device *dev, const char *format, ...);
4286 __printf(2, 3)
4287 void netdev_err(const struct net_device *dev, const char *format, ...);
4288 __printf(2, 3)
4289 void netdev_warn(const struct net_device *dev, const char *format, ...);
4290 __printf(2, 3)
4291 void netdev_notice(const struct net_device *dev, const char *format, ...);
4292 __printf(2, 3)
4293 void netdev_info(const struct net_device *dev, const char *format, ...);
4294
4295 #define MODULE_ALIAS_NETDEV(device) \
4296 MODULE_ALIAS("netdev-" device)
4297
4298 #if defined(CONFIG_DYNAMIC_DEBUG)
4299 #define netdev_dbg(__dev, format, args...) \
4300 do { \
4301 dynamic_netdev_dbg(__dev, format, ##args); \
4302 } while (0)
4303 #elif defined(DEBUG)
4304 #define netdev_dbg(__dev, format, args...) \
4305 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4306 #else
4307 #define netdev_dbg(__dev, format, args...) \
4308 ({ \
4309 if (0) \
4310 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4311 })
4312 #endif
4313
4314 #if defined(VERBOSE_DEBUG)
4315 #define netdev_vdbg netdev_dbg
4316 #else
4317
4318 #define netdev_vdbg(dev, format, args...) \
4319 ({ \
4320 if (0) \
4321 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4322 0; \
4323 })
4324 #endif
4325
4326 /*
4327 * netdev_WARN() acts like dev_printk(), but with the key difference
4328 * of using a WARN/WARN_ON to get the message out, including the
4329 * file/line information and a backtrace.
4330 */
4331 #define netdev_WARN(dev, format, args...) \
4332 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4333 netdev_reg_state(dev), ##args)
4334
4335 /* netif printk helpers, similar to netdev_printk */
4336
4337 #define netif_printk(priv, type, level, dev, fmt, args...) \
4338 do { \
4339 if (netif_msg_##type(priv)) \
4340 netdev_printk(level, (dev), fmt, ##args); \
4341 } while (0)
4342
4343 #define netif_level(level, priv, type, dev, fmt, args...) \
4344 do { \
4345 if (netif_msg_##type(priv)) \
4346 netdev_##level(dev, fmt, ##args); \
4347 } while (0)
4348
4349 #define netif_emerg(priv, type, dev, fmt, args...) \
4350 netif_level(emerg, priv, type, dev, fmt, ##args)
4351 #define netif_alert(priv, type, dev, fmt, args...) \
4352 netif_level(alert, priv, type, dev, fmt, ##args)
4353 #define netif_crit(priv, type, dev, fmt, args...) \
4354 netif_level(crit, priv, type, dev, fmt, ##args)
4355 #define netif_err(priv, type, dev, fmt, args...) \
4356 netif_level(err, priv, type, dev, fmt, ##args)
4357 #define netif_warn(priv, type, dev, fmt, args...) \
4358 netif_level(warn, priv, type, dev, fmt, ##args)
4359 #define netif_notice(priv, type, dev, fmt, args...) \
4360 netif_level(notice, priv, type, dev, fmt, ##args)
4361 #define netif_info(priv, type, dev, fmt, args...) \
4362 netif_level(info, priv, type, dev, fmt, ##args)
4363
4364 #if defined(CONFIG_DYNAMIC_DEBUG)
4365 #define netif_dbg(priv, type, netdev, format, args...) \
4366 do { \
4367 if (netif_msg_##type(priv)) \
4368 dynamic_netdev_dbg(netdev, format, ##args); \
4369 } while (0)
4370 #elif defined(DEBUG)
4371 #define netif_dbg(priv, type, dev, format, args...) \
4372 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4373 #else
4374 #define netif_dbg(priv, type, dev, format, args...) \
4375 ({ \
4376 if (0) \
4377 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4378 0; \
4379 })
4380 #endif
4381
4382 /* if @cond then downgrade to debug, else print at @level */
4383 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4384 do { \
4385 if (cond) \
4386 netif_dbg(priv, type, netdev, fmt, ##args); \
4387 else \
4388 netif_ ## level(priv, type, netdev, fmt, ##args); \
4389 } while (0)
4390
4391 #if defined(VERBOSE_DEBUG)
4392 #define netif_vdbg netif_dbg
4393 #else
4394 #define netif_vdbg(priv, type, dev, format, args...) \
4395 ({ \
4396 if (0) \
4397 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4398 0; \
4399 })
4400 #endif
4401
4402 /*
4403 * The list of packet types we will receive (as opposed to discard)
4404 * and the routines to invoke.
4405 *
4406 * Why 16. Because with 16 the only overlap we get on a hash of the
4407 * low nibble of the protocol value is RARP/SNAP/X.25.
4408 *
4409 * NOTE: That is no longer true with the addition of VLAN tags. Not
4410 * sure which should go first, but I bet it won't make much
4411 * difference if we are running VLANs. The good news is that
4412 * this protocol won't be in the list unless compiled in, so
4413 * the average user (w/out VLANs) will not be adversely affected.
4414 * --BLG
4415 *
4416 * 0800 IP
4417 * 8100 802.1Q VLAN
4418 * 0001 802.3
4419 * 0002 AX.25
4420 * 0004 802.2
4421 * 8035 RARP
4422 * 0005 SNAP
4423 * 0805 X.25
4424 * 0806 ARP
4425 * 8137 IPX
4426 * 0009 Localtalk
4427 * 86DD IPv6
4428 */
4429 #define PTYPE_HASH_SIZE (16)
4430 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4431
4432 #endif /* _LINUX_NETDEVICE_H */