]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame_incremental - include/linux/netdevice.h
bcm63xx_enet: add support Broadcom BCM6345 Ethernet
[mirror_ubuntu-zesty-kernel.git] / include / linux / netdevice.h
... / ...
CommitLineData
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
54struct netpoll_info;
55struct device;
56struct phy_device;
57/* 802.11 specific */
58struct wireless_dev;
59 /* source back-compat hooks */
60#define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
62
63extern 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
110enum 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};
116typedef 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 */
122static 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
163struct 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>
195extern struct static_key rps_needed;
196#endif
197
198struct neighbour;
199struct neigh_parms;
200struct sk_buff;
201
202struct 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
218struct 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
238struct 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
265struct 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
282enum 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 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
301extern int __init netdev_boot_setup(char *str);
302
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct 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 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
329};
330
331enum {
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash */
336};
337
338enum gro_result {
339 GRO_MERGED,
340 GRO_MERGED_FREE,
341 GRO_HELD,
342 GRO_NORMAL,
343 GRO_DROP,
344};
345typedef enum gro_result gro_result_t;
346
347/*
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * further.
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
355 *
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
358 *
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
361 *
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
365 *
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * do with the skb.
368 *
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
372 *
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
376 *
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
380 *
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
383 *
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
386 */
387
388enum rx_handler_result {
389 RX_HANDLER_CONSUMED,
390 RX_HANDLER_ANOTHER,
391 RX_HANDLER_EXACT,
392 RX_HANDLER_PASS,
393};
394typedef enum rx_handler_result rx_handler_result_t;
395typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
396
397extern void __napi_schedule(struct napi_struct *n);
398
399static inline bool napi_disable_pending(struct napi_struct *n)
400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
404/**
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
407 *
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
412 */
413static inline bool napi_schedule_prep(struct napi_struct *n)
414{
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
417}
418
419/**
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
432/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
433static inline bool napi_reschedule(struct napi_struct *napi)
434{
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
437 return true;
438 }
439 return false;
440}
441
442/**
443 * napi_complete - NAPI processing complete
444 * @n: napi context
445 *
446 * Mark NAPI processing as complete.
447 */
448extern void __napi_complete(struct napi_struct *n);
449extern void napi_complete(struct napi_struct *n);
450
451/**
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
454 *
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
457 */
458extern struct napi_struct *napi_by_id(unsigned int napi_id);
459
460/**
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
463 *
464 * generate a new napi_id and store a @napi under it in napi_hash
465 */
466extern void napi_hash_add(struct napi_struct *napi);
467
468/**
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
471 *
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
474 */
475extern void napi_hash_del(struct napi_struct *napi);
476
477/**
478 * napi_disable - prevent NAPI from scheduling
479 * @n: napi context
480 *
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
483 */
484static inline void napi_disable(struct napi_struct *n)
485{
486 set_bit(NAPI_STATE_DISABLE, &n->state);
487 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
488 msleep(1);
489 clear_bit(NAPI_STATE_DISABLE, &n->state);
490}
491
492/**
493 * napi_enable - enable NAPI scheduling
494 * @n: napi context
495 *
496 * Resume NAPI from being scheduled on this context.
497 * Must be paired with napi_disable.
498 */
499static inline void napi_enable(struct napi_struct *n)
500{
501 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
502 smp_mb__before_clear_bit();
503 clear_bit(NAPI_STATE_SCHED, &n->state);
504}
505
506#ifdef CONFIG_SMP
507/**
508 * napi_synchronize - wait until NAPI is not running
509 * @n: napi context
510 *
511 * Wait until NAPI is done being scheduled on this context.
512 * Waits till any outstanding processing completes but
513 * does not disable future activations.
514 */
515static inline void napi_synchronize(const struct napi_struct *n)
516{
517 while (test_bit(NAPI_STATE_SCHED, &n->state))
518 msleep(1);
519}
520#else
521# define napi_synchronize(n) barrier()
522#endif
523
524enum netdev_queue_state_t {
525 __QUEUE_STATE_DRV_XOFF,
526 __QUEUE_STATE_STACK_XOFF,
527 __QUEUE_STATE_FROZEN,
528#define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
529 (1 << __QUEUE_STATE_STACK_XOFF))
530#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
531 (1 << __QUEUE_STATE_FROZEN))
532};
533/*
534 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
535 * netif_tx_* functions below are used to manipulate this flag. The
536 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
537 * queue independently. The netif_xmit_*stopped functions below are called
538 * to check if the queue has been stopped by the driver or stack (either
539 * of the XOFF bits are set in the state). Drivers should not need to call
540 * netif_xmit*stopped functions, they should only be using netif_tx_*.
541 */
542
543struct netdev_queue {
544/*
545 * read mostly part
546 */
547 struct net_device *dev;
548 struct Qdisc *qdisc;
549 struct Qdisc *qdisc_sleeping;
550#ifdef CONFIG_SYSFS
551 struct kobject kobj;
552#endif
553#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
554 int numa_node;
555#endif
556/*
557 * write mostly part
558 */
559 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
560 int xmit_lock_owner;
561 /*
562 * please use this field instead of dev->trans_start
563 */
564 unsigned long trans_start;
565
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
571
572 unsigned long state;
573
574#ifdef CONFIG_BQL
575 struct dql dql;
576#endif
577} ____cacheline_aligned_in_smp;
578
579static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
580{
581#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
582 return q->numa_node;
583#else
584 return NUMA_NO_NODE;
585#endif
586}
587
588static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 q->numa_node = node;
592#endif
593}
594
595#ifdef CONFIG_RPS
596/*
597 * This structure holds an RPS map which can be of variable length. The
598 * map is an array of CPUs.
599 */
600struct rps_map {
601 unsigned int len;
602 struct rcu_head rcu;
603 u16 cpus[0];
604};
605#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
606
607/*
608 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
609 * tail pointer for that CPU's input queue at the time of last enqueue, and
610 * a hardware filter index.
611 */
612struct rps_dev_flow {
613 u16 cpu;
614 u16 filter;
615 unsigned int last_qtail;
616};
617#define RPS_NO_FILTER 0xffff
618
619/*
620 * The rps_dev_flow_table structure contains a table of flow mappings.
621 */
622struct rps_dev_flow_table {
623 unsigned int mask;
624 struct rcu_head rcu;
625 struct rps_dev_flow flows[0];
626};
627#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
628 ((_num) * sizeof(struct rps_dev_flow)))
629
630/*
631 * The rps_sock_flow_table contains mappings of flows to the last CPU
632 * on which they were processed by the application (set in recvmsg).
633 */
634struct rps_sock_flow_table {
635 unsigned int mask;
636 u16 ents[0];
637};
638#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
639 ((_num) * sizeof(u16)))
640
641#define RPS_NO_CPU 0xffff
642
643static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
644 u32 hash)
645{
646 if (table && hash) {
647 unsigned int cpu, index = hash & table->mask;
648
649 /* We only give a hint, preemption can change cpu under us */
650 cpu = raw_smp_processor_id();
651
652 if (table->ents[index] != cpu)
653 table->ents[index] = cpu;
654 }
655}
656
657static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
658 u32 hash)
659{
660 if (table && hash)
661 table->ents[hash & table->mask] = RPS_NO_CPU;
662}
663
664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
666#ifdef CONFIG_RFS_ACCEL
667extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
668 u32 flow_id, u16 filter_id);
669#endif
670
671/* This structure contains an instance of an RX queue. */
672struct netdev_rx_queue {
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
675 struct kobject kobj;
676 struct net_device *dev;
677} ____cacheline_aligned_in_smp;
678#endif /* CONFIG_RPS */
679
680#ifdef CONFIG_XPS
681/*
682 * This structure holds an XPS map which can be of variable length. The
683 * map is an array of queues.
684 */
685struct xps_map {
686 unsigned int len;
687 unsigned int alloc_len;
688 struct rcu_head rcu;
689 u16 queues[0];
690};
691#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
692#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
693 / sizeof(u16))
694
695/*
696 * This structure holds all XPS maps for device. Maps are indexed by CPU.
697 */
698struct xps_dev_maps {
699 struct rcu_head rcu;
700 struct xps_map __rcu *cpu_map[0];
701};
702#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
703 (nr_cpu_ids * sizeof(struct xps_map *)))
704#endif /* CONFIG_XPS */
705
706#define TC_MAX_QUEUE 16
707#define TC_BITMASK 15
708/* HW offloaded queuing disciplines txq count and offset maps */
709struct netdev_tc_txq {
710 u16 count;
711 u16 offset;
712};
713
714#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
715/*
716 * This structure is to hold information about the device
717 * configured to run FCoE protocol stack.
718 */
719struct netdev_fcoe_hbainfo {
720 char manufacturer[64];
721 char serial_number[64];
722 char hardware_version[64];
723 char driver_version[64];
724 char optionrom_version[64];
725 char firmware_version[64];
726 char model[256];
727 char model_description[256];
728};
729#endif
730
731/*
732 * This structure defines the management hooks for network devices.
733 * The following hooks can be defined; unless noted otherwise, they are
734 * optional and can be filled with a null pointer.
735 *
736 * int (*ndo_init)(struct net_device *dev);
737 * This function is called once when network device is registered.
738 * The network device can use this to any late stage initializaton
739 * or semantic validattion. It can fail with an error code which will
740 * be propogated back to register_netdev
741 *
742 * void (*ndo_uninit)(struct net_device *dev);
743 * This function is called when device is unregistered or when registration
744 * fails. It is not called if init fails.
745 *
746 * int (*ndo_open)(struct net_device *dev);
747 * This function is called when network device transistions to the up
748 * state.
749 *
750 * int (*ndo_stop)(struct net_device *dev);
751 * This function is called when network device transistions to the down
752 * state.
753 *
754 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
755 * struct net_device *dev);
756 * Called when a packet needs to be transmitted.
757 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
758 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
759 * Required can not be NULL.
760 *
761 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
762 * Called to decide which queue to when device supports multiple
763 * transmit queues.
764 *
765 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
766 * This function is called to allow device receiver to make
767 * changes to configuration when multicast or promiscious is enabled.
768 *
769 * void (*ndo_set_rx_mode)(struct net_device *dev);
770 * This function is called device changes address list filtering.
771 * If driver handles unicast address filtering, it should set
772 * IFF_UNICAST_FLT to its priv_flags.
773 *
774 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
775 * This function is called when the Media Access Control address
776 * needs to be changed. If this interface is not defined, the
777 * mac address can not be changed.
778 *
779 * int (*ndo_validate_addr)(struct net_device *dev);
780 * Test if Media Access Control address is valid for the device.
781 *
782 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
783 * Called when a user request an ioctl which can't be handled by
784 * the generic interface code. If not defined ioctl's return
785 * not supported error code.
786 *
787 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
788 * Used to set network devices bus interface parameters. This interface
789 * is retained for legacy reason, new devices should use the bus
790 * interface (PCI) for low level management.
791 *
792 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
793 * Called when a user wants to change the Maximum Transfer Unit
794 * of a device. If not defined, any request to change MTU will
795 * will return an error.
796 *
797 * void (*ndo_tx_timeout)(struct net_device *dev);
798 * Callback uses when the transmitter has not made any progress
799 * for dev->watchdog ticks.
800 *
801 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
802 * struct rtnl_link_stats64 *storage);
803 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
804 * Called when a user wants to get the network device usage
805 * statistics. Drivers must do one of the following:
806 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
807 * rtnl_link_stats64 structure passed by the caller.
808 * 2. Define @ndo_get_stats to update a net_device_stats structure
809 * (which should normally be dev->stats) and return a pointer to
810 * it. The structure may be changed asynchronously only if each
811 * field is written atomically.
812 * 3. Update dev->stats asynchronously and atomically, and define
813 * neither operation.
814 *
815 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
816 * If device support VLAN filtering this function is called when a
817 * VLAN id is registered.
818 *
819 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
820 * If device support VLAN filtering this function is called when a
821 * VLAN id is unregistered.
822 *
823 * void (*ndo_poll_controller)(struct net_device *dev);
824 *
825 * SR-IOV management functions.
826 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
827 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
828 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
829 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
830 * int (*ndo_get_vf_config)(struct net_device *dev,
831 * int vf, struct ifla_vf_info *ivf);
832 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
833 * struct nlattr *port[]);
834 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
835 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
836 * Called to setup 'tc' number of traffic classes in the net device. This
837 * is always called from the stack with the rtnl lock held and netif tx
838 * queues stopped. This allows the netdevice to perform queue management
839 * safely.
840 *
841 * Fiber Channel over Ethernet (FCoE) offload functions.
842 * int (*ndo_fcoe_enable)(struct net_device *dev);
843 * Called when the FCoE protocol stack wants to start using LLD for FCoE
844 * so the underlying device can perform whatever needed configuration or
845 * initialization to support acceleration of FCoE traffic.
846 *
847 * int (*ndo_fcoe_disable)(struct net_device *dev);
848 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
849 * so the underlying device can perform whatever needed clean-ups to
850 * stop supporting acceleration of FCoE traffic.
851 *
852 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
853 * struct scatterlist *sgl, unsigned int sgc);
854 * Called when the FCoE Initiator wants to initialize an I/O that
855 * is a possible candidate for Direct Data Placement (DDP). The LLD can
856 * perform necessary setup and returns 1 to indicate the device is set up
857 * successfully to perform DDP on this I/O, otherwise this returns 0.
858 *
859 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
860 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
861 * indicated by the FC exchange id 'xid', so the underlying device can
862 * clean up and reuse resources for later DDP requests.
863 *
864 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
865 * struct scatterlist *sgl, unsigned int sgc);
866 * Called when the FCoE Target wants to initialize an I/O that
867 * is a possible candidate for Direct Data Placement (DDP). The LLD can
868 * perform necessary setup and returns 1 to indicate the device is set up
869 * successfully to perform DDP on this I/O, otherwise this returns 0.
870 *
871 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
872 * struct netdev_fcoe_hbainfo *hbainfo);
873 * Called when the FCoE Protocol stack wants information on the underlying
874 * device. This information is utilized by the FCoE protocol stack to
875 * register attributes with Fiber Channel management service as per the
876 * FC-GS Fabric Device Management Information(FDMI) specification.
877 *
878 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
879 * Called when the underlying device wants to override default World Wide
880 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
881 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
882 * protocol stack to use.
883 *
884 * RFS acceleration.
885 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
886 * u16 rxq_index, u32 flow_id);
887 * Set hardware filter for RFS. rxq_index is the target queue index;
888 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
889 * Return the filter ID on success, or a negative error code.
890 *
891 * Slave management functions (for bridge, bonding, etc).
892 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
893 * Called to make another netdev an underling.
894 *
895 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
896 * Called to release previously enslaved netdev.
897 *
898 * Feature/offload setting functions.
899 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
900 * netdev_features_t features);
901 * Adjusts the requested feature flags according to device-specific
902 * constraints, and returns the resulting flags. Must not modify
903 * the device state.
904 *
905 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
906 * Called to update device configuration to new features. Passed
907 * feature set might be less than what was returned by ndo_fix_features()).
908 * Must return >0 or -errno if it changed dev->features itself.
909 *
910 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
911 * struct net_device *dev,
912 * const unsigned char *addr, u16 flags)
913 * Adds an FDB entry to dev for addr.
914 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
915 * struct net_device *dev,
916 * const unsigned char *addr)
917 * Deletes the FDB entry from dev coresponding to addr.
918 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
919 * struct net_device *dev, int idx)
920 * Used to add FDB entries to dump requests. Implementers should add
921 * entries to skb and update idx with the number of entries.
922 *
923 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
924 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
925 * struct net_device *dev, u32 filter_mask)
926 *
927 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
928 * Called to change device carrier. Soft-devices (like dummy, team, etc)
929 * which do not represent real hardware may define this to allow their
930 * userspace components to manage their virtual carrier state. Devices
931 * that determine carrier state from physical hardware properties (eg
932 * network cables) or protocol-dependent mechanisms (eg
933 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
934 */
935struct net_device_ops {
936 int (*ndo_init)(struct net_device *dev);
937 void (*ndo_uninit)(struct net_device *dev);
938 int (*ndo_open)(struct net_device *dev);
939 int (*ndo_stop)(struct net_device *dev);
940 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
941 struct net_device *dev);
942 u16 (*ndo_select_queue)(struct net_device *dev,
943 struct sk_buff *skb);
944 void (*ndo_change_rx_flags)(struct net_device *dev,
945 int flags);
946 void (*ndo_set_rx_mode)(struct net_device *dev);
947 int (*ndo_set_mac_address)(struct net_device *dev,
948 void *addr);
949 int (*ndo_validate_addr)(struct net_device *dev);
950 int (*ndo_do_ioctl)(struct net_device *dev,
951 struct ifreq *ifr, int cmd);
952 int (*ndo_set_config)(struct net_device *dev,
953 struct ifmap *map);
954 int (*ndo_change_mtu)(struct net_device *dev,
955 int new_mtu);
956 int (*ndo_neigh_setup)(struct net_device *dev,
957 struct neigh_parms *);
958 void (*ndo_tx_timeout) (struct net_device *dev);
959
960 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
961 struct rtnl_link_stats64 *storage);
962 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
963
964 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
965 __be16 proto, u16 vid);
966 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
967 __be16 proto, u16 vid);
968#ifdef CONFIG_NET_POLL_CONTROLLER
969 void (*ndo_poll_controller)(struct net_device *dev);
970 int (*ndo_netpoll_setup)(struct net_device *dev,
971 struct netpoll_info *info,
972 gfp_t gfp);
973 void (*ndo_netpoll_cleanup)(struct net_device *dev);
974#endif
975#ifdef CONFIG_NET_LL_RX_POLL
976 int (*ndo_ll_poll)(struct napi_struct *dev);
977#endif
978 int (*ndo_set_vf_mac)(struct net_device *dev,
979 int queue, u8 *mac);
980 int (*ndo_set_vf_vlan)(struct net_device *dev,
981 int queue, u16 vlan, u8 qos);
982 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
983 int vf, int rate);
984 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
985 int vf, bool setting);
986 int (*ndo_get_vf_config)(struct net_device *dev,
987 int vf,
988 struct ifla_vf_info *ivf);
989 int (*ndo_set_vf_port)(struct net_device *dev,
990 int vf,
991 struct nlattr *port[]);
992 int (*ndo_get_vf_port)(struct net_device *dev,
993 int vf, struct sk_buff *skb);
994 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
995#if IS_ENABLED(CONFIG_FCOE)
996 int (*ndo_fcoe_enable)(struct net_device *dev);
997 int (*ndo_fcoe_disable)(struct net_device *dev);
998 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
999 u16 xid,
1000 struct scatterlist *sgl,
1001 unsigned int sgc);
1002 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1003 u16 xid);
1004 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1005 u16 xid,
1006 struct scatterlist *sgl,
1007 unsigned int sgc);
1008 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1009 struct netdev_fcoe_hbainfo *hbainfo);
1010#endif
1011
1012#if IS_ENABLED(CONFIG_LIBFCOE)
1013#define NETDEV_FCOE_WWNN 0
1014#define NETDEV_FCOE_WWPN 1
1015 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1016 u64 *wwn, int type);
1017#endif
1018
1019#ifdef CONFIG_RFS_ACCEL
1020 int (*ndo_rx_flow_steer)(struct net_device *dev,
1021 const struct sk_buff *skb,
1022 u16 rxq_index,
1023 u32 flow_id);
1024#endif
1025 int (*ndo_add_slave)(struct net_device *dev,
1026 struct net_device *slave_dev);
1027 int (*ndo_del_slave)(struct net_device *dev,
1028 struct net_device *slave_dev);
1029 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1030 netdev_features_t features);
1031 int (*ndo_set_features)(struct net_device *dev,
1032 netdev_features_t features);
1033 int (*ndo_neigh_construct)(struct neighbour *n);
1034 void (*ndo_neigh_destroy)(struct neighbour *n);
1035
1036 int (*ndo_fdb_add)(struct ndmsg *ndm,
1037 struct nlattr *tb[],
1038 struct net_device *dev,
1039 const unsigned char *addr,
1040 u16 flags);
1041 int (*ndo_fdb_del)(struct ndmsg *ndm,
1042 struct nlattr *tb[],
1043 struct net_device *dev,
1044 const unsigned char *addr);
1045 int (*ndo_fdb_dump)(struct sk_buff *skb,
1046 struct netlink_callback *cb,
1047 struct net_device *dev,
1048 int idx);
1049
1050 int (*ndo_bridge_setlink)(struct net_device *dev,
1051 struct nlmsghdr *nlh);
1052 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1053 u32 pid, u32 seq,
1054 struct net_device *dev,
1055 u32 filter_mask);
1056 int (*ndo_bridge_dellink)(struct net_device *dev,
1057 struct nlmsghdr *nlh);
1058 int (*ndo_change_carrier)(struct net_device *dev,
1059 bool new_carrier);
1060};
1061
1062/*
1063 * The DEVICE structure.
1064 * Actually, this whole structure is a big mistake. It mixes I/O
1065 * data with strictly "high-level" data, and it has to know about
1066 * almost every data structure used in the INET module.
1067 *
1068 * FIXME: cleanup struct net_device such that network protocol info
1069 * moves out.
1070 */
1071
1072struct net_device {
1073
1074 /*
1075 * This is the first field of the "visible" part of this structure
1076 * (i.e. as seen by users in the "Space.c" file). It is the name
1077 * of the interface.
1078 */
1079 char name[IFNAMSIZ];
1080
1081 /* device name hash chain, please keep it close to name[] */
1082 struct hlist_node name_hlist;
1083
1084 /* snmp alias */
1085 char *ifalias;
1086
1087 /*
1088 * I/O specific fields
1089 * FIXME: Merge these and struct ifmap into one
1090 */
1091 unsigned long mem_end; /* shared mem end */
1092 unsigned long mem_start; /* shared mem start */
1093 unsigned long base_addr; /* device I/O address */
1094 unsigned int irq; /* device IRQ number */
1095
1096 /*
1097 * Some hardware also needs these fields, but they are not
1098 * part of the usual set specified in Space.c.
1099 */
1100
1101 unsigned long state;
1102
1103 struct list_head dev_list;
1104 struct list_head napi_list;
1105 struct list_head unreg_list;
1106 struct list_head upper_dev_list; /* List of upper devices */
1107
1108
1109 /* currently active device features */
1110 netdev_features_t features;
1111 /* user-changeable features */
1112 netdev_features_t hw_features;
1113 /* user-requested features */
1114 netdev_features_t wanted_features;
1115 /* mask of features inheritable by VLAN devices */
1116 netdev_features_t vlan_features;
1117 /* mask of features inherited by encapsulating devices
1118 * This field indicates what encapsulation offloads
1119 * the hardware is capable of doing, and drivers will
1120 * need to set them appropriately.
1121 */
1122 netdev_features_t hw_enc_features;
1123 /* mask of fetures inheritable by MPLS */
1124 netdev_features_t mpls_features;
1125
1126 /* Interface index. Unique device identifier */
1127 int ifindex;
1128 int iflink;
1129
1130 struct net_device_stats stats;
1131 atomic_long_t rx_dropped; /* dropped packets by core network
1132 * Do not use this in drivers.
1133 */
1134
1135#ifdef CONFIG_WIRELESS_EXT
1136 /* List of functions to handle Wireless Extensions (instead of ioctl).
1137 * See <net/iw_handler.h> for details. Jean II */
1138 const struct iw_handler_def * wireless_handlers;
1139 /* Instance data managed by the core of Wireless Extensions. */
1140 struct iw_public_data * wireless_data;
1141#endif
1142 /* Management operations */
1143 const struct net_device_ops *netdev_ops;
1144 const struct ethtool_ops *ethtool_ops;
1145
1146 /* Hardware header description */
1147 const struct header_ops *header_ops;
1148
1149 unsigned int flags; /* interface flags (a la BSD) */
1150 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1151 * See if.h for definitions. */
1152 unsigned short gflags;
1153 unsigned short padded; /* How much padding added by alloc_netdev() */
1154
1155 unsigned char operstate; /* RFC2863 operstate */
1156 unsigned char link_mode; /* mapping policy to operstate */
1157
1158 unsigned char if_port; /* Selectable AUI, TP,..*/
1159 unsigned char dma; /* DMA channel */
1160
1161 unsigned int mtu; /* interface MTU value */
1162 unsigned short type; /* interface hardware type */
1163 unsigned short hard_header_len; /* hardware hdr length */
1164
1165 /* extra head- and tailroom the hardware may need, but not in all cases
1166 * can this be guaranteed, especially tailroom. Some cases also use
1167 * LL_MAX_HEADER instead to allocate the skb.
1168 */
1169 unsigned short needed_headroom;
1170 unsigned short needed_tailroom;
1171
1172 /* Interface address info. */
1173 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1174 unsigned char addr_assign_type; /* hw address assignment type */
1175 unsigned char addr_len; /* hardware address length */
1176 unsigned char neigh_priv_len;
1177 unsigned short dev_id; /* Used to differentiate devices
1178 * that share the same link
1179 * layer address
1180 */
1181 spinlock_t addr_list_lock;
1182 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1183 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1184 struct netdev_hw_addr_list dev_addrs; /* list of device
1185 * hw addresses
1186 */
1187#ifdef CONFIG_SYSFS
1188 struct kset *queues_kset;
1189#endif
1190
1191 bool uc_promisc;
1192 unsigned int promiscuity;
1193 unsigned int allmulti;
1194
1195
1196 /* Protocol specific pointers */
1197
1198#if IS_ENABLED(CONFIG_VLAN_8021Q)
1199 struct vlan_info __rcu *vlan_info; /* VLAN info */
1200#endif
1201#if IS_ENABLED(CONFIG_NET_DSA)
1202 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1203#endif
1204 void *atalk_ptr; /* AppleTalk link */
1205 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1206 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1207 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1208 void *ax25_ptr; /* AX.25 specific data */
1209 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1210 assign before registering */
1211
1212/*
1213 * Cache lines mostly used on receive path (including eth_type_trans())
1214 */
1215 unsigned long last_rx; /* Time of last Rx
1216 * This should not be set in
1217 * drivers, unless really needed,
1218 * because network stack (bonding)
1219 * use it if/when necessary, to
1220 * avoid dirtying this cache line.
1221 */
1222
1223 /* Interface address info used in eth_type_trans() */
1224 unsigned char *dev_addr; /* hw address, (before bcast
1225 because most packets are
1226 unicast) */
1227
1228
1229#ifdef CONFIG_RPS
1230 struct netdev_rx_queue *_rx;
1231
1232 /* Number of RX queues allocated at register_netdev() time */
1233 unsigned int num_rx_queues;
1234
1235 /* Number of RX queues currently active in device */
1236 unsigned int real_num_rx_queues;
1237
1238#endif
1239
1240 rx_handler_func_t __rcu *rx_handler;
1241 void __rcu *rx_handler_data;
1242
1243 struct netdev_queue __rcu *ingress_queue;
1244 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1245
1246
1247/*
1248 * Cache lines mostly used on transmit path
1249 */
1250 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1251
1252 /* Number of TX queues allocated at alloc_netdev_mq() time */
1253 unsigned int num_tx_queues;
1254
1255 /* Number of TX queues currently active in device */
1256 unsigned int real_num_tx_queues;
1257
1258 /* root qdisc from userspace point of view */
1259 struct Qdisc *qdisc;
1260
1261 unsigned long tx_queue_len; /* Max frames per queue allowed */
1262 spinlock_t tx_global_lock;
1263
1264#ifdef CONFIG_XPS
1265 struct xps_dev_maps __rcu *xps_maps;
1266#endif
1267#ifdef CONFIG_RFS_ACCEL
1268 /* CPU reverse-mapping for RX completion interrupts, indexed
1269 * by RX queue number. Assigned by driver. This must only be
1270 * set if the ndo_rx_flow_steer operation is defined. */
1271 struct cpu_rmap *rx_cpu_rmap;
1272#endif
1273
1274 /* These may be needed for future network-power-down code. */
1275
1276 /*
1277 * trans_start here is expensive for high speed devices on SMP,
1278 * please use netdev_queue->trans_start instead.
1279 */
1280 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1281
1282 int watchdog_timeo; /* used by dev_watchdog() */
1283 struct timer_list watchdog_timer;
1284
1285 /* Number of references to this device */
1286 int __percpu *pcpu_refcnt;
1287
1288 /* delayed register/unregister */
1289 struct list_head todo_list;
1290 /* device index hash chain */
1291 struct hlist_node index_hlist;
1292
1293 struct list_head link_watch_list;
1294
1295 /* register/unregister state machine */
1296 enum { NETREG_UNINITIALIZED=0,
1297 NETREG_REGISTERED, /* completed register_netdevice */
1298 NETREG_UNREGISTERING, /* called unregister_netdevice */
1299 NETREG_UNREGISTERED, /* completed unregister todo */
1300 NETREG_RELEASED, /* called free_netdev */
1301 NETREG_DUMMY, /* dummy device for NAPI poll */
1302 } reg_state:8;
1303
1304 bool dismantle; /* device is going do be freed */
1305
1306 enum {
1307 RTNL_LINK_INITIALIZED,
1308 RTNL_LINK_INITIALIZING,
1309 } rtnl_link_state:16;
1310
1311 /* Called from unregister, can be used to call free_netdev */
1312 void (*destructor)(struct net_device *dev);
1313
1314#ifdef CONFIG_NETPOLL
1315 struct netpoll_info __rcu *npinfo;
1316#endif
1317
1318#ifdef CONFIG_NET_NS
1319 /* Network namespace this network device is inside */
1320 struct net *nd_net;
1321#endif
1322
1323 /* mid-layer private */
1324 union {
1325 void *ml_priv;
1326 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1327 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1328 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1329 struct pcpu_vstats __percpu *vstats; /* veth stats */
1330 };
1331 /* GARP */
1332 struct garp_port __rcu *garp_port;
1333 /* MRP */
1334 struct mrp_port __rcu *mrp_port;
1335
1336 /* class/net/name entry */
1337 struct device dev;
1338 /* space for optional device, statistics, and wireless sysfs groups */
1339 const struct attribute_group *sysfs_groups[4];
1340
1341 /* rtnetlink link ops */
1342 const struct rtnl_link_ops *rtnl_link_ops;
1343
1344 /* for setting kernel sock attribute on TCP connection setup */
1345#define GSO_MAX_SIZE 65536
1346 unsigned int gso_max_size;
1347#define GSO_MAX_SEGS 65535
1348 u16 gso_max_segs;
1349
1350#ifdef CONFIG_DCB
1351 /* Data Center Bridging netlink ops */
1352 const struct dcbnl_rtnl_ops *dcbnl_ops;
1353#endif
1354 u8 num_tc;
1355 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1356 u8 prio_tc_map[TC_BITMASK + 1];
1357
1358#if IS_ENABLED(CONFIG_FCOE)
1359 /* max exchange id for FCoE LRO by ddp */
1360 unsigned int fcoe_ddp_xid;
1361#endif
1362#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1363 struct netprio_map __rcu *priomap;
1364#endif
1365 /* phy device may attach itself for hardware timestamping */
1366 struct phy_device *phydev;
1367
1368 struct lock_class_key *qdisc_tx_busylock;
1369
1370 /* group the device belongs to */
1371 int group;
1372
1373 struct pm_qos_request pm_qos_req;
1374};
1375#define to_net_dev(d) container_of(d, struct net_device, dev)
1376
1377#define NETDEV_ALIGN 32
1378
1379static inline
1380int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1381{
1382 return dev->prio_tc_map[prio & TC_BITMASK];
1383}
1384
1385static inline
1386int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1387{
1388 if (tc >= dev->num_tc)
1389 return -EINVAL;
1390
1391 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1392 return 0;
1393}
1394
1395static inline
1396void netdev_reset_tc(struct net_device *dev)
1397{
1398 dev->num_tc = 0;
1399 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1400 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1401}
1402
1403static inline
1404int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1405{
1406 if (tc >= dev->num_tc)
1407 return -EINVAL;
1408
1409 dev->tc_to_txq[tc].count = count;
1410 dev->tc_to_txq[tc].offset = offset;
1411 return 0;
1412}
1413
1414static inline
1415int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1416{
1417 if (num_tc > TC_MAX_QUEUE)
1418 return -EINVAL;
1419
1420 dev->num_tc = num_tc;
1421 return 0;
1422}
1423
1424static inline
1425int netdev_get_num_tc(struct net_device *dev)
1426{
1427 return dev->num_tc;
1428}
1429
1430static inline
1431struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1432 unsigned int index)
1433{
1434 return &dev->_tx[index];
1435}
1436
1437static inline void netdev_for_each_tx_queue(struct net_device *dev,
1438 void (*f)(struct net_device *,
1439 struct netdev_queue *,
1440 void *),
1441 void *arg)
1442{
1443 unsigned int i;
1444
1445 for (i = 0; i < dev->num_tx_queues; i++)
1446 f(dev, &dev->_tx[i], arg);
1447}
1448
1449extern struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1450 struct sk_buff *skb);
1451extern u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
1452
1453/*
1454 * Net namespace inlines
1455 */
1456static inline
1457struct net *dev_net(const struct net_device *dev)
1458{
1459 return read_pnet(&dev->nd_net);
1460}
1461
1462static inline
1463void dev_net_set(struct net_device *dev, struct net *net)
1464{
1465#ifdef CONFIG_NET_NS
1466 release_net(dev->nd_net);
1467 dev->nd_net = hold_net(net);
1468#endif
1469}
1470
1471static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1472{
1473#ifdef CONFIG_NET_DSA_TAG_DSA
1474 if (dev->dsa_ptr != NULL)
1475 return dsa_uses_dsa_tags(dev->dsa_ptr);
1476#endif
1477
1478 return 0;
1479}
1480
1481static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1482{
1483#ifdef CONFIG_NET_DSA_TAG_TRAILER
1484 if (dev->dsa_ptr != NULL)
1485 return dsa_uses_trailer_tags(dev->dsa_ptr);
1486#endif
1487
1488 return 0;
1489}
1490
1491/**
1492 * netdev_priv - access network device private data
1493 * @dev: network device
1494 *
1495 * Get network device private data
1496 */
1497static inline void *netdev_priv(const struct net_device *dev)
1498{
1499 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1500}
1501
1502/* Set the sysfs physical device reference for the network logical device
1503 * if set prior to registration will cause a symlink during initialization.
1504 */
1505#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1506
1507/* Set the sysfs device type for the network logical device to allow
1508 * fin grained indentification of different network device types. For
1509 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1510 */
1511#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1512
1513/* Default NAPI poll() weight
1514 * Device drivers are strongly advised to not use bigger value
1515 */
1516#define NAPI_POLL_WEIGHT 64
1517
1518/**
1519 * netif_napi_add - initialize a napi context
1520 * @dev: network device
1521 * @napi: napi context
1522 * @poll: polling function
1523 * @weight: default weight
1524 *
1525 * netif_napi_add() must be used to initialize a napi context prior to calling
1526 * *any* of the other napi related functions.
1527 */
1528void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1529 int (*poll)(struct napi_struct *, int), int weight);
1530
1531/**
1532 * netif_napi_del - remove a napi context
1533 * @napi: napi context
1534 *
1535 * netif_napi_del() removes a napi context from the network device napi list
1536 */
1537void netif_napi_del(struct napi_struct *napi);
1538
1539struct napi_gro_cb {
1540 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1541 void *frag0;
1542
1543 /* Length of frag0. */
1544 unsigned int frag0_len;
1545
1546 /* This indicates where we are processing relative to skb->data. */
1547 int data_offset;
1548
1549 /* This is non-zero if the packet cannot be merged with the new skb. */
1550 int flush;
1551
1552 /* Number of segments aggregated. */
1553 u16 count;
1554
1555 /* This is non-zero if the packet may be of the same flow. */
1556 u8 same_flow;
1557
1558 /* Free the skb? */
1559 u8 free;
1560#define NAPI_GRO_FREE 1
1561#define NAPI_GRO_FREE_STOLEN_HEAD 2
1562
1563 /* jiffies when first packet was created/queued */
1564 unsigned long age;
1565
1566 /* Used in ipv6_gro_receive() */
1567 int proto;
1568
1569 /* used in skb_gro_receive() slow path */
1570 struct sk_buff *last;
1571};
1572
1573#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1574
1575struct packet_type {
1576 __be16 type; /* This is really htons(ether_type). */
1577 struct net_device *dev; /* NULL is wildcarded here */
1578 int (*func) (struct sk_buff *,
1579 struct net_device *,
1580 struct packet_type *,
1581 struct net_device *);
1582 bool (*id_match)(struct packet_type *ptype,
1583 struct sock *sk);
1584 void *af_packet_priv;
1585 struct list_head list;
1586};
1587
1588struct offload_callbacks {
1589 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1590 netdev_features_t features);
1591 int (*gso_send_check)(struct sk_buff *skb);
1592 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1593 struct sk_buff *skb);
1594 int (*gro_complete)(struct sk_buff *skb);
1595};
1596
1597struct packet_offload {
1598 __be16 type; /* This is really htons(ether_type). */
1599 struct offload_callbacks callbacks;
1600 struct list_head list;
1601};
1602
1603#include <linux/notifier.h>
1604
1605/* netdevice notifier chain. Please remember to update the rtnetlink
1606 * notification exclusion list in rtnetlink_event() when adding new
1607 * types.
1608 */
1609#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1610#define NETDEV_DOWN 0x0002
1611#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1612 detected a hardware crash and restarted
1613 - we can use this eg to kick tcp sessions
1614 once done */
1615#define NETDEV_CHANGE 0x0004 /* Notify device state change */
1616#define NETDEV_REGISTER 0x0005
1617#define NETDEV_UNREGISTER 0x0006
1618#define NETDEV_CHANGEMTU 0x0007
1619#define NETDEV_CHANGEADDR 0x0008
1620#define NETDEV_GOING_DOWN 0x0009
1621#define NETDEV_CHANGENAME 0x000A
1622#define NETDEV_FEAT_CHANGE 0x000B
1623#define NETDEV_BONDING_FAILOVER 0x000C
1624#define NETDEV_PRE_UP 0x000D
1625#define NETDEV_PRE_TYPE_CHANGE 0x000E
1626#define NETDEV_POST_TYPE_CHANGE 0x000F
1627#define NETDEV_POST_INIT 0x0010
1628#define NETDEV_UNREGISTER_FINAL 0x0011
1629#define NETDEV_RELEASE 0x0012
1630#define NETDEV_NOTIFY_PEERS 0x0013
1631#define NETDEV_JOIN 0x0014
1632#define NETDEV_CHANGEUPPER 0x0015
1633
1634extern int register_netdevice_notifier(struct notifier_block *nb);
1635extern int unregister_netdevice_notifier(struct notifier_block *nb);
1636
1637struct netdev_notifier_info {
1638 struct net_device *dev;
1639};
1640
1641struct netdev_notifier_change_info {
1642 struct netdev_notifier_info info; /* must be first */
1643 unsigned int flags_changed;
1644};
1645
1646static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1647 struct net_device *dev)
1648{
1649 info->dev = dev;
1650}
1651
1652static inline struct net_device *
1653netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1654{
1655 return info->dev;
1656}
1657
1658extern int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
1659 struct netdev_notifier_info *info);
1660extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1661
1662
1663extern rwlock_t dev_base_lock; /* Device list lock */
1664
1665extern seqcount_t devnet_rename_seq; /* Device rename seq */
1666
1667
1668#define for_each_netdev(net, d) \
1669 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1670#define for_each_netdev_reverse(net, d) \
1671 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1672#define for_each_netdev_rcu(net, d) \
1673 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1674#define for_each_netdev_safe(net, d, n) \
1675 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1676#define for_each_netdev_continue(net, d) \
1677 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1678#define for_each_netdev_continue_rcu(net, d) \
1679 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1680#define for_each_netdev_in_bond_rcu(bond, slave) \
1681 for_each_netdev_rcu(&init_net, slave) \
1682 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1683#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1684
1685static inline struct net_device *next_net_device(struct net_device *dev)
1686{
1687 struct list_head *lh;
1688 struct net *net;
1689
1690 net = dev_net(dev);
1691 lh = dev->dev_list.next;
1692 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1693}
1694
1695static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1696{
1697 struct list_head *lh;
1698 struct net *net;
1699
1700 net = dev_net(dev);
1701 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1702 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1703}
1704
1705static inline struct net_device *first_net_device(struct net *net)
1706{
1707 return list_empty(&net->dev_base_head) ? NULL :
1708 net_device_entry(net->dev_base_head.next);
1709}
1710
1711static inline struct net_device *first_net_device_rcu(struct net *net)
1712{
1713 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1714
1715 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1716}
1717
1718extern int netdev_boot_setup_check(struct net_device *dev);
1719extern unsigned long netdev_boot_base(const char *prefix, int unit);
1720extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1721 const char *hwaddr);
1722extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1723extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1724extern void dev_add_pack(struct packet_type *pt);
1725extern void dev_remove_pack(struct packet_type *pt);
1726extern void __dev_remove_pack(struct packet_type *pt);
1727extern void dev_add_offload(struct packet_offload *po);
1728extern void dev_remove_offload(struct packet_offload *po);
1729extern void __dev_remove_offload(struct packet_offload *po);
1730
1731extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1732 unsigned short mask);
1733extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1734extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1735extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1736extern int dev_alloc_name(struct net_device *dev, const char *name);
1737extern int dev_open(struct net_device *dev);
1738extern int dev_close(struct net_device *dev);
1739extern void dev_disable_lro(struct net_device *dev);
1740extern int dev_loopback_xmit(struct sk_buff *newskb);
1741extern int dev_queue_xmit(struct sk_buff *skb);
1742extern int register_netdevice(struct net_device *dev);
1743extern void unregister_netdevice_queue(struct net_device *dev,
1744 struct list_head *head);
1745extern void unregister_netdevice_many(struct list_head *head);
1746static inline void unregister_netdevice(struct net_device *dev)
1747{
1748 unregister_netdevice_queue(dev, NULL);
1749}
1750
1751extern int netdev_refcnt_read(const struct net_device *dev);
1752extern void free_netdev(struct net_device *dev);
1753extern void synchronize_net(void);
1754extern int init_dummy_netdev(struct net_device *dev);
1755
1756extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1757extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1758extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1759extern int dev_restart(struct net_device *dev);
1760#ifdef CONFIG_NETPOLL_TRAP
1761extern int netpoll_trap(void);
1762#endif
1763extern int skb_gro_receive(struct sk_buff **head,
1764 struct sk_buff *skb);
1765
1766static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1767{
1768 return NAPI_GRO_CB(skb)->data_offset;
1769}
1770
1771static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1772{
1773 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1774}
1775
1776static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1777{
1778 NAPI_GRO_CB(skb)->data_offset += len;
1779}
1780
1781static inline void *skb_gro_header_fast(struct sk_buff *skb,
1782 unsigned int offset)
1783{
1784 return NAPI_GRO_CB(skb)->frag0 + offset;
1785}
1786
1787static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1788{
1789 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1790}
1791
1792static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1793 unsigned int offset)
1794{
1795 if (!pskb_may_pull(skb, hlen))
1796 return NULL;
1797
1798 NAPI_GRO_CB(skb)->frag0 = NULL;
1799 NAPI_GRO_CB(skb)->frag0_len = 0;
1800 return skb->data + offset;
1801}
1802
1803static inline void *skb_gro_mac_header(struct sk_buff *skb)
1804{
1805 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1806}
1807
1808static inline void *skb_gro_network_header(struct sk_buff *skb)
1809{
1810 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1811 skb_network_offset(skb);
1812}
1813
1814static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1815 unsigned short type,
1816 const void *daddr, const void *saddr,
1817 unsigned int len)
1818{
1819 if (!dev->header_ops || !dev->header_ops->create)
1820 return 0;
1821
1822 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1823}
1824
1825static inline int dev_parse_header(const struct sk_buff *skb,
1826 unsigned char *haddr)
1827{
1828 const struct net_device *dev = skb->dev;
1829
1830 if (!dev->header_ops || !dev->header_ops->parse)
1831 return 0;
1832 return dev->header_ops->parse(skb, haddr);
1833}
1834
1835typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1836extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1837static inline int unregister_gifconf(unsigned int family)
1838{
1839 return register_gifconf(family, NULL);
1840}
1841
1842#ifdef CONFIG_NET_FLOW_LIMIT
1843#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
1844struct sd_flow_limit {
1845 u64 count;
1846 unsigned int num_buckets;
1847 unsigned int history_head;
1848 u16 history[FLOW_LIMIT_HISTORY];
1849 u8 buckets[];
1850};
1851
1852extern int netdev_flow_limit_table_len;
1853#endif /* CONFIG_NET_FLOW_LIMIT */
1854
1855/*
1856 * Incoming packets are placed on per-cpu queues
1857 */
1858struct softnet_data {
1859 struct Qdisc *output_queue;
1860 struct Qdisc **output_queue_tailp;
1861 struct list_head poll_list;
1862 struct sk_buff *completion_queue;
1863 struct sk_buff_head process_queue;
1864
1865 /* stats */
1866 unsigned int processed;
1867 unsigned int time_squeeze;
1868 unsigned int cpu_collision;
1869 unsigned int received_rps;
1870
1871#ifdef CONFIG_RPS
1872 struct softnet_data *rps_ipi_list;
1873
1874 /* Elements below can be accessed between CPUs for RPS */
1875 struct call_single_data csd ____cacheline_aligned_in_smp;
1876 struct softnet_data *rps_ipi_next;
1877 unsigned int cpu;
1878 unsigned int input_queue_head;
1879 unsigned int input_queue_tail;
1880#endif
1881 unsigned int dropped;
1882 struct sk_buff_head input_pkt_queue;
1883 struct napi_struct backlog;
1884
1885#ifdef CONFIG_NET_FLOW_LIMIT
1886 struct sd_flow_limit __rcu *flow_limit;
1887#endif
1888};
1889
1890static inline void input_queue_head_incr(struct softnet_data *sd)
1891{
1892#ifdef CONFIG_RPS
1893 sd->input_queue_head++;
1894#endif
1895}
1896
1897static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1898 unsigned int *qtail)
1899{
1900#ifdef CONFIG_RPS
1901 *qtail = ++sd->input_queue_tail;
1902#endif
1903}
1904
1905DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1906
1907extern void __netif_schedule(struct Qdisc *q);
1908
1909static inline void netif_schedule_queue(struct netdev_queue *txq)
1910{
1911 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1912 __netif_schedule(txq->qdisc);
1913}
1914
1915static inline void netif_tx_schedule_all(struct net_device *dev)
1916{
1917 unsigned int i;
1918
1919 for (i = 0; i < dev->num_tx_queues; i++)
1920 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1921}
1922
1923static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1924{
1925 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1926}
1927
1928/**
1929 * netif_start_queue - allow transmit
1930 * @dev: network device
1931 *
1932 * Allow upper layers to call the device hard_start_xmit routine.
1933 */
1934static inline void netif_start_queue(struct net_device *dev)
1935{
1936 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1937}
1938
1939static inline void netif_tx_start_all_queues(struct net_device *dev)
1940{
1941 unsigned int i;
1942
1943 for (i = 0; i < dev->num_tx_queues; i++) {
1944 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1945 netif_tx_start_queue(txq);
1946 }
1947}
1948
1949static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1950{
1951#ifdef CONFIG_NETPOLL_TRAP
1952 if (netpoll_trap()) {
1953 netif_tx_start_queue(dev_queue);
1954 return;
1955 }
1956#endif
1957 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1958 __netif_schedule(dev_queue->qdisc);
1959}
1960
1961/**
1962 * netif_wake_queue - restart transmit
1963 * @dev: network device
1964 *
1965 * Allow upper layers to call the device hard_start_xmit routine.
1966 * Used for flow control when transmit resources are available.
1967 */
1968static inline void netif_wake_queue(struct net_device *dev)
1969{
1970 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1971}
1972
1973static inline void netif_tx_wake_all_queues(struct net_device *dev)
1974{
1975 unsigned int i;
1976
1977 for (i = 0; i < dev->num_tx_queues; i++) {
1978 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1979 netif_tx_wake_queue(txq);
1980 }
1981}
1982
1983static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1984{
1985 if (WARN_ON(!dev_queue)) {
1986 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1987 return;
1988 }
1989 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1990}
1991
1992/**
1993 * netif_stop_queue - stop transmitted packets
1994 * @dev: network device
1995 *
1996 * Stop upper layers calling the device hard_start_xmit routine.
1997 * Used for flow control when transmit resources are unavailable.
1998 */
1999static inline void netif_stop_queue(struct net_device *dev)
2000{
2001 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2002}
2003
2004static inline void netif_tx_stop_all_queues(struct net_device *dev)
2005{
2006 unsigned int i;
2007
2008 for (i = 0; i < dev->num_tx_queues; i++) {
2009 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2010 netif_tx_stop_queue(txq);
2011 }
2012}
2013
2014static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2015{
2016 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2017}
2018
2019/**
2020 * netif_queue_stopped - test if transmit queue is flowblocked
2021 * @dev: network device
2022 *
2023 * Test if transmit queue on device is currently unable to send.
2024 */
2025static inline bool netif_queue_stopped(const struct net_device *dev)
2026{
2027 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2028}
2029
2030static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2031{
2032 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2033}
2034
2035static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2036{
2037 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2038}
2039
2040static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2041 unsigned int bytes)
2042{
2043#ifdef CONFIG_BQL
2044 dql_queued(&dev_queue->dql, bytes);
2045
2046 if (likely(dql_avail(&dev_queue->dql) >= 0))
2047 return;
2048
2049 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2050
2051 /*
2052 * The XOFF flag must be set before checking the dql_avail below,
2053 * because in netdev_tx_completed_queue we update the dql_completed
2054 * before checking the XOFF flag.
2055 */
2056 smp_mb();
2057
2058 /* check again in case another CPU has just made room avail */
2059 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2060 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2061#endif
2062}
2063
2064static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2065{
2066 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2067}
2068
2069static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2070 unsigned int pkts, unsigned int bytes)
2071{
2072#ifdef CONFIG_BQL
2073 if (unlikely(!bytes))
2074 return;
2075
2076 dql_completed(&dev_queue->dql, bytes);
2077
2078 /*
2079 * Without the memory barrier there is a small possiblity that
2080 * netdev_tx_sent_queue will miss the update and cause the queue to
2081 * be stopped forever
2082 */
2083 smp_mb();
2084
2085 if (dql_avail(&dev_queue->dql) < 0)
2086 return;
2087
2088 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2089 netif_schedule_queue(dev_queue);
2090#endif
2091}
2092
2093static inline void netdev_completed_queue(struct net_device *dev,
2094 unsigned int pkts, unsigned int bytes)
2095{
2096 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2097}
2098
2099static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2100{
2101#ifdef CONFIG_BQL
2102 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2103 dql_reset(&q->dql);
2104#endif
2105}
2106
2107static inline void netdev_reset_queue(struct net_device *dev_queue)
2108{
2109 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2110}
2111
2112/**
2113 * netif_running - test if up
2114 * @dev: network device
2115 *
2116 * Test if the device has been brought up.
2117 */
2118static inline bool netif_running(const struct net_device *dev)
2119{
2120 return test_bit(__LINK_STATE_START, &dev->state);
2121}
2122
2123/*
2124 * Routines to manage the subqueues on a device. We only need start
2125 * stop, and a check if it's stopped. All other device management is
2126 * done at the overall netdevice level.
2127 * Also test the device if we're multiqueue.
2128 */
2129
2130/**
2131 * netif_start_subqueue - allow sending packets on subqueue
2132 * @dev: network device
2133 * @queue_index: sub queue index
2134 *
2135 * Start individual transmit queue of a device with multiple transmit queues.
2136 */
2137static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2138{
2139 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2140
2141 netif_tx_start_queue(txq);
2142}
2143
2144/**
2145 * netif_stop_subqueue - stop sending packets on subqueue
2146 * @dev: network device
2147 * @queue_index: sub queue index
2148 *
2149 * Stop individual transmit queue of a device with multiple transmit queues.
2150 */
2151static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2152{
2153 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2154#ifdef CONFIG_NETPOLL_TRAP
2155 if (netpoll_trap())
2156 return;
2157#endif
2158 netif_tx_stop_queue(txq);
2159}
2160
2161/**
2162 * netif_subqueue_stopped - test status of subqueue
2163 * @dev: network device
2164 * @queue_index: sub queue index
2165 *
2166 * Check individual transmit queue of a device with multiple transmit queues.
2167 */
2168static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2169 u16 queue_index)
2170{
2171 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2172
2173 return netif_tx_queue_stopped(txq);
2174}
2175
2176static inline bool netif_subqueue_stopped(const struct net_device *dev,
2177 struct sk_buff *skb)
2178{
2179 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2180}
2181
2182/**
2183 * netif_wake_subqueue - allow sending packets on subqueue
2184 * @dev: network device
2185 * @queue_index: sub queue index
2186 *
2187 * Resume individual transmit queue of a device with multiple transmit queues.
2188 */
2189static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2190{
2191 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2192#ifdef CONFIG_NETPOLL_TRAP
2193 if (netpoll_trap())
2194 return;
2195#endif
2196 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2197 __netif_schedule(txq->qdisc);
2198}
2199
2200#ifdef CONFIG_XPS
2201extern int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask,
2202 u16 index);
2203#else
2204static inline int netif_set_xps_queue(struct net_device *dev,
2205 struct cpumask *mask,
2206 u16 index)
2207{
2208 return 0;
2209}
2210#endif
2211
2212/*
2213 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2214 * as a distribution range limit for the returned value.
2215 */
2216static inline u16 skb_tx_hash(const struct net_device *dev,
2217 const struct sk_buff *skb)
2218{
2219 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2220}
2221
2222/**
2223 * netif_is_multiqueue - test if device has multiple transmit queues
2224 * @dev: network device
2225 *
2226 * Check if device has multiple transmit queues
2227 */
2228static inline bool netif_is_multiqueue(const struct net_device *dev)
2229{
2230 return dev->num_tx_queues > 1;
2231}
2232
2233extern int netif_set_real_num_tx_queues(struct net_device *dev,
2234 unsigned int txq);
2235
2236#ifdef CONFIG_RPS
2237extern int netif_set_real_num_rx_queues(struct net_device *dev,
2238 unsigned int rxq);
2239#else
2240static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2241 unsigned int rxq)
2242{
2243 return 0;
2244}
2245#endif
2246
2247static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2248 const struct net_device *from_dev)
2249{
2250 int err;
2251
2252 err = netif_set_real_num_tx_queues(to_dev,
2253 from_dev->real_num_tx_queues);
2254 if (err)
2255 return err;
2256#ifdef CONFIG_RPS
2257 return netif_set_real_num_rx_queues(to_dev,
2258 from_dev->real_num_rx_queues);
2259#else
2260 return 0;
2261#endif
2262}
2263
2264#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2265extern int netif_get_num_default_rss_queues(void);
2266
2267/* Use this variant when it is known for sure that it
2268 * is executing from hardware interrupt context or with hardware interrupts
2269 * disabled.
2270 */
2271extern void dev_kfree_skb_irq(struct sk_buff *skb);
2272
2273/* Use this variant in places where it could be invoked
2274 * from either hardware interrupt or other context, with hardware interrupts
2275 * either disabled or enabled.
2276 */
2277extern void dev_kfree_skb_any(struct sk_buff *skb);
2278
2279extern int netif_rx(struct sk_buff *skb);
2280extern int netif_rx_ni(struct sk_buff *skb);
2281extern int netif_receive_skb(struct sk_buff *skb);
2282extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2283 struct sk_buff *skb);
2284extern void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2285extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2286extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2287
2288static inline void napi_free_frags(struct napi_struct *napi)
2289{
2290 kfree_skb(napi->skb);
2291 napi->skb = NULL;
2292}
2293
2294extern int netdev_rx_handler_register(struct net_device *dev,
2295 rx_handler_func_t *rx_handler,
2296 void *rx_handler_data);
2297extern void netdev_rx_handler_unregister(struct net_device *dev);
2298
2299extern bool dev_valid_name(const char *name);
2300extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2301extern int dev_ethtool(struct net *net, struct ifreq *);
2302extern unsigned int dev_get_flags(const struct net_device *);
2303extern int __dev_change_flags(struct net_device *, unsigned int flags);
2304extern int dev_change_flags(struct net_device *, unsigned int);
2305extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2306extern int dev_change_name(struct net_device *, const char *);
2307extern int dev_set_alias(struct net_device *, const char *, size_t);
2308extern int dev_change_net_namespace(struct net_device *,
2309 struct net *, const char *);
2310extern int dev_set_mtu(struct net_device *, int);
2311extern void dev_set_group(struct net_device *, int);
2312extern int dev_set_mac_address(struct net_device *,
2313 struct sockaddr *);
2314extern int dev_change_carrier(struct net_device *,
2315 bool new_carrier);
2316extern int dev_hard_start_xmit(struct sk_buff *skb,
2317 struct net_device *dev,
2318 struct netdev_queue *txq);
2319extern int dev_forward_skb(struct net_device *dev,
2320 struct sk_buff *skb);
2321
2322extern int netdev_budget;
2323
2324/* Called by rtnetlink.c:rtnl_unlock() */
2325extern void netdev_run_todo(void);
2326
2327/**
2328 * dev_put - release reference to device
2329 * @dev: network device
2330 *
2331 * Release reference to device to allow it to be freed.
2332 */
2333static inline void dev_put(struct net_device *dev)
2334{
2335 this_cpu_dec(*dev->pcpu_refcnt);
2336}
2337
2338/**
2339 * dev_hold - get reference to device
2340 * @dev: network device
2341 *
2342 * Hold reference to device to keep it from being freed.
2343 */
2344static inline void dev_hold(struct net_device *dev)
2345{
2346 this_cpu_inc(*dev->pcpu_refcnt);
2347}
2348
2349/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2350 * and _off may be called from IRQ context, but it is caller
2351 * who is responsible for serialization of these calls.
2352 *
2353 * The name carrier is inappropriate, these functions should really be
2354 * called netif_lowerlayer_*() because they represent the state of any
2355 * kind of lower layer not just hardware media.
2356 */
2357
2358extern void linkwatch_init_dev(struct net_device *dev);
2359extern void linkwatch_fire_event(struct net_device *dev);
2360extern void linkwatch_forget_dev(struct net_device *dev);
2361
2362/**
2363 * netif_carrier_ok - test if carrier present
2364 * @dev: network device
2365 *
2366 * Check if carrier is present on device
2367 */
2368static inline bool netif_carrier_ok(const struct net_device *dev)
2369{
2370 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2371}
2372
2373extern unsigned long dev_trans_start(struct net_device *dev);
2374
2375extern void __netdev_watchdog_up(struct net_device *dev);
2376
2377extern void netif_carrier_on(struct net_device *dev);
2378
2379extern void netif_carrier_off(struct net_device *dev);
2380
2381/**
2382 * netif_dormant_on - mark device as dormant.
2383 * @dev: network device
2384 *
2385 * Mark device as dormant (as per RFC2863).
2386 *
2387 * The dormant state indicates that the relevant interface is not
2388 * actually in a condition to pass packets (i.e., it is not 'up') but is
2389 * in a "pending" state, waiting for some external event. For "on-
2390 * demand" interfaces, this new state identifies the situation where the
2391 * interface is waiting for events to place it in the up state.
2392 *
2393 */
2394static inline void netif_dormant_on(struct net_device *dev)
2395{
2396 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2397 linkwatch_fire_event(dev);
2398}
2399
2400/**
2401 * netif_dormant_off - set device as not dormant.
2402 * @dev: network device
2403 *
2404 * Device is not in dormant state.
2405 */
2406static inline void netif_dormant_off(struct net_device *dev)
2407{
2408 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2409 linkwatch_fire_event(dev);
2410}
2411
2412/**
2413 * netif_dormant - test if carrier present
2414 * @dev: network device
2415 *
2416 * Check if carrier is present on device
2417 */
2418static inline bool netif_dormant(const struct net_device *dev)
2419{
2420 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2421}
2422
2423
2424/**
2425 * netif_oper_up - test if device is operational
2426 * @dev: network device
2427 *
2428 * Check if carrier is operational
2429 */
2430static inline bool netif_oper_up(const struct net_device *dev)
2431{
2432 return (dev->operstate == IF_OPER_UP ||
2433 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2434}
2435
2436/**
2437 * netif_device_present - is device available or removed
2438 * @dev: network device
2439 *
2440 * Check if device has not been removed from system.
2441 */
2442static inline bool netif_device_present(struct net_device *dev)
2443{
2444 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2445}
2446
2447extern void netif_device_detach(struct net_device *dev);
2448
2449extern void netif_device_attach(struct net_device *dev);
2450
2451/*
2452 * Network interface message level settings
2453 */
2454
2455enum {
2456 NETIF_MSG_DRV = 0x0001,
2457 NETIF_MSG_PROBE = 0x0002,
2458 NETIF_MSG_LINK = 0x0004,
2459 NETIF_MSG_TIMER = 0x0008,
2460 NETIF_MSG_IFDOWN = 0x0010,
2461 NETIF_MSG_IFUP = 0x0020,
2462 NETIF_MSG_RX_ERR = 0x0040,
2463 NETIF_MSG_TX_ERR = 0x0080,
2464 NETIF_MSG_TX_QUEUED = 0x0100,
2465 NETIF_MSG_INTR = 0x0200,
2466 NETIF_MSG_TX_DONE = 0x0400,
2467 NETIF_MSG_RX_STATUS = 0x0800,
2468 NETIF_MSG_PKTDATA = 0x1000,
2469 NETIF_MSG_HW = 0x2000,
2470 NETIF_MSG_WOL = 0x4000,
2471};
2472
2473#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2474#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2475#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2476#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2477#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2478#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2479#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2480#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2481#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2482#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2483#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2484#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2485#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2486#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2487#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2488
2489static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2490{
2491 /* use default */
2492 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2493 return default_msg_enable_bits;
2494 if (debug_value == 0) /* no output */
2495 return 0;
2496 /* set low N bits */
2497 return (1 << debug_value) - 1;
2498}
2499
2500static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2501{
2502 spin_lock(&txq->_xmit_lock);
2503 txq->xmit_lock_owner = cpu;
2504}
2505
2506static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2507{
2508 spin_lock_bh(&txq->_xmit_lock);
2509 txq->xmit_lock_owner = smp_processor_id();
2510}
2511
2512static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2513{
2514 bool ok = spin_trylock(&txq->_xmit_lock);
2515 if (likely(ok))
2516 txq->xmit_lock_owner = smp_processor_id();
2517 return ok;
2518}
2519
2520static inline void __netif_tx_unlock(struct netdev_queue *txq)
2521{
2522 txq->xmit_lock_owner = -1;
2523 spin_unlock(&txq->_xmit_lock);
2524}
2525
2526static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2527{
2528 txq->xmit_lock_owner = -1;
2529 spin_unlock_bh(&txq->_xmit_lock);
2530}
2531
2532static inline void txq_trans_update(struct netdev_queue *txq)
2533{
2534 if (txq->xmit_lock_owner != -1)
2535 txq->trans_start = jiffies;
2536}
2537
2538/**
2539 * netif_tx_lock - grab network device transmit lock
2540 * @dev: network device
2541 *
2542 * Get network device transmit lock
2543 */
2544static inline void netif_tx_lock(struct net_device *dev)
2545{
2546 unsigned int i;
2547 int cpu;
2548
2549 spin_lock(&dev->tx_global_lock);
2550 cpu = smp_processor_id();
2551 for (i = 0; i < dev->num_tx_queues; i++) {
2552 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2553
2554 /* We are the only thread of execution doing a
2555 * freeze, but we have to grab the _xmit_lock in
2556 * order to synchronize with threads which are in
2557 * the ->hard_start_xmit() handler and already
2558 * checked the frozen bit.
2559 */
2560 __netif_tx_lock(txq, cpu);
2561 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2562 __netif_tx_unlock(txq);
2563 }
2564}
2565
2566static inline void netif_tx_lock_bh(struct net_device *dev)
2567{
2568 local_bh_disable();
2569 netif_tx_lock(dev);
2570}
2571
2572static inline void netif_tx_unlock(struct net_device *dev)
2573{
2574 unsigned int i;
2575
2576 for (i = 0; i < dev->num_tx_queues; i++) {
2577 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2578
2579 /* No need to grab the _xmit_lock here. If the
2580 * queue is not stopped for another reason, we
2581 * force a schedule.
2582 */
2583 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2584 netif_schedule_queue(txq);
2585 }
2586 spin_unlock(&dev->tx_global_lock);
2587}
2588
2589static inline void netif_tx_unlock_bh(struct net_device *dev)
2590{
2591 netif_tx_unlock(dev);
2592 local_bh_enable();
2593}
2594
2595#define HARD_TX_LOCK(dev, txq, cpu) { \
2596 if ((dev->features & NETIF_F_LLTX) == 0) { \
2597 __netif_tx_lock(txq, cpu); \
2598 } \
2599}
2600
2601#define HARD_TX_UNLOCK(dev, txq) { \
2602 if ((dev->features & NETIF_F_LLTX) == 0) { \
2603 __netif_tx_unlock(txq); \
2604 } \
2605}
2606
2607static inline void netif_tx_disable(struct net_device *dev)
2608{
2609 unsigned int i;
2610 int cpu;
2611
2612 local_bh_disable();
2613 cpu = smp_processor_id();
2614 for (i = 0; i < dev->num_tx_queues; i++) {
2615 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2616
2617 __netif_tx_lock(txq, cpu);
2618 netif_tx_stop_queue(txq);
2619 __netif_tx_unlock(txq);
2620 }
2621 local_bh_enable();
2622}
2623
2624static inline void netif_addr_lock(struct net_device *dev)
2625{
2626 spin_lock(&dev->addr_list_lock);
2627}
2628
2629static inline void netif_addr_lock_nested(struct net_device *dev)
2630{
2631 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2632}
2633
2634static inline void netif_addr_lock_bh(struct net_device *dev)
2635{
2636 spin_lock_bh(&dev->addr_list_lock);
2637}
2638
2639static inline void netif_addr_unlock(struct net_device *dev)
2640{
2641 spin_unlock(&dev->addr_list_lock);
2642}
2643
2644static inline void netif_addr_unlock_bh(struct net_device *dev)
2645{
2646 spin_unlock_bh(&dev->addr_list_lock);
2647}
2648
2649/*
2650 * dev_addrs walker. Should be used only for read access. Call with
2651 * rcu_read_lock held.
2652 */
2653#define for_each_dev_addr(dev, ha) \
2654 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2655
2656/* These functions live elsewhere (drivers/net/net_init.c, but related) */
2657
2658extern void ether_setup(struct net_device *dev);
2659
2660/* Support for loadable net-drivers */
2661extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2662 void (*setup)(struct net_device *),
2663 unsigned int txqs, unsigned int rxqs);
2664#define alloc_netdev(sizeof_priv, name, setup) \
2665 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2666
2667#define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2668 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2669
2670extern int register_netdev(struct net_device *dev);
2671extern void unregister_netdev(struct net_device *dev);
2672
2673/* General hardware address lists handling functions */
2674extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2675 struct netdev_hw_addr_list *from_list,
2676 int addr_len, unsigned char addr_type);
2677extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2678 struct netdev_hw_addr_list *from_list,
2679 int addr_len, unsigned char addr_type);
2680extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2681 struct netdev_hw_addr_list *from_list,
2682 int addr_len);
2683extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2684 struct netdev_hw_addr_list *from_list,
2685 int addr_len);
2686extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2687extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2688
2689/* Functions used for device addresses handling */
2690extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2691 unsigned char addr_type);
2692extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2693 unsigned char addr_type);
2694extern int dev_addr_add_multiple(struct net_device *to_dev,
2695 struct net_device *from_dev,
2696 unsigned char addr_type);
2697extern int dev_addr_del_multiple(struct net_device *to_dev,
2698 struct net_device *from_dev,
2699 unsigned char addr_type);
2700extern void dev_addr_flush(struct net_device *dev);
2701extern int dev_addr_init(struct net_device *dev);
2702
2703/* Functions used for unicast addresses handling */
2704extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2705extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2706extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2707extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2708extern int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2709extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2710extern void dev_uc_flush(struct net_device *dev);
2711extern void dev_uc_init(struct net_device *dev);
2712
2713/* Functions used for multicast addresses handling */
2714extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2715extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2716extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2717extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2718extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2719extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2720extern int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2721extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2722extern void dev_mc_flush(struct net_device *dev);
2723extern void dev_mc_init(struct net_device *dev);
2724
2725/* Functions used for secondary unicast and multicast support */
2726extern void dev_set_rx_mode(struct net_device *dev);
2727extern void __dev_set_rx_mode(struct net_device *dev);
2728extern int dev_set_promiscuity(struct net_device *dev, int inc);
2729extern int dev_set_allmulti(struct net_device *dev, int inc);
2730extern void netdev_state_change(struct net_device *dev);
2731extern void netdev_notify_peers(struct net_device *dev);
2732extern void netdev_features_change(struct net_device *dev);
2733/* Load a device via the kmod */
2734extern void dev_load(struct net *net, const char *name);
2735extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2736 struct rtnl_link_stats64 *storage);
2737extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2738 const struct net_device_stats *netdev_stats);
2739
2740extern int netdev_max_backlog;
2741extern int netdev_tstamp_prequeue;
2742extern int weight_p;
2743extern int bpf_jit_enable;
2744
2745extern bool netdev_has_upper_dev(struct net_device *dev,
2746 struct net_device *upper_dev);
2747extern bool netdev_has_any_upper_dev(struct net_device *dev);
2748extern struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2749extern struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2750extern int netdev_upper_dev_link(struct net_device *dev,
2751 struct net_device *upper_dev);
2752extern int netdev_master_upper_dev_link(struct net_device *dev,
2753 struct net_device *upper_dev);
2754extern void netdev_upper_dev_unlink(struct net_device *dev,
2755 struct net_device *upper_dev);
2756extern int skb_checksum_help(struct sk_buff *skb);
2757extern struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2758 netdev_features_t features, bool tx_path);
2759extern struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2760 netdev_features_t features);
2761
2762static inline
2763struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2764{
2765 return __skb_gso_segment(skb, features, true);
2766}
2767__be16 skb_network_protocol(struct sk_buff *skb);
2768
2769static inline bool can_checksum_protocol(netdev_features_t features,
2770 __be16 protocol)
2771{
2772 return ((features & NETIF_F_GEN_CSUM) ||
2773 ((features & NETIF_F_V4_CSUM) &&
2774 protocol == htons(ETH_P_IP)) ||
2775 ((features & NETIF_F_V6_CSUM) &&
2776 protocol == htons(ETH_P_IPV6)) ||
2777 ((features & NETIF_F_FCOE_CRC) &&
2778 protocol == htons(ETH_P_FCOE)));
2779}
2780
2781#ifdef CONFIG_BUG
2782extern void netdev_rx_csum_fault(struct net_device *dev);
2783#else
2784static inline void netdev_rx_csum_fault(struct net_device *dev)
2785{
2786}
2787#endif
2788/* rx skb timestamps */
2789extern void net_enable_timestamp(void);
2790extern void net_disable_timestamp(void);
2791
2792#ifdef CONFIG_PROC_FS
2793extern int __init dev_proc_init(void);
2794#else
2795#define dev_proc_init() 0
2796#endif
2797
2798extern int netdev_class_create_file(struct class_attribute *class_attr);
2799extern void netdev_class_remove_file(struct class_attribute *class_attr);
2800
2801extern struct kobj_ns_type_operations net_ns_type_operations;
2802
2803extern const char *netdev_drivername(const struct net_device *dev);
2804
2805extern void linkwatch_run_queue(void);
2806
2807static inline netdev_features_t netdev_get_wanted_features(
2808 struct net_device *dev)
2809{
2810 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2811}
2812netdev_features_t netdev_increment_features(netdev_features_t all,
2813 netdev_features_t one, netdev_features_t mask);
2814
2815/* Allow TSO being used on stacked device :
2816 * Performing the GSO segmentation before last device
2817 * is a performance improvement.
2818 */
2819static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
2820 netdev_features_t mask)
2821{
2822 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
2823}
2824
2825int __netdev_update_features(struct net_device *dev);
2826void netdev_update_features(struct net_device *dev);
2827void netdev_change_features(struct net_device *dev);
2828
2829void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2830 struct net_device *dev);
2831
2832netdev_features_t netif_skb_features(struct sk_buff *skb);
2833
2834static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2835{
2836 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2837
2838 /* check flags correspondence */
2839 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2840 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2841 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2842 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2843 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2844 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2845
2846 return (features & feature) == feature;
2847}
2848
2849static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2850{
2851 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2852 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2853}
2854
2855static inline bool netif_needs_gso(struct sk_buff *skb,
2856 netdev_features_t features)
2857{
2858 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2859 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2860 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2861}
2862
2863static inline void netif_set_gso_max_size(struct net_device *dev,
2864 unsigned int size)
2865{
2866 dev->gso_max_size = size;
2867}
2868
2869static inline bool netif_is_bond_master(struct net_device *dev)
2870{
2871 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
2872}
2873
2874static inline bool netif_is_bond_slave(struct net_device *dev)
2875{
2876 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2877}
2878
2879static inline bool netif_supports_nofcs(struct net_device *dev)
2880{
2881 return dev->priv_flags & IFF_SUPP_NOFCS;
2882}
2883
2884extern struct pernet_operations __net_initdata loopback_net_ops;
2885
2886/* Logging, debugging and troubleshooting/diagnostic helpers. */
2887
2888/* netdev_printk helpers, similar to dev_printk */
2889
2890static inline const char *netdev_name(const struct net_device *dev)
2891{
2892 if (dev->reg_state != NETREG_REGISTERED)
2893 return "(unregistered net_device)";
2894 return dev->name;
2895}
2896
2897extern __printf(3, 4)
2898int netdev_printk(const char *level, const struct net_device *dev,
2899 const char *format, ...);
2900extern __printf(2, 3)
2901int netdev_emerg(const struct net_device *dev, const char *format, ...);
2902extern __printf(2, 3)
2903int netdev_alert(const struct net_device *dev, const char *format, ...);
2904extern __printf(2, 3)
2905int netdev_crit(const struct net_device *dev, const char *format, ...);
2906extern __printf(2, 3)
2907int netdev_err(const struct net_device *dev, const char *format, ...);
2908extern __printf(2, 3)
2909int netdev_warn(const struct net_device *dev, const char *format, ...);
2910extern __printf(2, 3)
2911int netdev_notice(const struct net_device *dev, const char *format, ...);
2912extern __printf(2, 3)
2913int netdev_info(const struct net_device *dev, const char *format, ...);
2914
2915#define MODULE_ALIAS_NETDEV(device) \
2916 MODULE_ALIAS("netdev-" device)
2917
2918#if defined(CONFIG_DYNAMIC_DEBUG)
2919#define netdev_dbg(__dev, format, args...) \
2920do { \
2921 dynamic_netdev_dbg(__dev, format, ##args); \
2922} while (0)
2923#elif defined(DEBUG)
2924#define netdev_dbg(__dev, format, args...) \
2925 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2926#else
2927#define netdev_dbg(__dev, format, args...) \
2928({ \
2929 if (0) \
2930 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2931 0; \
2932})
2933#endif
2934
2935#if defined(VERBOSE_DEBUG)
2936#define netdev_vdbg netdev_dbg
2937#else
2938
2939#define netdev_vdbg(dev, format, args...) \
2940({ \
2941 if (0) \
2942 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2943 0; \
2944})
2945#endif
2946
2947/*
2948 * netdev_WARN() acts like dev_printk(), but with the key difference
2949 * of using a WARN/WARN_ON to get the message out, including the
2950 * file/line information and a backtrace.
2951 */
2952#define netdev_WARN(dev, format, args...) \
2953 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2954
2955/* netif printk helpers, similar to netdev_printk */
2956
2957#define netif_printk(priv, type, level, dev, fmt, args...) \
2958do { \
2959 if (netif_msg_##type(priv)) \
2960 netdev_printk(level, (dev), fmt, ##args); \
2961} while (0)
2962
2963#define netif_level(level, priv, type, dev, fmt, args...) \
2964do { \
2965 if (netif_msg_##type(priv)) \
2966 netdev_##level(dev, fmt, ##args); \
2967} while (0)
2968
2969#define netif_emerg(priv, type, dev, fmt, args...) \
2970 netif_level(emerg, priv, type, dev, fmt, ##args)
2971#define netif_alert(priv, type, dev, fmt, args...) \
2972 netif_level(alert, priv, type, dev, fmt, ##args)
2973#define netif_crit(priv, type, dev, fmt, args...) \
2974 netif_level(crit, priv, type, dev, fmt, ##args)
2975#define netif_err(priv, type, dev, fmt, args...) \
2976 netif_level(err, priv, type, dev, fmt, ##args)
2977#define netif_warn(priv, type, dev, fmt, args...) \
2978 netif_level(warn, priv, type, dev, fmt, ##args)
2979#define netif_notice(priv, type, dev, fmt, args...) \
2980 netif_level(notice, priv, type, dev, fmt, ##args)
2981#define netif_info(priv, type, dev, fmt, args...) \
2982 netif_level(info, priv, type, dev, fmt, ##args)
2983
2984#if defined(CONFIG_DYNAMIC_DEBUG)
2985#define netif_dbg(priv, type, netdev, format, args...) \
2986do { \
2987 if (netif_msg_##type(priv)) \
2988 dynamic_netdev_dbg(netdev, format, ##args); \
2989} while (0)
2990#elif defined(DEBUG)
2991#define netif_dbg(priv, type, dev, format, args...) \
2992 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2993#else
2994#define netif_dbg(priv, type, dev, format, args...) \
2995({ \
2996 if (0) \
2997 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2998 0; \
2999})
3000#endif
3001
3002#if defined(VERBOSE_DEBUG)
3003#define netif_vdbg netif_dbg
3004#else
3005#define netif_vdbg(priv, type, dev, format, args...) \
3006({ \
3007 if (0) \
3008 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3009 0; \
3010})
3011#endif
3012
3013/*
3014 * The list of packet types we will receive (as opposed to discard)
3015 * and the routines to invoke.
3016 *
3017 * Why 16. Because with 16 the only overlap we get on a hash of the
3018 * low nibble of the protocol value is RARP/SNAP/X.25.
3019 *
3020 * NOTE: That is no longer true with the addition of VLAN tags. Not
3021 * sure which should go first, but I bet it won't make much
3022 * difference if we are running VLANs. The good news is that
3023 * this protocol won't be in the list unless compiled in, so
3024 * the average user (w/out VLANs) will not be adversely affected.
3025 * --BLG
3026 *
3027 * 0800 IP
3028 * 8100 802.1Q VLAN
3029 * 0001 802.3
3030 * 0002 AX.25
3031 * 0004 802.2
3032 * 8035 RARP
3033 * 0005 SNAP
3034 * 0805 X.25
3035 * 0806 ARP
3036 * 8137 IPX
3037 * 0009 Localtalk
3038 * 86DD IPv6
3039 */
3040#define PTYPE_HASH_SIZE (16)
3041#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3042
3043#endif /* _LINUX_NETDEVICE_H */