]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/netdevice.h
Merge git://git.infradead.org/users/willy/linux-nvme
[mirror_ubuntu-artful-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 #elif IS_ENABLED(CONFIG_TR)
148 # define LL_MAX_HEADER 48
149 #else
150 # define LL_MAX_HEADER 32
151 #endif
152
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
156 #else
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 #endif
159
160 /*
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
163 */
164
165 struct net_device_stats {
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
175 unsigned long collisions;
176 unsigned long rx_length_errors;
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
189 };
190
191
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
194
195 #ifdef CONFIG_RPS
196 #include <linux/static_key.h>
197 extern struct static_key rps_needed;
198 #endif
199
200 struct neighbour;
201 struct neigh_parms;
202 struct sk_buff;
203
204 struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
208 #define NETDEV_HW_ADDR_T_LAN 1
209 #define NETDEV_HW_ADDR_T_SAN 2
210 #define NETDEV_HW_ADDR_T_SLAVE 3
211 #define NETDEV_HW_ADDR_T_UNICAST 4
212 #define NETDEV_HW_ADDR_T_MULTICAST 5
213 bool synced;
214 bool global_use;
215 int refcount;
216 struct rcu_head rcu_head;
217 };
218
219 struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222 };
223
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
228
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238
239 struct hh_cache {
240 u16 hh_len;
241 u16 __pad;
242 seqlock_t hh_lock;
243
244 /* cached hardware header; allow for machine alignment needs. */
245 #define HH_DATA_MOD 16
246 #define HH_DATA_OFF(__len) \
247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
248 #define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 };
252
253 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261 #define LL_RESERVED_SPACE(dev) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265
266 struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
269 const void *saddr, unsigned int len);
270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
271 int (*rebuild)(struct sk_buff *skb);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 };
277
278 /* These flag bits are private to the generic network queueing
279 * layer, they may not be explicitly referenced by any other
280 * code.
281 */
282
283 enum netdev_state_t {
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
286 __LINK_STATE_NOCARRIER,
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
289 };
290
291
292 /*
293 * This structure holds at boot time configured netdevice settings. They
294 * are then used in the device probing.
295 */
296 struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299 };
300 #define NETDEV_BOOT_SETUP_MAX 8
301
302 extern int __init netdev_boot_setup(char *str);
303
304 /*
305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307 struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
311 * to the per-cpu poll_list, and whoever clears that bit
312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
318 unsigned int gro_count;
319 int (*poll)(struct napi_struct *, int);
320 #ifdef CONFIG_NETPOLL
321 spinlock_t poll_lock;
322 int poll_owner;
323 #endif
324 struct net_device *dev;
325 struct sk_buff *gro_list;
326 struct sk_buff *skb;
327 struct list_head dev_list;
328 };
329
330 enum {
331 NAPI_STATE_SCHED, /* Poll is scheduled */
332 NAPI_STATE_DISABLE, /* Disable pending */
333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
334 };
335
336 enum gro_result {
337 GRO_MERGED,
338 GRO_MERGED_FREE,
339 GRO_HELD,
340 GRO_NORMAL,
341 GRO_DROP,
342 };
343 typedef enum gro_result gro_result_t;
344
345 /*
346 * enum rx_handler_result - Possible return values for rx_handlers.
347 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
348 * further.
349 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
350 * case skb->dev was changed by rx_handler.
351 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
352 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
353 *
354 * rx_handlers are functions called from inside __netif_receive_skb(), to do
355 * special processing of the skb, prior to delivery to protocol handlers.
356 *
357 * Currently, a net_device can only have a single rx_handler registered. Trying
358 * to register a second rx_handler will return -EBUSY.
359 *
360 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
361 * To unregister a rx_handler on a net_device, use
362 * netdev_rx_handler_unregister().
363 *
364 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
365 * do with the skb.
366 *
367 * If the rx_handler consumed to skb in some way, it should return
368 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
369 * the skb to be delivered in some other ways.
370 *
371 * If the rx_handler changed skb->dev, to divert the skb to another
372 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
373 * new device will be called if it exists.
374 *
375 * If the rx_handler consider the skb should be ignored, it should return
376 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
377 * are registered on exact device (ptype->dev == skb->dev).
378 *
379 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
380 * delivered, it should return RX_HANDLER_PASS.
381 *
382 * A device without a registered rx_handler will behave as if rx_handler
383 * returned RX_HANDLER_PASS.
384 */
385
386 enum rx_handler_result {
387 RX_HANDLER_CONSUMED,
388 RX_HANDLER_ANOTHER,
389 RX_HANDLER_EXACT,
390 RX_HANDLER_PASS,
391 };
392 typedef enum rx_handler_result rx_handler_result_t;
393 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
394
395 extern void __napi_schedule(struct napi_struct *n);
396
397 static inline bool napi_disable_pending(struct napi_struct *n)
398 {
399 return test_bit(NAPI_STATE_DISABLE, &n->state);
400 }
401
402 /**
403 * napi_schedule_prep - check if napi can be scheduled
404 * @n: napi context
405 *
406 * Test if NAPI routine is already running, and if not mark
407 * it as running. This is used as a condition variable
408 * insure only one NAPI poll instance runs. We also make
409 * sure there is no pending NAPI disable.
410 */
411 static inline bool napi_schedule_prep(struct napi_struct *n)
412 {
413 return !napi_disable_pending(n) &&
414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
415 }
416
417 /**
418 * napi_schedule - schedule NAPI poll
419 * @n: napi context
420 *
421 * Schedule NAPI poll routine to be called if it is not already
422 * running.
423 */
424 static inline void napi_schedule(struct napi_struct *n)
425 {
426 if (napi_schedule_prep(n))
427 __napi_schedule(n);
428 }
429
430 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
431 static inline bool napi_reschedule(struct napi_struct *napi)
432 {
433 if (napi_schedule_prep(napi)) {
434 __napi_schedule(napi);
435 return true;
436 }
437 return false;
438 }
439
440 /**
441 * napi_complete - NAPI processing complete
442 * @n: napi context
443 *
444 * Mark NAPI processing as complete.
445 */
446 extern void __napi_complete(struct napi_struct *n);
447 extern void napi_complete(struct napi_struct *n);
448
449 /**
450 * napi_disable - prevent NAPI from scheduling
451 * @n: napi context
452 *
453 * Stop NAPI from being scheduled on this context.
454 * Waits till any outstanding processing completes.
455 */
456 static inline void napi_disable(struct napi_struct *n)
457 {
458 set_bit(NAPI_STATE_DISABLE, &n->state);
459 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
460 msleep(1);
461 clear_bit(NAPI_STATE_DISABLE, &n->state);
462 }
463
464 /**
465 * napi_enable - enable NAPI scheduling
466 * @n: napi context
467 *
468 * Resume NAPI from being scheduled on this context.
469 * Must be paired with napi_disable.
470 */
471 static inline void napi_enable(struct napi_struct *n)
472 {
473 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
474 smp_mb__before_clear_bit();
475 clear_bit(NAPI_STATE_SCHED, &n->state);
476 }
477
478 #ifdef CONFIG_SMP
479 /**
480 * napi_synchronize - wait until NAPI is not running
481 * @n: napi context
482 *
483 * Wait until NAPI is done being scheduled on this context.
484 * Waits till any outstanding processing completes but
485 * does not disable future activations.
486 */
487 static inline void napi_synchronize(const struct napi_struct *n)
488 {
489 while (test_bit(NAPI_STATE_SCHED, &n->state))
490 msleep(1);
491 }
492 #else
493 # define napi_synchronize(n) barrier()
494 #endif
495
496 enum netdev_queue_state_t {
497 __QUEUE_STATE_DRV_XOFF,
498 __QUEUE_STATE_STACK_XOFF,
499 __QUEUE_STATE_FROZEN,
500 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
501 (1 << __QUEUE_STATE_STACK_XOFF))
502 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
503 (1 << __QUEUE_STATE_FROZEN))
504 };
505 /*
506 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
507 * netif_tx_* functions below are used to manipulate this flag. The
508 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
509 * queue independently. The netif_xmit_*stopped functions below are called
510 * to check if the queue has been stopped by the driver or stack (either
511 * of the XOFF bits are set in the state). Drivers should not need to call
512 * netif_xmit*stopped functions, they should only be using netif_tx_*.
513 */
514
515 struct netdev_queue {
516 /*
517 * read mostly part
518 */
519 struct net_device *dev;
520 struct Qdisc *qdisc;
521 struct Qdisc *qdisc_sleeping;
522 #ifdef CONFIG_SYSFS
523 struct kobject kobj;
524 #endif
525 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
526 int numa_node;
527 #endif
528 /*
529 * write mostly part
530 */
531 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
532 int xmit_lock_owner;
533 /*
534 * please use this field instead of dev->trans_start
535 */
536 unsigned long trans_start;
537
538 /*
539 * Number of TX timeouts for this queue
540 * (/sys/class/net/DEV/Q/trans_timeout)
541 */
542 unsigned long trans_timeout;
543
544 unsigned long state;
545
546 #ifdef CONFIG_BQL
547 struct dql dql;
548 #endif
549 } ____cacheline_aligned_in_smp;
550
551 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
552 {
553 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
554 return q->numa_node;
555 #else
556 return NUMA_NO_NODE;
557 #endif
558 }
559
560 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
561 {
562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 q->numa_node = node;
564 #endif
565 }
566
567 #ifdef CONFIG_RPS
568 /*
569 * This structure holds an RPS map which can be of variable length. The
570 * map is an array of CPUs.
571 */
572 struct rps_map {
573 unsigned int len;
574 struct rcu_head rcu;
575 u16 cpus[0];
576 };
577 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
578
579 /*
580 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
581 * tail pointer for that CPU's input queue at the time of last enqueue, and
582 * a hardware filter index.
583 */
584 struct rps_dev_flow {
585 u16 cpu;
586 u16 filter;
587 unsigned int last_qtail;
588 };
589 #define RPS_NO_FILTER 0xffff
590
591 /*
592 * The rps_dev_flow_table structure contains a table of flow mappings.
593 */
594 struct rps_dev_flow_table {
595 unsigned int mask;
596 struct rcu_head rcu;
597 struct work_struct free_work;
598 struct rps_dev_flow flows[0];
599 };
600 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
601 ((_num) * sizeof(struct rps_dev_flow)))
602
603 /*
604 * The rps_sock_flow_table contains mappings of flows to the last CPU
605 * on which they were processed by the application (set in recvmsg).
606 */
607 struct rps_sock_flow_table {
608 unsigned int mask;
609 u16 ents[0];
610 };
611 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
612 ((_num) * sizeof(u16)))
613
614 #define RPS_NO_CPU 0xffff
615
616 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
617 u32 hash)
618 {
619 if (table && hash) {
620 unsigned int cpu, index = hash & table->mask;
621
622 /* We only give a hint, preemption can change cpu under us */
623 cpu = raw_smp_processor_id();
624
625 if (table->ents[index] != cpu)
626 table->ents[index] = cpu;
627 }
628 }
629
630 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
631 u32 hash)
632 {
633 if (table && hash)
634 table->ents[hash & table->mask] = RPS_NO_CPU;
635 }
636
637 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
638
639 #ifdef CONFIG_RFS_ACCEL
640 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
641 u32 flow_id, u16 filter_id);
642 #endif
643
644 /* This structure contains an instance of an RX queue. */
645 struct netdev_rx_queue {
646 struct rps_map __rcu *rps_map;
647 struct rps_dev_flow_table __rcu *rps_flow_table;
648 struct kobject kobj;
649 struct net_device *dev;
650 } ____cacheline_aligned_in_smp;
651 #endif /* CONFIG_RPS */
652
653 #ifdef CONFIG_XPS
654 /*
655 * This structure holds an XPS map which can be of variable length. The
656 * map is an array of queues.
657 */
658 struct xps_map {
659 unsigned int len;
660 unsigned int alloc_len;
661 struct rcu_head rcu;
662 u16 queues[0];
663 };
664 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
665 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
666 / sizeof(u16))
667
668 /*
669 * This structure holds all XPS maps for device. Maps are indexed by CPU.
670 */
671 struct xps_dev_maps {
672 struct rcu_head rcu;
673 struct xps_map __rcu *cpu_map[0];
674 };
675 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
676 (nr_cpu_ids * sizeof(struct xps_map *)))
677 #endif /* CONFIG_XPS */
678
679 #define TC_MAX_QUEUE 16
680 #define TC_BITMASK 15
681 /* HW offloaded queuing disciplines txq count and offset maps */
682 struct netdev_tc_txq {
683 u16 count;
684 u16 offset;
685 };
686
687 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
688 /*
689 * This structure is to hold information about the device
690 * configured to run FCoE protocol stack.
691 */
692 struct netdev_fcoe_hbainfo {
693 char manufacturer[64];
694 char serial_number[64];
695 char hardware_version[64];
696 char driver_version[64];
697 char optionrom_version[64];
698 char firmware_version[64];
699 char model[256];
700 char model_description[256];
701 };
702 #endif
703
704 /*
705 * This structure defines the management hooks for network devices.
706 * The following hooks can be defined; unless noted otherwise, they are
707 * optional and can be filled with a null pointer.
708 *
709 * int (*ndo_init)(struct net_device *dev);
710 * This function is called once when network device is registered.
711 * The network device can use this to any late stage initializaton
712 * or semantic validattion. It can fail with an error code which will
713 * be propogated back to register_netdev
714 *
715 * void (*ndo_uninit)(struct net_device *dev);
716 * This function is called when device is unregistered or when registration
717 * fails. It is not called if init fails.
718 *
719 * int (*ndo_open)(struct net_device *dev);
720 * This function is called when network device transistions to the up
721 * state.
722 *
723 * int (*ndo_stop)(struct net_device *dev);
724 * This function is called when network device transistions to the down
725 * state.
726 *
727 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
728 * struct net_device *dev);
729 * Called when a packet needs to be transmitted.
730 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
731 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
732 * Required can not be NULL.
733 *
734 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
735 * Called to decide which queue to when device supports multiple
736 * transmit queues.
737 *
738 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
739 * This function is called to allow device receiver to make
740 * changes to configuration when multicast or promiscious is enabled.
741 *
742 * void (*ndo_set_rx_mode)(struct net_device *dev);
743 * This function is called device changes address list filtering.
744 * If driver handles unicast address filtering, it should set
745 * IFF_UNICAST_FLT to its priv_flags.
746 *
747 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
748 * This function is called when the Media Access Control address
749 * needs to be changed. If this interface is not defined, the
750 * mac address can not be changed.
751 *
752 * int (*ndo_validate_addr)(struct net_device *dev);
753 * Test if Media Access Control address is valid for the device.
754 *
755 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
756 * Called when a user request an ioctl which can't be handled by
757 * the generic interface code. If not defined ioctl's return
758 * not supported error code.
759 *
760 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
761 * Used to set network devices bus interface parameters. This interface
762 * is retained for legacy reason, new devices should use the bus
763 * interface (PCI) for low level management.
764 *
765 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
766 * Called when a user wants to change the Maximum Transfer Unit
767 * of a device. If not defined, any request to change MTU will
768 * will return an error.
769 *
770 * void (*ndo_tx_timeout)(struct net_device *dev);
771 * Callback uses when the transmitter has not made any progress
772 * for dev->watchdog ticks.
773 *
774 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
775 * struct rtnl_link_stats64 *storage);
776 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
777 * Called when a user wants to get the network device usage
778 * statistics. Drivers must do one of the following:
779 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
780 * rtnl_link_stats64 structure passed by the caller.
781 * 2. Define @ndo_get_stats to update a net_device_stats structure
782 * (which should normally be dev->stats) and return a pointer to
783 * it. The structure may be changed asynchronously only if each
784 * field is written atomically.
785 * 3. Update dev->stats asynchronously and atomically, and define
786 * neither operation.
787 *
788 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, unsigned short vid);
789 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
790 * this function is called when a VLAN id is registered.
791 *
792 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
793 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
794 * this function is called when a VLAN id is unregistered.
795 *
796 * void (*ndo_poll_controller)(struct net_device *dev);
797 *
798 * SR-IOV management functions.
799 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
800 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
801 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
802 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
803 * int (*ndo_get_vf_config)(struct net_device *dev,
804 * int vf, struct ifla_vf_info *ivf);
805 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
806 * struct nlattr *port[]);
807 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
808 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
809 * Called to setup 'tc' number of traffic classes in the net device. This
810 * is always called from the stack with the rtnl lock held and netif tx
811 * queues stopped. This allows the netdevice to perform queue management
812 * safely.
813 *
814 * Fiber Channel over Ethernet (FCoE) offload functions.
815 * int (*ndo_fcoe_enable)(struct net_device *dev);
816 * Called when the FCoE protocol stack wants to start using LLD for FCoE
817 * so the underlying device can perform whatever needed configuration or
818 * initialization to support acceleration of FCoE traffic.
819 *
820 * int (*ndo_fcoe_disable)(struct net_device *dev);
821 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
822 * so the underlying device can perform whatever needed clean-ups to
823 * stop supporting acceleration of FCoE traffic.
824 *
825 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
826 * struct scatterlist *sgl, unsigned int sgc);
827 * Called when the FCoE Initiator wants to initialize an I/O that
828 * is a possible candidate for Direct Data Placement (DDP). The LLD can
829 * perform necessary setup and returns 1 to indicate the device is set up
830 * successfully to perform DDP on this I/O, otherwise this returns 0.
831 *
832 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
833 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
834 * indicated by the FC exchange id 'xid', so the underlying device can
835 * clean up and reuse resources for later DDP requests.
836 *
837 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
838 * struct scatterlist *sgl, unsigned int sgc);
839 * Called when the FCoE Target wants to initialize an I/O that
840 * is a possible candidate for Direct Data Placement (DDP). The LLD can
841 * perform necessary setup and returns 1 to indicate the device is set up
842 * successfully to perform DDP on this I/O, otherwise this returns 0.
843 *
844 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
845 * struct netdev_fcoe_hbainfo *hbainfo);
846 * Called when the FCoE Protocol stack wants information on the underlying
847 * device. This information is utilized by the FCoE protocol stack to
848 * register attributes with Fiber Channel management service as per the
849 * FC-GS Fabric Device Management Information(FDMI) specification.
850 *
851 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
852 * Called when the underlying device wants to override default World Wide
853 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
854 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
855 * protocol stack to use.
856 *
857 * RFS acceleration.
858 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
859 * u16 rxq_index, u32 flow_id);
860 * Set hardware filter for RFS. rxq_index is the target queue index;
861 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
862 * Return the filter ID on success, or a negative error code.
863 *
864 * Slave management functions (for bridge, bonding, etc).
865 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
866 * Called to make another netdev an underling.
867 *
868 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
869 * Called to release previously enslaved netdev.
870 *
871 * Feature/offload setting functions.
872 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
873 * netdev_features_t features);
874 * Adjusts the requested feature flags according to device-specific
875 * constraints, and returns the resulting flags. Must not modify
876 * the device state.
877 *
878 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
879 * Called to update device configuration to new features. Passed
880 * feature set might be less than what was returned by ndo_fix_features()).
881 * Must return >0 or -errno if it changed dev->features itself.
882 *
883 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
884 * struct net_device *dev,
885 * const unsigned char *addr, u16 flags)
886 * Adds an FDB entry to dev for addr.
887 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
888 * struct net_device *dev,
889 * const unsigned char *addr)
890 * Deletes the FDB entry from dev coresponding to addr.
891 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
892 * struct net_device *dev, int idx)
893 * Used to add FDB entries to dump requests. Implementers should add
894 * entries to skb and update idx with the number of entries.
895 *
896 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
897 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
898 * struct net_device *dev)
899 *
900 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
901 * Called to change device carrier. Soft-devices (like dummy, team, etc)
902 * which do not represent real hardware may define this to allow their
903 * userspace components to manage their virtual carrier state. Devices
904 * that determine carrier state from physical hardware properties (eg
905 * network cables) or protocol-dependent mechanisms (eg
906 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
907 */
908 struct net_device_ops {
909 int (*ndo_init)(struct net_device *dev);
910 void (*ndo_uninit)(struct net_device *dev);
911 int (*ndo_open)(struct net_device *dev);
912 int (*ndo_stop)(struct net_device *dev);
913 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
914 struct net_device *dev);
915 u16 (*ndo_select_queue)(struct net_device *dev,
916 struct sk_buff *skb);
917 void (*ndo_change_rx_flags)(struct net_device *dev,
918 int flags);
919 void (*ndo_set_rx_mode)(struct net_device *dev);
920 int (*ndo_set_mac_address)(struct net_device *dev,
921 void *addr);
922 int (*ndo_validate_addr)(struct net_device *dev);
923 int (*ndo_do_ioctl)(struct net_device *dev,
924 struct ifreq *ifr, int cmd);
925 int (*ndo_set_config)(struct net_device *dev,
926 struct ifmap *map);
927 int (*ndo_change_mtu)(struct net_device *dev,
928 int new_mtu);
929 int (*ndo_neigh_setup)(struct net_device *dev,
930 struct neigh_parms *);
931 void (*ndo_tx_timeout) (struct net_device *dev);
932
933 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
934 struct rtnl_link_stats64 *storage);
935 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
936
937 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
938 unsigned short vid);
939 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
940 unsigned short vid);
941 #ifdef CONFIG_NET_POLL_CONTROLLER
942 void (*ndo_poll_controller)(struct net_device *dev);
943 int (*ndo_netpoll_setup)(struct net_device *dev,
944 struct netpoll_info *info,
945 gfp_t gfp);
946 void (*ndo_netpoll_cleanup)(struct net_device *dev);
947 #endif
948 int (*ndo_set_vf_mac)(struct net_device *dev,
949 int queue, u8 *mac);
950 int (*ndo_set_vf_vlan)(struct net_device *dev,
951 int queue, u16 vlan, u8 qos);
952 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
953 int vf, int rate);
954 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
955 int vf, bool setting);
956 int (*ndo_get_vf_config)(struct net_device *dev,
957 int vf,
958 struct ifla_vf_info *ivf);
959 int (*ndo_set_vf_port)(struct net_device *dev,
960 int vf,
961 struct nlattr *port[]);
962 int (*ndo_get_vf_port)(struct net_device *dev,
963 int vf, struct sk_buff *skb);
964 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
965 #if IS_ENABLED(CONFIG_FCOE)
966 int (*ndo_fcoe_enable)(struct net_device *dev);
967 int (*ndo_fcoe_disable)(struct net_device *dev);
968 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
969 u16 xid,
970 struct scatterlist *sgl,
971 unsigned int sgc);
972 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
973 u16 xid);
974 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
975 u16 xid,
976 struct scatterlist *sgl,
977 unsigned int sgc);
978 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
979 struct netdev_fcoe_hbainfo *hbainfo);
980 #endif
981
982 #if IS_ENABLED(CONFIG_LIBFCOE)
983 #define NETDEV_FCOE_WWNN 0
984 #define NETDEV_FCOE_WWPN 1
985 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
986 u64 *wwn, int type);
987 #endif
988
989 #ifdef CONFIG_RFS_ACCEL
990 int (*ndo_rx_flow_steer)(struct net_device *dev,
991 const struct sk_buff *skb,
992 u16 rxq_index,
993 u32 flow_id);
994 #endif
995 int (*ndo_add_slave)(struct net_device *dev,
996 struct net_device *slave_dev);
997 int (*ndo_del_slave)(struct net_device *dev,
998 struct net_device *slave_dev);
999 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1000 netdev_features_t features);
1001 int (*ndo_set_features)(struct net_device *dev,
1002 netdev_features_t features);
1003 int (*ndo_neigh_construct)(struct neighbour *n);
1004 void (*ndo_neigh_destroy)(struct neighbour *n);
1005
1006 int (*ndo_fdb_add)(struct ndmsg *ndm,
1007 struct nlattr *tb[],
1008 struct net_device *dev,
1009 const unsigned char *addr,
1010 u16 flags);
1011 int (*ndo_fdb_del)(struct ndmsg *ndm,
1012 struct nlattr *tb[],
1013 struct net_device *dev,
1014 const unsigned char *addr);
1015 int (*ndo_fdb_dump)(struct sk_buff *skb,
1016 struct netlink_callback *cb,
1017 struct net_device *dev,
1018 int idx);
1019
1020 int (*ndo_bridge_setlink)(struct net_device *dev,
1021 struct nlmsghdr *nlh);
1022 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1023 u32 pid, u32 seq,
1024 struct net_device *dev,
1025 u32 filter_mask);
1026 int (*ndo_bridge_dellink)(struct net_device *dev,
1027 struct nlmsghdr *nlh);
1028 int (*ndo_change_carrier)(struct net_device *dev,
1029 bool new_carrier);
1030 };
1031
1032 /*
1033 * The DEVICE structure.
1034 * Actually, this whole structure is a big mistake. It mixes I/O
1035 * data with strictly "high-level" data, and it has to know about
1036 * almost every data structure used in the INET module.
1037 *
1038 * FIXME: cleanup struct net_device such that network protocol info
1039 * moves out.
1040 */
1041
1042 struct net_device {
1043
1044 /*
1045 * This is the first field of the "visible" part of this structure
1046 * (i.e. as seen by users in the "Space.c" file). It is the name
1047 * of the interface.
1048 */
1049 char name[IFNAMSIZ];
1050
1051 /* device name hash chain, please keep it close to name[] */
1052 struct hlist_node name_hlist;
1053
1054 /* snmp alias */
1055 char *ifalias;
1056
1057 /*
1058 * I/O specific fields
1059 * FIXME: Merge these and struct ifmap into one
1060 */
1061 unsigned long mem_end; /* shared mem end */
1062 unsigned long mem_start; /* shared mem start */
1063 unsigned long base_addr; /* device I/O address */
1064 unsigned int irq; /* device IRQ number */
1065
1066 /*
1067 * Some hardware also needs these fields, but they are not
1068 * part of the usual set specified in Space.c.
1069 */
1070
1071 unsigned long state;
1072
1073 struct list_head dev_list;
1074 struct list_head napi_list;
1075 struct list_head unreg_list;
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
1092 /* Interface index. Unique device identifier */
1093 int ifindex;
1094 int iflink;
1095
1096 struct net_device_stats stats;
1097 atomic_long_t rx_dropped; /* dropped packets by core network
1098 * Do not use this in drivers.
1099 */
1100
1101 #ifdef CONFIG_WIRELESS_EXT
1102 /* List of functions to handle Wireless Extensions (instead of ioctl).
1103 * See <net/iw_handler.h> for details. Jean II */
1104 const struct iw_handler_def * wireless_handlers;
1105 /* Instance data managed by the core of Wireless Extensions. */
1106 struct iw_public_data * wireless_data;
1107 #endif
1108 /* Management operations */
1109 const struct net_device_ops *netdev_ops;
1110 const struct ethtool_ops *ethtool_ops;
1111
1112 /* Hardware header description */
1113 const struct header_ops *header_ops;
1114
1115 unsigned int flags; /* interface flags (a la BSD) */
1116 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1117 * See if.h for definitions. */
1118 unsigned short gflags;
1119 unsigned short padded; /* How much padding added by alloc_netdev() */
1120
1121 unsigned char operstate; /* RFC2863 operstate */
1122 unsigned char link_mode; /* mapping policy to operstate */
1123
1124 unsigned char if_port; /* Selectable AUI, TP,..*/
1125 unsigned char dma; /* DMA channel */
1126
1127 unsigned int mtu; /* interface MTU value */
1128 unsigned short type; /* interface hardware type */
1129 unsigned short hard_header_len; /* hardware hdr length */
1130
1131 /* extra head- and tailroom the hardware may need, but not in all cases
1132 * can this be guaranteed, especially tailroom. Some cases also use
1133 * LL_MAX_HEADER instead to allocate the skb.
1134 */
1135 unsigned short needed_headroom;
1136 unsigned short needed_tailroom;
1137
1138 /* Interface address info. */
1139 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1140 unsigned char addr_assign_type; /* hw address assignment type */
1141 unsigned char addr_len; /* hardware address length */
1142 unsigned char neigh_priv_len;
1143 unsigned short dev_id; /* for shared network cards */
1144
1145 spinlock_t addr_list_lock;
1146 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1147 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1148 bool uc_promisc;
1149 unsigned int promiscuity;
1150 unsigned int allmulti;
1151
1152
1153 /* Protocol specific pointers */
1154
1155 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1156 struct vlan_info __rcu *vlan_info; /* VLAN info */
1157 #endif
1158 #if IS_ENABLED(CONFIG_NET_DSA)
1159 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1160 #endif
1161 void *atalk_ptr; /* AppleTalk link */
1162 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1163 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1164 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1165 void *ax25_ptr; /* AX.25 specific data */
1166 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1167 assign before registering */
1168
1169 /*
1170 * Cache lines mostly used on receive path (including eth_type_trans())
1171 */
1172 unsigned long last_rx; /* Time of last Rx
1173 * This should not be set in
1174 * drivers, unless really needed,
1175 * because network stack (bonding)
1176 * use it if/when necessary, to
1177 * avoid dirtying this cache line.
1178 */
1179
1180 struct list_head upper_dev_list; /* List of upper devices */
1181
1182 /* Interface address info used in eth_type_trans() */
1183 unsigned char *dev_addr; /* hw address, (before bcast
1184 because most packets are
1185 unicast) */
1186
1187 struct netdev_hw_addr_list dev_addrs; /* list of device
1188 hw addresses */
1189
1190 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1191
1192 #ifdef CONFIG_SYSFS
1193 struct kset *queues_kset;
1194 #endif
1195
1196 #ifdef CONFIG_RPS
1197 struct netdev_rx_queue *_rx;
1198
1199 /* Number of RX queues allocated at register_netdev() time */
1200 unsigned int num_rx_queues;
1201
1202 /* Number of RX queues currently active in device */
1203 unsigned int real_num_rx_queues;
1204
1205 #ifdef CONFIG_RFS_ACCEL
1206 /* CPU reverse-mapping for RX completion interrupts, indexed
1207 * by RX queue number. Assigned by driver. This must only be
1208 * set if the ndo_rx_flow_steer operation is defined. */
1209 struct cpu_rmap *rx_cpu_rmap;
1210 #endif
1211 #endif
1212
1213 rx_handler_func_t __rcu *rx_handler;
1214 void __rcu *rx_handler_data;
1215
1216 struct netdev_queue __rcu *ingress_queue;
1217
1218 /*
1219 * Cache lines mostly used on transmit path
1220 */
1221 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1222
1223 /* Number of TX queues allocated at alloc_netdev_mq() time */
1224 unsigned int num_tx_queues;
1225
1226 /* Number of TX queues currently active in device */
1227 unsigned int real_num_tx_queues;
1228
1229 /* root qdisc from userspace point of view */
1230 struct Qdisc *qdisc;
1231
1232 unsigned long tx_queue_len; /* Max frames per queue allowed */
1233 spinlock_t tx_global_lock;
1234
1235 #ifdef CONFIG_XPS
1236 struct xps_dev_maps __rcu *xps_maps;
1237 #endif
1238
1239 /* These may be needed for future network-power-down code. */
1240
1241 /*
1242 * trans_start here is expensive for high speed devices on SMP,
1243 * please use netdev_queue->trans_start instead.
1244 */
1245 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1246
1247 int watchdog_timeo; /* used by dev_watchdog() */
1248 struct timer_list watchdog_timer;
1249
1250 /* Number of references to this device */
1251 int __percpu *pcpu_refcnt;
1252
1253 /* delayed register/unregister */
1254 struct list_head todo_list;
1255 /* device index hash chain */
1256 struct hlist_node index_hlist;
1257
1258 struct list_head link_watch_list;
1259
1260 /* register/unregister state machine */
1261 enum { NETREG_UNINITIALIZED=0,
1262 NETREG_REGISTERED, /* completed register_netdevice */
1263 NETREG_UNREGISTERING, /* called unregister_netdevice */
1264 NETREG_UNREGISTERED, /* completed unregister todo */
1265 NETREG_RELEASED, /* called free_netdev */
1266 NETREG_DUMMY, /* dummy device for NAPI poll */
1267 } reg_state:8;
1268
1269 bool dismantle; /* device is going do be freed */
1270
1271 enum {
1272 RTNL_LINK_INITIALIZED,
1273 RTNL_LINK_INITIALIZING,
1274 } rtnl_link_state:16;
1275
1276 /* Called from unregister, can be used to call free_netdev */
1277 void (*destructor)(struct net_device *dev);
1278
1279 #ifdef CONFIG_NETPOLL
1280 struct netpoll_info __rcu *npinfo;
1281 #endif
1282
1283 #ifdef CONFIG_NET_NS
1284 /* Network namespace this network device is inside */
1285 struct net *nd_net;
1286 #endif
1287
1288 /* mid-layer private */
1289 union {
1290 void *ml_priv;
1291 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1292 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1293 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1294 struct pcpu_vstats __percpu *vstats; /* veth stats */
1295 };
1296 /* GARP */
1297 struct garp_port __rcu *garp_port;
1298 /* MRP */
1299 struct mrp_port __rcu *mrp_port;
1300
1301 /* class/net/name entry */
1302 struct device dev;
1303 /* space for optional device, statistics, and wireless sysfs groups */
1304 const struct attribute_group *sysfs_groups[4];
1305
1306 /* rtnetlink link ops */
1307 const struct rtnl_link_ops *rtnl_link_ops;
1308
1309 /* for setting kernel sock attribute on TCP connection setup */
1310 #define GSO_MAX_SIZE 65536
1311 unsigned int gso_max_size;
1312 #define GSO_MAX_SEGS 65535
1313 u16 gso_max_segs;
1314
1315 #ifdef CONFIG_DCB
1316 /* Data Center Bridging netlink ops */
1317 const struct dcbnl_rtnl_ops *dcbnl_ops;
1318 #endif
1319 u8 num_tc;
1320 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1321 u8 prio_tc_map[TC_BITMASK + 1];
1322
1323 #if IS_ENABLED(CONFIG_FCOE)
1324 /* max exchange id for FCoE LRO by ddp */
1325 unsigned int fcoe_ddp_xid;
1326 #endif
1327 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1328 struct netprio_map __rcu *priomap;
1329 #endif
1330 /* phy device may attach itself for hardware timestamping */
1331 struct phy_device *phydev;
1332
1333 struct lock_class_key *qdisc_tx_busylock;
1334
1335 /* group the device belongs to */
1336 int group;
1337
1338 struct pm_qos_request pm_qos_req;
1339 };
1340 #define to_net_dev(d) container_of(d, struct net_device, dev)
1341
1342 #define NETDEV_ALIGN 32
1343
1344 static inline
1345 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1346 {
1347 return dev->prio_tc_map[prio & TC_BITMASK];
1348 }
1349
1350 static inline
1351 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1352 {
1353 if (tc >= dev->num_tc)
1354 return -EINVAL;
1355
1356 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1357 return 0;
1358 }
1359
1360 static inline
1361 void netdev_reset_tc(struct net_device *dev)
1362 {
1363 dev->num_tc = 0;
1364 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1365 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1366 }
1367
1368 static inline
1369 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1370 {
1371 if (tc >= dev->num_tc)
1372 return -EINVAL;
1373
1374 dev->tc_to_txq[tc].count = count;
1375 dev->tc_to_txq[tc].offset = offset;
1376 return 0;
1377 }
1378
1379 static inline
1380 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1381 {
1382 if (num_tc > TC_MAX_QUEUE)
1383 return -EINVAL;
1384
1385 dev->num_tc = num_tc;
1386 return 0;
1387 }
1388
1389 static inline
1390 int netdev_get_num_tc(struct net_device *dev)
1391 {
1392 return dev->num_tc;
1393 }
1394
1395 static inline
1396 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1397 unsigned int index)
1398 {
1399 return &dev->_tx[index];
1400 }
1401
1402 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1403 void (*f)(struct net_device *,
1404 struct netdev_queue *,
1405 void *),
1406 void *arg)
1407 {
1408 unsigned int i;
1409
1410 for (i = 0; i < dev->num_tx_queues; i++)
1411 f(dev, &dev->_tx[i], arg);
1412 }
1413
1414 extern struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1415 struct sk_buff *skb);
1416 extern u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
1417
1418 /*
1419 * Net namespace inlines
1420 */
1421 static inline
1422 struct net *dev_net(const struct net_device *dev)
1423 {
1424 return read_pnet(&dev->nd_net);
1425 }
1426
1427 static inline
1428 void dev_net_set(struct net_device *dev, struct net *net)
1429 {
1430 #ifdef CONFIG_NET_NS
1431 release_net(dev->nd_net);
1432 dev->nd_net = hold_net(net);
1433 #endif
1434 }
1435
1436 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1437 {
1438 #ifdef CONFIG_NET_DSA_TAG_DSA
1439 if (dev->dsa_ptr != NULL)
1440 return dsa_uses_dsa_tags(dev->dsa_ptr);
1441 #endif
1442
1443 return 0;
1444 }
1445
1446 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1447 {
1448 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1449 if (dev->dsa_ptr != NULL)
1450 return dsa_uses_trailer_tags(dev->dsa_ptr);
1451 #endif
1452
1453 return 0;
1454 }
1455
1456 /**
1457 * netdev_priv - access network device private data
1458 * @dev: network device
1459 *
1460 * Get network device private data
1461 */
1462 static inline void *netdev_priv(const struct net_device *dev)
1463 {
1464 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1465 }
1466
1467 /* Set the sysfs physical device reference for the network logical device
1468 * if set prior to registration will cause a symlink during initialization.
1469 */
1470 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1471
1472 /* Set the sysfs device type for the network logical device to allow
1473 * fin grained indentification of different network device types. For
1474 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1475 */
1476 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1477
1478 /**
1479 * netif_napi_add - initialize a napi context
1480 * @dev: network device
1481 * @napi: napi context
1482 * @poll: polling function
1483 * @weight: default weight
1484 *
1485 * netif_napi_add() must be used to initialize a napi context prior to calling
1486 * *any* of the other napi related functions.
1487 */
1488 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1489 int (*poll)(struct napi_struct *, int), int weight);
1490
1491 /**
1492 * netif_napi_del - remove a napi context
1493 * @napi: napi context
1494 *
1495 * netif_napi_del() removes a napi context from the network device napi list
1496 */
1497 void netif_napi_del(struct napi_struct *napi);
1498
1499 struct napi_gro_cb {
1500 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1501 void *frag0;
1502
1503 /* Length of frag0. */
1504 unsigned int frag0_len;
1505
1506 /* This indicates where we are processing relative to skb->data. */
1507 int data_offset;
1508
1509 /* This is non-zero if the packet cannot be merged with the new skb. */
1510 int flush;
1511
1512 /* Number of segments aggregated. */
1513 u16 count;
1514
1515 /* This is non-zero if the packet may be of the same flow. */
1516 u8 same_flow;
1517
1518 /* Free the skb? */
1519 u8 free;
1520 #define NAPI_GRO_FREE 1
1521 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1522
1523 /* jiffies when first packet was created/queued */
1524 unsigned long age;
1525
1526 /* Used in ipv6_gro_receive() */
1527 int proto;
1528
1529 /* used in skb_gro_receive() slow path */
1530 struct sk_buff *last;
1531 };
1532
1533 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1534
1535 struct packet_type {
1536 __be16 type; /* This is really htons(ether_type). */
1537 struct net_device *dev; /* NULL is wildcarded here */
1538 int (*func) (struct sk_buff *,
1539 struct net_device *,
1540 struct packet_type *,
1541 struct net_device *);
1542 bool (*id_match)(struct packet_type *ptype,
1543 struct sock *sk);
1544 void *af_packet_priv;
1545 struct list_head list;
1546 };
1547
1548 struct offload_callbacks {
1549 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1550 netdev_features_t features);
1551 int (*gso_send_check)(struct sk_buff *skb);
1552 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1553 struct sk_buff *skb);
1554 int (*gro_complete)(struct sk_buff *skb);
1555 };
1556
1557 struct packet_offload {
1558 __be16 type; /* This is really htons(ether_type). */
1559 struct offload_callbacks callbacks;
1560 struct list_head list;
1561 };
1562
1563 #include <linux/notifier.h>
1564
1565 /* netdevice notifier chain. Please remember to update the rtnetlink
1566 * notification exclusion list in rtnetlink_event() when adding new
1567 * types.
1568 */
1569 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1570 #define NETDEV_DOWN 0x0002
1571 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1572 detected a hardware crash and restarted
1573 - we can use this eg to kick tcp sessions
1574 once done */
1575 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1576 #define NETDEV_REGISTER 0x0005
1577 #define NETDEV_UNREGISTER 0x0006
1578 #define NETDEV_CHANGEMTU 0x0007
1579 #define NETDEV_CHANGEADDR 0x0008
1580 #define NETDEV_GOING_DOWN 0x0009
1581 #define NETDEV_CHANGENAME 0x000A
1582 #define NETDEV_FEAT_CHANGE 0x000B
1583 #define NETDEV_BONDING_FAILOVER 0x000C
1584 #define NETDEV_PRE_UP 0x000D
1585 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1586 #define NETDEV_POST_TYPE_CHANGE 0x000F
1587 #define NETDEV_POST_INIT 0x0010
1588 #define NETDEV_UNREGISTER_FINAL 0x0011
1589 #define NETDEV_RELEASE 0x0012
1590 #define NETDEV_NOTIFY_PEERS 0x0013
1591 #define NETDEV_JOIN 0x0014
1592
1593 extern int register_netdevice_notifier(struct notifier_block *nb);
1594 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1595 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1596
1597
1598 extern rwlock_t dev_base_lock; /* Device list lock */
1599
1600 extern seqcount_t devnet_rename_seq; /* Device rename seq */
1601
1602
1603 #define for_each_netdev(net, d) \
1604 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1605 #define for_each_netdev_reverse(net, d) \
1606 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1607 #define for_each_netdev_rcu(net, d) \
1608 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1609 #define for_each_netdev_safe(net, d, n) \
1610 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1611 #define for_each_netdev_continue(net, d) \
1612 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1613 #define for_each_netdev_continue_rcu(net, d) \
1614 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1615 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1616
1617 static inline struct net_device *next_net_device(struct net_device *dev)
1618 {
1619 struct list_head *lh;
1620 struct net *net;
1621
1622 net = dev_net(dev);
1623 lh = dev->dev_list.next;
1624 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1625 }
1626
1627 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1628 {
1629 struct list_head *lh;
1630 struct net *net;
1631
1632 net = dev_net(dev);
1633 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1634 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1635 }
1636
1637 static inline struct net_device *first_net_device(struct net *net)
1638 {
1639 return list_empty(&net->dev_base_head) ? NULL :
1640 net_device_entry(net->dev_base_head.next);
1641 }
1642
1643 static inline struct net_device *first_net_device_rcu(struct net *net)
1644 {
1645 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1646
1647 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1648 }
1649
1650 extern int netdev_boot_setup_check(struct net_device *dev);
1651 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1652 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1653 const char *hwaddr);
1654 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1655 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1656 extern void dev_add_pack(struct packet_type *pt);
1657 extern void dev_remove_pack(struct packet_type *pt);
1658 extern void __dev_remove_pack(struct packet_type *pt);
1659 extern void dev_add_offload(struct packet_offload *po);
1660 extern void dev_remove_offload(struct packet_offload *po);
1661 extern void __dev_remove_offload(struct packet_offload *po);
1662
1663 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1664 unsigned short mask);
1665 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1666 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1667 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1668 extern int dev_alloc_name(struct net_device *dev, const char *name);
1669 extern int dev_open(struct net_device *dev);
1670 extern int dev_close(struct net_device *dev);
1671 extern void dev_disable_lro(struct net_device *dev);
1672 extern int dev_loopback_xmit(struct sk_buff *newskb);
1673 extern int dev_queue_xmit(struct sk_buff *skb);
1674 extern int register_netdevice(struct net_device *dev);
1675 extern void unregister_netdevice_queue(struct net_device *dev,
1676 struct list_head *head);
1677 extern void unregister_netdevice_many(struct list_head *head);
1678 static inline void unregister_netdevice(struct net_device *dev)
1679 {
1680 unregister_netdevice_queue(dev, NULL);
1681 }
1682
1683 extern int netdev_refcnt_read(const struct net_device *dev);
1684 extern void free_netdev(struct net_device *dev);
1685 extern void synchronize_net(void);
1686 extern int init_dummy_netdev(struct net_device *dev);
1687 extern void netdev_resync_ops(struct net_device *dev);
1688
1689 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1690 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1691 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1692 extern int dev_restart(struct net_device *dev);
1693 #ifdef CONFIG_NETPOLL_TRAP
1694 extern int netpoll_trap(void);
1695 #endif
1696 extern int skb_gro_receive(struct sk_buff **head,
1697 struct sk_buff *skb);
1698
1699 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1700 {
1701 return NAPI_GRO_CB(skb)->data_offset;
1702 }
1703
1704 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1705 {
1706 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1707 }
1708
1709 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1710 {
1711 NAPI_GRO_CB(skb)->data_offset += len;
1712 }
1713
1714 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1715 unsigned int offset)
1716 {
1717 return NAPI_GRO_CB(skb)->frag0 + offset;
1718 }
1719
1720 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1721 {
1722 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1723 }
1724
1725 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1726 unsigned int offset)
1727 {
1728 if (!pskb_may_pull(skb, hlen))
1729 return NULL;
1730
1731 NAPI_GRO_CB(skb)->frag0 = NULL;
1732 NAPI_GRO_CB(skb)->frag0_len = 0;
1733 return skb->data + offset;
1734 }
1735
1736 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1737 {
1738 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1739 }
1740
1741 static inline void *skb_gro_network_header(struct sk_buff *skb)
1742 {
1743 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1744 skb_network_offset(skb);
1745 }
1746
1747 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1748 unsigned short type,
1749 const void *daddr, const void *saddr,
1750 unsigned int len)
1751 {
1752 if (!dev->header_ops || !dev->header_ops->create)
1753 return 0;
1754
1755 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1756 }
1757
1758 static inline int dev_parse_header(const struct sk_buff *skb,
1759 unsigned char *haddr)
1760 {
1761 const struct net_device *dev = skb->dev;
1762
1763 if (!dev->header_ops || !dev->header_ops->parse)
1764 return 0;
1765 return dev->header_ops->parse(skb, haddr);
1766 }
1767
1768 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1769 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1770 static inline int unregister_gifconf(unsigned int family)
1771 {
1772 return register_gifconf(family, NULL);
1773 }
1774
1775 /*
1776 * Incoming packets are placed on per-cpu queues
1777 */
1778 struct softnet_data {
1779 struct Qdisc *output_queue;
1780 struct Qdisc **output_queue_tailp;
1781 struct list_head poll_list;
1782 struct sk_buff *completion_queue;
1783 struct sk_buff_head process_queue;
1784
1785 /* stats */
1786 unsigned int processed;
1787 unsigned int time_squeeze;
1788 unsigned int cpu_collision;
1789 unsigned int received_rps;
1790
1791 #ifdef CONFIG_RPS
1792 struct softnet_data *rps_ipi_list;
1793
1794 /* Elements below can be accessed between CPUs for RPS */
1795 struct call_single_data csd ____cacheline_aligned_in_smp;
1796 struct softnet_data *rps_ipi_next;
1797 unsigned int cpu;
1798 unsigned int input_queue_head;
1799 unsigned int input_queue_tail;
1800 #endif
1801 unsigned int dropped;
1802 struct sk_buff_head input_pkt_queue;
1803 struct napi_struct backlog;
1804 };
1805
1806 static inline void input_queue_head_incr(struct softnet_data *sd)
1807 {
1808 #ifdef CONFIG_RPS
1809 sd->input_queue_head++;
1810 #endif
1811 }
1812
1813 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1814 unsigned int *qtail)
1815 {
1816 #ifdef CONFIG_RPS
1817 *qtail = ++sd->input_queue_tail;
1818 #endif
1819 }
1820
1821 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1822
1823 extern void __netif_schedule(struct Qdisc *q);
1824
1825 static inline void netif_schedule_queue(struct netdev_queue *txq)
1826 {
1827 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1828 __netif_schedule(txq->qdisc);
1829 }
1830
1831 static inline void netif_tx_schedule_all(struct net_device *dev)
1832 {
1833 unsigned int i;
1834
1835 for (i = 0; i < dev->num_tx_queues; i++)
1836 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1837 }
1838
1839 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1840 {
1841 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1842 }
1843
1844 /**
1845 * netif_start_queue - allow transmit
1846 * @dev: network device
1847 *
1848 * Allow upper layers to call the device hard_start_xmit routine.
1849 */
1850 static inline void netif_start_queue(struct net_device *dev)
1851 {
1852 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1853 }
1854
1855 static inline void netif_tx_start_all_queues(struct net_device *dev)
1856 {
1857 unsigned int i;
1858
1859 for (i = 0; i < dev->num_tx_queues; i++) {
1860 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1861 netif_tx_start_queue(txq);
1862 }
1863 }
1864
1865 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1866 {
1867 #ifdef CONFIG_NETPOLL_TRAP
1868 if (netpoll_trap()) {
1869 netif_tx_start_queue(dev_queue);
1870 return;
1871 }
1872 #endif
1873 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1874 __netif_schedule(dev_queue->qdisc);
1875 }
1876
1877 /**
1878 * netif_wake_queue - restart transmit
1879 * @dev: network device
1880 *
1881 * Allow upper layers to call the device hard_start_xmit routine.
1882 * Used for flow control when transmit resources are available.
1883 */
1884 static inline void netif_wake_queue(struct net_device *dev)
1885 {
1886 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1887 }
1888
1889 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1890 {
1891 unsigned int i;
1892
1893 for (i = 0; i < dev->num_tx_queues; i++) {
1894 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1895 netif_tx_wake_queue(txq);
1896 }
1897 }
1898
1899 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1900 {
1901 if (WARN_ON(!dev_queue)) {
1902 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1903 return;
1904 }
1905 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1906 }
1907
1908 /**
1909 * netif_stop_queue - stop transmitted packets
1910 * @dev: network device
1911 *
1912 * Stop upper layers calling the device hard_start_xmit routine.
1913 * Used for flow control when transmit resources are unavailable.
1914 */
1915 static inline void netif_stop_queue(struct net_device *dev)
1916 {
1917 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1918 }
1919
1920 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1921 {
1922 unsigned int i;
1923
1924 for (i = 0; i < dev->num_tx_queues; i++) {
1925 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1926 netif_tx_stop_queue(txq);
1927 }
1928 }
1929
1930 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1931 {
1932 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1933 }
1934
1935 /**
1936 * netif_queue_stopped - test if transmit queue is flowblocked
1937 * @dev: network device
1938 *
1939 * Test if transmit queue on device is currently unable to send.
1940 */
1941 static inline bool netif_queue_stopped(const struct net_device *dev)
1942 {
1943 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1944 }
1945
1946 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
1947 {
1948 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1949 }
1950
1951 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
1952 {
1953 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
1954 }
1955
1956 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
1957 unsigned int bytes)
1958 {
1959 #ifdef CONFIG_BQL
1960 dql_queued(&dev_queue->dql, bytes);
1961
1962 if (likely(dql_avail(&dev_queue->dql) >= 0))
1963 return;
1964
1965 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1966
1967 /*
1968 * The XOFF flag must be set before checking the dql_avail below,
1969 * because in netdev_tx_completed_queue we update the dql_completed
1970 * before checking the XOFF flag.
1971 */
1972 smp_mb();
1973
1974 /* check again in case another CPU has just made room avail */
1975 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
1976 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1977 #endif
1978 }
1979
1980 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
1981 {
1982 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
1983 }
1984
1985 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
1986 unsigned int pkts, unsigned int bytes)
1987 {
1988 #ifdef CONFIG_BQL
1989 if (unlikely(!bytes))
1990 return;
1991
1992 dql_completed(&dev_queue->dql, bytes);
1993
1994 /*
1995 * Without the memory barrier there is a small possiblity that
1996 * netdev_tx_sent_queue will miss the update and cause the queue to
1997 * be stopped forever
1998 */
1999 smp_mb();
2000
2001 if (dql_avail(&dev_queue->dql) < 0)
2002 return;
2003
2004 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2005 netif_schedule_queue(dev_queue);
2006 #endif
2007 }
2008
2009 static inline void netdev_completed_queue(struct net_device *dev,
2010 unsigned int pkts, unsigned int bytes)
2011 {
2012 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2013 }
2014
2015 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2016 {
2017 #ifdef CONFIG_BQL
2018 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2019 dql_reset(&q->dql);
2020 #endif
2021 }
2022
2023 static inline void netdev_reset_queue(struct net_device *dev_queue)
2024 {
2025 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2026 }
2027
2028 /**
2029 * netif_running - test if up
2030 * @dev: network device
2031 *
2032 * Test if the device has been brought up.
2033 */
2034 static inline bool netif_running(const struct net_device *dev)
2035 {
2036 return test_bit(__LINK_STATE_START, &dev->state);
2037 }
2038
2039 /*
2040 * Routines to manage the subqueues on a device. We only need start
2041 * stop, and a check if it's stopped. All other device management is
2042 * done at the overall netdevice level.
2043 * Also test the device if we're multiqueue.
2044 */
2045
2046 /**
2047 * netif_start_subqueue - allow sending packets on subqueue
2048 * @dev: network device
2049 * @queue_index: sub queue index
2050 *
2051 * Start individual transmit queue of a device with multiple transmit queues.
2052 */
2053 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2054 {
2055 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2056
2057 netif_tx_start_queue(txq);
2058 }
2059
2060 /**
2061 * netif_stop_subqueue - stop sending packets on subqueue
2062 * @dev: network device
2063 * @queue_index: sub queue index
2064 *
2065 * Stop individual transmit queue of a device with multiple transmit queues.
2066 */
2067 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2068 {
2069 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2070 #ifdef CONFIG_NETPOLL_TRAP
2071 if (netpoll_trap())
2072 return;
2073 #endif
2074 netif_tx_stop_queue(txq);
2075 }
2076
2077 /**
2078 * netif_subqueue_stopped - test status of subqueue
2079 * @dev: network device
2080 * @queue_index: sub queue index
2081 *
2082 * Check individual transmit queue of a device with multiple transmit queues.
2083 */
2084 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2085 u16 queue_index)
2086 {
2087 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2088
2089 return netif_tx_queue_stopped(txq);
2090 }
2091
2092 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2093 struct sk_buff *skb)
2094 {
2095 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2096 }
2097
2098 /**
2099 * netif_wake_subqueue - allow sending packets on subqueue
2100 * @dev: network device
2101 * @queue_index: sub queue index
2102 *
2103 * Resume individual transmit queue of a device with multiple transmit queues.
2104 */
2105 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2106 {
2107 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2108 #ifdef CONFIG_NETPOLL_TRAP
2109 if (netpoll_trap())
2110 return;
2111 #endif
2112 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2113 __netif_schedule(txq->qdisc);
2114 }
2115
2116 #ifdef CONFIG_XPS
2117 extern int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask,
2118 u16 index);
2119 #else
2120 static inline int netif_set_xps_queue(struct net_device *dev,
2121 struct cpumask *mask,
2122 u16 index)
2123 {
2124 return 0;
2125 }
2126 #endif
2127
2128 /*
2129 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2130 * as a distribution range limit for the returned value.
2131 */
2132 static inline u16 skb_tx_hash(const struct net_device *dev,
2133 const struct sk_buff *skb)
2134 {
2135 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2136 }
2137
2138 /**
2139 * netif_is_multiqueue - test if device has multiple transmit queues
2140 * @dev: network device
2141 *
2142 * Check if device has multiple transmit queues
2143 */
2144 static inline bool netif_is_multiqueue(const struct net_device *dev)
2145 {
2146 return dev->num_tx_queues > 1;
2147 }
2148
2149 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2150 unsigned int txq);
2151
2152 #ifdef CONFIG_RPS
2153 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2154 unsigned int rxq);
2155 #else
2156 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2157 unsigned int rxq)
2158 {
2159 return 0;
2160 }
2161 #endif
2162
2163 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2164 const struct net_device *from_dev)
2165 {
2166 int err;
2167
2168 err = netif_set_real_num_tx_queues(to_dev,
2169 from_dev->real_num_tx_queues);
2170 if (err)
2171 return err;
2172 #ifdef CONFIG_RPS
2173 return netif_set_real_num_rx_queues(to_dev,
2174 from_dev->real_num_rx_queues);
2175 #else
2176 return 0;
2177 #endif
2178 }
2179
2180 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2181 extern int netif_get_num_default_rss_queues(void);
2182
2183 /* Use this variant when it is known for sure that it
2184 * is executing from hardware interrupt context or with hardware interrupts
2185 * disabled.
2186 */
2187 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2188
2189 /* Use this variant in places where it could be invoked
2190 * from either hardware interrupt or other context, with hardware interrupts
2191 * either disabled or enabled.
2192 */
2193 extern void dev_kfree_skb_any(struct sk_buff *skb);
2194
2195 extern int netif_rx(struct sk_buff *skb);
2196 extern int netif_rx_ni(struct sk_buff *skb);
2197 extern int netif_receive_skb(struct sk_buff *skb);
2198 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2199 struct sk_buff *skb);
2200 extern void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2201 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2202 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2203
2204 static inline void napi_free_frags(struct napi_struct *napi)
2205 {
2206 kfree_skb(napi->skb);
2207 napi->skb = NULL;
2208 }
2209
2210 extern int netdev_rx_handler_register(struct net_device *dev,
2211 rx_handler_func_t *rx_handler,
2212 void *rx_handler_data);
2213 extern void netdev_rx_handler_unregister(struct net_device *dev);
2214
2215 extern bool dev_valid_name(const char *name);
2216 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2217 extern int dev_ethtool(struct net *net, struct ifreq *);
2218 extern unsigned int dev_get_flags(const struct net_device *);
2219 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2220 extern int dev_change_flags(struct net_device *, unsigned int);
2221 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2222 extern int dev_change_name(struct net_device *, const char *);
2223 extern int dev_set_alias(struct net_device *, const char *, size_t);
2224 extern int dev_change_net_namespace(struct net_device *,
2225 struct net *, const char *);
2226 extern int dev_set_mtu(struct net_device *, int);
2227 extern void dev_set_group(struct net_device *, int);
2228 extern int dev_set_mac_address(struct net_device *,
2229 struct sockaddr *);
2230 extern int dev_change_carrier(struct net_device *,
2231 bool new_carrier);
2232 extern int dev_hard_start_xmit(struct sk_buff *skb,
2233 struct net_device *dev,
2234 struct netdev_queue *txq);
2235 extern int dev_forward_skb(struct net_device *dev,
2236 struct sk_buff *skb);
2237
2238 extern int netdev_budget;
2239
2240 /* Called by rtnetlink.c:rtnl_unlock() */
2241 extern void netdev_run_todo(void);
2242
2243 /**
2244 * dev_put - release reference to device
2245 * @dev: network device
2246 *
2247 * Release reference to device to allow it to be freed.
2248 */
2249 static inline void dev_put(struct net_device *dev)
2250 {
2251 this_cpu_dec(*dev->pcpu_refcnt);
2252 }
2253
2254 /**
2255 * dev_hold - get reference to device
2256 * @dev: network device
2257 *
2258 * Hold reference to device to keep it from being freed.
2259 */
2260 static inline void dev_hold(struct net_device *dev)
2261 {
2262 this_cpu_inc(*dev->pcpu_refcnt);
2263 }
2264
2265 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2266 * and _off may be called from IRQ context, but it is caller
2267 * who is responsible for serialization of these calls.
2268 *
2269 * The name carrier is inappropriate, these functions should really be
2270 * called netif_lowerlayer_*() because they represent the state of any
2271 * kind of lower layer not just hardware media.
2272 */
2273
2274 extern void linkwatch_init_dev(struct net_device *dev);
2275 extern void linkwatch_fire_event(struct net_device *dev);
2276 extern void linkwatch_forget_dev(struct net_device *dev);
2277
2278 /**
2279 * netif_carrier_ok - test if carrier present
2280 * @dev: network device
2281 *
2282 * Check if carrier is present on device
2283 */
2284 static inline bool netif_carrier_ok(const struct net_device *dev)
2285 {
2286 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2287 }
2288
2289 extern unsigned long dev_trans_start(struct net_device *dev);
2290
2291 extern void __netdev_watchdog_up(struct net_device *dev);
2292
2293 extern void netif_carrier_on(struct net_device *dev);
2294
2295 extern void netif_carrier_off(struct net_device *dev);
2296
2297 /**
2298 * netif_dormant_on - mark device as dormant.
2299 * @dev: network device
2300 *
2301 * Mark device as dormant (as per RFC2863).
2302 *
2303 * The dormant state indicates that the relevant interface is not
2304 * actually in a condition to pass packets (i.e., it is not 'up') but is
2305 * in a "pending" state, waiting for some external event. For "on-
2306 * demand" interfaces, this new state identifies the situation where the
2307 * interface is waiting for events to place it in the up state.
2308 *
2309 */
2310 static inline void netif_dormant_on(struct net_device *dev)
2311 {
2312 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2313 linkwatch_fire_event(dev);
2314 }
2315
2316 /**
2317 * netif_dormant_off - set device as not dormant.
2318 * @dev: network device
2319 *
2320 * Device is not in dormant state.
2321 */
2322 static inline void netif_dormant_off(struct net_device *dev)
2323 {
2324 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2325 linkwatch_fire_event(dev);
2326 }
2327
2328 /**
2329 * netif_dormant - test if carrier present
2330 * @dev: network device
2331 *
2332 * Check if carrier is present on device
2333 */
2334 static inline bool netif_dormant(const struct net_device *dev)
2335 {
2336 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2337 }
2338
2339
2340 /**
2341 * netif_oper_up - test if device is operational
2342 * @dev: network device
2343 *
2344 * Check if carrier is operational
2345 */
2346 static inline bool netif_oper_up(const struct net_device *dev)
2347 {
2348 return (dev->operstate == IF_OPER_UP ||
2349 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2350 }
2351
2352 /**
2353 * netif_device_present - is device available or removed
2354 * @dev: network device
2355 *
2356 * Check if device has not been removed from system.
2357 */
2358 static inline bool netif_device_present(struct net_device *dev)
2359 {
2360 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2361 }
2362
2363 extern void netif_device_detach(struct net_device *dev);
2364
2365 extern void netif_device_attach(struct net_device *dev);
2366
2367 /*
2368 * Network interface message level settings
2369 */
2370
2371 enum {
2372 NETIF_MSG_DRV = 0x0001,
2373 NETIF_MSG_PROBE = 0x0002,
2374 NETIF_MSG_LINK = 0x0004,
2375 NETIF_MSG_TIMER = 0x0008,
2376 NETIF_MSG_IFDOWN = 0x0010,
2377 NETIF_MSG_IFUP = 0x0020,
2378 NETIF_MSG_RX_ERR = 0x0040,
2379 NETIF_MSG_TX_ERR = 0x0080,
2380 NETIF_MSG_TX_QUEUED = 0x0100,
2381 NETIF_MSG_INTR = 0x0200,
2382 NETIF_MSG_TX_DONE = 0x0400,
2383 NETIF_MSG_RX_STATUS = 0x0800,
2384 NETIF_MSG_PKTDATA = 0x1000,
2385 NETIF_MSG_HW = 0x2000,
2386 NETIF_MSG_WOL = 0x4000,
2387 };
2388
2389 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2390 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2391 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2392 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2393 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2394 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2395 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2396 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2397 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2398 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2399 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2400 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2401 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2402 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2403 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2404
2405 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2406 {
2407 /* use default */
2408 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2409 return default_msg_enable_bits;
2410 if (debug_value == 0) /* no output */
2411 return 0;
2412 /* set low N bits */
2413 return (1 << debug_value) - 1;
2414 }
2415
2416 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2417 {
2418 spin_lock(&txq->_xmit_lock);
2419 txq->xmit_lock_owner = cpu;
2420 }
2421
2422 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2423 {
2424 spin_lock_bh(&txq->_xmit_lock);
2425 txq->xmit_lock_owner = smp_processor_id();
2426 }
2427
2428 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2429 {
2430 bool ok = spin_trylock(&txq->_xmit_lock);
2431 if (likely(ok))
2432 txq->xmit_lock_owner = smp_processor_id();
2433 return ok;
2434 }
2435
2436 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2437 {
2438 txq->xmit_lock_owner = -1;
2439 spin_unlock(&txq->_xmit_lock);
2440 }
2441
2442 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2443 {
2444 txq->xmit_lock_owner = -1;
2445 spin_unlock_bh(&txq->_xmit_lock);
2446 }
2447
2448 static inline void txq_trans_update(struct netdev_queue *txq)
2449 {
2450 if (txq->xmit_lock_owner != -1)
2451 txq->trans_start = jiffies;
2452 }
2453
2454 /**
2455 * netif_tx_lock - grab network device transmit lock
2456 * @dev: network device
2457 *
2458 * Get network device transmit lock
2459 */
2460 static inline void netif_tx_lock(struct net_device *dev)
2461 {
2462 unsigned int i;
2463 int cpu;
2464
2465 spin_lock(&dev->tx_global_lock);
2466 cpu = smp_processor_id();
2467 for (i = 0; i < dev->num_tx_queues; i++) {
2468 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2469
2470 /* We are the only thread of execution doing a
2471 * freeze, but we have to grab the _xmit_lock in
2472 * order to synchronize with threads which are in
2473 * the ->hard_start_xmit() handler and already
2474 * checked the frozen bit.
2475 */
2476 __netif_tx_lock(txq, cpu);
2477 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2478 __netif_tx_unlock(txq);
2479 }
2480 }
2481
2482 static inline void netif_tx_lock_bh(struct net_device *dev)
2483 {
2484 local_bh_disable();
2485 netif_tx_lock(dev);
2486 }
2487
2488 static inline void netif_tx_unlock(struct net_device *dev)
2489 {
2490 unsigned int i;
2491
2492 for (i = 0; i < dev->num_tx_queues; i++) {
2493 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2494
2495 /* No need to grab the _xmit_lock here. If the
2496 * queue is not stopped for another reason, we
2497 * force a schedule.
2498 */
2499 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2500 netif_schedule_queue(txq);
2501 }
2502 spin_unlock(&dev->tx_global_lock);
2503 }
2504
2505 static inline void netif_tx_unlock_bh(struct net_device *dev)
2506 {
2507 netif_tx_unlock(dev);
2508 local_bh_enable();
2509 }
2510
2511 #define HARD_TX_LOCK(dev, txq, cpu) { \
2512 if ((dev->features & NETIF_F_LLTX) == 0) { \
2513 __netif_tx_lock(txq, cpu); \
2514 } \
2515 }
2516
2517 #define HARD_TX_UNLOCK(dev, txq) { \
2518 if ((dev->features & NETIF_F_LLTX) == 0) { \
2519 __netif_tx_unlock(txq); \
2520 } \
2521 }
2522
2523 static inline void netif_tx_disable(struct net_device *dev)
2524 {
2525 unsigned int i;
2526 int cpu;
2527
2528 local_bh_disable();
2529 cpu = smp_processor_id();
2530 for (i = 0; i < dev->num_tx_queues; i++) {
2531 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2532
2533 __netif_tx_lock(txq, cpu);
2534 netif_tx_stop_queue(txq);
2535 __netif_tx_unlock(txq);
2536 }
2537 local_bh_enable();
2538 }
2539
2540 static inline void netif_addr_lock(struct net_device *dev)
2541 {
2542 spin_lock(&dev->addr_list_lock);
2543 }
2544
2545 static inline void netif_addr_lock_nested(struct net_device *dev)
2546 {
2547 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2548 }
2549
2550 static inline void netif_addr_lock_bh(struct net_device *dev)
2551 {
2552 spin_lock_bh(&dev->addr_list_lock);
2553 }
2554
2555 static inline void netif_addr_unlock(struct net_device *dev)
2556 {
2557 spin_unlock(&dev->addr_list_lock);
2558 }
2559
2560 static inline void netif_addr_unlock_bh(struct net_device *dev)
2561 {
2562 spin_unlock_bh(&dev->addr_list_lock);
2563 }
2564
2565 /*
2566 * dev_addrs walker. Should be used only for read access. Call with
2567 * rcu_read_lock held.
2568 */
2569 #define for_each_dev_addr(dev, ha) \
2570 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2571
2572 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2573
2574 extern void ether_setup(struct net_device *dev);
2575
2576 /* Support for loadable net-drivers */
2577 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2578 void (*setup)(struct net_device *),
2579 unsigned int txqs, unsigned int rxqs);
2580 #define alloc_netdev(sizeof_priv, name, setup) \
2581 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2582
2583 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2584 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2585
2586 extern int register_netdev(struct net_device *dev);
2587 extern void unregister_netdev(struct net_device *dev);
2588
2589 /* General hardware address lists handling functions */
2590 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2591 struct netdev_hw_addr_list *from_list,
2592 int addr_len, unsigned char addr_type);
2593 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2594 struct netdev_hw_addr_list *from_list,
2595 int addr_len, unsigned char addr_type);
2596 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2597 struct netdev_hw_addr_list *from_list,
2598 int addr_len);
2599 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2600 struct netdev_hw_addr_list *from_list,
2601 int addr_len);
2602 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2603 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2604
2605 /* Functions used for device addresses handling */
2606 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2607 unsigned char addr_type);
2608 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2609 unsigned char addr_type);
2610 extern int dev_addr_add_multiple(struct net_device *to_dev,
2611 struct net_device *from_dev,
2612 unsigned char addr_type);
2613 extern int dev_addr_del_multiple(struct net_device *to_dev,
2614 struct net_device *from_dev,
2615 unsigned char addr_type);
2616 extern void dev_addr_flush(struct net_device *dev);
2617 extern int dev_addr_init(struct net_device *dev);
2618
2619 /* Functions used for unicast addresses handling */
2620 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2621 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2622 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2623 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2624 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2625 extern void dev_uc_flush(struct net_device *dev);
2626 extern void dev_uc_init(struct net_device *dev);
2627
2628 /* Functions used for multicast addresses handling */
2629 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2630 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2631 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2632 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2633 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2634 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2635 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2636 extern void dev_mc_flush(struct net_device *dev);
2637 extern void dev_mc_init(struct net_device *dev);
2638
2639 /* Functions used for secondary unicast and multicast support */
2640 extern void dev_set_rx_mode(struct net_device *dev);
2641 extern void __dev_set_rx_mode(struct net_device *dev);
2642 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2643 extern int dev_set_allmulti(struct net_device *dev, int inc);
2644 extern void netdev_state_change(struct net_device *dev);
2645 extern void netdev_notify_peers(struct net_device *dev);
2646 extern void netdev_features_change(struct net_device *dev);
2647 /* Load a device via the kmod */
2648 extern void dev_load(struct net *net, const char *name);
2649 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2650 struct rtnl_link_stats64 *storage);
2651 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2652 const struct net_device_stats *netdev_stats);
2653
2654 extern int netdev_max_backlog;
2655 extern int netdev_tstamp_prequeue;
2656 extern int weight_p;
2657 extern int bpf_jit_enable;
2658
2659 extern bool netdev_has_upper_dev(struct net_device *dev,
2660 struct net_device *upper_dev);
2661 extern bool netdev_has_any_upper_dev(struct net_device *dev);
2662 extern struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2663 extern struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2664 extern int netdev_upper_dev_link(struct net_device *dev,
2665 struct net_device *upper_dev);
2666 extern int netdev_master_upper_dev_link(struct net_device *dev,
2667 struct net_device *upper_dev);
2668 extern void netdev_upper_dev_unlink(struct net_device *dev,
2669 struct net_device *upper_dev);
2670 extern int skb_checksum_help(struct sk_buff *skb);
2671 extern struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2672 netdev_features_t features, bool tx_path);
2673 extern struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2674 netdev_features_t features);
2675
2676 static inline
2677 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2678 {
2679 return __skb_gso_segment(skb, features, true);
2680 }
2681
2682 #ifdef CONFIG_BUG
2683 extern void netdev_rx_csum_fault(struct net_device *dev);
2684 #else
2685 static inline void netdev_rx_csum_fault(struct net_device *dev)
2686 {
2687 }
2688 #endif
2689 /* rx skb timestamps */
2690 extern void net_enable_timestamp(void);
2691 extern void net_disable_timestamp(void);
2692
2693 #ifdef CONFIG_PROC_FS
2694 extern int __init dev_proc_init(void);
2695 #else
2696 #define dev_proc_init() 0
2697 #endif
2698
2699 extern int netdev_class_create_file(struct class_attribute *class_attr);
2700 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2701
2702 extern struct kobj_ns_type_operations net_ns_type_operations;
2703
2704 extern const char *netdev_drivername(const struct net_device *dev);
2705
2706 extern void linkwatch_run_queue(void);
2707
2708 static inline netdev_features_t netdev_get_wanted_features(
2709 struct net_device *dev)
2710 {
2711 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2712 }
2713 netdev_features_t netdev_increment_features(netdev_features_t all,
2714 netdev_features_t one, netdev_features_t mask);
2715 int __netdev_update_features(struct net_device *dev);
2716 void netdev_update_features(struct net_device *dev);
2717 void netdev_change_features(struct net_device *dev);
2718
2719 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2720 struct net_device *dev);
2721
2722 netdev_features_t netif_skb_features(struct sk_buff *skb);
2723
2724 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2725 {
2726 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2727
2728 /* check flags correspondence */
2729 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2730 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2731 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2732 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2733 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2734 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2735
2736 return (features & feature) == feature;
2737 }
2738
2739 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2740 {
2741 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2742 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2743 }
2744
2745 static inline bool netif_needs_gso(struct sk_buff *skb,
2746 netdev_features_t features)
2747 {
2748 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2749 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2750 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2751 }
2752
2753 static inline void netif_set_gso_max_size(struct net_device *dev,
2754 unsigned int size)
2755 {
2756 dev->gso_max_size = size;
2757 }
2758
2759 static inline bool netif_is_bond_slave(struct net_device *dev)
2760 {
2761 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2762 }
2763
2764 static inline bool netif_supports_nofcs(struct net_device *dev)
2765 {
2766 return dev->priv_flags & IFF_SUPP_NOFCS;
2767 }
2768
2769 extern struct pernet_operations __net_initdata loopback_net_ops;
2770
2771 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2772
2773 /* netdev_printk helpers, similar to dev_printk */
2774
2775 static inline const char *netdev_name(const struct net_device *dev)
2776 {
2777 if (dev->reg_state != NETREG_REGISTERED)
2778 return "(unregistered net_device)";
2779 return dev->name;
2780 }
2781
2782 extern __printf(3, 4)
2783 int netdev_printk(const char *level, const struct net_device *dev,
2784 const char *format, ...);
2785 extern __printf(2, 3)
2786 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2787 extern __printf(2, 3)
2788 int netdev_alert(const struct net_device *dev, const char *format, ...);
2789 extern __printf(2, 3)
2790 int netdev_crit(const struct net_device *dev, const char *format, ...);
2791 extern __printf(2, 3)
2792 int netdev_err(const struct net_device *dev, const char *format, ...);
2793 extern __printf(2, 3)
2794 int netdev_warn(const struct net_device *dev, const char *format, ...);
2795 extern __printf(2, 3)
2796 int netdev_notice(const struct net_device *dev, const char *format, ...);
2797 extern __printf(2, 3)
2798 int netdev_info(const struct net_device *dev, const char *format, ...);
2799
2800 #define MODULE_ALIAS_NETDEV(device) \
2801 MODULE_ALIAS("netdev-" device)
2802
2803 #if defined(CONFIG_DYNAMIC_DEBUG)
2804 #define netdev_dbg(__dev, format, args...) \
2805 do { \
2806 dynamic_netdev_dbg(__dev, format, ##args); \
2807 } while (0)
2808 #elif defined(DEBUG)
2809 #define netdev_dbg(__dev, format, args...) \
2810 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2811 #else
2812 #define netdev_dbg(__dev, format, args...) \
2813 ({ \
2814 if (0) \
2815 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2816 0; \
2817 })
2818 #endif
2819
2820 #if defined(VERBOSE_DEBUG)
2821 #define netdev_vdbg netdev_dbg
2822 #else
2823
2824 #define netdev_vdbg(dev, format, args...) \
2825 ({ \
2826 if (0) \
2827 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2828 0; \
2829 })
2830 #endif
2831
2832 /*
2833 * netdev_WARN() acts like dev_printk(), but with the key difference
2834 * of using a WARN/WARN_ON to get the message out, including the
2835 * file/line information and a backtrace.
2836 */
2837 #define netdev_WARN(dev, format, args...) \
2838 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2839
2840 /* netif printk helpers, similar to netdev_printk */
2841
2842 #define netif_printk(priv, type, level, dev, fmt, args...) \
2843 do { \
2844 if (netif_msg_##type(priv)) \
2845 netdev_printk(level, (dev), fmt, ##args); \
2846 } while (0)
2847
2848 #define netif_level(level, priv, type, dev, fmt, args...) \
2849 do { \
2850 if (netif_msg_##type(priv)) \
2851 netdev_##level(dev, fmt, ##args); \
2852 } while (0)
2853
2854 #define netif_emerg(priv, type, dev, fmt, args...) \
2855 netif_level(emerg, priv, type, dev, fmt, ##args)
2856 #define netif_alert(priv, type, dev, fmt, args...) \
2857 netif_level(alert, priv, type, dev, fmt, ##args)
2858 #define netif_crit(priv, type, dev, fmt, args...) \
2859 netif_level(crit, priv, type, dev, fmt, ##args)
2860 #define netif_err(priv, type, dev, fmt, args...) \
2861 netif_level(err, priv, type, dev, fmt, ##args)
2862 #define netif_warn(priv, type, dev, fmt, args...) \
2863 netif_level(warn, priv, type, dev, fmt, ##args)
2864 #define netif_notice(priv, type, dev, fmt, args...) \
2865 netif_level(notice, priv, type, dev, fmt, ##args)
2866 #define netif_info(priv, type, dev, fmt, args...) \
2867 netif_level(info, priv, type, dev, fmt, ##args)
2868
2869 #if defined(CONFIG_DYNAMIC_DEBUG)
2870 #define netif_dbg(priv, type, netdev, format, args...) \
2871 do { \
2872 if (netif_msg_##type(priv)) \
2873 dynamic_netdev_dbg(netdev, format, ##args); \
2874 } while (0)
2875 #elif defined(DEBUG)
2876 #define netif_dbg(priv, type, dev, format, args...) \
2877 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2878 #else
2879 #define netif_dbg(priv, type, dev, format, args...) \
2880 ({ \
2881 if (0) \
2882 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2883 0; \
2884 })
2885 #endif
2886
2887 #if defined(VERBOSE_DEBUG)
2888 #define netif_vdbg netif_dbg
2889 #else
2890 #define netif_vdbg(priv, type, dev, format, args...) \
2891 ({ \
2892 if (0) \
2893 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2894 0; \
2895 })
2896 #endif
2897
2898 /*
2899 * The list of packet types we will receive (as opposed to discard)
2900 * and the routines to invoke.
2901 *
2902 * Why 16. Because with 16 the only overlap we get on a hash of the
2903 * low nibble of the protocol value is RARP/SNAP/X.25.
2904 *
2905 * NOTE: That is no longer true with the addition of VLAN tags. Not
2906 * sure which should go first, but I bet it won't make much
2907 * difference if we are running VLANs. The good news is that
2908 * this protocol won't be in the list unless compiled in, so
2909 * the average user (w/out VLANs) will not be adversely affected.
2910 * --BLG
2911 *
2912 * 0800 IP
2913 * 8100 802.1Q VLAN
2914 * 0001 802.3
2915 * 0002 AX.25
2916 * 0004 802.2
2917 * 8035 RARP
2918 * 0005 SNAP
2919 * 0805 X.25
2920 * 0806 ARP
2921 * 8137 IPX
2922 * 0009 Localtalk
2923 * 86DD IPv6
2924 */
2925 #define PTYPE_HASH_SIZE (16)
2926 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
2927
2928 #endif /* _LINUX_NETDEVICE_H */