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/timer.h>
29#include <linux/bug.h>
30#include <linux/delay.h>
31#include <linux/atomic.h>
32#include <linux/prefetch.h>
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
36#include <linux/percpu.h>
37#include <linux/rculist.h>
38#include <linux/workqueue.h>
39#include <linux/dynamic_queue_limits.h>
40
41#include <linux/ethtool.h>
42#include <net/net_namespace.h>
43#ifdef CONFIG_DCB
44#include <net/dcbnl.h>
45#endif
46#include <net/netprio_cgroup.h>
47#include <net/xdp.h>
48
49#include <linux/netdev_features.h>
50#include <linux/neighbour.h>
51#include <uapi/linux/netdevice.h>
52#include <uapi/linux/if_bonding.h>
53#include <uapi/linux/pkt_cls.h>
54#include <linux/hashtable.h>
55
56struct netpoll_info;
57struct device;
58struct phy_device;
59struct dsa_port;
60
61/* 802.11 specific */
62struct wireless_dev;
63/* 802.15.4 specific */
64struct wpan_dev;
65struct mpls_dev;
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
68struct bpf_prog;
69struct xdp_buff;
70
71void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
73
74/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
78/*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
92 * others are propagated to higher layers.
93 */
94
95/* qdisc ->enqueue() return codes. */
96#define NET_XMIT_SUCCESS 0x00
97#define NET_XMIT_DROP 0x01 /* skb dropped */
98#define NET_XMIT_CN 0x02 /* congestion notification */
99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107/* Driver transmit return codes */
108#define NETDEV_TX_MASK 0xf0
109
110enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114};
115typedef enum netdev_tx netdev_tx_t;
116
117/*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121static inline bool dev_xmit_complete(int rc)
122{
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133}
134
135/*
136 * Compute the worst-case header length according to the protocols
137 * used.
138 */
139
140#if defined(CONFIG_HYPERV_NET)
141# define LL_MAX_HEADER 128
142#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
143# if defined(CONFIG_MAC80211_MESH)
144# define LL_MAX_HEADER 128
145# else
146# define LL_MAX_HEADER 96
147# endif
148#else
149# define LL_MAX_HEADER 32
150#endif
151
152#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
154#define MAX_HEADER LL_MAX_HEADER
155#else
156#define MAX_HEADER (LL_MAX_HEADER + 48)
157#endif
158
159/*
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
162 */
163
164struct net_device_stats {
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
174 unsigned long collisions;
175 unsigned long rx_length_errors;
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188};
189
190
191#include <linux/cache.h>
192#include <linux/skbuff.h>
193
194#ifdef CONFIG_RPS
195#include <linux/static_key.h>
196extern struct static_key rps_needed;
197extern struct static_key rfs_needed;
198#endif
199
200struct neighbour;
201struct neigh_parms;
202struct sk_buff;
203
204struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
208#define NETDEV_HW_ADDR_T_LAN 1
209#define NETDEV_HW_ADDR_T_SAN 2
210#define NETDEV_HW_ADDR_T_SLAVE 3
211#define NETDEV_HW_ADDR_T_UNICAST 4
212#define NETDEV_HW_ADDR_T_MULTICAST 5
213 bool global_use;
214 int sync_cnt;
215 int refcount;
216 int synced;
217 struct rcu_head rcu_head;
218};
219
220struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223};
224
225#define netdev_hw_addr_list_count(l) ((l)->count)
226#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227#define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
229
230#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232#define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234
235#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
237#define netdev_for_each_mc_addr(ha, dev) \
238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
239
240struct hh_cache {
241 unsigned int hh_len;
242 seqlock_t hh_lock;
243
244 /* cached hardware header; allow for machine alignment needs. */
245#define HH_DATA_MOD 16
246#define HH_DATA_OFF(__len) \
247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
248#define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251};
252
253/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261#define LL_RESERVED_SPACE(dev) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265
266struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
269 const void *saddr, unsigned int len);
270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275 bool (*validate)(const char *ll_header, unsigned int len);
276};
277
278/* These flag bits are private to the generic network queueing
279 * layer; they may not be explicitly referenced by any other
280 * code.
281 */
282
283enum netdev_state_t {
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
286 __LINK_STATE_NOCARRIER,
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
289};
290
291
292/*
293 * This structure holds boot-time configured netdevice settings. They
294 * are then used in the device probing.
295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
302int __init netdev_boot_setup(char *str);
303
304/*
305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
311 * to the per-CPU poll_list, and whoever clears that bit
312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
318 unsigned int gro_count;
319 int (*poll)(struct napi_struct *, int);
320#ifdef CONFIG_NETPOLL
321 int poll_owner;
322#endif
323 struct net_device *dev;
324 struct sk_buff *gro_list;
325 struct sk_buff *skb;
326 struct hrtimer timer;
327 struct list_head dev_list;
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
330};
331
332enum {
333 NAPI_STATE_SCHED, /* Poll is scheduled */
334 NAPI_STATE_MISSED, /* reschedule a napi */
335 NAPI_STATE_DISABLE, /* Disable pending */
336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340};
341
342enum {
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
350};
351
352enum gro_result {
353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
358 GRO_CONSUMED,
359};
360typedef enum gro_result gro_result_t;
361
362/*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
384 * If the rx_handler consumed the skb in some way, it should return
385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
386 * the skb to be delivered in some other way.
387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
392 * If the rx_handler decides the skb should be ignored, it should return
393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
394 * are registered on exact device (ptype->dev == skb->dev).
395 *
396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408};
409typedef enum rx_handler_result rx_handler_result_t;
410typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
411
412void __napi_schedule(struct napi_struct *n);
413void __napi_schedule_irqoff(struct napi_struct *n);
414
415static inline bool napi_disable_pending(struct napi_struct *n)
416{
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418}
419
420bool napi_schedule_prep(struct napi_struct *n);
421
422/**
423 * napi_schedule - schedule NAPI poll
424 * @n: NAPI context
425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429static inline void napi_schedule(struct napi_struct *n)
430{
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433}
434
435/**
436 * napi_schedule_irqoff - schedule NAPI poll
437 * @n: NAPI context
438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441static inline void napi_schedule_irqoff(struct napi_struct *n)
442{
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445}
446
447/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
448static inline bool napi_reschedule(struct napi_struct *napi)
449{
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
452 return true;
453 }
454 return false;
455}
456
457bool napi_complete_done(struct napi_struct *n, int work_done);
458/**
459 * napi_complete - NAPI processing complete
460 * @n: NAPI context
461 *
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
464 * Return false if device should avoid rearming interrupts.
465 */
466static inline bool napi_complete(struct napi_struct *n)
467{
468 return napi_complete_done(n, 0);
469}
470
471/**
472 * napi_hash_del - remove a NAPI from global table
473 * @napi: NAPI context
474 *
475 * Warning: caller must observe RCU grace period
476 * before freeing memory containing @napi, if
477 * this function returns true.
478 * Note: core networking stack automatically calls it
479 * from netif_napi_del().
480 * Drivers might want to call this helper to combine all
481 * the needed RCU grace periods into a single one.
482 */
483bool napi_hash_del(struct napi_struct *napi);
484
485/**
486 * napi_disable - prevent NAPI from scheduling
487 * @n: NAPI context
488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
492void napi_disable(struct napi_struct *n);
493
494/**
495 * napi_enable - enable NAPI scheduling
496 * @n: NAPI context
497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501static inline void napi_enable(struct napi_struct *n)
502{
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
504 smp_mb__before_atomic();
505 clear_bit(NAPI_STATE_SCHED, &n->state);
506 clear_bit(NAPI_STATE_NPSVC, &n->state);
507}
508
509/**
510 * napi_synchronize - wait until NAPI is not running
511 * @n: NAPI context
512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517static inline void napi_synchronize(const struct napi_struct *n)
518{
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
524}
525
526enum netdev_queue_state_t {
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
529 __QUEUE_STATE_FROZEN,
530};
531
532#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
542/*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
551
552struct netdev_queue {
553/*
554 * read-mostly part
555 */
556 struct net_device *dev;
557 struct Qdisc __rcu *qdisc;
558 struct Qdisc *qdisc_sleeping;
559#ifdef CONFIG_SYSFS
560 struct kobject kobj;
561#endif
562#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564#endif
565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
571/*
572 * write-mostly part
573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
576 /*
577 * Time (in jiffies) of last Tx
578 */
579 unsigned long trans_start;
580
581 unsigned long state;
582
583#ifdef CONFIG_BQL
584 struct dql dql;
585#endif
586} ____cacheline_aligned_in_smp;
587
588static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 return q->numa_node;
592#else
593 return NUMA_NO_NODE;
594#endif
595}
596
597static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 q->numa_node = node;
601#endif
602}
603
604#ifdef CONFIG_RPS
605/*
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
608 */
609struct rps_map {
610 unsigned int len;
611 struct rcu_head rcu;
612 u16 cpus[0];
613};
614#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
615
616/*
617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
620 */
621struct rps_dev_flow {
622 u16 cpu;
623 u16 filter;
624 unsigned int last_qtail;
625};
626#define RPS_NO_FILTER 0xffff
627
628/*
629 * The rps_dev_flow_table structure contains a table of flow mappings.
630 */
631struct rps_dev_flow_table {
632 unsigned int mask;
633 struct rcu_head rcu;
634 struct rps_dev_flow flows[0];
635};
636#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
637 ((_num) * sizeof(struct rps_dev_flow)))
638
639/*
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
644 * rps_cpu_mask is used to partition the space, depending on number of
645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
647 * meaning we use 32-6=26 bits for the hash.
648 */
649struct rps_sock_flow_table {
650 u32 mask;
651
652 u32 ents[0] ____cacheline_aligned_in_smp;
653};
654#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655
656#define RPS_NO_CPU 0xffff
657
658extern u32 rps_cpu_mask;
659extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660
661static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 u32 hash)
663{
664 if (table && hash) {
665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
667
668 /* We only give a hint, preemption can change CPU under us */
669 val |= raw_smp_processor_id();
670
671 if (table->ents[index] != val)
672 table->ents[index] = val;
673 }
674}
675
676#ifdef CONFIG_RFS_ACCEL
677bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 u16 filter_id);
679#endif
680#endif /* CONFIG_RPS */
681
682/* This structure contains an instance of an RX queue. */
683struct netdev_rx_queue {
684#ifdef CONFIG_RPS
685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
687#endif
688 struct kobject kobj;
689 struct net_device *dev;
690 struct xdp_rxq_info xdp_rxq;
691} ____cacheline_aligned_in_smp;
692
693/*
694 * RX queue sysfs structures and functions.
695 */
696struct rx_queue_attribute {
697 struct attribute attr;
698 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 const char *buf, size_t len);
701};
702
703#ifdef CONFIG_XPS
704/*
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
707 */
708struct xps_map {
709 unsigned int len;
710 unsigned int alloc_len;
711 struct rcu_head rcu;
712 u16 queues[0];
713};
714#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
715#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
717
718/*
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 */
721struct xps_dev_maps {
722 struct rcu_head rcu;
723 struct xps_map __rcu *cpu_map[0];
724};
725#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
727#endif /* CONFIG_XPS */
728
729#define TC_MAX_QUEUE 16
730#define TC_BITMASK 15
731/* HW offloaded queuing disciplines txq count and offset maps */
732struct netdev_tc_txq {
733 u16 count;
734 u16 offset;
735};
736
737#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738/*
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
741 */
742struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
749 char model[256];
750 char model_description[256];
751};
752#endif
753
754#define MAX_PHYS_ITEM_ID_LEN 32
755
756/* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
758 */
759struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
761 unsigned char id_len;
762};
763
764static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
766{
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
769}
770
771typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
773
774enum tc_setup_type {
775 TC_SETUP_QDISC_MQPRIO,
776 TC_SETUP_CLSU32,
777 TC_SETUP_CLSFLOWER,
778 TC_SETUP_CLSMATCHALL,
779 TC_SETUP_CLSBPF,
780 TC_SETUP_BLOCK,
781 TC_SETUP_QDISC_CBS,
782 TC_SETUP_QDISC_RED,
783 TC_SETUP_QDISC_PRIO,
784};
785
786/* These structures hold the attributes of bpf state that are being passed
787 * to the netdevice through the bpf op.
788 */
789enum bpf_netdev_command {
790 /* Set or clear a bpf program used in the earliest stages of packet
791 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
792 * is responsible for calling bpf_prog_put on any old progs that are
793 * stored. In case of error, the callee need not release the new prog
794 * reference, but on success it takes ownership and must bpf_prog_put
795 * when it is no longer used.
796 */
797 XDP_SETUP_PROG,
798 XDP_SETUP_PROG_HW,
799 /* Check if a bpf program is set on the device. The callee should
800 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
801 * is equivalent to XDP_ATTACHED_DRV.
802 */
803 XDP_QUERY_PROG,
804 /* BPF program for offload callbacks, invoked at program load time. */
805 BPF_OFFLOAD_VERIFIER_PREP,
806 BPF_OFFLOAD_TRANSLATE,
807 BPF_OFFLOAD_DESTROY,
808 BPF_OFFLOAD_MAP_ALLOC,
809 BPF_OFFLOAD_MAP_FREE,
810};
811
812struct bpf_prog_offload_ops;
813struct netlink_ext_ack;
814
815struct netdev_bpf {
816 enum bpf_netdev_command command;
817 union {
818 /* XDP_SETUP_PROG */
819 struct {
820 u32 flags;
821 struct bpf_prog *prog;
822 struct netlink_ext_ack *extack;
823 };
824 /* XDP_QUERY_PROG */
825 struct {
826 u8 prog_attached;
827 u32 prog_id;
828 /* flags with which program was installed */
829 u32 prog_flags;
830 };
831 /* BPF_OFFLOAD_VERIFIER_PREP */
832 struct {
833 struct bpf_prog *prog;
834 const struct bpf_prog_offload_ops *ops; /* callee set */
835 } verifier;
836 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
837 struct {
838 struct bpf_prog *prog;
839 } offload;
840 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
841 struct {
842 struct bpf_offloaded_map *offmap;
843 };
844 };
845};
846
847#ifdef CONFIG_XFRM_OFFLOAD
848struct xfrmdev_ops {
849 int (*xdo_dev_state_add) (struct xfrm_state *x);
850 void (*xdo_dev_state_delete) (struct xfrm_state *x);
851 void (*xdo_dev_state_free) (struct xfrm_state *x);
852 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
853 struct xfrm_state *x);
854 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
855};
856#endif
857
858struct dev_ifalias {
859 struct rcu_head rcuhead;
860 char ifalias[];
861};
862
863/*
864 * This structure defines the management hooks for network devices.
865 * The following hooks can be defined; unless noted otherwise, they are
866 * optional and can be filled with a null pointer.
867 *
868 * int (*ndo_init)(struct net_device *dev);
869 * This function is called once when a network device is registered.
870 * The network device can use this for any late stage initialization
871 * or semantic validation. It can fail with an error code which will
872 * be propagated back to register_netdev.
873 *
874 * void (*ndo_uninit)(struct net_device *dev);
875 * This function is called when device is unregistered or when registration
876 * fails. It is not called if init fails.
877 *
878 * int (*ndo_open)(struct net_device *dev);
879 * This function is called when a network device transitions to the up
880 * state.
881 *
882 * int (*ndo_stop)(struct net_device *dev);
883 * This function is called when a network device transitions to the down
884 * state.
885 *
886 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
887 * struct net_device *dev);
888 * Called when a packet needs to be transmitted.
889 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
890 * the queue before that can happen; it's for obsolete devices and weird
891 * corner cases, but the stack really does a non-trivial amount
892 * of useless work if you return NETDEV_TX_BUSY.
893 * Required; cannot be NULL.
894 *
895 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
896 * struct net_device *dev
897 * netdev_features_t features);
898 * Called by core transmit path to determine if device is capable of
899 * performing offload operations on a given packet. This is to give
900 * the device an opportunity to implement any restrictions that cannot
901 * be otherwise expressed by feature flags. The check is called with
902 * the set of features that the stack has calculated and it returns
903 * those the driver believes to be appropriate.
904 *
905 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
906 * void *accel_priv, select_queue_fallback_t fallback);
907 * Called to decide which queue to use when device supports multiple
908 * transmit queues.
909 *
910 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
911 * This function is called to allow device receiver to make
912 * changes to configuration when multicast or promiscuous is enabled.
913 *
914 * void (*ndo_set_rx_mode)(struct net_device *dev);
915 * This function is called device changes address list filtering.
916 * If driver handles unicast address filtering, it should set
917 * IFF_UNICAST_FLT in its priv_flags.
918 *
919 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
920 * This function is called when the Media Access Control address
921 * needs to be changed. If this interface is not defined, the
922 * MAC address can not be changed.
923 *
924 * int (*ndo_validate_addr)(struct net_device *dev);
925 * Test if Media Access Control address is valid for the device.
926 *
927 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
928 * Called when a user requests an ioctl which can't be handled by
929 * the generic interface code. If not defined ioctls return
930 * not supported error code.
931 *
932 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
933 * Used to set network devices bus interface parameters. This interface
934 * is retained for legacy reasons; new devices should use the bus
935 * interface (PCI) for low level management.
936 *
937 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
938 * Called when a user wants to change the Maximum Transfer Unit
939 * of a device.
940 *
941 * void (*ndo_tx_timeout)(struct net_device *dev);
942 * Callback used when the transmitter has not made any progress
943 * for dev->watchdog ticks.
944 *
945 * void (*ndo_get_stats64)(struct net_device *dev,
946 * struct rtnl_link_stats64 *storage);
947 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
948 * Called when a user wants to get the network device usage
949 * statistics. Drivers must do one of the following:
950 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
951 * rtnl_link_stats64 structure passed by the caller.
952 * 2. Define @ndo_get_stats to update a net_device_stats structure
953 * (which should normally be dev->stats) and return a pointer to
954 * it. The structure may be changed asynchronously only if each
955 * field is written atomically.
956 * 3. Update dev->stats asynchronously and atomically, and define
957 * neither operation.
958 *
959 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
960 * Return true if this device supports offload stats of this attr_id.
961 *
962 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
963 * void *attr_data)
964 * Get statistics for offload operations by attr_id. Write it into the
965 * attr_data pointer.
966 *
967 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
968 * If device supports VLAN filtering this function is called when a
969 * VLAN id is registered.
970 *
971 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
972 * If device supports VLAN filtering this function is called when a
973 * VLAN id is unregistered.
974 *
975 * void (*ndo_poll_controller)(struct net_device *dev);
976 *
977 * SR-IOV management functions.
978 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
979 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
980 * u8 qos, __be16 proto);
981 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
982 * int max_tx_rate);
983 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
984 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
985 * int (*ndo_get_vf_config)(struct net_device *dev,
986 * int vf, struct ifla_vf_info *ivf);
987 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
988 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
989 * struct nlattr *port[]);
990 *
991 * Enable or disable the VF ability to query its RSS Redirection Table and
992 * Hash Key. This is needed since on some devices VF share this information
993 * with PF and querying it may introduce a theoretical security risk.
994 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
995 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
996 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
997 * void *type_data);
998 * Called to setup any 'tc' scheduler, classifier or action on @dev.
999 * This is always called from the stack with the rtnl lock held and netif
1000 * tx queues stopped. This allows the netdevice to perform queue
1001 * management safely.
1002 *
1003 * Fiber Channel over Ethernet (FCoE) offload functions.
1004 * int (*ndo_fcoe_enable)(struct net_device *dev);
1005 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1006 * so the underlying device can perform whatever needed configuration or
1007 * initialization to support acceleration of FCoE traffic.
1008 *
1009 * int (*ndo_fcoe_disable)(struct net_device *dev);
1010 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1011 * so the underlying device can perform whatever needed clean-ups to
1012 * stop supporting acceleration of FCoE traffic.
1013 *
1014 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1015 * struct scatterlist *sgl, unsigned int sgc);
1016 * Called when the FCoE Initiator wants to initialize an I/O that
1017 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1018 * perform necessary setup and returns 1 to indicate the device is set up
1019 * successfully to perform DDP on this I/O, otherwise this returns 0.
1020 *
1021 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1022 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1023 * indicated by the FC exchange id 'xid', so the underlying device can
1024 * clean up and reuse resources for later DDP requests.
1025 *
1026 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1027 * struct scatterlist *sgl, unsigned int sgc);
1028 * Called when the FCoE Target wants to initialize an I/O that
1029 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1030 * perform necessary setup and returns 1 to indicate the device is set up
1031 * successfully to perform DDP on this I/O, otherwise this returns 0.
1032 *
1033 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1034 * struct netdev_fcoe_hbainfo *hbainfo);
1035 * Called when the FCoE Protocol stack wants information on the underlying
1036 * device. This information is utilized by the FCoE protocol stack to
1037 * register attributes with Fiber Channel management service as per the
1038 * FC-GS Fabric Device Management Information(FDMI) specification.
1039 *
1040 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1041 * Called when the underlying device wants to override default World Wide
1042 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1043 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1044 * protocol stack to use.
1045 *
1046 * RFS acceleration.
1047 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1048 * u16 rxq_index, u32 flow_id);
1049 * Set hardware filter for RFS. rxq_index is the target queue index;
1050 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1051 * Return the filter ID on success, or a negative error code.
1052 *
1053 * Slave management functions (for bridge, bonding, etc).
1054 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1055 * Called to make another netdev an underling.
1056 *
1057 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1058 * Called to release previously enslaved netdev.
1059 *
1060 * Feature/offload setting functions.
1061 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1062 * netdev_features_t features);
1063 * Adjusts the requested feature flags according to device-specific
1064 * constraints, and returns the resulting flags. Must not modify
1065 * the device state.
1066 *
1067 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1068 * Called to update device configuration to new features. Passed
1069 * feature set might be less than what was returned by ndo_fix_features()).
1070 * Must return >0 or -errno if it changed dev->features itself.
1071 *
1072 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1073 * struct net_device *dev,
1074 * const unsigned char *addr, u16 vid, u16 flags)
1075 * Adds an FDB entry to dev for addr.
1076 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1077 * struct net_device *dev,
1078 * const unsigned char *addr, u16 vid)
1079 * Deletes the FDB entry from dev coresponding to addr.
1080 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1081 * struct net_device *dev, struct net_device *filter_dev,
1082 * int *idx)
1083 * Used to add FDB entries to dump requests. Implementers should add
1084 * entries to skb and update idx with the number of entries.
1085 *
1086 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1087 * u16 flags)
1088 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1089 * struct net_device *dev, u32 filter_mask,
1090 * int nlflags)
1091 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1092 * u16 flags);
1093 *
1094 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1095 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1096 * which do not represent real hardware may define this to allow their
1097 * userspace components to manage their virtual carrier state. Devices
1098 * that determine carrier state from physical hardware properties (eg
1099 * network cables) or protocol-dependent mechanisms (eg
1100 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1101 *
1102 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1103 * struct netdev_phys_item_id *ppid);
1104 * Called to get ID of physical port of this device. If driver does
1105 * not implement this, it is assumed that the hw is not able to have
1106 * multiple net devices on single physical port.
1107 *
1108 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1109 * struct udp_tunnel_info *ti);
1110 * Called by UDP tunnel to notify a driver about the UDP port and socket
1111 * address family that a UDP tunnel is listnening to. It is called only
1112 * when a new port starts listening. The operation is protected by the
1113 * RTNL.
1114 *
1115 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1116 * struct udp_tunnel_info *ti);
1117 * Called by UDP tunnel to notify the driver about a UDP port and socket
1118 * address family that the UDP tunnel is not listening to anymore. The
1119 * operation is protected by the RTNL.
1120 *
1121 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1122 * struct net_device *dev)
1123 * Called by upper layer devices to accelerate switching or other
1124 * station functionality into hardware. 'pdev is the lowerdev
1125 * to use for the offload and 'dev' is the net device that will
1126 * back the offload. Returns a pointer to the private structure
1127 * the upper layer will maintain.
1128 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1129 * Called by upper layer device to delete the station created
1130 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1131 * the station and priv is the structure returned by the add
1132 * operation.
1133 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1134 * int queue_index, u32 maxrate);
1135 * Called when a user wants to set a max-rate limitation of specific
1136 * TX queue.
1137 * int (*ndo_get_iflink)(const struct net_device *dev);
1138 * Called to get the iflink value of this device.
1139 * void (*ndo_change_proto_down)(struct net_device *dev,
1140 * bool proto_down);
1141 * This function is used to pass protocol port error state information
1142 * to the switch driver. The switch driver can react to the proto_down
1143 * by doing a phys down on the associated switch port.
1144 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1145 * This function is used to get egress tunnel information for given skb.
1146 * This is useful for retrieving outer tunnel header parameters while
1147 * sampling packet.
1148 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1149 * This function is used to specify the headroom that the skb must
1150 * consider when allocation skb during packet reception. Setting
1151 * appropriate rx headroom value allows avoiding skb head copy on
1152 * forward. Setting a negative value resets the rx headroom to the
1153 * default value.
1154 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1155 * This function is used to set or query state related to XDP on the
1156 * netdevice and manage BPF offload. See definition of
1157 * enum bpf_netdev_command for details.
1158 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1159 * This function is used to submit a XDP packet for transmit on a
1160 * netdevice.
1161 * void (*ndo_xdp_flush)(struct net_device *dev);
1162 * This function is used to inform the driver to flush a particular
1163 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1164 */
1165struct net_device_ops {
1166 int (*ndo_init)(struct net_device *dev);
1167 void (*ndo_uninit)(struct net_device *dev);
1168 int (*ndo_open)(struct net_device *dev);
1169 int (*ndo_stop)(struct net_device *dev);
1170 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1171 struct net_device *dev);
1172 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1173 struct net_device *dev,
1174 netdev_features_t features);
1175 u16 (*ndo_select_queue)(struct net_device *dev,
1176 struct sk_buff *skb,
1177 void *accel_priv,
1178 select_queue_fallback_t fallback);
1179 void (*ndo_change_rx_flags)(struct net_device *dev,
1180 int flags);
1181 void (*ndo_set_rx_mode)(struct net_device *dev);
1182 int (*ndo_set_mac_address)(struct net_device *dev,
1183 void *addr);
1184 int (*ndo_validate_addr)(struct net_device *dev);
1185 int (*ndo_do_ioctl)(struct net_device *dev,
1186 struct ifreq *ifr, int cmd);
1187 int (*ndo_set_config)(struct net_device *dev,
1188 struct ifmap *map);
1189 int (*ndo_change_mtu)(struct net_device *dev,
1190 int new_mtu);
1191 int (*ndo_neigh_setup)(struct net_device *dev,
1192 struct neigh_parms *);
1193 void (*ndo_tx_timeout) (struct net_device *dev);
1194
1195 void (*ndo_get_stats64)(struct net_device *dev,
1196 struct rtnl_link_stats64 *storage);
1197 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1198 int (*ndo_get_offload_stats)(int attr_id,
1199 const struct net_device *dev,
1200 void *attr_data);
1201 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1202
1203 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1204 __be16 proto, u16 vid);
1205 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1206 __be16 proto, u16 vid);
1207#ifdef CONFIG_NET_POLL_CONTROLLER
1208 void (*ndo_poll_controller)(struct net_device *dev);
1209 int (*ndo_netpoll_setup)(struct net_device *dev,
1210 struct netpoll_info *info);
1211 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1212#endif
1213 int (*ndo_set_vf_mac)(struct net_device *dev,
1214 int queue, u8 *mac);
1215 int (*ndo_set_vf_vlan)(struct net_device *dev,
1216 int queue, u16 vlan,
1217 u8 qos, __be16 proto);
1218 int (*ndo_set_vf_rate)(struct net_device *dev,
1219 int vf, int min_tx_rate,
1220 int max_tx_rate);
1221 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1222 int vf, bool setting);
1223 int (*ndo_set_vf_trust)(struct net_device *dev,
1224 int vf, bool setting);
1225 int (*ndo_get_vf_config)(struct net_device *dev,
1226 int vf,
1227 struct ifla_vf_info *ivf);
1228 int (*ndo_set_vf_link_state)(struct net_device *dev,
1229 int vf, int link_state);
1230 int (*ndo_get_vf_stats)(struct net_device *dev,
1231 int vf,
1232 struct ifla_vf_stats
1233 *vf_stats);
1234 int (*ndo_set_vf_port)(struct net_device *dev,
1235 int vf,
1236 struct nlattr *port[]);
1237 int (*ndo_get_vf_port)(struct net_device *dev,
1238 int vf, struct sk_buff *skb);
1239 int (*ndo_set_vf_guid)(struct net_device *dev,
1240 int vf, u64 guid,
1241 int guid_type);
1242 int (*ndo_set_vf_rss_query_en)(
1243 struct net_device *dev,
1244 int vf, bool setting);
1245 int (*ndo_setup_tc)(struct net_device *dev,
1246 enum tc_setup_type type,
1247 void *type_data);
1248#if IS_ENABLED(CONFIG_FCOE)
1249 int (*ndo_fcoe_enable)(struct net_device *dev);
1250 int (*ndo_fcoe_disable)(struct net_device *dev);
1251 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1252 u16 xid,
1253 struct scatterlist *sgl,
1254 unsigned int sgc);
1255 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1256 u16 xid);
1257 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1258 u16 xid,
1259 struct scatterlist *sgl,
1260 unsigned int sgc);
1261 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1262 struct netdev_fcoe_hbainfo *hbainfo);
1263#endif
1264
1265#if IS_ENABLED(CONFIG_LIBFCOE)
1266#define NETDEV_FCOE_WWNN 0
1267#define NETDEV_FCOE_WWPN 1
1268 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1269 u64 *wwn, int type);
1270#endif
1271
1272#ifdef CONFIG_RFS_ACCEL
1273 int (*ndo_rx_flow_steer)(struct net_device *dev,
1274 const struct sk_buff *skb,
1275 u16 rxq_index,
1276 u32 flow_id);
1277#endif
1278 int (*ndo_add_slave)(struct net_device *dev,
1279 struct net_device *slave_dev,
1280 struct netlink_ext_ack *extack);
1281 int (*ndo_del_slave)(struct net_device *dev,
1282 struct net_device *slave_dev);
1283 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1284 netdev_features_t features);
1285 int (*ndo_set_features)(struct net_device *dev,
1286 netdev_features_t features);
1287 int (*ndo_neigh_construct)(struct net_device *dev,
1288 struct neighbour *n);
1289 void (*ndo_neigh_destroy)(struct net_device *dev,
1290 struct neighbour *n);
1291
1292 int (*ndo_fdb_add)(struct ndmsg *ndm,
1293 struct nlattr *tb[],
1294 struct net_device *dev,
1295 const unsigned char *addr,
1296 u16 vid,
1297 u16 flags);
1298 int (*ndo_fdb_del)(struct ndmsg *ndm,
1299 struct nlattr *tb[],
1300 struct net_device *dev,
1301 const unsigned char *addr,
1302 u16 vid);
1303 int (*ndo_fdb_dump)(struct sk_buff *skb,
1304 struct netlink_callback *cb,
1305 struct net_device *dev,
1306 struct net_device *filter_dev,
1307 int *idx);
1308
1309 int (*ndo_bridge_setlink)(struct net_device *dev,
1310 struct nlmsghdr *nlh,
1311 u16 flags);
1312 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1313 u32 pid, u32 seq,
1314 struct net_device *dev,
1315 u32 filter_mask,
1316 int nlflags);
1317 int (*ndo_bridge_dellink)(struct net_device *dev,
1318 struct nlmsghdr *nlh,
1319 u16 flags);
1320 int (*ndo_change_carrier)(struct net_device *dev,
1321 bool new_carrier);
1322 int (*ndo_get_phys_port_id)(struct net_device *dev,
1323 struct netdev_phys_item_id *ppid);
1324 int (*ndo_get_phys_port_name)(struct net_device *dev,
1325 char *name, size_t len);
1326 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1327 struct udp_tunnel_info *ti);
1328 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1329 struct udp_tunnel_info *ti);
1330 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1331 struct net_device *dev);
1332 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1333 void *priv);
1334
1335 int (*ndo_get_lock_subclass)(struct net_device *dev);
1336 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1337 int queue_index,
1338 u32 maxrate);
1339 int (*ndo_get_iflink)(const struct net_device *dev);
1340 int (*ndo_change_proto_down)(struct net_device *dev,
1341 bool proto_down);
1342 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1343 struct sk_buff *skb);
1344 void (*ndo_set_rx_headroom)(struct net_device *dev,
1345 int needed_headroom);
1346 int (*ndo_bpf)(struct net_device *dev,
1347 struct netdev_bpf *bpf);
1348 int (*ndo_xdp_xmit)(struct net_device *dev,
1349 struct xdp_buff *xdp);
1350 void (*ndo_xdp_flush)(struct net_device *dev);
1351};
1352
1353/**
1354 * enum net_device_priv_flags - &struct net_device priv_flags
1355 *
1356 * These are the &struct net_device, they are only set internally
1357 * by drivers and used in the kernel. These flags are invisible to
1358 * userspace; this means that the order of these flags can change
1359 * during any kernel release.
1360 *
1361 * You should have a pretty good reason to be extending these flags.
1362 *
1363 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1364 * @IFF_EBRIDGE: Ethernet bridging device
1365 * @IFF_BONDING: bonding master or slave
1366 * @IFF_ISATAP: ISATAP interface (RFC4214)
1367 * @IFF_WAN_HDLC: WAN HDLC device
1368 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1369 * release skb->dst
1370 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1371 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1372 * @IFF_MACVLAN_PORT: device used as macvlan port
1373 * @IFF_BRIDGE_PORT: device used as bridge port
1374 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1375 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1376 * @IFF_UNICAST_FLT: Supports unicast filtering
1377 * @IFF_TEAM_PORT: device used as team port
1378 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1379 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1380 * change when it's running
1381 * @IFF_MACVLAN: Macvlan device
1382 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1383 * underlying stacked devices
1384 * @IFF_IPVLAN_MASTER: IPvlan master device
1385 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1386 * @IFF_L3MDEV_MASTER: device is an L3 master device
1387 * @IFF_NO_QUEUE: device can run without qdisc attached
1388 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1389 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1390 * @IFF_TEAM: device is a team device
1391 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1392 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1393 * entity (i.e. the master device for bridged veth)
1394 * @IFF_MACSEC: device is a MACsec device
1395 */
1396enum netdev_priv_flags {
1397 IFF_802_1Q_VLAN = 1<<0,
1398 IFF_EBRIDGE = 1<<1,
1399 IFF_BONDING = 1<<2,
1400 IFF_ISATAP = 1<<3,
1401 IFF_WAN_HDLC = 1<<4,
1402 IFF_XMIT_DST_RELEASE = 1<<5,
1403 IFF_DONT_BRIDGE = 1<<6,
1404 IFF_DISABLE_NETPOLL = 1<<7,
1405 IFF_MACVLAN_PORT = 1<<8,
1406 IFF_BRIDGE_PORT = 1<<9,
1407 IFF_OVS_DATAPATH = 1<<10,
1408 IFF_TX_SKB_SHARING = 1<<11,
1409 IFF_UNICAST_FLT = 1<<12,
1410 IFF_TEAM_PORT = 1<<13,
1411 IFF_SUPP_NOFCS = 1<<14,
1412 IFF_LIVE_ADDR_CHANGE = 1<<15,
1413 IFF_MACVLAN = 1<<16,
1414 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1415 IFF_IPVLAN_MASTER = 1<<18,
1416 IFF_IPVLAN_SLAVE = 1<<19,
1417 IFF_L3MDEV_MASTER = 1<<20,
1418 IFF_NO_QUEUE = 1<<21,
1419 IFF_OPENVSWITCH = 1<<22,
1420 IFF_L3MDEV_SLAVE = 1<<23,
1421 IFF_TEAM = 1<<24,
1422 IFF_RXFH_CONFIGURED = 1<<25,
1423 IFF_PHONY_HEADROOM = 1<<26,
1424 IFF_MACSEC = 1<<27,
1425};
1426
1427#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1428#define IFF_EBRIDGE IFF_EBRIDGE
1429#define IFF_BONDING IFF_BONDING
1430#define IFF_ISATAP IFF_ISATAP
1431#define IFF_WAN_HDLC IFF_WAN_HDLC
1432#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1433#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1434#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1435#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1436#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1437#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1438#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1439#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1440#define IFF_TEAM_PORT IFF_TEAM_PORT
1441#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1442#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1443#define IFF_MACVLAN IFF_MACVLAN
1444#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1445#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1446#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1447#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1448#define IFF_NO_QUEUE IFF_NO_QUEUE
1449#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1450#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1451#define IFF_TEAM IFF_TEAM
1452#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1453#define IFF_MACSEC IFF_MACSEC
1454
1455/**
1456 * struct net_device - The DEVICE structure.
1457 *
1458 * Actually, this whole structure is a big mistake. It mixes I/O
1459 * data with strictly "high-level" data, and it has to know about
1460 * almost every data structure used in the INET module.
1461 *
1462 * @name: This is the first field of the "visible" part of this structure
1463 * (i.e. as seen by users in the "Space.c" file). It is the name
1464 * of the interface.
1465 *
1466 * @name_hlist: Device name hash chain, please keep it close to name[]
1467 * @ifalias: SNMP alias
1468 * @mem_end: Shared memory end
1469 * @mem_start: Shared memory start
1470 * @base_addr: Device I/O address
1471 * @irq: Device IRQ number
1472 *
1473 * @state: Generic network queuing layer state, see netdev_state_t
1474 * @dev_list: The global list of network devices
1475 * @napi_list: List entry used for polling NAPI devices
1476 * @unreg_list: List entry when we are unregistering the
1477 * device; see the function unregister_netdev
1478 * @close_list: List entry used when we are closing the device
1479 * @ptype_all: Device-specific packet handlers for all protocols
1480 * @ptype_specific: Device-specific, protocol-specific packet handlers
1481 *
1482 * @adj_list: Directly linked devices, like slaves for bonding
1483 * @features: Currently active device features
1484 * @hw_features: User-changeable features
1485 *
1486 * @wanted_features: User-requested features
1487 * @vlan_features: Mask of features inheritable by VLAN devices
1488 *
1489 * @hw_enc_features: Mask of features inherited by encapsulating devices
1490 * This field indicates what encapsulation
1491 * offloads the hardware is capable of doing,
1492 * and drivers will need to set them appropriately.
1493 *
1494 * @mpls_features: Mask of features inheritable by MPLS
1495 *
1496 * @ifindex: interface index
1497 * @group: The group the device belongs to
1498 *
1499 * @stats: Statistics struct, which was left as a legacy, use
1500 * rtnl_link_stats64 instead
1501 *
1502 * @rx_dropped: Dropped packets by core network,
1503 * do not use this in drivers
1504 * @tx_dropped: Dropped packets by core network,
1505 * do not use this in drivers
1506 * @rx_nohandler: nohandler dropped packets by core network on
1507 * inactive devices, do not use this in drivers
1508 * @carrier_up_count: Number of times the carrier has been up
1509 * @carrier_down_count: Number of times the carrier has been down
1510 *
1511 * @wireless_handlers: List of functions to handle Wireless Extensions,
1512 * instead of ioctl,
1513 * see <net/iw_handler.h> for details.
1514 * @wireless_data: Instance data managed by the core of wireless extensions
1515 *
1516 * @netdev_ops: Includes several pointers to callbacks,
1517 * if one wants to override the ndo_*() functions
1518 * @ethtool_ops: Management operations
1519 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1520 * discovery handling. Necessary for e.g. 6LoWPAN.
1521 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1522 * of Layer 2 headers.
1523 *
1524 * @flags: Interface flags (a la BSD)
1525 * @priv_flags: Like 'flags' but invisible to userspace,
1526 * see if.h for the definitions
1527 * @gflags: Global flags ( kept as legacy )
1528 * @padded: How much padding added by alloc_netdev()
1529 * @operstate: RFC2863 operstate
1530 * @link_mode: Mapping policy to operstate
1531 * @if_port: Selectable AUI, TP, ...
1532 * @dma: DMA channel
1533 * @mtu: Interface MTU value
1534 * @min_mtu: Interface Minimum MTU value
1535 * @max_mtu: Interface Maximum MTU value
1536 * @type: Interface hardware type
1537 * @hard_header_len: Maximum hardware header length.
1538 * @min_header_len: Minimum hardware header length
1539 *
1540 * @needed_headroom: Extra headroom the hardware may need, but not in all
1541 * cases can this be guaranteed
1542 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1543 * cases can this be guaranteed. Some cases also use
1544 * LL_MAX_HEADER instead to allocate the skb
1545 *
1546 * interface address info:
1547 *
1548 * @perm_addr: Permanent hw address
1549 * @addr_assign_type: Hw address assignment type
1550 * @addr_len: Hardware address length
1551 * @neigh_priv_len: Used in neigh_alloc()
1552 * @dev_id: Used to differentiate devices that share
1553 * the same link layer address
1554 * @dev_port: Used to differentiate devices that share
1555 * the same function
1556 * @addr_list_lock: XXX: need comments on this one
1557 * @uc_promisc: Counter that indicates promiscuous mode
1558 * has been enabled due to the need to listen to
1559 * additional unicast addresses in a device that
1560 * does not implement ndo_set_rx_mode()
1561 * @uc: unicast mac addresses
1562 * @mc: multicast mac addresses
1563 * @dev_addrs: list of device hw addresses
1564 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1565 * @promiscuity: Number of times the NIC is told to work in
1566 * promiscuous mode; if it becomes 0 the NIC will
1567 * exit promiscuous mode
1568 * @allmulti: Counter, enables or disables allmulticast mode
1569 *
1570 * @vlan_info: VLAN info
1571 * @dsa_ptr: dsa specific data
1572 * @tipc_ptr: TIPC specific data
1573 * @atalk_ptr: AppleTalk link
1574 * @ip_ptr: IPv4 specific data
1575 * @dn_ptr: DECnet specific data
1576 * @ip6_ptr: IPv6 specific data
1577 * @ax25_ptr: AX.25 specific data
1578 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1579 *
1580 * @dev_addr: Hw address (before bcast,
1581 * because most packets are unicast)
1582 *
1583 * @_rx: Array of RX queues
1584 * @num_rx_queues: Number of RX queues
1585 * allocated at register_netdev() time
1586 * @real_num_rx_queues: Number of RX queues currently active in device
1587 *
1588 * @rx_handler: handler for received packets
1589 * @rx_handler_data: XXX: need comments on this one
1590 * @miniq_ingress: ingress/clsact qdisc specific data for
1591 * ingress processing
1592 * @ingress_queue: XXX: need comments on this one
1593 * @broadcast: hw bcast address
1594 *
1595 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1596 * indexed by RX queue number. Assigned by driver.
1597 * This must only be set if the ndo_rx_flow_steer
1598 * operation is defined
1599 * @index_hlist: Device index hash chain
1600 *
1601 * @_tx: Array of TX queues
1602 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1603 * @real_num_tx_queues: Number of TX queues currently active in device
1604 * @qdisc: Root qdisc from userspace point of view
1605 * @tx_queue_len: Max frames per queue allowed
1606 * @tx_global_lock: XXX: need comments on this one
1607 *
1608 * @xps_maps: XXX: need comments on this one
1609 * @miniq_egress: clsact qdisc specific data for
1610 * egress processing
1611 * @watchdog_timeo: Represents the timeout that is used by
1612 * the watchdog (see dev_watchdog())
1613 * @watchdog_timer: List of timers
1614 *
1615 * @pcpu_refcnt: Number of references to this device
1616 * @todo_list: Delayed register/unregister
1617 * @link_watch_list: XXX: need comments on this one
1618 *
1619 * @reg_state: Register/unregister state machine
1620 * @dismantle: Device is going to be freed
1621 * @rtnl_link_state: This enum represents the phases of creating
1622 * a new link
1623 *
1624 * @needs_free_netdev: Should unregister perform free_netdev?
1625 * @priv_destructor: Called from unregister
1626 * @npinfo: XXX: need comments on this one
1627 * @nd_net: Network namespace this network device is inside
1628 *
1629 * @ml_priv: Mid-layer private
1630 * @lstats: Loopback statistics
1631 * @tstats: Tunnel statistics
1632 * @dstats: Dummy statistics
1633 * @vstats: Virtual ethernet statistics
1634 *
1635 * @garp_port: GARP
1636 * @mrp_port: MRP
1637 *
1638 * @dev: Class/net/name entry
1639 * @sysfs_groups: Space for optional device, statistics and wireless
1640 * sysfs groups
1641 *
1642 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1643 * @rtnl_link_ops: Rtnl_link_ops
1644 *
1645 * @gso_max_size: Maximum size of generic segmentation offload
1646 * @gso_max_segs: Maximum number of segments that can be passed to the
1647 * NIC for GSO
1648 *
1649 * @dcbnl_ops: Data Center Bridging netlink ops
1650 * @num_tc: Number of traffic classes in the net device
1651 * @tc_to_txq: XXX: need comments on this one
1652 * @prio_tc_map: XXX: need comments on this one
1653 *
1654 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1655 *
1656 * @priomap: XXX: need comments on this one
1657 * @phydev: Physical device may attach itself
1658 * for hardware timestamping
1659 *
1660 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1661 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1662 *
1663 * @proto_down: protocol port state information can be sent to the
1664 * switch driver and used to set the phys state of the
1665 * switch port.
1666 *
1667 * FIXME: cleanup struct net_device such that network protocol info
1668 * moves out.
1669 */
1670
1671struct net_device {
1672 char name[IFNAMSIZ];
1673 struct hlist_node name_hlist;
1674 struct dev_ifalias __rcu *ifalias;
1675 /*
1676 * I/O specific fields
1677 * FIXME: Merge these and struct ifmap into one
1678 */
1679 unsigned long mem_end;
1680 unsigned long mem_start;
1681 unsigned long base_addr;
1682 int irq;
1683
1684 /*
1685 * Some hardware also needs these fields (state,dev_list,
1686 * napi_list,unreg_list,close_list) but they are not
1687 * part of the usual set specified in Space.c.
1688 */
1689
1690 unsigned long state;
1691
1692 struct list_head dev_list;
1693 struct list_head napi_list;
1694 struct list_head unreg_list;
1695 struct list_head close_list;
1696 struct list_head ptype_all;
1697 struct list_head ptype_specific;
1698
1699 struct {
1700 struct list_head upper;
1701 struct list_head lower;
1702 } adj_list;
1703
1704 netdev_features_t features;
1705 netdev_features_t hw_features;
1706 netdev_features_t wanted_features;
1707 netdev_features_t vlan_features;
1708 netdev_features_t hw_enc_features;
1709 netdev_features_t mpls_features;
1710 netdev_features_t gso_partial_features;
1711
1712 int ifindex;
1713 int group;
1714
1715 struct net_device_stats stats;
1716
1717 atomic_long_t rx_dropped;
1718 atomic_long_t tx_dropped;
1719 atomic_long_t rx_nohandler;
1720
1721 /* Stats to monitor link on/off, flapping */
1722 atomic_t carrier_up_count;
1723 atomic_t carrier_down_count;
1724
1725#ifdef CONFIG_WIRELESS_EXT
1726 const struct iw_handler_def *wireless_handlers;
1727 struct iw_public_data *wireless_data;
1728#endif
1729 const struct net_device_ops *netdev_ops;
1730 const struct ethtool_ops *ethtool_ops;
1731#ifdef CONFIG_NET_SWITCHDEV
1732 const struct switchdev_ops *switchdev_ops;
1733#endif
1734#ifdef CONFIG_NET_L3_MASTER_DEV
1735 const struct l3mdev_ops *l3mdev_ops;
1736#endif
1737#if IS_ENABLED(CONFIG_IPV6)
1738 const struct ndisc_ops *ndisc_ops;
1739#endif
1740
1741#ifdef CONFIG_XFRM_OFFLOAD
1742 const struct xfrmdev_ops *xfrmdev_ops;
1743#endif
1744
1745 const struct header_ops *header_ops;
1746
1747 unsigned int flags;
1748 unsigned int priv_flags;
1749
1750 unsigned short gflags;
1751 unsigned short padded;
1752
1753 unsigned char operstate;
1754 unsigned char link_mode;
1755
1756 unsigned char if_port;
1757 unsigned char dma;
1758
1759 unsigned int mtu;
1760 unsigned int min_mtu;
1761 unsigned int max_mtu;
1762 unsigned short type;
1763 unsigned short hard_header_len;
1764 unsigned char min_header_len;
1765
1766 unsigned short needed_headroom;
1767 unsigned short needed_tailroom;
1768
1769 /* Interface address info. */
1770 unsigned char perm_addr[MAX_ADDR_LEN];
1771 unsigned char addr_assign_type;
1772 unsigned char addr_len;
1773 unsigned short neigh_priv_len;
1774 unsigned short dev_id;
1775 unsigned short dev_port;
1776 spinlock_t addr_list_lock;
1777 unsigned char name_assign_type;
1778 bool uc_promisc;
1779 struct netdev_hw_addr_list uc;
1780 struct netdev_hw_addr_list mc;
1781 struct netdev_hw_addr_list dev_addrs;
1782
1783#ifdef CONFIG_SYSFS
1784 struct kset *queues_kset;
1785#endif
1786 unsigned int promiscuity;
1787 unsigned int allmulti;
1788
1789
1790 /* Protocol-specific pointers */
1791
1792#if IS_ENABLED(CONFIG_VLAN_8021Q)
1793 struct vlan_info __rcu *vlan_info;
1794#endif
1795#if IS_ENABLED(CONFIG_NET_DSA)
1796 struct dsa_port *dsa_ptr;
1797#endif
1798#if IS_ENABLED(CONFIG_TIPC)
1799 struct tipc_bearer __rcu *tipc_ptr;
1800#endif
1801 void *atalk_ptr;
1802 struct in_device __rcu *ip_ptr;
1803 struct dn_dev __rcu *dn_ptr;
1804 struct inet6_dev __rcu *ip6_ptr;
1805 void *ax25_ptr;
1806 struct wireless_dev *ieee80211_ptr;
1807 struct wpan_dev *ieee802154_ptr;
1808#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1809 struct mpls_dev __rcu *mpls_ptr;
1810#endif
1811
1812/*
1813 * Cache lines mostly used on receive path (including eth_type_trans())
1814 */
1815 /* Interface address info used in eth_type_trans() */
1816 unsigned char *dev_addr;
1817
1818 struct netdev_rx_queue *_rx;
1819 unsigned int num_rx_queues;
1820 unsigned int real_num_rx_queues;
1821
1822 struct bpf_prog __rcu *xdp_prog;
1823 unsigned long gro_flush_timeout;
1824 rx_handler_func_t __rcu *rx_handler;
1825 void __rcu *rx_handler_data;
1826
1827#ifdef CONFIG_NET_CLS_ACT
1828 struct mini_Qdisc __rcu *miniq_ingress;
1829#endif
1830 struct netdev_queue __rcu *ingress_queue;
1831#ifdef CONFIG_NETFILTER_INGRESS
1832 struct nf_hook_entries __rcu *nf_hooks_ingress;
1833#endif
1834
1835 unsigned char broadcast[MAX_ADDR_LEN];
1836#ifdef CONFIG_RFS_ACCEL
1837 struct cpu_rmap *rx_cpu_rmap;
1838#endif
1839 struct hlist_node index_hlist;
1840
1841/*
1842 * Cache lines mostly used on transmit path
1843 */
1844 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1845 unsigned int num_tx_queues;
1846 unsigned int real_num_tx_queues;
1847 struct Qdisc *qdisc;
1848#ifdef CONFIG_NET_SCHED
1849 DECLARE_HASHTABLE (qdisc_hash, 4);
1850#endif
1851 unsigned int tx_queue_len;
1852 spinlock_t tx_global_lock;
1853 int watchdog_timeo;
1854
1855#ifdef CONFIG_XPS
1856 struct xps_dev_maps __rcu *xps_maps;
1857#endif
1858#ifdef CONFIG_NET_CLS_ACT
1859 struct mini_Qdisc __rcu *miniq_egress;
1860#endif
1861
1862 /* These may be needed for future network-power-down code. */
1863 struct timer_list watchdog_timer;
1864
1865 int __percpu *pcpu_refcnt;
1866 struct list_head todo_list;
1867
1868 struct list_head link_watch_list;
1869
1870 enum { NETREG_UNINITIALIZED=0,
1871 NETREG_REGISTERED, /* completed register_netdevice */
1872 NETREG_UNREGISTERING, /* called unregister_netdevice */
1873 NETREG_UNREGISTERED, /* completed unregister todo */
1874 NETREG_RELEASED, /* called free_netdev */
1875 NETREG_DUMMY, /* dummy device for NAPI poll */
1876 } reg_state:8;
1877
1878 bool dismantle;
1879
1880 enum {
1881 RTNL_LINK_INITIALIZED,
1882 RTNL_LINK_INITIALIZING,
1883 } rtnl_link_state:16;
1884
1885 bool needs_free_netdev;
1886 void (*priv_destructor)(struct net_device *dev);
1887
1888#ifdef CONFIG_NETPOLL
1889 struct netpoll_info __rcu *npinfo;
1890#endif
1891
1892 possible_net_t nd_net;
1893
1894 /* mid-layer private */
1895 union {
1896 void *ml_priv;
1897 struct pcpu_lstats __percpu *lstats;
1898 struct pcpu_sw_netstats __percpu *tstats;
1899 struct pcpu_dstats __percpu *dstats;
1900 struct pcpu_vstats __percpu *vstats;
1901 };
1902
1903#if IS_ENABLED(CONFIG_GARP)
1904 struct garp_port __rcu *garp_port;
1905#endif
1906#if IS_ENABLED(CONFIG_MRP)
1907 struct mrp_port __rcu *mrp_port;
1908#endif
1909
1910 struct device dev;
1911 const struct attribute_group *sysfs_groups[4];
1912 const struct attribute_group *sysfs_rx_queue_group;
1913
1914 const struct rtnl_link_ops *rtnl_link_ops;
1915
1916 /* for setting kernel sock attribute on TCP connection setup */
1917#define GSO_MAX_SIZE 65536
1918 unsigned int gso_max_size;
1919#define GSO_MAX_SEGS 65535
1920 u16 gso_max_segs;
1921
1922#ifdef CONFIG_DCB
1923 const struct dcbnl_rtnl_ops *dcbnl_ops;
1924#endif
1925 u8 num_tc;
1926 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1927 u8 prio_tc_map[TC_BITMASK + 1];
1928
1929#if IS_ENABLED(CONFIG_FCOE)
1930 unsigned int fcoe_ddp_xid;
1931#endif
1932#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1933 struct netprio_map __rcu *priomap;
1934#endif
1935 struct phy_device *phydev;
1936 struct lock_class_key *qdisc_tx_busylock;
1937 struct lock_class_key *qdisc_running_key;
1938 bool proto_down;
1939};
1940#define to_net_dev(d) container_of(d, struct net_device, dev)
1941
1942static inline bool netif_elide_gro(const struct net_device *dev)
1943{
1944 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1945 return true;
1946 return false;
1947}
1948
1949#define NETDEV_ALIGN 32
1950
1951static inline
1952int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1953{
1954 return dev->prio_tc_map[prio & TC_BITMASK];
1955}
1956
1957static inline
1958int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1959{
1960 if (tc >= dev->num_tc)
1961 return -EINVAL;
1962
1963 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1964 return 0;
1965}
1966
1967int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1968void netdev_reset_tc(struct net_device *dev);
1969int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1970int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1971
1972static inline
1973int netdev_get_num_tc(struct net_device *dev)
1974{
1975 return dev->num_tc;
1976}
1977
1978static inline
1979struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1980 unsigned int index)
1981{
1982 return &dev->_tx[index];
1983}
1984
1985static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1986 const struct sk_buff *skb)
1987{
1988 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1989}
1990
1991static inline void netdev_for_each_tx_queue(struct net_device *dev,
1992 void (*f)(struct net_device *,
1993 struct netdev_queue *,
1994 void *),
1995 void *arg)
1996{
1997 unsigned int i;
1998
1999 for (i = 0; i < dev->num_tx_queues; i++)
2000 f(dev, &dev->_tx[i], arg);
2001}
2002
2003#define netdev_lockdep_set_classes(dev) \
2004{ \
2005 static struct lock_class_key qdisc_tx_busylock_key; \
2006 static struct lock_class_key qdisc_running_key; \
2007 static struct lock_class_key qdisc_xmit_lock_key; \
2008 static struct lock_class_key dev_addr_list_lock_key; \
2009 unsigned int i; \
2010 \
2011 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2012 (dev)->qdisc_running_key = &qdisc_running_key; \
2013 lockdep_set_class(&(dev)->addr_list_lock, \
2014 &dev_addr_list_lock_key); \
2015 for (i = 0; i < (dev)->num_tx_queues; i++) \
2016 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2017 &qdisc_xmit_lock_key); \
2018}
2019
2020struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2021 struct sk_buff *skb,
2022 void *accel_priv);
2023
2024/* returns the headroom that the master device needs to take in account
2025 * when forwarding to this dev
2026 */
2027static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2028{
2029 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2030}
2031
2032static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2033{
2034 if (dev->netdev_ops->ndo_set_rx_headroom)
2035 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2036}
2037
2038/* set the device rx headroom to the dev's default */
2039static inline void netdev_reset_rx_headroom(struct net_device *dev)
2040{
2041 netdev_set_rx_headroom(dev, -1);
2042}
2043
2044/*
2045 * Net namespace inlines
2046 */
2047static inline
2048struct net *dev_net(const struct net_device *dev)
2049{
2050 return read_pnet(&dev->nd_net);
2051}
2052
2053static inline
2054void dev_net_set(struct net_device *dev, struct net *net)
2055{
2056 write_pnet(&dev->nd_net, net);
2057}
2058
2059/**
2060 * netdev_priv - access network device private data
2061 * @dev: network device
2062 *
2063 * Get network device private data
2064 */
2065static inline void *netdev_priv(const struct net_device *dev)
2066{
2067 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2068}
2069
2070/* Set the sysfs physical device reference for the network logical device
2071 * if set prior to registration will cause a symlink during initialization.
2072 */
2073#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2074
2075/* Set the sysfs device type for the network logical device to allow
2076 * fine-grained identification of different network device types. For
2077 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2078 */
2079#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2080
2081/* Default NAPI poll() weight
2082 * Device drivers are strongly advised to not use bigger value
2083 */
2084#define NAPI_POLL_WEIGHT 64
2085
2086/**
2087 * netif_napi_add - initialize a NAPI context
2088 * @dev: network device
2089 * @napi: NAPI context
2090 * @poll: polling function
2091 * @weight: default weight
2092 *
2093 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2094 * *any* of the other NAPI-related functions.
2095 */
2096void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2097 int (*poll)(struct napi_struct *, int), int weight);
2098
2099/**
2100 * netif_tx_napi_add - initialize a NAPI context
2101 * @dev: network device
2102 * @napi: NAPI context
2103 * @poll: polling function
2104 * @weight: default weight
2105 *
2106 * This variant of netif_napi_add() should be used from drivers using NAPI
2107 * to exclusively poll a TX queue.
2108 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2109 */
2110static inline void netif_tx_napi_add(struct net_device *dev,
2111 struct napi_struct *napi,
2112 int (*poll)(struct napi_struct *, int),
2113 int weight)
2114{
2115 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2116 netif_napi_add(dev, napi, poll, weight);
2117}
2118
2119/**
2120 * netif_napi_del - remove a NAPI context
2121 * @napi: NAPI context
2122 *
2123 * netif_napi_del() removes a NAPI context from the network device NAPI list
2124 */
2125void netif_napi_del(struct napi_struct *napi);
2126
2127struct napi_gro_cb {
2128 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2129 void *frag0;
2130
2131 /* Length of frag0. */
2132 unsigned int frag0_len;
2133
2134 /* This indicates where we are processing relative to skb->data. */
2135 int data_offset;
2136
2137 /* This is non-zero if the packet cannot be merged with the new skb. */
2138 u16 flush;
2139
2140 /* Save the IP ID here and check when we get to the transport layer */
2141 u16 flush_id;
2142
2143 /* Number of segments aggregated. */
2144 u16 count;
2145
2146 /* Start offset for remote checksum offload */
2147 u16 gro_remcsum_start;
2148
2149 /* jiffies when first packet was created/queued */
2150 unsigned long age;
2151
2152 /* Used in ipv6_gro_receive() and foo-over-udp */
2153 u16 proto;
2154
2155 /* This is non-zero if the packet may be of the same flow. */
2156 u8 same_flow:1;
2157
2158 /* Used in tunnel GRO receive */
2159 u8 encap_mark:1;
2160
2161 /* GRO checksum is valid */
2162 u8 csum_valid:1;
2163
2164 /* Number of checksums via CHECKSUM_UNNECESSARY */
2165 u8 csum_cnt:3;
2166
2167 /* Free the skb? */
2168 u8 free:2;
2169#define NAPI_GRO_FREE 1
2170#define NAPI_GRO_FREE_STOLEN_HEAD 2
2171
2172 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2173 u8 is_ipv6:1;
2174
2175 /* Used in GRE, set in fou/gue_gro_receive */
2176 u8 is_fou:1;
2177
2178 /* Used to determine if flush_id can be ignored */
2179 u8 is_atomic:1;
2180
2181 /* Number of gro_receive callbacks this packet already went through */
2182 u8 recursion_counter:4;
2183
2184 /* 1 bit hole */
2185
2186 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2187 __wsum csum;
2188
2189 /* used in skb_gro_receive() slow path */
2190 struct sk_buff *last;
2191};
2192
2193#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2194
2195#define GRO_RECURSION_LIMIT 15
2196static inline int gro_recursion_inc_test(struct sk_buff *skb)
2197{
2198 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2199}
2200
2201typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2202static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2203 struct sk_buff **head,
2204 struct sk_buff *skb)
2205{
2206 if (unlikely(gro_recursion_inc_test(skb))) {
2207 NAPI_GRO_CB(skb)->flush |= 1;
2208 return NULL;
2209 }
2210
2211 return cb(head, skb);
2212}
2213
2214typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2215 struct sk_buff *);
2216static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2217 struct sock *sk,
2218 struct sk_buff **head,
2219 struct sk_buff *skb)
2220{
2221 if (unlikely(gro_recursion_inc_test(skb))) {
2222 NAPI_GRO_CB(skb)->flush |= 1;
2223 return NULL;
2224 }
2225
2226 return cb(sk, head, skb);
2227}
2228
2229struct packet_type {
2230 __be16 type; /* This is really htons(ether_type). */
2231 struct net_device *dev; /* NULL is wildcarded here */
2232 int (*func) (struct sk_buff *,
2233 struct net_device *,
2234 struct packet_type *,
2235 struct net_device *);
2236 bool (*id_match)(struct packet_type *ptype,
2237 struct sock *sk);
2238 void *af_packet_priv;
2239 struct list_head list;
2240};
2241
2242struct offload_callbacks {
2243 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2244 netdev_features_t features);
2245 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2246 struct sk_buff *skb);
2247 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2248};
2249
2250struct packet_offload {
2251 __be16 type; /* This is really htons(ether_type). */
2252 u16 priority;
2253 struct offload_callbacks callbacks;
2254 struct list_head list;
2255};
2256
2257/* often modified stats are per-CPU, other are shared (netdev->stats) */
2258struct pcpu_sw_netstats {
2259 u64 rx_packets;
2260 u64 rx_bytes;
2261 u64 tx_packets;
2262 u64 tx_bytes;
2263 struct u64_stats_sync syncp;
2264};
2265
2266#define __netdev_alloc_pcpu_stats(type, gfp) \
2267({ \
2268 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2269 if (pcpu_stats) { \
2270 int __cpu; \
2271 for_each_possible_cpu(__cpu) { \
2272 typeof(type) *stat; \
2273 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2274 u64_stats_init(&stat->syncp); \
2275 } \
2276 } \
2277 pcpu_stats; \
2278})
2279
2280#define netdev_alloc_pcpu_stats(type) \
2281 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2282
2283enum netdev_lag_tx_type {
2284 NETDEV_LAG_TX_TYPE_UNKNOWN,
2285 NETDEV_LAG_TX_TYPE_RANDOM,
2286 NETDEV_LAG_TX_TYPE_BROADCAST,
2287 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2288 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2289 NETDEV_LAG_TX_TYPE_HASH,
2290};
2291
2292struct netdev_lag_upper_info {
2293 enum netdev_lag_tx_type tx_type;
2294};
2295
2296struct netdev_lag_lower_state_info {
2297 u8 link_up : 1,
2298 tx_enabled : 1;
2299};
2300
2301#include <linux/notifier.h>
2302
2303/* netdevice notifier chain. Please remember to update the rtnetlink
2304 * notification exclusion list in rtnetlink_event() when adding new
2305 * types.
2306 */
2307#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2308#define NETDEV_DOWN 0x0002
2309#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2310 detected a hardware crash and restarted
2311 - we can use this eg to kick tcp sessions
2312 once done */
2313#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2314#define NETDEV_REGISTER 0x0005
2315#define NETDEV_UNREGISTER 0x0006
2316#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2317#define NETDEV_CHANGEADDR 0x0008
2318#define NETDEV_GOING_DOWN 0x0009
2319#define NETDEV_CHANGENAME 0x000A
2320#define NETDEV_FEAT_CHANGE 0x000B
2321#define NETDEV_BONDING_FAILOVER 0x000C
2322#define NETDEV_PRE_UP 0x000D
2323#define NETDEV_PRE_TYPE_CHANGE 0x000E
2324#define NETDEV_POST_TYPE_CHANGE 0x000F
2325#define NETDEV_POST_INIT 0x0010
2326#define NETDEV_UNREGISTER_FINAL 0x0011
2327#define NETDEV_RELEASE 0x0012
2328#define NETDEV_NOTIFY_PEERS 0x0013
2329#define NETDEV_JOIN 0x0014
2330#define NETDEV_CHANGEUPPER 0x0015
2331#define NETDEV_RESEND_IGMP 0x0016
2332#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2333#define NETDEV_CHANGEINFODATA 0x0018
2334#define NETDEV_BONDING_INFO 0x0019
2335#define NETDEV_PRECHANGEUPPER 0x001A
2336#define NETDEV_CHANGELOWERSTATE 0x001B
2337#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2338#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2339#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2340
2341int register_netdevice_notifier(struct notifier_block *nb);
2342int unregister_netdevice_notifier(struct notifier_block *nb);
2343
2344struct netdev_notifier_info {
2345 struct net_device *dev;
2346 struct netlink_ext_ack *extack;
2347};
2348
2349struct netdev_notifier_change_info {
2350 struct netdev_notifier_info info; /* must be first */
2351 unsigned int flags_changed;
2352};
2353
2354struct netdev_notifier_changeupper_info {
2355 struct netdev_notifier_info info; /* must be first */
2356 struct net_device *upper_dev; /* new upper dev */
2357 bool master; /* is upper dev master */
2358 bool linking; /* is the notification for link or unlink */
2359 void *upper_info; /* upper dev info */
2360};
2361
2362struct netdev_notifier_changelowerstate_info {
2363 struct netdev_notifier_info info; /* must be first */
2364 void *lower_state_info; /* is lower dev state */
2365};
2366
2367static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2368 struct net_device *dev)
2369{
2370 info->dev = dev;
2371 info->extack = NULL;
2372}
2373
2374static inline struct net_device *
2375netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2376{
2377 return info->dev;
2378}
2379
2380static inline struct netlink_ext_ack *
2381netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2382{
2383 return info->extack;
2384}
2385
2386int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2387
2388
2389extern rwlock_t dev_base_lock; /* Device list lock */
2390
2391#define for_each_netdev(net, d) \
2392 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2393#define for_each_netdev_reverse(net, d) \
2394 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2395#define for_each_netdev_rcu(net, d) \
2396 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2397#define for_each_netdev_safe(net, d, n) \
2398 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2399#define for_each_netdev_continue(net, d) \
2400 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2401#define for_each_netdev_continue_rcu(net, d) \
2402 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2403#define for_each_netdev_in_bond_rcu(bond, slave) \
2404 for_each_netdev_rcu(&init_net, slave) \
2405 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2406#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2407
2408static inline struct net_device *next_net_device(struct net_device *dev)
2409{
2410 struct list_head *lh;
2411 struct net *net;
2412
2413 net = dev_net(dev);
2414 lh = dev->dev_list.next;
2415 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2416}
2417
2418static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2419{
2420 struct list_head *lh;
2421 struct net *net;
2422
2423 net = dev_net(dev);
2424 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2425 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2426}
2427
2428static inline struct net_device *first_net_device(struct net *net)
2429{
2430 return list_empty(&net->dev_base_head) ? NULL :
2431 net_device_entry(net->dev_base_head.next);
2432}
2433
2434static inline struct net_device *first_net_device_rcu(struct net *net)
2435{
2436 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2437
2438 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2439}
2440
2441int netdev_boot_setup_check(struct net_device *dev);
2442unsigned long netdev_boot_base(const char *prefix, int unit);
2443struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2444 const char *hwaddr);
2445struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2446struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2447void dev_add_pack(struct packet_type *pt);
2448void dev_remove_pack(struct packet_type *pt);
2449void __dev_remove_pack(struct packet_type *pt);
2450void dev_add_offload(struct packet_offload *po);
2451void dev_remove_offload(struct packet_offload *po);
2452
2453int dev_get_iflink(const struct net_device *dev);
2454int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2455struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2456 unsigned short mask);
2457struct net_device *dev_get_by_name(struct net *net, const char *name);
2458struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2459struct net_device *__dev_get_by_name(struct net *net, const char *name);
2460int dev_alloc_name(struct net_device *dev, const char *name);
2461int dev_open(struct net_device *dev);
2462void dev_close(struct net_device *dev);
2463void dev_close_many(struct list_head *head, bool unlink);
2464void dev_disable_lro(struct net_device *dev);
2465int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2466int dev_queue_xmit(struct sk_buff *skb);
2467int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2468int register_netdevice(struct net_device *dev);
2469void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2470void unregister_netdevice_many(struct list_head *head);
2471static inline void unregister_netdevice(struct net_device *dev)
2472{
2473 unregister_netdevice_queue(dev, NULL);
2474}
2475
2476int netdev_refcnt_read(const struct net_device *dev);
2477void free_netdev(struct net_device *dev);
2478void netdev_freemem(struct net_device *dev);
2479void synchronize_net(void);
2480int init_dummy_netdev(struct net_device *dev);
2481
2482DECLARE_PER_CPU(int, xmit_recursion);
2483#define XMIT_RECURSION_LIMIT 10
2484
2485static inline int dev_recursion_level(void)
2486{
2487 return this_cpu_read(xmit_recursion);
2488}
2489
2490struct net_device *dev_get_by_index(struct net *net, int ifindex);
2491struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2492struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2493struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2494int netdev_get_name(struct net *net, char *name, int ifindex);
2495int dev_restart(struct net_device *dev);
2496int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2497
2498static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2499{
2500 return NAPI_GRO_CB(skb)->data_offset;
2501}
2502
2503static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2504{
2505 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2506}
2507
2508static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2509{
2510 NAPI_GRO_CB(skb)->data_offset += len;
2511}
2512
2513static inline void *skb_gro_header_fast(struct sk_buff *skb,
2514 unsigned int offset)
2515{
2516 return NAPI_GRO_CB(skb)->frag0 + offset;
2517}
2518
2519static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2520{
2521 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2522}
2523
2524static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2525{
2526 NAPI_GRO_CB(skb)->frag0 = NULL;
2527 NAPI_GRO_CB(skb)->frag0_len = 0;
2528}
2529
2530static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2531 unsigned int offset)
2532{
2533 if (!pskb_may_pull(skb, hlen))
2534 return NULL;
2535
2536 skb_gro_frag0_invalidate(skb);
2537 return skb->data + offset;
2538}
2539
2540static inline void *skb_gro_network_header(struct sk_buff *skb)
2541{
2542 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2543 skb_network_offset(skb);
2544}
2545
2546static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2547 const void *start, unsigned int len)
2548{
2549 if (NAPI_GRO_CB(skb)->csum_valid)
2550 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2551 csum_partial(start, len, 0));
2552}
2553
2554/* GRO checksum functions. These are logical equivalents of the normal
2555 * checksum functions (in skbuff.h) except that they operate on the GRO
2556 * offsets and fields in sk_buff.
2557 */
2558
2559__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2560
2561static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2562{
2563 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2564}
2565
2566static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2567 bool zero_okay,
2568 __sum16 check)
2569{
2570 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2571 skb_checksum_start_offset(skb) <
2572 skb_gro_offset(skb)) &&
2573 !skb_at_gro_remcsum_start(skb) &&
2574 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2575 (!zero_okay || check));
2576}
2577
2578static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2579 __wsum psum)
2580{
2581 if (NAPI_GRO_CB(skb)->csum_valid &&
2582 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2583 return 0;
2584
2585 NAPI_GRO_CB(skb)->csum = psum;
2586
2587 return __skb_gro_checksum_complete(skb);
2588}
2589
2590static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2591{
2592 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2593 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2594 NAPI_GRO_CB(skb)->csum_cnt--;
2595 } else {
2596 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2597 * verified a new top level checksum or an encapsulated one
2598 * during GRO. This saves work if we fallback to normal path.
2599 */
2600 __skb_incr_checksum_unnecessary(skb);
2601 }
2602}
2603
2604#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2605 compute_pseudo) \
2606({ \
2607 __sum16 __ret = 0; \
2608 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2609 __ret = __skb_gro_checksum_validate_complete(skb, \
2610 compute_pseudo(skb, proto)); \
2611 if (!__ret) \
2612 skb_gro_incr_csum_unnecessary(skb); \
2613 __ret; \
2614})
2615
2616#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2617 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2618
2619#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2620 compute_pseudo) \
2621 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2622
2623#define skb_gro_checksum_simple_validate(skb) \
2624 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2625
2626static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2627{
2628 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2629 !NAPI_GRO_CB(skb)->csum_valid);
2630}
2631
2632static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2633 __sum16 check, __wsum pseudo)
2634{
2635 NAPI_GRO_CB(skb)->csum = ~pseudo;
2636 NAPI_GRO_CB(skb)->csum_valid = 1;
2637}
2638
2639#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2640do { \
2641 if (__skb_gro_checksum_convert_check(skb)) \
2642 __skb_gro_checksum_convert(skb, check, \
2643 compute_pseudo(skb, proto)); \
2644} while (0)
2645
2646struct gro_remcsum {
2647 int offset;
2648 __wsum delta;
2649};
2650
2651static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2652{
2653 grc->offset = 0;
2654 grc->delta = 0;
2655}
2656
2657static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2658 unsigned int off, size_t hdrlen,
2659 int start, int offset,
2660 struct gro_remcsum *grc,
2661 bool nopartial)
2662{
2663 __wsum delta;
2664 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2665
2666 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2667
2668 if (!nopartial) {
2669 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2670 return ptr;
2671 }
2672
2673 ptr = skb_gro_header_fast(skb, off);
2674 if (skb_gro_header_hard(skb, off + plen)) {
2675 ptr = skb_gro_header_slow(skb, off + plen, off);
2676 if (!ptr)
2677 return NULL;
2678 }
2679
2680 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2681 start, offset);
2682
2683 /* Adjust skb->csum since we changed the packet */
2684 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2685
2686 grc->offset = off + hdrlen + offset;
2687 grc->delta = delta;
2688
2689 return ptr;
2690}
2691
2692static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2693 struct gro_remcsum *grc)
2694{
2695 void *ptr;
2696 size_t plen = grc->offset + sizeof(u16);
2697
2698 if (!grc->delta)
2699 return;
2700
2701 ptr = skb_gro_header_fast(skb, grc->offset);
2702 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2703 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2704 if (!ptr)
2705 return;
2706 }
2707
2708 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2709}
2710
2711#ifdef CONFIG_XFRM_OFFLOAD
2712static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2713{
2714 if (PTR_ERR(pp) != -EINPROGRESS)
2715 NAPI_GRO_CB(skb)->flush |= flush;
2716}
2717#else
2718static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2719{
2720 NAPI_GRO_CB(skb)->flush |= flush;
2721}
2722#endif
2723
2724static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2725 unsigned short type,
2726 const void *daddr, const void *saddr,
2727 unsigned int len)
2728{
2729 if (!dev->header_ops || !dev->header_ops->create)
2730 return 0;
2731
2732 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2733}
2734
2735static inline int dev_parse_header(const struct sk_buff *skb,
2736 unsigned char *haddr)
2737{
2738 const struct net_device *dev = skb->dev;
2739
2740 if (!dev->header_ops || !dev->header_ops->parse)
2741 return 0;
2742 return dev->header_ops->parse(skb, haddr);
2743}
2744
2745/* ll_header must have at least hard_header_len allocated */
2746static inline bool dev_validate_header(const struct net_device *dev,
2747 char *ll_header, int len)
2748{
2749 if (likely(len >= dev->hard_header_len))
2750 return true;
2751 if (len < dev->min_header_len)
2752 return false;
2753
2754 if (capable(CAP_SYS_RAWIO)) {
2755 memset(ll_header + len, 0, dev->hard_header_len - len);
2756 return true;
2757 }
2758
2759 if (dev->header_ops && dev->header_ops->validate)
2760 return dev->header_ops->validate(ll_header, len);
2761
2762 return false;
2763}
2764
2765typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2766 int len, int size);
2767int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2768static inline int unregister_gifconf(unsigned int family)
2769{
2770 return register_gifconf(family, NULL);
2771}
2772
2773#ifdef CONFIG_NET_FLOW_LIMIT
2774#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2775struct sd_flow_limit {
2776 u64 count;
2777 unsigned int num_buckets;
2778 unsigned int history_head;
2779 u16 history[FLOW_LIMIT_HISTORY];
2780 u8 buckets[];
2781};
2782
2783extern int netdev_flow_limit_table_len;
2784#endif /* CONFIG_NET_FLOW_LIMIT */
2785
2786/*
2787 * Incoming packets are placed on per-CPU queues
2788 */
2789struct softnet_data {
2790 struct list_head poll_list;
2791 struct sk_buff_head process_queue;
2792
2793 /* stats */
2794 unsigned int processed;
2795 unsigned int time_squeeze;
2796 unsigned int received_rps;
2797#ifdef CONFIG_RPS
2798 struct softnet_data *rps_ipi_list;
2799#endif
2800#ifdef CONFIG_NET_FLOW_LIMIT
2801 struct sd_flow_limit __rcu *flow_limit;
2802#endif
2803 struct Qdisc *output_queue;
2804 struct Qdisc **output_queue_tailp;
2805 struct sk_buff *completion_queue;
2806#ifdef CONFIG_XFRM_OFFLOAD
2807 struct sk_buff_head xfrm_backlog;
2808#endif
2809#ifdef CONFIG_RPS
2810 /* input_queue_head should be written by cpu owning this struct,
2811 * and only read by other cpus. Worth using a cache line.
2812 */
2813 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2814
2815 /* Elements below can be accessed between CPUs for RPS/RFS */
2816 call_single_data_t csd ____cacheline_aligned_in_smp;
2817 struct softnet_data *rps_ipi_next;
2818 unsigned int cpu;
2819 unsigned int input_queue_tail;
2820#endif
2821 unsigned int dropped;
2822 struct sk_buff_head input_pkt_queue;
2823 struct napi_struct backlog;
2824
2825};
2826
2827static inline void input_queue_head_incr(struct softnet_data *sd)
2828{
2829#ifdef CONFIG_RPS
2830 sd->input_queue_head++;
2831#endif
2832}
2833
2834static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2835 unsigned int *qtail)
2836{
2837#ifdef CONFIG_RPS
2838 *qtail = ++sd->input_queue_tail;
2839#endif
2840}
2841
2842DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2843
2844void __netif_schedule(struct Qdisc *q);
2845void netif_schedule_queue(struct netdev_queue *txq);
2846
2847static inline void netif_tx_schedule_all(struct net_device *dev)
2848{
2849 unsigned int i;
2850
2851 for (i = 0; i < dev->num_tx_queues; i++)
2852 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2853}
2854
2855static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2856{
2857 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2858}
2859
2860/**
2861 * netif_start_queue - allow transmit
2862 * @dev: network device
2863 *
2864 * Allow upper layers to call the device hard_start_xmit routine.
2865 */
2866static inline void netif_start_queue(struct net_device *dev)
2867{
2868 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2869}
2870
2871static inline void netif_tx_start_all_queues(struct net_device *dev)
2872{
2873 unsigned int i;
2874
2875 for (i = 0; i < dev->num_tx_queues; i++) {
2876 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2877 netif_tx_start_queue(txq);
2878 }
2879}
2880
2881void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2882
2883/**
2884 * netif_wake_queue - restart transmit
2885 * @dev: network device
2886 *
2887 * Allow upper layers to call the device hard_start_xmit routine.
2888 * Used for flow control when transmit resources are available.
2889 */
2890static inline void netif_wake_queue(struct net_device *dev)
2891{
2892 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2893}
2894
2895static inline void netif_tx_wake_all_queues(struct net_device *dev)
2896{
2897 unsigned int i;
2898
2899 for (i = 0; i < dev->num_tx_queues; i++) {
2900 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2901 netif_tx_wake_queue(txq);
2902 }
2903}
2904
2905static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2906{
2907 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2908}
2909
2910/**
2911 * netif_stop_queue - stop transmitted packets
2912 * @dev: network device
2913 *
2914 * Stop upper layers calling the device hard_start_xmit routine.
2915 * Used for flow control when transmit resources are unavailable.
2916 */
2917static inline void netif_stop_queue(struct net_device *dev)
2918{
2919 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2920}
2921
2922void netif_tx_stop_all_queues(struct net_device *dev);
2923
2924static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2925{
2926 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2927}
2928
2929/**
2930 * netif_queue_stopped - test if transmit queue is flowblocked
2931 * @dev: network device
2932 *
2933 * Test if transmit queue on device is currently unable to send.
2934 */
2935static inline bool netif_queue_stopped(const struct net_device *dev)
2936{
2937 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2938}
2939
2940static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2941{
2942 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2943}
2944
2945static inline bool
2946netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2947{
2948 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2949}
2950
2951static inline bool
2952netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2953{
2954 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2955}
2956
2957/**
2958 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2959 * @dev_queue: pointer to transmit queue
2960 *
2961 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2962 * to give appropriate hint to the CPU.
2963 */
2964static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2965{
2966#ifdef CONFIG_BQL
2967 prefetchw(&dev_queue->dql.num_queued);
2968#endif
2969}
2970
2971/**
2972 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2973 * @dev_queue: pointer to transmit queue
2974 *
2975 * BQL enabled drivers might use this helper in their TX completion path,
2976 * to give appropriate hint to the CPU.
2977 */
2978static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2979{
2980#ifdef CONFIG_BQL
2981 prefetchw(&dev_queue->dql.limit);
2982#endif
2983}
2984
2985static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2986 unsigned int bytes)
2987{
2988#ifdef CONFIG_BQL
2989 dql_queued(&dev_queue->dql, bytes);
2990
2991 if (likely(dql_avail(&dev_queue->dql) >= 0))
2992 return;
2993
2994 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2995
2996 /*
2997 * The XOFF flag must be set before checking the dql_avail below,
2998 * because in netdev_tx_completed_queue we update the dql_completed
2999 * before checking the XOFF flag.
3000 */
3001 smp_mb();
3002
3003 /* check again in case another CPU has just made room avail */
3004 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3005 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3006#endif
3007}
3008
3009/**
3010 * netdev_sent_queue - report the number of bytes queued to hardware
3011 * @dev: network device
3012 * @bytes: number of bytes queued to the hardware device queue
3013 *
3014 * Report the number of bytes queued for sending/completion to the network
3015 * device hardware queue. @bytes should be a good approximation and should
3016 * exactly match netdev_completed_queue() @bytes
3017 */
3018static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3019{
3020 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3021}
3022
3023static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3024 unsigned int pkts, unsigned int bytes)
3025{
3026#ifdef CONFIG_BQL
3027 if (unlikely(!bytes))
3028 return;
3029
3030 dql_completed(&dev_queue->dql, bytes);
3031
3032 /*
3033 * Without the memory barrier there is a small possiblity that
3034 * netdev_tx_sent_queue will miss the update and cause the queue to
3035 * be stopped forever
3036 */
3037 smp_mb();
3038
3039 if (dql_avail(&dev_queue->dql) < 0)
3040 return;
3041
3042 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3043 netif_schedule_queue(dev_queue);
3044#endif
3045}
3046
3047/**
3048 * netdev_completed_queue - report bytes and packets completed by device
3049 * @dev: network device
3050 * @pkts: actual number of packets sent over the medium
3051 * @bytes: actual number of bytes sent over the medium
3052 *
3053 * Report the number of bytes and packets transmitted by the network device
3054 * hardware queue over the physical medium, @bytes must exactly match the
3055 * @bytes amount passed to netdev_sent_queue()
3056 */
3057static inline void netdev_completed_queue(struct net_device *dev,
3058 unsigned int pkts, unsigned int bytes)
3059{
3060 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3061}
3062
3063static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3064{
3065#ifdef CONFIG_BQL
3066 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3067 dql_reset(&q->dql);
3068#endif
3069}
3070
3071/**
3072 * netdev_reset_queue - reset the packets and bytes count of a network device
3073 * @dev_queue: network device
3074 *
3075 * Reset the bytes and packet count of a network device and clear the
3076 * software flow control OFF bit for this network device
3077 */
3078static inline void netdev_reset_queue(struct net_device *dev_queue)
3079{
3080 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3081}
3082
3083/**
3084 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3085 * @dev: network device
3086 * @queue_index: given tx queue index
3087 *
3088 * Returns 0 if given tx queue index >= number of device tx queues,
3089 * otherwise returns the originally passed tx queue index.
3090 */
3091static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3092{
3093 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3094 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3095 dev->name, queue_index,
3096 dev->real_num_tx_queues);
3097 return 0;
3098 }
3099
3100 return queue_index;
3101}
3102
3103/**
3104 * netif_running - test if up
3105 * @dev: network device
3106 *
3107 * Test if the device has been brought up.
3108 */
3109static inline bool netif_running(const struct net_device *dev)
3110{
3111 return test_bit(__LINK_STATE_START, &dev->state);
3112}
3113
3114/*
3115 * Routines to manage the subqueues on a device. We only need start,
3116 * stop, and a check if it's stopped. All other device management is
3117 * done at the overall netdevice level.
3118 * Also test the device if we're multiqueue.
3119 */
3120
3121/**
3122 * netif_start_subqueue - allow sending packets on subqueue
3123 * @dev: network device
3124 * @queue_index: sub queue index
3125 *
3126 * Start individual transmit queue of a device with multiple transmit queues.
3127 */
3128static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3129{
3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3131
3132 netif_tx_start_queue(txq);
3133}
3134
3135/**
3136 * netif_stop_subqueue - stop sending packets on subqueue
3137 * @dev: network device
3138 * @queue_index: sub queue index
3139 *
3140 * Stop individual transmit queue of a device with multiple transmit queues.
3141 */
3142static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3143{
3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3145 netif_tx_stop_queue(txq);
3146}
3147
3148/**
3149 * netif_subqueue_stopped - test status of subqueue
3150 * @dev: network device
3151 * @queue_index: sub queue index
3152 *
3153 * Check individual transmit queue of a device with multiple transmit queues.
3154 */
3155static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3156 u16 queue_index)
3157{
3158 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3159
3160 return netif_tx_queue_stopped(txq);
3161}
3162
3163static inline bool netif_subqueue_stopped(const struct net_device *dev,
3164 struct sk_buff *skb)
3165{
3166 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3167}
3168
3169/**
3170 * netif_wake_subqueue - allow sending packets on subqueue
3171 * @dev: network device
3172 * @queue_index: sub queue index
3173 *
3174 * Resume individual transmit queue of a device with multiple transmit queues.
3175 */
3176static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3177{
3178 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3179
3180 netif_tx_wake_queue(txq);
3181}
3182
3183#ifdef CONFIG_XPS
3184int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3185 u16 index);
3186#else
3187static inline int netif_set_xps_queue(struct net_device *dev,
3188 const struct cpumask *mask,
3189 u16 index)
3190{
3191 return 0;
3192}
3193#endif
3194
3195u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3196 unsigned int num_tx_queues);
3197
3198/*
3199 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3200 * as a distribution range limit for the returned value.
3201 */
3202static inline u16 skb_tx_hash(const struct net_device *dev,
3203 struct sk_buff *skb)
3204{
3205 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3206}
3207
3208/**
3209 * netif_is_multiqueue - test if device has multiple transmit queues
3210 * @dev: network device
3211 *
3212 * Check if device has multiple transmit queues
3213 */
3214static inline bool netif_is_multiqueue(const struct net_device *dev)
3215{
3216 return dev->num_tx_queues > 1;
3217}
3218
3219int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3220
3221#ifdef CONFIG_SYSFS
3222int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3223#else
3224static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3225 unsigned int rxq)
3226{
3227 return 0;
3228}
3229#endif
3230
3231static inline struct netdev_rx_queue *
3232__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3233{
3234 return dev->_rx + rxq;
3235}
3236
3237#ifdef CONFIG_SYSFS
3238static inline unsigned int get_netdev_rx_queue_index(
3239 struct netdev_rx_queue *queue)
3240{
3241 struct net_device *dev = queue->dev;
3242 int index = queue - dev->_rx;
3243
3244 BUG_ON(index >= dev->num_rx_queues);
3245 return index;
3246}
3247#endif
3248
3249#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3250int netif_get_num_default_rss_queues(void);
3251
3252enum skb_free_reason {
3253 SKB_REASON_CONSUMED,
3254 SKB_REASON_DROPPED,
3255};
3256
3257void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3258void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3259
3260/*
3261 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3262 * interrupt context or with hardware interrupts being disabled.
3263 * (in_irq() || irqs_disabled())
3264 *
3265 * We provide four helpers that can be used in following contexts :
3266 *
3267 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3268 * replacing kfree_skb(skb)
3269 *
3270 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3271 * Typically used in place of consume_skb(skb) in TX completion path
3272 *
3273 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3274 * replacing kfree_skb(skb)
3275 *
3276 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3277 * and consumed a packet. Used in place of consume_skb(skb)
3278 */
3279static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3280{
3281 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3282}
3283
3284static inline void dev_consume_skb_irq(struct sk_buff *skb)
3285{
3286 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3287}
3288
3289static inline void dev_kfree_skb_any(struct sk_buff *skb)
3290{
3291 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3292}
3293
3294static inline void dev_consume_skb_any(struct sk_buff *skb)
3295{
3296 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3297}
3298
3299void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3300int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3301int netif_rx(struct sk_buff *skb);
3302int netif_rx_ni(struct sk_buff *skb);
3303int netif_receive_skb(struct sk_buff *skb);
3304int netif_receive_skb_core(struct sk_buff *skb);
3305gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3306void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3307struct sk_buff *napi_get_frags(struct napi_struct *napi);
3308gro_result_t napi_gro_frags(struct napi_struct *napi);
3309struct packet_offload *gro_find_receive_by_type(__be16 type);
3310struct packet_offload *gro_find_complete_by_type(__be16 type);
3311
3312static inline void napi_free_frags(struct napi_struct *napi)
3313{
3314 kfree_skb(napi->skb);
3315 napi->skb = NULL;
3316}
3317
3318bool netdev_is_rx_handler_busy(struct net_device *dev);
3319int netdev_rx_handler_register(struct net_device *dev,
3320 rx_handler_func_t *rx_handler,
3321 void *rx_handler_data);
3322void netdev_rx_handler_unregister(struct net_device *dev);
3323
3324bool dev_valid_name(const char *name);
3325int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3326 bool *need_copyout);
3327int dev_ifconf(struct net *net, struct ifconf *, int);
3328int dev_ethtool(struct net *net, struct ifreq *);
3329unsigned int dev_get_flags(const struct net_device *);
3330int __dev_change_flags(struct net_device *, unsigned int flags);
3331int dev_change_flags(struct net_device *, unsigned int);
3332void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3333 unsigned int gchanges);
3334int dev_change_name(struct net_device *, const char *);
3335int dev_set_alias(struct net_device *, const char *, size_t);
3336int dev_get_alias(const struct net_device *, char *, size_t);
3337int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3338int __dev_set_mtu(struct net_device *, int);
3339int dev_set_mtu(struct net_device *, int);
3340int dev_change_tx_queue_len(struct net_device *, unsigned long);
3341void dev_set_group(struct net_device *, int);
3342int dev_set_mac_address(struct net_device *, struct sockaddr *);
3343int dev_change_carrier(struct net_device *, bool new_carrier);
3344int dev_get_phys_port_id(struct net_device *dev,
3345 struct netdev_phys_item_id *ppid);
3346int dev_get_phys_port_name(struct net_device *dev,
3347 char *name, size_t len);
3348int dev_change_proto_down(struct net_device *dev, bool proto_down);
3349struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3350struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3351 struct netdev_queue *txq, int *ret);
3352
3353typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3354int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3355 int fd, u32 flags);
3356void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3357 struct netdev_bpf *xdp);
3358
3359int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3360int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3361bool is_skb_forwardable(const struct net_device *dev,
3362 const struct sk_buff *skb);
3363
3364static __always_inline int ____dev_forward_skb(struct net_device *dev,
3365 struct sk_buff *skb)
3366{
3367 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3368 unlikely(!is_skb_forwardable(dev, skb))) {
3369 atomic_long_inc(&dev->rx_dropped);
3370 kfree_skb(skb);
3371 return NET_RX_DROP;
3372 }
3373
3374 skb_scrub_packet(skb, true);
3375 skb->priority = 0;
3376 return 0;
3377}
3378
3379void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3380
3381extern int netdev_budget;
3382extern unsigned int netdev_budget_usecs;
3383
3384/* Called by rtnetlink.c:rtnl_unlock() */
3385void netdev_run_todo(void);
3386
3387/**
3388 * dev_put - release reference to device
3389 * @dev: network device
3390 *
3391 * Release reference to device to allow it to be freed.
3392 */
3393static inline void dev_put(struct net_device *dev)
3394{
3395 this_cpu_dec(*dev->pcpu_refcnt);
3396}
3397
3398/**
3399 * dev_hold - get reference to device
3400 * @dev: network device
3401 *
3402 * Hold reference to device to keep it from being freed.
3403 */
3404static inline void dev_hold(struct net_device *dev)
3405{
3406 this_cpu_inc(*dev->pcpu_refcnt);
3407}
3408
3409/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3410 * and _off may be called from IRQ context, but it is caller
3411 * who is responsible for serialization of these calls.
3412 *
3413 * The name carrier is inappropriate, these functions should really be
3414 * called netif_lowerlayer_*() because they represent the state of any
3415 * kind of lower layer not just hardware media.
3416 */
3417
3418void linkwatch_init_dev(struct net_device *dev);
3419void linkwatch_fire_event(struct net_device *dev);
3420void linkwatch_forget_dev(struct net_device *dev);
3421
3422/**
3423 * netif_carrier_ok - test if carrier present
3424 * @dev: network device
3425 *
3426 * Check if carrier is present on device
3427 */
3428static inline bool netif_carrier_ok(const struct net_device *dev)
3429{
3430 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3431}
3432
3433unsigned long dev_trans_start(struct net_device *dev);
3434
3435void __netdev_watchdog_up(struct net_device *dev);
3436
3437void netif_carrier_on(struct net_device *dev);
3438
3439void netif_carrier_off(struct net_device *dev);
3440
3441/**
3442 * netif_dormant_on - mark device as dormant.
3443 * @dev: network device
3444 *
3445 * Mark device as dormant (as per RFC2863).
3446 *
3447 * The dormant state indicates that the relevant interface is not
3448 * actually in a condition to pass packets (i.e., it is not 'up') but is
3449 * in a "pending" state, waiting for some external event. For "on-
3450 * demand" interfaces, this new state identifies the situation where the
3451 * interface is waiting for events to place it in the up state.
3452 */
3453static inline void netif_dormant_on(struct net_device *dev)
3454{
3455 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3456 linkwatch_fire_event(dev);
3457}
3458
3459/**
3460 * netif_dormant_off - set device as not dormant.
3461 * @dev: network device
3462 *
3463 * Device is not in dormant state.
3464 */
3465static inline void netif_dormant_off(struct net_device *dev)
3466{
3467 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3468 linkwatch_fire_event(dev);
3469}
3470
3471/**
3472 * netif_dormant - test if device is dormant
3473 * @dev: network device
3474 *
3475 * Check if device is dormant.
3476 */
3477static inline bool netif_dormant(const struct net_device *dev)
3478{
3479 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3480}
3481
3482
3483/**
3484 * netif_oper_up - test if device is operational
3485 * @dev: network device
3486 *
3487 * Check if carrier is operational
3488 */
3489static inline bool netif_oper_up(const struct net_device *dev)
3490{
3491 return (dev->operstate == IF_OPER_UP ||
3492 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3493}
3494
3495/**
3496 * netif_device_present - is device available or removed
3497 * @dev: network device
3498 *
3499 * Check if device has not been removed from system.
3500 */
3501static inline bool netif_device_present(struct net_device *dev)
3502{
3503 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3504}
3505
3506void netif_device_detach(struct net_device *dev);
3507
3508void netif_device_attach(struct net_device *dev);
3509
3510/*
3511 * Network interface message level settings
3512 */
3513
3514enum {
3515 NETIF_MSG_DRV = 0x0001,
3516 NETIF_MSG_PROBE = 0x0002,
3517 NETIF_MSG_LINK = 0x0004,
3518 NETIF_MSG_TIMER = 0x0008,
3519 NETIF_MSG_IFDOWN = 0x0010,
3520 NETIF_MSG_IFUP = 0x0020,
3521 NETIF_MSG_RX_ERR = 0x0040,
3522 NETIF_MSG_TX_ERR = 0x0080,
3523 NETIF_MSG_TX_QUEUED = 0x0100,
3524 NETIF_MSG_INTR = 0x0200,
3525 NETIF_MSG_TX_DONE = 0x0400,
3526 NETIF_MSG_RX_STATUS = 0x0800,
3527 NETIF_MSG_PKTDATA = 0x1000,
3528 NETIF_MSG_HW = 0x2000,
3529 NETIF_MSG_WOL = 0x4000,
3530};
3531
3532#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3533#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3534#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3535#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3536#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3537#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3538#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3539#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3540#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3541#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3542#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3543#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3544#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3545#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3546#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3547
3548static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3549{
3550 /* use default */
3551 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3552 return default_msg_enable_bits;
3553 if (debug_value == 0) /* no output */
3554 return 0;
3555 /* set low N bits */
3556 return (1 << debug_value) - 1;
3557}
3558
3559static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3560{
3561 spin_lock(&txq->_xmit_lock);
3562 txq->xmit_lock_owner = cpu;
3563}
3564
3565static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3566{
3567 __acquire(&txq->_xmit_lock);
3568 return true;
3569}
3570
3571static inline void __netif_tx_release(struct netdev_queue *txq)
3572{
3573 __release(&txq->_xmit_lock);
3574}
3575
3576static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3577{
3578 spin_lock_bh(&txq->_xmit_lock);
3579 txq->xmit_lock_owner = smp_processor_id();
3580}
3581
3582static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3583{
3584 bool ok = spin_trylock(&txq->_xmit_lock);
3585 if (likely(ok))
3586 txq->xmit_lock_owner = smp_processor_id();
3587 return ok;
3588}
3589
3590static inline void __netif_tx_unlock(struct netdev_queue *txq)
3591{
3592 txq->xmit_lock_owner = -1;
3593 spin_unlock(&txq->_xmit_lock);
3594}
3595
3596static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3597{
3598 txq->xmit_lock_owner = -1;
3599 spin_unlock_bh(&txq->_xmit_lock);
3600}
3601
3602static inline void txq_trans_update(struct netdev_queue *txq)
3603{
3604 if (txq->xmit_lock_owner != -1)
3605 txq->trans_start = jiffies;
3606}
3607
3608/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3609static inline void netif_trans_update(struct net_device *dev)
3610{
3611 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3612
3613 if (txq->trans_start != jiffies)
3614 txq->trans_start = jiffies;
3615}
3616
3617/**
3618 * netif_tx_lock - grab network device transmit lock
3619 * @dev: network device
3620 *
3621 * Get network device transmit lock
3622 */
3623static inline void netif_tx_lock(struct net_device *dev)
3624{
3625 unsigned int i;
3626 int cpu;
3627
3628 spin_lock(&dev->tx_global_lock);
3629 cpu = smp_processor_id();
3630 for (i = 0; i < dev->num_tx_queues; i++) {
3631 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3632
3633 /* We are the only thread of execution doing a
3634 * freeze, but we have to grab the _xmit_lock in
3635 * order to synchronize with threads which are in
3636 * the ->hard_start_xmit() handler and already
3637 * checked the frozen bit.
3638 */
3639 __netif_tx_lock(txq, cpu);
3640 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3641 __netif_tx_unlock(txq);
3642 }
3643}
3644
3645static inline void netif_tx_lock_bh(struct net_device *dev)
3646{
3647 local_bh_disable();
3648 netif_tx_lock(dev);
3649}
3650
3651static inline void netif_tx_unlock(struct net_device *dev)
3652{
3653 unsigned int i;
3654
3655 for (i = 0; i < dev->num_tx_queues; i++) {
3656 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3657
3658 /* No need to grab the _xmit_lock here. If the
3659 * queue is not stopped for another reason, we
3660 * force a schedule.
3661 */
3662 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3663 netif_schedule_queue(txq);
3664 }
3665 spin_unlock(&dev->tx_global_lock);
3666}
3667
3668static inline void netif_tx_unlock_bh(struct net_device *dev)
3669{
3670 netif_tx_unlock(dev);
3671 local_bh_enable();
3672}
3673
3674#define HARD_TX_LOCK(dev, txq, cpu) { \
3675 if ((dev->features & NETIF_F_LLTX) == 0) { \
3676 __netif_tx_lock(txq, cpu); \
3677 } else { \
3678 __netif_tx_acquire(txq); \
3679 } \
3680}
3681
3682#define HARD_TX_TRYLOCK(dev, txq) \
3683 (((dev->features & NETIF_F_LLTX) == 0) ? \
3684 __netif_tx_trylock(txq) : \
3685 __netif_tx_acquire(txq))
3686
3687#define HARD_TX_UNLOCK(dev, txq) { \
3688 if ((dev->features & NETIF_F_LLTX) == 0) { \
3689 __netif_tx_unlock(txq); \
3690 } else { \
3691 __netif_tx_release(txq); \
3692 } \
3693}
3694
3695static inline void netif_tx_disable(struct net_device *dev)
3696{
3697 unsigned int i;
3698 int cpu;
3699
3700 local_bh_disable();
3701 cpu = smp_processor_id();
3702 for (i = 0; i < dev->num_tx_queues; i++) {
3703 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3704
3705 __netif_tx_lock(txq, cpu);
3706 netif_tx_stop_queue(txq);
3707 __netif_tx_unlock(txq);
3708 }
3709 local_bh_enable();
3710}
3711
3712static inline void netif_addr_lock(struct net_device *dev)
3713{
3714 spin_lock(&dev->addr_list_lock);
3715}
3716
3717static inline void netif_addr_lock_nested(struct net_device *dev)
3718{
3719 int subclass = SINGLE_DEPTH_NESTING;
3720
3721 if (dev->netdev_ops->ndo_get_lock_subclass)
3722 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3723
3724 spin_lock_nested(&dev->addr_list_lock, subclass);
3725}
3726
3727static inline void netif_addr_lock_bh(struct net_device *dev)
3728{
3729 spin_lock_bh(&dev->addr_list_lock);
3730}
3731
3732static inline void netif_addr_unlock(struct net_device *dev)
3733{
3734 spin_unlock(&dev->addr_list_lock);
3735}
3736
3737static inline void netif_addr_unlock_bh(struct net_device *dev)
3738{
3739 spin_unlock_bh(&dev->addr_list_lock);
3740}
3741
3742/*
3743 * dev_addrs walker. Should be used only for read access. Call with
3744 * rcu_read_lock held.
3745 */
3746#define for_each_dev_addr(dev, ha) \
3747 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3748
3749/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3750
3751void ether_setup(struct net_device *dev);
3752
3753/* Support for loadable net-drivers */
3754struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3755 unsigned char name_assign_type,
3756 void (*setup)(struct net_device *),
3757 unsigned int txqs, unsigned int rxqs);
3758int dev_get_valid_name(struct net *net, struct net_device *dev,
3759 const char *name);
3760
3761#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3762 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3763
3764#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3765 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3766 count)
3767
3768int register_netdev(struct net_device *dev);
3769void unregister_netdev(struct net_device *dev);
3770
3771/* General hardware address lists handling functions */
3772int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3773 struct netdev_hw_addr_list *from_list, int addr_len);
3774void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3775 struct netdev_hw_addr_list *from_list, int addr_len);
3776int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3777 struct net_device *dev,
3778 int (*sync)(struct net_device *, const unsigned char *),
3779 int (*unsync)(struct net_device *,
3780 const unsigned char *));
3781void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3782 struct net_device *dev,
3783 int (*unsync)(struct net_device *,
3784 const unsigned char *));
3785void __hw_addr_init(struct netdev_hw_addr_list *list);
3786
3787/* Functions used for device addresses handling */
3788int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3789 unsigned char addr_type);
3790int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3791 unsigned char addr_type);
3792void dev_addr_flush(struct net_device *dev);
3793int dev_addr_init(struct net_device *dev);
3794
3795/* Functions used for unicast addresses handling */
3796int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3797int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3798int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3799int dev_uc_sync(struct net_device *to, struct net_device *from);
3800int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3801void dev_uc_unsync(struct net_device *to, struct net_device *from);
3802void dev_uc_flush(struct net_device *dev);
3803void dev_uc_init(struct net_device *dev);
3804
3805/**
3806 * __dev_uc_sync - Synchonize device's unicast list
3807 * @dev: device to sync
3808 * @sync: function to call if address should be added
3809 * @unsync: function to call if address should be removed
3810 *
3811 * Add newly added addresses to the interface, and release
3812 * addresses that have been deleted.
3813 */
3814static inline int __dev_uc_sync(struct net_device *dev,
3815 int (*sync)(struct net_device *,
3816 const unsigned char *),
3817 int (*unsync)(struct net_device *,
3818 const unsigned char *))
3819{
3820 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3821}
3822
3823/**
3824 * __dev_uc_unsync - Remove synchronized addresses from device
3825 * @dev: device to sync
3826 * @unsync: function to call if address should be removed
3827 *
3828 * Remove all addresses that were added to the device by dev_uc_sync().
3829 */
3830static inline void __dev_uc_unsync(struct net_device *dev,
3831 int (*unsync)(struct net_device *,
3832 const unsigned char *))
3833{
3834 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3835}
3836
3837/* Functions used for multicast addresses handling */
3838int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3839int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3840int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3841int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3842int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3843int dev_mc_sync(struct net_device *to, struct net_device *from);
3844int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3845void dev_mc_unsync(struct net_device *to, struct net_device *from);
3846void dev_mc_flush(struct net_device *dev);
3847void dev_mc_init(struct net_device *dev);
3848
3849/**
3850 * __dev_mc_sync - Synchonize device's multicast list
3851 * @dev: device to sync
3852 * @sync: function to call if address should be added
3853 * @unsync: function to call if address should be removed
3854 *
3855 * Add newly added addresses to the interface, and release
3856 * addresses that have been deleted.
3857 */
3858static inline int __dev_mc_sync(struct net_device *dev,
3859 int (*sync)(struct net_device *,
3860 const unsigned char *),
3861 int (*unsync)(struct net_device *,
3862 const unsigned char *))
3863{
3864 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3865}
3866
3867/**
3868 * __dev_mc_unsync - Remove synchronized addresses from device
3869 * @dev: device to sync
3870 * @unsync: function to call if address should be removed
3871 *
3872 * Remove all addresses that were added to the device by dev_mc_sync().
3873 */
3874static inline void __dev_mc_unsync(struct net_device *dev,
3875 int (*unsync)(struct net_device *,
3876 const unsigned char *))
3877{
3878 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3879}
3880
3881/* Functions used for secondary unicast and multicast support */
3882void dev_set_rx_mode(struct net_device *dev);
3883void __dev_set_rx_mode(struct net_device *dev);
3884int dev_set_promiscuity(struct net_device *dev, int inc);
3885int dev_set_allmulti(struct net_device *dev, int inc);
3886void netdev_state_change(struct net_device *dev);
3887void netdev_notify_peers(struct net_device *dev);
3888void netdev_features_change(struct net_device *dev);
3889/* Load a device via the kmod */
3890void dev_load(struct net *net, const char *name);
3891struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3892 struct rtnl_link_stats64 *storage);
3893void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3894 const struct net_device_stats *netdev_stats);
3895
3896extern int netdev_max_backlog;
3897extern int netdev_tstamp_prequeue;
3898extern int weight_p;
3899extern int dev_weight_rx_bias;
3900extern int dev_weight_tx_bias;
3901extern int dev_rx_weight;
3902extern int dev_tx_weight;
3903
3904bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3905struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3906 struct list_head **iter);
3907struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3908 struct list_head **iter);
3909
3910/* iterate through upper list, must be called under RCU read lock */
3911#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3912 for (iter = &(dev)->adj_list.upper, \
3913 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3914 updev; \
3915 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3916
3917int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3918 int (*fn)(struct net_device *upper_dev,
3919 void *data),
3920 void *data);
3921
3922bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3923 struct net_device *upper_dev);
3924
3925bool netdev_has_any_upper_dev(struct net_device *dev);
3926
3927void *netdev_lower_get_next_private(struct net_device *dev,
3928 struct list_head **iter);
3929void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3930 struct list_head **iter);
3931
3932#define netdev_for_each_lower_private(dev, priv, iter) \
3933 for (iter = (dev)->adj_list.lower.next, \
3934 priv = netdev_lower_get_next_private(dev, &(iter)); \
3935 priv; \
3936 priv = netdev_lower_get_next_private(dev, &(iter)))
3937
3938#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3939 for (iter = &(dev)->adj_list.lower, \
3940 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3941 priv; \
3942 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3943
3944void *netdev_lower_get_next(struct net_device *dev,
3945 struct list_head **iter);
3946
3947#define netdev_for_each_lower_dev(dev, ldev, iter) \
3948 for (iter = (dev)->adj_list.lower.next, \
3949 ldev = netdev_lower_get_next(dev, &(iter)); \
3950 ldev; \
3951 ldev = netdev_lower_get_next(dev, &(iter)))
3952
3953struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3954 struct list_head **iter);
3955struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3956 struct list_head **iter);
3957
3958int netdev_walk_all_lower_dev(struct net_device *dev,
3959 int (*fn)(struct net_device *lower_dev,
3960 void *data),
3961 void *data);
3962int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3963 int (*fn)(struct net_device *lower_dev,
3964 void *data),
3965 void *data);
3966
3967void *netdev_adjacent_get_private(struct list_head *adj_list);
3968void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3969struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3970struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3971int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3972 struct netlink_ext_ack *extack);
3973int netdev_master_upper_dev_link(struct net_device *dev,
3974 struct net_device *upper_dev,
3975 void *upper_priv, void *upper_info,
3976 struct netlink_ext_ack *extack);
3977void netdev_upper_dev_unlink(struct net_device *dev,
3978 struct net_device *upper_dev);
3979void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3980void *netdev_lower_dev_get_private(struct net_device *dev,
3981 struct net_device *lower_dev);
3982void netdev_lower_state_changed(struct net_device *lower_dev,
3983 void *lower_state_info);
3984
3985/* RSS keys are 40 or 52 bytes long */
3986#define NETDEV_RSS_KEY_LEN 52
3987extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3988void netdev_rss_key_fill(void *buffer, size_t len);
3989
3990int dev_get_nest_level(struct net_device *dev);
3991int skb_checksum_help(struct sk_buff *skb);
3992int skb_crc32c_csum_help(struct sk_buff *skb);
3993int skb_csum_hwoffload_help(struct sk_buff *skb,
3994 const netdev_features_t features);
3995
3996struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3997 netdev_features_t features, bool tx_path);
3998struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3999 netdev_features_t features);
4000
4001struct netdev_bonding_info {
4002 ifslave slave;
4003 ifbond master;
4004};
4005
4006struct netdev_notifier_bonding_info {
4007 struct netdev_notifier_info info; /* must be first */
4008 struct netdev_bonding_info bonding_info;
4009};
4010
4011void netdev_bonding_info_change(struct net_device *dev,
4012 struct netdev_bonding_info *bonding_info);
4013
4014static inline
4015struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4016{
4017 return __skb_gso_segment(skb, features, true);
4018}
4019__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4020
4021static inline bool can_checksum_protocol(netdev_features_t features,
4022 __be16 protocol)
4023{
4024 if (protocol == htons(ETH_P_FCOE))
4025 return !!(features & NETIF_F_FCOE_CRC);
4026
4027 /* Assume this is an IP checksum (not SCTP CRC) */
4028
4029 if (features & NETIF_F_HW_CSUM) {
4030 /* Can checksum everything */
4031 return true;
4032 }
4033
4034 switch (protocol) {
4035 case htons(ETH_P_IP):
4036 return !!(features & NETIF_F_IP_CSUM);
4037 case htons(ETH_P_IPV6):
4038 return !!(features & NETIF_F_IPV6_CSUM);
4039 default:
4040 return false;
4041 }
4042}
4043
4044#ifdef CONFIG_BUG
4045void netdev_rx_csum_fault(struct net_device *dev);
4046#else
4047static inline void netdev_rx_csum_fault(struct net_device *dev)
4048{
4049}
4050#endif
4051/* rx skb timestamps */
4052void net_enable_timestamp(void);
4053void net_disable_timestamp(void);
4054
4055#ifdef CONFIG_PROC_FS
4056int __init dev_proc_init(void);
4057#else
4058#define dev_proc_init() 0
4059#endif
4060
4061static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4062 struct sk_buff *skb, struct net_device *dev,
4063 bool more)
4064{
4065 skb->xmit_more = more ? 1 : 0;
4066 return ops->ndo_start_xmit(skb, dev);
4067}
4068
4069static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4070 struct netdev_queue *txq, bool more)
4071{
4072 const struct net_device_ops *ops = dev->netdev_ops;
4073 int rc;
4074
4075 rc = __netdev_start_xmit(ops, skb, dev, more);
4076 if (rc == NETDEV_TX_OK)
4077 txq_trans_update(txq);
4078
4079 return rc;
4080}
4081
4082int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4083 const void *ns);
4084void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4085 const void *ns);
4086
4087static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4088{
4089 return netdev_class_create_file_ns(class_attr, NULL);
4090}
4091
4092static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4093{
4094 netdev_class_remove_file_ns(class_attr, NULL);
4095}
4096
4097extern const struct kobj_ns_type_operations net_ns_type_operations;
4098
4099const char *netdev_drivername(const struct net_device *dev);
4100
4101void linkwatch_run_queue(void);
4102
4103static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4104 netdev_features_t f2)
4105{
4106 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4107 if (f1 & NETIF_F_HW_CSUM)
4108 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4109 else
4110 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4111 }
4112
4113 return f1 & f2;
4114}
4115
4116static inline netdev_features_t netdev_get_wanted_features(
4117 struct net_device *dev)
4118{
4119 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4120}
4121netdev_features_t netdev_increment_features(netdev_features_t all,
4122 netdev_features_t one, netdev_features_t mask);
4123
4124/* Allow TSO being used on stacked device :
4125 * Performing the GSO segmentation before last device
4126 * is a performance improvement.
4127 */
4128static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4129 netdev_features_t mask)
4130{
4131 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4132}
4133
4134int __netdev_update_features(struct net_device *dev);
4135void netdev_update_features(struct net_device *dev);
4136void netdev_change_features(struct net_device *dev);
4137
4138void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4139 struct net_device *dev);
4140
4141netdev_features_t passthru_features_check(struct sk_buff *skb,
4142 struct net_device *dev,
4143 netdev_features_t features);
4144netdev_features_t netif_skb_features(struct sk_buff *skb);
4145
4146static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4147{
4148 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4149
4150 /* check flags correspondence */
4151 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4152 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4153 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4154 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4155 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4156 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4157 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4158 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4159 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4160 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4161 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4162 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4163 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4164 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4165 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4166 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4167 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4168
4169 return (features & feature) == feature;
4170}
4171
4172static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4173{
4174 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4175 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4176}
4177
4178static inline bool netif_needs_gso(struct sk_buff *skb,
4179 netdev_features_t features)
4180{
4181 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4182 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4183 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4184}
4185
4186static inline void netif_set_gso_max_size(struct net_device *dev,
4187 unsigned int size)
4188{
4189 dev->gso_max_size = size;
4190}
4191
4192static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4193 int pulled_hlen, u16 mac_offset,
4194 int mac_len)
4195{
4196 skb->protocol = protocol;
4197 skb->encapsulation = 1;
4198 skb_push(skb, pulled_hlen);
4199 skb_reset_transport_header(skb);
4200 skb->mac_header = mac_offset;
4201 skb->network_header = skb->mac_header + mac_len;
4202 skb->mac_len = mac_len;
4203}
4204
4205static inline bool netif_is_macsec(const struct net_device *dev)
4206{
4207 return dev->priv_flags & IFF_MACSEC;
4208}
4209
4210static inline bool netif_is_macvlan(const struct net_device *dev)
4211{
4212 return dev->priv_flags & IFF_MACVLAN;
4213}
4214
4215static inline bool netif_is_macvlan_port(const struct net_device *dev)
4216{
4217 return dev->priv_flags & IFF_MACVLAN_PORT;
4218}
4219
4220static inline bool netif_is_ipvlan(const struct net_device *dev)
4221{
4222 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4223}
4224
4225static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4226{
4227 return dev->priv_flags & IFF_IPVLAN_MASTER;
4228}
4229
4230static inline bool netif_is_bond_master(const struct net_device *dev)
4231{
4232 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4233}
4234
4235static inline bool netif_is_bond_slave(const struct net_device *dev)
4236{
4237 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4238}
4239
4240static inline bool netif_supports_nofcs(struct net_device *dev)
4241{
4242 return dev->priv_flags & IFF_SUPP_NOFCS;
4243}
4244
4245static inline bool netif_is_l3_master(const struct net_device *dev)
4246{
4247 return dev->priv_flags & IFF_L3MDEV_MASTER;
4248}
4249
4250static inline bool netif_is_l3_slave(const struct net_device *dev)
4251{
4252 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4253}
4254
4255static inline bool netif_is_bridge_master(const struct net_device *dev)
4256{
4257 return dev->priv_flags & IFF_EBRIDGE;
4258}
4259
4260static inline bool netif_is_bridge_port(const struct net_device *dev)
4261{
4262 return dev->priv_flags & IFF_BRIDGE_PORT;
4263}
4264
4265static inline bool netif_is_ovs_master(const struct net_device *dev)
4266{
4267 return dev->priv_flags & IFF_OPENVSWITCH;
4268}
4269
4270static inline bool netif_is_ovs_port(const struct net_device *dev)
4271{
4272 return dev->priv_flags & IFF_OVS_DATAPATH;
4273}
4274
4275static inline bool netif_is_team_master(const struct net_device *dev)
4276{
4277 return dev->priv_flags & IFF_TEAM;
4278}
4279
4280static inline bool netif_is_team_port(const struct net_device *dev)
4281{
4282 return dev->priv_flags & IFF_TEAM_PORT;
4283}
4284
4285static inline bool netif_is_lag_master(const struct net_device *dev)
4286{
4287 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4288}
4289
4290static inline bool netif_is_lag_port(const struct net_device *dev)
4291{
4292 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4293}
4294
4295static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4296{
4297 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4298}
4299
4300/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4301static inline void netif_keep_dst(struct net_device *dev)
4302{
4303 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4304}
4305
4306/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4307static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4308{
4309 /* TODO: reserve and use an additional IFF bit, if we get more users */
4310 return dev->priv_flags & IFF_MACSEC;
4311}
4312
4313extern struct pernet_operations __net_initdata loopback_net_ops;
4314
4315/* Logging, debugging and troubleshooting/diagnostic helpers. */
4316
4317/* netdev_printk helpers, similar to dev_printk */
4318
4319static inline const char *netdev_name(const struct net_device *dev)
4320{
4321 if (!dev->name[0] || strchr(dev->name, '%'))
4322 return "(unnamed net_device)";
4323 return dev->name;
4324}
4325
4326static inline bool netdev_unregistering(const struct net_device *dev)
4327{
4328 return dev->reg_state == NETREG_UNREGISTERING;
4329}
4330
4331static inline const char *netdev_reg_state(const struct net_device *dev)
4332{
4333 switch (dev->reg_state) {
4334 case NETREG_UNINITIALIZED: return " (uninitialized)";
4335 case NETREG_REGISTERED: return "";
4336 case NETREG_UNREGISTERING: return " (unregistering)";
4337 case NETREG_UNREGISTERED: return " (unregistered)";
4338 case NETREG_RELEASED: return " (released)";
4339 case NETREG_DUMMY: return " (dummy)";
4340 }
4341
4342 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4343 return " (unknown)";
4344}
4345
4346__printf(3, 4)
4347void netdev_printk(const char *level, const struct net_device *dev,
4348 const char *format, ...);
4349__printf(2, 3)
4350void netdev_emerg(const struct net_device *dev, const char *format, ...);
4351__printf(2, 3)
4352void netdev_alert(const struct net_device *dev, const char *format, ...);
4353__printf(2, 3)
4354void netdev_crit(const struct net_device *dev, const char *format, ...);
4355__printf(2, 3)
4356void netdev_err(const struct net_device *dev, const char *format, ...);
4357__printf(2, 3)
4358void netdev_warn(const struct net_device *dev, const char *format, ...);
4359__printf(2, 3)
4360void netdev_notice(const struct net_device *dev, const char *format, ...);
4361__printf(2, 3)
4362void netdev_info(const struct net_device *dev, const char *format, ...);
4363
4364#define netdev_level_once(level, dev, fmt, ...) \
4365do { \
4366 static bool __print_once __read_mostly; \
4367 \
4368 if (!__print_once) { \
4369 __print_once = true; \
4370 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4371 } \
4372} while (0)
4373
4374#define netdev_emerg_once(dev, fmt, ...) \
4375 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4376#define netdev_alert_once(dev, fmt, ...) \
4377 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4378#define netdev_crit_once(dev, fmt, ...) \
4379 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4380#define netdev_err_once(dev, fmt, ...) \
4381 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4382#define netdev_warn_once(dev, fmt, ...) \
4383 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4384#define netdev_notice_once(dev, fmt, ...) \
4385 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4386#define netdev_info_once(dev, fmt, ...) \
4387 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4388
4389#define MODULE_ALIAS_NETDEV(device) \
4390 MODULE_ALIAS("netdev-" device)
4391
4392#if defined(CONFIG_DYNAMIC_DEBUG)
4393#define netdev_dbg(__dev, format, args...) \
4394do { \
4395 dynamic_netdev_dbg(__dev, format, ##args); \
4396} while (0)
4397#elif defined(DEBUG)
4398#define netdev_dbg(__dev, format, args...) \
4399 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4400#else
4401#define netdev_dbg(__dev, format, args...) \
4402({ \
4403 if (0) \
4404 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4405})
4406#endif
4407
4408#if defined(VERBOSE_DEBUG)
4409#define netdev_vdbg netdev_dbg
4410#else
4411
4412#define netdev_vdbg(dev, format, args...) \
4413({ \
4414 if (0) \
4415 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4416 0; \
4417})
4418#endif
4419
4420/*
4421 * netdev_WARN() acts like dev_printk(), but with the key difference
4422 * of using a WARN/WARN_ON to get the message out, including the
4423 * file/line information and a backtrace.
4424 */
4425#define netdev_WARN(dev, format, args...) \
4426 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4427 netdev_reg_state(dev), ##args)
4428
4429#define netdev_WARN_ONCE(dev, format, args...) \
4430 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4431 netdev_reg_state(dev), ##args)
4432
4433/* netif printk helpers, similar to netdev_printk */
4434
4435#define netif_printk(priv, type, level, dev, fmt, args...) \
4436do { \
4437 if (netif_msg_##type(priv)) \
4438 netdev_printk(level, (dev), fmt, ##args); \
4439} while (0)
4440
4441#define netif_level(level, priv, type, dev, fmt, args...) \
4442do { \
4443 if (netif_msg_##type(priv)) \
4444 netdev_##level(dev, fmt, ##args); \
4445} while (0)
4446
4447#define netif_emerg(priv, type, dev, fmt, args...) \
4448 netif_level(emerg, priv, type, dev, fmt, ##args)
4449#define netif_alert(priv, type, dev, fmt, args...) \
4450 netif_level(alert, priv, type, dev, fmt, ##args)
4451#define netif_crit(priv, type, dev, fmt, args...) \
4452 netif_level(crit, priv, type, dev, fmt, ##args)
4453#define netif_err(priv, type, dev, fmt, args...) \
4454 netif_level(err, priv, type, dev, fmt, ##args)
4455#define netif_warn(priv, type, dev, fmt, args...) \
4456 netif_level(warn, priv, type, dev, fmt, ##args)
4457#define netif_notice(priv, type, dev, fmt, args...) \
4458 netif_level(notice, priv, type, dev, fmt, ##args)
4459#define netif_info(priv, type, dev, fmt, args...) \
4460 netif_level(info, priv, type, dev, fmt, ##args)
4461
4462#if defined(CONFIG_DYNAMIC_DEBUG)
4463#define netif_dbg(priv, type, netdev, format, args...) \
4464do { \
4465 if (netif_msg_##type(priv)) \
4466 dynamic_netdev_dbg(netdev, format, ##args); \
4467} while (0)
4468#elif defined(DEBUG)
4469#define netif_dbg(priv, type, dev, format, args...) \
4470 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4471#else
4472#define netif_dbg(priv, type, dev, format, args...) \
4473({ \
4474 if (0) \
4475 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4476 0; \
4477})
4478#endif
4479
4480/* if @cond then downgrade to debug, else print at @level */
4481#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4482 do { \
4483 if (cond) \
4484 netif_dbg(priv, type, netdev, fmt, ##args); \
4485 else \
4486 netif_ ## level(priv, type, netdev, fmt, ##args); \
4487 } while (0)
4488
4489#if defined(VERBOSE_DEBUG)
4490#define netif_vdbg netif_dbg
4491#else
4492#define netif_vdbg(priv, type, dev, format, args...) \
4493({ \
4494 if (0) \
4495 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4496 0; \
4497})
4498#endif
4499
4500/*
4501 * The list of packet types we will receive (as opposed to discard)
4502 * and the routines to invoke.
4503 *
4504 * Why 16. Because with 16 the only overlap we get on a hash of the
4505 * low nibble of the protocol value is RARP/SNAP/X.25.
4506 *
4507 * 0800 IP
4508 * 0001 802.3
4509 * 0002 AX.25
4510 * 0004 802.2
4511 * 8035 RARP
4512 * 0005 SNAP
4513 * 0805 X.25
4514 * 0806 ARP
4515 * 8137 IPX
4516 * 0009 Localtalk
4517 * 86DD IPv6
4518 */
4519#define PTYPE_HASH_SIZE (16)
4520#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4521
4522#endif /* _LINUX_NETDEVICE_H */