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