Merge tag 'jfs-3.10' of git://github.com/kleikamp/linux-shaggy
[deliverable/linux.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53
54 struct netpoll_info;
55 struct device;
56 struct phy_device;
57 /* 802.11 specific */
58 struct wireless_dev;
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
62
63 extern void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
65
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
72
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
76
77 /*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
92 */
93
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
109
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
115 };
116 typedef enum netdev_tx netdev_tx_t;
117
118 /*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122 static inline bool dev_xmit_complete(int rc)
123 {
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134 }
135
136 /*
137 * Compute the worst case header length according to the protocols
138 * used.
139 */
140
141 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
144 # else
145 # define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
150
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
157
158 /*
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
161 */
162
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187 };
188
189
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
192
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 #endif
197
198 struct neighbour;
199 struct neigh_parms;
200 struct sk_buff;
201
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
211 bool global_use;
212 int sync_cnt;
213 int refcount;
214 int synced;
215 struct rcu_head rcu_head;
216 };
217
218 struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221 };
222
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
227
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237
238 struct hh_cache {
239 u16 hh_len;
240 u16 __pad;
241 seqlock_t hh_lock;
242
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251
252 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264
265 struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*rebuild)(struct sk_buff *skb);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275 };
276
277 /* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
279 * code.
280 */
281
282 enum netdev_state_t {
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
288 };
289
290
291 /*
292 * This structure holds at boot time configured netdevice settings. They
293 * are then used in the device probing.
294 */
295 struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300
301 extern int __init netdev_boot_setup(char *str);
302
303 /*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306 struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-cpu poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 spinlock_t poll_lock;
321 int poll_owner;
322 #endif
323 struct net_device *dev;
324 struct sk_buff *gro_list;
325 struct sk_buff *skb;
326 struct list_head dev_list;
327 };
328
329 enum {
330 NAPI_STATE_SCHED, /* Poll is scheduled */
331 NAPI_STATE_DISABLE, /* Disable pending */
332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
333 };
334
335 enum gro_result {
336 GRO_MERGED,
337 GRO_MERGED_FREE,
338 GRO_HELD,
339 GRO_NORMAL,
340 GRO_DROP,
341 };
342 typedef enum gro_result gro_result_t;
343
344 /*
345 * enum rx_handler_result - Possible return values for rx_handlers.
346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * further.
348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
349 * case skb->dev was changed by rx_handler.
350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
351 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
352 *
353 * rx_handlers are functions called from inside __netif_receive_skb(), to do
354 * special processing of the skb, prior to delivery to protocol handlers.
355 *
356 * Currently, a net_device can only have a single rx_handler registered. Trying
357 * to register a second rx_handler will return -EBUSY.
358 *
359 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
360 * To unregister a rx_handler on a net_device, use
361 * netdev_rx_handler_unregister().
362 *
363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * do with the skb.
365 *
366 * If the rx_handler consumed to skb in some way, it should return
367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
368 * the skb to be delivered in some other ways.
369 *
370 * If the rx_handler changed skb->dev, to divert the skb to another
371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
372 * new device will be called if it exists.
373 *
374 * If the rx_handler consider the skb should be ignored, it should return
375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
376 * are registered on exact device (ptype->dev == skb->dev).
377 *
378 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
379 * delivered, it should return RX_HANDLER_PASS.
380 *
381 * A device without a registered rx_handler will behave as if rx_handler
382 * returned RX_HANDLER_PASS.
383 */
384
385 enum rx_handler_result {
386 RX_HANDLER_CONSUMED,
387 RX_HANDLER_ANOTHER,
388 RX_HANDLER_EXACT,
389 RX_HANDLER_PASS,
390 };
391 typedef enum rx_handler_result rx_handler_result_t;
392 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
393
394 extern void __napi_schedule(struct napi_struct *n);
395
396 static inline bool napi_disable_pending(struct napi_struct *n)
397 {
398 return test_bit(NAPI_STATE_DISABLE, &n->state);
399 }
400
401 /**
402 * napi_schedule_prep - check if napi can be scheduled
403 * @n: napi context
404 *
405 * Test if NAPI routine is already running, and if not mark
406 * it as running. This is used as a condition variable
407 * insure only one NAPI poll instance runs. We also make
408 * sure there is no pending NAPI disable.
409 */
410 static inline bool napi_schedule_prep(struct napi_struct *n)
411 {
412 return !napi_disable_pending(n) &&
413 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
414 }
415
416 /**
417 * napi_schedule - schedule NAPI poll
418 * @n: napi context
419 *
420 * Schedule NAPI poll routine to be called if it is not already
421 * running.
422 */
423 static inline void napi_schedule(struct napi_struct *n)
424 {
425 if (napi_schedule_prep(n))
426 __napi_schedule(n);
427 }
428
429 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
430 static inline bool napi_reschedule(struct napi_struct *napi)
431 {
432 if (napi_schedule_prep(napi)) {
433 __napi_schedule(napi);
434 return true;
435 }
436 return false;
437 }
438
439 /**
440 * napi_complete - NAPI processing complete
441 * @n: napi context
442 *
443 * Mark NAPI processing as complete.
444 */
445 extern void __napi_complete(struct napi_struct *n);
446 extern void napi_complete(struct napi_struct *n);
447
448 /**
449 * napi_disable - prevent NAPI from scheduling
450 * @n: napi context
451 *
452 * Stop NAPI from being scheduled on this context.
453 * Waits till any outstanding processing completes.
454 */
455 static inline void napi_disable(struct napi_struct *n)
456 {
457 set_bit(NAPI_STATE_DISABLE, &n->state);
458 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
459 msleep(1);
460 clear_bit(NAPI_STATE_DISABLE, &n->state);
461 }
462
463 /**
464 * napi_enable - enable NAPI scheduling
465 * @n: napi context
466 *
467 * Resume NAPI from being scheduled on this context.
468 * Must be paired with napi_disable.
469 */
470 static inline void napi_enable(struct napi_struct *n)
471 {
472 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
473 smp_mb__before_clear_bit();
474 clear_bit(NAPI_STATE_SCHED, &n->state);
475 }
476
477 #ifdef CONFIG_SMP
478 /**
479 * napi_synchronize - wait until NAPI is not running
480 * @n: napi context
481 *
482 * Wait until NAPI is done being scheduled on this context.
483 * Waits till any outstanding processing completes but
484 * does not disable future activations.
485 */
486 static inline void napi_synchronize(const struct napi_struct *n)
487 {
488 while (test_bit(NAPI_STATE_SCHED, &n->state))
489 msleep(1);
490 }
491 #else
492 # define napi_synchronize(n) barrier()
493 #endif
494
495 enum netdev_queue_state_t {
496 __QUEUE_STATE_DRV_XOFF,
497 __QUEUE_STATE_STACK_XOFF,
498 __QUEUE_STATE_FROZEN,
499 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
500 (1 << __QUEUE_STATE_STACK_XOFF))
501 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
502 (1 << __QUEUE_STATE_FROZEN))
503 };
504 /*
505 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
506 * netif_tx_* functions below are used to manipulate this flag. The
507 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
508 * queue independently. The netif_xmit_*stopped functions below are called
509 * to check if the queue has been stopped by the driver or stack (either
510 * of the XOFF bits are set in the state). Drivers should not need to call
511 * netif_xmit*stopped functions, they should only be using netif_tx_*.
512 */
513
514 struct netdev_queue {
515 /*
516 * read mostly part
517 */
518 struct net_device *dev;
519 struct Qdisc *qdisc;
520 struct Qdisc *qdisc_sleeping;
521 #ifdef CONFIG_SYSFS
522 struct kobject kobj;
523 #endif
524 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
525 int numa_node;
526 #endif
527 /*
528 * write mostly part
529 */
530 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
531 int xmit_lock_owner;
532 /*
533 * please use this field instead of dev->trans_start
534 */
535 unsigned long trans_start;
536
537 /*
538 * Number of TX timeouts for this queue
539 * (/sys/class/net/DEV/Q/trans_timeout)
540 */
541 unsigned long trans_timeout;
542
543 unsigned long state;
544
545 #ifdef CONFIG_BQL
546 struct dql dql;
547 #endif
548 } ____cacheline_aligned_in_smp;
549
550 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
551 {
552 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
553 return q->numa_node;
554 #else
555 return NUMA_NO_NODE;
556 #endif
557 }
558
559 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
560 {
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 q->numa_node = node;
563 #endif
564 }
565
566 #ifdef CONFIG_RPS
567 /*
568 * This structure holds an RPS map which can be of variable length. The
569 * map is an array of CPUs.
570 */
571 struct rps_map {
572 unsigned int len;
573 struct rcu_head rcu;
574 u16 cpus[0];
575 };
576 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
577
578 /*
579 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
580 * tail pointer for that CPU's input queue at the time of last enqueue, and
581 * a hardware filter index.
582 */
583 struct rps_dev_flow {
584 u16 cpu;
585 u16 filter;
586 unsigned int last_qtail;
587 };
588 #define RPS_NO_FILTER 0xffff
589
590 /*
591 * The rps_dev_flow_table structure contains a table of flow mappings.
592 */
593 struct rps_dev_flow_table {
594 unsigned int mask;
595 struct rcu_head rcu;
596 struct work_struct free_work;
597 struct rps_dev_flow flows[0];
598 };
599 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
600 ((_num) * sizeof(struct rps_dev_flow)))
601
602 /*
603 * The rps_sock_flow_table contains mappings of flows to the last CPU
604 * on which they were processed by the application (set in recvmsg).
605 */
606 struct rps_sock_flow_table {
607 unsigned int mask;
608 u16 ents[0];
609 };
610 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
611 ((_num) * sizeof(u16)))
612
613 #define RPS_NO_CPU 0xffff
614
615 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
616 u32 hash)
617 {
618 if (table && hash) {
619 unsigned int cpu, index = hash & table->mask;
620
621 /* We only give a hint, preemption can change cpu under us */
622 cpu = raw_smp_processor_id();
623
624 if (table->ents[index] != cpu)
625 table->ents[index] = cpu;
626 }
627 }
628
629 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
630 u32 hash)
631 {
632 if (table && hash)
633 table->ents[hash & table->mask] = RPS_NO_CPU;
634 }
635
636 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
637
638 #ifdef CONFIG_RFS_ACCEL
639 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
640 u32 flow_id, u16 filter_id);
641 #endif
642
643 /* This structure contains an instance of an RX queue. */
644 struct netdev_rx_queue {
645 struct rps_map __rcu *rps_map;
646 struct rps_dev_flow_table __rcu *rps_flow_table;
647 struct kobject kobj;
648 struct net_device *dev;
649 } ____cacheline_aligned_in_smp;
650 #endif /* CONFIG_RPS */
651
652 #ifdef CONFIG_XPS
653 /*
654 * This structure holds an XPS map which can be of variable length. The
655 * map is an array of queues.
656 */
657 struct xps_map {
658 unsigned int len;
659 unsigned int alloc_len;
660 struct rcu_head rcu;
661 u16 queues[0];
662 };
663 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
664 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
665 / sizeof(u16))
666
667 /*
668 * This structure holds all XPS maps for device. Maps are indexed by CPU.
669 */
670 struct xps_dev_maps {
671 struct rcu_head rcu;
672 struct xps_map __rcu *cpu_map[0];
673 };
674 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
675 (nr_cpu_ids * sizeof(struct xps_map *)))
676 #endif /* CONFIG_XPS */
677
678 #define TC_MAX_QUEUE 16
679 #define TC_BITMASK 15
680 /* HW offloaded queuing disciplines txq count and offset maps */
681 struct netdev_tc_txq {
682 u16 count;
683 u16 offset;
684 };
685
686 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
687 /*
688 * This structure is to hold information about the device
689 * configured to run FCoE protocol stack.
690 */
691 struct netdev_fcoe_hbainfo {
692 char manufacturer[64];
693 char serial_number[64];
694 char hardware_version[64];
695 char driver_version[64];
696 char optionrom_version[64];
697 char firmware_version[64];
698 char model[256];
699 char model_description[256];
700 };
701 #endif
702
703 /*
704 * This structure defines the management hooks for network devices.
705 * The following hooks can be defined; unless noted otherwise, they are
706 * optional and can be filled with a null pointer.
707 *
708 * int (*ndo_init)(struct net_device *dev);
709 * This function is called once when network device is registered.
710 * The network device can use this to any late stage initializaton
711 * or semantic validattion. It can fail with an error code which will
712 * be propogated back to register_netdev
713 *
714 * void (*ndo_uninit)(struct net_device *dev);
715 * This function is called when device is unregistered or when registration
716 * fails. It is not called if init fails.
717 *
718 * int (*ndo_open)(struct net_device *dev);
719 * This function is called when network device transistions to the up
720 * state.
721 *
722 * int (*ndo_stop)(struct net_device *dev);
723 * This function is called when network device transistions to the down
724 * state.
725 *
726 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
727 * struct net_device *dev);
728 * Called when a packet needs to be transmitted.
729 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
730 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
731 * Required can not be NULL.
732 *
733 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
734 * Called to decide which queue to when device supports multiple
735 * transmit queues.
736 *
737 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
738 * This function is called to allow device receiver to make
739 * changes to configuration when multicast or promiscious is enabled.
740 *
741 * void (*ndo_set_rx_mode)(struct net_device *dev);
742 * This function is called device changes address list filtering.
743 * If driver handles unicast address filtering, it should set
744 * IFF_UNICAST_FLT to its priv_flags.
745 *
746 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
747 * This function is called when the Media Access Control address
748 * needs to be changed. If this interface is not defined, the
749 * mac address can not be changed.
750 *
751 * int (*ndo_validate_addr)(struct net_device *dev);
752 * Test if Media Access Control address is valid for the device.
753 *
754 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
755 * Called when a user request an ioctl which can't be handled by
756 * the generic interface code. If not defined ioctl's return
757 * not supported error code.
758 *
759 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
760 * Used to set network devices bus interface parameters. This interface
761 * is retained for legacy reason, new devices should use the bus
762 * interface (PCI) for low level management.
763 *
764 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
765 * Called when a user wants to change the Maximum Transfer Unit
766 * of a device. If not defined, any request to change MTU will
767 * will return an error.
768 *
769 * void (*ndo_tx_timeout)(struct net_device *dev);
770 * Callback uses when the transmitter has not made any progress
771 * for dev->watchdog ticks.
772 *
773 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
774 * struct rtnl_link_stats64 *storage);
775 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
776 * Called when a user wants to get the network device usage
777 * statistics. Drivers must do one of the following:
778 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
779 * rtnl_link_stats64 structure passed by the caller.
780 * 2. Define @ndo_get_stats to update a net_device_stats structure
781 * (which should normally be dev->stats) and return a pointer to
782 * it. The structure may be changed asynchronously only if each
783 * field is written atomically.
784 * 3. Update dev->stats asynchronously and atomically, and define
785 * neither operation.
786 *
787 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
788 * If device support VLAN filtering this function is called when a
789 * VLAN id is registered.
790 *
791 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
792 * If device support VLAN filtering this function is called when a
793 * VLAN id is unregistered.
794 *
795 * void (*ndo_poll_controller)(struct net_device *dev);
796 *
797 * SR-IOV management functions.
798 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
799 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
800 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
801 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
802 * int (*ndo_get_vf_config)(struct net_device *dev,
803 * int vf, struct ifla_vf_info *ivf);
804 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
805 * struct nlattr *port[]);
806 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
807 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
808 * Called to setup 'tc' number of traffic classes in the net device. This
809 * is always called from the stack with the rtnl lock held and netif tx
810 * queues stopped. This allows the netdevice to perform queue management
811 * safely.
812 *
813 * Fiber Channel over Ethernet (FCoE) offload functions.
814 * int (*ndo_fcoe_enable)(struct net_device *dev);
815 * Called when the FCoE protocol stack wants to start using LLD for FCoE
816 * so the underlying device can perform whatever needed configuration or
817 * initialization to support acceleration of FCoE traffic.
818 *
819 * int (*ndo_fcoe_disable)(struct net_device *dev);
820 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
821 * so the underlying device can perform whatever needed clean-ups to
822 * stop supporting acceleration of FCoE traffic.
823 *
824 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
825 * struct scatterlist *sgl, unsigned int sgc);
826 * Called when the FCoE Initiator wants to initialize an I/O that
827 * is a possible candidate for Direct Data Placement (DDP). The LLD can
828 * perform necessary setup and returns 1 to indicate the device is set up
829 * successfully to perform DDP on this I/O, otherwise this returns 0.
830 *
831 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
832 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
833 * indicated by the FC exchange id 'xid', so the underlying device can
834 * clean up and reuse resources for later DDP requests.
835 *
836 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
837 * struct scatterlist *sgl, unsigned int sgc);
838 * Called when the FCoE Target wants to initialize an I/O that
839 * is a possible candidate for Direct Data Placement (DDP). The LLD can
840 * perform necessary setup and returns 1 to indicate the device is set up
841 * successfully to perform DDP on this I/O, otherwise this returns 0.
842 *
843 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
844 * struct netdev_fcoe_hbainfo *hbainfo);
845 * Called when the FCoE Protocol stack wants information on the underlying
846 * device. This information is utilized by the FCoE protocol stack to
847 * register attributes with Fiber Channel management service as per the
848 * FC-GS Fabric Device Management Information(FDMI) specification.
849 *
850 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
851 * Called when the underlying device wants to override default World Wide
852 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
853 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
854 * protocol stack to use.
855 *
856 * RFS acceleration.
857 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
858 * u16 rxq_index, u32 flow_id);
859 * Set hardware filter for RFS. rxq_index is the target queue index;
860 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
861 * Return the filter ID on success, or a negative error code.
862 *
863 * Slave management functions (for bridge, bonding, etc).
864 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
865 * Called to make another netdev an underling.
866 *
867 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
868 * Called to release previously enslaved netdev.
869 *
870 * Feature/offload setting functions.
871 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
872 * netdev_features_t features);
873 * Adjusts the requested feature flags according to device-specific
874 * constraints, and returns the resulting flags. Must not modify
875 * the device state.
876 *
877 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
878 * Called to update device configuration to new features. Passed
879 * feature set might be less than what was returned by ndo_fix_features()).
880 * Must return >0 or -errno if it changed dev->features itself.
881 *
882 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
883 * struct net_device *dev,
884 * const unsigned char *addr, u16 flags)
885 * Adds an FDB entry to dev for addr.
886 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
887 * struct net_device *dev,
888 * const unsigned char *addr)
889 * Deletes the FDB entry from dev coresponding to addr.
890 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
891 * struct net_device *dev, int idx)
892 * Used to add FDB entries to dump requests. Implementers should add
893 * entries to skb and update idx with the number of entries.
894 *
895 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
896 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
897 * struct net_device *dev, u32 filter_mask)
898 *
899 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
900 * Called to change device carrier. Soft-devices (like dummy, team, etc)
901 * which do not represent real hardware may define this to allow their
902 * userspace components to manage their virtual carrier state. Devices
903 * that determine carrier state from physical hardware properties (eg
904 * network cables) or protocol-dependent mechanisms (eg
905 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
906 */
907 struct net_device_ops {
908 int (*ndo_init)(struct net_device *dev);
909 void (*ndo_uninit)(struct net_device *dev);
910 int (*ndo_open)(struct net_device *dev);
911 int (*ndo_stop)(struct net_device *dev);
912 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
913 struct net_device *dev);
914 u16 (*ndo_select_queue)(struct net_device *dev,
915 struct sk_buff *skb);
916 void (*ndo_change_rx_flags)(struct net_device *dev,
917 int flags);
918 void (*ndo_set_rx_mode)(struct net_device *dev);
919 int (*ndo_set_mac_address)(struct net_device *dev,
920 void *addr);
921 int (*ndo_validate_addr)(struct net_device *dev);
922 int (*ndo_do_ioctl)(struct net_device *dev,
923 struct ifreq *ifr, int cmd);
924 int (*ndo_set_config)(struct net_device *dev,
925 struct ifmap *map);
926 int (*ndo_change_mtu)(struct net_device *dev,
927 int new_mtu);
928 int (*ndo_neigh_setup)(struct net_device *dev,
929 struct neigh_parms *);
930 void (*ndo_tx_timeout) (struct net_device *dev);
931
932 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
933 struct rtnl_link_stats64 *storage);
934 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
935
936 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
937 __be16 proto, u16 vid);
938 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
939 __be16 proto, u16 vid);
940 #ifdef CONFIG_NET_POLL_CONTROLLER
941 void (*ndo_poll_controller)(struct net_device *dev);
942 int (*ndo_netpoll_setup)(struct net_device *dev,
943 struct netpoll_info *info,
944 gfp_t gfp);
945 void (*ndo_netpoll_cleanup)(struct net_device *dev);
946 #endif
947 int (*ndo_set_vf_mac)(struct net_device *dev,
948 int queue, u8 *mac);
949 int (*ndo_set_vf_vlan)(struct net_device *dev,
950 int queue, u16 vlan, u8 qos);
951 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
952 int vf, int rate);
953 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
954 int vf, bool setting);
955 int (*ndo_get_vf_config)(struct net_device *dev,
956 int vf,
957 struct ifla_vf_info *ivf);
958 int (*ndo_set_vf_port)(struct net_device *dev,
959 int vf,
960 struct nlattr *port[]);
961 int (*ndo_get_vf_port)(struct net_device *dev,
962 int vf, struct sk_buff *skb);
963 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
964 #if IS_ENABLED(CONFIG_FCOE)
965 int (*ndo_fcoe_enable)(struct net_device *dev);
966 int (*ndo_fcoe_disable)(struct net_device *dev);
967 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
968 u16 xid,
969 struct scatterlist *sgl,
970 unsigned int sgc);
971 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
972 u16 xid);
973 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
974 u16 xid,
975 struct scatterlist *sgl,
976 unsigned int sgc);
977 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
978 struct netdev_fcoe_hbainfo *hbainfo);
979 #endif
980
981 #if IS_ENABLED(CONFIG_LIBFCOE)
982 #define NETDEV_FCOE_WWNN 0
983 #define NETDEV_FCOE_WWPN 1
984 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
985 u64 *wwn, int type);
986 #endif
987
988 #ifdef CONFIG_RFS_ACCEL
989 int (*ndo_rx_flow_steer)(struct net_device *dev,
990 const struct sk_buff *skb,
991 u16 rxq_index,
992 u32 flow_id);
993 #endif
994 int (*ndo_add_slave)(struct net_device *dev,
995 struct net_device *slave_dev);
996 int (*ndo_del_slave)(struct net_device *dev,
997 struct net_device *slave_dev);
998 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
999 netdev_features_t features);
1000 int (*ndo_set_features)(struct net_device *dev,
1001 netdev_features_t features);
1002 int (*ndo_neigh_construct)(struct neighbour *n);
1003 void (*ndo_neigh_destroy)(struct neighbour *n);
1004
1005 int (*ndo_fdb_add)(struct ndmsg *ndm,
1006 struct nlattr *tb[],
1007 struct net_device *dev,
1008 const unsigned char *addr,
1009 u16 flags);
1010 int (*ndo_fdb_del)(struct ndmsg *ndm,
1011 struct nlattr *tb[],
1012 struct net_device *dev,
1013 const unsigned char *addr);
1014 int (*ndo_fdb_dump)(struct sk_buff *skb,
1015 struct netlink_callback *cb,
1016 struct net_device *dev,
1017 int idx);
1018
1019 int (*ndo_bridge_setlink)(struct net_device *dev,
1020 struct nlmsghdr *nlh);
1021 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1022 u32 pid, u32 seq,
1023 struct net_device *dev,
1024 u32 filter_mask);
1025 int (*ndo_bridge_dellink)(struct net_device *dev,
1026 struct nlmsghdr *nlh);
1027 int (*ndo_change_carrier)(struct net_device *dev,
1028 bool new_carrier);
1029 };
1030
1031 /*
1032 * The DEVICE structure.
1033 * Actually, this whole structure is a big mistake. It mixes I/O
1034 * data with strictly "high-level" data, and it has to know about
1035 * almost every data structure used in the INET module.
1036 *
1037 * FIXME: cleanup struct net_device such that network protocol info
1038 * moves out.
1039 */
1040
1041 struct net_device {
1042
1043 /*
1044 * This is the first field of the "visible" part of this structure
1045 * (i.e. as seen by users in the "Space.c" file). It is the name
1046 * of the interface.
1047 */
1048 char name[IFNAMSIZ];
1049
1050 /* device name hash chain, please keep it close to name[] */
1051 struct hlist_node name_hlist;
1052
1053 /* snmp alias */
1054 char *ifalias;
1055
1056 /*
1057 * I/O specific fields
1058 * FIXME: Merge these and struct ifmap into one
1059 */
1060 unsigned long mem_end; /* shared mem end */
1061 unsigned long mem_start; /* shared mem start */
1062 unsigned long base_addr; /* device I/O address */
1063 unsigned int irq; /* device IRQ number */
1064
1065 /*
1066 * Some hardware also needs these fields, but they are not
1067 * part of the usual set specified in Space.c.
1068 */
1069
1070 unsigned long state;
1071
1072 struct list_head dev_list;
1073 struct list_head napi_list;
1074 struct list_head unreg_list;
1075 struct list_head upper_dev_list; /* List of upper devices */
1076
1077
1078 /* currently active device features */
1079 netdev_features_t features;
1080 /* user-changeable features */
1081 netdev_features_t hw_features;
1082 /* user-requested features */
1083 netdev_features_t wanted_features;
1084 /* mask of features inheritable by VLAN devices */
1085 netdev_features_t vlan_features;
1086 /* mask of features inherited by encapsulating devices
1087 * This field indicates what encapsulation offloads
1088 * the hardware is capable of doing, and drivers will
1089 * need to set them appropriately.
1090 */
1091 netdev_features_t hw_enc_features;
1092
1093 /* Interface index. Unique device identifier */
1094 int ifindex;
1095 int iflink;
1096
1097 struct net_device_stats stats;
1098 atomic_long_t rx_dropped; /* dropped packets by core network
1099 * Do not use this in drivers.
1100 */
1101
1102 #ifdef CONFIG_WIRELESS_EXT
1103 /* List of functions to handle Wireless Extensions (instead of ioctl).
1104 * See <net/iw_handler.h> for details. Jean II */
1105 const struct iw_handler_def * wireless_handlers;
1106 /* Instance data managed by the core of Wireless Extensions. */
1107 struct iw_public_data * wireless_data;
1108 #endif
1109 /* Management operations */
1110 const struct net_device_ops *netdev_ops;
1111 const struct ethtool_ops *ethtool_ops;
1112
1113 /* Hardware header description */
1114 const struct header_ops *header_ops;
1115
1116 unsigned int flags; /* interface flags (a la BSD) */
1117 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1118 * See if.h for definitions. */
1119 unsigned short gflags;
1120 unsigned short padded; /* How much padding added by alloc_netdev() */
1121
1122 unsigned char operstate; /* RFC2863 operstate */
1123 unsigned char link_mode; /* mapping policy to operstate */
1124
1125 unsigned char if_port; /* Selectable AUI, TP,..*/
1126 unsigned char dma; /* DMA channel */
1127
1128 unsigned int mtu; /* interface MTU value */
1129 unsigned short type; /* interface hardware type */
1130 unsigned short hard_header_len; /* hardware hdr length */
1131
1132 /* extra head- and tailroom the hardware may need, but not in all cases
1133 * can this be guaranteed, especially tailroom. Some cases also use
1134 * LL_MAX_HEADER instead to allocate the skb.
1135 */
1136 unsigned short needed_headroom;
1137 unsigned short needed_tailroom;
1138
1139 /* Interface address info. */
1140 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1141 unsigned char addr_assign_type; /* hw address assignment type */
1142 unsigned char addr_len; /* hardware address length */
1143 unsigned char neigh_priv_len;
1144 unsigned short dev_id; /* for shared network cards */
1145
1146 spinlock_t addr_list_lock;
1147 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1148 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1149 struct netdev_hw_addr_list dev_addrs; /* list of device
1150 * hw addresses
1151 */
1152 #ifdef CONFIG_SYSFS
1153 struct kset *queues_kset;
1154 #endif
1155
1156 bool uc_promisc;
1157 unsigned int promiscuity;
1158 unsigned int allmulti;
1159
1160
1161 /* Protocol specific pointers */
1162
1163 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1164 struct vlan_info __rcu *vlan_info; /* VLAN info */
1165 #endif
1166 #if IS_ENABLED(CONFIG_NET_DSA)
1167 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1168 #endif
1169 void *atalk_ptr; /* AppleTalk link */
1170 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1171 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1172 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1173 void *ax25_ptr; /* AX.25 specific data */
1174 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1175 assign before registering */
1176
1177 /*
1178 * Cache lines mostly used on receive path (including eth_type_trans())
1179 */
1180 unsigned long last_rx; /* Time of last Rx
1181 * This should not be set in
1182 * drivers, unless really needed,
1183 * because network stack (bonding)
1184 * use it if/when necessary, to
1185 * avoid dirtying this cache line.
1186 */
1187
1188 /* Interface address info used in eth_type_trans() */
1189 unsigned char *dev_addr; /* hw address, (before bcast
1190 because most packets are
1191 unicast) */
1192
1193
1194 #ifdef CONFIG_RPS
1195 struct netdev_rx_queue *_rx;
1196
1197 /* Number of RX queues allocated at register_netdev() time */
1198 unsigned int num_rx_queues;
1199
1200 /* Number of RX queues currently active in device */
1201 unsigned int real_num_rx_queues;
1202
1203 #endif
1204
1205 rx_handler_func_t __rcu *rx_handler;
1206 void __rcu *rx_handler_data;
1207
1208 struct netdev_queue __rcu *ingress_queue;
1209 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1210
1211
1212 /*
1213 * Cache lines mostly used on transmit path
1214 */
1215 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1216
1217 /* Number of TX queues allocated at alloc_netdev_mq() time */
1218 unsigned int num_tx_queues;
1219
1220 /* Number of TX queues currently active in device */
1221 unsigned int real_num_tx_queues;
1222
1223 /* root qdisc from userspace point of view */
1224 struct Qdisc *qdisc;
1225
1226 unsigned long tx_queue_len; /* Max frames per queue allowed */
1227 spinlock_t tx_global_lock;
1228
1229 #ifdef CONFIG_XPS
1230 struct xps_dev_maps __rcu *xps_maps;
1231 #endif
1232 #ifdef CONFIG_RFS_ACCEL
1233 /* CPU reverse-mapping for RX completion interrupts, indexed
1234 * by RX queue number. Assigned by driver. This must only be
1235 * set if the ndo_rx_flow_steer operation is defined. */
1236 struct cpu_rmap *rx_cpu_rmap;
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 /* Default NAPI poll() weight
1479 * Device drivers are strongly advised to not use bigger value
1480 */
1481 #define NAPI_POLL_WEIGHT 64
1482
1483 /**
1484 * netif_napi_add - initialize a napi context
1485 * @dev: network device
1486 * @napi: napi context
1487 * @poll: polling function
1488 * @weight: default weight
1489 *
1490 * netif_napi_add() must be used to initialize a napi context prior to calling
1491 * *any* of the other napi related functions.
1492 */
1493 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1494 int (*poll)(struct napi_struct *, int), int weight);
1495
1496 /**
1497 * netif_napi_del - remove a napi context
1498 * @napi: napi context
1499 *
1500 * netif_napi_del() removes a napi context from the network device napi list
1501 */
1502 void netif_napi_del(struct napi_struct *napi);
1503
1504 struct napi_gro_cb {
1505 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1506 void *frag0;
1507
1508 /* Length of frag0. */
1509 unsigned int frag0_len;
1510
1511 /* This indicates where we are processing relative to skb->data. */
1512 int data_offset;
1513
1514 /* This is non-zero if the packet cannot be merged with the new skb. */
1515 int flush;
1516
1517 /* Number of segments aggregated. */
1518 u16 count;
1519
1520 /* This is non-zero if the packet may be of the same flow. */
1521 u8 same_flow;
1522
1523 /* Free the skb? */
1524 u8 free;
1525 #define NAPI_GRO_FREE 1
1526 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1527
1528 /* jiffies when first packet was created/queued */
1529 unsigned long age;
1530
1531 /* Used in ipv6_gro_receive() */
1532 int proto;
1533
1534 /* used in skb_gro_receive() slow path */
1535 struct sk_buff *last;
1536 };
1537
1538 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1539
1540 struct packet_type {
1541 __be16 type; /* This is really htons(ether_type). */
1542 struct net_device *dev; /* NULL is wildcarded here */
1543 int (*func) (struct sk_buff *,
1544 struct net_device *,
1545 struct packet_type *,
1546 struct net_device *);
1547 bool (*id_match)(struct packet_type *ptype,
1548 struct sock *sk);
1549 void *af_packet_priv;
1550 struct list_head list;
1551 };
1552
1553 struct offload_callbacks {
1554 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1555 netdev_features_t features);
1556 int (*gso_send_check)(struct sk_buff *skb);
1557 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1558 struct sk_buff *skb);
1559 int (*gro_complete)(struct sk_buff *skb);
1560 };
1561
1562 struct packet_offload {
1563 __be16 type; /* This is really htons(ether_type). */
1564 struct offload_callbacks callbacks;
1565 struct list_head list;
1566 };
1567
1568 #include <linux/notifier.h>
1569
1570 /* netdevice notifier chain. Please remember to update the rtnetlink
1571 * notification exclusion list in rtnetlink_event() when adding new
1572 * types.
1573 */
1574 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1575 #define NETDEV_DOWN 0x0002
1576 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1577 detected a hardware crash and restarted
1578 - we can use this eg to kick tcp sessions
1579 once done */
1580 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1581 #define NETDEV_REGISTER 0x0005
1582 #define NETDEV_UNREGISTER 0x0006
1583 #define NETDEV_CHANGEMTU 0x0007
1584 #define NETDEV_CHANGEADDR 0x0008
1585 #define NETDEV_GOING_DOWN 0x0009
1586 #define NETDEV_CHANGENAME 0x000A
1587 #define NETDEV_FEAT_CHANGE 0x000B
1588 #define NETDEV_BONDING_FAILOVER 0x000C
1589 #define NETDEV_PRE_UP 0x000D
1590 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1591 #define NETDEV_POST_TYPE_CHANGE 0x000F
1592 #define NETDEV_POST_INIT 0x0010
1593 #define NETDEV_UNREGISTER_FINAL 0x0011
1594 #define NETDEV_RELEASE 0x0012
1595 #define NETDEV_NOTIFY_PEERS 0x0013
1596 #define NETDEV_JOIN 0x0014
1597
1598 extern int register_netdevice_notifier(struct notifier_block *nb);
1599 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1600 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1601
1602
1603 extern rwlock_t dev_base_lock; /* Device list lock */
1604
1605 extern seqcount_t devnet_rename_seq; /* Device rename seq */
1606
1607
1608 #define for_each_netdev(net, d) \
1609 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1610 #define for_each_netdev_reverse(net, d) \
1611 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1612 #define for_each_netdev_rcu(net, d) \
1613 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1614 #define for_each_netdev_safe(net, d, n) \
1615 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1616 #define for_each_netdev_continue(net, d) \
1617 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1618 #define for_each_netdev_continue_rcu(net, d) \
1619 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1620 #define for_each_netdev_in_bond_rcu(bond, slave) \
1621 for_each_netdev_rcu(&init_net, slave) \
1622 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1623 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1624
1625 static inline struct net_device *next_net_device(struct net_device *dev)
1626 {
1627 struct list_head *lh;
1628 struct net *net;
1629
1630 net = dev_net(dev);
1631 lh = dev->dev_list.next;
1632 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1633 }
1634
1635 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1636 {
1637 struct list_head *lh;
1638 struct net *net;
1639
1640 net = dev_net(dev);
1641 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1642 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1643 }
1644
1645 static inline struct net_device *first_net_device(struct net *net)
1646 {
1647 return list_empty(&net->dev_base_head) ? NULL :
1648 net_device_entry(net->dev_base_head.next);
1649 }
1650
1651 static inline struct net_device *first_net_device_rcu(struct net *net)
1652 {
1653 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1654
1655 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1656 }
1657
1658 extern int netdev_boot_setup_check(struct net_device *dev);
1659 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1660 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1661 const char *hwaddr);
1662 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1663 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1664 extern void dev_add_pack(struct packet_type *pt);
1665 extern void dev_remove_pack(struct packet_type *pt);
1666 extern void __dev_remove_pack(struct packet_type *pt);
1667 extern void dev_add_offload(struct packet_offload *po);
1668 extern void dev_remove_offload(struct packet_offload *po);
1669 extern void __dev_remove_offload(struct packet_offload *po);
1670
1671 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1672 unsigned short mask);
1673 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1674 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1675 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1676 extern int dev_alloc_name(struct net_device *dev, const char *name);
1677 extern int dev_open(struct net_device *dev);
1678 extern int dev_close(struct net_device *dev);
1679 extern void dev_disable_lro(struct net_device *dev);
1680 extern int dev_loopback_xmit(struct sk_buff *newskb);
1681 extern int dev_queue_xmit(struct sk_buff *skb);
1682 extern int register_netdevice(struct net_device *dev);
1683 extern void unregister_netdevice_queue(struct net_device *dev,
1684 struct list_head *head);
1685 extern void unregister_netdevice_many(struct list_head *head);
1686 static inline void unregister_netdevice(struct net_device *dev)
1687 {
1688 unregister_netdevice_queue(dev, NULL);
1689 }
1690
1691 extern int netdev_refcnt_read(const struct net_device *dev);
1692 extern void free_netdev(struct net_device *dev);
1693 extern void synchronize_net(void);
1694 extern int init_dummy_netdev(struct net_device *dev);
1695
1696 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1697 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1698 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1699 extern int dev_restart(struct net_device *dev);
1700 #ifdef CONFIG_NETPOLL_TRAP
1701 extern int netpoll_trap(void);
1702 #endif
1703 extern int skb_gro_receive(struct sk_buff **head,
1704 struct sk_buff *skb);
1705
1706 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1707 {
1708 return NAPI_GRO_CB(skb)->data_offset;
1709 }
1710
1711 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1712 {
1713 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1714 }
1715
1716 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1717 {
1718 NAPI_GRO_CB(skb)->data_offset += len;
1719 }
1720
1721 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1722 unsigned int offset)
1723 {
1724 return NAPI_GRO_CB(skb)->frag0 + offset;
1725 }
1726
1727 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1728 {
1729 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1730 }
1731
1732 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1733 unsigned int offset)
1734 {
1735 if (!pskb_may_pull(skb, hlen))
1736 return NULL;
1737
1738 NAPI_GRO_CB(skb)->frag0 = NULL;
1739 NAPI_GRO_CB(skb)->frag0_len = 0;
1740 return skb->data + offset;
1741 }
1742
1743 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1744 {
1745 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1746 }
1747
1748 static inline void *skb_gro_network_header(struct sk_buff *skb)
1749 {
1750 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1751 skb_network_offset(skb);
1752 }
1753
1754 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1755 unsigned short type,
1756 const void *daddr, const void *saddr,
1757 unsigned int len)
1758 {
1759 if (!dev->header_ops || !dev->header_ops->create)
1760 return 0;
1761
1762 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1763 }
1764
1765 static inline int dev_parse_header(const struct sk_buff *skb,
1766 unsigned char *haddr)
1767 {
1768 const struct net_device *dev = skb->dev;
1769
1770 if (!dev->header_ops || !dev->header_ops->parse)
1771 return 0;
1772 return dev->header_ops->parse(skb, haddr);
1773 }
1774
1775 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1776 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1777 static inline int unregister_gifconf(unsigned int family)
1778 {
1779 return register_gifconf(family, NULL);
1780 }
1781
1782 /*
1783 * Incoming packets are placed on per-cpu queues
1784 */
1785 struct softnet_data {
1786 struct Qdisc *output_queue;
1787 struct Qdisc **output_queue_tailp;
1788 struct list_head poll_list;
1789 struct sk_buff *completion_queue;
1790 struct sk_buff_head process_queue;
1791
1792 /* stats */
1793 unsigned int processed;
1794 unsigned int time_squeeze;
1795 unsigned int cpu_collision;
1796 unsigned int received_rps;
1797
1798 #ifdef CONFIG_RPS
1799 struct softnet_data *rps_ipi_list;
1800
1801 /* Elements below can be accessed between CPUs for RPS */
1802 struct call_single_data csd ____cacheline_aligned_in_smp;
1803 struct softnet_data *rps_ipi_next;
1804 unsigned int cpu;
1805 unsigned int input_queue_head;
1806 unsigned int input_queue_tail;
1807 #endif
1808 unsigned int dropped;
1809 struct sk_buff_head input_pkt_queue;
1810 struct napi_struct backlog;
1811 };
1812
1813 static inline void input_queue_head_incr(struct softnet_data *sd)
1814 {
1815 #ifdef CONFIG_RPS
1816 sd->input_queue_head++;
1817 #endif
1818 }
1819
1820 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1821 unsigned int *qtail)
1822 {
1823 #ifdef CONFIG_RPS
1824 *qtail = ++sd->input_queue_tail;
1825 #endif
1826 }
1827
1828 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1829
1830 extern void __netif_schedule(struct Qdisc *q);
1831
1832 static inline void netif_schedule_queue(struct netdev_queue *txq)
1833 {
1834 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1835 __netif_schedule(txq->qdisc);
1836 }
1837
1838 static inline void netif_tx_schedule_all(struct net_device *dev)
1839 {
1840 unsigned int i;
1841
1842 for (i = 0; i < dev->num_tx_queues; i++)
1843 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1844 }
1845
1846 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1847 {
1848 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1849 }
1850
1851 /**
1852 * netif_start_queue - allow transmit
1853 * @dev: network device
1854 *
1855 * Allow upper layers to call the device hard_start_xmit routine.
1856 */
1857 static inline void netif_start_queue(struct net_device *dev)
1858 {
1859 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1860 }
1861
1862 static inline void netif_tx_start_all_queues(struct net_device *dev)
1863 {
1864 unsigned int i;
1865
1866 for (i = 0; i < dev->num_tx_queues; i++) {
1867 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1868 netif_tx_start_queue(txq);
1869 }
1870 }
1871
1872 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1873 {
1874 #ifdef CONFIG_NETPOLL_TRAP
1875 if (netpoll_trap()) {
1876 netif_tx_start_queue(dev_queue);
1877 return;
1878 }
1879 #endif
1880 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1881 __netif_schedule(dev_queue->qdisc);
1882 }
1883
1884 /**
1885 * netif_wake_queue - restart transmit
1886 * @dev: network device
1887 *
1888 * Allow upper layers to call the device hard_start_xmit routine.
1889 * Used for flow control when transmit resources are available.
1890 */
1891 static inline void netif_wake_queue(struct net_device *dev)
1892 {
1893 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1894 }
1895
1896 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1897 {
1898 unsigned int i;
1899
1900 for (i = 0; i < dev->num_tx_queues; i++) {
1901 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1902 netif_tx_wake_queue(txq);
1903 }
1904 }
1905
1906 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1907 {
1908 if (WARN_ON(!dev_queue)) {
1909 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1910 return;
1911 }
1912 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1913 }
1914
1915 /**
1916 * netif_stop_queue - stop transmitted packets
1917 * @dev: network device
1918 *
1919 * Stop upper layers calling the device hard_start_xmit routine.
1920 * Used for flow control when transmit resources are unavailable.
1921 */
1922 static inline void netif_stop_queue(struct net_device *dev)
1923 {
1924 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1925 }
1926
1927 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1928 {
1929 unsigned int i;
1930
1931 for (i = 0; i < dev->num_tx_queues; i++) {
1932 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1933 netif_tx_stop_queue(txq);
1934 }
1935 }
1936
1937 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1938 {
1939 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1940 }
1941
1942 /**
1943 * netif_queue_stopped - test if transmit queue is flowblocked
1944 * @dev: network device
1945 *
1946 * Test if transmit queue on device is currently unable to send.
1947 */
1948 static inline bool netif_queue_stopped(const struct net_device *dev)
1949 {
1950 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1951 }
1952
1953 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
1954 {
1955 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1956 }
1957
1958 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
1959 {
1960 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
1961 }
1962
1963 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
1964 unsigned int bytes)
1965 {
1966 #ifdef CONFIG_BQL
1967 dql_queued(&dev_queue->dql, bytes);
1968
1969 if (likely(dql_avail(&dev_queue->dql) >= 0))
1970 return;
1971
1972 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1973
1974 /*
1975 * The XOFF flag must be set before checking the dql_avail below,
1976 * because in netdev_tx_completed_queue we update the dql_completed
1977 * before checking the XOFF flag.
1978 */
1979 smp_mb();
1980
1981 /* check again in case another CPU has just made room avail */
1982 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
1983 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1984 #endif
1985 }
1986
1987 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
1988 {
1989 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
1990 }
1991
1992 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
1993 unsigned int pkts, unsigned int bytes)
1994 {
1995 #ifdef CONFIG_BQL
1996 if (unlikely(!bytes))
1997 return;
1998
1999 dql_completed(&dev_queue->dql, bytes);
2000
2001 /*
2002 * Without the memory barrier there is a small possiblity that
2003 * netdev_tx_sent_queue will miss the update and cause the queue to
2004 * be stopped forever
2005 */
2006 smp_mb();
2007
2008 if (dql_avail(&dev_queue->dql) < 0)
2009 return;
2010
2011 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2012 netif_schedule_queue(dev_queue);
2013 #endif
2014 }
2015
2016 static inline void netdev_completed_queue(struct net_device *dev,
2017 unsigned int pkts, unsigned int bytes)
2018 {
2019 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2020 }
2021
2022 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2023 {
2024 #ifdef CONFIG_BQL
2025 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2026 dql_reset(&q->dql);
2027 #endif
2028 }
2029
2030 static inline void netdev_reset_queue(struct net_device *dev_queue)
2031 {
2032 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2033 }
2034
2035 /**
2036 * netif_running - test if up
2037 * @dev: network device
2038 *
2039 * Test if the device has been brought up.
2040 */
2041 static inline bool netif_running(const struct net_device *dev)
2042 {
2043 return test_bit(__LINK_STATE_START, &dev->state);
2044 }
2045
2046 /*
2047 * Routines to manage the subqueues on a device. We only need start
2048 * stop, and a check if it's stopped. All other device management is
2049 * done at the overall netdevice level.
2050 * Also test the device if we're multiqueue.
2051 */
2052
2053 /**
2054 * netif_start_subqueue - allow sending packets on subqueue
2055 * @dev: network device
2056 * @queue_index: sub queue index
2057 *
2058 * Start individual transmit queue of a device with multiple transmit queues.
2059 */
2060 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2061 {
2062 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2063
2064 netif_tx_start_queue(txq);
2065 }
2066
2067 /**
2068 * netif_stop_subqueue - stop sending packets on subqueue
2069 * @dev: network device
2070 * @queue_index: sub queue index
2071 *
2072 * Stop individual transmit queue of a device with multiple transmit queues.
2073 */
2074 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2075 {
2076 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2077 #ifdef CONFIG_NETPOLL_TRAP
2078 if (netpoll_trap())
2079 return;
2080 #endif
2081 netif_tx_stop_queue(txq);
2082 }
2083
2084 /**
2085 * netif_subqueue_stopped - test status of subqueue
2086 * @dev: network device
2087 * @queue_index: sub queue index
2088 *
2089 * Check individual transmit queue of a device with multiple transmit queues.
2090 */
2091 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2092 u16 queue_index)
2093 {
2094 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2095
2096 return netif_tx_queue_stopped(txq);
2097 }
2098
2099 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2100 struct sk_buff *skb)
2101 {
2102 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2103 }
2104
2105 /**
2106 * netif_wake_subqueue - allow sending packets on subqueue
2107 * @dev: network device
2108 * @queue_index: sub queue index
2109 *
2110 * Resume individual transmit queue of a device with multiple transmit queues.
2111 */
2112 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2113 {
2114 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2115 #ifdef CONFIG_NETPOLL_TRAP
2116 if (netpoll_trap())
2117 return;
2118 #endif
2119 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2120 __netif_schedule(txq->qdisc);
2121 }
2122
2123 #ifdef CONFIG_XPS
2124 extern int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask,
2125 u16 index);
2126 #else
2127 static inline int netif_set_xps_queue(struct net_device *dev,
2128 struct cpumask *mask,
2129 u16 index)
2130 {
2131 return 0;
2132 }
2133 #endif
2134
2135 /*
2136 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2137 * as a distribution range limit for the returned value.
2138 */
2139 static inline u16 skb_tx_hash(const struct net_device *dev,
2140 const struct sk_buff *skb)
2141 {
2142 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2143 }
2144
2145 /**
2146 * netif_is_multiqueue - test if device has multiple transmit queues
2147 * @dev: network device
2148 *
2149 * Check if device has multiple transmit queues
2150 */
2151 static inline bool netif_is_multiqueue(const struct net_device *dev)
2152 {
2153 return dev->num_tx_queues > 1;
2154 }
2155
2156 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2157 unsigned int txq);
2158
2159 #ifdef CONFIG_RPS
2160 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2161 unsigned int rxq);
2162 #else
2163 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2164 unsigned int rxq)
2165 {
2166 return 0;
2167 }
2168 #endif
2169
2170 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2171 const struct net_device *from_dev)
2172 {
2173 int err;
2174
2175 err = netif_set_real_num_tx_queues(to_dev,
2176 from_dev->real_num_tx_queues);
2177 if (err)
2178 return err;
2179 #ifdef CONFIG_RPS
2180 return netif_set_real_num_rx_queues(to_dev,
2181 from_dev->real_num_rx_queues);
2182 #else
2183 return 0;
2184 #endif
2185 }
2186
2187 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2188 extern int netif_get_num_default_rss_queues(void);
2189
2190 /* Use this variant when it is known for sure that it
2191 * is executing from hardware interrupt context or with hardware interrupts
2192 * disabled.
2193 */
2194 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2195
2196 /* Use this variant in places where it could be invoked
2197 * from either hardware interrupt or other context, with hardware interrupts
2198 * either disabled or enabled.
2199 */
2200 extern void dev_kfree_skb_any(struct sk_buff *skb);
2201
2202 extern int netif_rx(struct sk_buff *skb);
2203 extern int netif_rx_ni(struct sk_buff *skb);
2204 extern int netif_receive_skb(struct sk_buff *skb);
2205 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2206 struct sk_buff *skb);
2207 extern void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2208 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2209 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2210
2211 static inline void napi_free_frags(struct napi_struct *napi)
2212 {
2213 kfree_skb(napi->skb);
2214 napi->skb = NULL;
2215 }
2216
2217 extern int netdev_rx_handler_register(struct net_device *dev,
2218 rx_handler_func_t *rx_handler,
2219 void *rx_handler_data);
2220 extern void netdev_rx_handler_unregister(struct net_device *dev);
2221
2222 extern bool dev_valid_name(const char *name);
2223 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2224 extern int dev_ethtool(struct net *net, struct ifreq *);
2225 extern unsigned int dev_get_flags(const struct net_device *);
2226 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2227 extern int dev_change_flags(struct net_device *, unsigned int);
2228 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2229 extern int dev_change_name(struct net_device *, const char *);
2230 extern int dev_set_alias(struct net_device *, const char *, size_t);
2231 extern int dev_change_net_namespace(struct net_device *,
2232 struct net *, const char *);
2233 extern int dev_set_mtu(struct net_device *, int);
2234 extern void dev_set_group(struct net_device *, int);
2235 extern int dev_set_mac_address(struct net_device *,
2236 struct sockaddr *);
2237 extern int dev_change_carrier(struct net_device *,
2238 bool new_carrier);
2239 extern int dev_hard_start_xmit(struct sk_buff *skb,
2240 struct net_device *dev,
2241 struct netdev_queue *txq);
2242 extern int dev_forward_skb(struct net_device *dev,
2243 struct sk_buff *skb);
2244
2245 extern int netdev_budget;
2246
2247 /* Called by rtnetlink.c:rtnl_unlock() */
2248 extern void netdev_run_todo(void);
2249
2250 /**
2251 * dev_put - release reference to device
2252 * @dev: network device
2253 *
2254 * Release reference to device to allow it to be freed.
2255 */
2256 static inline void dev_put(struct net_device *dev)
2257 {
2258 this_cpu_dec(*dev->pcpu_refcnt);
2259 }
2260
2261 /**
2262 * dev_hold - get reference to device
2263 * @dev: network device
2264 *
2265 * Hold reference to device to keep it from being freed.
2266 */
2267 static inline void dev_hold(struct net_device *dev)
2268 {
2269 this_cpu_inc(*dev->pcpu_refcnt);
2270 }
2271
2272 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2273 * and _off may be called from IRQ context, but it is caller
2274 * who is responsible for serialization of these calls.
2275 *
2276 * The name carrier is inappropriate, these functions should really be
2277 * called netif_lowerlayer_*() because they represent the state of any
2278 * kind of lower layer not just hardware media.
2279 */
2280
2281 extern void linkwatch_init_dev(struct net_device *dev);
2282 extern void linkwatch_fire_event(struct net_device *dev);
2283 extern void linkwatch_forget_dev(struct net_device *dev);
2284
2285 /**
2286 * netif_carrier_ok - test if carrier present
2287 * @dev: network device
2288 *
2289 * Check if carrier is present on device
2290 */
2291 static inline bool netif_carrier_ok(const struct net_device *dev)
2292 {
2293 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2294 }
2295
2296 extern unsigned long dev_trans_start(struct net_device *dev);
2297
2298 extern void __netdev_watchdog_up(struct net_device *dev);
2299
2300 extern void netif_carrier_on(struct net_device *dev);
2301
2302 extern void netif_carrier_off(struct net_device *dev);
2303
2304 /**
2305 * netif_dormant_on - mark device as dormant.
2306 * @dev: network device
2307 *
2308 * Mark device as dormant (as per RFC2863).
2309 *
2310 * The dormant state indicates that the relevant interface is not
2311 * actually in a condition to pass packets (i.e., it is not 'up') but is
2312 * in a "pending" state, waiting for some external event. For "on-
2313 * demand" interfaces, this new state identifies the situation where the
2314 * interface is waiting for events to place it in the up state.
2315 *
2316 */
2317 static inline void netif_dormant_on(struct net_device *dev)
2318 {
2319 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2320 linkwatch_fire_event(dev);
2321 }
2322
2323 /**
2324 * netif_dormant_off - set device as not dormant.
2325 * @dev: network device
2326 *
2327 * Device is not in dormant state.
2328 */
2329 static inline void netif_dormant_off(struct net_device *dev)
2330 {
2331 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2332 linkwatch_fire_event(dev);
2333 }
2334
2335 /**
2336 * netif_dormant - test if carrier present
2337 * @dev: network device
2338 *
2339 * Check if carrier is present on device
2340 */
2341 static inline bool netif_dormant(const struct net_device *dev)
2342 {
2343 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2344 }
2345
2346
2347 /**
2348 * netif_oper_up - test if device is operational
2349 * @dev: network device
2350 *
2351 * Check if carrier is operational
2352 */
2353 static inline bool netif_oper_up(const struct net_device *dev)
2354 {
2355 return (dev->operstate == IF_OPER_UP ||
2356 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2357 }
2358
2359 /**
2360 * netif_device_present - is device available or removed
2361 * @dev: network device
2362 *
2363 * Check if device has not been removed from system.
2364 */
2365 static inline bool netif_device_present(struct net_device *dev)
2366 {
2367 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2368 }
2369
2370 extern void netif_device_detach(struct net_device *dev);
2371
2372 extern void netif_device_attach(struct net_device *dev);
2373
2374 /*
2375 * Network interface message level settings
2376 */
2377
2378 enum {
2379 NETIF_MSG_DRV = 0x0001,
2380 NETIF_MSG_PROBE = 0x0002,
2381 NETIF_MSG_LINK = 0x0004,
2382 NETIF_MSG_TIMER = 0x0008,
2383 NETIF_MSG_IFDOWN = 0x0010,
2384 NETIF_MSG_IFUP = 0x0020,
2385 NETIF_MSG_RX_ERR = 0x0040,
2386 NETIF_MSG_TX_ERR = 0x0080,
2387 NETIF_MSG_TX_QUEUED = 0x0100,
2388 NETIF_MSG_INTR = 0x0200,
2389 NETIF_MSG_TX_DONE = 0x0400,
2390 NETIF_MSG_RX_STATUS = 0x0800,
2391 NETIF_MSG_PKTDATA = 0x1000,
2392 NETIF_MSG_HW = 0x2000,
2393 NETIF_MSG_WOL = 0x4000,
2394 };
2395
2396 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2397 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2398 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2399 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2400 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2401 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2402 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2403 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2404 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2405 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2406 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2407 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2408 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2409 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2410 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2411
2412 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2413 {
2414 /* use default */
2415 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2416 return default_msg_enable_bits;
2417 if (debug_value == 0) /* no output */
2418 return 0;
2419 /* set low N bits */
2420 return (1 << debug_value) - 1;
2421 }
2422
2423 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2424 {
2425 spin_lock(&txq->_xmit_lock);
2426 txq->xmit_lock_owner = cpu;
2427 }
2428
2429 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2430 {
2431 spin_lock_bh(&txq->_xmit_lock);
2432 txq->xmit_lock_owner = smp_processor_id();
2433 }
2434
2435 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2436 {
2437 bool ok = spin_trylock(&txq->_xmit_lock);
2438 if (likely(ok))
2439 txq->xmit_lock_owner = smp_processor_id();
2440 return ok;
2441 }
2442
2443 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2444 {
2445 txq->xmit_lock_owner = -1;
2446 spin_unlock(&txq->_xmit_lock);
2447 }
2448
2449 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2450 {
2451 txq->xmit_lock_owner = -1;
2452 spin_unlock_bh(&txq->_xmit_lock);
2453 }
2454
2455 static inline void txq_trans_update(struct netdev_queue *txq)
2456 {
2457 if (txq->xmit_lock_owner != -1)
2458 txq->trans_start = jiffies;
2459 }
2460
2461 /**
2462 * netif_tx_lock - grab network device transmit lock
2463 * @dev: network device
2464 *
2465 * Get network device transmit lock
2466 */
2467 static inline void netif_tx_lock(struct net_device *dev)
2468 {
2469 unsigned int i;
2470 int cpu;
2471
2472 spin_lock(&dev->tx_global_lock);
2473 cpu = smp_processor_id();
2474 for (i = 0; i < dev->num_tx_queues; i++) {
2475 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2476
2477 /* We are the only thread of execution doing a
2478 * freeze, but we have to grab the _xmit_lock in
2479 * order to synchronize with threads which are in
2480 * the ->hard_start_xmit() handler and already
2481 * checked the frozen bit.
2482 */
2483 __netif_tx_lock(txq, cpu);
2484 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2485 __netif_tx_unlock(txq);
2486 }
2487 }
2488
2489 static inline void netif_tx_lock_bh(struct net_device *dev)
2490 {
2491 local_bh_disable();
2492 netif_tx_lock(dev);
2493 }
2494
2495 static inline void netif_tx_unlock(struct net_device *dev)
2496 {
2497 unsigned int i;
2498
2499 for (i = 0; i < dev->num_tx_queues; i++) {
2500 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2501
2502 /* No need to grab the _xmit_lock here. If the
2503 * queue is not stopped for another reason, we
2504 * force a schedule.
2505 */
2506 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2507 netif_schedule_queue(txq);
2508 }
2509 spin_unlock(&dev->tx_global_lock);
2510 }
2511
2512 static inline void netif_tx_unlock_bh(struct net_device *dev)
2513 {
2514 netif_tx_unlock(dev);
2515 local_bh_enable();
2516 }
2517
2518 #define HARD_TX_LOCK(dev, txq, cpu) { \
2519 if ((dev->features & NETIF_F_LLTX) == 0) { \
2520 __netif_tx_lock(txq, cpu); \
2521 } \
2522 }
2523
2524 #define HARD_TX_UNLOCK(dev, txq) { \
2525 if ((dev->features & NETIF_F_LLTX) == 0) { \
2526 __netif_tx_unlock(txq); \
2527 } \
2528 }
2529
2530 static inline void netif_tx_disable(struct net_device *dev)
2531 {
2532 unsigned int i;
2533 int cpu;
2534
2535 local_bh_disable();
2536 cpu = smp_processor_id();
2537 for (i = 0; i < dev->num_tx_queues; i++) {
2538 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2539
2540 __netif_tx_lock(txq, cpu);
2541 netif_tx_stop_queue(txq);
2542 __netif_tx_unlock(txq);
2543 }
2544 local_bh_enable();
2545 }
2546
2547 static inline void netif_addr_lock(struct net_device *dev)
2548 {
2549 spin_lock(&dev->addr_list_lock);
2550 }
2551
2552 static inline void netif_addr_lock_nested(struct net_device *dev)
2553 {
2554 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2555 }
2556
2557 static inline void netif_addr_lock_bh(struct net_device *dev)
2558 {
2559 spin_lock_bh(&dev->addr_list_lock);
2560 }
2561
2562 static inline void netif_addr_unlock(struct net_device *dev)
2563 {
2564 spin_unlock(&dev->addr_list_lock);
2565 }
2566
2567 static inline void netif_addr_unlock_bh(struct net_device *dev)
2568 {
2569 spin_unlock_bh(&dev->addr_list_lock);
2570 }
2571
2572 /*
2573 * dev_addrs walker. Should be used only for read access. Call with
2574 * rcu_read_lock held.
2575 */
2576 #define for_each_dev_addr(dev, ha) \
2577 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2578
2579 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2580
2581 extern void ether_setup(struct net_device *dev);
2582
2583 /* Support for loadable net-drivers */
2584 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2585 void (*setup)(struct net_device *),
2586 unsigned int txqs, unsigned int rxqs);
2587 #define alloc_netdev(sizeof_priv, name, setup) \
2588 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2589
2590 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2591 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2592
2593 extern int register_netdev(struct net_device *dev);
2594 extern void unregister_netdev(struct net_device *dev);
2595
2596 /* General hardware address lists handling functions */
2597 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2598 struct netdev_hw_addr_list *from_list,
2599 int addr_len, unsigned char addr_type);
2600 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2601 struct netdev_hw_addr_list *from_list,
2602 int addr_len, unsigned char addr_type);
2603 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2604 struct netdev_hw_addr_list *from_list,
2605 int addr_len);
2606 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2607 struct netdev_hw_addr_list *from_list,
2608 int addr_len);
2609 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2610 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2611
2612 /* Functions used for device addresses handling */
2613 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2614 unsigned char addr_type);
2615 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2616 unsigned char addr_type);
2617 extern int dev_addr_add_multiple(struct net_device *to_dev,
2618 struct net_device *from_dev,
2619 unsigned char addr_type);
2620 extern int dev_addr_del_multiple(struct net_device *to_dev,
2621 struct net_device *from_dev,
2622 unsigned char addr_type);
2623 extern void dev_addr_flush(struct net_device *dev);
2624 extern int dev_addr_init(struct net_device *dev);
2625
2626 /* Functions used for unicast addresses handling */
2627 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2628 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2629 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2630 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2631 extern int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2632 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2633 extern void dev_uc_flush(struct net_device *dev);
2634 extern void dev_uc_init(struct net_device *dev);
2635
2636 /* Functions used for multicast addresses handling */
2637 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2638 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2639 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2640 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2641 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2642 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2643 extern int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2644 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2645 extern void dev_mc_flush(struct net_device *dev);
2646 extern void dev_mc_init(struct net_device *dev);
2647
2648 /* Functions used for secondary unicast and multicast support */
2649 extern void dev_set_rx_mode(struct net_device *dev);
2650 extern void __dev_set_rx_mode(struct net_device *dev);
2651 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2652 extern int dev_set_allmulti(struct net_device *dev, int inc);
2653 extern void netdev_state_change(struct net_device *dev);
2654 extern void netdev_notify_peers(struct net_device *dev);
2655 extern void netdev_features_change(struct net_device *dev);
2656 /* Load a device via the kmod */
2657 extern void dev_load(struct net *net, const char *name);
2658 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2659 struct rtnl_link_stats64 *storage);
2660 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2661 const struct net_device_stats *netdev_stats);
2662
2663 extern int netdev_max_backlog;
2664 extern int netdev_tstamp_prequeue;
2665 extern int weight_p;
2666 extern int bpf_jit_enable;
2667
2668 extern bool netdev_has_upper_dev(struct net_device *dev,
2669 struct net_device *upper_dev);
2670 extern bool netdev_has_any_upper_dev(struct net_device *dev);
2671 extern struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2672 extern struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2673 extern int netdev_upper_dev_link(struct net_device *dev,
2674 struct net_device *upper_dev);
2675 extern int netdev_master_upper_dev_link(struct net_device *dev,
2676 struct net_device *upper_dev);
2677 extern void netdev_upper_dev_unlink(struct net_device *dev,
2678 struct net_device *upper_dev);
2679 extern int skb_checksum_help(struct sk_buff *skb);
2680 extern struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2681 netdev_features_t features, bool tx_path);
2682 extern struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2683 netdev_features_t features);
2684
2685 static inline
2686 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2687 {
2688 return __skb_gso_segment(skb, features, true);
2689 }
2690 __be16 skb_network_protocol(struct sk_buff *skb);
2691
2692 static inline bool can_checksum_protocol(netdev_features_t features,
2693 __be16 protocol)
2694 {
2695 return ((features & NETIF_F_GEN_CSUM) ||
2696 ((features & NETIF_F_V4_CSUM) &&
2697 protocol == htons(ETH_P_IP)) ||
2698 ((features & NETIF_F_V6_CSUM) &&
2699 protocol == htons(ETH_P_IPV6)) ||
2700 ((features & NETIF_F_FCOE_CRC) &&
2701 protocol == htons(ETH_P_FCOE)));
2702 }
2703
2704 #ifdef CONFIG_BUG
2705 extern void netdev_rx_csum_fault(struct net_device *dev);
2706 #else
2707 static inline void netdev_rx_csum_fault(struct net_device *dev)
2708 {
2709 }
2710 #endif
2711 /* rx skb timestamps */
2712 extern void net_enable_timestamp(void);
2713 extern void net_disable_timestamp(void);
2714
2715 #ifdef CONFIG_PROC_FS
2716 extern int __init dev_proc_init(void);
2717 #else
2718 #define dev_proc_init() 0
2719 #endif
2720
2721 extern int netdev_class_create_file(struct class_attribute *class_attr);
2722 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2723
2724 extern struct kobj_ns_type_operations net_ns_type_operations;
2725
2726 extern const char *netdev_drivername(const struct net_device *dev);
2727
2728 extern void linkwatch_run_queue(void);
2729
2730 static inline netdev_features_t netdev_get_wanted_features(
2731 struct net_device *dev)
2732 {
2733 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2734 }
2735 netdev_features_t netdev_increment_features(netdev_features_t all,
2736 netdev_features_t one, netdev_features_t mask);
2737 int __netdev_update_features(struct net_device *dev);
2738 void netdev_update_features(struct net_device *dev);
2739 void netdev_change_features(struct net_device *dev);
2740
2741 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2742 struct net_device *dev);
2743
2744 netdev_features_t netif_skb_features(struct sk_buff *skb);
2745
2746 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2747 {
2748 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2749
2750 /* check flags correspondence */
2751 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2752 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2753 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2754 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2755 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2756 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2757
2758 return (features & feature) == feature;
2759 }
2760
2761 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2762 {
2763 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2764 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2765 }
2766
2767 static inline bool netif_needs_gso(struct sk_buff *skb,
2768 netdev_features_t features)
2769 {
2770 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2771 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2772 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2773 }
2774
2775 static inline void netif_set_gso_max_size(struct net_device *dev,
2776 unsigned int size)
2777 {
2778 dev->gso_max_size = size;
2779 }
2780
2781 static inline bool netif_is_bond_master(struct net_device *dev)
2782 {
2783 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
2784 }
2785
2786 static inline bool netif_is_bond_slave(struct net_device *dev)
2787 {
2788 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2789 }
2790
2791 static inline bool netif_supports_nofcs(struct net_device *dev)
2792 {
2793 return dev->priv_flags & IFF_SUPP_NOFCS;
2794 }
2795
2796 extern struct pernet_operations __net_initdata loopback_net_ops;
2797
2798 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2799
2800 /* netdev_printk helpers, similar to dev_printk */
2801
2802 static inline const char *netdev_name(const struct net_device *dev)
2803 {
2804 if (dev->reg_state != NETREG_REGISTERED)
2805 return "(unregistered net_device)";
2806 return dev->name;
2807 }
2808
2809 extern __printf(3, 4)
2810 int netdev_printk(const char *level, const struct net_device *dev,
2811 const char *format, ...);
2812 extern __printf(2, 3)
2813 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2814 extern __printf(2, 3)
2815 int netdev_alert(const struct net_device *dev, const char *format, ...);
2816 extern __printf(2, 3)
2817 int netdev_crit(const struct net_device *dev, const char *format, ...);
2818 extern __printf(2, 3)
2819 int netdev_err(const struct net_device *dev, const char *format, ...);
2820 extern __printf(2, 3)
2821 int netdev_warn(const struct net_device *dev, const char *format, ...);
2822 extern __printf(2, 3)
2823 int netdev_notice(const struct net_device *dev, const char *format, ...);
2824 extern __printf(2, 3)
2825 int netdev_info(const struct net_device *dev, const char *format, ...);
2826
2827 #define MODULE_ALIAS_NETDEV(device) \
2828 MODULE_ALIAS("netdev-" device)
2829
2830 #if defined(CONFIG_DYNAMIC_DEBUG)
2831 #define netdev_dbg(__dev, format, args...) \
2832 do { \
2833 dynamic_netdev_dbg(__dev, format, ##args); \
2834 } while (0)
2835 #elif defined(DEBUG)
2836 #define netdev_dbg(__dev, format, args...) \
2837 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2838 #else
2839 #define netdev_dbg(__dev, format, args...) \
2840 ({ \
2841 if (0) \
2842 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2843 0; \
2844 })
2845 #endif
2846
2847 #if defined(VERBOSE_DEBUG)
2848 #define netdev_vdbg netdev_dbg
2849 #else
2850
2851 #define netdev_vdbg(dev, format, args...) \
2852 ({ \
2853 if (0) \
2854 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2855 0; \
2856 })
2857 #endif
2858
2859 /*
2860 * netdev_WARN() acts like dev_printk(), but with the key difference
2861 * of using a WARN/WARN_ON to get the message out, including the
2862 * file/line information and a backtrace.
2863 */
2864 #define netdev_WARN(dev, format, args...) \
2865 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2866
2867 /* netif printk helpers, similar to netdev_printk */
2868
2869 #define netif_printk(priv, type, level, dev, fmt, args...) \
2870 do { \
2871 if (netif_msg_##type(priv)) \
2872 netdev_printk(level, (dev), fmt, ##args); \
2873 } while (0)
2874
2875 #define netif_level(level, priv, type, dev, fmt, args...) \
2876 do { \
2877 if (netif_msg_##type(priv)) \
2878 netdev_##level(dev, fmt, ##args); \
2879 } while (0)
2880
2881 #define netif_emerg(priv, type, dev, fmt, args...) \
2882 netif_level(emerg, priv, type, dev, fmt, ##args)
2883 #define netif_alert(priv, type, dev, fmt, args...) \
2884 netif_level(alert, priv, type, dev, fmt, ##args)
2885 #define netif_crit(priv, type, dev, fmt, args...) \
2886 netif_level(crit, priv, type, dev, fmt, ##args)
2887 #define netif_err(priv, type, dev, fmt, args...) \
2888 netif_level(err, priv, type, dev, fmt, ##args)
2889 #define netif_warn(priv, type, dev, fmt, args...) \
2890 netif_level(warn, priv, type, dev, fmt, ##args)
2891 #define netif_notice(priv, type, dev, fmt, args...) \
2892 netif_level(notice, priv, type, dev, fmt, ##args)
2893 #define netif_info(priv, type, dev, fmt, args...) \
2894 netif_level(info, priv, type, dev, fmt, ##args)
2895
2896 #if defined(CONFIG_DYNAMIC_DEBUG)
2897 #define netif_dbg(priv, type, netdev, format, args...) \
2898 do { \
2899 if (netif_msg_##type(priv)) \
2900 dynamic_netdev_dbg(netdev, format, ##args); \
2901 } while (0)
2902 #elif defined(DEBUG)
2903 #define netif_dbg(priv, type, dev, format, args...) \
2904 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2905 #else
2906 #define netif_dbg(priv, type, dev, format, args...) \
2907 ({ \
2908 if (0) \
2909 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2910 0; \
2911 })
2912 #endif
2913
2914 #if defined(VERBOSE_DEBUG)
2915 #define netif_vdbg netif_dbg
2916 #else
2917 #define netif_vdbg(priv, type, dev, format, args...) \
2918 ({ \
2919 if (0) \
2920 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2921 0; \
2922 })
2923 #endif
2924
2925 /*
2926 * The list of packet types we will receive (as opposed to discard)
2927 * and the routines to invoke.
2928 *
2929 * Why 16. Because with 16 the only overlap we get on a hash of the
2930 * low nibble of the protocol value is RARP/SNAP/X.25.
2931 *
2932 * NOTE: That is no longer true with the addition of VLAN tags. Not
2933 * sure which should go first, but I bet it won't make much
2934 * difference if we are running VLANs. The good news is that
2935 * this protocol won't be in the list unless compiled in, so
2936 * the average user (w/out VLANs) will not be adversely affected.
2937 * --BLG
2938 *
2939 * 0800 IP
2940 * 8100 802.1Q VLAN
2941 * 0001 802.3
2942 * 0002 AX.25
2943 * 0004 802.2
2944 * 8035 RARP
2945 * 0005 SNAP
2946 * 0805 X.25
2947 * 0806 ARP
2948 * 8137 IPX
2949 * 0009 Localtalk
2950 * 86DD IPv6
2951 */
2952 #define PTYPE_HASH_SIZE (16)
2953 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
2954
2955 #endif /* _LINUX_NETDEVICE_H */
This page took 0.096344 seconds and 5 git commands to generate.