]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - include/linux/netdevice.h
net: add rx_nohandler stat counter
[mirror_ubuntu-zesty-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 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
70
71 /*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
103
104 enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 };
110 typedef enum netdev_tx netdev_tx_t;
111
112 /*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116 static inline bool dev_xmit_complete(int rc)
117 {
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128 }
129
130 /*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135 #if defined(CONFIG_HYPERV_NET)
136 # define LL_MAX_HEADER 128
137 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
138 # if defined(CONFIG_MAC80211_MESH)
139 # define LL_MAX_HEADER 128
140 # else
141 # define LL_MAX_HEADER 96
142 # endif
143 #else
144 # define LL_MAX_HEADER 32
145 #endif
146
147 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
148 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
149 #define MAX_HEADER LL_MAX_HEADER
150 #else
151 #define MAX_HEADER (LL_MAX_HEADER + 48)
152 #endif
153
154 /*
155 * Old network device statistics. Fields are native words
156 * (unsigned long) so they can be read and written atomically.
157 */
158
159 struct net_device_stats {
160 unsigned long rx_packets;
161 unsigned long tx_packets;
162 unsigned long rx_bytes;
163 unsigned long tx_bytes;
164 unsigned long rx_errors;
165 unsigned long tx_errors;
166 unsigned long rx_dropped;
167 unsigned long tx_dropped;
168 unsigned long multicast;
169 unsigned long collisions;
170 unsigned long rx_length_errors;
171 unsigned long rx_over_errors;
172 unsigned long rx_crc_errors;
173 unsigned long rx_frame_errors;
174 unsigned long rx_fifo_errors;
175 unsigned long rx_missed_errors;
176 unsigned long tx_aborted_errors;
177 unsigned long tx_carrier_errors;
178 unsigned long tx_fifo_errors;
179 unsigned long tx_heartbeat_errors;
180 unsigned long tx_window_errors;
181 unsigned long rx_compressed;
182 unsigned long tx_compressed;
183 };
184
185
186 #include <linux/cache.h>
187 #include <linux/skbuff.h>
188
189 #ifdef CONFIG_RPS
190 #include <linux/static_key.h>
191 extern struct static_key rps_needed;
192 #endif
193
194 struct neighbour;
195 struct neigh_parms;
196 struct sk_buff;
197
198 struct netdev_hw_addr {
199 struct list_head list;
200 unsigned char addr[MAX_ADDR_LEN];
201 unsigned char type;
202 #define NETDEV_HW_ADDR_T_LAN 1
203 #define NETDEV_HW_ADDR_T_SAN 2
204 #define NETDEV_HW_ADDR_T_SLAVE 3
205 #define NETDEV_HW_ADDR_T_UNICAST 4
206 #define NETDEV_HW_ADDR_T_MULTICAST 5
207 bool global_use;
208 int sync_cnt;
209 int refcount;
210 int synced;
211 struct rcu_head rcu_head;
212 };
213
214 struct netdev_hw_addr_list {
215 struct list_head list;
216 int count;
217 };
218
219 #define netdev_hw_addr_list_count(l) ((l)->count)
220 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
221 #define netdev_hw_addr_list_for_each(ha, l) \
222 list_for_each_entry(ha, &(l)->list, list)
223
224 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
225 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
226 #define netdev_for_each_uc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
228
229 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
230 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
231 #define netdev_for_each_mc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
233
234 struct hh_cache {
235 u16 hh_len;
236 u16 __pad;
237 seqlock_t hh_lock;
238
239 /* cached hardware header; allow for machine alignment needs. */
240 #define HH_DATA_MOD 16
241 #define HH_DATA_OFF(__len) \
242 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
243 #define HH_DATA_ALIGN(__len) \
244 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
245 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 };
247
248 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
249 * Alternative is:
250 * dev->hard_header_len ? (dev->hard_header_len +
251 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
252 *
253 * We could use other alignment values, but we must maintain the
254 * relationship HH alignment <= LL alignment.
255 */
256 #define LL_RESERVED_SPACE(dev) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260
261 struct header_ops {
262 int (*create) (struct sk_buff *skb, struct net_device *dev,
263 unsigned short type, const void *daddr,
264 const void *saddr, unsigned int len);
265 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
266 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
267 void (*cache_update)(struct hh_cache *hh,
268 const struct net_device *dev,
269 const unsigned char *haddr);
270 };
271
272 /* These flag bits are private to the generic network queueing
273 * layer, they may not be explicitly referenced by any other
274 * code.
275 */
276
277 enum netdev_state_t {
278 __LINK_STATE_START,
279 __LINK_STATE_PRESENT,
280 __LINK_STATE_NOCARRIER,
281 __LINK_STATE_LINKWATCH_PENDING,
282 __LINK_STATE_DORMANT,
283 };
284
285
286 /*
287 * This structure holds at boot time configured netdevice settings. They
288 * are then used in the device probing.
289 */
290 struct netdev_boot_setup {
291 char name[IFNAMSIZ];
292 struct ifmap map;
293 };
294 #define NETDEV_BOOT_SETUP_MAX 8
295
296 int __init netdev_boot_setup(char *str);
297
298 /*
299 * Structure for NAPI scheduling similar to tasklet but with weighting
300 */
301 struct napi_struct {
302 /* The poll_list must only be managed by the entity which
303 * changes the state of the NAPI_STATE_SCHED bit. This means
304 * whoever atomically sets that bit can add this napi_struct
305 * to the per-cpu poll_list, and whoever clears that bit
306 * can remove from the list right before clearing the bit.
307 */
308 struct list_head poll_list;
309
310 unsigned long state;
311 int weight;
312 unsigned int gro_count;
313 int (*poll)(struct napi_struct *, int);
314 #ifdef CONFIG_NETPOLL
315 spinlock_t poll_lock;
316 int poll_owner;
317 #endif
318 struct net_device *dev;
319 struct sk_buff *gro_list;
320 struct sk_buff *skb;
321 struct hrtimer timer;
322 struct list_head dev_list;
323 struct hlist_node napi_hash_node;
324 unsigned int napi_id;
325 };
326
327 enum {
328 NAPI_STATE_SCHED, /* Poll is scheduled */
329 NAPI_STATE_DISABLE, /* Disable pending */
330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
333 };
334
335 enum gro_result {
336 GRO_MERGED,
337 GRO_MERGED_FREE,
338 GRO_HELD,
339 GRO_NORMAL,
340 GRO_DROP,
341 };
342 typedef enum gro_result gro_result_t;
343
344 /*
345 * enum rx_handler_result - Possible return values for rx_handlers.
346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * further.
348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
349 * case skb->dev was changed by rx_handler.
350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
351 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
352 *
353 * rx_handlers are functions called from inside __netif_receive_skb(), to do
354 * special processing of the skb, prior to delivery to protocol handlers.
355 *
356 * Currently, a net_device can only have a single rx_handler registered. Trying
357 * to register a second rx_handler will return -EBUSY.
358 *
359 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
360 * To unregister a rx_handler on a net_device, use
361 * netdev_rx_handler_unregister().
362 *
363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * do with the skb.
365 *
366 * If the rx_handler consumed to skb in some way, it should return
367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
368 * the skb to be delivered in some other ways.
369 *
370 * If the rx_handler changed skb->dev, to divert the skb to another
371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
372 * new device will be called if it exists.
373 *
374 * If the rx_handler consider the skb should be ignored, it should return
375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
376 * are registered on exact device (ptype->dev == skb->dev).
377 *
378 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
379 * delivered, it should return RX_HANDLER_PASS.
380 *
381 * A device without a registered rx_handler will behave as if rx_handler
382 * returned RX_HANDLER_PASS.
383 */
384
385 enum rx_handler_result {
386 RX_HANDLER_CONSUMED,
387 RX_HANDLER_ANOTHER,
388 RX_HANDLER_EXACT,
389 RX_HANDLER_PASS,
390 };
391 typedef enum rx_handler_result rx_handler_result_t;
392 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
393
394 void __napi_schedule(struct napi_struct *n);
395 void __napi_schedule_irqoff(struct napi_struct *n);
396
397 static inline bool napi_disable_pending(struct napi_struct *n)
398 {
399 return test_bit(NAPI_STATE_DISABLE, &n->state);
400 }
401
402 /**
403 * napi_schedule_prep - check if napi can be scheduled
404 * @n: napi context
405 *
406 * Test if NAPI routine is already running, and if not mark
407 * it as running. This is used as a condition variable
408 * insure only one NAPI poll instance runs. We also make
409 * sure there is no pending NAPI disable.
410 */
411 static inline bool napi_schedule_prep(struct napi_struct *n)
412 {
413 return !napi_disable_pending(n) &&
414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
415 }
416
417 /**
418 * napi_schedule - schedule NAPI poll
419 * @n: napi context
420 *
421 * Schedule NAPI poll routine to be called if it is not already
422 * running.
423 */
424 static inline void napi_schedule(struct napi_struct *n)
425 {
426 if (napi_schedule_prep(n))
427 __napi_schedule(n);
428 }
429
430 /**
431 * napi_schedule_irqoff - schedule NAPI poll
432 * @n: napi context
433 *
434 * Variant of napi_schedule(), assuming hard irqs are masked.
435 */
436 static inline void napi_schedule_irqoff(struct napi_struct *n)
437 {
438 if (napi_schedule_prep(n))
439 __napi_schedule_irqoff(n);
440 }
441
442 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
443 static inline bool napi_reschedule(struct napi_struct *napi)
444 {
445 if (napi_schedule_prep(napi)) {
446 __napi_schedule(napi);
447 return true;
448 }
449 return false;
450 }
451
452 void __napi_complete(struct napi_struct *n);
453 void napi_complete_done(struct napi_struct *n, int work_done);
454 /**
455 * napi_complete - NAPI processing complete
456 * @n: napi context
457 *
458 * Mark NAPI processing as complete.
459 * Consider using napi_complete_done() instead.
460 */
461 static inline void napi_complete(struct napi_struct *n)
462 {
463 return napi_complete_done(n, 0);
464 }
465
466 /**
467 * napi_hash_add - add a NAPI to global hashtable
468 * @napi: napi context
469 *
470 * generate a new napi_id and store a @napi under it in napi_hash
471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
472 * Note: This is normally automatically done from netif_napi_add(),
473 * so might disappear in a future linux version.
474 */
475 void napi_hash_add(struct napi_struct *napi);
476
477 /**
478 * napi_hash_del - remove a NAPI from global table
479 * @napi: napi context
480 *
481 * Warning: caller must observe rcu grace period
482 * before freeing memory containing @napi, if
483 * this function returns true.
484 * Note: core networking stack automatically calls it
485 * from netif_napi_del()
486 * Drivers might want to call this helper to combine all
487 * the needed rcu grace periods into a single one.
488 */
489 bool napi_hash_del(struct napi_struct *napi);
490
491 /**
492 * napi_disable - prevent NAPI from scheduling
493 * @n: napi context
494 *
495 * Stop NAPI from being scheduled on this context.
496 * Waits till any outstanding processing completes.
497 */
498 void napi_disable(struct napi_struct *n);
499
500 /**
501 * napi_enable - enable NAPI scheduling
502 * @n: napi context
503 *
504 * Resume NAPI from being scheduled on this context.
505 * Must be paired with napi_disable.
506 */
507 static inline void napi_enable(struct napi_struct *n)
508 {
509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
510 smp_mb__before_atomic();
511 clear_bit(NAPI_STATE_SCHED, &n->state);
512 clear_bit(NAPI_STATE_NPSVC, &n->state);
513 }
514
515 /**
516 * napi_synchronize - wait until NAPI is not running
517 * @n: napi context
518 *
519 * Wait until NAPI is done being scheduled on this context.
520 * Waits till any outstanding processing completes but
521 * does not disable future activations.
522 */
523 static inline void napi_synchronize(const struct napi_struct *n)
524 {
525 if (IS_ENABLED(CONFIG_SMP))
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 msleep(1);
528 else
529 barrier();
530 }
531
532 enum netdev_queue_state_t {
533 __QUEUE_STATE_DRV_XOFF,
534 __QUEUE_STATE_STACK_XOFF,
535 __QUEUE_STATE_FROZEN,
536 };
537
538 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
539 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
540 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
541
542 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
543 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
544 QUEUE_STATE_FROZEN)
545 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
546 QUEUE_STATE_FROZEN)
547
548 /*
549 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
550 * netif_tx_* functions below are used to manipulate this flag. The
551 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
552 * queue independently. The netif_xmit_*stopped functions below are called
553 * to check if the queue has been stopped by the driver or stack (either
554 * of the XOFF bits are set in the state). Drivers should not need to call
555 * netif_xmit*stopped functions, they should only be using netif_tx_*.
556 */
557
558 struct netdev_queue {
559 /*
560 * read mostly part
561 */
562 struct net_device *dev;
563 struct Qdisc __rcu *qdisc;
564 struct Qdisc *qdisc_sleeping;
565 #ifdef CONFIG_SYSFS
566 struct kobject kobj;
567 #endif
568 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
569 int numa_node;
570 #endif
571 /*
572 * write mostly part
573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
576 /*
577 * please use this field instead of dev->trans_start
578 */
579 unsigned long trans_start;
580
581 /*
582 * Number of TX timeouts for this queue
583 * (/sys/class/net/DEV/Q/trans_timeout)
584 */
585 unsigned long trans_timeout;
586
587 unsigned long state;
588
589 #ifdef CONFIG_BQL
590 struct dql dql;
591 #endif
592 unsigned long tx_maxrate;
593 } ____cacheline_aligned_in_smp;
594
595 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
596 {
597 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 return q->numa_node;
599 #else
600 return NUMA_NO_NODE;
601 #endif
602 }
603
604 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
605 {
606 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 q->numa_node = node;
608 #endif
609 }
610
611 #ifdef CONFIG_RPS
612 /*
613 * This structure holds an RPS map which can be of variable length. The
614 * map is an array of CPUs.
615 */
616 struct rps_map {
617 unsigned int len;
618 struct rcu_head rcu;
619 u16 cpus[0];
620 };
621 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
622
623 /*
624 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
625 * tail pointer for that CPU's input queue at the time of last enqueue, and
626 * a hardware filter index.
627 */
628 struct rps_dev_flow {
629 u16 cpu;
630 u16 filter;
631 unsigned int last_qtail;
632 };
633 #define RPS_NO_FILTER 0xffff
634
635 /*
636 * The rps_dev_flow_table structure contains a table of flow mappings.
637 */
638 struct rps_dev_flow_table {
639 unsigned int mask;
640 struct rcu_head rcu;
641 struct rps_dev_flow flows[0];
642 };
643 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
644 ((_num) * sizeof(struct rps_dev_flow)))
645
646 /*
647 * The rps_sock_flow_table contains mappings of flows to the last CPU
648 * on which they were processed by the application (set in recvmsg).
649 * Each entry is a 32bit value. Upper part is the high order bits
650 * of flow hash, lower part is cpu number.
651 * rps_cpu_mask is used to partition the space, depending on number of
652 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
653 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
654 * meaning we use 32-6=26 bits for the hash.
655 */
656 struct rps_sock_flow_table {
657 u32 mask;
658
659 u32 ents[0] ____cacheline_aligned_in_smp;
660 };
661 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
662
663 #define RPS_NO_CPU 0xffff
664
665 extern u32 rps_cpu_mask;
666 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667
668 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
669 u32 hash)
670 {
671 if (table && hash) {
672 unsigned int index = hash & table->mask;
673 u32 val = hash & ~rps_cpu_mask;
674
675 /* We only give a hint, preemption can change cpu under us */
676 val |= raw_smp_processor_id();
677
678 if (table->ents[index] != val)
679 table->ents[index] = val;
680 }
681 }
682
683 #ifdef CONFIG_RFS_ACCEL
684 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
685 u16 filter_id);
686 #endif
687 #endif /* CONFIG_RPS */
688
689 /* This structure contains an instance of an RX queue. */
690 struct netdev_rx_queue {
691 #ifdef CONFIG_RPS
692 struct rps_map __rcu *rps_map;
693 struct rps_dev_flow_table __rcu *rps_flow_table;
694 #endif
695 struct kobject kobj;
696 struct net_device *dev;
697 } ____cacheline_aligned_in_smp;
698
699 /*
700 * RX queue sysfs structures and functions.
701 */
702 struct rx_queue_attribute {
703 struct attribute attr;
704 ssize_t (*show)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, char *buf);
706 ssize_t (*store)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, const char *buf, size_t len);
708 };
709
710 #ifdef CONFIG_XPS
711 /*
712 * This structure holds an XPS map which can be of variable length. The
713 * map is an array of queues.
714 */
715 struct xps_map {
716 unsigned int len;
717 unsigned int alloc_len;
718 struct rcu_head rcu;
719 u16 queues[0];
720 };
721 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
722 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
723 - sizeof(struct xps_map)) / sizeof(u16))
724
725 /*
726 * This structure holds all XPS maps for device. Maps are indexed by CPU.
727 */
728 struct xps_dev_maps {
729 struct rcu_head rcu;
730 struct xps_map __rcu *cpu_map[0];
731 };
732 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
733 (nr_cpu_ids * sizeof(struct xps_map *)))
734 #endif /* CONFIG_XPS */
735
736 #define TC_MAX_QUEUE 16
737 #define TC_BITMASK 15
738 /* HW offloaded queuing disciplines txq count and offset maps */
739 struct netdev_tc_txq {
740 u16 count;
741 u16 offset;
742 };
743
744 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
745 /*
746 * This structure is to hold information about the device
747 * configured to run FCoE protocol stack.
748 */
749 struct netdev_fcoe_hbainfo {
750 char manufacturer[64];
751 char serial_number[64];
752 char hardware_version[64];
753 char driver_version[64];
754 char optionrom_version[64];
755 char firmware_version[64];
756 char model[256];
757 char model_description[256];
758 };
759 #endif
760
761 #define MAX_PHYS_ITEM_ID_LEN 32
762
763 /* This structure holds a unique identifier to identify some
764 * physical item (port for example) used by a netdevice.
765 */
766 struct netdev_phys_item_id {
767 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
768 unsigned char id_len;
769 };
770
771 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
772 struct netdev_phys_item_id *b)
773 {
774 return a->id_len == b->id_len &&
775 memcmp(a->id, b->id, a->id_len) == 0;
776 }
777
778 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
779 struct sk_buff *skb);
780
781 /*
782 * This structure defines the management hooks for network devices.
783 * The following hooks can be defined; unless noted otherwise, they are
784 * optional and can be filled with a null pointer.
785 *
786 * int (*ndo_init)(struct net_device *dev);
787 * This function is called once when network device is registered.
788 * The network device can use this to any late stage initializaton
789 * or semantic validattion. It can fail with an error code which will
790 * be propogated back to register_netdev
791 *
792 * void (*ndo_uninit)(struct net_device *dev);
793 * This function is called when device is unregistered or when registration
794 * fails. It is not called if init fails.
795 *
796 * int (*ndo_open)(struct net_device *dev);
797 * This function is called when network device transistions to the up
798 * state.
799 *
800 * int (*ndo_stop)(struct net_device *dev);
801 * This function is called when network device transistions to the down
802 * state.
803 *
804 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
805 * struct net_device *dev);
806 * Called when a packet needs to be transmitted.
807 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
808 * the queue before that can happen; it's for obsolete devices and weird
809 * corner cases, but the stack really does a non-trivial amount
810 * of useless work if you return NETDEV_TX_BUSY.
811 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
812 * Required can not be NULL.
813 *
814 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
815 * netdev_features_t features);
816 * Adjusts the requested feature flags according to device-specific
817 * constraints, and returns the resulting flags. Must not modify
818 * the device state.
819 *
820 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
821 * void *accel_priv, select_queue_fallback_t fallback);
822 * Called to decide which queue to when device supports multiple
823 * transmit queues.
824 *
825 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
826 * This function is called to allow device receiver to make
827 * changes to configuration when multicast or promiscious is enabled.
828 *
829 * void (*ndo_set_rx_mode)(struct net_device *dev);
830 * This function is called device changes address list filtering.
831 * If driver handles unicast address filtering, it should set
832 * IFF_UNICAST_FLT to its priv_flags.
833 *
834 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
835 * This function is called when the Media Access Control address
836 * needs to be changed. If this interface is not defined, the
837 * mac address can not be changed.
838 *
839 * int (*ndo_validate_addr)(struct net_device *dev);
840 * Test if Media Access Control address is valid for the device.
841 *
842 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
843 * Called when a user request an ioctl which can't be handled by
844 * the generic interface code. If not defined ioctl's return
845 * not supported error code.
846 *
847 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
848 * Used to set network devices bus interface parameters. This interface
849 * is retained for legacy reason, new devices should use the bus
850 * interface (PCI) for low level management.
851 *
852 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
853 * Called when a user wants to change the Maximum Transfer Unit
854 * of a device. If not defined, any request to change MTU will
855 * will return an error.
856 *
857 * void (*ndo_tx_timeout)(struct net_device *dev);
858 * Callback uses when the transmitter has not made any progress
859 * for dev->watchdog ticks.
860 *
861 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
862 * struct rtnl_link_stats64 *storage);
863 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
864 * Called when a user wants to get the network device usage
865 * statistics. Drivers must do one of the following:
866 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
867 * rtnl_link_stats64 structure passed by the caller.
868 * 2. Define @ndo_get_stats to update a net_device_stats structure
869 * (which should normally be dev->stats) and return a pointer to
870 * it. The structure may be changed asynchronously only if each
871 * field is written atomically.
872 * 3. Update dev->stats asynchronously and atomically, and define
873 * neither operation.
874 *
875 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
876 * If device support VLAN filtering this function is called when a
877 * VLAN id is registered.
878 *
879 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
880 * If device support VLAN filtering this function is called when a
881 * VLAN id is unregistered.
882 *
883 * void (*ndo_poll_controller)(struct net_device *dev);
884 *
885 * SR-IOV management functions.
886 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
887 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
888 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
889 * int max_tx_rate);
890 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
891 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
892 * int (*ndo_get_vf_config)(struct net_device *dev,
893 * int vf, struct ifla_vf_info *ivf);
894 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
895 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
896 * struct nlattr *port[]);
897 *
898 * Enable or disable the VF ability to query its RSS Redirection Table and
899 * Hash Key. This is needed since on some devices VF share this information
900 * with PF and querying it may adduce a theoretical security risk.
901 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
902 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
903 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
904 * Called to setup 'tc' number of traffic classes in the net device. This
905 * is always called from the stack with the rtnl lock held and netif tx
906 * queues stopped. This allows the netdevice to perform queue management
907 * safely.
908 *
909 * Fiber Channel over Ethernet (FCoE) offload functions.
910 * int (*ndo_fcoe_enable)(struct net_device *dev);
911 * Called when the FCoE protocol stack wants to start using LLD for FCoE
912 * so the underlying device can perform whatever needed configuration or
913 * initialization to support acceleration of FCoE traffic.
914 *
915 * int (*ndo_fcoe_disable)(struct net_device *dev);
916 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
917 * so the underlying device can perform whatever needed clean-ups to
918 * stop supporting acceleration of FCoE traffic.
919 *
920 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
921 * struct scatterlist *sgl, unsigned int sgc);
922 * Called when the FCoE Initiator wants to initialize an I/O that
923 * is a possible candidate for Direct Data Placement (DDP). The LLD can
924 * perform necessary setup and returns 1 to indicate the device is set up
925 * successfully to perform DDP on this I/O, otherwise this returns 0.
926 *
927 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
928 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
929 * indicated by the FC exchange id 'xid', so the underlying device can
930 * clean up and reuse resources for later DDP requests.
931 *
932 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
933 * struct scatterlist *sgl, unsigned int sgc);
934 * Called when the FCoE Target wants to initialize an I/O that
935 * is a possible candidate for Direct Data Placement (DDP). The LLD can
936 * perform necessary setup and returns 1 to indicate the device is set up
937 * successfully to perform DDP on this I/O, otherwise this returns 0.
938 *
939 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
940 * struct netdev_fcoe_hbainfo *hbainfo);
941 * Called when the FCoE Protocol stack wants information on the underlying
942 * device. This information is utilized by the FCoE protocol stack to
943 * register attributes with Fiber Channel management service as per the
944 * FC-GS Fabric Device Management Information(FDMI) specification.
945 *
946 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
947 * Called when the underlying device wants to override default World Wide
948 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
949 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
950 * protocol stack to use.
951 *
952 * RFS acceleration.
953 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
954 * u16 rxq_index, u32 flow_id);
955 * Set hardware filter for RFS. rxq_index is the target queue index;
956 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
957 * Return the filter ID on success, or a negative error code.
958 *
959 * Slave management functions (for bridge, bonding, etc).
960 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
961 * Called to make another netdev an underling.
962 *
963 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
964 * Called to release previously enslaved netdev.
965 *
966 * Feature/offload setting functions.
967 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
968 * Called to update device configuration to new features. Passed
969 * feature set might be less than what was returned by ndo_fix_features()).
970 * Must return >0 or -errno if it changed dev->features itself.
971 *
972 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
973 * struct net_device *dev,
974 * const unsigned char *addr, u16 vid, u16 flags)
975 * Adds an FDB entry to dev for addr.
976 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
977 * struct net_device *dev,
978 * const unsigned char *addr, u16 vid)
979 * Deletes the FDB entry from dev coresponding to addr.
980 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
981 * struct net_device *dev, struct net_device *filter_dev,
982 * int idx)
983 * Used to add FDB entries to dump requests. Implementers should add
984 * entries to skb and update idx with the number of entries.
985 *
986 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
987 * u16 flags)
988 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
989 * struct net_device *dev, u32 filter_mask,
990 * int nlflags)
991 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
992 * u16 flags);
993 *
994 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
995 * Called to change device carrier. Soft-devices (like dummy, team, etc)
996 * which do not represent real hardware may define this to allow their
997 * userspace components to manage their virtual carrier state. Devices
998 * that determine carrier state from physical hardware properties (eg
999 * network cables) or protocol-dependent mechanisms (eg
1000 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1001 *
1002 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1003 * struct netdev_phys_item_id *ppid);
1004 * Called to get ID of physical port of this device. If driver does
1005 * not implement this, it is assumed that the hw is not able to have
1006 * multiple net devices on single physical port.
1007 *
1008 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1009 * sa_family_t sa_family, __be16 port);
1010 * Called by vxlan to notiy a driver about the UDP port and socket
1011 * address family that vxlan is listnening to. It is called only when
1012 * a new port starts listening. The operation is protected by the
1013 * vxlan_net->sock_lock.
1014 *
1015 * void (*ndo_add_geneve_port)(struct net_device *dev,
1016 * sa_family_t sa_family, __be16 port);
1017 * Called by geneve to notify a driver about the UDP port and socket
1018 * address family that geneve is listnening to. It is called only when
1019 * a new port starts listening. The operation is protected by the
1020 * geneve_net->sock_lock.
1021 *
1022 * void (*ndo_del_geneve_port)(struct net_device *dev,
1023 * sa_family_t sa_family, __be16 port);
1024 * Called by geneve to notify the driver about a UDP port and socket
1025 * address family that geneve is not listening to anymore. The operation
1026 * is protected by the geneve_net->sock_lock.
1027 *
1028 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1029 * sa_family_t sa_family, __be16 port);
1030 * Called by vxlan to notify the driver about a UDP port and socket
1031 * address family that vxlan is not listening to anymore. The operation
1032 * is protected by the vxlan_net->sock_lock.
1033 *
1034 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1035 * struct net_device *dev)
1036 * Called by upper layer devices to accelerate switching or other
1037 * station functionality into hardware. 'pdev is the lowerdev
1038 * to use for the offload and 'dev' is the net device that will
1039 * back the offload. Returns a pointer to the private structure
1040 * the upper layer will maintain.
1041 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1042 * Called by upper layer device to delete the station created
1043 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1044 * the station and priv is the structure returned by the add
1045 * operation.
1046 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1047 * struct net_device *dev,
1048 * void *priv);
1049 * Callback to use for xmit over the accelerated station. This
1050 * is used in place of ndo_start_xmit on accelerated net
1051 * devices.
1052 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1053 * struct net_device *dev
1054 * netdev_features_t features);
1055 * Called by core transmit path to determine if device is capable of
1056 * performing offload operations on a given packet. This is to give
1057 * the device an opportunity to implement any restrictions that cannot
1058 * be otherwise expressed by feature flags. The check is called with
1059 * the set of features that the stack has calculated and it returns
1060 * those the driver believes to be appropriate.
1061 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1062 * int queue_index, u32 maxrate);
1063 * Called when a user wants to set a max-rate limitation of specific
1064 * TX queue.
1065 * int (*ndo_get_iflink)(const struct net_device *dev);
1066 * Called to get the iflink value of this device.
1067 * void (*ndo_change_proto_down)(struct net_device *dev,
1068 * bool proto_down);
1069 * This function is used to pass protocol port error state information
1070 * to the switch driver. The switch driver can react to the proto_down
1071 * by doing a phys down on the associated switch port.
1072 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1073 * This function is used to get egress tunnel information for given skb.
1074 * This is useful for retrieving outer tunnel header parameters while
1075 * sampling packet.
1076 *
1077 */
1078 struct net_device_ops {
1079 int (*ndo_init)(struct net_device *dev);
1080 void (*ndo_uninit)(struct net_device *dev);
1081 int (*ndo_open)(struct net_device *dev);
1082 int (*ndo_stop)(struct net_device *dev);
1083 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1084 struct net_device *dev);
1085 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1086 struct net_device *dev,
1087 netdev_features_t features);
1088 u16 (*ndo_select_queue)(struct net_device *dev,
1089 struct sk_buff *skb,
1090 void *accel_priv,
1091 select_queue_fallback_t fallback);
1092 void (*ndo_change_rx_flags)(struct net_device *dev,
1093 int flags);
1094 void (*ndo_set_rx_mode)(struct net_device *dev);
1095 int (*ndo_set_mac_address)(struct net_device *dev,
1096 void *addr);
1097 int (*ndo_validate_addr)(struct net_device *dev);
1098 int (*ndo_do_ioctl)(struct net_device *dev,
1099 struct ifreq *ifr, int cmd);
1100 int (*ndo_set_config)(struct net_device *dev,
1101 struct ifmap *map);
1102 int (*ndo_change_mtu)(struct net_device *dev,
1103 int new_mtu);
1104 int (*ndo_neigh_setup)(struct net_device *dev,
1105 struct neigh_parms *);
1106 void (*ndo_tx_timeout) (struct net_device *dev);
1107
1108 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1109 struct rtnl_link_stats64 *storage);
1110 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1111
1112 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1113 __be16 proto, u16 vid);
1114 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1115 __be16 proto, u16 vid);
1116 #ifdef CONFIG_NET_POLL_CONTROLLER
1117 void (*ndo_poll_controller)(struct net_device *dev);
1118 int (*ndo_netpoll_setup)(struct net_device *dev,
1119 struct netpoll_info *info);
1120 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1121 #endif
1122 #ifdef CONFIG_NET_RX_BUSY_POLL
1123 int (*ndo_busy_poll)(struct napi_struct *dev);
1124 #endif
1125 int (*ndo_set_vf_mac)(struct net_device *dev,
1126 int queue, u8 *mac);
1127 int (*ndo_set_vf_vlan)(struct net_device *dev,
1128 int queue, u16 vlan, u8 qos);
1129 int (*ndo_set_vf_rate)(struct net_device *dev,
1130 int vf, int min_tx_rate,
1131 int max_tx_rate);
1132 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1133 int vf, bool setting);
1134 int (*ndo_set_vf_trust)(struct net_device *dev,
1135 int vf, bool setting);
1136 int (*ndo_get_vf_config)(struct net_device *dev,
1137 int vf,
1138 struct ifla_vf_info *ivf);
1139 int (*ndo_set_vf_link_state)(struct net_device *dev,
1140 int vf, int link_state);
1141 int (*ndo_get_vf_stats)(struct net_device *dev,
1142 int vf,
1143 struct ifla_vf_stats
1144 *vf_stats);
1145 int (*ndo_set_vf_port)(struct net_device *dev,
1146 int vf,
1147 struct nlattr *port[]);
1148 int (*ndo_get_vf_port)(struct net_device *dev,
1149 int vf, struct sk_buff *skb);
1150 int (*ndo_set_vf_rss_query_en)(
1151 struct net_device *dev,
1152 int vf, bool setting);
1153 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1154 #if IS_ENABLED(CONFIG_FCOE)
1155 int (*ndo_fcoe_enable)(struct net_device *dev);
1156 int (*ndo_fcoe_disable)(struct net_device *dev);
1157 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1158 u16 xid,
1159 struct scatterlist *sgl,
1160 unsigned int sgc);
1161 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1162 u16 xid);
1163 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1164 u16 xid,
1165 struct scatterlist *sgl,
1166 unsigned int sgc);
1167 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1168 struct netdev_fcoe_hbainfo *hbainfo);
1169 #endif
1170
1171 #if IS_ENABLED(CONFIG_LIBFCOE)
1172 #define NETDEV_FCOE_WWNN 0
1173 #define NETDEV_FCOE_WWPN 1
1174 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1175 u64 *wwn, int type);
1176 #endif
1177
1178 #ifdef CONFIG_RFS_ACCEL
1179 int (*ndo_rx_flow_steer)(struct net_device *dev,
1180 const struct sk_buff *skb,
1181 u16 rxq_index,
1182 u32 flow_id);
1183 #endif
1184 int (*ndo_add_slave)(struct net_device *dev,
1185 struct net_device *slave_dev);
1186 int (*ndo_del_slave)(struct net_device *dev,
1187 struct net_device *slave_dev);
1188 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1189 netdev_features_t features);
1190 int (*ndo_set_features)(struct net_device *dev,
1191 netdev_features_t features);
1192 int (*ndo_neigh_construct)(struct neighbour *n);
1193 void (*ndo_neigh_destroy)(struct neighbour *n);
1194
1195 int (*ndo_fdb_add)(struct ndmsg *ndm,
1196 struct nlattr *tb[],
1197 struct net_device *dev,
1198 const unsigned char *addr,
1199 u16 vid,
1200 u16 flags);
1201 int (*ndo_fdb_del)(struct ndmsg *ndm,
1202 struct nlattr *tb[],
1203 struct net_device *dev,
1204 const unsigned char *addr,
1205 u16 vid);
1206 int (*ndo_fdb_dump)(struct sk_buff *skb,
1207 struct netlink_callback *cb,
1208 struct net_device *dev,
1209 struct net_device *filter_dev,
1210 int idx);
1211
1212 int (*ndo_bridge_setlink)(struct net_device *dev,
1213 struct nlmsghdr *nlh,
1214 u16 flags);
1215 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1216 u32 pid, u32 seq,
1217 struct net_device *dev,
1218 u32 filter_mask,
1219 int nlflags);
1220 int (*ndo_bridge_dellink)(struct net_device *dev,
1221 struct nlmsghdr *nlh,
1222 u16 flags);
1223 int (*ndo_change_carrier)(struct net_device *dev,
1224 bool new_carrier);
1225 int (*ndo_get_phys_port_id)(struct net_device *dev,
1226 struct netdev_phys_item_id *ppid);
1227 int (*ndo_get_phys_port_name)(struct net_device *dev,
1228 char *name, size_t len);
1229 void (*ndo_add_vxlan_port)(struct net_device *dev,
1230 sa_family_t sa_family,
1231 __be16 port);
1232 void (*ndo_del_vxlan_port)(struct net_device *dev,
1233 sa_family_t sa_family,
1234 __be16 port);
1235 void (*ndo_add_geneve_port)(struct net_device *dev,
1236 sa_family_t sa_family,
1237 __be16 port);
1238 void (*ndo_del_geneve_port)(struct net_device *dev,
1239 sa_family_t sa_family,
1240 __be16 port);
1241 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1242 struct net_device *dev);
1243 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1244 void *priv);
1245
1246 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1247 struct net_device *dev,
1248 void *priv);
1249 int (*ndo_get_lock_subclass)(struct net_device *dev);
1250 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1251 int queue_index,
1252 u32 maxrate);
1253 int (*ndo_get_iflink)(const struct net_device *dev);
1254 int (*ndo_change_proto_down)(struct net_device *dev,
1255 bool proto_down);
1256 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1257 struct sk_buff *skb);
1258 };
1259
1260 /**
1261 * enum net_device_priv_flags - &struct net_device priv_flags
1262 *
1263 * These are the &struct net_device, they are only set internally
1264 * by drivers and used in the kernel. These flags are invisible to
1265 * userspace, this means that the order of these flags can change
1266 * during any kernel release.
1267 *
1268 * You should have a pretty good reason to be extending these flags.
1269 *
1270 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1271 * @IFF_EBRIDGE: Ethernet bridging device
1272 * @IFF_BONDING: bonding master or slave
1273 * @IFF_ISATAP: ISATAP interface (RFC4214)
1274 * @IFF_WAN_HDLC: WAN HDLC device
1275 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1276 * release skb->dst
1277 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1278 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1279 * @IFF_MACVLAN_PORT: device used as macvlan port
1280 * @IFF_BRIDGE_PORT: device used as bridge port
1281 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1282 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1283 * @IFF_UNICAST_FLT: Supports unicast filtering
1284 * @IFF_TEAM_PORT: device used as team port
1285 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1286 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1287 * change when it's running
1288 * @IFF_MACVLAN: Macvlan device
1289 * @IFF_L3MDEV_MASTER: device is an L3 master device
1290 * @IFF_NO_QUEUE: device can run without qdisc attached
1291 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1292 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1293 * @IFF_TEAM: device is a team device
1294 */
1295 enum netdev_priv_flags {
1296 IFF_802_1Q_VLAN = 1<<0,
1297 IFF_EBRIDGE = 1<<1,
1298 IFF_BONDING = 1<<2,
1299 IFF_ISATAP = 1<<3,
1300 IFF_WAN_HDLC = 1<<4,
1301 IFF_XMIT_DST_RELEASE = 1<<5,
1302 IFF_DONT_BRIDGE = 1<<6,
1303 IFF_DISABLE_NETPOLL = 1<<7,
1304 IFF_MACVLAN_PORT = 1<<8,
1305 IFF_BRIDGE_PORT = 1<<9,
1306 IFF_OVS_DATAPATH = 1<<10,
1307 IFF_TX_SKB_SHARING = 1<<11,
1308 IFF_UNICAST_FLT = 1<<12,
1309 IFF_TEAM_PORT = 1<<13,
1310 IFF_SUPP_NOFCS = 1<<14,
1311 IFF_LIVE_ADDR_CHANGE = 1<<15,
1312 IFF_MACVLAN = 1<<16,
1313 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1314 IFF_IPVLAN_MASTER = 1<<18,
1315 IFF_IPVLAN_SLAVE = 1<<19,
1316 IFF_L3MDEV_MASTER = 1<<20,
1317 IFF_NO_QUEUE = 1<<21,
1318 IFF_OPENVSWITCH = 1<<22,
1319 IFF_L3MDEV_SLAVE = 1<<23,
1320 IFF_TEAM = 1<<24,
1321 };
1322
1323 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1324 #define IFF_EBRIDGE IFF_EBRIDGE
1325 #define IFF_BONDING IFF_BONDING
1326 #define IFF_ISATAP IFF_ISATAP
1327 #define IFF_WAN_HDLC IFF_WAN_HDLC
1328 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1329 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1330 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1331 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1332 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1333 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1334 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1335 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1336 #define IFF_TEAM_PORT IFF_TEAM_PORT
1337 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1338 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1339 #define IFF_MACVLAN IFF_MACVLAN
1340 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1341 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1342 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1343 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1344 #define IFF_NO_QUEUE IFF_NO_QUEUE
1345 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1346 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1347 #define IFF_TEAM IFF_TEAM
1348
1349 /**
1350 * struct net_device - The DEVICE structure.
1351 * Actually, this whole structure is a big mistake. It mixes I/O
1352 * data with strictly "high-level" data, and it has to know about
1353 * almost every data structure used in the INET module.
1354 *
1355 * @name: This is the first field of the "visible" part of this structure
1356 * (i.e. as seen by users in the "Space.c" file). It is the name
1357 * of the interface.
1358 *
1359 * @name_hlist: Device name hash chain, please keep it close to name[]
1360 * @ifalias: SNMP alias
1361 * @mem_end: Shared memory end
1362 * @mem_start: Shared memory start
1363 * @base_addr: Device I/O address
1364 * @irq: Device IRQ number
1365 *
1366 * @carrier_changes: Stats to monitor carrier on<->off transitions
1367 *
1368 * @state: Generic network queuing layer state, see netdev_state_t
1369 * @dev_list: The global list of network devices
1370 * @napi_list: List entry, that is used for polling napi devices
1371 * @unreg_list: List entry, that is used, when we are unregistering the
1372 * device, see the function unregister_netdev
1373 * @close_list: List entry, that is used, when we are closing the device
1374 *
1375 * @adj_list: Directly linked devices, like slaves for bonding
1376 * @all_adj_list: All linked devices, *including* neighbours
1377 * @features: Currently active device features
1378 * @hw_features: User-changeable features
1379 *
1380 * @wanted_features: User-requested features
1381 * @vlan_features: Mask of features inheritable by VLAN devices
1382 *
1383 * @hw_enc_features: Mask of features inherited by encapsulating devices
1384 * This field indicates what encapsulation
1385 * offloads the hardware is capable of doing,
1386 * and drivers will need to set them appropriately.
1387 *
1388 * @mpls_features: Mask of features inheritable by MPLS
1389 *
1390 * @ifindex: interface index
1391 * @group: The group, that the device belongs to
1392 *
1393 * @stats: Statistics struct, which was left as a legacy, use
1394 * rtnl_link_stats64 instead
1395 *
1396 * @rx_dropped: Dropped packets by core network,
1397 * do not use this in drivers
1398 * @tx_dropped: Dropped packets by core network,
1399 * do not use this in drivers
1400 * @rx_nohandler: nohandler dropped packets by core network on
1401 * inactive devices, do not use this in drivers
1402 *
1403 * @wireless_handlers: List of functions to handle Wireless Extensions,
1404 * instead of ioctl,
1405 * see <net/iw_handler.h> for details.
1406 * @wireless_data: Instance data managed by the core of wireless extensions
1407 *
1408 * @netdev_ops: Includes several pointers to callbacks,
1409 * if one wants to override the ndo_*() functions
1410 * @ethtool_ops: Management operations
1411 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1412 * of Layer 2 headers.
1413 *
1414 * @flags: Interface flags (a la BSD)
1415 * @priv_flags: Like 'flags' but invisible to userspace,
1416 * see if.h for the definitions
1417 * @gflags: Global flags ( kept as legacy )
1418 * @padded: How much padding added by alloc_netdev()
1419 * @operstate: RFC2863 operstate
1420 * @link_mode: Mapping policy to operstate
1421 * @if_port: Selectable AUI, TP, ...
1422 * @dma: DMA channel
1423 * @mtu: Interface MTU value
1424 * @type: Interface hardware type
1425 * @hard_header_len: Hardware header length, which means that this is the
1426 * minimum size of a packet.
1427 *
1428 * @needed_headroom: Extra headroom the hardware may need, but not in all
1429 * cases can this be guaranteed
1430 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1431 * cases can this be guaranteed. Some cases also use
1432 * LL_MAX_HEADER instead to allocate the skb
1433 *
1434 * interface address info:
1435 *
1436 * @perm_addr: Permanent hw address
1437 * @addr_assign_type: Hw address assignment type
1438 * @addr_len: Hardware address length
1439 * @neigh_priv_len; Used in neigh_alloc(),
1440 * initialized only in atm/clip.c
1441 * @dev_id: Used to differentiate devices that share
1442 * the same link layer address
1443 * @dev_port: Used to differentiate devices that share
1444 * the same function
1445 * @addr_list_lock: XXX: need comments on this one
1446 * @uc_promisc: Counter, that indicates, that promiscuous mode
1447 * has been enabled due to the need to listen to
1448 * additional unicast addresses in a device that
1449 * does not implement ndo_set_rx_mode()
1450 * @uc: unicast mac addresses
1451 * @mc: multicast mac addresses
1452 * @dev_addrs: list of device hw addresses
1453 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1454 * @promiscuity: Number of times, the NIC is told to work in
1455 * Promiscuous mode, if it becomes 0 the NIC will
1456 * exit from working in Promiscuous mode
1457 * @allmulti: Counter, enables or disables allmulticast mode
1458 *
1459 * @vlan_info: VLAN info
1460 * @dsa_ptr: dsa specific data
1461 * @tipc_ptr: TIPC specific data
1462 * @atalk_ptr: AppleTalk link
1463 * @ip_ptr: IPv4 specific data
1464 * @dn_ptr: DECnet specific data
1465 * @ip6_ptr: IPv6 specific data
1466 * @ax25_ptr: AX.25 specific data
1467 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1468 *
1469 * @last_rx: Time of last Rx
1470 * @dev_addr: Hw address (before bcast,
1471 * because most packets are unicast)
1472 *
1473 * @_rx: Array of RX queues
1474 * @num_rx_queues: Number of RX queues
1475 * allocated at register_netdev() time
1476 * @real_num_rx_queues: Number of RX queues currently active in device
1477 *
1478 * @rx_handler: handler for received packets
1479 * @rx_handler_data: XXX: need comments on this one
1480 * @ingress_queue: XXX: need comments on this one
1481 * @broadcast: hw bcast address
1482 *
1483 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1484 * indexed by RX queue number. Assigned by driver.
1485 * This must only be set if the ndo_rx_flow_steer
1486 * operation is defined
1487 * @index_hlist: Device index hash chain
1488 *
1489 * @_tx: Array of TX queues
1490 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1491 * @real_num_tx_queues: Number of TX queues currently active in device
1492 * @qdisc: Root qdisc from userspace point of view
1493 * @tx_queue_len: Max frames per queue allowed
1494 * @tx_global_lock: XXX: need comments on this one
1495 *
1496 * @xps_maps: XXX: need comments on this one
1497 *
1498 * @offload_fwd_mark: Offload device fwding mark
1499 *
1500 * @trans_start: Time (in jiffies) of last Tx
1501 * @watchdog_timeo: Represents the timeout that is used by
1502 * the watchdog ( see dev_watchdog() )
1503 * @watchdog_timer: List of timers
1504 *
1505 * @pcpu_refcnt: Number of references to this device
1506 * @todo_list: Delayed register/unregister
1507 * @link_watch_list: XXX: need comments on this one
1508 *
1509 * @reg_state: Register/unregister state machine
1510 * @dismantle: Device is going to be freed
1511 * @rtnl_link_state: This enum represents the phases of creating
1512 * a new link
1513 *
1514 * @destructor: Called from unregister,
1515 * can be used to call free_netdev
1516 * @npinfo: XXX: need comments on this one
1517 * @nd_net: Network namespace this network device is inside
1518 *
1519 * @ml_priv: Mid-layer private
1520 * @lstats: Loopback statistics
1521 * @tstats: Tunnel statistics
1522 * @dstats: Dummy statistics
1523 * @vstats: Virtual ethernet statistics
1524 *
1525 * @garp_port: GARP
1526 * @mrp_port: MRP
1527 *
1528 * @dev: Class/net/name entry
1529 * @sysfs_groups: Space for optional device, statistics and wireless
1530 * sysfs groups
1531 *
1532 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1533 * @rtnl_link_ops: Rtnl_link_ops
1534 *
1535 * @gso_max_size: Maximum size of generic segmentation offload
1536 * @gso_max_segs: Maximum number of segments that can be passed to the
1537 * NIC for GSO
1538 * @gso_min_segs: Minimum number of segments that can be passed to the
1539 * NIC for GSO
1540 *
1541 * @dcbnl_ops: Data Center Bridging netlink ops
1542 * @num_tc: Number of traffic classes in the net device
1543 * @tc_to_txq: XXX: need comments on this one
1544 * @prio_tc_map XXX: need comments on this one
1545 *
1546 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1547 *
1548 * @priomap: XXX: need comments on this one
1549 * @phydev: Physical device may attach itself
1550 * for hardware timestamping
1551 *
1552 * @qdisc_tx_busylock: XXX: need comments on this one
1553 *
1554 * @proto_down: protocol port state information can be sent to the
1555 * switch driver and used to set the phys state of the
1556 * switch port.
1557 *
1558 * FIXME: cleanup struct net_device such that network protocol info
1559 * moves out.
1560 */
1561
1562 struct net_device {
1563 char name[IFNAMSIZ];
1564 struct hlist_node name_hlist;
1565 char *ifalias;
1566 /*
1567 * I/O specific fields
1568 * FIXME: Merge these and struct ifmap into one
1569 */
1570 unsigned long mem_end;
1571 unsigned long mem_start;
1572 unsigned long base_addr;
1573 int irq;
1574
1575 atomic_t carrier_changes;
1576
1577 /*
1578 * Some hardware also needs these fields (state,dev_list,
1579 * napi_list,unreg_list,close_list) but they are not
1580 * part of the usual set specified in Space.c.
1581 */
1582
1583 unsigned long state;
1584
1585 struct list_head dev_list;
1586 struct list_head napi_list;
1587 struct list_head unreg_list;
1588 struct list_head close_list;
1589 struct list_head ptype_all;
1590 struct list_head ptype_specific;
1591
1592 struct {
1593 struct list_head upper;
1594 struct list_head lower;
1595 } adj_list;
1596
1597 struct {
1598 struct list_head upper;
1599 struct list_head lower;
1600 } all_adj_list;
1601
1602 netdev_features_t features;
1603 netdev_features_t hw_features;
1604 netdev_features_t wanted_features;
1605 netdev_features_t vlan_features;
1606 netdev_features_t hw_enc_features;
1607 netdev_features_t mpls_features;
1608
1609 int ifindex;
1610 int group;
1611
1612 struct net_device_stats stats;
1613
1614 atomic_long_t rx_dropped;
1615 atomic_long_t tx_dropped;
1616 atomic_long_t rx_nohandler;
1617
1618 #ifdef CONFIG_WIRELESS_EXT
1619 const struct iw_handler_def * wireless_handlers;
1620 struct iw_public_data * wireless_data;
1621 #endif
1622 const struct net_device_ops *netdev_ops;
1623 const struct ethtool_ops *ethtool_ops;
1624 #ifdef CONFIG_NET_SWITCHDEV
1625 const struct switchdev_ops *switchdev_ops;
1626 #endif
1627 #ifdef CONFIG_NET_L3_MASTER_DEV
1628 const struct l3mdev_ops *l3mdev_ops;
1629 #endif
1630
1631 const struct header_ops *header_ops;
1632
1633 unsigned int flags;
1634 unsigned int priv_flags;
1635
1636 unsigned short gflags;
1637 unsigned short padded;
1638
1639 unsigned char operstate;
1640 unsigned char link_mode;
1641
1642 unsigned char if_port;
1643 unsigned char dma;
1644
1645 unsigned int mtu;
1646 unsigned short type;
1647 unsigned short hard_header_len;
1648
1649 unsigned short needed_headroom;
1650 unsigned short needed_tailroom;
1651
1652 /* Interface address info. */
1653 unsigned char perm_addr[MAX_ADDR_LEN];
1654 unsigned char addr_assign_type;
1655 unsigned char addr_len;
1656 unsigned short neigh_priv_len;
1657 unsigned short dev_id;
1658 unsigned short dev_port;
1659 spinlock_t addr_list_lock;
1660 unsigned char name_assign_type;
1661 bool uc_promisc;
1662 struct netdev_hw_addr_list uc;
1663 struct netdev_hw_addr_list mc;
1664 struct netdev_hw_addr_list dev_addrs;
1665
1666 #ifdef CONFIG_SYSFS
1667 struct kset *queues_kset;
1668 #endif
1669 unsigned int promiscuity;
1670 unsigned int allmulti;
1671
1672
1673 /* Protocol specific pointers */
1674
1675 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1676 struct vlan_info __rcu *vlan_info;
1677 #endif
1678 #if IS_ENABLED(CONFIG_NET_DSA)
1679 struct dsa_switch_tree *dsa_ptr;
1680 #endif
1681 #if IS_ENABLED(CONFIG_TIPC)
1682 struct tipc_bearer __rcu *tipc_ptr;
1683 #endif
1684 void *atalk_ptr;
1685 struct in_device __rcu *ip_ptr;
1686 struct dn_dev __rcu *dn_ptr;
1687 struct inet6_dev __rcu *ip6_ptr;
1688 void *ax25_ptr;
1689 struct wireless_dev *ieee80211_ptr;
1690 struct wpan_dev *ieee802154_ptr;
1691 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1692 struct mpls_dev __rcu *mpls_ptr;
1693 #endif
1694
1695 /*
1696 * Cache lines mostly used on receive path (including eth_type_trans())
1697 */
1698 unsigned long last_rx;
1699
1700 /* Interface address info used in eth_type_trans() */
1701 unsigned char *dev_addr;
1702
1703
1704 #ifdef CONFIG_SYSFS
1705 struct netdev_rx_queue *_rx;
1706
1707 unsigned int num_rx_queues;
1708 unsigned int real_num_rx_queues;
1709
1710 #endif
1711
1712 unsigned long gro_flush_timeout;
1713 rx_handler_func_t __rcu *rx_handler;
1714 void __rcu *rx_handler_data;
1715
1716 #ifdef CONFIG_NET_CLS_ACT
1717 struct tcf_proto __rcu *ingress_cl_list;
1718 #endif
1719 struct netdev_queue __rcu *ingress_queue;
1720 #ifdef CONFIG_NETFILTER_INGRESS
1721 struct list_head nf_hooks_ingress;
1722 #endif
1723
1724 unsigned char broadcast[MAX_ADDR_LEN];
1725 #ifdef CONFIG_RFS_ACCEL
1726 struct cpu_rmap *rx_cpu_rmap;
1727 #endif
1728 struct hlist_node index_hlist;
1729
1730 /*
1731 * Cache lines mostly used on transmit path
1732 */
1733 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1734 unsigned int num_tx_queues;
1735 unsigned int real_num_tx_queues;
1736 struct Qdisc *qdisc;
1737 unsigned long tx_queue_len;
1738 spinlock_t tx_global_lock;
1739 int watchdog_timeo;
1740
1741 #ifdef CONFIG_XPS
1742 struct xps_dev_maps __rcu *xps_maps;
1743 #endif
1744 #ifdef CONFIG_NET_CLS_ACT
1745 struct tcf_proto __rcu *egress_cl_list;
1746 #endif
1747 #ifdef CONFIG_NET_SWITCHDEV
1748 u32 offload_fwd_mark;
1749 #endif
1750
1751 /* These may be needed for future network-power-down code. */
1752
1753 /*
1754 * trans_start here is expensive for high speed devices on SMP,
1755 * please use netdev_queue->trans_start instead.
1756 */
1757 unsigned long trans_start;
1758
1759 struct timer_list watchdog_timer;
1760
1761 int __percpu *pcpu_refcnt;
1762 struct list_head todo_list;
1763
1764 struct list_head link_watch_list;
1765
1766 enum { NETREG_UNINITIALIZED=0,
1767 NETREG_REGISTERED, /* completed register_netdevice */
1768 NETREG_UNREGISTERING, /* called unregister_netdevice */
1769 NETREG_UNREGISTERED, /* completed unregister todo */
1770 NETREG_RELEASED, /* called free_netdev */
1771 NETREG_DUMMY, /* dummy device for NAPI poll */
1772 } reg_state:8;
1773
1774 bool dismantle;
1775
1776 enum {
1777 RTNL_LINK_INITIALIZED,
1778 RTNL_LINK_INITIALIZING,
1779 } rtnl_link_state:16;
1780
1781 void (*destructor)(struct net_device *dev);
1782
1783 #ifdef CONFIG_NETPOLL
1784 struct netpoll_info __rcu *npinfo;
1785 #endif
1786
1787 possible_net_t nd_net;
1788
1789 /* mid-layer private */
1790 union {
1791 void *ml_priv;
1792 struct pcpu_lstats __percpu *lstats;
1793 struct pcpu_sw_netstats __percpu *tstats;
1794 struct pcpu_dstats __percpu *dstats;
1795 struct pcpu_vstats __percpu *vstats;
1796 };
1797
1798 struct garp_port __rcu *garp_port;
1799 struct mrp_port __rcu *mrp_port;
1800
1801 struct device dev;
1802 const struct attribute_group *sysfs_groups[4];
1803 const struct attribute_group *sysfs_rx_queue_group;
1804
1805 const struct rtnl_link_ops *rtnl_link_ops;
1806
1807 /* for setting kernel sock attribute on TCP connection setup */
1808 #define GSO_MAX_SIZE 65536
1809 unsigned int gso_max_size;
1810 #define GSO_MAX_SEGS 65535
1811 u16 gso_max_segs;
1812 u16 gso_min_segs;
1813 #ifdef CONFIG_DCB
1814 const struct dcbnl_rtnl_ops *dcbnl_ops;
1815 #endif
1816 u8 num_tc;
1817 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1818 u8 prio_tc_map[TC_BITMASK + 1];
1819
1820 #if IS_ENABLED(CONFIG_FCOE)
1821 unsigned int fcoe_ddp_xid;
1822 #endif
1823 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1824 struct netprio_map __rcu *priomap;
1825 #endif
1826 struct phy_device *phydev;
1827 struct lock_class_key *qdisc_tx_busylock;
1828 bool proto_down;
1829 };
1830 #define to_net_dev(d) container_of(d, struct net_device, dev)
1831
1832 #define NETDEV_ALIGN 32
1833
1834 static inline
1835 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1836 {
1837 return dev->prio_tc_map[prio & TC_BITMASK];
1838 }
1839
1840 static inline
1841 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1842 {
1843 if (tc >= dev->num_tc)
1844 return -EINVAL;
1845
1846 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1847 return 0;
1848 }
1849
1850 static inline
1851 void netdev_reset_tc(struct net_device *dev)
1852 {
1853 dev->num_tc = 0;
1854 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1855 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1856 }
1857
1858 static inline
1859 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1860 {
1861 if (tc >= dev->num_tc)
1862 return -EINVAL;
1863
1864 dev->tc_to_txq[tc].count = count;
1865 dev->tc_to_txq[tc].offset = offset;
1866 return 0;
1867 }
1868
1869 static inline
1870 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1871 {
1872 if (num_tc > TC_MAX_QUEUE)
1873 return -EINVAL;
1874
1875 dev->num_tc = num_tc;
1876 return 0;
1877 }
1878
1879 static inline
1880 int netdev_get_num_tc(struct net_device *dev)
1881 {
1882 return dev->num_tc;
1883 }
1884
1885 static inline
1886 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1887 unsigned int index)
1888 {
1889 return &dev->_tx[index];
1890 }
1891
1892 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1893 const struct sk_buff *skb)
1894 {
1895 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1896 }
1897
1898 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1899 void (*f)(struct net_device *,
1900 struct netdev_queue *,
1901 void *),
1902 void *arg)
1903 {
1904 unsigned int i;
1905
1906 for (i = 0; i < dev->num_tx_queues; i++)
1907 f(dev, &dev->_tx[i], arg);
1908 }
1909
1910 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1911 struct sk_buff *skb,
1912 void *accel_priv);
1913
1914 /*
1915 * Net namespace inlines
1916 */
1917 static inline
1918 struct net *dev_net(const struct net_device *dev)
1919 {
1920 return read_pnet(&dev->nd_net);
1921 }
1922
1923 static inline
1924 void dev_net_set(struct net_device *dev, struct net *net)
1925 {
1926 write_pnet(&dev->nd_net, net);
1927 }
1928
1929 static inline bool netdev_uses_dsa(struct net_device *dev)
1930 {
1931 #if IS_ENABLED(CONFIG_NET_DSA)
1932 if (dev->dsa_ptr != NULL)
1933 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1934 #endif
1935 return false;
1936 }
1937
1938 /**
1939 * netdev_priv - access network device private data
1940 * @dev: network device
1941 *
1942 * Get network device private data
1943 */
1944 static inline void *netdev_priv(const struct net_device *dev)
1945 {
1946 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1947 }
1948
1949 /* Set the sysfs physical device reference for the network logical device
1950 * if set prior to registration will cause a symlink during initialization.
1951 */
1952 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1953
1954 /* Set the sysfs device type for the network logical device to allow
1955 * fine-grained identification of different network device types. For
1956 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1957 */
1958 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1959
1960 /* Default NAPI poll() weight
1961 * Device drivers are strongly advised to not use bigger value
1962 */
1963 #define NAPI_POLL_WEIGHT 64
1964
1965 /**
1966 * netif_napi_add - initialize a napi context
1967 * @dev: network device
1968 * @napi: napi context
1969 * @poll: polling function
1970 * @weight: default weight
1971 *
1972 * netif_napi_add() must be used to initialize a napi context prior to calling
1973 * *any* of the other napi related functions.
1974 */
1975 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1976 int (*poll)(struct napi_struct *, int), int weight);
1977
1978 /**
1979 * netif_tx_napi_add - initialize a napi context
1980 * @dev: network device
1981 * @napi: napi context
1982 * @poll: polling function
1983 * @weight: default weight
1984 *
1985 * This variant of netif_napi_add() should be used from drivers using NAPI
1986 * to exclusively poll a TX queue.
1987 * This will avoid we add it into napi_hash[], thus polluting this hash table.
1988 */
1989 static inline void netif_tx_napi_add(struct net_device *dev,
1990 struct napi_struct *napi,
1991 int (*poll)(struct napi_struct *, int),
1992 int weight)
1993 {
1994 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
1995 netif_napi_add(dev, napi, poll, weight);
1996 }
1997
1998 /**
1999 * netif_napi_del - remove a napi context
2000 * @napi: napi context
2001 *
2002 * netif_napi_del() removes a napi context from the network device napi list
2003 */
2004 void netif_napi_del(struct napi_struct *napi);
2005
2006 struct napi_gro_cb {
2007 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2008 void *frag0;
2009
2010 /* Length of frag0. */
2011 unsigned int frag0_len;
2012
2013 /* This indicates where we are processing relative to skb->data. */
2014 int data_offset;
2015
2016 /* This is non-zero if the packet cannot be merged with the new skb. */
2017 u16 flush;
2018
2019 /* Save the IP ID here and check when we get to the transport layer */
2020 u16 flush_id;
2021
2022 /* Number of segments aggregated. */
2023 u16 count;
2024
2025 /* Start offset for remote checksum offload */
2026 u16 gro_remcsum_start;
2027
2028 /* jiffies when first packet was created/queued */
2029 unsigned long age;
2030
2031 /* Used in ipv6_gro_receive() and foo-over-udp */
2032 u16 proto;
2033
2034 /* This is non-zero if the packet may be of the same flow. */
2035 u8 same_flow:1;
2036
2037 /* Used in udp_gro_receive */
2038 u8 udp_mark:1;
2039
2040 /* GRO checksum is valid */
2041 u8 csum_valid:1;
2042
2043 /* Number of checksums via CHECKSUM_UNNECESSARY */
2044 u8 csum_cnt:3;
2045
2046 /* Free the skb? */
2047 u8 free:2;
2048 #define NAPI_GRO_FREE 1
2049 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2050
2051 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2052 u8 is_ipv6:1;
2053
2054 /* 7 bit hole */
2055
2056 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2057 __wsum csum;
2058
2059 /* used in skb_gro_receive() slow path */
2060 struct sk_buff *last;
2061 };
2062
2063 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2064
2065 struct packet_type {
2066 __be16 type; /* This is really htons(ether_type). */
2067 struct net_device *dev; /* NULL is wildcarded here */
2068 int (*func) (struct sk_buff *,
2069 struct net_device *,
2070 struct packet_type *,
2071 struct net_device *);
2072 bool (*id_match)(struct packet_type *ptype,
2073 struct sock *sk);
2074 void *af_packet_priv;
2075 struct list_head list;
2076 };
2077
2078 struct offload_callbacks {
2079 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2080 netdev_features_t features);
2081 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2082 struct sk_buff *skb);
2083 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2084 };
2085
2086 struct packet_offload {
2087 __be16 type; /* This is really htons(ether_type). */
2088 u16 priority;
2089 struct offload_callbacks callbacks;
2090 struct list_head list;
2091 };
2092
2093 struct udp_offload;
2094
2095 struct udp_offload_callbacks {
2096 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2097 struct sk_buff *skb,
2098 struct udp_offload *uoff);
2099 int (*gro_complete)(struct sk_buff *skb,
2100 int nhoff,
2101 struct udp_offload *uoff);
2102 };
2103
2104 struct udp_offload {
2105 __be16 port;
2106 u8 ipproto;
2107 struct udp_offload_callbacks callbacks;
2108 };
2109
2110 /* often modified stats are per cpu, other are shared (netdev->stats) */
2111 struct pcpu_sw_netstats {
2112 u64 rx_packets;
2113 u64 rx_bytes;
2114 u64 tx_packets;
2115 u64 tx_bytes;
2116 struct u64_stats_sync syncp;
2117 };
2118
2119 #define __netdev_alloc_pcpu_stats(type, gfp) \
2120 ({ \
2121 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2122 if (pcpu_stats) { \
2123 int __cpu; \
2124 for_each_possible_cpu(__cpu) { \
2125 typeof(type) *stat; \
2126 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2127 u64_stats_init(&stat->syncp); \
2128 } \
2129 } \
2130 pcpu_stats; \
2131 })
2132
2133 #define netdev_alloc_pcpu_stats(type) \
2134 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2135
2136 enum netdev_lag_tx_type {
2137 NETDEV_LAG_TX_TYPE_UNKNOWN,
2138 NETDEV_LAG_TX_TYPE_RANDOM,
2139 NETDEV_LAG_TX_TYPE_BROADCAST,
2140 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2141 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2142 NETDEV_LAG_TX_TYPE_HASH,
2143 };
2144
2145 struct netdev_lag_upper_info {
2146 enum netdev_lag_tx_type tx_type;
2147 };
2148
2149 struct netdev_lag_lower_state_info {
2150 u8 link_up : 1,
2151 tx_enabled : 1;
2152 };
2153
2154 #include <linux/notifier.h>
2155
2156 /* netdevice notifier chain. Please remember to update the rtnetlink
2157 * notification exclusion list in rtnetlink_event() when adding new
2158 * types.
2159 */
2160 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2161 #define NETDEV_DOWN 0x0002
2162 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2163 detected a hardware crash and restarted
2164 - we can use this eg to kick tcp sessions
2165 once done */
2166 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2167 #define NETDEV_REGISTER 0x0005
2168 #define NETDEV_UNREGISTER 0x0006
2169 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2170 #define NETDEV_CHANGEADDR 0x0008
2171 #define NETDEV_GOING_DOWN 0x0009
2172 #define NETDEV_CHANGENAME 0x000A
2173 #define NETDEV_FEAT_CHANGE 0x000B
2174 #define NETDEV_BONDING_FAILOVER 0x000C
2175 #define NETDEV_PRE_UP 0x000D
2176 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2177 #define NETDEV_POST_TYPE_CHANGE 0x000F
2178 #define NETDEV_POST_INIT 0x0010
2179 #define NETDEV_UNREGISTER_FINAL 0x0011
2180 #define NETDEV_RELEASE 0x0012
2181 #define NETDEV_NOTIFY_PEERS 0x0013
2182 #define NETDEV_JOIN 0x0014
2183 #define NETDEV_CHANGEUPPER 0x0015
2184 #define NETDEV_RESEND_IGMP 0x0016
2185 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2186 #define NETDEV_CHANGEINFODATA 0x0018
2187 #define NETDEV_BONDING_INFO 0x0019
2188 #define NETDEV_PRECHANGEUPPER 0x001A
2189 #define NETDEV_CHANGELOWERSTATE 0x001B
2190
2191 int register_netdevice_notifier(struct notifier_block *nb);
2192 int unregister_netdevice_notifier(struct notifier_block *nb);
2193
2194 struct netdev_notifier_info {
2195 struct net_device *dev;
2196 };
2197
2198 struct netdev_notifier_change_info {
2199 struct netdev_notifier_info info; /* must be first */
2200 unsigned int flags_changed;
2201 };
2202
2203 struct netdev_notifier_changeupper_info {
2204 struct netdev_notifier_info info; /* must be first */
2205 struct net_device *upper_dev; /* new upper dev */
2206 bool master; /* is upper dev master */
2207 bool linking; /* is the nofication for link or unlink */
2208 void *upper_info; /* upper dev info */
2209 };
2210
2211 struct netdev_notifier_changelowerstate_info {
2212 struct netdev_notifier_info info; /* must be first */
2213 void *lower_state_info; /* is lower dev state */
2214 };
2215
2216 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2217 struct net_device *dev)
2218 {
2219 info->dev = dev;
2220 }
2221
2222 static inline struct net_device *
2223 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2224 {
2225 return info->dev;
2226 }
2227
2228 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2229
2230
2231 extern rwlock_t dev_base_lock; /* Device list lock */
2232
2233 #define for_each_netdev(net, d) \
2234 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2235 #define for_each_netdev_reverse(net, d) \
2236 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2237 #define for_each_netdev_rcu(net, d) \
2238 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2239 #define for_each_netdev_safe(net, d, n) \
2240 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2241 #define for_each_netdev_continue(net, d) \
2242 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2243 #define for_each_netdev_continue_rcu(net, d) \
2244 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2245 #define for_each_netdev_in_bond_rcu(bond, slave) \
2246 for_each_netdev_rcu(&init_net, slave) \
2247 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2248 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2249
2250 static inline struct net_device *next_net_device(struct net_device *dev)
2251 {
2252 struct list_head *lh;
2253 struct net *net;
2254
2255 net = dev_net(dev);
2256 lh = dev->dev_list.next;
2257 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2258 }
2259
2260 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2261 {
2262 struct list_head *lh;
2263 struct net *net;
2264
2265 net = dev_net(dev);
2266 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2267 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2268 }
2269
2270 static inline struct net_device *first_net_device(struct net *net)
2271 {
2272 return list_empty(&net->dev_base_head) ? NULL :
2273 net_device_entry(net->dev_base_head.next);
2274 }
2275
2276 static inline struct net_device *first_net_device_rcu(struct net *net)
2277 {
2278 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2279
2280 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2281 }
2282
2283 int netdev_boot_setup_check(struct net_device *dev);
2284 unsigned long netdev_boot_base(const char *prefix, int unit);
2285 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2286 const char *hwaddr);
2287 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2288 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2289 void dev_add_pack(struct packet_type *pt);
2290 void dev_remove_pack(struct packet_type *pt);
2291 void __dev_remove_pack(struct packet_type *pt);
2292 void dev_add_offload(struct packet_offload *po);
2293 void dev_remove_offload(struct packet_offload *po);
2294
2295 int dev_get_iflink(const struct net_device *dev);
2296 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2297 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2298 unsigned short mask);
2299 struct net_device *dev_get_by_name(struct net *net, const char *name);
2300 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2301 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2302 int dev_alloc_name(struct net_device *dev, const char *name);
2303 int dev_open(struct net_device *dev);
2304 int dev_close(struct net_device *dev);
2305 int dev_close_many(struct list_head *head, bool unlink);
2306 void dev_disable_lro(struct net_device *dev);
2307 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2308 int dev_queue_xmit(struct sk_buff *skb);
2309 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2310 int register_netdevice(struct net_device *dev);
2311 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2312 void unregister_netdevice_many(struct list_head *head);
2313 static inline void unregister_netdevice(struct net_device *dev)
2314 {
2315 unregister_netdevice_queue(dev, NULL);
2316 }
2317
2318 int netdev_refcnt_read(const struct net_device *dev);
2319 void free_netdev(struct net_device *dev);
2320 void netdev_freemem(struct net_device *dev);
2321 void synchronize_net(void);
2322 int init_dummy_netdev(struct net_device *dev);
2323
2324 DECLARE_PER_CPU(int, xmit_recursion);
2325 static inline int dev_recursion_level(void)
2326 {
2327 return this_cpu_read(xmit_recursion);
2328 }
2329
2330 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2331 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2332 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2333 int netdev_get_name(struct net *net, char *name, int ifindex);
2334 int dev_restart(struct net_device *dev);
2335 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2336
2337 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2338 {
2339 return NAPI_GRO_CB(skb)->data_offset;
2340 }
2341
2342 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2343 {
2344 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2345 }
2346
2347 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2348 {
2349 NAPI_GRO_CB(skb)->data_offset += len;
2350 }
2351
2352 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2353 unsigned int offset)
2354 {
2355 return NAPI_GRO_CB(skb)->frag0 + offset;
2356 }
2357
2358 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2359 {
2360 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2361 }
2362
2363 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2364 unsigned int offset)
2365 {
2366 if (!pskb_may_pull(skb, hlen))
2367 return NULL;
2368
2369 NAPI_GRO_CB(skb)->frag0 = NULL;
2370 NAPI_GRO_CB(skb)->frag0_len = 0;
2371 return skb->data + offset;
2372 }
2373
2374 static inline void *skb_gro_network_header(struct sk_buff *skb)
2375 {
2376 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2377 skb_network_offset(skb);
2378 }
2379
2380 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2381 const void *start, unsigned int len)
2382 {
2383 if (NAPI_GRO_CB(skb)->csum_valid)
2384 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2385 csum_partial(start, len, 0));
2386 }
2387
2388 /* GRO checksum functions. These are logical equivalents of the normal
2389 * checksum functions (in skbuff.h) except that they operate on the GRO
2390 * offsets and fields in sk_buff.
2391 */
2392
2393 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2394
2395 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2396 {
2397 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2398 }
2399
2400 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2401 bool zero_okay,
2402 __sum16 check)
2403 {
2404 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2405 skb_checksum_start_offset(skb) <
2406 skb_gro_offset(skb)) &&
2407 !skb_at_gro_remcsum_start(skb) &&
2408 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2409 (!zero_okay || check));
2410 }
2411
2412 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2413 __wsum psum)
2414 {
2415 if (NAPI_GRO_CB(skb)->csum_valid &&
2416 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2417 return 0;
2418
2419 NAPI_GRO_CB(skb)->csum = psum;
2420
2421 return __skb_gro_checksum_complete(skb);
2422 }
2423
2424 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2425 {
2426 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2427 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2428 NAPI_GRO_CB(skb)->csum_cnt--;
2429 } else {
2430 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2431 * verified a new top level checksum or an encapsulated one
2432 * during GRO. This saves work if we fallback to normal path.
2433 */
2434 __skb_incr_checksum_unnecessary(skb);
2435 }
2436 }
2437
2438 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2439 compute_pseudo) \
2440 ({ \
2441 __sum16 __ret = 0; \
2442 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2443 __ret = __skb_gro_checksum_validate_complete(skb, \
2444 compute_pseudo(skb, proto)); \
2445 if (__ret) \
2446 __skb_mark_checksum_bad(skb); \
2447 else \
2448 skb_gro_incr_csum_unnecessary(skb); \
2449 __ret; \
2450 })
2451
2452 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2453 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2454
2455 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2456 compute_pseudo) \
2457 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2458
2459 #define skb_gro_checksum_simple_validate(skb) \
2460 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2461
2462 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2463 {
2464 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2465 !NAPI_GRO_CB(skb)->csum_valid);
2466 }
2467
2468 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2469 __sum16 check, __wsum pseudo)
2470 {
2471 NAPI_GRO_CB(skb)->csum = ~pseudo;
2472 NAPI_GRO_CB(skb)->csum_valid = 1;
2473 }
2474
2475 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2476 do { \
2477 if (__skb_gro_checksum_convert_check(skb)) \
2478 __skb_gro_checksum_convert(skb, check, \
2479 compute_pseudo(skb, proto)); \
2480 } while (0)
2481
2482 struct gro_remcsum {
2483 int offset;
2484 __wsum delta;
2485 };
2486
2487 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2488 {
2489 grc->offset = 0;
2490 grc->delta = 0;
2491 }
2492
2493 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2494 unsigned int off, size_t hdrlen,
2495 int start, int offset,
2496 struct gro_remcsum *grc,
2497 bool nopartial)
2498 {
2499 __wsum delta;
2500 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2501
2502 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2503
2504 if (!nopartial) {
2505 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2506 return ptr;
2507 }
2508
2509 ptr = skb_gro_header_fast(skb, off);
2510 if (skb_gro_header_hard(skb, off + plen)) {
2511 ptr = skb_gro_header_slow(skb, off + plen, off);
2512 if (!ptr)
2513 return NULL;
2514 }
2515
2516 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2517 start, offset);
2518
2519 /* Adjust skb->csum since we changed the packet */
2520 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2521
2522 grc->offset = off + hdrlen + offset;
2523 grc->delta = delta;
2524
2525 return ptr;
2526 }
2527
2528 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2529 struct gro_remcsum *grc)
2530 {
2531 void *ptr;
2532 size_t plen = grc->offset + sizeof(u16);
2533
2534 if (!grc->delta)
2535 return;
2536
2537 ptr = skb_gro_header_fast(skb, grc->offset);
2538 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2539 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2540 if (!ptr)
2541 return;
2542 }
2543
2544 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2545 }
2546
2547 struct skb_csum_offl_spec {
2548 __u16 ipv4_okay:1,
2549 ipv6_okay:1,
2550 encap_okay:1,
2551 ip_options_okay:1,
2552 ext_hdrs_okay:1,
2553 tcp_okay:1,
2554 udp_okay:1,
2555 sctp_okay:1,
2556 vlan_okay:1,
2557 no_encapped_ipv6:1,
2558 no_not_encapped:1;
2559 };
2560
2561 bool __skb_csum_offload_chk(struct sk_buff *skb,
2562 const struct skb_csum_offl_spec *spec,
2563 bool *csum_encapped,
2564 bool csum_help);
2565
2566 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2567 const struct skb_csum_offl_spec *spec,
2568 bool *csum_encapped,
2569 bool csum_help)
2570 {
2571 if (skb->ip_summed != CHECKSUM_PARTIAL)
2572 return false;
2573
2574 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2575 }
2576
2577 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2578 const struct skb_csum_offl_spec *spec)
2579 {
2580 bool csum_encapped;
2581
2582 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2583 }
2584
2585 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2586 {
2587 static const struct skb_csum_offl_spec csum_offl_spec = {
2588 .ipv4_okay = 1,
2589 .ip_options_okay = 1,
2590 .ipv6_okay = 1,
2591 .vlan_okay = 1,
2592 .tcp_okay = 1,
2593 .udp_okay = 1,
2594 };
2595
2596 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2597 }
2598
2599 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2600 {
2601 static const struct skb_csum_offl_spec csum_offl_spec = {
2602 .ipv4_okay = 1,
2603 .ip_options_okay = 1,
2604 .tcp_okay = 1,
2605 .udp_okay = 1,
2606 .vlan_okay = 1,
2607 };
2608
2609 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2610 }
2611
2612 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2613 unsigned short type,
2614 const void *daddr, const void *saddr,
2615 unsigned int len)
2616 {
2617 if (!dev->header_ops || !dev->header_ops->create)
2618 return 0;
2619
2620 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2621 }
2622
2623 static inline int dev_parse_header(const struct sk_buff *skb,
2624 unsigned char *haddr)
2625 {
2626 const struct net_device *dev = skb->dev;
2627
2628 if (!dev->header_ops || !dev->header_ops->parse)
2629 return 0;
2630 return dev->header_ops->parse(skb, haddr);
2631 }
2632
2633 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2634 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2635 static inline int unregister_gifconf(unsigned int family)
2636 {
2637 return register_gifconf(family, NULL);
2638 }
2639
2640 #ifdef CONFIG_NET_FLOW_LIMIT
2641 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2642 struct sd_flow_limit {
2643 u64 count;
2644 unsigned int num_buckets;
2645 unsigned int history_head;
2646 u16 history[FLOW_LIMIT_HISTORY];
2647 u8 buckets[];
2648 };
2649
2650 extern int netdev_flow_limit_table_len;
2651 #endif /* CONFIG_NET_FLOW_LIMIT */
2652
2653 /*
2654 * Incoming packets are placed on per-cpu queues
2655 */
2656 struct softnet_data {
2657 struct list_head poll_list;
2658 struct sk_buff_head process_queue;
2659
2660 /* stats */
2661 unsigned int processed;
2662 unsigned int time_squeeze;
2663 unsigned int cpu_collision;
2664 unsigned int received_rps;
2665 #ifdef CONFIG_RPS
2666 struct softnet_data *rps_ipi_list;
2667 #endif
2668 #ifdef CONFIG_NET_FLOW_LIMIT
2669 struct sd_flow_limit __rcu *flow_limit;
2670 #endif
2671 struct Qdisc *output_queue;
2672 struct Qdisc **output_queue_tailp;
2673 struct sk_buff *completion_queue;
2674
2675 #ifdef CONFIG_RPS
2676 /* Elements below can be accessed between CPUs for RPS */
2677 struct call_single_data csd ____cacheline_aligned_in_smp;
2678 struct softnet_data *rps_ipi_next;
2679 unsigned int cpu;
2680 unsigned int input_queue_head;
2681 unsigned int input_queue_tail;
2682 #endif
2683 unsigned int dropped;
2684 struct sk_buff_head input_pkt_queue;
2685 struct napi_struct backlog;
2686
2687 };
2688
2689 static inline void input_queue_head_incr(struct softnet_data *sd)
2690 {
2691 #ifdef CONFIG_RPS
2692 sd->input_queue_head++;
2693 #endif
2694 }
2695
2696 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2697 unsigned int *qtail)
2698 {
2699 #ifdef CONFIG_RPS
2700 *qtail = ++sd->input_queue_tail;
2701 #endif
2702 }
2703
2704 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2705
2706 void __netif_schedule(struct Qdisc *q);
2707 void netif_schedule_queue(struct netdev_queue *txq);
2708
2709 static inline void netif_tx_schedule_all(struct net_device *dev)
2710 {
2711 unsigned int i;
2712
2713 for (i = 0; i < dev->num_tx_queues; i++)
2714 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2715 }
2716
2717 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2718 {
2719 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2720 }
2721
2722 /**
2723 * netif_start_queue - allow transmit
2724 * @dev: network device
2725 *
2726 * Allow upper layers to call the device hard_start_xmit routine.
2727 */
2728 static inline void netif_start_queue(struct net_device *dev)
2729 {
2730 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2731 }
2732
2733 static inline void netif_tx_start_all_queues(struct net_device *dev)
2734 {
2735 unsigned int i;
2736
2737 for (i = 0; i < dev->num_tx_queues; i++) {
2738 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2739 netif_tx_start_queue(txq);
2740 }
2741 }
2742
2743 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2744
2745 /**
2746 * netif_wake_queue - restart transmit
2747 * @dev: network device
2748 *
2749 * Allow upper layers to call the device hard_start_xmit routine.
2750 * Used for flow control when transmit resources are available.
2751 */
2752 static inline void netif_wake_queue(struct net_device *dev)
2753 {
2754 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2755 }
2756
2757 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2758 {
2759 unsigned int i;
2760
2761 for (i = 0; i < dev->num_tx_queues; i++) {
2762 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2763 netif_tx_wake_queue(txq);
2764 }
2765 }
2766
2767 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2768 {
2769 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2770 }
2771
2772 /**
2773 * netif_stop_queue - stop transmitted packets
2774 * @dev: network device
2775 *
2776 * Stop upper layers calling the device hard_start_xmit routine.
2777 * Used for flow control when transmit resources are unavailable.
2778 */
2779 static inline void netif_stop_queue(struct net_device *dev)
2780 {
2781 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2782 }
2783
2784 void netif_tx_stop_all_queues(struct net_device *dev);
2785
2786 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2787 {
2788 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2789 }
2790
2791 /**
2792 * netif_queue_stopped - test if transmit queue is flowblocked
2793 * @dev: network device
2794 *
2795 * Test if transmit queue on device is currently unable to send.
2796 */
2797 static inline bool netif_queue_stopped(const struct net_device *dev)
2798 {
2799 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2800 }
2801
2802 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2803 {
2804 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2805 }
2806
2807 static inline bool
2808 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2809 {
2810 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2811 }
2812
2813 static inline bool
2814 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2815 {
2816 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2817 }
2818
2819 /**
2820 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2821 * @dev_queue: pointer to transmit queue
2822 *
2823 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2824 * to give appropriate hint to the cpu.
2825 */
2826 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2827 {
2828 #ifdef CONFIG_BQL
2829 prefetchw(&dev_queue->dql.num_queued);
2830 #endif
2831 }
2832
2833 /**
2834 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2835 * @dev_queue: pointer to transmit queue
2836 *
2837 * BQL enabled drivers might use this helper in their TX completion path,
2838 * to give appropriate hint to the cpu.
2839 */
2840 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2841 {
2842 #ifdef CONFIG_BQL
2843 prefetchw(&dev_queue->dql.limit);
2844 #endif
2845 }
2846
2847 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2848 unsigned int bytes)
2849 {
2850 #ifdef CONFIG_BQL
2851 dql_queued(&dev_queue->dql, bytes);
2852
2853 if (likely(dql_avail(&dev_queue->dql) >= 0))
2854 return;
2855
2856 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2857
2858 /*
2859 * The XOFF flag must be set before checking the dql_avail below,
2860 * because in netdev_tx_completed_queue we update the dql_completed
2861 * before checking the XOFF flag.
2862 */
2863 smp_mb();
2864
2865 /* check again in case another CPU has just made room avail */
2866 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2867 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2868 #endif
2869 }
2870
2871 /**
2872 * netdev_sent_queue - report the number of bytes queued to hardware
2873 * @dev: network device
2874 * @bytes: number of bytes queued to the hardware device queue
2875 *
2876 * Report the number of bytes queued for sending/completion to the network
2877 * device hardware queue. @bytes should be a good approximation and should
2878 * exactly match netdev_completed_queue() @bytes
2879 */
2880 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2881 {
2882 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2883 }
2884
2885 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2886 unsigned int pkts, unsigned int bytes)
2887 {
2888 #ifdef CONFIG_BQL
2889 if (unlikely(!bytes))
2890 return;
2891
2892 dql_completed(&dev_queue->dql, bytes);
2893
2894 /*
2895 * Without the memory barrier there is a small possiblity that
2896 * netdev_tx_sent_queue will miss the update and cause the queue to
2897 * be stopped forever
2898 */
2899 smp_mb();
2900
2901 if (dql_avail(&dev_queue->dql) < 0)
2902 return;
2903
2904 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2905 netif_schedule_queue(dev_queue);
2906 #endif
2907 }
2908
2909 /**
2910 * netdev_completed_queue - report bytes and packets completed by device
2911 * @dev: network device
2912 * @pkts: actual number of packets sent over the medium
2913 * @bytes: actual number of bytes sent over the medium
2914 *
2915 * Report the number of bytes and packets transmitted by the network device
2916 * hardware queue over the physical medium, @bytes must exactly match the
2917 * @bytes amount passed to netdev_sent_queue()
2918 */
2919 static inline void netdev_completed_queue(struct net_device *dev,
2920 unsigned int pkts, unsigned int bytes)
2921 {
2922 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2923 }
2924
2925 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2926 {
2927 #ifdef CONFIG_BQL
2928 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2929 dql_reset(&q->dql);
2930 #endif
2931 }
2932
2933 /**
2934 * netdev_reset_queue - reset the packets and bytes count of a network device
2935 * @dev_queue: network device
2936 *
2937 * Reset the bytes and packet count of a network device and clear the
2938 * software flow control OFF bit for this network device
2939 */
2940 static inline void netdev_reset_queue(struct net_device *dev_queue)
2941 {
2942 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2943 }
2944
2945 /**
2946 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2947 * @dev: network device
2948 * @queue_index: given tx queue index
2949 *
2950 * Returns 0 if given tx queue index >= number of device tx queues,
2951 * otherwise returns the originally passed tx queue index.
2952 */
2953 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2954 {
2955 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2956 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2957 dev->name, queue_index,
2958 dev->real_num_tx_queues);
2959 return 0;
2960 }
2961
2962 return queue_index;
2963 }
2964
2965 /**
2966 * netif_running - test if up
2967 * @dev: network device
2968 *
2969 * Test if the device has been brought up.
2970 */
2971 static inline bool netif_running(const struct net_device *dev)
2972 {
2973 return test_bit(__LINK_STATE_START, &dev->state);
2974 }
2975
2976 /*
2977 * Routines to manage the subqueues on a device. We only need start
2978 * stop, and a check if it's stopped. All other device management is
2979 * done at the overall netdevice level.
2980 * Also test the device if we're multiqueue.
2981 */
2982
2983 /**
2984 * netif_start_subqueue - allow sending packets on subqueue
2985 * @dev: network device
2986 * @queue_index: sub queue index
2987 *
2988 * Start individual transmit queue of a device with multiple transmit queues.
2989 */
2990 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2991 {
2992 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2993
2994 netif_tx_start_queue(txq);
2995 }
2996
2997 /**
2998 * netif_stop_subqueue - stop sending packets on subqueue
2999 * @dev: network device
3000 * @queue_index: sub queue index
3001 *
3002 * Stop individual transmit queue of a device with multiple transmit queues.
3003 */
3004 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3005 {
3006 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3007 netif_tx_stop_queue(txq);
3008 }
3009
3010 /**
3011 * netif_subqueue_stopped - test status of subqueue
3012 * @dev: network device
3013 * @queue_index: sub queue index
3014 *
3015 * Check individual transmit queue of a device with multiple transmit queues.
3016 */
3017 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3018 u16 queue_index)
3019 {
3020 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3021
3022 return netif_tx_queue_stopped(txq);
3023 }
3024
3025 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3026 struct sk_buff *skb)
3027 {
3028 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3029 }
3030
3031 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3032
3033 #ifdef CONFIG_XPS
3034 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3035 u16 index);
3036 #else
3037 static inline int netif_set_xps_queue(struct net_device *dev,
3038 const struct cpumask *mask,
3039 u16 index)
3040 {
3041 return 0;
3042 }
3043 #endif
3044
3045 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3046 unsigned int num_tx_queues);
3047
3048 /*
3049 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3050 * as a distribution range limit for the returned value.
3051 */
3052 static inline u16 skb_tx_hash(const struct net_device *dev,
3053 struct sk_buff *skb)
3054 {
3055 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3056 }
3057
3058 /**
3059 * netif_is_multiqueue - test if device has multiple transmit queues
3060 * @dev: network device
3061 *
3062 * Check if device has multiple transmit queues
3063 */
3064 static inline bool netif_is_multiqueue(const struct net_device *dev)
3065 {
3066 return dev->num_tx_queues > 1;
3067 }
3068
3069 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3070
3071 #ifdef CONFIG_SYSFS
3072 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3073 #else
3074 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3075 unsigned int rxq)
3076 {
3077 return 0;
3078 }
3079 #endif
3080
3081 #ifdef CONFIG_SYSFS
3082 static inline unsigned int get_netdev_rx_queue_index(
3083 struct netdev_rx_queue *queue)
3084 {
3085 struct net_device *dev = queue->dev;
3086 int index = queue - dev->_rx;
3087
3088 BUG_ON(index >= dev->num_rx_queues);
3089 return index;
3090 }
3091 #endif
3092
3093 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3094 int netif_get_num_default_rss_queues(void);
3095
3096 enum skb_free_reason {
3097 SKB_REASON_CONSUMED,
3098 SKB_REASON_DROPPED,
3099 };
3100
3101 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3102 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3103
3104 /*
3105 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3106 * interrupt context or with hardware interrupts being disabled.
3107 * (in_irq() || irqs_disabled())
3108 *
3109 * We provide four helpers that can be used in following contexts :
3110 *
3111 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3112 * replacing kfree_skb(skb)
3113 *
3114 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3115 * Typically used in place of consume_skb(skb) in TX completion path
3116 *
3117 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3118 * replacing kfree_skb(skb)
3119 *
3120 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3121 * and consumed a packet. Used in place of consume_skb(skb)
3122 */
3123 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3124 {
3125 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3126 }
3127
3128 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3129 {
3130 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3131 }
3132
3133 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3134 {
3135 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3136 }
3137
3138 static inline void dev_consume_skb_any(struct sk_buff *skb)
3139 {
3140 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3141 }
3142
3143 int netif_rx(struct sk_buff *skb);
3144 int netif_rx_ni(struct sk_buff *skb);
3145 int netif_receive_skb(struct sk_buff *skb);
3146 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3147 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3148 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3149 gro_result_t napi_gro_frags(struct napi_struct *napi);
3150 struct packet_offload *gro_find_receive_by_type(__be16 type);
3151 struct packet_offload *gro_find_complete_by_type(__be16 type);
3152
3153 static inline void napi_free_frags(struct napi_struct *napi)
3154 {
3155 kfree_skb(napi->skb);
3156 napi->skb = NULL;
3157 }
3158
3159 int netdev_rx_handler_register(struct net_device *dev,
3160 rx_handler_func_t *rx_handler,
3161 void *rx_handler_data);
3162 void netdev_rx_handler_unregister(struct net_device *dev);
3163
3164 bool dev_valid_name(const char *name);
3165 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3166 int dev_ethtool(struct net *net, struct ifreq *);
3167 unsigned int dev_get_flags(const struct net_device *);
3168 int __dev_change_flags(struct net_device *, unsigned int flags);
3169 int dev_change_flags(struct net_device *, unsigned int);
3170 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3171 unsigned int gchanges);
3172 int dev_change_name(struct net_device *, const char *);
3173 int dev_set_alias(struct net_device *, const char *, size_t);
3174 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3175 int dev_set_mtu(struct net_device *, int);
3176 void dev_set_group(struct net_device *, int);
3177 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3178 int dev_change_carrier(struct net_device *, bool new_carrier);
3179 int dev_get_phys_port_id(struct net_device *dev,
3180 struct netdev_phys_item_id *ppid);
3181 int dev_get_phys_port_name(struct net_device *dev,
3182 char *name, size_t len);
3183 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3184 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3185 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3186 struct netdev_queue *txq, int *ret);
3187 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3188 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3189 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3190
3191 extern int netdev_budget;
3192
3193 /* Called by rtnetlink.c:rtnl_unlock() */
3194 void netdev_run_todo(void);
3195
3196 /**
3197 * dev_put - release reference to device
3198 * @dev: network device
3199 *
3200 * Release reference to device to allow it to be freed.
3201 */
3202 static inline void dev_put(struct net_device *dev)
3203 {
3204 this_cpu_dec(*dev->pcpu_refcnt);
3205 }
3206
3207 /**
3208 * dev_hold - get reference to device
3209 * @dev: network device
3210 *
3211 * Hold reference to device to keep it from being freed.
3212 */
3213 static inline void dev_hold(struct net_device *dev)
3214 {
3215 this_cpu_inc(*dev->pcpu_refcnt);
3216 }
3217
3218 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3219 * and _off may be called from IRQ context, but it is caller
3220 * who is responsible for serialization of these calls.
3221 *
3222 * The name carrier is inappropriate, these functions should really be
3223 * called netif_lowerlayer_*() because they represent the state of any
3224 * kind of lower layer not just hardware media.
3225 */
3226
3227 void linkwatch_init_dev(struct net_device *dev);
3228 void linkwatch_fire_event(struct net_device *dev);
3229 void linkwatch_forget_dev(struct net_device *dev);
3230
3231 /**
3232 * netif_carrier_ok - test if carrier present
3233 * @dev: network device
3234 *
3235 * Check if carrier is present on device
3236 */
3237 static inline bool netif_carrier_ok(const struct net_device *dev)
3238 {
3239 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3240 }
3241
3242 unsigned long dev_trans_start(struct net_device *dev);
3243
3244 void __netdev_watchdog_up(struct net_device *dev);
3245
3246 void netif_carrier_on(struct net_device *dev);
3247
3248 void netif_carrier_off(struct net_device *dev);
3249
3250 /**
3251 * netif_dormant_on - mark device as dormant.
3252 * @dev: network device
3253 *
3254 * Mark device as dormant (as per RFC2863).
3255 *
3256 * The dormant state indicates that the relevant interface is not
3257 * actually in a condition to pass packets (i.e., it is not 'up') but is
3258 * in a "pending" state, waiting for some external event. For "on-
3259 * demand" interfaces, this new state identifies the situation where the
3260 * interface is waiting for events to place it in the up state.
3261 *
3262 */
3263 static inline void netif_dormant_on(struct net_device *dev)
3264 {
3265 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3266 linkwatch_fire_event(dev);
3267 }
3268
3269 /**
3270 * netif_dormant_off - set device as not dormant.
3271 * @dev: network device
3272 *
3273 * Device is not in dormant state.
3274 */
3275 static inline void netif_dormant_off(struct net_device *dev)
3276 {
3277 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3278 linkwatch_fire_event(dev);
3279 }
3280
3281 /**
3282 * netif_dormant - test if carrier present
3283 * @dev: network device
3284 *
3285 * Check if carrier is present on device
3286 */
3287 static inline bool netif_dormant(const struct net_device *dev)
3288 {
3289 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3290 }
3291
3292
3293 /**
3294 * netif_oper_up - test if device is operational
3295 * @dev: network device
3296 *
3297 * Check if carrier is operational
3298 */
3299 static inline bool netif_oper_up(const struct net_device *dev)
3300 {
3301 return (dev->operstate == IF_OPER_UP ||
3302 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3303 }
3304
3305 /**
3306 * netif_device_present - is device available or removed
3307 * @dev: network device
3308 *
3309 * Check if device has not been removed from system.
3310 */
3311 static inline bool netif_device_present(struct net_device *dev)
3312 {
3313 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3314 }
3315
3316 void netif_device_detach(struct net_device *dev);
3317
3318 void netif_device_attach(struct net_device *dev);
3319
3320 /*
3321 * Network interface message level settings
3322 */
3323
3324 enum {
3325 NETIF_MSG_DRV = 0x0001,
3326 NETIF_MSG_PROBE = 0x0002,
3327 NETIF_MSG_LINK = 0x0004,
3328 NETIF_MSG_TIMER = 0x0008,
3329 NETIF_MSG_IFDOWN = 0x0010,
3330 NETIF_MSG_IFUP = 0x0020,
3331 NETIF_MSG_RX_ERR = 0x0040,
3332 NETIF_MSG_TX_ERR = 0x0080,
3333 NETIF_MSG_TX_QUEUED = 0x0100,
3334 NETIF_MSG_INTR = 0x0200,
3335 NETIF_MSG_TX_DONE = 0x0400,
3336 NETIF_MSG_RX_STATUS = 0x0800,
3337 NETIF_MSG_PKTDATA = 0x1000,
3338 NETIF_MSG_HW = 0x2000,
3339 NETIF_MSG_WOL = 0x4000,
3340 };
3341
3342 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3343 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3344 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3345 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3346 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3347 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3348 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3349 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3350 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3351 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3352 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3353 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3354 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3355 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3356 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3357
3358 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3359 {
3360 /* use default */
3361 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3362 return default_msg_enable_bits;
3363 if (debug_value == 0) /* no output */
3364 return 0;
3365 /* set low N bits */
3366 return (1 << debug_value) - 1;
3367 }
3368
3369 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3370 {
3371 spin_lock(&txq->_xmit_lock);
3372 txq->xmit_lock_owner = cpu;
3373 }
3374
3375 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3376 {
3377 spin_lock_bh(&txq->_xmit_lock);
3378 txq->xmit_lock_owner = smp_processor_id();
3379 }
3380
3381 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3382 {
3383 bool ok = spin_trylock(&txq->_xmit_lock);
3384 if (likely(ok))
3385 txq->xmit_lock_owner = smp_processor_id();
3386 return ok;
3387 }
3388
3389 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3390 {
3391 txq->xmit_lock_owner = -1;
3392 spin_unlock(&txq->_xmit_lock);
3393 }
3394
3395 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3396 {
3397 txq->xmit_lock_owner = -1;
3398 spin_unlock_bh(&txq->_xmit_lock);
3399 }
3400
3401 static inline void txq_trans_update(struct netdev_queue *txq)
3402 {
3403 if (txq->xmit_lock_owner != -1)
3404 txq->trans_start = jiffies;
3405 }
3406
3407 /**
3408 * netif_tx_lock - grab network device transmit lock
3409 * @dev: network device
3410 *
3411 * Get network device transmit lock
3412 */
3413 static inline void netif_tx_lock(struct net_device *dev)
3414 {
3415 unsigned int i;
3416 int cpu;
3417
3418 spin_lock(&dev->tx_global_lock);
3419 cpu = smp_processor_id();
3420 for (i = 0; i < dev->num_tx_queues; i++) {
3421 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3422
3423 /* We are the only thread of execution doing a
3424 * freeze, but we have to grab the _xmit_lock in
3425 * order to synchronize with threads which are in
3426 * the ->hard_start_xmit() handler and already
3427 * checked the frozen bit.
3428 */
3429 __netif_tx_lock(txq, cpu);
3430 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3431 __netif_tx_unlock(txq);
3432 }
3433 }
3434
3435 static inline void netif_tx_lock_bh(struct net_device *dev)
3436 {
3437 local_bh_disable();
3438 netif_tx_lock(dev);
3439 }
3440
3441 static inline void netif_tx_unlock(struct net_device *dev)
3442 {
3443 unsigned int i;
3444
3445 for (i = 0; i < dev->num_tx_queues; i++) {
3446 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3447
3448 /* No need to grab the _xmit_lock here. If the
3449 * queue is not stopped for another reason, we
3450 * force a schedule.
3451 */
3452 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3453 netif_schedule_queue(txq);
3454 }
3455 spin_unlock(&dev->tx_global_lock);
3456 }
3457
3458 static inline void netif_tx_unlock_bh(struct net_device *dev)
3459 {
3460 netif_tx_unlock(dev);
3461 local_bh_enable();
3462 }
3463
3464 #define HARD_TX_LOCK(dev, txq, cpu) { \
3465 if ((dev->features & NETIF_F_LLTX) == 0) { \
3466 __netif_tx_lock(txq, cpu); \
3467 } \
3468 }
3469
3470 #define HARD_TX_TRYLOCK(dev, txq) \
3471 (((dev->features & NETIF_F_LLTX) == 0) ? \
3472 __netif_tx_trylock(txq) : \
3473 true )
3474
3475 #define HARD_TX_UNLOCK(dev, txq) { \
3476 if ((dev->features & NETIF_F_LLTX) == 0) { \
3477 __netif_tx_unlock(txq); \
3478 } \
3479 }
3480
3481 static inline void netif_tx_disable(struct net_device *dev)
3482 {
3483 unsigned int i;
3484 int cpu;
3485
3486 local_bh_disable();
3487 cpu = smp_processor_id();
3488 for (i = 0; i < dev->num_tx_queues; i++) {
3489 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3490
3491 __netif_tx_lock(txq, cpu);
3492 netif_tx_stop_queue(txq);
3493 __netif_tx_unlock(txq);
3494 }
3495 local_bh_enable();
3496 }
3497
3498 static inline void netif_addr_lock(struct net_device *dev)
3499 {
3500 spin_lock(&dev->addr_list_lock);
3501 }
3502
3503 static inline void netif_addr_lock_nested(struct net_device *dev)
3504 {
3505 int subclass = SINGLE_DEPTH_NESTING;
3506
3507 if (dev->netdev_ops->ndo_get_lock_subclass)
3508 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3509
3510 spin_lock_nested(&dev->addr_list_lock, subclass);
3511 }
3512
3513 static inline void netif_addr_lock_bh(struct net_device *dev)
3514 {
3515 spin_lock_bh(&dev->addr_list_lock);
3516 }
3517
3518 static inline void netif_addr_unlock(struct net_device *dev)
3519 {
3520 spin_unlock(&dev->addr_list_lock);
3521 }
3522
3523 static inline void netif_addr_unlock_bh(struct net_device *dev)
3524 {
3525 spin_unlock_bh(&dev->addr_list_lock);
3526 }
3527
3528 /*
3529 * dev_addrs walker. Should be used only for read access. Call with
3530 * rcu_read_lock held.
3531 */
3532 #define for_each_dev_addr(dev, ha) \
3533 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3534
3535 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3536
3537 void ether_setup(struct net_device *dev);
3538
3539 /* Support for loadable net-drivers */
3540 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3541 unsigned char name_assign_type,
3542 void (*setup)(struct net_device *),
3543 unsigned int txqs, unsigned int rxqs);
3544 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3545 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3546
3547 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3548 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3549 count)
3550
3551 int register_netdev(struct net_device *dev);
3552 void unregister_netdev(struct net_device *dev);
3553
3554 /* General hardware address lists handling functions */
3555 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3556 struct netdev_hw_addr_list *from_list, int addr_len);
3557 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3558 struct netdev_hw_addr_list *from_list, int addr_len);
3559 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3560 struct net_device *dev,
3561 int (*sync)(struct net_device *, const unsigned char *),
3562 int (*unsync)(struct net_device *,
3563 const unsigned char *));
3564 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3565 struct net_device *dev,
3566 int (*unsync)(struct net_device *,
3567 const unsigned char *));
3568 void __hw_addr_init(struct netdev_hw_addr_list *list);
3569
3570 /* Functions used for device addresses handling */
3571 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3572 unsigned char addr_type);
3573 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3574 unsigned char addr_type);
3575 void dev_addr_flush(struct net_device *dev);
3576 int dev_addr_init(struct net_device *dev);
3577
3578 /* Functions used for unicast addresses handling */
3579 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3580 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3581 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3582 int dev_uc_sync(struct net_device *to, struct net_device *from);
3583 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3584 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3585 void dev_uc_flush(struct net_device *dev);
3586 void dev_uc_init(struct net_device *dev);
3587
3588 /**
3589 * __dev_uc_sync - Synchonize device's unicast list
3590 * @dev: device to sync
3591 * @sync: function to call if address should be added
3592 * @unsync: function to call if address should be removed
3593 *
3594 * Add newly added addresses to the interface, and release
3595 * addresses that have been deleted.
3596 **/
3597 static inline int __dev_uc_sync(struct net_device *dev,
3598 int (*sync)(struct net_device *,
3599 const unsigned char *),
3600 int (*unsync)(struct net_device *,
3601 const unsigned char *))
3602 {
3603 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3604 }
3605
3606 /**
3607 * __dev_uc_unsync - Remove synchronized addresses from device
3608 * @dev: device to sync
3609 * @unsync: function to call if address should be removed
3610 *
3611 * Remove all addresses that were added to the device by dev_uc_sync().
3612 **/
3613 static inline void __dev_uc_unsync(struct net_device *dev,
3614 int (*unsync)(struct net_device *,
3615 const unsigned char *))
3616 {
3617 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3618 }
3619
3620 /* Functions used for multicast addresses handling */
3621 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3622 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3623 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3624 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3625 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3626 int dev_mc_sync(struct net_device *to, struct net_device *from);
3627 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3628 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3629 void dev_mc_flush(struct net_device *dev);
3630 void dev_mc_init(struct net_device *dev);
3631
3632 /**
3633 * __dev_mc_sync - Synchonize device's multicast list
3634 * @dev: device to sync
3635 * @sync: function to call if address should be added
3636 * @unsync: function to call if address should be removed
3637 *
3638 * Add newly added addresses to the interface, and release
3639 * addresses that have been deleted.
3640 **/
3641 static inline int __dev_mc_sync(struct net_device *dev,
3642 int (*sync)(struct net_device *,
3643 const unsigned char *),
3644 int (*unsync)(struct net_device *,
3645 const unsigned char *))
3646 {
3647 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3648 }
3649
3650 /**
3651 * __dev_mc_unsync - Remove synchronized addresses from device
3652 * @dev: device to sync
3653 * @unsync: function to call if address should be removed
3654 *
3655 * Remove all addresses that were added to the device by dev_mc_sync().
3656 **/
3657 static inline void __dev_mc_unsync(struct net_device *dev,
3658 int (*unsync)(struct net_device *,
3659 const unsigned char *))
3660 {
3661 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3662 }
3663
3664 /* Functions used for secondary unicast and multicast support */
3665 void dev_set_rx_mode(struct net_device *dev);
3666 void __dev_set_rx_mode(struct net_device *dev);
3667 int dev_set_promiscuity(struct net_device *dev, int inc);
3668 int dev_set_allmulti(struct net_device *dev, int inc);
3669 void netdev_state_change(struct net_device *dev);
3670 void netdev_notify_peers(struct net_device *dev);
3671 void netdev_features_change(struct net_device *dev);
3672 /* Load a device via the kmod */
3673 void dev_load(struct net *net, const char *name);
3674 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3675 struct rtnl_link_stats64 *storage);
3676 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3677 const struct net_device_stats *netdev_stats);
3678
3679 extern int netdev_max_backlog;
3680 extern int netdev_tstamp_prequeue;
3681 extern int weight_p;
3682 extern int bpf_jit_enable;
3683
3684 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3685 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3686 struct list_head **iter);
3687 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3688 struct list_head **iter);
3689
3690 /* iterate through upper list, must be called under RCU read lock */
3691 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3692 for (iter = &(dev)->adj_list.upper, \
3693 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3694 updev; \
3695 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3696
3697 /* iterate through upper list, must be called under RCU read lock */
3698 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3699 for (iter = &(dev)->all_adj_list.upper, \
3700 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3701 updev; \
3702 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3703
3704 void *netdev_lower_get_next_private(struct net_device *dev,
3705 struct list_head **iter);
3706 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3707 struct list_head **iter);
3708
3709 #define netdev_for_each_lower_private(dev, priv, iter) \
3710 for (iter = (dev)->adj_list.lower.next, \
3711 priv = netdev_lower_get_next_private(dev, &(iter)); \
3712 priv; \
3713 priv = netdev_lower_get_next_private(dev, &(iter)))
3714
3715 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3716 for (iter = &(dev)->adj_list.lower, \
3717 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3718 priv; \
3719 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3720
3721 void *netdev_lower_get_next(struct net_device *dev,
3722 struct list_head **iter);
3723 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3724 for (iter = &(dev)->adj_list.lower, \
3725 ldev = netdev_lower_get_next(dev, &(iter)); \
3726 ldev; \
3727 ldev = netdev_lower_get_next(dev, &(iter)))
3728
3729 void *netdev_adjacent_get_private(struct list_head *adj_list);
3730 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3731 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3732 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3733 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3734 int netdev_master_upper_dev_link(struct net_device *dev,
3735 struct net_device *upper_dev,
3736 void *upper_priv, void *upper_info);
3737 void netdev_upper_dev_unlink(struct net_device *dev,
3738 struct net_device *upper_dev);
3739 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3740 void *netdev_lower_dev_get_private(struct net_device *dev,
3741 struct net_device *lower_dev);
3742 void netdev_lower_state_changed(struct net_device *lower_dev,
3743 void *lower_state_info);
3744
3745 /* RSS keys are 40 or 52 bytes long */
3746 #define NETDEV_RSS_KEY_LEN 52
3747 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3748 void netdev_rss_key_fill(void *buffer, size_t len);
3749
3750 int dev_get_nest_level(struct net_device *dev,
3751 bool (*type_check)(const struct net_device *dev));
3752 int skb_checksum_help(struct sk_buff *skb);
3753 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3754 netdev_features_t features, bool tx_path);
3755 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3756 netdev_features_t features);
3757
3758 struct netdev_bonding_info {
3759 ifslave slave;
3760 ifbond master;
3761 };
3762
3763 struct netdev_notifier_bonding_info {
3764 struct netdev_notifier_info info; /* must be first */
3765 struct netdev_bonding_info bonding_info;
3766 };
3767
3768 void netdev_bonding_info_change(struct net_device *dev,
3769 struct netdev_bonding_info *bonding_info);
3770
3771 static inline
3772 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3773 {
3774 return __skb_gso_segment(skb, features, true);
3775 }
3776 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3777
3778 static inline bool can_checksum_protocol(netdev_features_t features,
3779 __be16 protocol)
3780 {
3781 if (protocol == htons(ETH_P_FCOE))
3782 return !!(features & NETIF_F_FCOE_CRC);
3783
3784 /* Assume this is an IP checksum (not SCTP CRC) */
3785
3786 if (features & NETIF_F_HW_CSUM) {
3787 /* Can checksum everything */
3788 return true;
3789 }
3790
3791 switch (protocol) {
3792 case htons(ETH_P_IP):
3793 return !!(features & NETIF_F_IP_CSUM);
3794 case htons(ETH_P_IPV6):
3795 return !!(features & NETIF_F_IPV6_CSUM);
3796 default:
3797 return false;
3798 }
3799 }
3800
3801 /* Map an ethertype into IP protocol if possible */
3802 static inline int eproto_to_ipproto(int eproto)
3803 {
3804 switch (eproto) {
3805 case htons(ETH_P_IP):
3806 return IPPROTO_IP;
3807 case htons(ETH_P_IPV6):
3808 return IPPROTO_IPV6;
3809 default:
3810 return -1;
3811 }
3812 }
3813
3814 #ifdef CONFIG_BUG
3815 void netdev_rx_csum_fault(struct net_device *dev);
3816 #else
3817 static inline void netdev_rx_csum_fault(struct net_device *dev)
3818 {
3819 }
3820 #endif
3821 /* rx skb timestamps */
3822 void net_enable_timestamp(void);
3823 void net_disable_timestamp(void);
3824
3825 #ifdef CONFIG_PROC_FS
3826 int __init dev_proc_init(void);
3827 #else
3828 #define dev_proc_init() 0
3829 #endif
3830
3831 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3832 struct sk_buff *skb, struct net_device *dev,
3833 bool more)
3834 {
3835 skb->xmit_more = more ? 1 : 0;
3836 return ops->ndo_start_xmit(skb, dev);
3837 }
3838
3839 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3840 struct netdev_queue *txq, bool more)
3841 {
3842 const struct net_device_ops *ops = dev->netdev_ops;
3843 int rc;
3844
3845 rc = __netdev_start_xmit(ops, skb, dev, more);
3846 if (rc == NETDEV_TX_OK)
3847 txq_trans_update(txq);
3848
3849 return rc;
3850 }
3851
3852 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3853 const void *ns);
3854 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3855 const void *ns);
3856
3857 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3858 {
3859 return netdev_class_create_file_ns(class_attr, NULL);
3860 }
3861
3862 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3863 {
3864 netdev_class_remove_file_ns(class_attr, NULL);
3865 }
3866
3867 extern struct kobj_ns_type_operations net_ns_type_operations;
3868
3869 const char *netdev_drivername(const struct net_device *dev);
3870
3871 void linkwatch_run_queue(void);
3872
3873 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3874 netdev_features_t f2)
3875 {
3876 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3877 if (f1 & NETIF_F_HW_CSUM)
3878 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3879 else
3880 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3881 }
3882
3883 return f1 & f2;
3884 }
3885
3886 static inline netdev_features_t netdev_get_wanted_features(
3887 struct net_device *dev)
3888 {
3889 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3890 }
3891 netdev_features_t netdev_increment_features(netdev_features_t all,
3892 netdev_features_t one, netdev_features_t mask);
3893
3894 /* Allow TSO being used on stacked device :
3895 * Performing the GSO segmentation before last device
3896 * is a performance improvement.
3897 */
3898 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3899 netdev_features_t mask)
3900 {
3901 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3902 }
3903
3904 int __netdev_update_features(struct net_device *dev);
3905 void netdev_update_features(struct net_device *dev);
3906 void netdev_change_features(struct net_device *dev);
3907
3908 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3909 struct net_device *dev);
3910
3911 netdev_features_t passthru_features_check(struct sk_buff *skb,
3912 struct net_device *dev,
3913 netdev_features_t features);
3914 netdev_features_t netif_skb_features(struct sk_buff *skb);
3915
3916 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3917 {
3918 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3919
3920 /* check flags correspondence */
3921 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3922 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3923 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3924 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3925 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3926 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3927 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3928 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3929 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3930 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3931 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3932 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3933 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3934
3935 return (features & feature) == feature;
3936 }
3937
3938 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3939 {
3940 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3941 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3942 }
3943
3944 static inline bool netif_needs_gso(struct sk_buff *skb,
3945 netdev_features_t features)
3946 {
3947 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3948 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3949 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3950 }
3951
3952 static inline void netif_set_gso_max_size(struct net_device *dev,
3953 unsigned int size)
3954 {
3955 dev->gso_max_size = size;
3956 }
3957
3958 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3959 int pulled_hlen, u16 mac_offset,
3960 int mac_len)
3961 {
3962 skb->protocol = protocol;
3963 skb->encapsulation = 1;
3964 skb_push(skb, pulled_hlen);
3965 skb_reset_transport_header(skb);
3966 skb->mac_header = mac_offset;
3967 skb->network_header = skb->mac_header + mac_len;
3968 skb->mac_len = mac_len;
3969 }
3970
3971 static inline bool netif_is_macvlan(const struct net_device *dev)
3972 {
3973 return dev->priv_flags & IFF_MACVLAN;
3974 }
3975
3976 static inline bool netif_is_macvlan_port(const struct net_device *dev)
3977 {
3978 return dev->priv_flags & IFF_MACVLAN_PORT;
3979 }
3980
3981 static inline bool netif_is_ipvlan(const struct net_device *dev)
3982 {
3983 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3984 }
3985
3986 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
3987 {
3988 return dev->priv_flags & IFF_IPVLAN_MASTER;
3989 }
3990
3991 static inline bool netif_is_bond_master(const struct net_device *dev)
3992 {
3993 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3994 }
3995
3996 static inline bool netif_is_bond_slave(const struct net_device *dev)
3997 {
3998 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3999 }
4000
4001 static inline bool netif_supports_nofcs(struct net_device *dev)
4002 {
4003 return dev->priv_flags & IFF_SUPP_NOFCS;
4004 }
4005
4006 static inline bool netif_is_l3_master(const struct net_device *dev)
4007 {
4008 return dev->priv_flags & IFF_L3MDEV_MASTER;
4009 }
4010
4011 static inline bool netif_is_l3_slave(const struct net_device *dev)
4012 {
4013 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4014 }
4015
4016 static inline bool netif_is_bridge_master(const struct net_device *dev)
4017 {
4018 return dev->priv_flags & IFF_EBRIDGE;
4019 }
4020
4021 static inline bool netif_is_bridge_port(const struct net_device *dev)
4022 {
4023 return dev->priv_flags & IFF_BRIDGE_PORT;
4024 }
4025
4026 static inline bool netif_is_ovs_master(const struct net_device *dev)
4027 {
4028 return dev->priv_flags & IFF_OPENVSWITCH;
4029 }
4030
4031 static inline bool netif_is_team_master(const struct net_device *dev)
4032 {
4033 return dev->priv_flags & IFF_TEAM;
4034 }
4035
4036 static inline bool netif_is_team_port(const struct net_device *dev)
4037 {
4038 return dev->priv_flags & IFF_TEAM_PORT;
4039 }
4040
4041 static inline bool netif_is_lag_master(const struct net_device *dev)
4042 {
4043 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4044 }
4045
4046 static inline bool netif_is_lag_port(const struct net_device *dev)
4047 {
4048 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4049 }
4050
4051 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4052 static inline void netif_keep_dst(struct net_device *dev)
4053 {
4054 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4055 }
4056
4057 extern struct pernet_operations __net_initdata loopback_net_ops;
4058
4059 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4060
4061 /* netdev_printk helpers, similar to dev_printk */
4062
4063 static inline const char *netdev_name(const struct net_device *dev)
4064 {
4065 if (!dev->name[0] || strchr(dev->name, '%'))
4066 return "(unnamed net_device)";
4067 return dev->name;
4068 }
4069
4070 static inline const char *netdev_reg_state(const struct net_device *dev)
4071 {
4072 switch (dev->reg_state) {
4073 case NETREG_UNINITIALIZED: return " (uninitialized)";
4074 case NETREG_REGISTERED: return "";
4075 case NETREG_UNREGISTERING: return " (unregistering)";
4076 case NETREG_UNREGISTERED: return " (unregistered)";
4077 case NETREG_RELEASED: return " (released)";
4078 case NETREG_DUMMY: return " (dummy)";
4079 }
4080
4081 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4082 return " (unknown)";
4083 }
4084
4085 __printf(3, 4)
4086 void netdev_printk(const char *level, const struct net_device *dev,
4087 const char *format, ...);
4088 __printf(2, 3)
4089 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4090 __printf(2, 3)
4091 void netdev_alert(const struct net_device *dev, const char *format, ...);
4092 __printf(2, 3)
4093 void netdev_crit(const struct net_device *dev, const char *format, ...);
4094 __printf(2, 3)
4095 void netdev_err(const struct net_device *dev, const char *format, ...);
4096 __printf(2, 3)
4097 void netdev_warn(const struct net_device *dev, const char *format, ...);
4098 __printf(2, 3)
4099 void netdev_notice(const struct net_device *dev, const char *format, ...);
4100 __printf(2, 3)
4101 void netdev_info(const struct net_device *dev, const char *format, ...);
4102
4103 #define MODULE_ALIAS_NETDEV(device) \
4104 MODULE_ALIAS("netdev-" device)
4105
4106 #if defined(CONFIG_DYNAMIC_DEBUG)
4107 #define netdev_dbg(__dev, format, args...) \
4108 do { \
4109 dynamic_netdev_dbg(__dev, format, ##args); \
4110 } while (0)
4111 #elif defined(DEBUG)
4112 #define netdev_dbg(__dev, format, args...) \
4113 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4114 #else
4115 #define netdev_dbg(__dev, format, args...) \
4116 ({ \
4117 if (0) \
4118 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4119 })
4120 #endif
4121
4122 #if defined(VERBOSE_DEBUG)
4123 #define netdev_vdbg netdev_dbg
4124 #else
4125
4126 #define netdev_vdbg(dev, format, args...) \
4127 ({ \
4128 if (0) \
4129 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4130 0; \
4131 })
4132 #endif
4133
4134 /*
4135 * netdev_WARN() acts like dev_printk(), but with the key difference
4136 * of using a WARN/WARN_ON to get the message out, including the
4137 * file/line information and a backtrace.
4138 */
4139 #define netdev_WARN(dev, format, args...) \
4140 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4141 netdev_reg_state(dev), ##args)
4142
4143 /* netif printk helpers, similar to netdev_printk */
4144
4145 #define netif_printk(priv, type, level, dev, fmt, args...) \
4146 do { \
4147 if (netif_msg_##type(priv)) \
4148 netdev_printk(level, (dev), fmt, ##args); \
4149 } while (0)
4150
4151 #define netif_level(level, priv, type, dev, fmt, args...) \
4152 do { \
4153 if (netif_msg_##type(priv)) \
4154 netdev_##level(dev, fmt, ##args); \
4155 } while (0)
4156
4157 #define netif_emerg(priv, type, dev, fmt, args...) \
4158 netif_level(emerg, priv, type, dev, fmt, ##args)
4159 #define netif_alert(priv, type, dev, fmt, args...) \
4160 netif_level(alert, priv, type, dev, fmt, ##args)
4161 #define netif_crit(priv, type, dev, fmt, args...) \
4162 netif_level(crit, priv, type, dev, fmt, ##args)
4163 #define netif_err(priv, type, dev, fmt, args...) \
4164 netif_level(err, priv, type, dev, fmt, ##args)
4165 #define netif_warn(priv, type, dev, fmt, args...) \
4166 netif_level(warn, priv, type, dev, fmt, ##args)
4167 #define netif_notice(priv, type, dev, fmt, args...) \
4168 netif_level(notice, priv, type, dev, fmt, ##args)
4169 #define netif_info(priv, type, dev, fmt, args...) \
4170 netif_level(info, priv, type, dev, fmt, ##args)
4171
4172 #if defined(CONFIG_DYNAMIC_DEBUG)
4173 #define netif_dbg(priv, type, netdev, format, args...) \
4174 do { \
4175 if (netif_msg_##type(priv)) \
4176 dynamic_netdev_dbg(netdev, format, ##args); \
4177 } while (0)
4178 #elif defined(DEBUG)
4179 #define netif_dbg(priv, type, dev, format, args...) \
4180 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4181 #else
4182 #define netif_dbg(priv, type, dev, format, args...) \
4183 ({ \
4184 if (0) \
4185 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4186 0; \
4187 })
4188 #endif
4189
4190 #if defined(VERBOSE_DEBUG)
4191 #define netif_vdbg netif_dbg
4192 #else
4193 #define netif_vdbg(priv, type, dev, format, args...) \
4194 ({ \
4195 if (0) \
4196 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4197 0; \
4198 })
4199 #endif
4200
4201 /*
4202 * The list of packet types we will receive (as opposed to discard)
4203 * and the routines to invoke.
4204 *
4205 * Why 16. Because with 16 the only overlap we get on a hash of the
4206 * low nibble of the protocol value is RARP/SNAP/X.25.
4207 *
4208 * NOTE: That is no longer true with the addition of VLAN tags. Not
4209 * sure which should go first, but I bet it won't make much
4210 * difference if we are running VLANs. The good news is that
4211 * this protocol won't be in the list unless compiled in, so
4212 * the average user (w/out VLANs) will not be adversely affected.
4213 * --BLG
4214 *
4215 * 0800 IP
4216 * 8100 802.1Q VLAN
4217 * 0001 802.3
4218 * 0002 AX.25
4219 * 0004 802.2
4220 * 8035 RARP
4221 * 0005 SNAP
4222 * 0805 X.25
4223 * 0806 ARP
4224 * 8137 IPX
4225 * 0009 Localtalk
4226 * 86DD IPv6
4227 */
4228 #define PTYPE_HASH_SIZE (16)
4229 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4230
4231 #endif /* _LINUX_NETDEVICE_H */