]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/netdevice.h
net: always pass struct netdev_notifier_info to netdevice notifiers
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.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
54 struct netpoll_info;
55 struct device;
56 struct phy_device;
57 /* 802.11 specific */
58 struct wireless_dev;
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
62
63 extern void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
65
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
72
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
76
77 /*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
92 */
93
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
109
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
115 };
116 typedef enum netdev_tx netdev_tx_t;
117
118 /*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122 static inline bool dev_xmit_complete(int rc)
123 {
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134 }
135
136 /*
137 * Compute the worst case header length according to the protocols
138 * used.
139 */
140
141 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
144 # else
145 # define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
150
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
157
158 /*
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
161 */
162
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187 };
188
189
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
192
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 #endif
197
198 struct neighbour;
199 struct neigh_parms;
200 struct sk_buff;
201
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
211 bool global_use;
212 int sync_cnt;
213 int refcount;
214 int synced;
215 struct rcu_head rcu_head;
216 };
217
218 struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221 };
222
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
227
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237
238 struct hh_cache {
239 u16 hh_len;
240 u16 __pad;
241 seqlock_t hh_lock;
242
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251
252 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264
265 struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*rebuild)(struct sk_buff *skb);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275 };
276
277 /* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
279 * code.
280 */
281
282 enum netdev_state_t {
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
288 };
289
290
291 /*
292 * This structure holds at boot time configured netdevice settings. They
293 * are then used in the device probing.
294 */
295 struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300
301 extern int __init netdev_boot_setup(char *str);
302
303 /*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306 struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-cpu poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 spinlock_t poll_lock;
321 int poll_owner;
322 #endif
323 struct net_device *dev;
324 struct sk_buff *gro_list;
325 struct sk_buff *skb;
326 struct list_head dev_list;
327 };
328
329 enum {
330 NAPI_STATE_SCHED, /* Poll is scheduled */
331 NAPI_STATE_DISABLE, /* Disable pending */
332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
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 extern void __napi_schedule(struct napi_struct *n);
395
396 static inline bool napi_disable_pending(struct napi_struct *n)
397 {
398 return test_bit(NAPI_STATE_DISABLE, &n->state);
399 }
400
401 /**
402 * napi_schedule_prep - check if napi can be scheduled
403 * @n: napi context
404 *
405 * Test if NAPI routine is already running, and if not mark
406 * it as running. This is used as a condition variable
407 * insure only one NAPI poll instance runs. We also make
408 * sure there is no pending NAPI disable.
409 */
410 static inline bool napi_schedule_prep(struct napi_struct *n)
411 {
412 return !napi_disable_pending(n) &&
413 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
414 }
415
416 /**
417 * napi_schedule - schedule NAPI poll
418 * @n: napi context
419 *
420 * Schedule NAPI poll routine to be called if it is not already
421 * running.
422 */
423 static inline void napi_schedule(struct napi_struct *n)
424 {
425 if (napi_schedule_prep(n))
426 __napi_schedule(n);
427 }
428
429 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
430 static inline bool napi_reschedule(struct napi_struct *napi)
431 {
432 if (napi_schedule_prep(napi)) {
433 __napi_schedule(napi);
434 return true;
435 }
436 return false;
437 }
438
439 /**
440 * napi_complete - NAPI processing complete
441 * @n: napi context
442 *
443 * Mark NAPI processing as complete.
444 */
445 extern void __napi_complete(struct napi_struct *n);
446 extern void napi_complete(struct napi_struct *n);
447
448 /**
449 * napi_disable - prevent NAPI from scheduling
450 * @n: napi context
451 *
452 * Stop NAPI from being scheduled on this context.
453 * Waits till any outstanding processing completes.
454 */
455 static inline void napi_disable(struct napi_struct *n)
456 {
457 set_bit(NAPI_STATE_DISABLE, &n->state);
458 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
459 msleep(1);
460 clear_bit(NAPI_STATE_DISABLE, &n->state);
461 }
462
463 /**
464 * napi_enable - enable NAPI scheduling
465 * @n: napi context
466 *
467 * Resume NAPI from being scheduled on this context.
468 * Must be paired with napi_disable.
469 */
470 static inline void napi_enable(struct napi_struct *n)
471 {
472 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
473 smp_mb__before_clear_bit();
474 clear_bit(NAPI_STATE_SCHED, &n->state);
475 }
476
477 #ifdef CONFIG_SMP
478 /**
479 * napi_synchronize - wait until NAPI is not running
480 * @n: napi context
481 *
482 * Wait until NAPI is done being scheduled on this context.
483 * Waits till any outstanding processing completes but
484 * does not disable future activations.
485 */
486 static inline void napi_synchronize(const struct napi_struct *n)
487 {
488 while (test_bit(NAPI_STATE_SCHED, &n->state))
489 msleep(1);
490 }
491 #else
492 # define napi_synchronize(n) barrier()
493 #endif
494
495 enum netdev_queue_state_t {
496 __QUEUE_STATE_DRV_XOFF,
497 __QUEUE_STATE_STACK_XOFF,
498 __QUEUE_STATE_FROZEN,
499 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
500 (1 << __QUEUE_STATE_STACK_XOFF))
501 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
502 (1 << __QUEUE_STATE_FROZEN))
503 };
504 /*
505 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
506 * netif_tx_* functions below are used to manipulate this flag. The
507 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
508 * queue independently. The netif_xmit_*stopped functions below are called
509 * to check if the queue has been stopped by the driver or stack (either
510 * of the XOFF bits are set in the state). Drivers should not need to call
511 * netif_xmit*stopped functions, they should only be using netif_tx_*.
512 */
513
514 struct netdev_queue {
515 /*
516 * read mostly part
517 */
518 struct net_device *dev;
519 struct Qdisc *qdisc;
520 struct Qdisc *qdisc_sleeping;
521 #ifdef CONFIG_SYSFS
522 struct kobject kobj;
523 #endif
524 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
525 int numa_node;
526 #endif
527 /*
528 * write mostly part
529 */
530 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
531 int xmit_lock_owner;
532 /*
533 * please use this field instead of dev->trans_start
534 */
535 unsigned long trans_start;
536
537 /*
538 * Number of TX timeouts for this queue
539 * (/sys/class/net/DEV/Q/trans_timeout)
540 */
541 unsigned long trans_timeout;
542
543 unsigned long state;
544
545 #ifdef CONFIG_BQL
546 struct dql dql;
547 #endif
548 } ____cacheline_aligned_in_smp;
549
550 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
551 {
552 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
553 return q->numa_node;
554 #else
555 return NUMA_NO_NODE;
556 #endif
557 }
558
559 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
560 {
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 q->numa_node = node;
563 #endif
564 }
565
566 #ifdef CONFIG_RPS
567 /*
568 * This structure holds an RPS map which can be of variable length. The
569 * map is an array of CPUs.
570 */
571 struct rps_map {
572 unsigned int len;
573 struct rcu_head rcu;
574 u16 cpus[0];
575 };
576 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
577
578 /*
579 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
580 * tail pointer for that CPU's input queue at the time of last enqueue, and
581 * a hardware filter index.
582 */
583 struct rps_dev_flow {
584 u16 cpu;
585 u16 filter;
586 unsigned int last_qtail;
587 };
588 #define RPS_NO_FILTER 0xffff
589
590 /*
591 * The rps_dev_flow_table structure contains a table of flow mappings.
592 */
593 struct rps_dev_flow_table {
594 unsigned int mask;
595 struct rcu_head rcu;
596 struct rps_dev_flow flows[0];
597 };
598 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
599 ((_num) * sizeof(struct rps_dev_flow)))
600
601 /*
602 * The rps_sock_flow_table contains mappings of flows to the last CPU
603 * on which they were processed by the application (set in recvmsg).
604 */
605 struct rps_sock_flow_table {
606 unsigned int mask;
607 u16 ents[0];
608 };
609 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
610 ((_num) * sizeof(u16)))
611
612 #define RPS_NO_CPU 0xffff
613
614 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
615 u32 hash)
616 {
617 if (table && hash) {
618 unsigned int cpu, index = hash & table->mask;
619
620 /* We only give a hint, preemption can change cpu under us */
621 cpu = raw_smp_processor_id();
622
623 if (table->ents[index] != cpu)
624 table->ents[index] = cpu;
625 }
626 }
627
628 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
629 u32 hash)
630 {
631 if (table && hash)
632 table->ents[hash & table->mask] = RPS_NO_CPU;
633 }
634
635 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
636
637 #ifdef CONFIG_RFS_ACCEL
638 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
639 u32 flow_id, u16 filter_id);
640 #endif
641
642 /* This structure contains an instance of an RX queue. */
643 struct netdev_rx_queue {
644 struct rps_map __rcu *rps_map;
645 struct rps_dev_flow_table __rcu *rps_flow_table;
646 struct kobject kobj;
647 struct net_device *dev;
648 } ____cacheline_aligned_in_smp;
649 #endif /* CONFIG_RPS */
650
651 #ifdef CONFIG_XPS
652 /*
653 * This structure holds an XPS map which can be of variable length. The
654 * map is an array of queues.
655 */
656 struct xps_map {
657 unsigned int len;
658 unsigned int alloc_len;
659 struct rcu_head rcu;
660 u16 queues[0];
661 };
662 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
663 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
664 / sizeof(u16))
665
666 /*
667 * This structure holds all XPS maps for device. Maps are indexed by CPU.
668 */
669 struct xps_dev_maps {
670 struct rcu_head rcu;
671 struct xps_map __rcu *cpu_map[0];
672 };
673 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
674 (nr_cpu_ids * sizeof(struct xps_map *)))
675 #endif /* CONFIG_XPS */
676
677 #define TC_MAX_QUEUE 16
678 #define TC_BITMASK 15
679 /* HW offloaded queuing disciplines txq count and offset maps */
680 struct netdev_tc_txq {
681 u16 count;
682 u16 offset;
683 };
684
685 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
686 /*
687 * This structure is to hold information about the device
688 * configured to run FCoE protocol stack.
689 */
690 struct netdev_fcoe_hbainfo {
691 char manufacturer[64];
692 char serial_number[64];
693 char hardware_version[64];
694 char driver_version[64];
695 char optionrom_version[64];
696 char firmware_version[64];
697 char model[256];
698 char model_description[256];
699 };
700 #endif
701
702 /*
703 * This structure defines the management hooks for network devices.
704 * The following hooks can be defined; unless noted otherwise, they are
705 * optional and can be filled with a null pointer.
706 *
707 * int (*ndo_init)(struct net_device *dev);
708 * This function is called once when network device is registered.
709 * The network device can use this to any late stage initializaton
710 * or semantic validattion. It can fail with an error code which will
711 * be propogated back to register_netdev
712 *
713 * void (*ndo_uninit)(struct net_device *dev);
714 * This function is called when device is unregistered or when registration
715 * fails. It is not called if init fails.
716 *
717 * int (*ndo_open)(struct net_device *dev);
718 * This function is called when network device transistions to the up
719 * state.
720 *
721 * int (*ndo_stop)(struct net_device *dev);
722 * This function is called when network device transistions to the down
723 * state.
724 *
725 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
726 * struct net_device *dev);
727 * Called when a packet needs to be transmitted.
728 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
729 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
730 * Required can not be NULL.
731 *
732 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
733 * Called to decide which queue to when device supports multiple
734 * transmit queues.
735 *
736 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
737 * This function is called to allow device receiver to make
738 * changes to configuration when multicast or promiscious is enabled.
739 *
740 * void (*ndo_set_rx_mode)(struct net_device *dev);
741 * This function is called device changes address list filtering.
742 * If driver handles unicast address filtering, it should set
743 * IFF_UNICAST_FLT to its priv_flags.
744 *
745 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
746 * This function is called when the Media Access Control address
747 * needs to be changed. If this interface is not defined, the
748 * mac address can not be changed.
749 *
750 * int (*ndo_validate_addr)(struct net_device *dev);
751 * Test if Media Access Control address is valid for the device.
752 *
753 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
754 * Called when a user request an ioctl which can't be handled by
755 * the generic interface code. If not defined ioctl's return
756 * not supported error code.
757 *
758 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
759 * Used to set network devices bus interface parameters. This interface
760 * is retained for legacy reason, new devices should use the bus
761 * interface (PCI) for low level management.
762 *
763 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
764 * Called when a user wants to change the Maximum Transfer Unit
765 * of a device. If not defined, any request to change MTU will
766 * will return an error.
767 *
768 * void (*ndo_tx_timeout)(struct net_device *dev);
769 * Callback uses when the transmitter has not made any progress
770 * for dev->watchdog ticks.
771 *
772 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
773 * struct rtnl_link_stats64 *storage);
774 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
775 * Called when a user wants to get the network device usage
776 * statistics. Drivers must do one of the following:
777 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
778 * rtnl_link_stats64 structure passed by the caller.
779 * 2. Define @ndo_get_stats to update a net_device_stats structure
780 * (which should normally be dev->stats) and return a pointer to
781 * it. The structure may be changed asynchronously only if each
782 * field is written atomically.
783 * 3. Update dev->stats asynchronously and atomically, and define
784 * neither operation.
785 *
786 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
787 * If device support VLAN filtering this function is called when a
788 * VLAN id is registered.
789 *
790 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
791 * If device support VLAN filtering this function is called when a
792 * VLAN id is unregistered.
793 *
794 * void (*ndo_poll_controller)(struct net_device *dev);
795 *
796 * SR-IOV management functions.
797 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
798 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
799 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
800 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
801 * int (*ndo_get_vf_config)(struct net_device *dev,
802 * int vf, struct ifla_vf_info *ivf);
803 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
804 * struct nlattr *port[]);
805 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
806 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
807 * Called to setup 'tc' number of traffic classes in the net device. This
808 * is always called from the stack with the rtnl lock held and netif tx
809 * queues stopped. This allows the netdevice to perform queue management
810 * safely.
811 *
812 * Fiber Channel over Ethernet (FCoE) offload functions.
813 * int (*ndo_fcoe_enable)(struct net_device *dev);
814 * Called when the FCoE protocol stack wants to start using LLD for FCoE
815 * so the underlying device can perform whatever needed configuration or
816 * initialization to support acceleration of FCoE traffic.
817 *
818 * int (*ndo_fcoe_disable)(struct net_device *dev);
819 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
820 * so the underlying device can perform whatever needed clean-ups to
821 * stop supporting acceleration of FCoE traffic.
822 *
823 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
824 * struct scatterlist *sgl, unsigned int sgc);
825 * Called when the FCoE Initiator wants to initialize an I/O that
826 * is a possible candidate for Direct Data Placement (DDP). The LLD can
827 * perform necessary setup and returns 1 to indicate the device is set up
828 * successfully to perform DDP on this I/O, otherwise this returns 0.
829 *
830 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
831 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
832 * indicated by the FC exchange id 'xid', so the underlying device can
833 * clean up and reuse resources for later DDP requests.
834 *
835 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
836 * struct scatterlist *sgl, unsigned int sgc);
837 * Called when the FCoE Target wants to initialize an I/O that
838 * is a possible candidate for Direct Data Placement (DDP). The LLD can
839 * perform necessary setup and returns 1 to indicate the device is set up
840 * successfully to perform DDP on this I/O, otherwise this returns 0.
841 *
842 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
843 * struct netdev_fcoe_hbainfo *hbainfo);
844 * Called when the FCoE Protocol stack wants information on the underlying
845 * device. This information is utilized by the FCoE protocol stack to
846 * register attributes with Fiber Channel management service as per the
847 * FC-GS Fabric Device Management Information(FDMI) specification.
848 *
849 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
850 * Called when the underlying device wants to override default World Wide
851 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
852 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
853 * protocol stack to use.
854 *
855 * RFS acceleration.
856 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
857 * u16 rxq_index, u32 flow_id);
858 * Set hardware filter for RFS. rxq_index is the target queue index;
859 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
860 * Return the filter ID on success, or a negative error code.
861 *
862 * Slave management functions (for bridge, bonding, etc).
863 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
864 * Called to make another netdev an underling.
865 *
866 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
867 * Called to release previously enslaved netdev.
868 *
869 * Feature/offload setting functions.
870 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
871 * netdev_features_t features);
872 * Adjusts the requested feature flags according to device-specific
873 * constraints, and returns the resulting flags. Must not modify
874 * the device state.
875 *
876 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
877 * Called to update device configuration to new features. Passed
878 * feature set might be less than what was returned by ndo_fix_features()).
879 * Must return >0 or -errno if it changed dev->features itself.
880 *
881 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
882 * struct net_device *dev,
883 * const unsigned char *addr, u16 flags)
884 * Adds an FDB entry to dev for addr.
885 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
886 * struct net_device *dev,
887 * const unsigned char *addr)
888 * Deletes the FDB entry from dev coresponding to addr.
889 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
890 * struct net_device *dev, int idx)
891 * Used to add FDB entries to dump requests. Implementers should add
892 * entries to skb and update idx with the number of entries.
893 *
894 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
895 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
896 * struct net_device *dev, u32 filter_mask)
897 *
898 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
899 * Called to change device carrier. Soft-devices (like dummy, team, etc)
900 * which do not represent real hardware may define this to allow their
901 * userspace components to manage their virtual carrier state. Devices
902 * that determine carrier state from physical hardware properties (eg
903 * network cables) or protocol-dependent mechanisms (eg
904 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
905 */
906 struct net_device_ops {
907 int (*ndo_init)(struct net_device *dev);
908 void (*ndo_uninit)(struct net_device *dev);
909 int (*ndo_open)(struct net_device *dev);
910 int (*ndo_stop)(struct net_device *dev);
911 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
912 struct net_device *dev);
913 u16 (*ndo_select_queue)(struct net_device *dev,
914 struct sk_buff *skb);
915 void (*ndo_change_rx_flags)(struct net_device *dev,
916 int flags);
917 void (*ndo_set_rx_mode)(struct net_device *dev);
918 int (*ndo_set_mac_address)(struct net_device *dev,
919 void *addr);
920 int (*ndo_validate_addr)(struct net_device *dev);
921 int (*ndo_do_ioctl)(struct net_device *dev,
922 struct ifreq *ifr, int cmd);
923 int (*ndo_set_config)(struct net_device *dev,
924 struct ifmap *map);
925 int (*ndo_change_mtu)(struct net_device *dev,
926 int new_mtu);
927 int (*ndo_neigh_setup)(struct net_device *dev,
928 struct neigh_parms *);
929 void (*ndo_tx_timeout) (struct net_device *dev);
930
931 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
932 struct rtnl_link_stats64 *storage);
933 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
934
935 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
936 __be16 proto, u16 vid);
937 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
938 __be16 proto, u16 vid);
939 #ifdef CONFIG_NET_POLL_CONTROLLER
940 void (*ndo_poll_controller)(struct net_device *dev);
941 int (*ndo_netpoll_setup)(struct net_device *dev,
942 struct netpoll_info *info,
943 gfp_t gfp);
944 void (*ndo_netpoll_cleanup)(struct net_device *dev);
945 #endif
946 int (*ndo_set_vf_mac)(struct net_device *dev,
947 int queue, u8 *mac);
948 int (*ndo_set_vf_vlan)(struct net_device *dev,
949 int queue, u16 vlan, u8 qos);
950 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
951 int vf, int rate);
952 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
953 int vf, bool setting);
954 int (*ndo_get_vf_config)(struct net_device *dev,
955 int vf,
956 struct ifla_vf_info *ivf);
957 int (*ndo_set_vf_port)(struct net_device *dev,
958 int vf,
959 struct nlattr *port[]);
960 int (*ndo_get_vf_port)(struct net_device *dev,
961 int vf, struct sk_buff *skb);
962 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
963 #if IS_ENABLED(CONFIG_FCOE)
964 int (*ndo_fcoe_enable)(struct net_device *dev);
965 int (*ndo_fcoe_disable)(struct net_device *dev);
966 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
967 u16 xid,
968 struct scatterlist *sgl,
969 unsigned int sgc);
970 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
971 u16 xid);
972 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
973 u16 xid,
974 struct scatterlist *sgl,
975 unsigned int sgc);
976 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
977 struct netdev_fcoe_hbainfo *hbainfo);
978 #endif
979
980 #if IS_ENABLED(CONFIG_LIBFCOE)
981 #define NETDEV_FCOE_WWNN 0
982 #define NETDEV_FCOE_WWPN 1
983 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
984 u64 *wwn, int type);
985 #endif
986
987 #ifdef CONFIG_RFS_ACCEL
988 int (*ndo_rx_flow_steer)(struct net_device *dev,
989 const struct sk_buff *skb,
990 u16 rxq_index,
991 u32 flow_id);
992 #endif
993 int (*ndo_add_slave)(struct net_device *dev,
994 struct net_device *slave_dev);
995 int (*ndo_del_slave)(struct net_device *dev,
996 struct net_device *slave_dev);
997 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
998 netdev_features_t features);
999 int (*ndo_set_features)(struct net_device *dev,
1000 netdev_features_t features);
1001 int (*ndo_neigh_construct)(struct neighbour *n);
1002 void (*ndo_neigh_destroy)(struct neighbour *n);
1003
1004 int (*ndo_fdb_add)(struct ndmsg *ndm,
1005 struct nlattr *tb[],
1006 struct net_device *dev,
1007 const unsigned char *addr,
1008 u16 flags);
1009 int (*ndo_fdb_del)(struct ndmsg *ndm,
1010 struct nlattr *tb[],
1011 struct net_device *dev,
1012 const unsigned char *addr);
1013 int (*ndo_fdb_dump)(struct sk_buff *skb,
1014 struct netlink_callback *cb,
1015 struct net_device *dev,
1016 int idx);
1017
1018 int (*ndo_bridge_setlink)(struct net_device *dev,
1019 struct nlmsghdr *nlh);
1020 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1021 u32 pid, u32 seq,
1022 struct net_device *dev,
1023 u32 filter_mask);
1024 int (*ndo_bridge_dellink)(struct net_device *dev,
1025 struct nlmsghdr *nlh);
1026 int (*ndo_change_carrier)(struct net_device *dev,
1027 bool new_carrier);
1028 };
1029
1030 /*
1031 * The DEVICE structure.
1032 * Actually, this whole structure is a big mistake. It mixes I/O
1033 * data with strictly "high-level" data, and it has to know about
1034 * almost every data structure used in the INET module.
1035 *
1036 * FIXME: cleanup struct net_device such that network protocol info
1037 * moves out.
1038 */
1039
1040 struct net_device {
1041
1042 /*
1043 * This is the first field of the "visible" part of this structure
1044 * (i.e. as seen by users in the "Space.c" file). It is the name
1045 * of the interface.
1046 */
1047 char name[IFNAMSIZ];
1048
1049 /* device name hash chain, please keep it close to name[] */
1050 struct hlist_node name_hlist;
1051
1052 /* snmp alias */
1053 char *ifalias;
1054
1055 /*
1056 * I/O specific fields
1057 * FIXME: Merge these and struct ifmap into one
1058 */
1059 unsigned long mem_end; /* shared mem end */
1060 unsigned long mem_start; /* shared mem start */
1061 unsigned long base_addr; /* device I/O address */
1062 unsigned int irq; /* device IRQ number */
1063
1064 /*
1065 * Some hardware also needs these fields, but they are not
1066 * part of the usual set specified in Space.c.
1067 */
1068
1069 unsigned long state;
1070
1071 struct list_head dev_list;
1072 struct list_head napi_list;
1073 struct list_head unreg_list;
1074 struct list_head upper_dev_list; /* List of upper devices */
1075
1076
1077 /* currently active device features */
1078 netdev_features_t features;
1079 /* user-changeable features */
1080 netdev_features_t hw_features;
1081 /* user-requested features */
1082 netdev_features_t wanted_features;
1083 /* mask of features inheritable by VLAN devices */
1084 netdev_features_t vlan_features;
1085 /* mask of features inherited by encapsulating devices
1086 * This field indicates what encapsulation offloads
1087 * the hardware is capable of doing, and drivers will
1088 * need to set them appropriately.
1089 */
1090 netdev_features_t hw_enc_features;
1091 /* mask of fetures inheritable by MPLS */
1092 netdev_features_t mpls_features;
1093
1094 /* Interface index. Unique device identifier */
1095 int ifindex;
1096 int iflink;
1097
1098 struct net_device_stats stats;
1099 atomic_long_t rx_dropped; /* dropped packets by core network
1100 * Do not use this in drivers.
1101 */
1102
1103 #ifdef CONFIG_WIRELESS_EXT
1104 /* List of functions to handle Wireless Extensions (instead of ioctl).
1105 * See <net/iw_handler.h> for details. Jean II */
1106 const struct iw_handler_def * wireless_handlers;
1107 /* Instance data managed by the core of Wireless Extensions. */
1108 struct iw_public_data * wireless_data;
1109 #endif
1110 /* Management operations */
1111 const struct net_device_ops *netdev_ops;
1112 const struct ethtool_ops *ethtool_ops;
1113
1114 /* Hardware header description */
1115 const struct header_ops *header_ops;
1116
1117 unsigned int flags; /* interface flags (a la BSD) */
1118 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1119 * See if.h for definitions. */
1120 unsigned short gflags;
1121 unsigned short padded; /* How much padding added by alloc_netdev() */
1122
1123 unsigned char operstate; /* RFC2863 operstate */
1124 unsigned char link_mode; /* mapping policy to operstate */
1125
1126 unsigned char if_port; /* Selectable AUI, TP,..*/
1127 unsigned char dma; /* DMA channel */
1128
1129 unsigned int mtu; /* interface MTU value */
1130 unsigned short type; /* interface hardware type */
1131 unsigned short hard_header_len; /* hardware hdr length */
1132
1133 /* extra head- and tailroom the hardware may need, but not in all cases
1134 * can this be guaranteed, especially tailroom. Some cases also use
1135 * LL_MAX_HEADER instead to allocate the skb.
1136 */
1137 unsigned short needed_headroom;
1138 unsigned short needed_tailroom;
1139
1140 /* Interface address info. */
1141 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1142 unsigned char addr_assign_type; /* hw address assignment type */
1143 unsigned char addr_len; /* hardware address length */
1144 unsigned char neigh_priv_len;
1145 unsigned short dev_id; /* for shared network cards */
1146
1147 spinlock_t addr_list_lock;
1148 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1149 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1150 struct netdev_hw_addr_list dev_addrs; /* list of device
1151 * hw addresses
1152 */
1153 #ifdef CONFIG_SYSFS
1154 struct kset *queues_kset;
1155 #endif
1156
1157 bool uc_promisc;
1158 unsigned int promiscuity;
1159 unsigned int allmulti;
1160
1161
1162 /* Protocol specific pointers */
1163
1164 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1165 struct vlan_info __rcu *vlan_info; /* VLAN info */
1166 #endif
1167 #if IS_ENABLED(CONFIG_NET_DSA)
1168 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1169 #endif
1170 void *atalk_ptr; /* AppleTalk link */
1171 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1172 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1173 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1174 void *ax25_ptr; /* AX.25 specific data */
1175 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1176 assign before registering */
1177
1178 /*
1179 * Cache lines mostly used on receive path (including eth_type_trans())
1180 */
1181 unsigned long last_rx; /* Time of last Rx
1182 * This should not be set in
1183 * drivers, unless really needed,
1184 * because network stack (bonding)
1185 * use it if/when necessary, to
1186 * avoid dirtying this cache line.
1187 */
1188
1189 /* Interface address info used in eth_type_trans() */
1190 unsigned char *dev_addr; /* hw address, (before bcast
1191 because most packets are
1192 unicast) */
1193
1194
1195 #ifdef CONFIG_RPS
1196 struct netdev_rx_queue *_rx;
1197
1198 /* Number of RX queues allocated at register_netdev() time */
1199 unsigned int num_rx_queues;
1200
1201 /* Number of RX queues currently active in device */
1202 unsigned int real_num_rx_queues;
1203
1204 #endif
1205
1206 rx_handler_func_t __rcu *rx_handler;
1207 void __rcu *rx_handler_data;
1208
1209 struct netdev_queue __rcu *ingress_queue;
1210 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1211
1212
1213 /*
1214 * Cache lines mostly used on transmit path
1215 */
1216 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1217
1218 /* Number of TX queues allocated at alloc_netdev_mq() time */
1219 unsigned int num_tx_queues;
1220
1221 /* Number of TX queues currently active in device */
1222 unsigned int real_num_tx_queues;
1223
1224 /* root qdisc from userspace point of view */
1225 struct Qdisc *qdisc;
1226
1227 unsigned long tx_queue_len; /* Max frames per queue allowed */
1228 spinlock_t tx_global_lock;
1229
1230 #ifdef CONFIG_XPS
1231 struct xps_dev_maps __rcu *xps_maps;
1232 #endif
1233 #ifdef CONFIG_RFS_ACCEL
1234 /* CPU reverse-mapping for RX completion interrupts, indexed
1235 * by RX queue number. Assigned by driver. This must only be
1236 * set if the ndo_rx_flow_steer operation is defined. */
1237 struct cpu_rmap *rx_cpu_rmap;
1238 #endif
1239
1240 /* These may be needed for future network-power-down code. */
1241
1242 /*
1243 * trans_start here is expensive for high speed devices on SMP,
1244 * please use netdev_queue->trans_start instead.
1245 */
1246 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1247
1248 int watchdog_timeo; /* used by dev_watchdog() */
1249 struct timer_list watchdog_timer;
1250
1251 /* Number of references to this device */
1252 int __percpu *pcpu_refcnt;
1253
1254 /* delayed register/unregister */
1255 struct list_head todo_list;
1256 /* device index hash chain */
1257 struct hlist_node index_hlist;
1258
1259 struct list_head link_watch_list;
1260
1261 /* register/unregister state machine */
1262 enum { NETREG_UNINITIALIZED=0,
1263 NETREG_REGISTERED, /* completed register_netdevice */
1264 NETREG_UNREGISTERING, /* called unregister_netdevice */
1265 NETREG_UNREGISTERED, /* completed unregister todo */
1266 NETREG_RELEASED, /* called free_netdev */
1267 NETREG_DUMMY, /* dummy device for NAPI poll */
1268 } reg_state:8;
1269
1270 bool dismantle; /* device is going do be freed */
1271
1272 enum {
1273 RTNL_LINK_INITIALIZED,
1274 RTNL_LINK_INITIALIZING,
1275 } rtnl_link_state:16;
1276
1277 /* Called from unregister, can be used to call free_netdev */
1278 void (*destructor)(struct net_device *dev);
1279
1280 #ifdef CONFIG_NETPOLL
1281 struct netpoll_info __rcu *npinfo;
1282 #endif
1283
1284 #ifdef CONFIG_NET_NS
1285 /* Network namespace this network device is inside */
1286 struct net *nd_net;
1287 #endif
1288
1289 /* mid-layer private */
1290 union {
1291 void *ml_priv;
1292 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1293 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1294 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1295 struct pcpu_vstats __percpu *vstats; /* veth stats */
1296 };
1297 /* GARP */
1298 struct garp_port __rcu *garp_port;
1299 /* MRP */
1300 struct mrp_port __rcu *mrp_port;
1301
1302 /* class/net/name entry */
1303 struct device dev;
1304 /* space for optional device, statistics, and wireless sysfs groups */
1305 const struct attribute_group *sysfs_groups[4];
1306
1307 /* rtnetlink link ops */
1308 const struct rtnl_link_ops *rtnl_link_ops;
1309
1310 /* for setting kernel sock attribute on TCP connection setup */
1311 #define GSO_MAX_SIZE 65536
1312 unsigned int gso_max_size;
1313 #define GSO_MAX_SEGS 65535
1314 u16 gso_max_segs;
1315
1316 #ifdef CONFIG_DCB
1317 /* Data Center Bridging netlink ops */
1318 const struct dcbnl_rtnl_ops *dcbnl_ops;
1319 #endif
1320 u8 num_tc;
1321 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1322 u8 prio_tc_map[TC_BITMASK + 1];
1323
1324 #if IS_ENABLED(CONFIG_FCOE)
1325 /* max exchange id for FCoE LRO by ddp */
1326 unsigned int fcoe_ddp_xid;
1327 #endif
1328 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1329 struct netprio_map __rcu *priomap;
1330 #endif
1331 /* phy device may attach itself for hardware timestamping */
1332 struct phy_device *phydev;
1333
1334 struct lock_class_key *qdisc_tx_busylock;
1335
1336 /* group the device belongs to */
1337 int group;
1338
1339 struct pm_qos_request pm_qos_req;
1340 };
1341 #define to_net_dev(d) container_of(d, struct net_device, dev)
1342
1343 #define NETDEV_ALIGN 32
1344
1345 static inline
1346 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1347 {
1348 return dev->prio_tc_map[prio & TC_BITMASK];
1349 }
1350
1351 static inline
1352 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1353 {
1354 if (tc >= dev->num_tc)
1355 return -EINVAL;
1356
1357 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1358 return 0;
1359 }
1360
1361 static inline
1362 void netdev_reset_tc(struct net_device *dev)
1363 {
1364 dev->num_tc = 0;
1365 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1366 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1367 }
1368
1369 static inline
1370 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1371 {
1372 if (tc >= dev->num_tc)
1373 return -EINVAL;
1374
1375 dev->tc_to_txq[tc].count = count;
1376 dev->tc_to_txq[tc].offset = offset;
1377 return 0;
1378 }
1379
1380 static inline
1381 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1382 {
1383 if (num_tc > TC_MAX_QUEUE)
1384 return -EINVAL;
1385
1386 dev->num_tc = num_tc;
1387 return 0;
1388 }
1389
1390 static inline
1391 int netdev_get_num_tc(struct net_device *dev)
1392 {
1393 return dev->num_tc;
1394 }
1395
1396 static inline
1397 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1398 unsigned int index)
1399 {
1400 return &dev->_tx[index];
1401 }
1402
1403 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1404 void (*f)(struct net_device *,
1405 struct netdev_queue *,
1406 void *),
1407 void *arg)
1408 {
1409 unsigned int i;
1410
1411 for (i = 0; i < dev->num_tx_queues; i++)
1412 f(dev, &dev->_tx[i], arg);
1413 }
1414
1415 extern struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1416 struct sk_buff *skb);
1417 extern u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
1418
1419 /*
1420 * Net namespace inlines
1421 */
1422 static inline
1423 struct net *dev_net(const struct net_device *dev)
1424 {
1425 return read_pnet(&dev->nd_net);
1426 }
1427
1428 static inline
1429 void dev_net_set(struct net_device *dev, struct net *net)
1430 {
1431 #ifdef CONFIG_NET_NS
1432 release_net(dev->nd_net);
1433 dev->nd_net = hold_net(net);
1434 #endif
1435 }
1436
1437 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1438 {
1439 #ifdef CONFIG_NET_DSA_TAG_DSA
1440 if (dev->dsa_ptr != NULL)
1441 return dsa_uses_dsa_tags(dev->dsa_ptr);
1442 #endif
1443
1444 return 0;
1445 }
1446
1447 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1448 {
1449 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1450 if (dev->dsa_ptr != NULL)
1451 return dsa_uses_trailer_tags(dev->dsa_ptr);
1452 #endif
1453
1454 return 0;
1455 }
1456
1457 /**
1458 * netdev_priv - access network device private data
1459 * @dev: network device
1460 *
1461 * Get network device private data
1462 */
1463 static inline void *netdev_priv(const struct net_device *dev)
1464 {
1465 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1466 }
1467
1468 /* Set the sysfs physical device reference for the network logical device
1469 * if set prior to registration will cause a symlink during initialization.
1470 */
1471 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1472
1473 /* Set the sysfs device type for the network logical device to allow
1474 * fin grained indentification of different network device types. For
1475 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1476 */
1477 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1478
1479 /* Default NAPI poll() weight
1480 * Device drivers are strongly advised to not use bigger value
1481 */
1482 #define NAPI_POLL_WEIGHT 64
1483
1484 /**
1485 * netif_napi_add - initialize a napi context
1486 * @dev: network device
1487 * @napi: napi context
1488 * @poll: polling function
1489 * @weight: default weight
1490 *
1491 * netif_napi_add() must be used to initialize a napi context prior to calling
1492 * *any* of the other napi related functions.
1493 */
1494 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1495 int (*poll)(struct napi_struct *, int), int weight);
1496
1497 /**
1498 * netif_napi_del - remove a napi context
1499 * @napi: napi context
1500 *
1501 * netif_napi_del() removes a napi context from the network device napi list
1502 */
1503 void netif_napi_del(struct napi_struct *napi);
1504
1505 struct napi_gro_cb {
1506 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1507 void *frag0;
1508
1509 /* Length of frag0. */
1510 unsigned int frag0_len;
1511
1512 /* This indicates where we are processing relative to skb->data. */
1513 int data_offset;
1514
1515 /* This is non-zero if the packet cannot be merged with the new skb. */
1516 int flush;
1517
1518 /* Number of segments aggregated. */
1519 u16 count;
1520
1521 /* This is non-zero if the packet may be of the same flow. */
1522 u8 same_flow;
1523
1524 /* Free the skb? */
1525 u8 free;
1526 #define NAPI_GRO_FREE 1
1527 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1528
1529 /* jiffies when first packet was created/queued */
1530 unsigned long age;
1531
1532 /* Used in ipv6_gro_receive() */
1533 int proto;
1534
1535 /* used in skb_gro_receive() slow path */
1536 struct sk_buff *last;
1537 };
1538
1539 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1540
1541 struct packet_type {
1542 __be16 type; /* This is really htons(ether_type). */
1543 struct net_device *dev; /* NULL is wildcarded here */
1544 int (*func) (struct sk_buff *,
1545 struct net_device *,
1546 struct packet_type *,
1547 struct net_device *);
1548 bool (*id_match)(struct packet_type *ptype,
1549 struct sock *sk);
1550 void *af_packet_priv;
1551 struct list_head list;
1552 };
1553
1554 struct offload_callbacks {
1555 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1556 netdev_features_t features);
1557 int (*gso_send_check)(struct sk_buff *skb);
1558 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1559 struct sk_buff *skb);
1560 int (*gro_complete)(struct sk_buff *skb);
1561 };
1562
1563 struct packet_offload {
1564 __be16 type; /* This is really htons(ether_type). */
1565 struct offload_callbacks callbacks;
1566 struct list_head list;
1567 };
1568
1569 #include <linux/notifier.h>
1570
1571 /* netdevice notifier chain. Please remember to update the rtnetlink
1572 * notification exclusion list in rtnetlink_event() when adding new
1573 * types.
1574 */
1575 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1576 #define NETDEV_DOWN 0x0002
1577 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1578 detected a hardware crash and restarted
1579 - we can use this eg to kick tcp sessions
1580 once done */
1581 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1582 #define NETDEV_REGISTER 0x0005
1583 #define NETDEV_UNREGISTER 0x0006
1584 #define NETDEV_CHANGEMTU 0x0007
1585 #define NETDEV_CHANGEADDR 0x0008
1586 #define NETDEV_GOING_DOWN 0x0009
1587 #define NETDEV_CHANGENAME 0x000A
1588 #define NETDEV_FEAT_CHANGE 0x000B
1589 #define NETDEV_BONDING_FAILOVER 0x000C
1590 #define NETDEV_PRE_UP 0x000D
1591 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1592 #define NETDEV_POST_TYPE_CHANGE 0x000F
1593 #define NETDEV_POST_INIT 0x0010
1594 #define NETDEV_UNREGISTER_FINAL 0x0011
1595 #define NETDEV_RELEASE 0x0012
1596 #define NETDEV_NOTIFY_PEERS 0x0013
1597 #define NETDEV_JOIN 0x0014
1598 #define NETDEV_CHANGEUPPER 0x0015
1599
1600 extern int register_netdevice_notifier(struct notifier_block *nb);
1601 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1602
1603 struct netdev_notifier_info {
1604 struct net_device *dev;
1605 };
1606
1607 struct netdev_notifier_change_info {
1608 struct netdev_notifier_info info; /* must be first */
1609 unsigned int flags_changed;
1610 };
1611
1612 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1613 struct net_device *dev)
1614 {
1615 info->dev = dev;
1616 }
1617
1618 static inline struct net_device *
1619 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1620 {
1621 return info->dev;
1622 }
1623
1624 extern int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
1625 struct netdev_notifier_info *info);
1626 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1627
1628
1629 extern rwlock_t dev_base_lock; /* Device list lock */
1630
1631 extern seqcount_t devnet_rename_seq; /* Device rename seq */
1632
1633
1634 #define for_each_netdev(net, d) \
1635 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1636 #define for_each_netdev_reverse(net, d) \
1637 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1638 #define for_each_netdev_rcu(net, d) \
1639 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1640 #define for_each_netdev_safe(net, d, n) \
1641 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1642 #define for_each_netdev_continue(net, d) \
1643 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1644 #define for_each_netdev_continue_rcu(net, d) \
1645 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1646 #define for_each_netdev_in_bond_rcu(bond, slave) \
1647 for_each_netdev_rcu(&init_net, slave) \
1648 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1649 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1650
1651 static inline struct net_device *next_net_device(struct net_device *dev)
1652 {
1653 struct list_head *lh;
1654 struct net *net;
1655
1656 net = dev_net(dev);
1657 lh = dev->dev_list.next;
1658 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1659 }
1660
1661 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1662 {
1663 struct list_head *lh;
1664 struct net *net;
1665
1666 net = dev_net(dev);
1667 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1668 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1669 }
1670
1671 static inline struct net_device *first_net_device(struct net *net)
1672 {
1673 return list_empty(&net->dev_base_head) ? NULL :
1674 net_device_entry(net->dev_base_head.next);
1675 }
1676
1677 static inline struct net_device *first_net_device_rcu(struct net *net)
1678 {
1679 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1680
1681 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1682 }
1683
1684 extern int netdev_boot_setup_check(struct net_device *dev);
1685 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1686 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1687 const char *hwaddr);
1688 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1689 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1690 extern void dev_add_pack(struct packet_type *pt);
1691 extern void dev_remove_pack(struct packet_type *pt);
1692 extern void __dev_remove_pack(struct packet_type *pt);
1693 extern void dev_add_offload(struct packet_offload *po);
1694 extern void dev_remove_offload(struct packet_offload *po);
1695 extern void __dev_remove_offload(struct packet_offload *po);
1696
1697 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1698 unsigned short mask);
1699 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1700 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1701 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1702 extern int dev_alloc_name(struct net_device *dev, const char *name);
1703 extern int dev_open(struct net_device *dev);
1704 extern int dev_close(struct net_device *dev);
1705 extern void dev_disable_lro(struct net_device *dev);
1706 extern int dev_loopback_xmit(struct sk_buff *newskb);
1707 extern int dev_queue_xmit(struct sk_buff *skb);
1708 extern int register_netdevice(struct net_device *dev);
1709 extern void unregister_netdevice_queue(struct net_device *dev,
1710 struct list_head *head);
1711 extern void unregister_netdevice_many(struct list_head *head);
1712 static inline void unregister_netdevice(struct net_device *dev)
1713 {
1714 unregister_netdevice_queue(dev, NULL);
1715 }
1716
1717 extern int netdev_refcnt_read(const struct net_device *dev);
1718 extern void free_netdev(struct net_device *dev);
1719 extern void synchronize_net(void);
1720 extern int init_dummy_netdev(struct net_device *dev);
1721
1722 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1723 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1724 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1725 extern int dev_restart(struct net_device *dev);
1726 #ifdef CONFIG_NETPOLL_TRAP
1727 extern int netpoll_trap(void);
1728 #endif
1729 extern int skb_gro_receive(struct sk_buff **head,
1730 struct sk_buff *skb);
1731
1732 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1733 {
1734 return NAPI_GRO_CB(skb)->data_offset;
1735 }
1736
1737 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1738 {
1739 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1740 }
1741
1742 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1743 {
1744 NAPI_GRO_CB(skb)->data_offset += len;
1745 }
1746
1747 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1748 unsigned int offset)
1749 {
1750 return NAPI_GRO_CB(skb)->frag0 + offset;
1751 }
1752
1753 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1754 {
1755 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1756 }
1757
1758 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1759 unsigned int offset)
1760 {
1761 if (!pskb_may_pull(skb, hlen))
1762 return NULL;
1763
1764 NAPI_GRO_CB(skb)->frag0 = NULL;
1765 NAPI_GRO_CB(skb)->frag0_len = 0;
1766 return skb->data + offset;
1767 }
1768
1769 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1770 {
1771 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1772 }
1773
1774 static inline void *skb_gro_network_header(struct sk_buff *skb)
1775 {
1776 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1777 skb_network_offset(skb);
1778 }
1779
1780 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1781 unsigned short type,
1782 const void *daddr, const void *saddr,
1783 unsigned int len)
1784 {
1785 if (!dev->header_ops || !dev->header_ops->create)
1786 return 0;
1787
1788 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1789 }
1790
1791 static inline int dev_parse_header(const struct sk_buff *skb,
1792 unsigned char *haddr)
1793 {
1794 const struct net_device *dev = skb->dev;
1795
1796 if (!dev->header_ops || !dev->header_ops->parse)
1797 return 0;
1798 return dev->header_ops->parse(skb, haddr);
1799 }
1800
1801 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1802 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1803 static inline int unregister_gifconf(unsigned int family)
1804 {
1805 return register_gifconf(family, NULL);
1806 }
1807
1808 #ifdef CONFIG_NET_FLOW_LIMIT
1809 #define FLOW_LIMIT_HISTORY (1 << 8) /* must be ^2 */
1810 struct sd_flow_limit {
1811 u64 count;
1812 unsigned int num_buckets;
1813 unsigned int history_head;
1814 u16 history[FLOW_LIMIT_HISTORY];
1815 u8 buckets[];
1816 };
1817
1818 extern int netdev_flow_limit_table_len;
1819 #endif /* CONFIG_NET_FLOW_LIMIT */
1820
1821 /*
1822 * Incoming packets are placed on per-cpu queues
1823 */
1824 struct softnet_data {
1825 struct Qdisc *output_queue;
1826 struct Qdisc **output_queue_tailp;
1827 struct list_head poll_list;
1828 struct sk_buff *completion_queue;
1829 struct sk_buff_head process_queue;
1830
1831 /* stats */
1832 unsigned int processed;
1833 unsigned int time_squeeze;
1834 unsigned int cpu_collision;
1835 unsigned int received_rps;
1836
1837 #ifdef CONFIG_RPS
1838 struct softnet_data *rps_ipi_list;
1839
1840 /* Elements below can be accessed between CPUs for RPS */
1841 struct call_single_data csd ____cacheline_aligned_in_smp;
1842 struct softnet_data *rps_ipi_next;
1843 unsigned int cpu;
1844 unsigned int input_queue_head;
1845 unsigned int input_queue_tail;
1846 #endif
1847 unsigned int dropped;
1848 struct sk_buff_head input_pkt_queue;
1849 struct napi_struct backlog;
1850
1851 #ifdef CONFIG_NET_FLOW_LIMIT
1852 struct sd_flow_limit *flow_limit;
1853 #endif
1854 };
1855
1856 static inline void input_queue_head_incr(struct softnet_data *sd)
1857 {
1858 #ifdef CONFIG_RPS
1859 sd->input_queue_head++;
1860 #endif
1861 }
1862
1863 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1864 unsigned int *qtail)
1865 {
1866 #ifdef CONFIG_RPS
1867 *qtail = ++sd->input_queue_tail;
1868 #endif
1869 }
1870
1871 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1872
1873 extern void __netif_schedule(struct Qdisc *q);
1874
1875 static inline void netif_schedule_queue(struct netdev_queue *txq)
1876 {
1877 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1878 __netif_schedule(txq->qdisc);
1879 }
1880
1881 static inline void netif_tx_schedule_all(struct net_device *dev)
1882 {
1883 unsigned int i;
1884
1885 for (i = 0; i < dev->num_tx_queues; i++)
1886 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1887 }
1888
1889 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1890 {
1891 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1892 }
1893
1894 /**
1895 * netif_start_queue - allow transmit
1896 * @dev: network device
1897 *
1898 * Allow upper layers to call the device hard_start_xmit routine.
1899 */
1900 static inline void netif_start_queue(struct net_device *dev)
1901 {
1902 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1903 }
1904
1905 static inline void netif_tx_start_all_queues(struct net_device *dev)
1906 {
1907 unsigned int i;
1908
1909 for (i = 0; i < dev->num_tx_queues; i++) {
1910 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1911 netif_tx_start_queue(txq);
1912 }
1913 }
1914
1915 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1916 {
1917 #ifdef CONFIG_NETPOLL_TRAP
1918 if (netpoll_trap()) {
1919 netif_tx_start_queue(dev_queue);
1920 return;
1921 }
1922 #endif
1923 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1924 __netif_schedule(dev_queue->qdisc);
1925 }
1926
1927 /**
1928 * netif_wake_queue - restart transmit
1929 * @dev: network device
1930 *
1931 * Allow upper layers to call the device hard_start_xmit routine.
1932 * Used for flow control when transmit resources are available.
1933 */
1934 static inline void netif_wake_queue(struct net_device *dev)
1935 {
1936 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1937 }
1938
1939 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1940 {
1941 unsigned int i;
1942
1943 for (i = 0; i < dev->num_tx_queues; i++) {
1944 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1945 netif_tx_wake_queue(txq);
1946 }
1947 }
1948
1949 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1950 {
1951 if (WARN_ON(!dev_queue)) {
1952 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1953 return;
1954 }
1955 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1956 }
1957
1958 /**
1959 * netif_stop_queue - stop transmitted packets
1960 * @dev: network device
1961 *
1962 * Stop upper layers calling the device hard_start_xmit routine.
1963 * Used for flow control when transmit resources are unavailable.
1964 */
1965 static inline void netif_stop_queue(struct net_device *dev)
1966 {
1967 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1968 }
1969
1970 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1971 {
1972 unsigned int i;
1973
1974 for (i = 0; i < dev->num_tx_queues; i++) {
1975 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1976 netif_tx_stop_queue(txq);
1977 }
1978 }
1979
1980 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1981 {
1982 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1983 }
1984
1985 /**
1986 * netif_queue_stopped - test if transmit queue is flowblocked
1987 * @dev: network device
1988 *
1989 * Test if transmit queue on device is currently unable to send.
1990 */
1991 static inline bool netif_queue_stopped(const struct net_device *dev)
1992 {
1993 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1994 }
1995
1996 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
1997 {
1998 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1999 }
2000
2001 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2002 {
2003 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2004 }
2005
2006 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2007 unsigned int bytes)
2008 {
2009 #ifdef CONFIG_BQL
2010 dql_queued(&dev_queue->dql, bytes);
2011
2012 if (likely(dql_avail(&dev_queue->dql) >= 0))
2013 return;
2014
2015 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2016
2017 /*
2018 * The XOFF flag must be set before checking the dql_avail below,
2019 * because in netdev_tx_completed_queue we update the dql_completed
2020 * before checking the XOFF flag.
2021 */
2022 smp_mb();
2023
2024 /* check again in case another CPU has just made room avail */
2025 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2026 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2027 #endif
2028 }
2029
2030 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2031 {
2032 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2033 }
2034
2035 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2036 unsigned int pkts, unsigned int bytes)
2037 {
2038 #ifdef CONFIG_BQL
2039 if (unlikely(!bytes))
2040 return;
2041
2042 dql_completed(&dev_queue->dql, bytes);
2043
2044 /*
2045 * Without the memory barrier there is a small possiblity that
2046 * netdev_tx_sent_queue will miss the update and cause the queue to
2047 * be stopped forever
2048 */
2049 smp_mb();
2050
2051 if (dql_avail(&dev_queue->dql) < 0)
2052 return;
2053
2054 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2055 netif_schedule_queue(dev_queue);
2056 #endif
2057 }
2058
2059 static inline void netdev_completed_queue(struct net_device *dev,
2060 unsigned int pkts, unsigned int bytes)
2061 {
2062 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2063 }
2064
2065 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2066 {
2067 #ifdef CONFIG_BQL
2068 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2069 dql_reset(&q->dql);
2070 #endif
2071 }
2072
2073 static inline void netdev_reset_queue(struct net_device *dev_queue)
2074 {
2075 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2076 }
2077
2078 /**
2079 * netif_running - test if up
2080 * @dev: network device
2081 *
2082 * Test if the device has been brought up.
2083 */
2084 static inline bool netif_running(const struct net_device *dev)
2085 {
2086 return test_bit(__LINK_STATE_START, &dev->state);
2087 }
2088
2089 /*
2090 * Routines to manage the subqueues on a device. We only need start
2091 * stop, and a check if it's stopped. All other device management is
2092 * done at the overall netdevice level.
2093 * Also test the device if we're multiqueue.
2094 */
2095
2096 /**
2097 * netif_start_subqueue - allow sending packets on subqueue
2098 * @dev: network device
2099 * @queue_index: sub queue index
2100 *
2101 * Start individual transmit queue of a device with multiple transmit queues.
2102 */
2103 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2104 {
2105 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2106
2107 netif_tx_start_queue(txq);
2108 }
2109
2110 /**
2111 * netif_stop_subqueue - stop sending packets on subqueue
2112 * @dev: network device
2113 * @queue_index: sub queue index
2114 *
2115 * Stop individual transmit queue of a device with multiple transmit queues.
2116 */
2117 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2118 {
2119 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2120 #ifdef CONFIG_NETPOLL_TRAP
2121 if (netpoll_trap())
2122 return;
2123 #endif
2124 netif_tx_stop_queue(txq);
2125 }
2126
2127 /**
2128 * netif_subqueue_stopped - test status of subqueue
2129 * @dev: network device
2130 * @queue_index: sub queue index
2131 *
2132 * Check individual transmit queue of a device with multiple transmit queues.
2133 */
2134 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2135 u16 queue_index)
2136 {
2137 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2138
2139 return netif_tx_queue_stopped(txq);
2140 }
2141
2142 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2143 struct sk_buff *skb)
2144 {
2145 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2146 }
2147
2148 /**
2149 * netif_wake_subqueue - allow sending packets on subqueue
2150 * @dev: network device
2151 * @queue_index: sub queue index
2152 *
2153 * Resume individual transmit queue of a device with multiple transmit queues.
2154 */
2155 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2156 {
2157 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2158 #ifdef CONFIG_NETPOLL_TRAP
2159 if (netpoll_trap())
2160 return;
2161 #endif
2162 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2163 __netif_schedule(txq->qdisc);
2164 }
2165
2166 #ifdef CONFIG_XPS
2167 extern int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask,
2168 u16 index);
2169 #else
2170 static inline int netif_set_xps_queue(struct net_device *dev,
2171 struct cpumask *mask,
2172 u16 index)
2173 {
2174 return 0;
2175 }
2176 #endif
2177
2178 /*
2179 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2180 * as a distribution range limit for the returned value.
2181 */
2182 static inline u16 skb_tx_hash(const struct net_device *dev,
2183 const struct sk_buff *skb)
2184 {
2185 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2186 }
2187
2188 /**
2189 * netif_is_multiqueue - test if device has multiple transmit queues
2190 * @dev: network device
2191 *
2192 * Check if device has multiple transmit queues
2193 */
2194 static inline bool netif_is_multiqueue(const struct net_device *dev)
2195 {
2196 return dev->num_tx_queues > 1;
2197 }
2198
2199 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2200 unsigned int txq);
2201
2202 #ifdef CONFIG_RPS
2203 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2204 unsigned int rxq);
2205 #else
2206 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2207 unsigned int rxq)
2208 {
2209 return 0;
2210 }
2211 #endif
2212
2213 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2214 const struct net_device *from_dev)
2215 {
2216 int err;
2217
2218 err = netif_set_real_num_tx_queues(to_dev,
2219 from_dev->real_num_tx_queues);
2220 if (err)
2221 return err;
2222 #ifdef CONFIG_RPS
2223 return netif_set_real_num_rx_queues(to_dev,
2224 from_dev->real_num_rx_queues);
2225 #else
2226 return 0;
2227 #endif
2228 }
2229
2230 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2231 extern int netif_get_num_default_rss_queues(void);
2232
2233 /* Use this variant when it is known for sure that it
2234 * is executing from hardware interrupt context or with hardware interrupts
2235 * disabled.
2236 */
2237 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2238
2239 /* Use this variant in places where it could be invoked
2240 * from either hardware interrupt or other context, with hardware interrupts
2241 * either disabled or enabled.
2242 */
2243 extern void dev_kfree_skb_any(struct sk_buff *skb);
2244
2245 extern int netif_rx(struct sk_buff *skb);
2246 extern int netif_rx_ni(struct sk_buff *skb);
2247 extern int netif_receive_skb(struct sk_buff *skb);
2248 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2249 struct sk_buff *skb);
2250 extern void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2251 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2252 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2253
2254 static inline void napi_free_frags(struct napi_struct *napi)
2255 {
2256 kfree_skb(napi->skb);
2257 napi->skb = NULL;
2258 }
2259
2260 extern int netdev_rx_handler_register(struct net_device *dev,
2261 rx_handler_func_t *rx_handler,
2262 void *rx_handler_data);
2263 extern void netdev_rx_handler_unregister(struct net_device *dev);
2264
2265 extern bool dev_valid_name(const char *name);
2266 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2267 extern int dev_ethtool(struct net *net, struct ifreq *);
2268 extern unsigned int dev_get_flags(const struct net_device *);
2269 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2270 extern int dev_change_flags(struct net_device *, unsigned int);
2271 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2272 extern int dev_change_name(struct net_device *, const char *);
2273 extern int dev_set_alias(struct net_device *, const char *, size_t);
2274 extern int dev_change_net_namespace(struct net_device *,
2275 struct net *, const char *);
2276 extern int dev_set_mtu(struct net_device *, int);
2277 extern void dev_set_group(struct net_device *, int);
2278 extern int dev_set_mac_address(struct net_device *,
2279 struct sockaddr *);
2280 extern int dev_change_carrier(struct net_device *,
2281 bool new_carrier);
2282 extern int dev_hard_start_xmit(struct sk_buff *skb,
2283 struct net_device *dev,
2284 struct netdev_queue *txq);
2285 extern int dev_forward_skb(struct net_device *dev,
2286 struct sk_buff *skb);
2287
2288 extern int netdev_budget;
2289
2290 /* Called by rtnetlink.c:rtnl_unlock() */
2291 extern void netdev_run_todo(void);
2292
2293 /**
2294 * dev_put - release reference to device
2295 * @dev: network device
2296 *
2297 * Release reference to device to allow it to be freed.
2298 */
2299 static inline void dev_put(struct net_device *dev)
2300 {
2301 this_cpu_dec(*dev->pcpu_refcnt);
2302 }
2303
2304 /**
2305 * dev_hold - get reference to device
2306 * @dev: network device
2307 *
2308 * Hold reference to device to keep it from being freed.
2309 */
2310 static inline void dev_hold(struct net_device *dev)
2311 {
2312 this_cpu_inc(*dev->pcpu_refcnt);
2313 }
2314
2315 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2316 * and _off may be called from IRQ context, but it is caller
2317 * who is responsible for serialization of these calls.
2318 *
2319 * The name carrier is inappropriate, these functions should really be
2320 * called netif_lowerlayer_*() because they represent the state of any
2321 * kind of lower layer not just hardware media.
2322 */
2323
2324 extern void linkwatch_init_dev(struct net_device *dev);
2325 extern void linkwatch_fire_event(struct net_device *dev);
2326 extern void linkwatch_forget_dev(struct net_device *dev);
2327
2328 /**
2329 * netif_carrier_ok - test if carrier present
2330 * @dev: network device
2331 *
2332 * Check if carrier is present on device
2333 */
2334 static inline bool netif_carrier_ok(const struct net_device *dev)
2335 {
2336 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2337 }
2338
2339 extern unsigned long dev_trans_start(struct net_device *dev);
2340
2341 extern void __netdev_watchdog_up(struct net_device *dev);
2342
2343 extern void netif_carrier_on(struct net_device *dev);
2344
2345 extern void netif_carrier_off(struct net_device *dev);
2346
2347 /**
2348 * netif_dormant_on - mark device as dormant.
2349 * @dev: network device
2350 *
2351 * Mark device as dormant (as per RFC2863).
2352 *
2353 * The dormant state indicates that the relevant interface is not
2354 * actually in a condition to pass packets (i.e., it is not 'up') but is
2355 * in a "pending" state, waiting for some external event. For "on-
2356 * demand" interfaces, this new state identifies the situation where the
2357 * interface is waiting for events to place it in the up state.
2358 *
2359 */
2360 static inline void netif_dormant_on(struct net_device *dev)
2361 {
2362 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2363 linkwatch_fire_event(dev);
2364 }
2365
2366 /**
2367 * netif_dormant_off - set device as not dormant.
2368 * @dev: network device
2369 *
2370 * Device is not in dormant state.
2371 */
2372 static inline void netif_dormant_off(struct net_device *dev)
2373 {
2374 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2375 linkwatch_fire_event(dev);
2376 }
2377
2378 /**
2379 * netif_dormant - test if carrier present
2380 * @dev: network device
2381 *
2382 * Check if carrier is present on device
2383 */
2384 static inline bool netif_dormant(const struct net_device *dev)
2385 {
2386 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2387 }
2388
2389
2390 /**
2391 * netif_oper_up - test if device is operational
2392 * @dev: network device
2393 *
2394 * Check if carrier is operational
2395 */
2396 static inline bool netif_oper_up(const struct net_device *dev)
2397 {
2398 return (dev->operstate == IF_OPER_UP ||
2399 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2400 }
2401
2402 /**
2403 * netif_device_present - is device available or removed
2404 * @dev: network device
2405 *
2406 * Check if device has not been removed from system.
2407 */
2408 static inline bool netif_device_present(struct net_device *dev)
2409 {
2410 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2411 }
2412
2413 extern void netif_device_detach(struct net_device *dev);
2414
2415 extern void netif_device_attach(struct net_device *dev);
2416
2417 /*
2418 * Network interface message level settings
2419 */
2420
2421 enum {
2422 NETIF_MSG_DRV = 0x0001,
2423 NETIF_MSG_PROBE = 0x0002,
2424 NETIF_MSG_LINK = 0x0004,
2425 NETIF_MSG_TIMER = 0x0008,
2426 NETIF_MSG_IFDOWN = 0x0010,
2427 NETIF_MSG_IFUP = 0x0020,
2428 NETIF_MSG_RX_ERR = 0x0040,
2429 NETIF_MSG_TX_ERR = 0x0080,
2430 NETIF_MSG_TX_QUEUED = 0x0100,
2431 NETIF_MSG_INTR = 0x0200,
2432 NETIF_MSG_TX_DONE = 0x0400,
2433 NETIF_MSG_RX_STATUS = 0x0800,
2434 NETIF_MSG_PKTDATA = 0x1000,
2435 NETIF_MSG_HW = 0x2000,
2436 NETIF_MSG_WOL = 0x4000,
2437 };
2438
2439 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2440 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2441 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2442 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2443 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2444 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2445 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2446 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2447 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2448 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2449 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2450 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2451 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2452 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2453 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2454
2455 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2456 {
2457 /* use default */
2458 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2459 return default_msg_enable_bits;
2460 if (debug_value == 0) /* no output */
2461 return 0;
2462 /* set low N bits */
2463 return (1 << debug_value) - 1;
2464 }
2465
2466 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2467 {
2468 spin_lock(&txq->_xmit_lock);
2469 txq->xmit_lock_owner = cpu;
2470 }
2471
2472 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2473 {
2474 spin_lock_bh(&txq->_xmit_lock);
2475 txq->xmit_lock_owner = smp_processor_id();
2476 }
2477
2478 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2479 {
2480 bool ok = spin_trylock(&txq->_xmit_lock);
2481 if (likely(ok))
2482 txq->xmit_lock_owner = smp_processor_id();
2483 return ok;
2484 }
2485
2486 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2487 {
2488 txq->xmit_lock_owner = -1;
2489 spin_unlock(&txq->_xmit_lock);
2490 }
2491
2492 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2493 {
2494 txq->xmit_lock_owner = -1;
2495 spin_unlock_bh(&txq->_xmit_lock);
2496 }
2497
2498 static inline void txq_trans_update(struct netdev_queue *txq)
2499 {
2500 if (txq->xmit_lock_owner != -1)
2501 txq->trans_start = jiffies;
2502 }
2503
2504 /**
2505 * netif_tx_lock - grab network device transmit lock
2506 * @dev: network device
2507 *
2508 * Get network device transmit lock
2509 */
2510 static inline void netif_tx_lock(struct net_device *dev)
2511 {
2512 unsigned int i;
2513 int cpu;
2514
2515 spin_lock(&dev->tx_global_lock);
2516 cpu = smp_processor_id();
2517 for (i = 0; i < dev->num_tx_queues; i++) {
2518 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2519
2520 /* We are the only thread of execution doing a
2521 * freeze, but we have to grab the _xmit_lock in
2522 * order to synchronize with threads which are in
2523 * the ->hard_start_xmit() handler and already
2524 * checked the frozen bit.
2525 */
2526 __netif_tx_lock(txq, cpu);
2527 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2528 __netif_tx_unlock(txq);
2529 }
2530 }
2531
2532 static inline void netif_tx_lock_bh(struct net_device *dev)
2533 {
2534 local_bh_disable();
2535 netif_tx_lock(dev);
2536 }
2537
2538 static inline void netif_tx_unlock(struct net_device *dev)
2539 {
2540 unsigned int i;
2541
2542 for (i = 0; i < dev->num_tx_queues; i++) {
2543 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2544
2545 /* No need to grab the _xmit_lock here. If the
2546 * queue is not stopped for another reason, we
2547 * force a schedule.
2548 */
2549 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2550 netif_schedule_queue(txq);
2551 }
2552 spin_unlock(&dev->tx_global_lock);
2553 }
2554
2555 static inline void netif_tx_unlock_bh(struct net_device *dev)
2556 {
2557 netif_tx_unlock(dev);
2558 local_bh_enable();
2559 }
2560
2561 #define HARD_TX_LOCK(dev, txq, cpu) { \
2562 if ((dev->features & NETIF_F_LLTX) == 0) { \
2563 __netif_tx_lock(txq, cpu); \
2564 } \
2565 }
2566
2567 #define HARD_TX_UNLOCK(dev, txq) { \
2568 if ((dev->features & NETIF_F_LLTX) == 0) { \
2569 __netif_tx_unlock(txq); \
2570 } \
2571 }
2572
2573 static inline void netif_tx_disable(struct net_device *dev)
2574 {
2575 unsigned int i;
2576 int cpu;
2577
2578 local_bh_disable();
2579 cpu = smp_processor_id();
2580 for (i = 0; i < dev->num_tx_queues; i++) {
2581 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2582
2583 __netif_tx_lock(txq, cpu);
2584 netif_tx_stop_queue(txq);
2585 __netif_tx_unlock(txq);
2586 }
2587 local_bh_enable();
2588 }
2589
2590 static inline void netif_addr_lock(struct net_device *dev)
2591 {
2592 spin_lock(&dev->addr_list_lock);
2593 }
2594
2595 static inline void netif_addr_lock_nested(struct net_device *dev)
2596 {
2597 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2598 }
2599
2600 static inline void netif_addr_lock_bh(struct net_device *dev)
2601 {
2602 spin_lock_bh(&dev->addr_list_lock);
2603 }
2604
2605 static inline void netif_addr_unlock(struct net_device *dev)
2606 {
2607 spin_unlock(&dev->addr_list_lock);
2608 }
2609
2610 static inline void netif_addr_unlock_bh(struct net_device *dev)
2611 {
2612 spin_unlock_bh(&dev->addr_list_lock);
2613 }
2614
2615 /*
2616 * dev_addrs walker. Should be used only for read access. Call with
2617 * rcu_read_lock held.
2618 */
2619 #define for_each_dev_addr(dev, ha) \
2620 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2621
2622 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2623
2624 extern void ether_setup(struct net_device *dev);
2625
2626 /* Support for loadable net-drivers */
2627 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2628 void (*setup)(struct net_device *),
2629 unsigned int txqs, unsigned int rxqs);
2630 #define alloc_netdev(sizeof_priv, name, setup) \
2631 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2632
2633 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2634 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2635
2636 extern int register_netdev(struct net_device *dev);
2637 extern void unregister_netdev(struct net_device *dev);
2638
2639 /* General hardware address lists handling functions */
2640 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2641 struct netdev_hw_addr_list *from_list,
2642 int addr_len, unsigned char addr_type);
2643 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2644 struct netdev_hw_addr_list *from_list,
2645 int addr_len, unsigned char addr_type);
2646 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2647 struct netdev_hw_addr_list *from_list,
2648 int addr_len);
2649 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2650 struct netdev_hw_addr_list *from_list,
2651 int addr_len);
2652 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2653 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2654
2655 /* Functions used for device addresses handling */
2656 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2657 unsigned char addr_type);
2658 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2659 unsigned char addr_type);
2660 extern int dev_addr_add_multiple(struct net_device *to_dev,
2661 struct net_device *from_dev,
2662 unsigned char addr_type);
2663 extern int dev_addr_del_multiple(struct net_device *to_dev,
2664 struct net_device *from_dev,
2665 unsigned char addr_type);
2666 extern void dev_addr_flush(struct net_device *dev);
2667 extern int dev_addr_init(struct net_device *dev);
2668
2669 /* Functions used for unicast addresses handling */
2670 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2671 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2672 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2673 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2674 extern int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2675 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2676 extern void dev_uc_flush(struct net_device *dev);
2677 extern void dev_uc_init(struct net_device *dev);
2678
2679 /* Functions used for multicast addresses handling */
2680 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2681 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2682 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2683 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2684 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2685 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2686 extern int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2687 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2688 extern void dev_mc_flush(struct net_device *dev);
2689 extern void dev_mc_init(struct net_device *dev);
2690
2691 /* Functions used for secondary unicast and multicast support */
2692 extern void dev_set_rx_mode(struct net_device *dev);
2693 extern void __dev_set_rx_mode(struct net_device *dev);
2694 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2695 extern int dev_set_allmulti(struct net_device *dev, int inc);
2696 extern void netdev_state_change(struct net_device *dev);
2697 extern void netdev_notify_peers(struct net_device *dev);
2698 extern void netdev_features_change(struct net_device *dev);
2699 /* Load a device via the kmod */
2700 extern void dev_load(struct net *net, const char *name);
2701 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2702 struct rtnl_link_stats64 *storage);
2703 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2704 const struct net_device_stats *netdev_stats);
2705
2706 extern int netdev_max_backlog;
2707 extern int netdev_tstamp_prequeue;
2708 extern int weight_p;
2709 extern int bpf_jit_enable;
2710
2711 extern bool netdev_has_upper_dev(struct net_device *dev,
2712 struct net_device *upper_dev);
2713 extern bool netdev_has_any_upper_dev(struct net_device *dev);
2714 extern struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2715 extern struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2716 extern int netdev_upper_dev_link(struct net_device *dev,
2717 struct net_device *upper_dev);
2718 extern int netdev_master_upper_dev_link(struct net_device *dev,
2719 struct net_device *upper_dev);
2720 extern void netdev_upper_dev_unlink(struct net_device *dev,
2721 struct net_device *upper_dev);
2722 extern int skb_checksum_help(struct sk_buff *skb);
2723 extern struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2724 netdev_features_t features, bool tx_path);
2725 extern struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2726 netdev_features_t features);
2727
2728 static inline
2729 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2730 {
2731 return __skb_gso_segment(skb, features, true);
2732 }
2733 __be16 skb_network_protocol(struct sk_buff *skb);
2734
2735 static inline bool can_checksum_protocol(netdev_features_t features,
2736 __be16 protocol)
2737 {
2738 return ((features & NETIF_F_GEN_CSUM) ||
2739 ((features & NETIF_F_V4_CSUM) &&
2740 protocol == htons(ETH_P_IP)) ||
2741 ((features & NETIF_F_V6_CSUM) &&
2742 protocol == htons(ETH_P_IPV6)) ||
2743 ((features & NETIF_F_FCOE_CRC) &&
2744 protocol == htons(ETH_P_FCOE)));
2745 }
2746
2747 #ifdef CONFIG_BUG
2748 extern void netdev_rx_csum_fault(struct net_device *dev);
2749 #else
2750 static inline void netdev_rx_csum_fault(struct net_device *dev)
2751 {
2752 }
2753 #endif
2754 /* rx skb timestamps */
2755 extern void net_enable_timestamp(void);
2756 extern void net_disable_timestamp(void);
2757
2758 #ifdef CONFIG_PROC_FS
2759 extern int __init dev_proc_init(void);
2760 #else
2761 #define dev_proc_init() 0
2762 #endif
2763
2764 extern int netdev_class_create_file(struct class_attribute *class_attr);
2765 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2766
2767 extern struct kobj_ns_type_operations net_ns_type_operations;
2768
2769 extern const char *netdev_drivername(const struct net_device *dev);
2770
2771 extern void linkwatch_run_queue(void);
2772
2773 static inline netdev_features_t netdev_get_wanted_features(
2774 struct net_device *dev)
2775 {
2776 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2777 }
2778 netdev_features_t netdev_increment_features(netdev_features_t all,
2779 netdev_features_t one, netdev_features_t mask);
2780
2781 /* Allow TSO being used on stacked device :
2782 * Performing the GSO segmentation before last device
2783 * is a performance improvement.
2784 */
2785 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
2786 netdev_features_t mask)
2787 {
2788 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
2789 }
2790
2791 int __netdev_update_features(struct net_device *dev);
2792 void netdev_update_features(struct net_device *dev);
2793 void netdev_change_features(struct net_device *dev);
2794
2795 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2796 struct net_device *dev);
2797
2798 netdev_features_t netif_skb_features(struct sk_buff *skb);
2799
2800 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2801 {
2802 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2803
2804 /* check flags correspondence */
2805 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2806 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2807 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2808 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2809 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2810 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2811
2812 return (features & feature) == feature;
2813 }
2814
2815 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2816 {
2817 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2818 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2819 }
2820
2821 static inline bool netif_needs_gso(struct sk_buff *skb,
2822 netdev_features_t features)
2823 {
2824 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2825 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2826 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2827 }
2828
2829 static inline void netif_set_gso_max_size(struct net_device *dev,
2830 unsigned int size)
2831 {
2832 dev->gso_max_size = size;
2833 }
2834
2835 static inline bool netif_is_bond_master(struct net_device *dev)
2836 {
2837 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
2838 }
2839
2840 static inline bool netif_is_bond_slave(struct net_device *dev)
2841 {
2842 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2843 }
2844
2845 static inline bool netif_supports_nofcs(struct net_device *dev)
2846 {
2847 return dev->priv_flags & IFF_SUPP_NOFCS;
2848 }
2849
2850 extern struct pernet_operations __net_initdata loopback_net_ops;
2851
2852 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2853
2854 /* netdev_printk helpers, similar to dev_printk */
2855
2856 static inline const char *netdev_name(const struct net_device *dev)
2857 {
2858 if (dev->reg_state != NETREG_REGISTERED)
2859 return "(unregistered net_device)";
2860 return dev->name;
2861 }
2862
2863 extern __printf(3, 4)
2864 int netdev_printk(const char *level, const struct net_device *dev,
2865 const char *format, ...);
2866 extern __printf(2, 3)
2867 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2868 extern __printf(2, 3)
2869 int netdev_alert(const struct net_device *dev, const char *format, ...);
2870 extern __printf(2, 3)
2871 int netdev_crit(const struct net_device *dev, const char *format, ...);
2872 extern __printf(2, 3)
2873 int netdev_err(const struct net_device *dev, const char *format, ...);
2874 extern __printf(2, 3)
2875 int netdev_warn(const struct net_device *dev, const char *format, ...);
2876 extern __printf(2, 3)
2877 int netdev_notice(const struct net_device *dev, const char *format, ...);
2878 extern __printf(2, 3)
2879 int netdev_info(const struct net_device *dev, const char *format, ...);
2880
2881 #define MODULE_ALIAS_NETDEV(device) \
2882 MODULE_ALIAS("netdev-" device)
2883
2884 #if defined(CONFIG_DYNAMIC_DEBUG)
2885 #define netdev_dbg(__dev, format, args...) \
2886 do { \
2887 dynamic_netdev_dbg(__dev, format, ##args); \
2888 } while (0)
2889 #elif defined(DEBUG)
2890 #define netdev_dbg(__dev, format, args...) \
2891 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2892 #else
2893 #define netdev_dbg(__dev, format, args...) \
2894 ({ \
2895 if (0) \
2896 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2897 0; \
2898 })
2899 #endif
2900
2901 #if defined(VERBOSE_DEBUG)
2902 #define netdev_vdbg netdev_dbg
2903 #else
2904
2905 #define netdev_vdbg(dev, format, args...) \
2906 ({ \
2907 if (0) \
2908 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2909 0; \
2910 })
2911 #endif
2912
2913 /*
2914 * netdev_WARN() acts like dev_printk(), but with the key difference
2915 * of using a WARN/WARN_ON to get the message out, including the
2916 * file/line information and a backtrace.
2917 */
2918 #define netdev_WARN(dev, format, args...) \
2919 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2920
2921 /* netif printk helpers, similar to netdev_printk */
2922
2923 #define netif_printk(priv, type, level, dev, fmt, args...) \
2924 do { \
2925 if (netif_msg_##type(priv)) \
2926 netdev_printk(level, (dev), fmt, ##args); \
2927 } while (0)
2928
2929 #define netif_level(level, priv, type, dev, fmt, args...) \
2930 do { \
2931 if (netif_msg_##type(priv)) \
2932 netdev_##level(dev, fmt, ##args); \
2933 } while (0)
2934
2935 #define netif_emerg(priv, type, dev, fmt, args...) \
2936 netif_level(emerg, priv, type, dev, fmt, ##args)
2937 #define netif_alert(priv, type, dev, fmt, args...) \
2938 netif_level(alert, priv, type, dev, fmt, ##args)
2939 #define netif_crit(priv, type, dev, fmt, args...) \
2940 netif_level(crit, priv, type, dev, fmt, ##args)
2941 #define netif_err(priv, type, dev, fmt, args...) \
2942 netif_level(err, priv, type, dev, fmt, ##args)
2943 #define netif_warn(priv, type, dev, fmt, args...) \
2944 netif_level(warn, priv, type, dev, fmt, ##args)
2945 #define netif_notice(priv, type, dev, fmt, args...) \
2946 netif_level(notice, priv, type, dev, fmt, ##args)
2947 #define netif_info(priv, type, dev, fmt, args...) \
2948 netif_level(info, priv, type, dev, fmt, ##args)
2949
2950 #if defined(CONFIG_DYNAMIC_DEBUG)
2951 #define netif_dbg(priv, type, netdev, format, args...) \
2952 do { \
2953 if (netif_msg_##type(priv)) \
2954 dynamic_netdev_dbg(netdev, format, ##args); \
2955 } while (0)
2956 #elif defined(DEBUG)
2957 #define netif_dbg(priv, type, dev, format, args...) \
2958 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2959 #else
2960 #define netif_dbg(priv, type, dev, format, args...) \
2961 ({ \
2962 if (0) \
2963 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2964 0; \
2965 })
2966 #endif
2967
2968 #if defined(VERBOSE_DEBUG)
2969 #define netif_vdbg netif_dbg
2970 #else
2971 #define netif_vdbg(priv, type, dev, format, args...) \
2972 ({ \
2973 if (0) \
2974 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2975 0; \
2976 })
2977 #endif
2978
2979 /*
2980 * The list of packet types we will receive (as opposed to discard)
2981 * and the routines to invoke.
2982 *
2983 * Why 16. Because with 16 the only overlap we get on a hash of the
2984 * low nibble of the protocol value is RARP/SNAP/X.25.
2985 *
2986 * NOTE: That is no longer true with the addition of VLAN tags. Not
2987 * sure which should go first, but I bet it won't make much
2988 * difference if we are running VLANs. The good news is that
2989 * this protocol won't be in the list unless compiled in, so
2990 * the average user (w/out VLANs) will not be adversely affected.
2991 * --BLG
2992 *
2993 * 0800 IP
2994 * 8100 802.1Q VLAN
2995 * 0001 802.3
2996 * 0002 AX.25
2997 * 0004 802.2
2998 * 8035 RARP
2999 * 0005 SNAP
3000 * 0805 X.25
3001 * 0806 ARP
3002 * 8137 IPX
3003 * 0009 Localtalk
3004 * 86DD IPv6
3005 */
3006 #define PTYPE_HASH_SIZE (16)
3007 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3008
3009 #endif /* _LINUX_NETDEVICE_H */