net_device: Reorder members to fill holes
[deliverable/linux.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 <linux/prefetch.h>
34#include <asm/cache.h>
35#include <asm/byteorder.h>
36
37#include <linux/percpu.h>
38#include <linux/rculist.h>
39#include <linux/dmaengine.h>
40#include <linux/workqueue.h>
41#include <linux/dynamic_queue_limits.h>
42
43#include <linux/ethtool.h>
44#include <net/net_namespace.h>
45#include <net/dsa.h>
46#ifdef CONFIG_DCB
47#include <net/dcbnl.h>
48#endif
49#include <net/netprio_cgroup.h>
50
51#include <linux/netdev_features.h>
52#include <linux/neighbour.h>
53#include <uapi/linux/netdevice.h>
54#include <uapi/linux/if_bonding.h>
55
56struct netpoll_info;
57struct device;
58struct phy_device;
59/* 802.11 specific */
60struct wireless_dev;
61/* 802.15.4 specific */
62struct wpan_dev;
63
64void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
71/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
90#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101/* Driver transmit return codes */
102#define NETDEV_TX_MASK 0xf0
103
104enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109};
110typedef enum netdev_tx netdev_tx_t;
111
112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
141#else
142# define LL_MAX_HEADER 32
143#endif
144
145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
155 */
156
157struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
187#ifdef CONFIG_RPS
188#include <linux/static_key.h>
189extern struct static_key rps_needed;
190#endif
191
192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
204#define NETDEV_HW_ADDR_T_MULTICAST 5
205 bool global_use;
206 int sync_cnt;
207 int refcount;
208 int synced;
209 struct rcu_head rcu_head;
210};
211
212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224#define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226
227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229#define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
231
232struct hh_cache {
233 u16 hh_len;
234 u16 __pad;
235 seqlock_t hh_lock;
236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258
259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268};
269
270/* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
275enum netdev_state_t {
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
281};
282
283
284/*
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
287 */
288struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291};
292#define NETDEV_BOOT_SETUP_MAX 8
293
294int __init netdev_boot_setup(char *str);
295
296/*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312#ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
315#endif
316 struct net_device *dev;
317 struct sk_buff *gro_list;
318 struct sk_buff *skb;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
323};
324
325enum {
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
330};
331
332enum gro_result {
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338};
339typedef enum gro_result gro_result_t;
340
341/*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387};
388typedef enum rx_handler_result rx_handler_result_t;
389typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390
391void __napi_schedule(struct napi_struct *n);
392void __napi_schedule_irqoff(struct napi_struct *n);
393
394static inline bool napi_disable_pending(struct napi_struct *n)
395{
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397}
398
399/**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
407 */
408static inline bool napi_schedule_prep(struct napi_struct *n)
409{
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
412}
413
414/**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421static inline void napi_schedule(struct napi_struct *n)
422{
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425}
426
427/**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433static inline void napi_schedule_irqoff(struct napi_struct *n)
434{
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437}
438
439/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440static inline bool napi_reschedule(struct napi_struct *napi)
441{
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
444 return true;
445 }
446 return false;
447}
448
449void __napi_complete(struct napi_struct *n);
450void napi_complete_done(struct napi_struct *n, int work_done);
451/**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
457 */
458static inline void napi_complete(struct napi_struct *n)
459{
460 return napi_complete_done(n, 0);
461}
462
463/**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
470struct napi_struct *napi_by_id(unsigned int napi_id);
471
472/**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
478void napi_hash_add(struct napi_struct *napi);
479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
487void napi_hash_del(struct napi_struct *napi);
488
489/**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
496void napi_disable(struct napi_struct *n);
497
498/**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505static inline void napi_enable(struct napi_struct *n)
506{
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510}
511
512#ifdef CONFIG_SMP
513/**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521static inline void napi_synchronize(const struct napi_struct *n)
522{
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525}
526#else
527# define napi_synchronize(n) barrier()
528#endif
529
530enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
534};
535
536#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
546/*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
555
556struct netdev_queue {
557/*
558 * read mostly part
559 */
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
563#ifdef CONFIG_SYSFS
564 struct kobject kobj;
565#endif
566#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568#endif
569/*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
584
585 unsigned long state;
586
587#ifdef CONFIG_BQL
588 struct dql dql;
589#endif
590 unsigned long tx_maxrate;
591} ____cacheline_aligned_in_smp;
592
593static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594{
595#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597#else
598 return NUMA_NO_NODE;
599#endif
600}
601
602static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603{
604#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606#endif
607}
608
609#ifdef CONFIG_RPS
610/*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618};
619#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
620
621/*
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
625 */
626struct rps_dev_flow {
627 u16 cpu;
628 u16 filter;
629 unsigned int last_qtail;
630};
631#define RPS_NO_FILTER 0xffff
632
633/*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
639 struct rps_dev_flow flows[0];
640};
641#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
642 ((_num) * sizeof(struct rps_dev_flow)))
643
644/*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
653 */
654struct rps_sock_flow_table {
655 u32 mask;
656
657 u32 ents[0] ____cacheline_aligned_in_smp;
658};
659#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
660
661#define RPS_NO_CPU 0xffff
662
663extern u32 rps_cpu_mask;
664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
666static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668{
669 if (table && hash) {
670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
672
673 /* We only give a hint, preemption can change cpu under us */
674 val |= raw_smp_processor_id();
675
676 if (table->ents[index] != val)
677 table->ents[index] = val;
678 }
679}
680
681#ifdef CONFIG_RFS_ACCEL
682bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
684#endif
685#endif /* CONFIG_RPS */
686
687/* This structure contains an instance of an RX queue. */
688struct netdev_rx_queue {
689#ifdef CONFIG_RPS
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
692#endif
693 struct kobject kobj;
694 struct net_device *dev;
695} ____cacheline_aligned_in_smp;
696
697/*
698 * RX queue sysfs structures and functions.
699 */
700struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706};
707
708#ifdef CONFIG_XPS
709/*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718};
719#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
720#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723/*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726struct xps_dev_maps {
727 struct rcu_head rcu;
728 struct xps_map __rcu *cpu_map[0];
729};
730#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732#endif /* CONFIG_XPS */
733
734#define TC_MAX_QUEUE 16
735#define TC_BITMASK 15
736/* HW offloaded queuing disciplines txq count and offset maps */
737struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740};
741
742#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743/*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756};
757#endif
758
759#define MAX_PHYS_ITEM_ID_LEN 32
760
761/* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
763 */
764struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
766 unsigned char id_len;
767};
768
769typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
772/*
773 * This structure defines the management hooks for network devices.
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
776 *
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
782 *
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
786 *
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
789 * state.
790 *
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
793 * state.
794 *
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
797 * Called when a packet needs to be transmitted.
798 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
799 * the queue before that can happen; it's for obsolete devices and weird
800 * corner cases, but the stack really does a non-trivial amount
801 * of useless work if you return NETDEV_TX_BUSY.
802 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
803 * Required can not be NULL.
804 *
805 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
806 * void *accel_priv, select_queue_fallback_t fallback);
807 * Called to decide which queue to when device supports multiple
808 * transmit queues.
809 *
810 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
811 * This function is called to allow device receiver to make
812 * changes to configuration when multicast or promiscious is enabled.
813 *
814 * void (*ndo_set_rx_mode)(struct net_device *dev);
815 * This function is called device changes address list filtering.
816 * If driver handles unicast address filtering, it should set
817 * IFF_UNICAST_FLT to its priv_flags.
818 *
819 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
820 * This function is called when the Media Access Control address
821 * needs to be changed. If this interface is not defined, the
822 * mac address can not be changed.
823 *
824 * int (*ndo_validate_addr)(struct net_device *dev);
825 * Test if Media Access Control address is valid for the device.
826 *
827 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
828 * Called when a user request an ioctl which can't be handled by
829 * the generic interface code. If not defined ioctl's return
830 * not supported error code.
831 *
832 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
833 * Used to set network devices bus interface parameters. This interface
834 * is retained for legacy reason, new devices should use the bus
835 * interface (PCI) for low level management.
836 *
837 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
838 * Called when a user wants to change the Maximum Transfer Unit
839 * of a device. If not defined, any request to change MTU will
840 * will return an error.
841 *
842 * void (*ndo_tx_timeout)(struct net_device *dev);
843 * Callback uses when the transmitter has not made any progress
844 * for dev->watchdog ticks.
845 *
846 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
847 * struct rtnl_link_stats64 *storage);
848 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
849 * Called when a user wants to get the network device usage
850 * statistics. Drivers must do one of the following:
851 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
852 * rtnl_link_stats64 structure passed by the caller.
853 * 2. Define @ndo_get_stats to update a net_device_stats structure
854 * (which should normally be dev->stats) and return a pointer to
855 * it. The structure may be changed asynchronously only if each
856 * field is written atomically.
857 * 3. Update dev->stats asynchronously and atomically, and define
858 * neither operation.
859 *
860 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
861 * If device support VLAN filtering this function is called when a
862 * VLAN id is registered.
863 *
864 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
865 * If device support VLAN filtering this function is called when a
866 * VLAN id is unregistered.
867 *
868 * void (*ndo_poll_controller)(struct net_device *dev);
869 *
870 * SR-IOV management functions.
871 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
872 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
873 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
874 * int max_tx_rate);
875 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
876 * int (*ndo_get_vf_config)(struct net_device *dev,
877 * int vf, struct ifla_vf_info *ivf);
878 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
879 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
880 * struct nlattr *port[]);
881 *
882 * Enable or disable the VF ability to query its RSS Redirection Table and
883 * Hash Key. This is needed since on some devices VF share this information
884 * with PF and querying it may adduce a theoretical security risk.
885 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
886 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
887 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
888 * Called to setup 'tc' number of traffic classes in the net device. This
889 * is always called from the stack with the rtnl lock held and netif tx
890 * queues stopped. This allows the netdevice to perform queue management
891 * safely.
892 *
893 * Fiber Channel over Ethernet (FCoE) offload functions.
894 * int (*ndo_fcoe_enable)(struct net_device *dev);
895 * Called when the FCoE protocol stack wants to start using LLD for FCoE
896 * so the underlying device can perform whatever needed configuration or
897 * initialization to support acceleration of FCoE traffic.
898 *
899 * int (*ndo_fcoe_disable)(struct net_device *dev);
900 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
901 * so the underlying device can perform whatever needed clean-ups to
902 * stop supporting acceleration of FCoE traffic.
903 *
904 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
905 * struct scatterlist *sgl, unsigned int sgc);
906 * Called when the FCoE Initiator wants to initialize an I/O that
907 * is a possible candidate for Direct Data Placement (DDP). The LLD can
908 * perform necessary setup and returns 1 to indicate the device is set up
909 * successfully to perform DDP on this I/O, otherwise this returns 0.
910 *
911 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
912 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
913 * indicated by the FC exchange id 'xid', so the underlying device can
914 * clean up and reuse resources for later DDP requests.
915 *
916 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
917 * struct scatterlist *sgl, unsigned int sgc);
918 * Called when the FCoE Target wants to initialize an I/O that
919 * is a possible candidate for Direct Data Placement (DDP). The LLD can
920 * perform necessary setup and returns 1 to indicate the device is set up
921 * successfully to perform DDP on this I/O, otherwise this returns 0.
922 *
923 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
924 * struct netdev_fcoe_hbainfo *hbainfo);
925 * Called when the FCoE Protocol stack wants information on the underlying
926 * device. This information is utilized by the FCoE protocol stack to
927 * register attributes with Fiber Channel management service as per the
928 * FC-GS Fabric Device Management Information(FDMI) specification.
929 *
930 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
931 * Called when the underlying device wants to override default World Wide
932 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
933 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
934 * protocol stack to use.
935 *
936 * RFS acceleration.
937 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
938 * u16 rxq_index, u32 flow_id);
939 * Set hardware filter for RFS. rxq_index is the target queue index;
940 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
941 * Return the filter ID on success, or a negative error code.
942 *
943 * Slave management functions (for bridge, bonding, etc).
944 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
945 * Called to make another netdev an underling.
946 *
947 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
948 * Called to release previously enslaved netdev.
949 *
950 * Feature/offload setting functions.
951 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
952 * netdev_features_t features);
953 * Adjusts the requested feature flags according to device-specific
954 * constraints, and returns the resulting flags. Must not modify
955 * the device state.
956 *
957 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
958 * Called to update device configuration to new features. Passed
959 * feature set might be less than what was returned by ndo_fix_features()).
960 * Must return >0 or -errno if it changed dev->features itself.
961 *
962 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
963 * struct net_device *dev,
964 * const unsigned char *addr, u16 vid, u16 flags)
965 * Adds an FDB entry to dev for addr.
966 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
967 * struct net_device *dev,
968 * const unsigned char *addr, u16 vid)
969 * Deletes the FDB entry from dev coresponding to addr.
970 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
971 * struct net_device *dev, struct net_device *filter_dev,
972 * int idx)
973 * Used to add FDB entries to dump requests. Implementers should add
974 * entries to skb and update idx with the number of entries.
975 *
976 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
977 * u16 flags)
978 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
979 * struct net_device *dev, u32 filter_mask)
980 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
981 * u16 flags);
982 *
983 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
984 * Called to change device carrier. Soft-devices (like dummy, team, etc)
985 * which do not represent real hardware may define this to allow their
986 * userspace components to manage their virtual carrier state. Devices
987 * that determine carrier state from physical hardware properties (eg
988 * network cables) or protocol-dependent mechanisms (eg
989 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
990 *
991 * int (*ndo_get_phys_port_id)(struct net_device *dev,
992 * struct netdev_phys_item_id *ppid);
993 * Called to get ID of physical port of this device. If driver does
994 * not implement this, it is assumed that the hw is not able to have
995 * multiple net devices on single physical port.
996 *
997 * void (*ndo_add_vxlan_port)(struct net_device *dev,
998 * sa_family_t sa_family, __be16 port);
999 * Called by vxlan to notiy a driver about the UDP port and socket
1000 * address family that vxlan is listnening to. It is called only when
1001 * a new port starts listening. The operation is protected by the
1002 * vxlan_net->sock_lock.
1003 *
1004 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1005 * sa_family_t sa_family, __be16 port);
1006 * Called by vxlan to notify the driver about a UDP port and socket
1007 * address family that vxlan is not listening to anymore. The operation
1008 * is protected by the vxlan_net->sock_lock.
1009 *
1010 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1011 * struct net_device *dev)
1012 * Called by upper layer devices to accelerate switching or other
1013 * station functionality into hardware. 'pdev is the lowerdev
1014 * to use for the offload and 'dev' is the net device that will
1015 * back the offload. Returns a pointer to the private structure
1016 * the upper layer will maintain.
1017 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1018 * Called by upper layer device to delete the station created
1019 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1020 * the station and priv is the structure returned by the add
1021 * operation.
1022 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1023 * struct net_device *dev,
1024 * void *priv);
1025 * Callback to use for xmit over the accelerated station. This
1026 * is used in place of ndo_start_xmit on accelerated net
1027 * devices.
1028 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1029 * struct net_device *dev
1030 * netdev_features_t features);
1031 * Called by core transmit path to determine if device is capable of
1032 * performing offload operations on a given packet. This is to give
1033 * the device an opportunity to implement any restrictions that cannot
1034 * be otherwise expressed by feature flags. The check is called with
1035 * the set of features that the stack has calculated and it returns
1036 * those the driver believes to be appropriate.
1037 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1038 * int queue_index, u32 maxrate);
1039 * Called when a user wants to set a max-rate limitation of specific
1040 * TX queue.
1041 * int (*ndo_get_iflink)(const struct net_device *dev);
1042 * Called to get the iflink value of this device.
1043 */
1044struct net_device_ops {
1045 int (*ndo_init)(struct net_device *dev);
1046 void (*ndo_uninit)(struct net_device *dev);
1047 int (*ndo_open)(struct net_device *dev);
1048 int (*ndo_stop)(struct net_device *dev);
1049 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1050 struct net_device *dev);
1051 u16 (*ndo_select_queue)(struct net_device *dev,
1052 struct sk_buff *skb,
1053 void *accel_priv,
1054 select_queue_fallback_t fallback);
1055 void (*ndo_change_rx_flags)(struct net_device *dev,
1056 int flags);
1057 void (*ndo_set_rx_mode)(struct net_device *dev);
1058 int (*ndo_set_mac_address)(struct net_device *dev,
1059 void *addr);
1060 int (*ndo_validate_addr)(struct net_device *dev);
1061 int (*ndo_do_ioctl)(struct net_device *dev,
1062 struct ifreq *ifr, int cmd);
1063 int (*ndo_set_config)(struct net_device *dev,
1064 struct ifmap *map);
1065 int (*ndo_change_mtu)(struct net_device *dev,
1066 int new_mtu);
1067 int (*ndo_neigh_setup)(struct net_device *dev,
1068 struct neigh_parms *);
1069 void (*ndo_tx_timeout) (struct net_device *dev);
1070
1071 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1072 struct rtnl_link_stats64 *storage);
1073 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1074
1075 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1076 __be16 proto, u16 vid);
1077 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1078 __be16 proto, u16 vid);
1079#ifdef CONFIG_NET_POLL_CONTROLLER
1080 void (*ndo_poll_controller)(struct net_device *dev);
1081 int (*ndo_netpoll_setup)(struct net_device *dev,
1082 struct netpoll_info *info);
1083 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1084#endif
1085#ifdef CONFIG_NET_RX_BUSY_POLL
1086 int (*ndo_busy_poll)(struct napi_struct *dev);
1087#endif
1088 int (*ndo_set_vf_mac)(struct net_device *dev,
1089 int queue, u8 *mac);
1090 int (*ndo_set_vf_vlan)(struct net_device *dev,
1091 int queue, u16 vlan, u8 qos);
1092 int (*ndo_set_vf_rate)(struct net_device *dev,
1093 int vf, int min_tx_rate,
1094 int max_tx_rate);
1095 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1096 int vf, bool setting);
1097 int (*ndo_get_vf_config)(struct net_device *dev,
1098 int vf,
1099 struct ifla_vf_info *ivf);
1100 int (*ndo_set_vf_link_state)(struct net_device *dev,
1101 int vf, int link_state);
1102 int (*ndo_set_vf_port)(struct net_device *dev,
1103 int vf,
1104 struct nlattr *port[]);
1105 int (*ndo_get_vf_port)(struct net_device *dev,
1106 int vf, struct sk_buff *skb);
1107 int (*ndo_set_vf_rss_query_en)(
1108 struct net_device *dev,
1109 int vf, bool setting);
1110 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1111#if IS_ENABLED(CONFIG_FCOE)
1112 int (*ndo_fcoe_enable)(struct net_device *dev);
1113 int (*ndo_fcoe_disable)(struct net_device *dev);
1114 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1115 u16 xid,
1116 struct scatterlist *sgl,
1117 unsigned int sgc);
1118 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1119 u16 xid);
1120 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1121 u16 xid,
1122 struct scatterlist *sgl,
1123 unsigned int sgc);
1124 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1125 struct netdev_fcoe_hbainfo *hbainfo);
1126#endif
1127
1128#if IS_ENABLED(CONFIG_LIBFCOE)
1129#define NETDEV_FCOE_WWNN 0
1130#define NETDEV_FCOE_WWPN 1
1131 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1132 u64 *wwn, int type);
1133#endif
1134
1135#ifdef CONFIG_RFS_ACCEL
1136 int (*ndo_rx_flow_steer)(struct net_device *dev,
1137 const struct sk_buff *skb,
1138 u16 rxq_index,
1139 u32 flow_id);
1140#endif
1141 int (*ndo_add_slave)(struct net_device *dev,
1142 struct net_device *slave_dev);
1143 int (*ndo_del_slave)(struct net_device *dev,
1144 struct net_device *slave_dev);
1145 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1146 netdev_features_t features);
1147 int (*ndo_set_features)(struct net_device *dev,
1148 netdev_features_t features);
1149 int (*ndo_neigh_construct)(struct neighbour *n);
1150 void (*ndo_neigh_destroy)(struct neighbour *n);
1151
1152 int (*ndo_fdb_add)(struct ndmsg *ndm,
1153 struct nlattr *tb[],
1154 struct net_device *dev,
1155 const unsigned char *addr,
1156 u16 vid,
1157 u16 flags);
1158 int (*ndo_fdb_del)(struct ndmsg *ndm,
1159 struct nlattr *tb[],
1160 struct net_device *dev,
1161 const unsigned char *addr,
1162 u16 vid);
1163 int (*ndo_fdb_dump)(struct sk_buff *skb,
1164 struct netlink_callback *cb,
1165 struct net_device *dev,
1166 struct net_device *filter_dev,
1167 int idx);
1168
1169 int (*ndo_bridge_setlink)(struct net_device *dev,
1170 struct nlmsghdr *nlh,
1171 u16 flags);
1172 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1173 u32 pid, u32 seq,
1174 struct net_device *dev,
1175 u32 filter_mask);
1176 int (*ndo_bridge_dellink)(struct net_device *dev,
1177 struct nlmsghdr *nlh,
1178 u16 flags);
1179 int (*ndo_change_carrier)(struct net_device *dev,
1180 bool new_carrier);
1181 int (*ndo_get_phys_port_id)(struct net_device *dev,
1182 struct netdev_phys_item_id *ppid);
1183 int (*ndo_get_phys_port_name)(struct net_device *dev,
1184 char *name, size_t len);
1185 void (*ndo_add_vxlan_port)(struct net_device *dev,
1186 sa_family_t sa_family,
1187 __be16 port);
1188 void (*ndo_del_vxlan_port)(struct net_device *dev,
1189 sa_family_t sa_family,
1190 __be16 port);
1191
1192 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1193 struct net_device *dev);
1194 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1195 void *priv);
1196
1197 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1198 struct net_device *dev,
1199 void *priv);
1200 int (*ndo_get_lock_subclass)(struct net_device *dev);
1201 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1202 struct net_device *dev,
1203 netdev_features_t features);
1204 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1205 int queue_index,
1206 u32 maxrate);
1207 int (*ndo_get_iflink)(const struct net_device *dev);
1208};
1209
1210/**
1211 * enum net_device_priv_flags - &struct net_device priv_flags
1212 *
1213 * These are the &struct net_device, they are only set internally
1214 * by drivers and used in the kernel. These flags are invisible to
1215 * userspace, this means that the order of these flags can change
1216 * during any kernel release.
1217 *
1218 * You should have a pretty good reason to be extending these flags.
1219 *
1220 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1221 * @IFF_EBRIDGE: Ethernet bridging device
1222 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1223 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1224 * @IFF_MASTER_ALB: bonding master, balance-alb
1225 * @IFF_BONDING: bonding master or slave
1226 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1227 * @IFF_ISATAP: ISATAP interface (RFC4214)
1228 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1229 * @IFF_WAN_HDLC: WAN HDLC device
1230 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1231 * release skb->dst
1232 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1233 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1234 * @IFF_MACVLAN_PORT: device used as macvlan port
1235 * @IFF_BRIDGE_PORT: device used as bridge port
1236 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1237 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1238 * @IFF_UNICAST_FLT: Supports unicast filtering
1239 * @IFF_TEAM_PORT: device used as team port
1240 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1241 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1242 * change when it's running
1243 * @IFF_MACVLAN: Macvlan device
1244 */
1245enum netdev_priv_flags {
1246 IFF_802_1Q_VLAN = 1<<0,
1247 IFF_EBRIDGE = 1<<1,
1248 IFF_SLAVE_INACTIVE = 1<<2,
1249 IFF_MASTER_8023AD = 1<<3,
1250 IFF_MASTER_ALB = 1<<4,
1251 IFF_BONDING = 1<<5,
1252 IFF_SLAVE_NEEDARP = 1<<6,
1253 IFF_ISATAP = 1<<7,
1254 IFF_MASTER_ARPMON = 1<<8,
1255 IFF_WAN_HDLC = 1<<9,
1256 IFF_XMIT_DST_RELEASE = 1<<10,
1257 IFF_DONT_BRIDGE = 1<<11,
1258 IFF_DISABLE_NETPOLL = 1<<12,
1259 IFF_MACVLAN_PORT = 1<<13,
1260 IFF_BRIDGE_PORT = 1<<14,
1261 IFF_OVS_DATAPATH = 1<<15,
1262 IFF_TX_SKB_SHARING = 1<<16,
1263 IFF_UNICAST_FLT = 1<<17,
1264 IFF_TEAM_PORT = 1<<18,
1265 IFF_SUPP_NOFCS = 1<<19,
1266 IFF_LIVE_ADDR_CHANGE = 1<<20,
1267 IFF_MACVLAN = 1<<21,
1268 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1269 IFF_IPVLAN_MASTER = 1<<23,
1270 IFF_IPVLAN_SLAVE = 1<<24,
1271};
1272
1273#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1274#define IFF_EBRIDGE IFF_EBRIDGE
1275#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1276#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1277#define IFF_MASTER_ALB IFF_MASTER_ALB
1278#define IFF_BONDING IFF_BONDING
1279#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1280#define IFF_ISATAP IFF_ISATAP
1281#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1282#define IFF_WAN_HDLC IFF_WAN_HDLC
1283#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1284#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1285#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1286#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1287#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1288#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1289#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1290#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1291#define IFF_TEAM_PORT IFF_TEAM_PORT
1292#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1293#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1294#define IFF_MACVLAN IFF_MACVLAN
1295#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1296#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1297#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1298
1299/**
1300 * struct net_device - The DEVICE structure.
1301 * Actually, this whole structure is a big mistake. It mixes I/O
1302 * data with strictly "high-level" data, and it has to know about
1303 * almost every data structure used in the INET module.
1304 *
1305 * @name: This is the first field of the "visible" part of this structure
1306 * (i.e. as seen by users in the "Space.c" file). It is the name
1307 * of the interface.
1308 *
1309 * @name_hlist: Device name hash chain, please keep it close to name[]
1310 * @ifalias: SNMP alias
1311 * @mem_end: Shared memory end
1312 * @mem_start: Shared memory start
1313 * @base_addr: Device I/O address
1314 * @irq: Device IRQ number
1315 *
1316 * @carrier_changes: Stats to monitor carrier on<->off transitions
1317 *
1318 * @state: Generic network queuing layer state, see netdev_state_t
1319 * @dev_list: The global list of network devices
1320 * @napi_list: List entry, that is used for polling napi devices
1321 * @unreg_list: List entry, that is used, when we are unregistering the
1322 * device, see the function unregister_netdev
1323 * @close_list: List entry, that is used, when we are closing the device
1324 *
1325 * @adj_list: Directly linked devices, like slaves for bonding
1326 * @all_adj_list: All linked devices, *including* neighbours
1327 * @features: Currently active device features
1328 * @hw_features: User-changeable features
1329 *
1330 * @wanted_features: User-requested features
1331 * @vlan_features: Mask of features inheritable by VLAN devices
1332 *
1333 * @hw_enc_features: Mask of features inherited by encapsulating devices
1334 * This field indicates what encapsulation
1335 * offloads the hardware is capable of doing,
1336 * and drivers will need to set them appropriately.
1337 *
1338 * @mpls_features: Mask of features inheritable by MPLS
1339 *
1340 * @ifindex: interface index
1341 * @group: The group, that the device belongs to
1342 *
1343 * @stats: Statistics struct, which was left as a legacy, use
1344 * rtnl_link_stats64 instead
1345 *
1346 * @rx_dropped: Dropped packets by core network,
1347 * do not use this in drivers
1348 * @tx_dropped: Dropped packets by core network,
1349 * do not use this in drivers
1350 *
1351 * @wireless_handlers: List of functions to handle Wireless Extensions,
1352 * instead of ioctl,
1353 * see <net/iw_handler.h> for details.
1354 * @wireless_data: Instance data managed by the core of wireless extensions
1355 *
1356 * @netdev_ops: Includes several pointers to callbacks,
1357 * if one wants to override the ndo_*() functions
1358 * @ethtool_ops: Management operations
1359 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1360 * of Layer 2 headers.
1361 *
1362 * @flags: Interface flags (a la BSD)
1363 * @priv_flags: Like 'flags' but invisible to userspace,
1364 * see if.h for the definitions
1365 * @gflags: Global flags ( kept as legacy )
1366 * @padded: How much padding added by alloc_netdev()
1367 * @operstate: RFC2863 operstate
1368 * @link_mode: Mapping policy to operstate
1369 * @if_port: Selectable AUI, TP, ...
1370 * @dma: DMA channel
1371 * @mtu: Interface MTU value
1372 * @type: Interface hardware type
1373 * @hard_header_len: Hardware header length
1374 *
1375 * @needed_headroom: Extra headroom the hardware may need, but not in all
1376 * cases can this be guaranteed
1377 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1378 * cases can this be guaranteed. Some cases also use
1379 * LL_MAX_HEADER instead to allocate the skb
1380 *
1381 * interface address info:
1382 *
1383 * @perm_addr: Permanent hw address
1384 * @addr_assign_type: Hw address assignment type
1385 * @addr_len: Hardware address length
1386 * @neigh_priv_len; Used in neigh_alloc(),
1387 * initialized only in atm/clip.c
1388 * @dev_id: Used to differentiate devices that share
1389 * the same link layer address
1390 * @dev_port: Used to differentiate devices that share
1391 * the same function
1392 * @addr_list_lock: XXX: need comments on this one
1393 * @uc_promisc: Counter, that indicates, that promiscuous mode
1394 * has been enabled due to the need to listen to
1395 * additional unicast addresses in a device that
1396 * does not implement ndo_set_rx_mode()
1397 * @uc: unicast mac addresses
1398 * @mc: multicast mac addresses
1399 * @dev_addrs: list of device hw addresses
1400 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1401 * @promiscuity: Number of times, the NIC is told to work in
1402 * Promiscuous mode, if it becomes 0 the NIC will
1403 * exit from working in Promiscuous mode
1404 * @allmulti: Counter, enables or disables allmulticast mode
1405 *
1406 * @vlan_info: VLAN info
1407 * @dsa_ptr: dsa specific data
1408 * @tipc_ptr: TIPC specific data
1409 * @atalk_ptr: AppleTalk link
1410 * @ip_ptr: IPv4 specific data
1411 * @dn_ptr: DECnet specific data
1412 * @ip6_ptr: IPv6 specific data
1413 * @ax25_ptr: AX.25 specific data
1414 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1415 *
1416 * @last_rx: Time of last Rx
1417 * @dev_addr: Hw address (before bcast,
1418 * because most packets are unicast)
1419 *
1420 * @_rx: Array of RX queues
1421 * @num_rx_queues: Number of RX queues
1422 * allocated at register_netdev() time
1423 * @real_num_rx_queues: Number of RX queues currently active in device
1424 *
1425 * @rx_handler: handler for received packets
1426 * @rx_handler_data: XXX: need comments on this one
1427 * @ingress_queue: XXX: need comments on this one
1428 * @broadcast: hw bcast address
1429 *
1430 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1431 * indexed by RX queue number. Assigned by driver.
1432 * This must only be set if the ndo_rx_flow_steer
1433 * operation is defined
1434 * @index_hlist: Device index hash chain
1435 *
1436 * @_tx: Array of TX queues
1437 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1438 * @real_num_tx_queues: Number of TX queues currently active in device
1439 * @qdisc: Root qdisc from userspace point of view
1440 * @tx_queue_len: Max frames per queue allowed
1441 * @tx_global_lock: XXX: need comments on this one
1442 *
1443 * @xps_maps: XXX: need comments on this one
1444 *
1445 * @trans_start: Time (in jiffies) of last Tx
1446 * @watchdog_timeo: Represents the timeout that is used by
1447 * the watchdog ( see dev_watchdog() )
1448 * @watchdog_timer: List of timers
1449 *
1450 * @pcpu_refcnt: Number of references to this device
1451 * @todo_list: Delayed register/unregister
1452 * @link_watch_list: XXX: need comments on this one
1453 *
1454 * @reg_state: Register/unregister state machine
1455 * @dismantle: Device is going to be freed
1456 * @rtnl_link_state: This enum represents the phases of creating
1457 * a new link
1458 *
1459 * @destructor: Called from unregister,
1460 * can be used to call free_netdev
1461 * @npinfo: XXX: need comments on this one
1462 * @nd_net: Network namespace this network device is inside
1463 *
1464 * @ml_priv: Mid-layer private
1465 * @lstats: Loopback statistics
1466 * @tstats: Tunnel statistics
1467 * @dstats: Dummy statistics
1468 * @vstats: Virtual ethernet statistics
1469 *
1470 * @garp_port: GARP
1471 * @mrp_port: MRP
1472 *
1473 * @dev: Class/net/name entry
1474 * @sysfs_groups: Space for optional device, statistics and wireless
1475 * sysfs groups
1476 *
1477 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1478 * @rtnl_link_ops: Rtnl_link_ops
1479 *
1480 * @gso_max_size: Maximum size of generic segmentation offload
1481 * @gso_max_segs: Maximum number of segments that can be passed to the
1482 * NIC for GSO
1483 * @gso_min_segs: Minimum number of segments that can be passed to the
1484 * NIC for GSO
1485 *
1486 * @dcbnl_ops: Data Center Bridging netlink ops
1487 * @num_tc: Number of traffic classes in the net device
1488 * @tc_to_txq: XXX: need comments on this one
1489 * @prio_tc_map XXX: need comments on this one
1490 *
1491 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1492 *
1493 * @priomap: XXX: need comments on this one
1494 * @phydev: Physical device may attach itself
1495 * for hardware timestamping
1496 *
1497 * @qdisc_tx_busylock: XXX: need comments on this one
1498 *
1499 * @pm_qos_req: Power Management QoS object
1500 *
1501 * FIXME: cleanup struct net_device such that network protocol info
1502 * moves out.
1503 */
1504
1505struct net_device {
1506 char name[IFNAMSIZ];
1507 struct hlist_node name_hlist;
1508 char *ifalias;
1509 /*
1510 * I/O specific fields
1511 * FIXME: Merge these and struct ifmap into one
1512 */
1513 unsigned long mem_end;
1514 unsigned long mem_start;
1515 unsigned long base_addr;
1516 int irq;
1517
1518 atomic_t carrier_changes;
1519
1520 /*
1521 * Some hardware also needs these fields (state,dev_list,
1522 * napi_list,unreg_list,close_list) but they are not
1523 * part of the usual set specified in Space.c.
1524 */
1525
1526 unsigned long state;
1527
1528 struct list_head dev_list;
1529 struct list_head napi_list;
1530 struct list_head unreg_list;
1531 struct list_head close_list;
1532 struct list_head ptype_all;
1533 struct list_head ptype_specific;
1534
1535 struct {
1536 struct list_head upper;
1537 struct list_head lower;
1538 } adj_list;
1539
1540 struct {
1541 struct list_head upper;
1542 struct list_head lower;
1543 } all_adj_list;
1544
1545 netdev_features_t features;
1546 netdev_features_t hw_features;
1547 netdev_features_t wanted_features;
1548 netdev_features_t vlan_features;
1549 netdev_features_t hw_enc_features;
1550 netdev_features_t mpls_features;
1551
1552 int ifindex;
1553 int group;
1554
1555 struct net_device_stats stats;
1556
1557 atomic_long_t rx_dropped;
1558 atomic_long_t tx_dropped;
1559
1560#ifdef CONFIG_WIRELESS_EXT
1561 const struct iw_handler_def * wireless_handlers;
1562 struct iw_public_data * wireless_data;
1563#endif
1564 const struct net_device_ops *netdev_ops;
1565 const struct ethtool_ops *ethtool_ops;
1566#ifdef CONFIG_NET_SWITCHDEV
1567 const struct swdev_ops *swdev_ops;
1568#endif
1569
1570 const struct header_ops *header_ops;
1571
1572 unsigned int flags;
1573 unsigned int priv_flags;
1574
1575 unsigned short gflags;
1576 unsigned short padded;
1577
1578 unsigned char operstate;
1579 unsigned char link_mode;
1580
1581 unsigned char if_port;
1582 unsigned char dma;
1583
1584 unsigned int mtu;
1585 unsigned short type;
1586 unsigned short hard_header_len;
1587
1588 unsigned short needed_headroom;
1589 unsigned short needed_tailroom;
1590
1591 /* Interface address info. */
1592 unsigned char perm_addr[MAX_ADDR_LEN];
1593 unsigned char addr_assign_type;
1594 unsigned char addr_len;
1595 unsigned short neigh_priv_len;
1596 unsigned short dev_id;
1597 unsigned short dev_port;
1598 spinlock_t addr_list_lock;
1599 unsigned char name_assign_type;
1600 bool uc_promisc;
1601 struct netdev_hw_addr_list uc;
1602 struct netdev_hw_addr_list mc;
1603 struct netdev_hw_addr_list dev_addrs;
1604
1605#ifdef CONFIG_SYSFS
1606 struct kset *queues_kset;
1607#endif
1608 unsigned int promiscuity;
1609 unsigned int allmulti;
1610
1611
1612 /* Protocol specific pointers */
1613
1614#if IS_ENABLED(CONFIG_VLAN_8021Q)
1615 struct vlan_info __rcu *vlan_info;
1616#endif
1617#if IS_ENABLED(CONFIG_NET_DSA)
1618 struct dsa_switch_tree *dsa_ptr;
1619#endif
1620#if IS_ENABLED(CONFIG_TIPC)
1621 struct tipc_bearer __rcu *tipc_ptr;
1622#endif
1623 void *atalk_ptr;
1624 struct in_device __rcu *ip_ptr;
1625 struct dn_dev __rcu *dn_ptr;
1626 struct inet6_dev __rcu *ip6_ptr;
1627 void *ax25_ptr;
1628 struct wireless_dev *ieee80211_ptr;
1629 struct wpan_dev *ieee802154_ptr;
1630
1631/*
1632 * Cache lines mostly used on receive path (including eth_type_trans())
1633 */
1634 unsigned long last_rx;
1635
1636 /* Interface address info used in eth_type_trans() */
1637 unsigned char *dev_addr;
1638
1639
1640#ifdef CONFIG_SYSFS
1641 struct netdev_rx_queue *_rx;
1642
1643 unsigned int num_rx_queues;
1644 unsigned int real_num_rx_queues;
1645
1646#endif
1647
1648 unsigned long gro_flush_timeout;
1649 rx_handler_func_t __rcu *rx_handler;
1650 void __rcu *rx_handler_data;
1651
1652 struct netdev_queue __rcu *ingress_queue;
1653 unsigned char broadcast[MAX_ADDR_LEN];
1654#ifdef CONFIG_RFS_ACCEL
1655 struct cpu_rmap *rx_cpu_rmap;
1656#endif
1657 struct hlist_node index_hlist;
1658
1659/*
1660 * Cache lines mostly used on transmit path
1661 */
1662 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1663 unsigned int num_tx_queues;
1664 unsigned int real_num_tx_queues;
1665 struct Qdisc *qdisc;
1666 unsigned long tx_queue_len;
1667 spinlock_t tx_global_lock;
1668 int watchdog_timeo;
1669
1670#ifdef CONFIG_XPS
1671 struct xps_dev_maps __rcu *xps_maps;
1672#endif
1673
1674 /* These may be needed for future network-power-down code. */
1675
1676 /*
1677 * trans_start here is expensive for high speed devices on SMP,
1678 * please use netdev_queue->trans_start instead.
1679 */
1680 unsigned long trans_start;
1681
1682 struct timer_list watchdog_timer;
1683
1684 int __percpu *pcpu_refcnt;
1685 struct list_head todo_list;
1686
1687 struct list_head link_watch_list;
1688
1689 enum { NETREG_UNINITIALIZED=0,
1690 NETREG_REGISTERED, /* completed register_netdevice */
1691 NETREG_UNREGISTERING, /* called unregister_netdevice */
1692 NETREG_UNREGISTERED, /* completed unregister todo */
1693 NETREG_RELEASED, /* called free_netdev */
1694 NETREG_DUMMY, /* dummy device for NAPI poll */
1695 } reg_state:8;
1696
1697 bool dismantle;
1698
1699 enum {
1700 RTNL_LINK_INITIALIZED,
1701 RTNL_LINK_INITIALIZING,
1702 } rtnl_link_state:16;
1703
1704 void (*destructor)(struct net_device *dev);
1705
1706#ifdef CONFIG_NETPOLL
1707 struct netpoll_info __rcu *npinfo;
1708#endif
1709
1710 possible_net_t nd_net;
1711
1712 /* mid-layer private */
1713 union {
1714 void *ml_priv;
1715 struct pcpu_lstats __percpu *lstats;
1716 struct pcpu_sw_netstats __percpu *tstats;
1717 struct pcpu_dstats __percpu *dstats;
1718 struct pcpu_vstats __percpu *vstats;
1719 };
1720
1721 struct garp_port __rcu *garp_port;
1722 struct mrp_port __rcu *mrp_port;
1723
1724 struct device dev;
1725 const struct attribute_group *sysfs_groups[4];
1726 const struct attribute_group *sysfs_rx_queue_group;
1727
1728 const struct rtnl_link_ops *rtnl_link_ops;
1729
1730 /* for setting kernel sock attribute on TCP connection setup */
1731#define GSO_MAX_SIZE 65536
1732 unsigned int gso_max_size;
1733#define GSO_MAX_SEGS 65535
1734 u16 gso_max_segs;
1735 u16 gso_min_segs;
1736#ifdef CONFIG_DCB
1737 const struct dcbnl_rtnl_ops *dcbnl_ops;
1738#endif
1739 u8 num_tc;
1740 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1741 u8 prio_tc_map[TC_BITMASK + 1];
1742
1743#if IS_ENABLED(CONFIG_FCOE)
1744 unsigned int fcoe_ddp_xid;
1745#endif
1746#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1747 struct netprio_map __rcu *priomap;
1748#endif
1749 struct phy_device *phydev;
1750 struct lock_class_key *qdisc_tx_busylock;
1751};
1752#define to_net_dev(d) container_of(d, struct net_device, dev)
1753
1754#define NETDEV_ALIGN 32
1755
1756static inline
1757int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1758{
1759 return dev->prio_tc_map[prio & TC_BITMASK];
1760}
1761
1762static inline
1763int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1764{
1765 if (tc >= dev->num_tc)
1766 return -EINVAL;
1767
1768 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1769 return 0;
1770}
1771
1772static inline
1773void netdev_reset_tc(struct net_device *dev)
1774{
1775 dev->num_tc = 0;
1776 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1777 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1778}
1779
1780static inline
1781int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1782{
1783 if (tc >= dev->num_tc)
1784 return -EINVAL;
1785
1786 dev->tc_to_txq[tc].count = count;
1787 dev->tc_to_txq[tc].offset = offset;
1788 return 0;
1789}
1790
1791static inline
1792int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1793{
1794 if (num_tc > TC_MAX_QUEUE)
1795 return -EINVAL;
1796
1797 dev->num_tc = num_tc;
1798 return 0;
1799}
1800
1801static inline
1802int netdev_get_num_tc(struct net_device *dev)
1803{
1804 return dev->num_tc;
1805}
1806
1807static inline
1808struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1809 unsigned int index)
1810{
1811 return &dev->_tx[index];
1812}
1813
1814static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1815 const struct sk_buff *skb)
1816{
1817 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1818}
1819
1820static inline void netdev_for_each_tx_queue(struct net_device *dev,
1821 void (*f)(struct net_device *,
1822 struct netdev_queue *,
1823 void *),
1824 void *arg)
1825{
1826 unsigned int i;
1827
1828 for (i = 0; i < dev->num_tx_queues; i++)
1829 f(dev, &dev->_tx[i], arg);
1830}
1831
1832struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1833 struct sk_buff *skb,
1834 void *accel_priv);
1835
1836/*
1837 * Net namespace inlines
1838 */
1839static inline
1840struct net *dev_net(const struct net_device *dev)
1841{
1842 return read_pnet(&dev->nd_net);
1843}
1844
1845static inline
1846void dev_net_set(struct net_device *dev, struct net *net)
1847{
1848 write_pnet(&dev->nd_net, net);
1849}
1850
1851static inline bool netdev_uses_dsa(struct net_device *dev)
1852{
1853#if IS_ENABLED(CONFIG_NET_DSA)
1854 if (dev->dsa_ptr != NULL)
1855 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1856#endif
1857 return false;
1858}
1859
1860/**
1861 * netdev_priv - access network device private data
1862 * @dev: network device
1863 *
1864 * Get network device private data
1865 */
1866static inline void *netdev_priv(const struct net_device *dev)
1867{
1868 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1869}
1870
1871/* Set the sysfs physical device reference for the network logical device
1872 * if set prior to registration will cause a symlink during initialization.
1873 */
1874#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1875
1876/* Set the sysfs device type for the network logical device to allow
1877 * fine-grained identification of different network device types. For
1878 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1879 */
1880#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1881
1882/* Default NAPI poll() weight
1883 * Device drivers are strongly advised to not use bigger value
1884 */
1885#define NAPI_POLL_WEIGHT 64
1886
1887/**
1888 * netif_napi_add - initialize a napi context
1889 * @dev: network device
1890 * @napi: napi context
1891 * @poll: polling function
1892 * @weight: default weight
1893 *
1894 * netif_napi_add() must be used to initialize a napi context prior to calling
1895 * *any* of the other napi related functions.
1896 */
1897void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1898 int (*poll)(struct napi_struct *, int), int weight);
1899
1900/**
1901 * netif_napi_del - remove a napi context
1902 * @napi: napi context
1903 *
1904 * netif_napi_del() removes a napi context from the network device napi list
1905 */
1906void netif_napi_del(struct napi_struct *napi);
1907
1908struct napi_gro_cb {
1909 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1910 void *frag0;
1911
1912 /* Length of frag0. */
1913 unsigned int frag0_len;
1914
1915 /* This indicates where we are processing relative to skb->data. */
1916 int data_offset;
1917
1918 /* This is non-zero if the packet cannot be merged with the new skb. */
1919 u16 flush;
1920
1921 /* Save the IP ID here and check when we get to the transport layer */
1922 u16 flush_id;
1923
1924 /* Number of segments aggregated. */
1925 u16 count;
1926
1927 /* Start offset for remote checksum offload */
1928 u16 gro_remcsum_start;
1929
1930 /* jiffies when first packet was created/queued */
1931 unsigned long age;
1932
1933 /* Used in ipv6_gro_receive() and foo-over-udp */
1934 u16 proto;
1935
1936 /* This is non-zero if the packet may be of the same flow. */
1937 u8 same_flow:1;
1938
1939 /* Used in udp_gro_receive */
1940 u8 udp_mark:1;
1941
1942 /* GRO checksum is valid */
1943 u8 csum_valid:1;
1944
1945 /* Number of checksums via CHECKSUM_UNNECESSARY */
1946 u8 csum_cnt:3;
1947
1948 /* Free the skb? */
1949 u8 free:2;
1950#define NAPI_GRO_FREE 1
1951#define NAPI_GRO_FREE_STOLEN_HEAD 2
1952
1953 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1954 u8 is_ipv6:1;
1955
1956 /* 7 bit hole */
1957
1958 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1959 __wsum csum;
1960
1961 /* used in skb_gro_receive() slow path */
1962 struct sk_buff *last;
1963};
1964
1965#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1966
1967struct packet_type {
1968 __be16 type; /* This is really htons(ether_type). */
1969 struct net_device *dev; /* NULL is wildcarded here */
1970 int (*func) (struct sk_buff *,
1971 struct net_device *,
1972 struct packet_type *,
1973 struct net_device *);
1974 bool (*id_match)(struct packet_type *ptype,
1975 struct sock *sk);
1976 void *af_packet_priv;
1977 struct list_head list;
1978};
1979
1980struct offload_callbacks {
1981 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1982 netdev_features_t features);
1983 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1984 struct sk_buff *skb);
1985 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1986};
1987
1988struct packet_offload {
1989 __be16 type; /* This is really htons(ether_type). */
1990 struct offload_callbacks callbacks;
1991 struct list_head list;
1992};
1993
1994struct udp_offload;
1995
1996struct udp_offload_callbacks {
1997 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1998 struct sk_buff *skb,
1999 struct udp_offload *uoff);
2000 int (*gro_complete)(struct sk_buff *skb,
2001 int nhoff,
2002 struct udp_offload *uoff);
2003};
2004
2005struct udp_offload {
2006 __be16 port;
2007 u8 ipproto;
2008 struct udp_offload_callbacks callbacks;
2009};
2010
2011/* often modified stats are per cpu, other are shared (netdev->stats) */
2012struct pcpu_sw_netstats {
2013 u64 rx_packets;
2014 u64 rx_bytes;
2015 u64 tx_packets;
2016 u64 tx_bytes;
2017 struct u64_stats_sync syncp;
2018};
2019
2020#define netdev_alloc_pcpu_stats(type) \
2021({ \
2022 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2023 if (pcpu_stats) { \
2024 int i; \
2025 for_each_possible_cpu(i) { \
2026 typeof(type) *stat; \
2027 stat = per_cpu_ptr(pcpu_stats, i); \
2028 u64_stats_init(&stat->syncp); \
2029 } \
2030 } \
2031 pcpu_stats; \
2032})
2033
2034#include <linux/notifier.h>
2035
2036/* netdevice notifier chain. Please remember to update the rtnetlink
2037 * notification exclusion list in rtnetlink_event() when adding new
2038 * types.
2039 */
2040#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2041#define NETDEV_DOWN 0x0002
2042#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2043 detected a hardware crash and restarted
2044 - we can use this eg to kick tcp sessions
2045 once done */
2046#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2047#define NETDEV_REGISTER 0x0005
2048#define NETDEV_UNREGISTER 0x0006
2049#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2050#define NETDEV_CHANGEADDR 0x0008
2051#define NETDEV_GOING_DOWN 0x0009
2052#define NETDEV_CHANGENAME 0x000A
2053#define NETDEV_FEAT_CHANGE 0x000B
2054#define NETDEV_BONDING_FAILOVER 0x000C
2055#define NETDEV_PRE_UP 0x000D
2056#define NETDEV_PRE_TYPE_CHANGE 0x000E
2057#define NETDEV_POST_TYPE_CHANGE 0x000F
2058#define NETDEV_POST_INIT 0x0010
2059#define NETDEV_UNREGISTER_FINAL 0x0011
2060#define NETDEV_RELEASE 0x0012
2061#define NETDEV_NOTIFY_PEERS 0x0013
2062#define NETDEV_JOIN 0x0014
2063#define NETDEV_CHANGEUPPER 0x0015
2064#define NETDEV_RESEND_IGMP 0x0016
2065#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2066#define NETDEV_CHANGEINFODATA 0x0018
2067#define NETDEV_BONDING_INFO 0x0019
2068
2069int register_netdevice_notifier(struct notifier_block *nb);
2070int unregister_netdevice_notifier(struct notifier_block *nb);
2071
2072struct netdev_notifier_info {
2073 struct net_device *dev;
2074};
2075
2076struct netdev_notifier_change_info {
2077 struct netdev_notifier_info info; /* must be first */
2078 unsigned int flags_changed;
2079};
2080
2081static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2082 struct net_device *dev)
2083{
2084 info->dev = dev;
2085}
2086
2087static inline struct net_device *
2088netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2089{
2090 return info->dev;
2091}
2092
2093int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2094
2095
2096extern rwlock_t dev_base_lock; /* Device list lock */
2097
2098#define for_each_netdev(net, d) \
2099 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2100#define for_each_netdev_reverse(net, d) \
2101 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2102#define for_each_netdev_rcu(net, d) \
2103 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2104#define for_each_netdev_safe(net, d, n) \
2105 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2106#define for_each_netdev_continue(net, d) \
2107 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2108#define for_each_netdev_continue_rcu(net, d) \
2109 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2110#define for_each_netdev_in_bond_rcu(bond, slave) \
2111 for_each_netdev_rcu(&init_net, slave) \
2112 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2113#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2114
2115static inline struct net_device *next_net_device(struct net_device *dev)
2116{
2117 struct list_head *lh;
2118 struct net *net;
2119
2120 net = dev_net(dev);
2121 lh = dev->dev_list.next;
2122 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2123}
2124
2125static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2126{
2127 struct list_head *lh;
2128 struct net *net;
2129
2130 net = dev_net(dev);
2131 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2132 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2133}
2134
2135static inline struct net_device *first_net_device(struct net *net)
2136{
2137 return list_empty(&net->dev_base_head) ? NULL :
2138 net_device_entry(net->dev_base_head.next);
2139}
2140
2141static inline struct net_device *first_net_device_rcu(struct net *net)
2142{
2143 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2144
2145 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2146}
2147
2148int netdev_boot_setup_check(struct net_device *dev);
2149unsigned long netdev_boot_base(const char *prefix, int unit);
2150struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2151 const char *hwaddr);
2152struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2153struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2154void dev_add_pack(struct packet_type *pt);
2155void dev_remove_pack(struct packet_type *pt);
2156void __dev_remove_pack(struct packet_type *pt);
2157void dev_add_offload(struct packet_offload *po);
2158void dev_remove_offload(struct packet_offload *po);
2159
2160int dev_get_iflink(const struct net_device *dev);
2161struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2162 unsigned short mask);
2163struct net_device *dev_get_by_name(struct net *net, const char *name);
2164struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2165struct net_device *__dev_get_by_name(struct net *net, const char *name);
2166int dev_alloc_name(struct net_device *dev, const char *name);
2167int dev_open(struct net_device *dev);
2168int dev_close(struct net_device *dev);
2169int dev_close_many(struct list_head *head, bool unlink);
2170void dev_disable_lro(struct net_device *dev);
2171int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2172int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2173static inline int dev_queue_xmit(struct sk_buff *skb)
2174{
2175 return dev_queue_xmit_sk(skb->sk, skb);
2176}
2177int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2178int register_netdevice(struct net_device *dev);
2179void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2180void unregister_netdevice_many(struct list_head *head);
2181static inline void unregister_netdevice(struct net_device *dev)
2182{
2183 unregister_netdevice_queue(dev, NULL);
2184}
2185
2186int netdev_refcnt_read(const struct net_device *dev);
2187void free_netdev(struct net_device *dev);
2188void netdev_freemem(struct net_device *dev);
2189void synchronize_net(void);
2190int init_dummy_netdev(struct net_device *dev);
2191
2192DECLARE_PER_CPU(int, xmit_recursion);
2193static inline int dev_recursion_level(void)
2194{
2195 return this_cpu_read(xmit_recursion);
2196}
2197
2198struct net_device *dev_get_by_index(struct net *net, int ifindex);
2199struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2200struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2201int netdev_get_name(struct net *net, char *name, int ifindex);
2202int dev_restart(struct net_device *dev);
2203int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2204
2205static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2206{
2207 return NAPI_GRO_CB(skb)->data_offset;
2208}
2209
2210static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2211{
2212 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2213}
2214
2215static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2216{
2217 NAPI_GRO_CB(skb)->data_offset += len;
2218}
2219
2220static inline void *skb_gro_header_fast(struct sk_buff *skb,
2221 unsigned int offset)
2222{
2223 return NAPI_GRO_CB(skb)->frag0 + offset;
2224}
2225
2226static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2227{
2228 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2229}
2230
2231static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2232 unsigned int offset)
2233{
2234 if (!pskb_may_pull(skb, hlen))
2235 return NULL;
2236
2237 NAPI_GRO_CB(skb)->frag0 = NULL;
2238 NAPI_GRO_CB(skb)->frag0_len = 0;
2239 return skb->data + offset;
2240}
2241
2242static inline void *skb_gro_network_header(struct sk_buff *skb)
2243{
2244 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2245 skb_network_offset(skb);
2246}
2247
2248static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2249 const void *start, unsigned int len)
2250{
2251 if (NAPI_GRO_CB(skb)->csum_valid)
2252 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2253 csum_partial(start, len, 0));
2254}
2255
2256/* GRO checksum functions. These are logical equivalents of the normal
2257 * checksum functions (in skbuff.h) except that they operate on the GRO
2258 * offsets and fields in sk_buff.
2259 */
2260
2261__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2262
2263static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2264{
2265 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2266 skb_gro_offset(skb));
2267}
2268
2269static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2270 bool zero_okay,
2271 __sum16 check)
2272{
2273 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2274 skb_checksum_start_offset(skb) <
2275 skb_gro_offset(skb)) &&
2276 !skb_at_gro_remcsum_start(skb) &&
2277 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2278 (!zero_okay || check));
2279}
2280
2281static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2282 __wsum psum)
2283{
2284 if (NAPI_GRO_CB(skb)->csum_valid &&
2285 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2286 return 0;
2287
2288 NAPI_GRO_CB(skb)->csum = psum;
2289
2290 return __skb_gro_checksum_complete(skb);
2291}
2292
2293static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2294{
2295 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2296 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2297 NAPI_GRO_CB(skb)->csum_cnt--;
2298 } else {
2299 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2300 * verified a new top level checksum or an encapsulated one
2301 * during GRO. This saves work if we fallback to normal path.
2302 */
2303 __skb_incr_checksum_unnecessary(skb);
2304 }
2305}
2306
2307#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2308 compute_pseudo) \
2309({ \
2310 __sum16 __ret = 0; \
2311 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2312 __ret = __skb_gro_checksum_validate_complete(skb, \
2313 compute_pseudo(skb, proto)); \
2314 if (__ret) \
2315 __skb_mark_checksum_bad(skb); \
2316 else \
2317 skb_gro_incr_csum_unnecessary(skb); \
2318 __ret; \
2319})
2320
2321#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2322 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2323
2324#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2325 compute_pseudo) \
2326 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2327
2328#define skb_gro_checksum_simple_validate(skb) \
2329 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2330
2331static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2332{
2333 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2334 !NAPI_GRO_CB(skb)->csum_valid);
2335}
2336
2337static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2338 __sum16 check, __wsum pseudo)
2339{
2340 NAPI_GRO_CB(skb)->csum = ~pseudo;
2341 NAPI_GRO_CB(skb)->csum_valid = 1;
2342}
2343
2344#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2345do { \
2346 if (__skb_gro_checksum_convert_check(skb)) \
2347 __skb_gro_checksum_convert(skb, check, \
2348 compute_pseudo(skb, proto)); \
2349} while (0)
2350
2351struct gro_remcsum {
2352 int offset;
2353 __wsum delta;
2354};
2355
2356static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2357{
2358 grc->offset = 0;
2359 grc->delta = 0;
2360}
2361
2362static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2363 int start, int offset,
2364 struct gro_remcsum *grc,
2365 bool nopartial)
2366{
2367 __wsum delta;
2368
2369 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2370
2371 if (!nopartial) {
2372 NAPI_GRO_CB(skb)->gro_remcsum_start =
2373 ((unsigned char *)ptr + start) - skb->head;
2374 return;
2375 }
2376
2377 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2378
2379 /* Adjust skb->csum since we changed the packet */
2380 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2381
2382 grc->offset = (ptr + offset) - (void *)skb->head;
2383 grc->delta = delta;
2384}
2385
2386static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2387 struct gro_remcsum *grc)
2388{
2389 if (!grc->delta)
2390 return;
2391
2392 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2393}
2394
2395static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2396 unsigned short type,
2397 const void *daddr, const void *saddr,
2398 unsigned int len)
2399{
2400 if (!dev->header_ops || !dev->header_ops->create)
2401 return 0;
2402
2403 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2404}
2405
2406static inline int dev_parse_header(const struct sk_buff *skb,
2407 unsigned char *haddr)
2408{
2409 const struct net_device *dev = skb->dev;
2410
2411 if (!dev->header_ops || !dev->header_ops->parse)
2412 return 0;
2413 return dev->header_ops->parse(skb, haddr);
2414}
2415
2416typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2417int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2418static inline int unregister_gifconf(unsigned int family)
2419{
2420 return register_gifconf(family, NULL);
2421}
2422
2423#ifdef CONFIG_NET_FLOW_LIMIT
2424#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2425struct sd_flow_limit {
2426 u64 count;
2427 unsigned int num_buckets;
2428 unsigned int history_head;
2429 u16 history[FLOW_LIMIT_HISTORY];
2430 u8 buckets[];
2431};
2432
2433extern int netdev_flow_limit_table_len;
2434#endif /* CONFIG_NET_FLOW_LIMIT */
2435
2436/*
2437 * Incoming packets are placed on per-cpu queues
2438 */
2439struct softnet_data {
2440 struct list_head poll_list;
2441 struct sk_buff_head process_queue;
2442
2443 /* stats */
2444 unsigned int processed;
2445 unsigned int time_squeeze;
2446 unsigned int cpu_collision;
2447 unsigned int received_rps;
2448#ifdef CONFIG_RPS
2449 struct softnet_data *rps_ipi_list;
2450#endif
2451#ifdef CONFIG_NET_FLOW_LIMIT
2452 struct sd_flow_limit __rcu *flow_limit;
2453#endif
2454 struct Qdisc *output_queue;
2455 struct Qdisc **output_queue_tailp;
2456 struct sk_buff *completion_queue;
2457
2458#ifdef CONFIG_RPS
2459 /* Elements below can be accessed between CPUs for RPS */
2460 struct call_single_data csd ____cacheline_aligned_in_smp;
2461 struct softnet_data *rps_ipi_next;
2462 unsigned int cpu;
2463 unsigned int input_queue_head;
2464 unsigned int input_queue_tail;
2465#endif
2466 unsigned int dropped;
2467 struct sk_buff_head input_pkt_queue;
2468 struct napi_struct backlog;
2469
2470};
2471
2472static inline void input_queue_head_incr(struct softnet_data *sd)
2473{
2474#ifdef CONFIG_RPS
2475 sd->input_queue_head++;
2476#endif
2477}
2478
2479static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2480 unsigned int *qtail)
2481{
2482#ifdef CONFIG_RPS
2483 *qtail = ++sd->input_queue_tail;
2484#endif
2485}
2486
2487DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2488
2489void __netif_schedule(struct Qdisc *q);
2490void netif_schedule_queue(struct netdev_queue *txq);
2491
2492static inline void netif_tx_schedule_all(struct net_device *dev)
2493{
2494 unsigned int i;
2495
2496 for (i = 0; i < dev->num_tx_queues; i++)
2497 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2498}
2499
2500static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2501{
2502 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2503}
2504
2505/**
2506 * netif_start_queue - allow transmit
2507 * @dev: network device
2508 *
2509 * Allow upper layers to call the device hard_start_xmit routine.
2510 */
2511static inline void netif_start_queue(struct net_device *dev)
2512{
2513 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2514}
2515
2516static inline void netif_tx_start_all_queues(struct net_device *dev)
2517{
2518 unsigned int i;
2519
2520 for (i = 0; i < dev->num_tx_queues; i++) {
2521 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2522 netif_tx_start_queue(txq);
2523 }
2524}
2525
2526void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2527
2528/**
2529 * netif_wake_queue - restart transmit
2530 * @dev: network device
2531 *
2532 * Allow upper layers to call the device hard_start_xmit routine.
2533 * Used for flow control when transmit resources are available.
2534 */
2535static inline void netif_wake_queue(struct net_device *dev)
2536{
2537 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2538}
2539
2540static inline void netif_tx_wake_all_queues(struct net_device *dev)
2541{
2542 unsigned int i;
2543
2544 for (i = 0; i < dev->num_tx_queues; i++) {
2545 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2546 netif_tx_wake_queue(txq);
2547 }
2548}
2549
2550static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2551{
2552 if (WARN_ON(!dev_queue)) {
2553 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2554 return;
2555 }
2556 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2557}
2558
2559/**
2560 * netif_stop_queue - stop transmitted packets
2561 * @dev: network device
2562 *
2563 * Stop upper layers calling the device hard_start_xmit routine.
2564 * Used for flow control when transmit resources are unavailable.
2565 */
2566static inline void netif_stop_queue(struct net_device *dev)
2567{
2568 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2569}
2570
2571static inline void netif_tx_stop_all_queues(struct net_device *dev)
2572{
2573 unsigned int i;
2574
2575 for (i = 0; i < dev->num_tx_queues; i++) {
2576 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2577 netif_tx_stop_queue(txq);
2578 }
2579}
2580
2581static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2582{
2583 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2584}
2585
2586/**
2587 * netif_queue_stopped - test if transmit queue is flowblocked
2588 * @dev: network device
2589 *
2590 * Test if transmit queue on device is currently unable to send.
2591 */
2592static inline bool netif_queue_stopped(const struct net_device *dev)
2593{
2594 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2595}
2596
2597static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2598{
2599 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2600}
2601
2602static inline bool
2603netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2604{
2605 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2606}
2607
2608static inline bool
2609netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2610{
2611 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2612}
2613
2614/**
2615 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2616 * @dev_queue: pointer to transmit queue
2617 *
2618 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2619 * to give appropriate hint to the cpu.
2620 */
2621static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2622{
2623#ifdef CONFIG_BQL
2624 prefetchw(&dev_queue->dql.num_queued);
2625#endif
2626}
2627
2628/**
2629 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2630 * @dev_queue: pointer to transmit queue
2631 *
2632 * BQL enabled drivers might use this helper in their TX completion path,
2633 * to give appropriate hint to the cpu.
2634 */
2635static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2636{
2637#ifdef CONFIG_BQL
2638 prefetchw(&dev_queue->dql.limit);
2639#endif
2640}
2641
2642static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2643 unsigned int bytes)
2644{
2645#ifdef CONFIG_BQL
2646 dql_queued(&dev_queue->dql, bytes);
2647
2648 if (likely(dql_avail(&dev_queue->dql) >= 0))
2649 return;
2650
2651 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2652
2653 /*
2654 * The XOFF flag must be set before checking the dql_avail below,
2655 * because in netdev_tx_completed_queue we update the dql_completed
2656 * before checking the XOFF flag.
2657 */
2658 smp_mb();
2659
2660 /* check again in case another CPU has just made room avail */
2661 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2662 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2663#endif
2664}
2665
2666/**
2667 * netdev_sent_queue - report the number of bytes queued to hardware
2668 * @dev: network device
2669 * @bytes: number of bytes queued to the hardware device queue
2670 *
2671 * Report the number of bytes queued for sending/completion to the network
2672 * device hardware queue. @bytes should be a good approximation and should
2673 * exactly match netdev_completed_queue() @bytes
2674 */
2675static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2676{
2677 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2678}
2679
2680static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2681 unsigned int pkts, unsigned int bytes)
2682{
2683#ifdef CONFIG_BQL
2684 if (unlikely(!bytes))
2685 return;
2686
2687 dql_completed(&dev_queue->dql, bytes);
2688
2689 /*
2690 * Without the memory barrier there is a small possiblity that
2691 * netdev_tx_sent_queue will miss the update and cause the queue to
2692 * be stopped forever
2693 */
2694 smp_mb();
2695
2696 if (dql_avail(&dev_queue->dql) < 0)
2697 return;
2698
2699 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2700 netif_schedule_queue(dev_queue);
2701#endif
2702}
2703
2704/**
2705 * netdev_completed_queue - report bytes and packets completed by device
2706 * @dev: network device
2707 * @pkts: actual number of packets sent over the medium
2708 * @bytes: actual number of bytes sent over the medium
2709 *
2710 * Report the number of bytes and packets transmitted by the network device
2711 * hardware queue over the physical medium, @bytes must exactly match the
2712 * @bytes amount passed to netdev_sent_queue()
2713 */
2714static inline void netdev_completed_queue(struct net_device *dev,
2715 unsigned int pkts, unsigned int bytes)
2716{
2717 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2718}
2719
2720static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2721{
2722#ifdef CONFIG_BQL
2723 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2724 dql_reset(&q->dql);
2725#endif
2726}
2727
2728/**
2729 * netdev_reset_queue - reset the packets and bytes count of a network device
2730 * @dev_queue: network device
2731 *
2732 * Reset the bytes and packet count of a network device and clear the
2733 * software flow control OFF bit for this network device
2734 */
2735static inline void netdev_reset_queue(struct net_device *dev_queue)
2736{
2737 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2738}
2739
2740/**
2741 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2742 * @dev: network device
2743 * @queue_index: given tx queue index
2744 *
2745 * Returns 0 if given tx queue index >= number of device tx queues,
2746 * otherwise returns the originally passed tx queue index.
2747 */
2748static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2749{
2750 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2751 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2752 dev->name, queue_index,
2753 dev->real_num_tx_queues);
2754 return 0;
2755 }
2756
2757 return queue_index;
2758}
2759
2760/**
2761 * netif_running - test if up
2762 * @dev: network device
2763 *
2764 * Test if the device has been brought up.
2765 */
2766static inline bool netif_running(const struct net_device *dev)
2767{
2768 return test_bit(__LINK_STATE_START, &dev->state);
2769}
2770
2771/*
2772 * Routines to manage the subqueues on a device. We only need start
2773 * stop, and a check if it's stopped. All other device management is
2774 * done at the overall netdevice level.
2775 * Also test the device if we're multiqueue.
2776 */
2777
2778/**
2779 * netif_start_subqueue - allow sending packets on subqueue
2780 * @dev: network device
2781 * @queue_index: sub queue index
2782 *
2783 * Start individual transmit queue of a device with multiple transmit queues.
2784 */
2785static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2786{
2787 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2788
2789 netif_tx_start_queue(txq);
2790}
2791
2792/**
2793 * netif_stop_subqueue - stop sending packets on subqueue
2794 * @dev: network device
2795 * @queue_index: sub queue index
2796 *
2797 * Stop individual transmit queue of a device with multiple transmit queues.
2798 */
2799static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2800{
2801 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2802 netif_tx_stop_queue(txq);
2803}
2804
2805/**
2806 * netif_subqueue_stopped - test status of subqueue
2807 * @dev: network device
2808 * @queue_index: sub queue index
2809 *
2810 * Check individual transmit queue of a device with multiple transmit queues.
2811 */
2812static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2813 u16 queue_index)
2814{
2815 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2816
2817 return netif_tx_queue_stopped(txq);
2818}
2819
2820static inline bool netif_subqueue_stopped(const struct net_device *dev,
2821 struct sk_buff *skb)
2822{
2823 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2824}
2825
2826void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2827
2828#ifdef CONFIG_XPS
2829int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2830 u16 index);
2831#else
2832static inline int netif_set_xps_queue(struct net_device *dev,
2833 const struct cpumask *mask,
2834 u16 index)
2835{
2836 return 0;
2837}
2838#endif
2839
2840/*
2841 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2842 * as a distribution range limit for the returned value.
2843 */
2844static inline u16 skb_tx_hash(const struct net_device *dev,
2845 struct sk_buff *skb)
2846{
2847 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2848}
2849
2850/**
2851 * netif_is_multiqueue - test if device has multiple transmit queues
2852 * @dev: network device
2853 *
2854 * Check if device has multiple transmit queues
2855 */
2856static inline bool netif_is_multiqueue(const struct net_device *dev)
2857{
2858 return dev->num_tx_queues > 1;
2859}
2860
2861int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2862
2863#ifdef CONFIG_SYSFS
2864int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2865#else
2866static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2867 unsigned int rxq)
2868{
2869 return 0;
2870}
2871#endif
2872
2873#ifdef CONFIG_SYSFS
2874static inline unsigned int get_netdev_rx_queue_index(
2875 struct netdev_rx_queue *queue)
2876{
2877 struct net_device *dev = queue->dev;
2878 int index = queue - dev->_rx;
2879
2880 BUG_ON(index >= dev->num_rx_queues);
2881 return index;
2882}
2883#endif
2884
2885#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2886int netif_get_num_default_rss_queues(void);
2887
2888enum skb_free_reason {
2889 SKB_REASON_CONSUMED,
2890 SKB_REASON_DROPPED,
2891};
2892
2893void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2894void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2895
2896/*
2897 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2898 * interrupt context or with hardware interrupts being disabled.
2899 * (in_irq() || irqs_disabled())
2900 *
2901 * We provide four helpers that can be used in following contexts :
2902 *
2903 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2904 * replacing kfree_skb(skb)
2905 *
2906 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2907 * Typically used in place of consume_skb(skb) in TX completion path
2908 *
2909 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2910 * replacing kfree_skb(skb)
2911 *
2912 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2913 * and consumed a packet. Used in place of consume_skb(skb)
2914 */
2915static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2916{
2917 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2918}
2919
2920static inline void dev_consume_skb_irq(struct sk_buff *skb)
2921{
2922 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2923}
2924
2925static inline void dev_kfree_skb_any(struct sk_buff *skb)
2926{
2927 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2928}
2929
2930static inline void dev_consume_skb_any(struct sk_buff *skb)
2931{
2932 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2933}
2934
2935int netif_rx(struct sk_buff *skb);
2936int netif_rx_ni(struct sk_buff *skb);
2937int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
2938static inline int netif_receive_skb(struct sk_buff *skb)
2939{
2940 return netif_receive_skb_sk(skb->sk, skb);
2941}
2942gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2943void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2944struct sk_buff *napi_get_frags(struct napi_struct *napi);
2945gro_result_t napi_gro_frags(struct napi_struct *napi);
2946struct packet_offload *gro_find_receive_by_type(__be16 type);
2947struct packet_offload *gro_find_complete_by_type(__be16 type);
2948
2949static inline void napi_free_frags(struct napi_struct *napi)
2950{
2951 kfree_skb(napi->skb);
2952 napi->skb = NULL;
2953}
2954
2955int netdev_rx_handler_register(struct net_device *dev,
2956 rx_handler_func_t *rx_handler,
2957 void *rx_handler_data);
2958void netdev_rx_handler_unregister(struct net_device *dev);
2959
2960bool dev_valid_name(const char *name);
2961int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2962int dev_ethtool(struct net *net, struct ifreq *);
2963unsigned int dev_get_flags(const struct net_device *);
2964int __dev_change_flags(struct net_device *, unsigned int flags);
2965int dev_change_flags(struct net_device *, unsigned int);
2966void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2967 unsigned int gchanges);
2968int dev_change_name(struct net_device *, const char *);
2969int dev_set_alias(struct net_device *, const char *, size_t);
2970int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2971int dev_set_mtu(struct net_device *, int);
2972void dev_set_group(struct net_device *, int);
2973int dev_set_mac_address(struct net_device *, struct sockaddr *);
2974int dev_change_carrier(struct net_device *, bool new_carrier);
2975int dev_get_phys_port_id(struct net_device *dev,
2976 struct netdev_phys_item_id *ppid);
2977int dev_get_phys_port_name(struct net_device *dev,
2978 char *name, size_t len);
2979struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2980struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2981 struct netdev_queue *txq, int *ret);
2982int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2983int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2984bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2985
2986extern int netdev_budget;
2987
2988/* Called by rtnetlink.c:rtnl_unlock() */
2989void netdev_run_todo(void);
2990
2991/**
2992 * dev_put - release reference to device
2993 * @dev: network device
2994 *
2995 * Release reference to device to allow it to be freed.
2996 */
2997static inline void dev_put(struct net_device *dev)
2998{
2999 this_cpu_dec(*dev->pcpu_refcnt);
3000}
3001
3002/**
3003 * dev_hold - get reference to device
3004 * @dev: network device
3005 *
3006 * Hold reference to device to keep it from being freed.
3007 */
3008static inline void dev_hold(struct net_device *dev)
3009{
3010 this_cpu_inc(*dev->pcpu_refcnt);
3011}
3012
3013/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3014 * and _off may be called from IRQ context, but it is caller
3015 * who is responsible for serialization of these calls.
3016 *
3017 * The name carrier is inappropriate, these functions should really be
3018 * called netif_lowerlayer_*() because they represent the state of any
3019 * kind of lower layer not just hardware media.
3020 */
3021
3022void linkwatch_init_dev(struct net_device *dev);
3023void linkwatch_fire_event(struct net_device *dev);
3024void linkwatch_forget_dev(struct net_device *dev);
3025
3026/**
3027 * netif_carrier_ok - test if carrier present
3028 * @dev: network device
3029 *
3030 * Check if carrier is present on device
3031 */
3032static inline bool netif_carrier_ok(const struct net_device *dev)
3033{
3034 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3035}
3036
3037unsigned long dev_trans_start(struct net_device *dev);
3038
3039void __netdev_watchdog_up(struct net_device *dev);
3040
3041void netif_carrier_on(struct net_device *dev);
3042
3043void netif_carrier_off(struct net_device *dev);
3044
3045/**
3046 * netif_dormant_on - mark device as dormant.
3047 * @dev: network device
3048 *
3049 * Mark device as dormant (as per RFC2863).
3050 *
3051 * The dormant state indicates that the relevant interface is not
3052 * actually in a condition to pass packets (i.e., it is not 'up') but is
3053 * in a "pending" state, waiting for some external event. For "on-
3054 * demand" interfaces, this new state identifies the situation where the
3055 * interface is waiting for events to place it in the up state.
3056 *
3057 */
3058static inline void netif_dormant_on(struct net_device *dev)
3059{
3060 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3061 linkwatch_fire_event(dev);
3062}
3063
3064/**
3065 * netif_dormant_off - set device as not dormant.
3066 * @dev: network device
3067 *
3068 * Device is not in dormant state.
3069 */
3070static inline void netif_dormant_off(struct net_device *dev)
3071{
3072 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3073 linkwatch_fire_event(dev);
3074}
3075
3076/**
3077 * netif_dormant - test if carrier present
3078 * @dev: network device
3079 *
3080 * Check if carrier is present on device
3081 */
3082static inline bool netif_dormant(const struct net_device *dev)
3083{
3084 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3085}
3086
3087
3088/**
3089 * netif_oper_up - test if device is operational
3090 * @dev: network device
3091 *
3092 * Check if carrier is operational
3093 */
3094static inline bool netif_oper_up(const struct net_device *dev)
3095{
3096 return (dev->operstate == IF_OPER_UP ||
3097 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3098}
3099
3100/**
3101 * netif_device_present - is device available or removed
3102 * @dev: network device
3103 *
3104 * Check if device has not been removed from system.
3105 */
3106static inline bool netif_device_present(struct net_device *dev)
3107{
3108 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3109}
3110
3111void netif_device_detach(struct net_device *dev);
3112
3113void netif_device_attach(struct net_device *dev);
3114
3115/*
3116 * Network interface message level settings
3117 */
3118
3119enum {
3120 NETIF_MSG_DRV = 0x0001,
3121 NETIF_MSG_PROBE = 0x0002,
3122 NETIF_MSG_LINK = 0x0004,
3123 NETIF_MSG_TIMER = 0x0008,
3124 NETIF_MSG_IFDOWN = 0x0010,
3125 NETIF_MSG_IFUP = 0x0020,
3126 NETIF_MSG_RX_ERR = 0x0040,
3127 NETIF_MSG_TX_ERR = 0x0080,
3128 NETIF_MSG_TX_QUEUED = 0x0100,
3129 NETIF_MSG_INTR = 0x0200,
3130 NETIF_MSG_TX_DONE = 0x0400,
3131 NETIF_MSG_RX_STATUS = 0x0800,
3132 NETIF_MSG_PKTDATA = 0x1000,
3133 NETIF_MSG_HW = 0x2000,
3134 NETIF_MSG_WOL = 0x4000,
3135};
3136
3137#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3138#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3139#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3140#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3141#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3142#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3143#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3144#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3145#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3146#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3147#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3148#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3149#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3150#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3151#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3152
3153static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3154{
3155 /* use default */
3156 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3157 return default_msg_enable_bits;
3158 if (debug_value == 0) /* no output */
3159 return 0;
3160 /* set low N bits */
3161 return (1 << debug_value) - 1;
3162}
3163
3164static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3165{
3166 spin_lock(&txq->_xmit_lock);
3167 txq->xmit_lock_owner = cpu;
3168}
3169
3170static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3171{
3172 spin_lock_bh(&txq->_xmit_lock);
3173 txq->xmit_lock_owner = smp_processor_id();
3174}
3175
3176static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3177{
3178 bool ok = spin_trylock(&txq->_xmit_lock);
3179 if (likely(ok))
3180 txq->xmit_lock_owner = smp_processor_id();
3181 return ok;
3182}
3183
3184static inline void __netif_tx_unlock(struct netdev_queue *txq)
3185{
3186 txq->xmit_lock_owner = -1;
3187 spin_unlock(&txq->_xmit_lock);
3188}
3189
3190static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3191{
3192 txq->xmit_lock_owner = -1;
3193 spin_unlock_bh(&txq->_xmit_lock);
3194}
3195
3196static inline void txq_trans_update(struct netdev_queue *txq)
3197{
3198 if (txq->xmit_lock_owner != -1)
3199 txq->trans_start = jiffies;
3200}
3201
3202/**
3203 * netif_tx_lock - grab network device transmit lock
3204 * @dev: network device
3205 *
3206 * Get network device transmit lock
3207 */
3208static inline void netif_tx_lock(struct net_device *dev)
3209{
3210 unsigned int i;
3211 int cpu;
3212
3213 spin_lock(&dev->tx_global_lock);
3214 cpu = smp_processor_id();
3215 for (i = 0; i < dev->num_tx_queues; i++) {
3216 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3217
3218 /* We are the only thread of execution doing a
3219 * freeze, but we have to grab the _xmit_lock in
3220 * order to synchronize with threads which are in
3221 * the ->hard_start_xmit() handler and already
3222 * checked the frozen bit.
3223 */
3224 __netif_tx_lock(txq, cpu);
3225 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3226 __netif_tx_unlock(txq);
3227 }
3228}
3229
3230static inline void netif_tx_lock_bh(struct net_device *dev)
3231{
3232 local_bh_disable();
3233 netif_tx_lock(dev);
3234}
3235
3236static inline void netif_tx_unlock(struct net_device *dev)
3237{
3238 unsigned int i;
3239
3240 for (i = 0; i < dev->num_tx_queues; i++) {
3241 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3242
3243 /* No need to grab the _xmit_lock here. If the
3244 * queue is not stopped for another reason, we
3245 * force a schedule.
3246 */
3247 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3248 netif_schedule_queue(txq);
3249 }
3250 spin_unlock(&dev->tx_global_lock);
3251}
3252
3253static inline void netif_tx_unlock_bh(struct net_device *dev)
3254{
3255 netif_tx_unlock(dev);
3256 local_bh_enable();
3257}
3258
3259#define HARD_TX_LOCK(dev, txq, cpu) { \
3260 if ((dev->features & NETIF_F_LLTX) == 0) { \
3261 __netif_tx_lock(txq, cpu); \
3262 } \
3263}
3264
3265#define HARD_TX_TRYLOCK(dev, txq) \
3266 (((dev->features & NETIF_F_LLTX) == 0) ? \
3267 __netif_tx_trylock(txq) : \
3268 true )
3269
3270#define HARD_TX_UNLOCK(dev, txq) { \
3271 if ((dev->features & NETIF_F_LLTX) == 0) { \
3272 __netif_tx_unlock(txq); \
3273 } \
3274}
3275
3276static inline void netif_tx_disable(struct net_device *dev)
3277{
3278 unsigned int i;
3279 int cpu;
3280
3281 local_bh_disable();
3282 cpu = smp_processor_id();
3283 for (i = 0; i < dev->num_tx_queues; i++) {
3284 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3285
3286 __netif_tx_lock(txq, cpu);
3287 netif_tx_stop_queue(txq);
3288 __netif_tx_unlock(txq);
3289 }
3290 local_bh_enable();
3291}
3292
3293static inline void netif_addr_lock(struct net_device *dev)
3294{
3295 spin_lock(&dev->addr_list_lock);
3296}
3297
3298static inline void netif_addr_lock_nested(struct net_device *dev)
3299{
3300 int subclass = SINGLE_DEPTH_NESTING;
3301
3302 if (dev->netdev_ops->ndo_get_lock_subclass)
3303 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3304
3305 spin_lock_nested(&dev->addr_list_lock, subclass);
3306}
3307
3308static inline void netif_addr_lock_bh(struct net_device *dev)
3309{
3310 spin_lock_bh(&dev->addr_list_lock);
3311}
3312
3313static inline void netif_addr_unlock(struct net_device *dev)
3314{
3315 spin_unlock(&dev->addr_list_lock);
3316}
3317
3318static inline void netif_addr_unlock_bh(struct net_device *dev)
3319{
3320 spin_unlock_bh(&dev->addr_list_lock);
3321}
3322
3323/*
3324 * dev_addrs walker. Should be used only for read access. Call with
3325 * rcu_read_lock held.
3326 */
3327#define for_each_dev_addr(dev, ha) \
3328 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3329
3330/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3331
3332void ether_setup(struct net_device *dev);
3333
3334/* Support for loadable net-drivers */
3335struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3336 unsigned char name_assign_type,
3337 void (*setup)(struct net_device *),
3338 unsigned int txqs, unsigned int rxqs);
3339#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3340 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3341
3342#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3343 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3344 count)
3345
3346int register_netdev(struct net_device *dev);
3347void unregister_netdev(struct net_device *dev);
3348
3349/* General hardware address lists handling functions */
3350int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3351 struct netdev_hw_addr_list *from_list, int addr_len);
3352void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3353 struct netdev_hw_addr_list *from_list, int addr_len);
3354int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3355 struct net_device *dev,
3356 int (*sync)(struct net_device *, const unsigned char *),
3357 int (*unsync)(struct net_device *,
3358 const unsigned char *));
3359void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3360 struct net_device *dev,
3361 int (*unsync)(struct net_device *,
3362 const unsigned char *));
3363void __hw_addr_init(struct netdev_hw_addr_list *list);
3364
3365/* Functions used for device addresses handling */
3366int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3367 unsigned char addr_type);
3368int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3369 unsigned char addr_type);
3370void dev_addr_flush(struct net_device *dev);
3371int dev_addr_init(struct net_device *dev);
3372
3373/* Functions used for unicast addresses handling */
3374int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3375int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3376int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3377int dev_uc_sync(struct net_device *to, struct net_device *from);
3378int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3379void dev_uc_unsync(struct net_device *to, struct net_device *from);
3380void dev_uc_flush(struct net_device *dev);
3381void dev_uc_init(struct net_device *dev);
3382
3383/**
3384 * __dev_uc_sync - Synchonize device's unicast list
3385 * @dev: device to sync
3386 * @sync: function to call if address should be added
3387 * @unsync: function to call if address should be removed
3388 *
3389 * Add newly added addresses to the interface, and release
3390 * addresses that have been deleted.
3391 **/
3392static inline int __dev_uc_sync(struct net_device *dev,
3393 int (*sync)(struct net_device *,
3394 const unsigned char *),
3395 int (*unsync)(struct net_device *,
3396 const unsigned char *))
3397{
3398 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3399}
3400
3401/**
3402 * __dev_uc_unsync - Remove synchronized addresses from device
3403 * @dev: device to sync
3404 * @unsync: function to call if address should be removed
3405 *
3406 * Remove all addresses that were added to the device by dev_uc_sync().
3407 **/
3408static inline void __dev_uc_unsync(struct net_device *dev,
3409 int (*unsync)(struct net_device *,
3410 const unsigned char *))
3411{
3412 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3413}
3414
3415/* Functions used for multicast addresses handling */
3416int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3417int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3418int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3419int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3420int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3421int dev_mc_sync(struct net_device *to, struct net_device *from);
3422int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3423void dev_mc_unsync(struct net_device *to, struct net_device *from);
3424void dev_mc_flush(struct net_device *dev);
3425void dev_mc_init(struct net_device *dev);
3426
3427/**
3428 * __dev_mc_sync - Synchonize device's multicast list
3429 * @dev: device to sync
3430 * @sync: function to call if address should be added
3431 * @unsync: function to call if address should be removed
3432 *
3433 * Add newly added addresses to the interface, and release
3434 * addresses that have been deleted.
3435 **/
3436static inline int __dev_mc_sync(struct net_device *dev,
3437 int (*sync)(struct net_device *,
3438 const unsigned char *),
3439 int (*unsync)(struct net_device *,
3440 const unsigned char *))
3441{
3442 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3443}
3444
3445/**
3446 * __dev_mc_unsync - Remove synchronized addresses from device
3447 * @dev: device to sync
3448 * @unsync: function to call if address should be removed
3449 *
3450 * Remove all addresses that were added to the device by dev_mc_sync().
3451 **/
3452static inline void __dev_mc_unsync(struct net_device *dev,
3453 int (*unsync)(struct net_device *,
3454 const unsigned char *))
3455{
3456 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3457}
3458
3459/* Functions used for secondary unicast and multicast support */
3460void dev_set_rx_mode(struct net_device *dev);
3461void __dev_set_rx_mode(struct net_device *dev);
3462int dev_set_promiscuity(struct net_device *dev, int inc);
3463int dev_set_allmulti(struct net_device *dev, int inc);
3464void netdev_state_change(struct net_device *dev);
3465void netdev_notify_peers(struct net_device *dev);
3466void netdev_features_change(struct net_device *dev);
3467/* Load a device via the kmod */
3468void dev_load(struct net *net, const char *name);
3469struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3470 struct rtnl_link_stats64 *storage);
3471void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3472 const struct net_device_stats *netdev_stats);
3473
3474extern int netdev_max_backlog;
3475extern int netdev_tstamp_prequeue;
3476extern int weight_p;
3477extern int bpf_jit_enable;
3478
3479bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3480struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3481 struct list_head **iter);
3482struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3483 struct list_head **iter);
3484
3485/* iterate through upper list, must be called under RCU read lock */
3486#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3487 for (iter = &(dev)->adj_list.upper, \
3488 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3489 updev; \
3490 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3491
3492/* iterate through upper list, must be called under RCU read lock */
3493#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3494 for (iter = &(dev)->all_adj_list.upper, \
3495 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3496 updev; \
3497 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3498
3499void *netdev_lower_get_next_private(struct net_device *dev,
3500 struct list_head **iter);
3501void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3502 struct list_head **iter);
3503
3504#define netdev_for_each_lower_private(dev, priv, iter) \
3505 for (iter = (dev)->adj_list.lower.next, \
3506 priv = netdev_lower_get_next_private(dev, &(iter)); \
3507 priv; \
3508 priv = netdev_lower_get_next_private(dev, &(iter)))
3509
3510#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3511 for (iter = &(dev)->adj_list.lower, \
3512 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3513 priv; \
3514 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3515
3516void *netdev_lower_get_next(struct net_device *dev,
3517 struct list_head **iter);
3518#define netdev_for_each_lower_dev(dev, ldev, iter) \
3519 for (iter = &(dev)->adj_list.lower, \
3520 ldev = netdev_lower_get_next(dev, &(iter)); \
3521 ldev; \
3522 ldev = netdev_lower_get_next(dev, &(iter)))
3523
3524void *netdev_adjacent_get_private(struct list_head *adj_list);
3525void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3526struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3527struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3528int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3529int netdev_master_upper_dev_link(struct net_device *dev,
3530 struct net_device *upper_dev);
3531int netdev_master_upper_dev_link_private(struct net_device *dev,
3532 struct net_device *upper_dev,
3533 void *private);
3534void netdev_upper_dev_unlink(struct net_device *dev,
3535 struct net_device *upper_dev);
3536void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3537void *netdev_lower_dev_get_private(struct net_device *dev,
3538 struct net_device *lower_dev);
3539
3540/* RSS keys are 40 or 52 bytes long */
3541#define NETDEV_RSS_KEY_LEN 52
3542extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3543void netdev_rss_key_fill(void *buffer, size_t len);
3544
3545int dev_get_nest_level(struct net_device *dev,
3546 bool (*type_check)(struct net_device *dev));
3547int skb_checksum_help(struct sk_buff *skb);
3548struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3549 netdev_features_t features, bool tx_path);
3550struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3551 netdev_features_t features);
3552
3553struct netdev_bonding_info {
3554 ifslave slave;
3555 ifbond master;
3556};
3557
3558struct netdev_notifier_bonding_info {
3559 struct netdev_notifier_info info; /* must be first */
3560 struct netdev_bonding_info bonding_info;
3561};
3562
3563void netdev_bonding_info_change(struct net_device *dev,
3564 struct netdev_bonding_info *bonding_info);
3565
3566static inline
3567struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3568{
3569 return __skb_gso_segment(skb, features, true);
3570}
3571__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3572
3573static inline bool can_checksum_protocol(netdev_features_t features,
3574 __be16 protocol)
3575{
3576 return ((features & NETIF_F_GEN_CSUM) ||
3577 ((features & NETIF_F_V4_CSUM) &&
3578 protocol == htons(ETH_P_IP)) ||
3579 ((features & NETIF_F_V6_CSUM) &&
3580 protocol == htons(ETH_P_IPV6)) ||
3581 ((features & NETIF_F_FCOE_CRC) &&
3582 protocol == htons(ETH_P_FCOE)));
3583}
3584
3585#ifdef CONFIG_BUG
3586void netdev_rx_csum_fault(struct net_device *dev);
3587#else
3588static inline void netdev_rx_csum_fault(struct net_device *dev)
3589{
3590}
3591#endif
3592/* rx skb timestamps */
3593void net_enable_timestamp(void);
3594void net_disable_timestamp(void);
3595
3596#ifdef CONFIG_PROC_FS
3597int __init dev_proc_init(void);
3598#else
3599#define dev_proc_init() 0
3600#endif
3601
3602static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3603 struct sk_buff *skb, struct net_device *dev,
3604 bool more)
3605{
3606 skb->xmit_more = more ? 1 : 0;
3607 return ops->ndo_start_xmit(skb, dev);
3608}
3609
3610static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3611 struct netdev_queue *txq, bool more)
3612{
3613 const struct net_device_ops *ops = dev->netdev_ops;
3614 int rc;
3615
3616 rc = __netdev_start_xmit(ops, skb, dev, more);
3617 if (rc == NETDEV_TX_OK)
3618 txq_trans_update(txq);
3619
3620 return rc;
3621}
3622
3623int netdev_class_create_file_ns(struct class_attribute *class_attr,
3624 const void *ns);
3625void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3626 const void *ns);
3627
3628static inline int netdev_class_create_file(struct class_attribute *class_attr)
3629{
3630 return netdev_class_create_file_ns(class_attr, NULL);
3631}
3632
3633static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3634{
3635 netdev_class_remove_file_ns(class_attr, NULL);
3636}
3637
3638extern struct kobj_ns_type_operations net_ns_type_operations;
3639
3640const char *netdev_drivername(const struct net_device *dev);
3641
3642void linkwatch_run_queue(void);
3643
3644static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3645 netdev_features_t f2)
3646{
3647 if (f1 & NETIF_F_GEN_CSUM)
3648 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3649 if (f2 & NETIF_F_GEN_CSUM)
3650 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3651 f1 &= f2;
3652 if (f1 & NETIF_F_GEN_CSUM)
3653 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3654
3655 return f1;
3656}
3657
3658static inline netdev_features_t netdev_get_wanted_features(
3659 struct net_device *dev)
3660{
3661 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3662}
3663netdev_features_t netdev_increment_features(netdev_features_t all,
3664 netdev_features_t one, netdev_features_t mask);
3665
3666/* Allow TSO being used on stacked device :
3667 * Performing the GSO segmentation before last device
3668 * is a performance improvement.
3669 */
3670static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3671 netdev_features_t mask)
3672{
3673 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3674}
3675
3676int __netdev_update_features(struct net_device *dev);
3677void netdev_update_features(struct net_device *dev);
3678void netdev_change_features(struct net_device *dev);
3679
3680void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3681 struct net_device *dev);
3682
3683netdev_features_t passthru_features_check(struct sk_buff *skb,
3684 struct net_device *dev,
3685 netdev_features_t features);
3686netdev_features_t netif_skb_features(struct sk_buff *skb);
3687
3688static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3689{
3690 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3691
3692 /* check flags correspondence */
3693 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3694 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3695 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3696 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3697 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3701 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3702 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3703 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3704 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3705 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3706
3707 return (features & feature) == feature;
3708}
3709
3710static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3711{
3712 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3713 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3714}
3715
3716static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3717 netdev_features_t features)
3718{
3719 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3720 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3721 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3722}
3723
3724static inline void netif_set_gso_max_size(struct net_device *dev,
3725 unsigned int size)
3726{
3727 dev->gso_max_size = size;
3728}
3729
3730static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3731 int pulled_hlen, u16 mac_offset,
3732 int mac_len)
3733{
3734 skb->protocol = protocol;
3735 skb->encapsulation = 1;
3736 skb_push(skb, pulled_hlen);
3737 skb_reset_transport_header(skb);
3738 skb->mac_header = mac_offset;
3739 skb->network_header = skb->mac_header + mac_len;
3740 skb->mac_len = mac_len;
3741}
3742
3743static inline bool netif_is_macvlan(struct net_device *dev)
3744{
3745 return dev->priv_flags & IFF_MACVLAN;
3746}
3747
3748static inline bool netif_is_macvlan_port(struct net_device *dev)
3749{
3750 return dev->priv_flags & IFF_MACVLAN_PORT;
3751}
3752
3753static inline bool netif_is_ipvlan(struct net_device *dev)
3754{
3755 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3756}
3757
3758static inline bool netif_is_ipvlan_port(struct net_device *dev)
3759{
3760 return dev->priv_flags & IFF_IPVLAN_MASTER;
3761}
3762
3763static inline bool netif_is_bond_master(struct net_device *dev)
3764{
3765 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3766}
3767
3768static inline bool netif_is_bond_slave(struct net_device *dev)
3769{
3770 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3771}
3772
3773static inline bool netif_supports_nofcs(struct net_device *dev)
3774{
3775 return dev->priv_flags & IFF_SUPP_NOFCS;
3776}
3777
3778/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3779static inline void netif_keep_dst(struct net_device *dev)
3780{
3781 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3782}
3783
3784extern struct pernet_operations __net_initdata loopback_net_ops;
3785
3786/* Logging, debugging and troubleshooting/diagnostic helpers. */
3787
3788/* netdev_printk helpers, similar to dev_printk */
3789
3790static inline const char *netdev_name(const struct net_device *dev)
3791{
3792 if (!dev->name[0] || strchr(dev->name, '%'))
3793 return "(unnamed net_device)";
3794 return dev->name;
3795}
3796
3797static inline const char *netdev_reg_state(const struct net_device *dev)
3798{
3799 switch (dev->reg_state) {
3800 case NETREG_UNINITIALIZED: return " (uninitialized)";
3801 case NETREG_REGISTERED: return "";
3802 case NETREG_UNREGISTERING: return " (unregistering)";
3803 case NETREG_UNREGISTERED: return " (unregistered)";
3804 case NETREG_RELEASED: return " (released)";
3805 case NETREG_DUMMY: return " (dummy)";
3806 }
3807
3808 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3809 return " (unknown)";
3810}
3811
3812__printf(3, 4)
3813void netdev_printk(const char *level, const struct net_device *dev,
3814 const char *format, ...);
3815__printf(2, 3)
3816void netdev_emerg(const struct net_device *dev, const char *format, ...);
3817__printf(2, 3)
3818void netdev_alert(const struct net_device *dev, const char *format, ...);
3819__printf(2, 3)
3820void netdev_crit(const struct net_device *dev, const char *format, ...);
3821__printf(2, 3)
3822void netdev_err(const struct net_device *dev, const char *format, ...);
3823__printf(2, 3)
3824void netdev_warn(const struct net_device *dev, const char *format, ...);
3825__printf(2, 3)
3826void netdev_notice(const struct net_device *dev, const char *format, ...);
3827__printf(2, 3)
3828void netdev_info(const struct net_device *dev, const char *format, ...);
3829
3830#define MODULE_ALIAS_NETDEV(device) \
3831 MODULE_ALIAS("netdev-" device)
3832
3833#if defined(CONFIG_DYNAMIC_DEBUG)
3834#define netdev_dbg(__dev, format, args...) \
3835do { \
3836 dynamic_netdev_dbg(__dev, format, ##args); \
3837} while (0)
3838#elif defined(DEBUG)
3839#define netdev_dbg(__dev, format, args...) \
3840 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3841#else
3842#define netdev_dbg(__dev, format, args...) \
3843({ \
3844 if (0) \
3845 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3846})
3847#endif
3848
3849#if defined(VERBOSE_DEBUG)
3850#define netdev_vdbg netdev_dbg
3851#else
3852
3853#define netdev_vdbg(dev, format, args...) \
3854({ \
3855 if (0) \
3856 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3857 0; \
3858})
3859#endif
3860
3861/*
3862 * netdev_WARN() acts like dev_printk(), but with the key difference
3863 * of using a WARN/WARN_ON to get the message out, including the
3864 * file/line information and a backtrace.
3865 */
3866#define netdev_WARN(dev, format, args...) \
3867 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3868 netdev_reg_state(dev), ##args)
3869
3870/* netif printk helpers, similar to netdev_printk */
3871
3872#define netif_printk(priv, type, level, dev, fmt, args...) \
3873do { \
3874 if (netif_msg_##type(priv)) \
3875 netdev_printk(level, (dev), fmt, ##args); \
3876} while (0)
3877
3878#define netif_level(level, priv, type, dev, fmt, args...) \
3879do { \
3880 if (netif_msg_##type(priv)) \
3881 netdev_##level(dev, fmt, ##args); \
3882} while (0)
3883
3884#define netif_emerg(priv, type, dev, fmt, args...) \
3885 netif_level(emerg, priv, type, dev, fmt, ##args)
3886#define netif_alert(priv, type, dev, fmt, args...) \
3887 netif_level(alert, priv, type, dev, fmt, ##args)
3888#define netif_crit(priv, type, dev, fmt, args...) \
3889 netif_level(crit, priv, type, dev, fmt, ##args)
3890#define netif_err(priv, type, dev, fmt, args...) \
3891 netif_level(err, priv, type, dev, fmt, ##args)
3892#define netif_warn(priv, type, dev, fmt, args...) \
3893 netif_level(warn, priv, type, dev, fmt, ##args)
3894#define netif_notice(priv, type, dev, fmt, args...) \
3895 netif_level(notice, priv, type, dev, fmt, ##args)
3896#define netif_info(priv, type, dev, fmt, args...) \
3897 netif_level(info, priv, type, dev, fmt, ##args)
3898
3899#if defined(CONFIG_DYNAMIC_DEBUG)
3900#define netif_dbg(priv, type, netdev, format, args...) \
3901do { \
3902 if (netif_msg_##type(priv)) \
3903 dynamic_netdev_dbg(netdev, format, ##args); \
3904} while (0)
3905#elif defined(DEBUG)
3906#define netif_dbg(priv, type, dev, format, args...) \
3907 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3908#else
3909#define netif_dbg(priv, type, dev, format, args...) \
3910({ \
3911 if (0) \
3912 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3913 0; \
3914})
3915#endif
3916
3917#if defined(VERBOSE_DEBUG)
3918#define netif_vdbg netif_dbg
3919#else
3920#define netif_vdbg(priv, type, dev, format, args...) \
3921({ \
3922 if (0) \
3923 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3924 0; \
3925})
3926#endif
3927
3928/*
3929 * The list of packet types we will receive (as opposed to discard)
3930 * and the routines to invoke.
3931 *
3932 * Why 16. Because with 16 the only overlap we get on a hash of the
3933 * low nibble of the protocol value is RARP/SNAP/X.25.
3934 *
3935 * NOTE: That is no longer true with the addition of VLAN tags. Not
3936 * sure which should go first, but I bet it won't make much
3937 * difference if we are running VLANs. The good news is that
3938 * this protocol won't be in the list unless compiled in, so
3939 * the average user (w/out VLANs) will not be adversely affected.
3940 * --BLG
3941 *
3942 * 0800 IP
3943 * 8100 802.1Q VLAN
3944 * 0001 802.3
3945 * 0002 AX.25
3946 * 0004 802.2
3947 * 8035 RARP
3948 * 0005 SNAP
3949 * 0805 X.25
3950 * 0806 ARP
3951 * 8137 IPX
3952 * 0009 Localtalk
3953 * 86DD IPv6
3954 */
3955#define PTYPE_HASH_SIZE (16)
3956#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3957
3958#endif /* _LINUX_NETDEVICE_H */
This page took 0.044359 seconds and 5 git commands to generate.