Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <net/xdp.h>
44
45#include <linux/netdev_features.h>
46#include <linux/neighbour.h>
47#include <uapi/linux/netdevice.h>
48#include <uapi/linux/if_bonding.h>
49#include <uapi/linux/pkt_cls.h>
50#include <linux/hashtable.h>
51#include <linux/rbtree.h>
52#include <net/net_trackers.h>
53
54struct netpoll_info;
55struct device;
56struct ethtool_ops;
57struct phy_device;
58struct dsa_port;
59struct ip_tunnel_parm;
60struct macsec_context;
61struct macsec_ops;
62
63struct sfp_bus;
64/* 802.11 specific */
65struct wireless_dev;
66/* 802.15.4 specific */
67struct wpan_dev;
68struct mpls_dev;
69/* UDP Tunnel offloads */
70struct udp_tunnel_info;
71struct udp_tunnel_nic_info;
72struct udp_tunnel_nic;
73struct bpf_prog;
74struct xdp_buff;
75
76void synchronize_net(void);
77void netdev_set_default_ethtool_ops(struct net_device *dev,
78 const struct ethtool_ops *ops);
79
80/* Backlog congestion levels */
81#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
82#define NET_RX_DROP 1 /* packet dropped */
83
84#define MAX_NEST_DEV 8
85
86/*
87 * Transmit return codes: transmit return codes originate from three different
88 * namespaces:
89 *
90 * - qdisc return codes
91 * - driver transmit return codes
92 * - errno values
93 *
94 * Drivers are allowed to return any one of those in their hard_start_xmit()
95 * function. Real network devices commonly used with qdiscs should only return
96 * the driver transmit return codes though - when qdiscs are used, the actual
97 * transmission happens asynchronously, so the value is not propagated to
98 * higher layers. Virtual network devices transmit synchronously; in this case
99 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
100 * others are propagated to higher layers.
101 */
102
103/* qdisc ->enqueue() return codes. */
104#define NET_XMIT_SUCCESS 0x00
105#define NET_XMIT_DROP 0x01 /* skb dropped */
106#define NET_XMIT_CN 0x02 /* congestion notification */
107#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
108
109/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
110 * indicates that the device will soon be dropping packets, or already drops
111 * some packets of the same priority; prompting us to send less aggressively. */
112#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
113#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
114
115/* Driver transmit return codes */
116#define NETDEV_TX_MASK 0xf0
117
118enum netdev_tx {
119 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
120 NETDEV_TX_OK = 0x00, /* driver took care of packet */
121 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
122};
123typedef enum netdev_tx netdev_tx_t;
124
125/*
126 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
127 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
128 */
129static inline bool dev_xmit_complete(int rc)
130{
131 /*
132 * Positive cases with an skb consumed by a driver:
133 * - successful transmission (rc == NETDEV_TX_OK)
134 * - error while transmitting (rc < 0)
135 * - error while queueing to a different device (rc & NET_XMIT_MASK)
136 */
137 if (likely(rc < NET_XMIT_MASK))
138 return true;
139
140 return false;
141}
142
143/*
144 * Compute the worst-case header length according to the protocols
145 * used.
146 */
147
148#if defined(CONFIG_HYPERV_NET)
149# define LL_MAX_HEADER 128
150#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
151# if defined(CONFIG_MAC80211_MESH)
152# define LL_MAX_HEADER 128
153# else
154# define LL_MAX_HEADER 96
155# endif
156#else
157# define LL_MAX_HEADER 32
158#endif
159
160#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
161 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
162#define MAX_HEADER LL_MAX_HEADER
163#else
164#define MAX_HEADER (LL_MAX_HEADER + 48)
165#endif
166
167/*
168 * Old network device statistics. Fields are native words
169 * (unsigned long) so they can be read and written atomically.
170 */
171
172struct net_device_stats {
173 unsigned long rx_packets;
174 unsigned long tx_packets;
175 unsigned long rx_bytes;
176 unsigned long tx_bytes;
177 unsigned long rx_errors;
178 unsigned long tx_errors;
179 unsigned long rx_dropped;
180 unsigned long tx_dropped;
181 unsigned long multicast;
182 unsigned long collisions;
183 unsigned long rx_length_errors;
184 unsigned long rx_over_errors;
185 unsigned long rx_crc_errors;
186 unsigned long rx_frame_errors;
187 unsigned long rx_fifo_errors;
188 unsigned long rx_missed_errors;
189 unsigned long tx_aborted_errors;
190 unsigned long tx_carrier_errors;
191 unsigned long tx_fifo_errors;
192 unsigned long tx_heartbeat_errors;
193 unsigned long tx_window_errors;
194 unsigned long rx_compressed;
195 unsigned long tx_compressed;
196};
197
198/* per-cpu stats, allocated on demand.
199 * Try to fit them in a single cache line, for dev_get_stats() sake.
200 */
201struct net_device_core_stats {
202 unsigned long rx_dropped;
203 unsigned long tx_dropped;
204 unsigned long rx_nohandler;
205} __aligned(4 * sizeof(unsigned long));
206
207#include <linux/cache.h>
208#include <linux/skbuff.h>
209
210#ifdef CONFIG_RPS
211#include <linux/static_key.h>
212extern struct static_key_false rps_needed;
213extern struct static_key_false rfs_needed;
214#endif
215
216struct neighbour;
217struct neigh_parms;
218struct sk_buff;
219
220struct netdev_hw_addr {
221 struct list_head list;
222 struct rb_node node;
223 unsigned char addr[MAX_ADDR_LEN];
224 unsigned char type;
225#define NETDEV_HW_ADDR_T_LAN 1
226#define NETDEV_HW_ADDR_T_SAN 2
227#define NETDEV_HW_ADDR_T_UNICAST 3
228#define NETDEV_HW_ADDR_T_MULTICAST 4
229 bool global_use;
230 int sync_cnt;
231 int refcount;
232 int synced;
233 struct rcu_head rcu_head;
234};
235
236struct netdev_hw_addr_list {
237 struct list_head list;
238 int count;
239
240 /* Auxiliary tree for faster lookup on addition and deletion */
241 struct rb_root tree;
242};
243
244#define netdev_hw_addr_list_count(l) ((l)->count)
245#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
246#define netdev_hw_addr_list_for_each(ha, l) \
247 list_for_each_entry(ha, &(l)->list, list)
248
249#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
250#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
251#define netdev_for_each_uc_addr(ha, dev) \
252 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
253
254#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
255#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
256#define netdev_for_each_mc_addr(ha, dev) \
257 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
258
259struct hh_cache {
260 unsigned int hh_len;
261 seqlock_t hh_lock;
262
263 /* cached hardware header; allow for machine alignment needs. */
264#define HH_DATA_MOD 16
265#define HH_DATA_OFF(__len) \
266 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
267#define HH_DATA_ALIGN(__len) \
268 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
269 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
270};
271
272/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
273 * Alternative is:
274 * dev->hard_header_len ? (dev->hard_header_len +
275 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
276 *
277 * We could use other alignment values, but we must maintain the
278 * relationship HH alignment <= LL alignment.
279 */
280#define LL_RESERVED_SPACE(dev) \
281 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
282#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
283 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
284
285struct header_ops {
286 int (*create) (struct sk_buff *skb, struct net_device *dev,
287 unsigned short type, const void *daddr,
288 const void *saddr, unsigned int len);
289 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
290 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
291 void (*cache_update)(struct hh_cache *hh,
292 const struct net_device *dev,
293 const unsigned char *haddr);
294 bool (*validate)(const char *ll_header, unsigned int len);
295 __be16 (*parse_protocol)(const struct sk_buff *skb);
296};
297
298/* These flag bits are private to the generic network queueing
299 * layer; they may not be explicitly referenced by any other
300 * code.
301 */
302
303enum netdev_state_t {
304 __LINK_STATE_START,
305 __LINK_STATE_PRESENT,
306 __LINK_STATE_NOCARRIER,
307 __LINK_STATE_LINKWATCH_PENDING,
308 __LINK_STATE_DORMANT,
309 __LINK_STATE_TESTING,
310};
311
312struct gro_list {
313 struct list_head list;
314 int count;
315};
316
317/*
318 * size of gro hash buckets, must less than bit number of
319 * napi_struct::gro_bitmask
320 */
321#define GRO_HASH_BUCKETS 8
322
323/*
324 * Structure for NAPI scheduling similar to tasklet but with weighting
325 */
326struct napi_struct {
327 /* The poll_list must only be managed by the entity which
328 * changes the state of the NAPI_STATE_SCHED bit. This means
329 * whoever atomically sets that bit can add this napi_struct
330 * to the per-CPU poll_list, and whoever clears that bit
331 * can remove from the list right before clearing the bit.
332 */
333 struct list_head poll_list;
334
335 unsigned long state;
336 int weight;
337 int defer_hard_irqs_count;
338 unsigned long gro_bitmask;
339 int (*poll)(struct napi_struct *, int);
340#ifdef CONFIG_NETPOLL
341 int poll_owner;
342#endif
343 struct net_device *dev;
344 struct gro_list gro_hash[GRO_HASH_BUCKETS];
345 struct sk_buff *skb;
346 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
347 int rx_count; /* length of rx_list */
348 struct hrtimer timer;
349 struct list_head dev_list;
350 struct hlist_node napi_hash_node;
351 unsigned int napi_id;
352 struct task_struct *thread;
353};
354
355enum {
356 NAPI_STATE_SCHED, /* Poll is scheduled */
357 NAPI_STATE_MISSED, /* reschedule a napi */
358 NAPI_STATE_DISABLE, /* Disable pending */
359 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
360 NAPI_STATE_LISTED, /* NAPI added to system lists */
361 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
362 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
363 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
364 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
365 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
366};
367
368enum {
369 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
370 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
371 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
372 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
373 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
374 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
375 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
376 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
377 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
378 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
379};
380
381enum gro_result {
382 GRO_MERGED,
383 GRO_MERGED_FREE,
384 GRO_HELD,
385 GRO_NORMAL,
386 GRO_CONSUMED,
387};
388typedef enum gro_result gro_result_t;
389
390/*
391 * enum rx_handler_result - Possible return values for rx_handlers.
392 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
393 * further.
394 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
395 * case skb->dev was changed by rx_handler.
396 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
397 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
398 *
399 * rx_handlers are functions called from inside __netif_receive_skb(), to do
400 * special processing of the skb, prior to delivery to protocol handlers.
401 *
402 * Currently, a net_device can only have a single rx_handler registered. Trying
403 * to register a second rx_handler will return -EBUSY.
404 *
405 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
406 * To unregister a rx_handler on a net_device, use
407 * netdev_rx_handler_unregister().
408 *
409 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
410 * do with the skb.
411 *
412 * If the rx_handler consumed the skb in some way, it should return
413 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
414 * the skb to be delivered in some other way.
415 *
416 * If the rx_handler changed skb->dev, to divert the skb to another
417 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
418 * new device will be called if it exists.
419 *
420 * If the rx_handler decides the skb should be ignored, it should return
421 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
422 * are registered on exact device (ptype->dev == skb->dev).
423 *
424 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
425 * delivered, it should return RX_HANDLER_PASS.
426 *
427 * A device without a registered rx_handler will behave as if rx_handler
428 * returned RX_HANDLER_PASS.
429 */
430
431enum rx_handler_result {
432 RX_HANDLER_CONSUMED,
433 RX_HANDLER_ANOTHER,
434 RX_HANDLER_EXACT,
435 RX_HANDLER_PASS,
436};
437typedef enum rx_handler_result rx_handler_result_t;
438typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
439
440void __napi_schedule(struct napi_struct *n);
441void __napi_schedule_irqoff(struct napi_struct *n);
442
443static inline bool napi_disable_pending(struct napi_struct *n)
444{
445 return test_bit(NAPI_STATE_DISABLE, &n->state);
446}
447
448static inline bool napi_prefer_busy_poll(struct napi_struct *n)
449{
450 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
451}
452
453bool napi_schedule_prep(struct napi_struct *n);
454
455/**
456 * napi_schedule - schedule NAPI poll
457 * @n: NAPI context
458 *
459 * Schedule NAPI poll routine to be called if it is not already
460 * running.
461 */
462static inline void napi_schedule(struct napi_struct *n)
463{
464 if (napi_schedule_prep(n))
465 __napi_schedule(n);
466}
467
468/**
469 * napi_schedule_irqoff - schedule NAPI poll
470 * @n: NAPI context
471 *
472 * Variant of napi_schedule(), assuming hard irqs are masked.
473 */
474static inline void napi_schedule_irqoff(struct napi_struct *n)
475{
476 if (napi_schedule_prep(n))
477 __napi_schedule_irqoff(n);
478}
479
480/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
481static inline bool napi_reschedule(struct napi_struct *napi)
482{
483 if (napi_schedule_prep(napi)) {
484 __napi_schedule(napi);
485 return true;
486 }
487 return false;
488}
489
490bool napi_complete_done(struct napi_struct *n, int work_done);
491/**
492 * napi_complete - NAPI processing complete
493 * @n: NAPI context
494 *
495 * Mark NAPI processing as complete.
496 * Consider using napi_complete_done() instead.
497 * Return false if device should avoid rearming interrupts.
498 */
499static inline bool napi_complete(struct napi_struct *n)
500{
501 return napi_complete_done(n, 0);
502}
503
504int dev_set_threaded(struct net_device *dev, bool threaded);
505
506/**
507 * napi_disable - prevent NAPI from scheduling
508 * @n: NAPI context
509 *
510 * Stop NAPI from being scheduled on this context.
511 * Waits till any outstanding processing completes.
512 */
513void napi_disable(struct napi_struct *n);
514
515void napi_enable(struct napi_struct *n);
516
517/**
518 * napi_synchronize - wait until NAPI is not running
519 * @n: NAPI context
520 *
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
524 */
525static inline void napi_synchronize(const struct napi_struct *n)
526{
527 if (IS_ENABLED(CONFIG_SMP))
528 while (test_bit(NAPI_STATE_SCHED, &n->state))
529 msleep(1);
530 else
531 barrier();
532}
533
534/**
535 * napi_if_scheduled_mark_missed - if napi is running, set the
536 * NAPIF_STATE_MISSED
537 * @n: NAPI context
538 *
539 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
540 * NAPI is scheduled.
541 **/
542static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
543{
544 unsigned long val, new;
545
546 do {
547 val = READ_ONCE(n->state);
548 if (val & NAPIF_STATE_DISABLE)
549 return true;
550
551 if (!(val & NAPIF_STATE_SCHED))
552 return false;
553
554 new = val | NAPIF_STATE_MISSED;
555 } while (cmpxchg(&n->state, val, new) != val);
556
557 return true;
558}
559
560enum netdev_queue_state_t {
561 __QUEUE_STATE_DRV_XOFF,
562 __QUEUE_STATE_STACK_XOFF,
563 __QUEUE_STATE_FROZEN,
564};
565
566#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
567#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
568#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
569
570#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
571#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
572 QUEUE_STATE_FROZEN)
573#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
574 QUEUE_STATE_FROZEN)
575
576/*
577 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
578 * netif_tx_* functions below are used to manipulate this flag. The
579 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
580 * queue independently. The netif_xmit_*stopped functions below are called
581 * to check if the queue has been stopped by the driver or stack (either
582 * of the XOFF bits are set in the state). Drivers should not need to call
583 * netif_xmit*stopped functions, they should only be using netif_tx_*.
584 */
585
586struct netdev_queue {
587/*
588 * read-mostly part
589 */
590 struct net_device *dev;
591 netdevice_tracker dev_tracker;
592
593 struct Qdisc __rcu *qdisc;
594 struct Qdisc *qdisc_sleeping;
595#ifdef CONFIG_SYSFS
596 struct kobject kobj;
597#endif
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 int numa_node;
600#endif
601 unsigned long tx_maxrate;
602 /*
603 * Number of TX timeouts for this queue
604 * (/sys/class/net/DEV/Q/trans_timeout)
605 */
606 atomic_long_t trans_timeout;
607
608 /* Subordinate device that the queue has been assigned to */
609 struct net_device *sb_dev;
610#ifdef CONFIG_XDP_SOCKETS
611 struct xsk_buff_pool *pool;
612#endif
613/*
614 * write-mostly part
615 */
616 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
617 int xmit_lock_owner;
618 /*
619 * Time (in jiffies) of last Tx
620 */
621 unsigned long trans_start;
622
623 unsigned long state;
624
625#ifdef CONFIG_BQL
626 struct dql dql;
627#endif
628} ____cacheline_aligned_in_smp;
629
630extern int sysctl_fb_tunnels_only_for_init_net;
631extern int sysctl_devconf_inherit_init_net;
632
633/*
634 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
635 * == 1 : For initns only
636 * == 2 : For none.
637 */
638static inline bool net_has_fallback_tunnels(const struct net *net)
639{
640 return !IS_ENABLED(CONFIG_SYSCTL) ||
641 !sysctl_fb_tunnels_only_for_init_net ||
642 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
643}
644
645static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
646{
647#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
648 return q->numa_node;
649#else
650 return NUMA_NO_NODE;
651#endif
652}
653
654static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
655{
656#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 q->numa_node = node;
658#endif
659}
660
661#ifdef CONFIG_RPS
662/*
663 * This structure holds an RPS map which can be of variable length. The
664 * map is an array of CPUs.
665 */
666struct rps_map {
667 unsigned int len;
668 struct rcu_head rcu;
669 u16 cpus[];
670};
671#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
672
673/*
674 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
675 * tail pointer for that CPU's input queue at the time of last enqueue, and
676 * a hardware filter index.
677 */
678struct rps_dev_flow {
679 u16 cpu;
680 u16 filter;
681 unsigned int last_qtail;
682};
683#define RPS_NO_FILTER 0xffff
684
685/*
686 * The rps_dev_flow_table structure contains a table of flow mappings.
687 */
688struct rps_dev_flow_table {
689 unsigned int mask;
690 struct rcu_head rcu;
691 struct rps_dev_flow flows[];
692};
693#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
694 ((_num) * sizeof(struct rps_dev_flow)))
695
696/*
697 * The rps_sock_flow_table contains mappings of flows to the last CPU
698 * on which they were processed by the application (set in recvmsg).
699 * Each entry is a 32bit value. Upper part is the high-order bits
700 * of flow hash, lower part is CPU number.
701 * rps_cpu_mask is used to partition the space, depending on number of
702 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
703 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
704 * meaning we use 32-6=26 bits for the hash.
705 */
706struct rps_sock_flow_table {
707 u32 mask;
708
709 u32 ents[] ____cacheline_aligned_in_smp;
710};
711#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
712
713#define RPS_NO_CPU 0xffff
714
715extern u32 rps_cpu_mask;
716extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
717
718static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
719 u32 hash)
720{
721 if (table && hash) {
722 unsigned int index = hash & table->mask;
723 u32 val = hash & ~rps_cpu_mask;
724
725 /* We only give a hint, preemption can change CPU under us */
726 val |= raw_smp_processor_id();
727
728 if (table->ents[index] != val)
729 table->ents[index] = val;
730 }
731}
732
733#ifdef CONFIG_RFS_ACCEL
734bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
735 u16 filter_id);
736#endif
737#endif /* CONFIG_RPS */
738
739/* This structure contains an instance of an RX queue. */
740struct netdev_rx_queue {
741 struct xdp_rxq_info xdp_rxq;
742#ifdef CONFIG_RPS
743 struct rps_map __rcu *rps_map;
744 struct rps_dev_flow_table __rcu *rps_flow_table;
745#endif
746 struct kobject kobj;
747 struct net_device *dev;
748 netdevice_tracker dev_tracker;
749
750#ifdef CONFIG_XDP_SOCKETS
751 struct xsk_buff_pool *pool;
752#endif
753} ____cacheline_aligned_in_smp;
754
755/*
756 * RX queue sysfs structures and functions.
757 */
758struct rx_queue_attribute {
759 struct attribute attr;
760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
761 ssize_t (*store)(struct netdev_rx_queue *queue,
762 const char *buf, size_t len);
763};
764
765/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
766enum xps_map_type {
767 XPS_CPUS = 0,
768 XPS_RXQS,
769 XPS_MAPS_MAX,
770};
771
772#ifdef CONFIG_XPS
773/*
774 * This structure holds an XPS map which can be of variable length. The
775 * map is an array of queues.
776 */
777struct xps_map {
778 unsigned int len;
779 unsigned int alloc_len;
780 struct rcu_head rcu;
781 u16 queues[];
782};
783#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
784#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
785 - sizeof(struct xps_map)) / sizeof(u16))
786
787/*
788 * This structure holds all XPS maps for device. Maps are indexed by CPU.
789 *
790 * We keep track of the number of cpus/rxqs used when the struct is allocated,
791 * in nr_ids. This will help not accessing out-of-bound memory.
792 *
793 * We keep track of the number of traffic classes used when the struct is
794 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
795 * not crossing its upper bound, as the original dev->num_tc can be updated in
796 * the meantime.
797 */
798struct xps_dev_maps {
799 struct rcu_head rcu;
800 unsigned int nr_ids;
801 s16 num_tc;
802 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
803};
804
805#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
806 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
807
808#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
809 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
810
811#endif /* CONFIG_XPS */
812
813#define TC_MAX_QUEUE 16
814#define TC_BITMASK 15
815/* HW offloaded queuing disciplines txq count and offset maps */
816struct netdev_tc_txq {
817 u16 count;
818 u16 offset;
819};
820
821#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
822/*
823 * This structure is to hold information about the device
824 * configured to run FCoE protocol stack.
825 */
826struct netdev_fcoe_hbainfo {
827 char manufacturer[64];
828 char serial_number[64];
829 char hardware_version[64];
830 char driver_version[64];
831 char optionrom_version[64];
832 char firmware_version[64];
833 char model[256];
834 char model_description[256];
835};
836#endif
837
838#define MAX_PHYS_ITEM_ID_LEN 32
839
840/* This structure holds a unique identifier to identify some
841 * physical item (port for example) used by a netdevice.
842 */
843struct netdev_phys_item_id {
844 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
845 unsigned char id_len;
846};
847
848static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
849 struct netdev_phys_item_id *b)
850{
851 return a->id_len == b->id_len &&
852 memcmp(a->id, b->id, a->id_len) == 0;
853}
854
855typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
856 struct sk_buff *skb,
857 struct net_device *sb_dev);
858
859enum net_device_path_type {
860 DEV_PATH_ETHERNET = 0,
861 DEV_PATH_VLAN,
862 DEV_PATH_BRIDGE,
863 DEV_PATH_PPPOE,
864 DEV_PATH_DSA,
865};
866
867struct net_device_path {
868 enum net_device_path_type type;
869 const struct net_device *dev;
870 union {
871 struct {
872 u16 id;
873 __be16 proto;
874 u8 h_dest[ETH_ALEN];
875 } encap;
876 struct {
877 enum {
878 DEV_PATH_BR_VLAN_KEEP,
879 DEV_PATH_BR_VLAN_TAG,
880 DEV_PATH_BR_VLAN_UNTAG,
881 DEV_PATH_BR_VLAN_UNTAG_HW,
882 } vlan_mode;
883 u16 vlan_id;
884 __be16 vlan_proto;
885 } bridge;
886 struct {
887 int port;
888 u16 proto;
889 } dsa;
890 };
891};
892
893#define NET_DEVICE_PATH_STACK_MAX 5
894#define NET_DEVICE_PATH_VLAN_MAX 2
895
896struct net_device_path_stack {
897 int num_paths;
898 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
899};
900
901struct net_device_path_ctx {
902 const struct net_device *dev;
903 u8 daddr[ETH_ALEN];
904
905 int num_vlans;
906 struct {
907 u16 id;
908 __be16 proto;
909 } vlan[NET_DEVICE_PATH_VLAN_MAX];
910};
911
912enum tc_setup_type {
913 TC_SETUP_QDISC_MQPRIO,
914 TC_SETUP_CLSU32,
915 TC_SETUP_CLSFLOWER,
916 TC_SETUP_CLSMATCHALL,
917 TC_SETUP_CLSBPF,
918 TC_SETUP_BLOCK,
919 TC_SETUP_QDISC_CBS,
920 TC_SETUP_QDISC_RED,
921 TC_SETUP_QDISC_PRIO,
922 TC_SETUP_QDISC_MQ,
923 TC_SETUP_QDISC_ETF,
924 TC_SETUP_ROOT_QDISC,
925 TC_SETUP_QDISC_GRED,
926 TC_SETUP_QDISC_TAPRIO,
927 TC_SETUP_FT,
928 TC_SETUP_QDISC_ETS,
929 TC_SETUP_QDISC_TBF,
930 TC_SETUP_QDISC_FIFO,
931 TC_SETUP_QDISC_HTB,
932 TC_SETUP_ACT,
933};
934
935/* These structures hold the attributes of bpf state that are being passed
936 * to the netdevice through the bpf op.
937 */
938enum bpf_netdev_command {
939 /* Set or clear a bpf program used in the earliest stages of packet
940 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
941 * is responsible for calling bpf_prog_put on any old progs that are
942 * stored. In case of error, the callee need not release the new prog
943 * reference, but on success it takes ownership and must bpf_prog_put
944 * when it is no longer used.
945 */
946 XDP_SETUP_PROG,
947 XDP_SETUP_PROG_HW,
948 /* BPF program for offload callbacks, invoked at program load time. */
949 BPF_OFFLOAD_MAP_ALLOC,
950 BPF_OFFLOAD_MAP_FREE,
951 XDP_SETUP_XSK_POOL,
952};
953
954struct bpf_prog_offload_ops;
955struct netlink_ext_ack;
956struct xdp_umem;
957struct xdp_dev_bulk_queue;
958struct bpf_xdp_link;
959
960enum bpf_xdp_mode {
961 XDP_MODE_SKB = 0,
962 XDP_MODE_DRV = 1,
963 XDP_MODE_HW = 2,
964 __MAX_XDP_MODE
965};
966
967struct bpf_xdp_entity {
968 struct bpf_prog *prog;
969 struct bpf_xdp_link *link;
970};
971
972struct netdev_bpf {
973 enum bpf_netdev_command command;
974 union {
975 /* XDP_SETUP_PROG */
976 struct {
977 u32 flags;
978 struct bpf_prog *prog;
979 struct netlink_ext_ack *extack;
980 };
981 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
982 struct {
983 struct bpf_offloaded_map *offmap;
984 };
985 /* XDP_SETUP_XSK_POOL */
986 struct {
987 struct xsk_buff_pool *pool;
988 u16 queue_id;
989 } xsk;
990 };
991};
992
993/* Flags for ndo_xsk_wakeup. */
994#define XDP_WAKEUP_RX (1 << 0)
995#define XDP_WAKEUP_TX (1 << 1)
996
997#ifdef CONFIG_XFRM_OFFLOAD
998struct xfrmdev_ops {
999 int (*xdo_dev_state_add) (struct xfrm_state *x);
1000 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1001 void (*xdo_dev_state_free) (struct xfrm_state *x);
1002 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1003 struct xfrm_state *x);
1004 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1005};
1006#endif
1007
1008struct dev_ifalias {
1009 struct rcu_head rcuhead;
1010 char ifalias[];
1011};
1012
1013struct devlink;
1014struct tlsdev_ops;
1015
1016struct netdev_name_node {
1017 struct hlist_node hlist;
1018 struct list_head list;
1019 struct net_device *dev;
1020 const char *name;
1021};
1022
1023int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1024int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1025
1026struct netdev_net_notifier {
1027 struct list_head list;
1028 struct notifier_block *nb;
1029};
1030
1031/*
1032 * This structure defines the management hooks for network devices.
1033 * The following hooks can be defined; unless noted otherwise, they are
1034 * optional and can be filled with a null pointer.
1035 *
1036 * int (*ndo_init)(struct net_device *dev);
1037 * This function is called once when a network device is registered.
1038 * The network device can use this for any late stage initialization
1039 * or semantic validation. It can fail with an error code which will
1040 * be propagated back to register_netdev.
1041 *
1042 * void (*ndo_uninit)(struct net_device *dev);
1043 * This function is called when device is unregistered or when registration
1044 * fails. It is not called if init fails.
1045 *
1046 * int (*ndo_open)(struct net_device *dev);
1047 * This function is called when a network device transitions to the up
1048 * state.
1049 *
1050 * int (*ndo_stop)(struct net_device *dev);
1051 * This function is called when a network device transitions to the down
1052 * state.
1053 *
1054 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1055 * struct net_device *dev);
1056 * Called when a packet needs to be transmitted.
1057 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1058 * the queue before that can happen; it's for obsolete devices and weird
1059 * corner cases, but the stack really does a non-trivial amount
1060 * of useless work if you return NETDEV_TX_BUSY.
1061 * Required; cannot be NULL.
1062 *
1063 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1064 * struct net_device *dev
1065 * netdev_features_t features);
1066 * Called by core transmit path to determine if device is capable of
1067 * performing offload operations on a given packet. This is to give
1068 * the device an opportunity to implement any restrictions that cannot
1069 * be otherwise expressed by feature flags. The check is called with
1070 * the set of features that the stack has calculated and it returns
1071 * those the driver believes to be appropriate.
1072 *
1073 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1074 * struct net_device *sb_dev);
1075 * Called to decide which queue to use when device supports multiple
1076 * transmit queues.
1077 *
1078 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1079 * This function is called to allow device receiver to make
1080 * changes to configuration when multicast or promiscuous is enabled.
1081 *
1082 * void (*ndo_set_rx_mode)(struct net_device *dev);
1083 * This function is called device changes address list filtering.
1084 * If driver handles unicast address filtering, it should set
1085 * IFF_UNICAST_FLT in its priv_flags.
1086 *
1087 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1088 * This function is called when the Media Access Control address
1089 * needs to be changed. If this interface is not defined, the
1090 * MAC address can not be changed.
1091 *
1092 * int (*ndo_validate_addr)(struct net_device *dev);
1093 * Test if Media Access Control address is valid for the device.
1094 *
1095 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1096 * Old-style ioctl entry point. This is used internally by the
1097 * appletalk and ieee802154 subsystems but is no longer called by
1098 * the device ioctl handler.
1099 *
1100 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1101 * Used by the bonding driver for its device specific ioctls:
1102 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1103 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1104 *
1105 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1106 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1107 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1108 *
1109 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1110 * Used to set network devices bus interface parameters. This interface
1111 * is retained for legacy reasons; new devices should use the bus
1112 * interface (PCI) for low level management.
1113 *
1114 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1115 * Called when a user wants to change the Maximum Transfer Unit
1116 * of a device.
1117 *
1118 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1119 * Callback used when the transmitter has not made any progress
1120 * for dev->watchdog ticks.
1121 *
1122 * void (*ndo_get_stats64)(struct net_device *dev,
1123 * struct rtnl_link_stats64 *storage);
1124 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1125 * Called when a user wants to get the network device usage
1126 * statistics. Drivers must do one of the following:
1127 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1128 * rtnl_link_stats64 structure passed by the caller.
1129 * 2. Define @ndo_get_stats to update a net_device_stats structure
1130 * (which should normally be dev->stats) and return a pointer to
1131 * it. The structure may be changed asynchronously only if each
1132 * field is written atomically.
1133 * 3. Update dev->stats asynchronously and atomically, and define
1134 * neither operation.
1135 *
1136 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1137 * Return true if this device supports offload stats of this attr_id.
1138 *
1139 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1140 * void *attr_data)
1141 * Get statistics for offload operations by attr_id. Write it into the
1142 * attr_data pointer.
1143 *
1144 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1145 * If device supports VLAN filtering this function is called when a
1146 * VLAN id is registered.
1147 *
1148 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1149 * If device supports VLAN filtering this function is called when a
1150 * VLAN id is unregistered.
1151 *
1152 * void (*ndo_poll_controller)(struct net_device *dev);
1153 *
1154 * SR-IOV management functions.
1155 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1156 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1157 * u8 qos, __be16 proto);
1158 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1159 * int max_tx_rate);
1160 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1161 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1162 * int (*ndo_get_vf_config)(struct net_device *dev,
1163 * int vf, struct ifla_vf_info *ivf);
1164 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1165 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1166 * struct nlattr *port[]);
1167 *
1168 * Enable or disable the VF ability to query its RSS Redirection Table and
1169 * Hash Key. This is needed since on some devices VF share this information
1170 * with PF and querying it may introduce a theoretical security risk.
1171 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1172 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1173 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1174 * void *type_data);
1175 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1176 * This is always called from the stack with the rtnl lock held and netif
1177 * tx queues stopped. This allows the netdevice to perform queue
1178 * management safely.
1179 *
1180 * Fiber Channel over Ethernet (FCoE) offload functions.
1181 * int (*ndo_fcoe_enable)(struct net_device *dev);
1182 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1183 * so the underlying device can perform whatever needed configuration or
1184 * initialization to support acceleration of FCoE traffic.
1185 *
1186 * int (*ndo_fcoe_disable)(struct net_device *dev);
1187 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1188 * so the underlying device can perform whatever needed clean-ups to
1189 * stop supporting acceleration of FCoE traffic.
1190 *
1191 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1192 * struct scatterlist *sgl, unsigned int sgc);
1193 * Called when the FCoE Initiator wants to initialize an I/O that
1194 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1195 * perform necessary setup and returns 1 to indicate the device is set up
1196 * successfully to perform DDP on this I/O, otherwise this returns 0.
1197 *
1198 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1199 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1200 * indicated by the FC exchange id 'xid', so the underlying device can
1201 * clean up and reuse resources for later DDP requests.
1202 *
1203 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1204 * struct scatterlist *sgl, unsigned int sgc);
1205 * Called when the FCoE Target wants to initialize an I/O that
1206 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1207 * perform necessary setup and returns 1 to indicate the device is set up
1208 * successfully to perform DDP on this I/O, otherwise this returns 0.
1209 *
1210 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1211 * struct netdev_fcoe_hbainfo *hbainfo);
1212 * Called when the FCoE Protocol stack wants information on the underlying
1213 * device. This information is utilized by the FCoE protocol stack to
1214 * register attributes with Fiber Channel management service as per the
1215 * FC-GS Fabric Device Management Information(FDMI) specification.
1216 *
1217 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1218 * Called when the underlying device wants to override default World Wide
1219 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1220 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1221 * protocol stack to use.
1222 *
1223 * RFS acceleration.
1224 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1225 * u16 rxq_index, u32 flow_id);
1226 * Set hardware filter for RFS. rxq_index is the target queue index;
1227 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1228 * Return the filter ID on success, or a negative error code.
1229 *
1230 * Slave management functions (for bridge, bonding, etc).
1231 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1232 * Called to make another netdev an underling.
1233 *
1234 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1235 * Called to release previously enslaved netdev.
1236 *
1237 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1238 * struct sk_buff *skb,
1239 * bool all_slaves);
1240 * Get the xmit slave of master device. If all_slaves is true, function
1241 * assume all the slaves can transmit.
1242 *
1243 * Feature/offload setting functions.
1244 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1245 * netdev_features_t features);
1246 * Adjusts the requested feature flags according to device-specific
1247 * constraints, and returns the resulting flags. Must not modify
1248 * the device state.
1249 *
1250 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1251 * Called to update device configuration to new features. Passed
1252 * feature set might be less than what was returned by ndo_fix_features()).
1253 * Must return >0 or -errno if it changed dev->features itself.
1254 *
1255 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1256 * struct net_device *dev,
1257 * const unsigned char *addr, u16 vid, u16 flags,
1258 * struct netlink_ext_ack *extack);
1259 * Adds an FDB entry to dev for addr.
1260 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1261 * struct net_device *dev,
1262 * const unsigned char *addr, u16 vid)
1263 * Deletes the FDB entry from dev coresponding to addr.
1264 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1265 * struct net_device *dev, struct net_device *filter_dev,
1266 * int *idx)
1267 * Used to add FDB entries to dump requests. Implementers should add
1268 * entries to skb and update idx with the number of entries.
1269 *
1270 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1271 * u16 flags, struct netlink_ext_ack *extack)
1272 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1273 * struct net_device *dev, u32 filter_mask,
1274 * int nlflags)
1275 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1276 * u16 flags);
1277 *
1278 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1279 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1280 * which do not represent real hardware may define this to allow their
1281 * userspace components to manage their virtual carrier state. Devices
1282 * that determine carrier state from physical hardware properties (eg
1283 * network cables) or protocol-dependent mechanisms (eg
1284 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1285 *
1286 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1287 * struct netdev_phys_item_id *ppid);
1288 * Called to get ID of physical port of this device. If driver does
1289 * not implement this, it is assumed that the hw is not able to have
1290 * multiple net devices on single physical port.
1291 *
1292 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1293 * struct netdev_phys_item_id *ppid)
1294 * Called to get the parent ID of the physical port of this device.
1295 *
1296 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1297 * struct net_device *dev)
1298 * Called by upper layer devices to accelerate switching or other
1299 * station functionality into hardware. 'pdev is the lowerdev
1300 * to use for the offload and 'dev' is the net device that will
1301 * back the offload. Returns a pointer to the private structure
1302 * the upper layer will maintain.
1303 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1304 * Called by upper layer device to delete the station created
1305 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1306 * the station and priv is the structure returned by the add
1307 * operation.
1308 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1309 * int queue_index, u32 maxrate);
1310 * Called when a user wants to set a max-rate limitation of specific
1311 * TX queue.
1312 * int (*ndo_get_iflink)(const struct net_device *dev);
1313 * Called to get the iflink value of this device.
1314 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1315 * This function is used to get egress tunnel information for given skb.
1316 * This is useful for retrieving outer tunnel header parameters while
1317 * sampling packet.
1318 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1319 * This function is used to specify the headroom that the skb must
1320 * consider when allocation skb during packet reception. Setting
1321 * appropriate rx headroom value allows avoiding skb head copy on
1322 * forward. Setting a negative value resets the rx headroom to the
1323 * default value.
1324 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1325 * This function is used to set or query state related to XDP on the
1326 * netdevice and manage BPF offload. See definition of
1327 * enum bpf_netdev_command for details.
1328 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1329 * u32 flags);
1330 * This function is used to submit @n XDP packets for transmit on a
1331 * netdevice. Returns number of frames successfully transmitted, frames
1332 * that got dropped are freed/returned via xdp_return_frame().
1333 * Returns negative number, means general error invoking ndo, meaning
1334 * no frames were xmit'ed and core-caller will free all frames.
1335 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1336 * struct xdp_buff *xdp);
1337 * Get the xmit slave of master device based on the xdp_buff.
1338 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1339 * This function is used to wake up the softirq, ksoftirqd or kthread
1340 * responsible for sending and/or receiving packets on a specific
1341 * queue id bound to an AF_XDP socket. The flags field specifies if
1342 * only RX, only Tx, or both should be woken up using the flags
1343 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1344 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1345 * Get devlink port instance associated with a given netdev.
1346 * Called with a reference on the netdevice and devlink locks only,
1347 * rtnl_lock is not held.
1348 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1349 * int cmd);
1350 * Add, change, delete or get information on an IPv4 tunnel.
1351 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1352 * If a device is paired with a peer device, return the peer instance.
1353 * The caller must be under RCU read context.
1354 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1355 * Get the forwarding path to reach the real device from the HW destination address
1356 */
1357struct net_device_ops {
1358 int (*ndo_init)(struct net_device *dev);
1359 void (*ndo_uninit)(struct net_device *dev);
1360 int (*ndo_open)(struct net_device *dev);
1361 int (*ndo_stop)(struct net_device *dev);
1362 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1363 struct net_device *dev);
1364 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1365 struct net_device *dev,
1366 netdev_features_t features);
1367 u16 (*ndo_select_queue)(struct net_device *dev,
1368 struct sk_buff *skb,
1369 struct net_device *sb_dev);
1370 void (*ndo_change_rx_flags)(struct net_device *dev,
1371 int flags);
1372 void (*ndo_set_rx_mode)(struct net_device *dev);
1373 int (*ndo_set_mac_address)(struct net_device *dev,
1374 void *addr);
1375 int (*ndo_validate_addr)(struct net_device *dev);
1376 int (*ndo_do_ioctl)(struct net_device *dev,
1377 struct ifreq *ifr, int cmd);
1378 int (*ndo_eth_ioctl)(struct net_device *dev,
1379 struct ifreq *ifr, int cmd);
1380 int (*ndo_siocbond)(struct net_device *dev,
1381 struct ifreq *ifr, int cmd);
1382 int (*ndo_siocwandev)(struct net_device *dev,
1383 struct if_settings *ifs);
1384 int (*ndo_siocdevprivate)(struct net_device *dev,
1385 struct ifreq *ifr,
1386 void __user *data, int cmd);
1387 int (*ndo_set_config)(struct net_device *dev,
1388 struct ifmap *map);
1389 int (*ndo_change_mtu)(struct net_device *dev,
1390 int new_mtu);
1391 int (*ndo_neigh_setup)(struct net_device *dev,
1392 struct neigh_parms *);
1393 void (*ndo_tx_timeout) (struct net_device *dev,
1394 unsigned int txqueue);
1395
1396 void (*ndo_get_stats64)(struct net_device *dev,
1397 struct rtnl_link_stats64 *storage);
1398 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1399 int (*ndo_get_offload_stats)(int attr_id,
1400 const struct net_device *dev,
1401 void *attr_data);
1402 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1403
1404 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1405 __be16 proto, u16 vid);
1406 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1407 __be16 proto, u16 vid);
1408#ifdef CONFIG_NET_POLL_CONTROLLER
1409 void (*ndo_poll_controller)(struct net_device *dev);
1410 int (*ndo_netpoll_setup)(struct net_device *dev,
1411 struct netpoll_info *info);
1412 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1413#endif
1414 int (*ndo_set_vf_mac)(struct net_device *dev,
1415 int queue, u8 *mac);
1416 int (*ndo_set_vf_vlan)(struct net_device *dev,
1417 int queue, u16 vlan,
1418 u8 qos, __be16 proto);
1419 int (*ndo_set_vf_rate)(struct net_device *dev,
1420 int vf, int min_tx_rate,
1421 int max_tx_rate);
1422 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1423 int vf, bool setting);
1424 int (*ndo_set_vf_trust)(struct net_device *dev,
1425 int vf, bool setting);
1426 int (*ndo_get_vf_config)(struct net_device *dev,
1427 int vf,
1428 struct ifla_vf_info *ivf);
1429 int (*ndo_set_vf_link_state)(struct net_device *dev,
1430 int vf, int link_state);
1431 int (*ndo_get_vf_stats)(struct net_device *dev,
1432 int vf,
1433 struct ifla_vf_stats
1434 *vf_stats);
1435 int (*ndo_set_vf_port)(struct net_device *dev,
1436 int vf,
1437 struct nlattr *port[]);
1438 int (*ndo_get_vf_port)(struct net_device *dev,
1439 int vf, struct sk_buff *skb);
1440 int (*ndo_get_vf_guid)(struct net_device *dev,
1441 int vf,
1442 struct ifla_vf_guid *node_guid,
1443 struct ifla_vf_guid *port_guid);
1444 int (*ndo_set_vf_guid)(struct net_device *dev,
1445 int vf, u64 guid,
1446 int guid_type);
1447 int (*ndo_set_vf_rss_query_en)(
1448 struct net_device *dev,
1449 int vf, bool setting);
1450 int (*ndo_setup_tc)(struct net_device *dev,
1451 enum tc_setup_type type,
1452 void *type_data);
1453#if IS_ENABLED(CONFIG_FCOE)
1454 int (*ndo_fcoe_enable)(struct net_device *dev);
1455 int (*ndo_fcoe_disable)(struct net_device *dev);
1456 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1457 u16 xid,
1458 struct scatterlist *sgl,
1459 unsigned int sgc);
1460 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1461 u16 xid);
1462 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1463 u16 xid,
1464 struct scatterlist *sgl,
1465 unsigned int sgc);
1466 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1467 struct netdev_fcoe_hbainfo *hbainfo);
1468#endif
1469
1470#if IS_ENABLED(CONFIG_LIBFCOE)
1471#define NETDEV_FCOE_WWNN 0
1472#define NETDEV_FCOE_WWPN 1
1473 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1474 u64 *wwn, int type);
1475#endif
1476
1477#ifdef CONFIG_RFS_ACCEL
1478 int (*ndo_rx_flow_steer)(struct net_device *dev,
1479 const struct sk_buff *skb,
1480 u16 rxq_index,
1481 u32 flow_id);
1482#endif
1483 int (*ndo_add_slave)(struct net_device *dev,
1484 struct net_device *slave_dev,
1485 struct netlink_ext_ack *extack);
1486 int (*ndo_del_slave)(struct net_device *dev,
1487 struct net_device *slave_dev);
1488 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1489 struct sk_buff *skb,
1490 bool all_slaves);
1491 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1492 struct sock *sk);
1493 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1494 netdev_features_t features);
1495 int (*ndo_set_features)(struct net_device *dev,
1496 netdev_features_t features);
1497 int (*ndo_neigh_construct)(struct net_device *dev,
1498 struct neighbour *n);
1499 void (*ndo_neigh_destroy)(struct net_device *dev,
1500 struct neighbour *n);
1501
1502 int (*ndo_fdb_add)(struct ndmsg *ndm,
1503 struct nlattr *tb[],
1504 struct net_device *dev,
1505 const unsigned char *addr,
1506 u16 vid,
1507 u16 flags,
1508 struct netlink_ext_ack *extack);
1509 int (*ndo_fdb_del)(struct ndmsg *ndm,
1510 struct nlattr *tb[],
1511 struct net_device *dev,
1512 const unsigned char *addr,
1513 u16 vid);
1514 int (*ndo_fdb_dump)(struct sk_buff *skb,
1515 struct netlink_callback *cb,
1516 struct net_device *dev,
1517 struct net_device *filter_dev,
1518 int *idx);
1519 int (*ndo_fdb_get)(struct sk_buff *skb,
1520 struct nlattr *tb[],
1521 struct net_device *dev,
1522 const unsigned char *addr,
1523 u16 vid, u32 portid, u32 seq,
1524 struct netlink_ext_ack *extack);
1525 int (*ndo_bridge_setlink)(struct net_device *dev,
1526 struct nlmsghdr *nlh,
1527 u16 flags,
1528 struct netlink_ext_ack *extack);
1529 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1530 u32 pid, u32 seq,
1531 struct net_device *dev,
1532 u32 filter_mask,
1533 int nlflags);
1534 int (*ndo_bridge_dellink)(struct net_device *dev,
1535 struct nlmsghdr *nlh,
1536 u16 flags);
1537 int (*ndo_change_carrier)(struct net_device *dev,
1538 bool new_carrier);
1539 int (*ndo_get_phys_port_id)(struct net_device *dev,
1540 struct netdev_phys_item_id *ppid);
1541 int (*ndo_get_port_parent_id)(struct net_device *dev,
1542 struct netdev_phys_item_id *ppid);
1543 int (*ndo_get_phys_port_name)(struct net_device *dev,
1544 char *name, size_t len);
1545 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1546 struct net_device *dev);
1547 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1548 void *priv);
1549
1550 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1551 int queue_index,
1552 u32 maxrate);
1553 int (*ndo_get_iflink)(const struct net_device *dev);
1554 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1555 struct sk_buff *skb);
1556 void (*ndo_set_rx_headroom)(struct net_device *dev,
1557 int needed_headroom);
1558 int (*ndo_bpf)(struct net_device *dev,
1559 struct netdev_bpf *bpf);
1560 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1561 struct xdp_frame **xdp,
1562 u32 flags);
1563 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1564 struct xdp_buff *xdp);
1565 int (*ndo_xsk_wakeup)(struct net_device *dev,
1566 u32 queue_id, u32 flags);
1567 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1568 int (*ndo_tunnel_ctl)(struct net_device *dev,
1569 struct ip_tunnel_parm *p, int cmd);
1570 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1571 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1572 struct net_device_path *path);
1573};
1574
1575/**
1576 * enum netdev_priv_flags - &struct net_device priv_flags
1577 *
1578 * These are the &struct net_device, they are only set internally
1579 * by drivers and used in the kernel. These flags are invisible to
1580 * userspace; this means that the order of these flags can change
1581 * during any kernel release.
1582 *
1583 * You should have a pretty good reason to be extending these flags.
1584 *
1585 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1586 * @IFF_EBRIDGE: Ethernet bridging device
1587 * @IFF_BONDING: bonding master or slave
1588 * @IFF_ISATAP: ISATAP interface (RFC4214)
1589 * @IFF_WAN_HDLC: WAN HDLC device
1590 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1591 * release skb->dst
1592 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1593 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1594 * @IFF_MACVLAN_PORT: device used as macvlan port
1595 * @IFF_BRIDGE_PORT: device used as bridge port
1596 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1597 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1598 * @IFF_UNICAST_FLT: Supports unicast filtering
1599 * @IFF_TEAM_PORT: device used as team port
1600 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1601 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1602 * change when it's running
1603 * @IFF_MACVLAN: Macvlan device
1604 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1605 * underlying stacked devices
1606 * @IFF_L3MDEV_MASTER: device is an L3 master device
1607 * @IFF_NO_QUEUE: device can run without qdisc attached
1608 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1609 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1610 * @IFF_TEAM: device is a team device
1611 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1612 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1613 * entity (i.e. the master device for bridged veth)
1614 * @IFF_MACSEC: device is a MACsec device
1615 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1616 * @IFF_FAILOVER: device is a failover master device
1617 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1618 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1619 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1620 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1621 * skb_headlen(skb) == 0 (data starts from frag0)
1622 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1623 */
1624enum netdev_priv_flags {
1625 IFF_802_1Q_VLAN = 1<<0,
1626 IFF_EBRIDGE = 1<<1,
1627 IFF_BONDING = 1<<2,
1628 IFF_ISATAP = 1<<3,
1629 IFF_WAN_HDLC = 1<<4,
1630 IFF_XMIT_DST_RELEASE = 1<<5,
1631 IFF_DONT_BRIDGE = 1<<6,
1632 IFF_DISABLE_NETPOLL = 1<<7,
1633 IFF_MACVLAN_PORT = 1<<8,
1634 IFF_BRIDGE_PORT = 1<<9,
1635 IFF_OVS_DATAPATH = 1<<10,
1636 IFF_TX_SKB_SHARING = 1<<11,
1637 IFF_UNICAST_FLT = 1<<12,
1638 IFF_TEAM_PORT = 1<<13,
1639 IFF_SUPP_NOFCS = 1<<14,
1640 IFF_LIVE_ADDR_CHANGE = 1<<15,
1641 IFF_MACVLAN = 1<<16,
1642 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1643 IFF_L3MDEV_MASTER = 1<<18,
1644 IFF_NO_QUEUE = 1<<19,
1645 IFF_OPENVSWITCH = 1<<20,
1646 IFF_L3MDEV_SLAVE = 1<<21,
1647 IFF_TEAM = 1<<22,
1648 IFF_RXFH_CONFIGURED = 1<<23,
1649 IFF_PHONY_HEADROOM = 1<<24,
1650 IFF_MACSEC = 1<<25,
1651 IFF_NO_RX_HANDLER = 1<<26,
1652 IFF_FAILOVER = 1<<27,
1653 IFF_FAILOVER_SLAVE = 1<<28,
1654 IFF_L3MDEV_RX_HANDLER = 1<<29,
1655 IFF_LIVE_RENAME_OK = 1<<30,
1656 IFF_TX_SKB_NO_LINEAR = 1<<31,
1657 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1658};
1659
1660#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1661#define IFF_EBRIDGE IFF_EBRIDGE
1662#define IFF_BONDING IFF_BONDING
1663#define IFF_ISATAP IFF_ISATAP
1664#define IFF_WAN_HDLC IFF_WAN_HDLC
1665#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1666#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1667#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1668#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1669#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1670#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1671#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1672#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1673#define IFF_TEAM_PORT IFF_TEAM_PORT
1674#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1675#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1676#define IFF_MACVLAN IFF_MACVLAN
1677#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1678#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1679#define IFF_NO_QUEUE IFF_NO_QUEUE
1680#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1681#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1682#define IFF_TEAM IFF_TEAM
1683#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1684#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1685#define IFF_MACSEC IFF_MACSEC
1686#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1687#define IFF_FAILOVER IFF_FAILOVER
1688#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1689#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1690#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1691#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1692
1693/* Specifies the type of the struct net_device::ml_priv pointer */
1694enum netdev_ml_priv_type {
1695 ML_PRIV_NONE,
1696 ML_PRIV_CAN,
1697};
1698
1699/**
1700 * struct net_device - The DEVICE structure.
1701 *
1702 * Actually, this whole structure is a big mistake. It mixes I/O
1703 * data with strictly "high-level" data, and it has to know about
1704 * almost every data structure used in the INET module.
1705 *
1706 * @name: This is the first field of the "visible" part of this structure
1707 * (i.e. as seen by users in the "Space.c" file). It is the name
1708 * of the interface.
1709 *
1710 * @name_node: Name hashlist node
1711 * @ifalias: SNMP alias
1712 * @mem_end: Shared memory end
1713 * @mem_start: Shared memory start
1714 * @base_addr: Device I/O address
1715 * @irq: Device IRQ number
1716 *
1717 * @state: Generic network queuing layer state, see netdev_state_t
1718 * @dev_list: The global list of network devices
1719 * @napi_list: List entry used for polling NAPI devices
1720 * @unreg_list: List entry when we are unregistering the
1721 * device; see the function unregister_netdev
1722 * @close_list: List entry used when we are closing the device
1723 * @ptype_all: Device-specific packet handlers for all protocols
1724 * @ptype_specific: Device-specific, protocol-specific packet handlers
1725 *
1726 * @adj_list: Directly linked devices, like slaves for bonding
1727 * @features: Currently active device features
1728 * @hw_features: User-changeable features
1729 *
1730 * @wanted_features: User-requested features
1731 * @vlan_features: Mask of features inheritable by VLAN devices
1732 *
1733 * @hw_enc_features: Mask of features inherited by encapsulating devices
1734 * This field indicates what encapsulation
1735 * offloads the hardware is capable of doing,
1736 * and drivers will need to set them appropriately.
1737 *
1738 * @mpls_features: Mask of features inheritable by MPLS
1739 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1740 *
1741 * @ifindex: interface index
1742 * @group: The group the device belongs to
1743 *
1744 * @stats: Statistics struct, which was left as a legacy, use
1745 * rtnl_link_stats64 instead
1746 *
1747 * @core_stats: core networking counters,
1748 * do not use this in drivers
1749 * @carrier_up_count: Number of times the carrier has been up
1750 * @carrier_down_count: Number of times the carrier has been down
1751 *
1752 * @wireless_handlers: List of functions to handle Wireless Extensions,
1753 * instead of ioctl,
1754 * see <net/iw_handler.h> for details.
1755 * @wireless_data: Instance data managed by the core of wireless extensions
1756 *
1757 * @netdev_ops: Includes several pointers to callbacks,
1758 * if one wants to override the ndo_*() functions
1759 * @ethtool_ops: Management operations
1760 * @l3mdev_ops: Layer 3 master device operations
1761 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1762 * discovery handling. Necessary for e.g. 6LoWPAN.
1763 * @xfrmdev_ops: Transformation offload operations
1764 * @tlsdev_ops: Transport Layer Security offload operations
1765 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1766 * of Layer 2 headers.
1767 *
1768 * @flags: Interface flags (a la BSD)
1769 * @priv_flags: Like 'flags' but invisible to userspace,
1770 * see if.h for the definitions
1771 * @gflags: Global flags ( kept as legacy )
1772 * @padded: How much padding added by alloc_netdev()
1773 * @operstate: RFC2863 operstate
1774 * @link_mode: Mapping policy to operstate
1775 * @if_port: Selectable AUI, TP, ...
1776 * @dma: DMA channel
1777 * @mtu: Interface MTU value
1778 * @min_mtu: Interface Minimum MTU value
1779 * @max_mtu: Interface Maximum MTU value
1780 * @type: Interface hardware type
1781 * @hard_header_len: Maximum hardware header length.
1782 * @min_header_len: Minimum hardware header length
1783 *
1784 * @needed_headroom: Extra headroom the hardware may need, but not in all
1785 * cases can this be guaranteed
1786 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1787 * cases can this be guaranteed. Some cases also use
1788 * LL_MAX_HEADER instead to allocate the skb
1789 *
1790 * interface address info:
1791 *
1792 * @perm_addr: Permanent hw address
1793 * @addr_assign_type: Hw address assignment type
1794 * @addr_len: Hardware address length
1795 * @upper_level: Maximum depth level of upper devices.
1796 * @lower_level: Maximum depth level of lower devices.
1797 * @neigh_priv_len: Used in neigh_alloc()
1798 * @dev_id: Used to differentiate devices that share
1799 * the same link layer address
1800 * @dev_port: Used to differentiate devices that share
1801 * the same function
1802 * @addr_list_lock: XXX: need comments on this one
1803 * @name_assign_type: network interface name assignment type
1804 * @uc_promisc: Counter that indicates promiscuous mode
1805 * has been enabled due to the need to listen to
1806 * additional unicast addresses in a device that
1807 * does not implement ndo_set_rx_mode()
1808 * @uc: unicast mac addresses
1809 * @mc: multicast mac addresses
1810 * @dev_addrs: list of device hw addresses
1811 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1812 * @promiscuity: Number of times the NIC is told to work in
1813 * promiscuous mode; if it becomes 0 the NIC will
1814 * exit promiscuous mode
1815 * @allmulti: Counter, enables or disables allmulticast mode
1816 *
1817 * @vlan_info: VLAN info
1818 * @dsa_ptr: dsa specific data
1819 * @tipc_ptr: TIPC specific data
1820 * @atalk_ptr: AppleTalk link
1821 * @ip_ptr: IPv4 specific data
1822 * @dn_ptr: DECnet specific data
1823 * @ip6_ptr: IPv6 specific data
1824 * @ax25_ptr: AX.25 specific data
1825 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1826 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1827 * device struct
1828 * @mpls_ptr: mpls_dev struct pointer
1829 * @mctp_ptr: MCTP specific data
1830 *
1831 * @dev_addr: Hw address (before bcast,
1832 * because most packets are unicast)
1833 *
1834 * @_rx: Array of RX queues
1835 * @num_rx_queues: Number of RX queues
1836 * allocated at register_netdev() time
1837 * @real_num_rx_queues: Number of RX queues currently active in device
1838 * @xdp_prog: XDP sockets filter program pointer
1839 * @gro_flush_timeout: timeout for GRO layer in NAPI
1840 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1841 * allow to avoid NIC hard IRQ, on busy queues.
1842 *
1843 * @rx_handler: handler for received packets
1844 * @rx_handler_data: XXX: need comments on this one
1845 * @miniq_ingress: ingress/clsact qdisc specific data for
1846 * ingress processing
1847 * @ingress_queue: XXX: need comments on this one
1848 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1849 * @broadcast: hw bcast address
1850 *
1851 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1852 * indexed by RX queue number. Assigned by driver.
1853 * This must only be set if the ndo_rx_flow_steer
1854 * operation is defined
1855 * @index_hlist: Device index hash chain
1856 *
1857 * @_tx: Array of TX queues
1858 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1859 * @real_num_tx_queues: Number of TX queues currently active in device
1860 * @qdisc: Root qdisc from userspace point of view
1861 * @tx_queue_len: Max frames per queue allowed
1862 * @tx_global_lock: XXX: need comments on this one
1863 * @xdp_bulkq: XDP device bulk queue
1864 * @xps_maps: all CPUs/RXQs maps for XPS device
1865 *
1866 * @xps_maps: XXX: need comments on this one
1867 * @miniq_egress: clsact qdisc specific data for
1868 * egress processing
1869 * @nf_hooks_egress: netfilter hooks executed for egress packets
1870 * @qdisc_hash: qdisc hash table
1871 * @watchdog_timeo: Represents the timeout that is used by
1872 * the watchdog (see dev_watchdog())
1873 * @watchdog_timer: List of timers
1874 *
1875 * @proto_down_reason: reason a netdev interface is held down
1876 * @pcpu_refcnt: Number of references to this device
1877 * @dev_refcnt: Number of references to this device
1878 * @refcnt_tracker: Tracker directory for tracked references to this device
1879 * @todo_list: Delayed register/unregister
1880 * @link_watch_list: XXX: need comments on this one
1881 *
1882 * @reg_state: Register/unregister state machine
1883 * @dismantle: Device is going to be freed
1884 * @rtnl_link_state: This enum represents the phases of creating
1885 * a new link
1886 *
1887 * @needs_free_netdev: Should unregister perform free_netdev?
1888 * @priv_destructor: Called from unregister
1889 * @npinfo: XXX: need comments on this one
1890 * @nd_net: Network namespace this network device is inside
1891 *
1892 * @ml_priv: Mid-layer private
1893 * @ml_priv_type: Mid-layer private type
1894 * @lstats: Loopback statistics
1895 * @tstats: Tunnel statistics
1896 * @dstats: Dummy statistics
1897 * @vstats: Virtual ethernet statistics
1898 *
1899 * @garp_port: GARP
1900 * @mrp_port: MRP
1901 *
1902 * @dm_private: Drop monitor private
1903 *
1904 * @dev: Class/net/name entry
1905 * @sysfs_groups: Space for optional device, statistics and wireless
1906 * sysfs groups
1907 *
1908 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1909 * @rtnl_link_ops: Rtnl_link_ops
1910 *
1911 * @gso_max_size: Maximum size of generic segmentation offload
1912 * @gso_max_segs: Maximum number of segments that can be passed to the
1913 * NIC for GSO
1914 *
1915 * @dcbnl_ops: Data Center Bridging netlink ops
1916 * @num_tc: Number of traffic classes in the net device
1917 * @tc_to_txq: XXX: need comments on this one
1918 * @prio_tc_map: XXX: need comments on this one
1919 *
1920 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1921 *
1922 * @priomap: XXX: need comments on this one
1923 * @phydev: Physical device may attach itself
1924 * for hardware timestamping
1925 * @sfp_bus: attached &struct sfp_bus structure.
1926 *
1927 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1928 *
1929 * @proto_down: protocol port state information can be sent to the
1930 * switch driver and used to set the phys state of the
1931 * switch port.
1932 *
1933 * @wol_enabled: Wake-on-LAN is enabled
1934 *
1935 * @threaded: napi threaded mode is enabled
1936 *
1937 * @net_notifier_list: List of per-net netdev notifier block
1938 * that follow this device when it is moved
1939 * to another network namespace.
1940 *
1941 * @macsec_ops: MACsec offloading ops
1942 *
1943 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1944 * offload capabilities of the device
1945 * @udp_tunnel_nic: UDP tunnel offload state
1946 * @xdp_state: stores info on attached XDP BPF programs
1947 *
1948 * @nested_level: Used as a parameter of spin_lock_nested() of
1949 * dev->addr_list_lock.
1950 * @unlink_list: As netif_addr_lock() can be called recursively,
1951 * keep a list of interfaces to be deleted.
1952 * @gro_max_size: Maximum size of aggregated packet in generic
1953 * receive offload (GRO)
1954 *
1955 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
1956 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
1957 * @watchdog_dev_tracker: refcount tracker used by watchdog.
1958 * @dev_registered_tracker: tracker for reference held while
1959 * registered
1960 * @offload_xstats_l3: L3 HW stats for this netdevice.
1961 *
1962 * FIXME: cleanup struct net_device such that network protocol info
1963 * moves out.
1964 */
1965
1966struct net_device {
1967 char name[IFNAMSIZ];
1968 struct netdev_name_node *name_node;
1969 struct dev_ifalias __rcu *ifalias;
1970 /*
1971 * I/O specific fields
1972 * FIXME: Merge these and struct ifmap into one
1973 */
1974 unsigned long mem_end;
1975 unsigned long mem_start;
1976 unsigned long base_addr;
1977
1978 /*
1979 * Some hardware also needs these fields (state,dev_list,
1980 * napi_list,unreg_list,close_list) but they are not
1981 * part of the usual set specified in Space.c.
1982 */
1983
1984 unsigned long state;
1985
1986 struct list_head dev_list;
1987 struct list_head napi_list;
1988 struct list_head unreg_list;
1989 struct list_head close_list;
1990 struct list_head ptype_all;
1991 struct list_head ptype_specific;
1992
1993 struct {
1994 struct list_head upper;
1995 struct list_head lower;
1996 } adj_list;
1997
1998 /* Read-mostly cache-line for fast-path access */
1999 unsigned int flags;
2000 unsigned long long priv_flags;
2001 const struct net_device_ops *netdev_ops;
2002 int ifindex;
2003 unsigned short gflags;
2004 unsigned short hard_header_len;
2005
2006 /* Note : dev->mtu is often read without holding a lock.
2007 * Writers usually hold RTNL.
2008 * It is recommended to use READ_ONCE() to annotate the reads,
2009 * and to use WRITE_ONCE() to annotate the writes.
2010 */
2011 unsigned int mtu;
2012 unsigned short needed_headroom;
2013 unsigned short needed_tailroom;
2014
2015 netdev_features_t features;
2016 netdev_features_t hw_features;
2017 netdev_features_t wanted_features;
2018 netdev_features_t vlan_features;
2019 netdev_features_t hw_enc_features;
2020 netdev_features_t mpls_features;
2021 netdev_features_t gso_partial_features;
2022
2023 unsigned int min_mtu;
2024 unsigned int max_mtu;
2025 unsigned short type;
2026 unsigned char min_header_len;
2027 unsigned char name_assign_type;
2028
2029 int group;
2030
2031 struct net_device_stats stats; /* not used by modern drivers */
2032
2033 struct net_device_core_stats __percpu *core_stats;
2034
2035 /* Stats to monitor link on/off, flapping */
2036 atomic_t carrier_up_count;
2037 atomic_t carrier_down_count;
2038
2039#ifdef CONFIG_WIRELESS_EXT
2040 const struct iw_handler_def *wireless_handlers;
2041 struct iw_public_data *wireless_data;
2042#endif
2043 const struct ethtool_ops *ethtool_ops;
2044#ifdef CONFIG_NET_L3_MASTER_DEV
2045 const struct l3mdev_ops *l3mdev_ops;
2046#endif
2047#if IS_ENABLED(CONFIG_IPV6)
2048 const struct ndisc_ops *ndisc_ops;
2049#endif
2050
2051#ifdef CONFIG_XFRM_OFFLOAD
2052 const struct xfrmdev_ops *xfrmdev_ops;
2053#endif
2054
2055#if IS_ENABLED(CONFIG_TLS_DEVICE)
2056 const struct tlsdev_ops *tlsdev_ops;
2057#endif
2058
2059 const struct header_ops *header_ops;
2060
2061 unsigned char operstate;
2062 unsigned char link_mode;
2063
2064 unsigned char if_port;
2065 unsigned char dma;
2066
2067 /* Interface address info. */
2068 unsigned char perm_addr[MAX_ADDR_LEN];
2069 unsigned char addr_assign_type;
2070 unsigned char addr_len;
2071 unsigned char upper_level;
2072 unsigned char lower_level;
2073
2074 unsigned short neigh_priv_len;
2075 unsigned short dev_id;
2076 unsigned short dev_port;
2077 unsigned short padded;
2078
2079 spinlock_t addr_list_lock;
2080 int irq;
2081
2082 struct netdev_hw_addr_list uc;
2083 struct netdev_hw_addr_list mc;
2084 struct netdev_hw_addr_list dev_addrs;
2085
2086#ifdef CONFIG_SYSFS
2087 struct kset *queues_kset;
2088#endif
2089#ifdef CONFIG_LOCKDEP
2090 struct list_head unlink_list;
2091#endif
2092 unsigned int promiscuity;
2093 unsigned int allmulti;
2094 bool uc_promisc;
2095#ifdef CONFIG_LOCKDEP
2096 unsigned char nested_level;
2097#endif
2098
2099
2100 /* Protocol-specific pointers */
2101
2102#if IS_ENABLED(CONFIG_VLAN_8021Q)
2103 struct vlan_info __rcu *vlan_info;
2104#endif
2105#if IS_ENABLED(CONFIG_NET_DSA)
2106 struct dsa_port *dsa_ptr;
2107#endif
2108#if IS_ENABLED(CONFIG_TIPC)
2109 struct tipc_bearer __rcu *tipc_ptr;
2110#endif
2111#if IS_ENABLED(CONFIG_ATALK)
2112 void *atalk_ptr;
2113#endif
2114 struct in_device __rcu *ip_ptr;
2115#if IS_ENABLED(CONFIG_DECNET)
2116 struct dn_dev __rcu *dn_ptr;
2117#endif
2118 struct inet6_dev __rcu *ip6_ptr;
2119#if IS_ENABLED(CONFIG_AX25)
2120 void *ax25_ptr;
2121#endif
2122 struct wireless_dev *ieee80211_ptr;
2123 struct wpan_dev *ieee802154_ptr;
2124#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2125 struct mpls_dev __rcu *mpls_ptr;
2126#endif
2127#if IS_ENABLED(CONFIG_MCTP)
2128 struct mctp_dev __rcu *mctp_ptr;
2129#endif
2130
2131/*
2132 * Cache lines mostly used on receive path (including eth_type_trans())
2133 */
2134 /* Interface address info used in eth_type_trans() */
2135 const unsigned char *dev_addr;
2136
2137 struct netdev_rx_queue *_rx;
2138 unsigned int num_rx_queues;
2139 unsigned int real_num_rx_queues;
2140
2141 struct bpf_prog __rcu *xdp_prog;
2142 unsigned long gro_flush_timeout;
2143 int napi_defer_hard_irqs;
2144#define GRO_MAX_SIZE 65536
2145 unsigned int gro_max_size;
2146 rx_handler_func_t __rcu *rx_handler;
2147 void __rcu *rx_handler_data;
2148
2149#ifdef CONFIG_NET_CLS_ACT
2150 struct mini_Qdisc __rcu *miniq_ingress;
2151#endif
2152 struct netdev_queue __rcu *ingress_queue;
2153#ifdef CONFIG_NETFILTER_INGRESS
2154 struct nf_hook_entries __rcu *nf_hooks_ingress;
2155#endif
2156
2157 unsigned char broadcast[MAX_ADDR_LEN];
2158#ifdef CONFIG_RFS_ACCEL
2159 struct cpu_rmap *rx_cpu_rmap;
2160#endif
2161 struct hlist_node index_hlist;
2162
2163/*
2164 * Cache lines mostly used on transmit path
2165 */
2166 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2167 unsigned int num_tx_queues;
2168 unsigned int real_num_tx_queues;
2169 struct Qdisc __rcu *qdisc;
2170 unsigned int tx_queue_len;
2171 spinlock_t tx_global_lock;
2172
2173 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2174
2175#ifdef CONFIG_XPS
2176 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2177#endif
2178#ifdef CONFIG_NET_CLS_ACT
2179 struct mini_Qdisc __rcu *miniq_egress;
2180#endif
2181#ifdef CONFIG_NETFILTER_EGRESS
2182 struct nf_hook_entries __rcu *nf_hooks_egress;
2183#endif
2184
2185#ifdef CONFIG_NET_SCHED
2186 DECLARE_HASHTABLE (qdisc_hash, 4);
2187#endif
2188 /* These may be needed for future network-power-down code. */
2189 struct timer_list watchdog_timer;
2190 int watchdog_timeo;
2191
2192 u32 proto_down_reason;
2193
2194 struct list_head todo_list;
2195
2196#ifdef CONFIG_PCPU_DEV_REFCNT
2197 int __percpu *pcpu_refcnt;
2198#else
2199 refcount_t dev_refcnt;
2200#endif
2201 struct ref_tracker_dir refcnt_tracker;
2202
2203 struct list_head link_watch_list;
2204
2205 enum { NETREG_UNINITIALIZED=0,
2206 NETREG_REGISTERED, /* completed register_netdevice */
2207 NETREG_UNREGISTERING, /* called unregister_netdevice */
2208 NETREG_UNREGISTERED, /* completed unregister todo */
2209 NETREG_RELEASED, /* called free_netdev */
2210 NETREG_DUMMY, /* dummy device for NAPI poll */
2211 } reg_state:8;
2212
2213 bool dismantle;
2214
2215 enum {
2216 RTNL_LINK_INITIALIZED,
2217 RTNL_LINK_INITIALIZING,
2218 } rtnl_link_state:16;
2219
2220 bool needs_free_netdev;
2221 void (*priv_destructor)(struct net_device *dev);
2222
2223#ifdef CONFIG_NETPOLL
2224 struct netpoll_info __rcu *npinfo;
2225#endif
2226
2227 possible_net_t nd_net;
2228
2229 /* mid-layer private */
2230 void *ml_priv;
2231 enum netdev_ml_priv_type ml_priv_type;
2232
2233 union {
2234 struct pcpu_lstats __percpu *lstats;
2235 struct pcpu_sw_netstats __percpu *tstats;
2236 struct pcpu_dstats __percpu *dstats;
2237 };
2238
2239#if IS_ENABLED(CONFIG_GARP)
2240 struct garp_port __rcu *garp_port;
2241#endif
2242#if IS_ENABLED(CONFIG_MRP)
2243 struct mrp_port __rcu *mrp_port;
2244#endif
2245#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2246 struct dm_hw_stat_delta __rcu *dm_private;
2247#endif
2248 struct device dev;
2249 const struct attribute_group *sysfs_groups[4];
2250 const struct attribute_group *sysfs_rx_queue_group;
2251
2252 const struct rtnl_link_ops *rtnl_link_ops;
2253
2254 /* for setting kernel sock attribute on TCP connection setup */
2255#define GSO_MAX_SIZE 65536
2256 unsigned int gso_max_size;
2257#define GSO_MAX_SEGS 65535
2258 u16 gso_max_segs;
2259
2260#ifdef CONFIG_DCB
2261 const struct dcbnl_rtnl_ops *dcbnl_ops;
2262#endif
2263 s16 num_tc;
2264 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2265 u8 prio_tc_map[TC_BITMASK + 1];
2266
2267#if IS_ENABLED(CONFIG_FCOE)
2268 unsigned int fcoe_ddp_xid;
2269#endif
2270#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2271 struct netprio_map __rcu *priomap;
2272#endif
2273 struct phy_device *phydev;
2274 struct sfp_bus *sfp_bus;
2275 struct lock_class_key *qdisc_tx_busylock;
2276 bool proto_down;
2277 unsigned wol_enabled:1;
2278 unsigned threaded:1;
2279
2280 struct list_head net_notifier_list;
2281
2282#if IS_ENABLED(CONFIG_MACSEC)
2283 /* MACsec management functions */
2284 const struct macsec_ops *macsec_ops;
2285#endif
2286 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2287 struct udp_tunnel_nic *udp_tunnel_nic;
2288
2289 /* protected by rtnl_lock */
2290 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2291
2292 u8 dev_addr_shadow[MAX_ADDR_LEN];
2293 netdevice_tracker linkwatch_dev_tracker;
2294 netdevice_tracker watchdog_dev_tracker;
2295 netdevice_tracker dev_registered_tracker;
2296 struct rtnl_hw_stats64 *offload_xstats_l3;
2297};
2298#define to_net_dev(d) container_of(d, struct net_device, dev)
2299
2300static inline bool netif_elide_gro(const struct net_device *dev)
2301{
2302 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2303 return true;
2304 return false;
2305}
2306
2307#define NETDEV_ALIGN 32
2308
2309static inline
2310int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2311{
2312 return dev->prio_tc_map[prio & TC_BITMASK];
2313}
2314
2315static inline
2316int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2317{
2318 if (tc >= dev->num_tc)
2319 return -EINVAL;
2320
2321 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2322 return 0;
2323}
2324
2325int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2326void netdev_reset_tc(struct net_device *dev);
2327int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2328int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2329
2330static inline
2331int netdev_get_num_tc(struct net_device *dev)
2332{
2333 return dev->num_tc;
2334}
2335
2336static inline void net_prefetch(void *p)
2337{
2338 prefetch(p);
2339#if L1_CACHE_BYTES < 128
2340 prefetch((u8 *)p + L1_CACHE_BYTES);
2341#endif
2342}
2343
2344static inline void net_prefetchw(void *p)
2345{
2346 prefetchw(p);
2347#if L1_CACHE_BYTES < 128
2348 prefetchw((u8 *)p + L1_CACHE_BYTES);
2349#endif
2350}
2351
2352void netdev_unbind_sb_channel(struct net_device *dev,
2353 struct net_device *sb_dev);
2354int netdev_bind_sb_channel_queue(struct net_device *dev,
2355 struct net_device *sb_dev,
2356 u8 tc, u16 count, u16 offset);
2357int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2358static inline int netdev_get_sb_channel(struct net_device *dev)
2359{
2360 return max_t(int, -dev->num_tc, 0);
2361}
2362
2363static inline
2364struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2365 unsigned int index)
2366{
2367 return &dev->_tx[index];
2368}
2369
2370static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2371 const struct sk_buff *skb)
2372{
2373 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2374}
2375
2376static inline void netdev_for_each_tx_queue(struct net_device *dev,
2377 void (*f)(struct net_device *,
2378 struct netdev_queue *,
2379 void *),
2380 void *arg)
2381{
2382 unsigned int i;
2383
2384 for (i = 0; i < dev->num_tx_queues; i++)
2385 f(dev, &dev->_tx[i], arg);
2386}
2387
2388#define netdev_lockdep_set_classes(dev) \
2389{ \
2390 static struct lock_class_key qdisc_tx_busylock_key; \
2391 static struct lock_class_key qdisc_xmit_lock_key; \
2392 static struct lock_class_key dev_addr_list_lock_key; \
2393 unsigned int i; \
2394 \
2395 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2396 lockdep_set_class(&(dev)->addr_list_lock, \
2397 &dev_addr_list_lock_key); \
2398 for (i = 0; i < (dev)->num_tx_queues; i++) \
2399 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2400 &qdisc_xmit_lock_key); \
2401}
2402
2403u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2404 struct net_device *sb_dev);
2405struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2406 struct sk_buff *skb,
2407 struct net_device *sb_dev);
2408
2409/* returns the headroom that the master device needs to take in account
2410 * when forwarding to this dev
2411 */
2412static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2413{
2414 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2415}
2416
2417static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2418{
2419 if (dev->netdev_ops->ndo_set_rx_headroom)
2420 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2421}
2422
2423/* set the device rx headroom to the dev's default */
2424static inline void netdev_reset_rx_headroom(struct net_device *dev)
2425{
2426 netdev_set_rx_headroom(dev, -1);
2427}
2428
2429static inline void *netdev_get_ml_priv(struct net_device *dev,
2430 enum netdev_ml_priv_type type)
2431{
2432 if (dev->ml_priv_type != type)
2433 return NULL;
2434
2435 return dev->ml_priv;
2436}
2437
2438static inline void netdev_set_ml_priv(struct net_device *dev,
2439 void *ml_priv,
2440 enum netdev_ml_priv_type type)
2441{
2442 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2443 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2444 dev->ml_priv_type, type);
2445 WARN(!dev->ml_priv_type && dev->ml_priv,
2446 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2447
2448 dev->ml_priv = ml_priv;
2449 dev->ml_priv_type = type;
2450}
2451
2452/*
2453 * Net namespace inlines
2454 */
2455static inline
2456struct net *dev_net(const struct net_device *dev)
2457{
2458 return read_pnet(&dev->nd_net);
2459}
2460
2461static inline
2462void dev_net_set(struct net_device *dev, struct net *net)
2463{
2464 write_pnet(&dev->nd_net, net);
2465}
2466
2467/**
2468 * netdev_priv - access network device private data
2469 * @dev: network device
2470 *
2471 * Get network device private data
2472 */
2473static inline void *netdev_priv(const struct net_device *dev)
2474{
2475 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2476}
2477
2478/* Set the sysfs physical device reference for the network logical device
2479 * if set prior to registration will cause a symlink during initialization.
2480 */
2481#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2482
2483/* Set the sysfs device type for the network logical device to allow
2484 * fine-grained identification of different network device types. For
2485 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2486 */
2487#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2488
2489/* Default NAPI poll() weight
2490 * Device drivers are strongly advised to not use bigger value
2491 */
2492#define NAPI_POLL_WEIGHT 64
2493
2494/**
2495 * netif_napi_add - initialize a NAPI context
2496 * @dev: network device
2497 * @napi: NAPI context
2498 * @poll: polling function
2499 * @weight: default weight
2500 *
2501 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2502 * *any* of the other NAPI-related functions.
2503 */
2504void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2505 int (*poll)(struct napi_struct *, int), int weight);
2506
2507/**
2508 * netif_tx_napi_add - initialize a NAPI context
2509 * @dev: network device
2510 * @napi: NAPI context
2511 * @poll: polling function
2512 * @weight: default weight
2513 *
2514 * This variant of netif_napi_add() should be used from drivers using NAPI
2515 * to exclusively poll a TX queue.
2516 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2517 */
2518static inline void netif_tx_napi_add(struct net_device *dev,
2519 struct napi_struct *napi,
2520 int (*poll)(struct napi_struct *, int),
2521 int weight)
2522{
2523 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2524 netif_napi_add(dev, napi, poll, weight);
2525}
2526
2527/**
2528 * __netif_napi_del - remove a NAPI context
2529 * @napi: NAPI context
2530 *
2531 * Warning: caller must observe RCU grace period before freeing memory
2532 * containing @napi. Drivers might want to call this helper to combine
2533 * all the needed RCU grace periods into a single one.
2534 */
2535void __netif_napi_del(struct napi_struct *napi);
2536
2537/**
2538 * netif_napi_del - remove a NAPI context
2539 * @napi: NAPI context
2540 *
2541 * netif_napi_del() removes a NAPI context from the network device NAPI list
2542 */
2543static inline void netif_napi_del(struct napi_struct *napi)
2544{
2545 __netif_napi_del(napi);
2546 synchronize_net();
2547}
2548
2549struct packet_type {
2550 __be16 type; /* This is really htons(ether_type). */
2551 bool ignore_outgoing;
2552 struct net_device *dev; /* NULL is wildcarded here */
2553 netdevice_tracker dev_tracker;
2554 int (*func) (struct sk_buff *,
2555 struct net_device *,
2556 struct packet_type *,
2557 struct net_device *);
2558 void (*list_func) (struct list_head *,
2559 struct packet_type *,
2560 struct net_device *);
2561 bool (*id_match)(struct packet_type *ptype,
2562 struct sock *sk);
2563 struct net *af_packet_net;
2564 void *af_packet_priv;
2565 struct list_head list;
2566};
2567
2568struct offload_callbacks {
2569 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2570 netdev_features_t features);
2571 struct sk_buff *(*gro_receive)(struct list_head *head,
2572 struct sk_buff *skb);
2573 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2574};
2575
2576struct packet_offload {
2577 __be16 type; /* This is really htons(ether_type). */
2578 u16 priority;
2579 struct offload_callbacks callbacks;
2580 struct list_head list;
2581};
2582
2583/* often modified stats are per-CPU, other are shared (netdev->stats) */
2584struct pcpu_sw_netstats {
2585 u64 rx_packets;
2586 u64 rx_bytes;
2587 u64 tx_packets;
2588 u64 tx_bytes;
2589 struct u64_stats_sync syncp;
2590} __aligned(4 * sizeof(u64));
2591
2592struct pcpu_lstats {
2593 u64_stats_t packets;
2594 u64_stats_t bytes;
2595 struct u64_stats_sync syncp;
2596} __aligned(2 * sizeof(u64));
2597
2598void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2599
2600static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2601{
2602 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2603
2604 u64_stats_update_begin(&tstats->syncp);
2605 tstats->rx_bytes += len;
2606 tstats->rx_packets++;
2607 u64_stats_update_end(&tstats->syncp);
2608}
2609
2610static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2611 unsigned int packets,
2612 unsigned int len)
2613{
2614 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2615
2616 u64_stats_update_begin(&tstats->syncp);
2617 tstats->tx_bytes += len;
2618 tstats->tx_packets += packets;
2619 u64_stats_update_end(&tstats->syncp);
2620}
2621
2622static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2623{
2624 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2625
2626 u64_stats_update_begin(&lstats->syncp);
2627 u64_stats_add(&lstats->bytes, len);
2628 u64_stats_inc(&lstats->packets);
2629 u64_stats_update_end(&lstats->syncp);
2630}
2631
2632#define __netdev_alloc_pcpu_stats(type, gfp) \
2633({ \
2634 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2635 if (pcpu_stats) { \
2636 int __cpu; \
2637 for_each_possible_cpu(__cpu) { \
2638 typeof(type) *stat; \
2639 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2640 u64_stats_init(&stat->syncp); \
2641 } \
2642 } \
2643 pcpu_stats; \
2644})
2645
2646#define netdev_alloc_pcpu_stats(type) \
2647 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2648
2649#define devm_netdev_alloc_pcpu_stats(dev, type) \
2650({ \
2651 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2652 if (pcpu_stats) { \
2653 int __cpu; \
2654 for_each_possible_cpu(__cpu) { \
2655 typeof(type) *stat; \
2656 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2657 u64_stats_init(&stat->syncp); \
2658 } \
2659 } \
2660 pcpu_stats; \
2661})
2662
2663enum netdev_lag_tx_type {
2664 NETDEV_LAG_TX_TYPE_UNKNOWN,
2665 NETDEV_LAG_TX_TYPE_RANDOM,
2666 NETDEV_LAG_TX_TYPE_BROADCAST,
2667 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2668 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2669 NETDEV_LAG_TX_TYPE_HASH,
2670};
2671
2672enum netdev_lag_hash {
2673 NETDEV_LAG_HASH_NONE,
2674 NETDEV_LAG_HASH_L2,
2675 NETDEV_LAG_HASH_L34,
2676 NETDEV_LAG_HASH_L23,
2677 NETDEV_LAG_HASH_E23,
2678 NETDEV_LAG_HASH_E34,
2679 NETDEV_LAG_HASH_VLAN_SRCMAC,
2680 NETDEV_LAG_HASH_UNKNOWN,
2681};
2682
2683struct netdev_lag_upper_info {
2684 enum netdev_lag_tx_type tx_type;
2685 enum netdev_lag_hash hash_type;
2686};
2687
2688struct netdev_lag_lower_state_info {
2689 u8 link_up : 1,
2690 tx_enabled : 1;
2691};
2692
2693#include <linux/notifier.h>
2694
2695/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2696 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2697 * adding new types.
2698 */
2699enum netdev_cmd {
2700 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2701 NETDEV_DOWN,
2702 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2703 detected a hardware crash and restarted
2704 - we can use this eg to kick tcp sessions
2705 once done */
2706 NETDEV_CHANGE, /* Notify device state change */
2707 NETDEV_REGISTER,
2708 NETDEV_UNREGISTER,
2709 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2710 NETDEV_CHANGEADDR, /* notify after the address change */
2711 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2712 NETDEV_GOING_DOWN,
2713 NETDEV_CHANGENAME,
2714 NETDEV_FEAT_CHANGE,
2715 NETDEV_BONDING_FAILOVER,
2716 NETDEV_PRE_UP,
2717 NETDEV_PRE_TYPE_CHANGE,
2718 NETDEV_POST_TYPE_CHANGE,
2719 NETDEV_POST_INIT,
2720 NETDEV_RELEASE,
2721 NETDEV_NOTIFY_PEERS,
2722 NETDEV_JOIN,
2723 NETDEV_CHANGEUPPER,
2724 NETDEV_RESEND_IGMP,
2725 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2726 NETDEV_CHANGEINFODATA,
2727 NETDEV_BONDING_INFO,
2728 NETDEV_PRECHANGEUPPER,
2729 NETDEV_CHANGELOWERSTATE,
2730 NETDEV_UDP_TUNNEL_PUSH_INFO,
2731 NETDEV_UDP_TUNNEL_DROP_INFO,
2732 NETDEV_CHANGE_TX_QUEUE_LEN,
2733 NETDEV_CVLAN_FILTER_PUSH_INFO,
2734 NETDEV_CVLAN_FILTER_DROP_INFO,
2735 NETDEV_SVLAN_FILTER_PUSH_INFO,
2736 NETDEV_SVLAN_FILTER_DROP_INFO,
2737 NETDEV_OFFLOAD_XSTATS_ENABLE,
2738 NETDEV_OFFLOAD_XSTATS_DISABLE,
2739 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2740 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2741};
2742const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2743
2744int register_netdevice_notifier(struct notifier_block *nb);
2745int unregister_netdevice_notifier(struct notifier_block *nb);
2746int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2747int unregister_netdevice_notifier_net(struct net *net,
2748 struct notifier_block *nb);
2749int register_netdevice_notifier_dev_net(struct net_device *dev,
2750 struct notifier_block *nb,
2751 struct netdev_net_notifier *nn);
2752int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2753 struct notifier_block *nb,
2754 struct netdev_net_notifier *nn);
2755
2756struct netdev_notifier_info {
2757 struct net_device *dev;
2758 struct netlink_ext_ack *extack;
2759};
2760
2761struct netdev_notifier_info_ext {
2762 struct netdev_notifier_info info; /* must be first */
2763 union {
2764 u32 mtu;
2765 } ext;
2766};
2767
2768struct netdev_notifier_change_info {
2769 struct netdev_notifier_info info; /* must be first */
2770 unsigned int flags_changed;
2771};
2772
2773struct netdev_notifier_changeupper_info {
2774 struct netdev_notifier_info info; /* must be first */
2775 struct net_device *upper_dev; /* new upper dev */
2776 bool master; /* is upper dev master */
2777 bool linking; /* is the notification for link or unlink */
2778 void *upper_info; /* upper dev info */
2779};
2780
2781struct netdev_notifier_changelowerstate_info {
2782 struct netdev_notifier_info info; /* must be first */
2783 void *lower_state_info; /* is lower dev state */
2784};
2785
2786struct netdev_notifier_pre_changeaddr_info {
2787 struct netdev_notifier_info info; /* must be first */
2788 const unsigned char *dev_addr;
2789};
2790
2791enum netdev_offload_xstats_type {
2792 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2793};
2794
2795struct netdev_notifier_offload_xstats_info {
2796 struct netdev_notifier_info info; /* must be first */
2797 enum netdev_offload_xstats_type type;
2798
2799 union {
2800 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2801 struct netdev_notifier_offload_xstats_rd *report_delta;
2802 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2803 struct netdev_notifier_offload_xstats_ru *report_used;
2804 };
2805};
2806
2807int netdev_offload_xstats_enable(struct net_device *dev,
2808 enum netdev_offload_xstats_type type,
2809 struct netlink_ext_ack *extack);
2810int netdev_offload_xstats_disable(struct net_device *dev,
2811 enum netdev_offload_xstats_type type);
2812bool netdev_offload_xstats_enabled(const struct net_device *dev,
2813 enum netdev_offload_xstats_type type);
2814int netdev_offload_xstats_get(struct net_device *dev,
2815 enum netdev_offload_xstats_type type,
2816 struct rtnl_hw_stats64 *stats, bool *used,
2817 struct netlink_ext_ack *extack);
2818void
2819netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2820 const struct rtnl_hw_stats64 *stats);
2821void
2822netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2823void netdev_offload_xstats_push_delta(struct net_device *dev,
2824 enum netdev_offload_xstats_type type,
2825 const struct rtnl_hw_stats64 *stats);
2826
2827static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2828 struct net_device *dev)
2829{
2830 info->dev = dev;
2831 info->extack = NULL;
2832}
2833
2834static inline struct net_device *
2835netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2836{
2837 return info->dev;
2838}
2839
2840static inline struct netlink_ext_ack *
2841netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2842{
2843 return info->extack;
2844}
2845
2846int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2847
2848
2849extern rwlock_t dev_base_lock; /* Device list lock */
2850
2851#define for_each_netdev(net, d) \
2852 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2853#define for_each_netdev_reverse(net, d) \
2854 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2855#define for_each_netdev_rcu(net, d) \
2856 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2857#define for_each_netdev_safe(net, d, n) \
2858 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2859#define for_each_netdev_continue(net, d) \
2860 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2861#define for_each_netdev_continue_reverse(net, d) \
2862 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2863 dev_list)
2864#define for_each_netdev_continue_rcu(net, d) \
2865 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2866#define for_each_netdev_in_bond_rcu(bond, slave) \
2867 for_each_netdev_rcu(&init_net, slave) \
2868 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2869#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2870
2871static inline struct net_device *next_net_device(struct net_device *dev)
2872{
2873 struct list_head *lh;
2874 struct net *net;
2875
2876 net = dev_net(dev);
2877 lh = dev->dev_list.next;
2878 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2879}
2880
2881static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2882{
2883 struct list_head *lh;
2884 struct net *net;
2885
2886 net = dev_net(dev);
2887 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2888 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2889}
2890
2891static inline struct net_device *first_net_device(struct net *net)
2892{
2893 return list_empty(&net->dev_base_head) ? NULL :
2894 net_device_entry(net->dev_base_head.next);
2895}
2896
2897static inline struct net_device *first_net_device_rcu(struct net *net)
2898{
2899 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2900
2901 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2902}
2903
2904int netdev_boot_setup_check(struct net_device *dev);
2905struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2906 const char *hwaddr);
2907struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2908void dev_add_pack(struct packet_type *pt);
2909void dev_remove_pack(struct packet_type *pt);
2910void __dev_remove_pack(struct packet_type *pt);
2911void dev_add_offload(struct packet_offload *po);
2912void dev_remove_offload(struct packet_offload *po);
2913
2914int dev_get_iflink(const struct net_device *dev);
2915int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2916int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2917 struct net_device_path_stack *stack);
2918struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2919 unsigned short mask);
2920struct net_device *dev_get_by_name(struct net *net, const char *name);
2921struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2922struct net_device *__dev_get_by_name(struct net *net, const char *name);
2923bool netdev_name_in_use(struct net *net, const char *name);
2924int dev_alloc_name(struct net_device *dev, const char *name);
2925int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2926void dev_close(struct net_device *dev);
2927void dev_close_many(struct list_head *head, bool unlink);
2928void dev_disable_lro(struct net_device *dev);
2929int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2930u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2931 struct net_device *sb_dev);
2932u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2933 struct net_device *sb_dev);
2934
2935int dev_queue_xmit(struct sk_buff *skb);
2936int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2937int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2938
2939static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2940{
2941 int ret;
2942
2943 ret = __dev_direct_xmit(skb, queue_id);
2944 if (!dev_xmit_complete(ret))
2945 kfree_skb(skb);
2946 return ret;
2947}
2948
2949int register_netdevice(struct net_device *dev);
2950void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2951void unregister_netdevice_many(struct list_head *head);
2952static inline void unregister_netdevice(struct net_device *dev)
2953{
2954 unregister_netdevice_queue(dev, NULL);
2955}
2956
2957int netdev_refcnt_read(const struct net_device *dev);
2958void free_netdev(struct net_device *dev);
2959void netdev_freemem(struct net_device *dev);
2960int init_dummy_netdev(struct net_device *dev);
2961
2962struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2963 struct sk_buff *skb,
2964 bool all_slaves);
2965struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2966 struct sock *sk);
2967struct net_device *dev_get_by_index(struct net *net, int ifindex);
2968struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2969struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2970struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2971int netdev_get_name(struct net *net, char *name, int ifindex);
2972int dev_restart(struct net_device *dev);
2973
2974
2975static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2976 unsigned short type,
2977 const void *daddr, const void *saddr,
2978 unsigned int len)
2979{
2980 if (!dev->header_ops || !dev->header_ops->create)
2981 return 0;
2982
2983 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2984}
2985
2986static inline int dev_parse_header(const struct sk_buff *skb,
2987 unsigned char *haddr)
2988{
2989 const struct net_device *dev = skb->dev;
2990
2991 if (!dev->header_ops || !dev->header_ops->parse)
2992 return 0;
2993 return dev->header_ops->parse(skb, haddr);
2994}
2995
2996static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2997{
2998 const struct net_device *dev = skb->dev;
2999
3000 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3001 return 0;
3002 return dev->header_ops->parse_protocol(skb);
3003}
3004
3005/* ll_header must have at least hard_header_len allocated */
3006static inline bool dev_validate_header(const struct net_device *dev,
3007 char *ll_header, int len)
3008{
3009 if (likely(len >= dev->hard_header_len))
3010 return true;
3011 if (len < dev->min_header_len)
3012 return false;
3013
3014 if (capable(CAP_SYS_RAWIO)) {
3015 memset(ll_header + len, 0, dev->hard_header_len - len);
3016 return true;
3017 }
3018
3019 if (dev->header_ops && dev->header_ops->validate)
3020 return dev->header_ops->validate(ll_header, len);
3021
3022 return false;
3023}
3024
3025static inline bool dev_has_header(const struct net_device *dev)
3026{
3027 return dev->header_ops && dev->header_ops->create;
3028}
3029
3030#ifdef CONFIG_NET_FLOW_LIMIT
3031#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3032struct sd_flow_limit {
3033 u64 count;
3034 unsigned int num_buckets;
3035 unsigned int history_head;
3036 u16 history[FLOW_LIMIT_HISTORY];
3037 u8 buckets[];
3038};
3039
3040extern int netdev_flow_limit_table_len;
3041#endif /* CONFIG_NET_FLOW_LIMIT */
3042
3043/*
3044 * Incoming packets are placed on per-CPU queues
3045 */
3046struct softnet_data {
3047 struct list_head poll_list;
3048 struct sk_buff_head process_queue;
3049
3050 /* stats */
3051 unsigned int processed;
3052 unsigned int time_squeeze;
3053 unsigned int received_rps;
3054#ifdef CONFIG_RPS
3055 struct softnet_data *rps_ipi_list;
3056#endif
3057#ifdef CONFIG_NET_FLOW_LIMIT
3058 struct sd_flow_limit __rcu *flow_limit;
3059#endif
3060 struct Qdisc *output_queue;
3061 struct Qdisc **output_queue_tailp;
3062 struct sk_buff *completion_queue;
3063#ifdef CONFIG_XFRM_OFFLOAD
3064 struct sk_buff_head xfrm_backlog;
3065#endif
3066 /* written and read only by owning cpu: */
3067 struct {
3068 u16 recursion;
3069 u8 more;
3070 } xmit;
3071#ifdef CONFIG_RPS
3072 /* input_queue_head should be written by cpu owning this struct,
3073 * and only read by other cpus. Worth using a cache line.
3074 */
3075 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3076
3077 /* Elements below can be accessed between CPUs for RPS/RFS */
3078 call_single_data_t csd ____cacheline_aligned_in_smp;
3079 struct softnet_data *rps_ipi_next;
3080 unsigned int cpu;
3081 unsigned int input_queue_tail;
3082#endif
3083 unsigned int dropped;
3084 struct sk_buff_head input_pkt_queue;
3085 struct napi_struct backlog;
3086
3087};
3088
3089static inline void input_queue_head_incr(struct softnet_data *sd)
3090{
3091#ifdef CONFIG_RPS
3092 sd->input_queue_head++;
3093#endif
3094}
3095
3096static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3097 unsigned int *qtail)
3098{
3099#ifdef CONFIG_RPS
3100 *qtail = ++sd->input_queue_tail;
3101#endif
3102}
3103
3104DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3105
3106static inline int dev_recursion_level(void)
3107{
3108 return this_cpu_read(softnet_data.xmit.recursion);
3109}
3110
3111#define XMIT_RECURSION_LIMIT 8
3112static inline bool dev_xmit_recursion(void)
3113{
3114 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3115 XMIT_RECURSION_LIMIT);
3116}
3117
3118static inline void dev_xmit_recursion_inc(void)
3119{
3120 __this_cpu_inc(softnet_data.xmit.recursion);
3121}
3122
3123static inline void dev_xmit_recursion_dec(void)
3124{
3125 __this_cpu_dec(softnet_data.xmit.recursion);
3126}
3127
3128void __netif_schedule(struct Qdisc *q);
3129void netif_schedule_queue(struct netdev_queue *txq);
3130
3131static inline void netif_tx_schedule_all(struct net_device *dev)
3132{
3133 unsigned int i;
3134
3135 for (i = 0; i < dev->num_tx_queues; i++)
3136 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3137}
3138
3139static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3140{
3141 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3142}
3143
3144/**
3145 * netif_start_queue - allow transmit
3146 * @dev: network device
3147 *
3148 * Allow upper layers to call the device hard_start_xmit routine.
3149 */
3150static inline void netif_start_queue(struct net_device *dev)
3151{
3152 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3153}
3154
3155static inline void netif_tx_start_all_queues(struct net_device *dev)
3156{
3157 unsigned int i;
3158
3159 for (i = 0; i < dev->num_tx_queues; i++) {
3160 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3161 netif_tx_start_queue(txq);
3162 }
3163}
3164
3165void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3166
3167/**
3168 * netif_wake_queue - restart transmit
3169 * @dev: network device
3170 *
3171 * Allow upper layers to call the device hard_start_xmit routine.
3172 * Used for flow control when transmit resources are available.
3173 */
3174static inline void netif_wake_queue(struct net_device *dev)
3175{
3176 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3177}
3178
3179static inline void netif_tx_wake_all_queues(struct net_device *dev)
3180{
3181 unsigned int i;
3182
3183 for (i = 0; i < dev->num_tx_queues; i++) {
3184 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3185 netif_tx_wake_queue(txq);
3186 }
3187}
3188
3189static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3190{
3191 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3192}
3193
3194/**
3195 * netif_stop_queue - stop transmitted packets
3196 * @dev: network device
3197 *
3198 * Stop upper layers calling the device hard_start_xmit routine.
3199 * Used for flow control when transmit resources are unavailable.
3200 */
3201static inline void netif_stop_queue(struct net_device *dev)
3202{
3203 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3204}
3205
3206void netif_tx_stop_all_queues(struct net_device *dev);
3207
3208static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3209{
3210 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3211}
3212
3213/**
3214 * netif_queue_stopped - test if transmit queue is flowblocked
3215 * @dev: network device
3216 *
3217 * Test if transmit queue on device is currently unable to send.
3218 */
3219static inline bool netif_queue_stopped(const struct net_device *dev)
3220{
3221 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3222}
3223
3224static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3225{
3226 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3227}
3228
3229static inline bool
3230netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3231{
3232 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3233}
3234
3235static inline bool
3236netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3237{
3238 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3239}
3240
3241/**
3242 * netdev_queue_set_dql_min_limit - set dql minimum limit
3243 * @dev_queue: pointer to transmit queue
3244 * @min_limit: dql minimum limit
3245 *
3246 * Forces xmit_more() to return true until the minimum threshold
3247 * defined by @min_limit is reached (or until the tx queue is
3248 * empty). Warning: to be use with care, misuse will impact the
3249 * latency.
3250 */
3251static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3252 unsigned int min_limit)
3253{
3254#ifdef CONFIG_BQL
3255 dev_queue->dql.min_limit = min_limit;
3256#endif
3257}
3258
3259/**
3260 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3261 * @dev_queue: pointer to transmit queue
3262 *
3263 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3264 * to give appropriate hint to the CPU.
3265 */
3266static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3267{
3268#ifdef CONFIG_BQL
3269 prefetchw(&dev_queue->dql.num_queued);
3270#endif
3271}
3272
3273/**
3274 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3275 * @dev_queue: pointer to transmit queue
3276 *
3277 * BQL enabled drivers might use this helper in their TX completion path,
3278 * to give appropriate hint to the CPU.
3279 */
3280static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3281{
3282#ifdef CONFIG_BQL
3283 prefetchw(&dev_queue->dql.limit);
3284#endif
3285}
3286
3287static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3288 unsigned int bytes)
3289{
3290#ifdef CONFIG_BQL
3291 dql_queued(&dev_queue->dql, bytes);
3292
3293 if (likely(dql_avail(&dev_queue->dql) >= 0))
3294 return;
3295
3296 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3297
3298 /*
3299 * The XOFF flag must be set before checking the dql_avail below,
3300 * because in netdev_tx_completed_queue we update the dql_completed
3301 * before checking the XOFF flag.
3302 */
3303 smp_mb();
3304
3305 /* check again in case another CPU has just made room avail */
3306 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3307 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3308#endif
3309}
3310
3311/* Variant of netdev_tx_sent_queue() for drivers that are aware
3312 * that they should not test BQL status themselves.
3313 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3314 * skb of a batch.
3315 * Returns true if the doorbell must be used to kick the NIC.
3316 */
3317static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3318 unsigned int bytes,
3319 bool xmit_more)
3320{
3321 if (xmit_more) {
3322#ifdef CONFIG_BQL
3323 dql_queued(&dev_queue->dql, bytes);
3324#endif
3325 return netif_tx_queue_stopped(dev_queue);
3326 }
3327 netdev_tx_sent_queue(dev_queue, bytes);
3328 return true;
3329}
3330
3331/**
3332 * netdev_sent_queue - report the number of bytes queued to hardware
3333 * @dev: network device
3334 * @bytes: number of bytes queued to the hardware device queue
3335 *
3336 * Report the number of bytes queued for sending/completion to the network
3337 * device hardware queue. @bytes should be a good approximation and should
3338 * exactly match netdev_completed_queue() @bytes
3339 */
3340static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3341{
3342 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3343}
3344
3345static inline bool __netdev_sent_queue(struct net_device *dev,
3346 unsigned int bytes,
3347 bool xmit_more)
3348{
3349 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3350 xmit_more);
3351}
3352
3353static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3354 unsigned int pkts, unsigned int bytes)
3355{
3356#ifdef CONFIG_BQL
3357 if (unlikely(!bytes))
3358 return;
3359
3360 dql_completed(&dev_queue->dql, bytes);
3361
3362 /*
3363 * Without the memory barrier there is a small possiblity that
3364 * netdev_tx_sent_queue will miss the update and cause the queue to
3365 * be stopped forever
3366 */
3367 smp_mb();
3368
3369 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3370 return;
3371
3372 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3373 netif_schedule_queue(dev_queue);
3374#endif
3375}
3376
3377/**
3378 * netdev_completed_queue - report bytes and packets completed by device
3379 * @dev: network device
3380 * @pkts: actual number of packets sent over the medium
3381 * @bytes: actual number of bytes sent over the medium
3382 *
3383 * Report the number of bytes and packets transmitted by the network device
3384 * hardware queue over the physical medium, @bytes must exactly match the
3385 * @bytes amount passed to netdev_sent_queue()
3386 */
3387static inline void netdev_completed_queue(struct net_device *dev,
3388 unsigned int pkts, unsigned int bytes)
3389{
3390 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3391}
3392
3393static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3394{
3395#ifdef CONFIG_BQL
3396 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3397 dql_reset(&q->dql);
3398#endif
3399}
3400
3401/**
3402 * netdev_reset_queue - reset the packets and bytes count of a network device
3403 * @dev_queue: network device
3404 *
3405 * Reset the bytes and packet count of a network device and clear the
3406 * software flow control OFF bit for this network device
3407 */
3408static inline void netdev_reset_queue(struct net_device *dev_queue)
3409{
3410 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3411}
3412
3413/**
3414 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3415 * @dev: network device
3416 * @queue_index: given tx queue index
3417 *
3418 * Returns 0 if given tx queue index >= number of device tx queues,
3419 * otherwise returns the originally passed tx queue index.
3420 */
3421static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3422{
3423 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3424 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3425 dev->name, queue_index,
3426 dev->real_num_tx_queues);
3427 return 0;
3428 }
3429
3430 return queue_index;
3431}
3432
3433/**
3434 * netif_running - test if up
3435 * @dev: network device
3436 *
3437 * Test if the device has been brought up.
3438 */
3439static inline bool netif_running(const struct net_device *dev)
3440{
3441 return test_bit(__LINK_STATE_START, &dev->state);
3442}
3443
3444/*
3445 * Routines to manage the subqueues on a device. We only need start,
3446 * stop, and a check if it's stopped. All other device management is
3447 * done at the overall netdevice level.
3448 * Also test the device if we're multiqueue.
3449 */
3450
3451/**
3452 * netif_start_subqueue - allow sending packets on subqueue
3453 * @dev: network device
3454 * @queue_index: sub queue index
3455 *
3456 * Start individual transmit queue of a device with multiple transmit queues.
3457 */
3458static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3459{
3460 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3461
3462 netif_tx_start_queue(txq);
3463}
3464
3465/**
3466 * netif_stop_subqueue - stop sending packets on subqueue
3467 * @dev: network device
3468 * @queue_index: sub queue index
3469 *
3470 * Stop individual transmit queue of a device with multiple transmit queues.
3471 */
3472static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3473{
3474 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3475 netif_tx_stop_queue(txq);
3476}
3477
3478/**
3479 * __netif_subqueue_stopped - test status of subqueue
3480 * @dev: network device
3481 * @queue_index: sub queue index
3482 *
3483 * Check individual transmit queue of a device with multiple transmit queues.
3484 */
3485static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3486 u16 queue_index)
3487{
3488 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3489
3490 return netif_tx_queue_stopped(txq);
3491}
3492
3493/**
3494 * netif_subqueue_stopped - test status of subqueue
3495 * @dev: network device
3496 * @skb: sub queue buffer pointer
3497 *
3498 * Check individual transmit queue of a device with multiple transmit queues.
3499 */
3500static inline bool netif_subqueue_stopped(const struct net_device *dev,
3501 struct sk_buff *skb)
3502{
3503 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3504}
3505
3506/**
3507 * netif_wake_subqueue - allow sending packets on subqueue
3508 * @dev: network device
3509 * @queue_index: sub queue index
3510 *
3511 * Resume individual transmit queue of a device with multiple transmit queues.
3512 */
3513static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3514{
3515 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3516
3517 netif_tx_wake_queue(txq);
3518}
3519
3520#ifdef CONFIG_XPS
3521int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3522 u16 index);
3523int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3524 u16 index, enum xps_map_type type);
3525
3526/**
3527 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3528 * @j: CPU/Rx queue index
3529 * @mask: bitmask of all cpus/rx queues
3530 * @nr_bits: number of bits in the bitmask
3531 *
3532 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3533 */
3534static inline bool netif_attr_test_mask(unsigned long j,
3535 const unsigned long *mask,
3536 unsigned int nr_bits)
3537{
3538 cpu_max_bits_warn(j, nr_bits);
3539 return test_bit(j, mask);
3540}
3541
3542/**
3543 * netif_attr_test_online - Test for online CPU/Rx queue
3544 * @j: CPU/Rx queue index
3545 * @online_mask: bitmask for CPUs/Rx queues that are online
3546 * @nr_bits: number of bits in the bitmask
3547 *
3548 * Returns true if a CPU/Rx queue is online.
3549 */
3550static inline bool netif_attr_test_online(unsigned long j,
3551 const unsigned long *online_mask,
3552 unsigned int nr_bits)
3553{
3554 cpu_max_bits_warn(j, nr_bits);
3555
3556 if (online_mask)
3557 return test_bit(j, online_mask);
3558
3559 return (j < nr_bits);
3560}
3561
3562/**
3563 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3564 * @n: CPU/Rx queue index
3565 * @srcp: the cpumask/Rx queue mask pointer
3566 * @nr_bits: number of bits in the bitmask
3567 *
3568 * Returns >= nr_bits if no further CPUs/Rx queues set.
3569 */
3570static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3571 unsigned int nr_bits)
3572{
3573 /* -1 is a legal arg here. */
3574 if (n != -1)
3575 cpu_max_bits_warn(n, nr_bits);
3576
3577 if (srcp)
3578 return find_next_bit(srcp, nr_bits, n + 1);
3579
3580 return n + 1;
3581}
3582
3583/**
3584 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3585 * @n: CPU/Rx queue index
3586 * @src1p: the first CPUs/Rx queues mask pointer
3587 * @src2p: the second CPUs/Rx queues mask pointer
3588 * @nr_bits: number of bits in the bitmask
3589 *
3590 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3591 */
3592static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3593 const unsigned long *src2p,
3594 unsigned int nr_bits)
3595{
3596 /* -1 is a legal arg here. */
3597 if (n != -1)
3598 cpu_max_bits_warn(n, nr_bits);
3599
3600 if (src1p && src2p)
3601 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3602 else if (src1p)
3603 return find_next_bit(src1p, nr_bits, n + 1);
3604 else if (src2p)
3605 return find_next_bit(src2p, nr_bits, n + 1);
3606
3607 return n + 1;
3608}
3609#else
3610static inline int netif_set_xps_queue(struct net_device *dev,
3611 const struct cpumask *mask,
3612 u16 index)
3613{
3614 return 0;
3615}
3616
3617static inline int __netif_set_xps_queue(struct net_device *dev,
3618 const unsigned long *mask,
3619 u16 index, enum xps_map_type type)
3620{
3621 return 0;
3622}
3623#endif
3624
3625/**
3626 * netif_is_multiqueue - test if device has multiple transmit queues
3627 * @dev: network device
3628 *
3629 * Check if device has multiple transmit queues
3630 */
3631static inline bool netif_is_multiqueue(const struct net_device *dev)
3632{
3633 return dev->num_tx_queues > 1;
3634}
3635
3636int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3637
3638#ifdef CONFIG_SYSFS
3639int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3640#else
3641static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3642 unsigned int rxqs)
3643{
3644 dev->real_num_rx_queues = rxqs;
3645 return 0;
3646}
3647#endif
3648int netif_set_real_num_queues(struct net_device *dev,
3649 unsigned int txq, unsigned int rxq);
3650
3651static inline struct netdev_rx_queue *
3652__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3653{
3654 return dev->_rx + rxq;
3655}
3656
3657#ifdef CONFIG_SYSFS
3658static inline unsigned int get_netdev_rx_queue_index(
3659 struct netdev_rx_queue *queue)
3660{
3661 struct net_device *dev = queue->dev;
3662 int index = queue - dev->_rx;
3663
3664 BUG_ON(index >= dev->num_rx_queues);
3665 return index;
3666}
3667#endif
3668
3669int netif_get_num_default_rss_queues(void);
3670
3671enum skb_free_reason {
3672 SKB_REASON_CONSUMED,
3673 SKB_REASON_DROPPED,
3674};
3675
3676void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3677void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3678
3679/*
3680 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3681 * interrupt context or with hardware interrupts being disabled.
3682 * (in_hardirq() || irqs_disabled())
3683 *
3684 * We provide four helpers that can be used in following contexts :
3685 *
3686 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3687 * replacing kfree_skb(skb)
3688 *
3689 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3690 * Typically used in place of consume_skb(skb) in TX completion path
3691 *
3692 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3693 * replacing kfree_skb(skb)
3694 *
3695 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3696 * and consumed a packet. Used in place of consume_skb(skb)
3697 */
3698static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3699{
3700 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3701}
3702
3703static inline void dev_consume_skb_irq(struct sk_buff *skb)
3704{
3705 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3706}
3707
3708static inline void dev_kfree_skb_any(struct sk_buff *skb)
3709{
3710 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3711}
3712
3713static inline void dev_consume_skb_any(struct sk_buff *skb)
3714{
3715 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3716}
3717
3718u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3719 struct bpf_prog *xdp_prog);
3720void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3721int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3722int netif_rx(struct sk_buff *skb);
3723int __netif_rx(struct sk_buff *skb);
3724
3725int netif_receive_skb(struct sk_buff *skb);
3726int netif_receive_skb_core(struct sk_buff *skb);
3727void netif_receive_skb_list_internal(struct list_head *head);
3728void netif_receive_skb_list(struct list_head *head);
3729gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3730void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3731struct sk_buff *napi_get_frags(struct napi_struct *napi);
3732gro_result_t napi_gro_frags(struct napi_struct *napi);
3733struct packet_offload *gro_find_receive_by_type(__be16 type);
3734struct packet_offload *gro_find_complete_by_type(__be16 type);
3735
3736static inline void napi_free_frags(struct napi_struct *napi)
3737{
3738 kfree_skb(napi->skb);
3739 napi->skb = NULL;
3740}
3741
3742bool netdev_is_rx_handler_busy(struct net_device *dev);
3743int netdev_rx_handler_register(struct net_device *dev,
3744 rx_handler_func_t *rx_handler,
3745 void *rx_handler_data);
3746void netdev_rx_handler_unregister(struct net_device *dev);
3747
3748bool dev_valid_name(const char *name);
3749static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3750{
3751 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3752}
3753int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3754int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3755int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3756 void __user *data, bool *need_copyout);
3757int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3758int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3759unsigned int dev_get_flags(const struct net_device *);
3760int __dev_change_flags(struct net_device *dev, unsigned int flags,
3761 struct netlink_ext_ack *extack);
3762int dev_change_flags(struct net_device *dev, unsigned int flags,
3763 struct netlink_ext_ack *extack);
3764void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3765 unsigned int gchanges);
3766int dev_change_name(struct net_device *, const char *);
3767int dev_set_alias(struct net_device *, const char *, size_t);
3768int dev_get_alias(const struct net_device *, char *, size_t);
3769int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3770 const char *pat, int new_ifindex);
3771static inline
3772int dev_change_net_namespace(struct net_device *dev, struct net *net,
3773 const char *pat)
3774{
3775 return __dev_change_net_namespace(dev, net, pat, 0);
3776}
3777int __dev_set_mtu(struct net_device *, int);
3778int dev_validate_mtu(struct net_device *dev, int mtu,
3779 struct netlink_ext_ack *extack);
3780int dev_set_mtu_ext(struct net_device *dev, int mtu,
3781 struct netlink_ext_ack *extack);
3782int dev_set_mtu(struct net_device *, int);
3783int dev_change_tx_queue_len(struct net_device *, unsigned long);
3784void dev_set_group(struct net_device *, int);
3785int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3786 struct netlink_ext_ack *extack);
3787int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3788 struct netlink_ext_ack *extack);
3789int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3790 struct netlink_ext_ack *extack);
3791int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3792int dev_change_carrier(struct net_device *, bool new_carrier);
3793int dev_get_phys_port_id(struct net_device *dev,
3794 struct netdev_phys_item_id *ppid);
3795int dev_get_phys_port_name(struct net_device *dev,
3796 char *name, size_t len);
3797int dev_get_port_parent_id(struct net_device *dev,
3798 struct netdev_phys_item_id *ppid, bool recurse);
3799bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3800int dev_change_proto_down(struct net_device *dev, bool proto_down);
3801void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3802 u32 value);
3803struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3804struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3805 struct netdev_queue *txq, int *ret);
3806
3807typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3808int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3809 int fd, int expected_fd, u32 flags);
3810int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3811u8 dev_xdp_prog_count(struct net_device *dev);
3812u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3813
3814int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3815int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3816int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3817bool is_skb_forwardable(const struct net_device *dev,
3818 const struct sk_buff *skb);
3819
3820static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3821 const struct sk_buff *skb,
3822 const bool check_mtu)
3823{
3824 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3825 unsigned int len;
3826
3827 if (!(dev->flags & IFF_UP))
3828 return false;
3829
3830 if (!check_mtu)
3831 return true;
3832
3833 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3834 if (skb->len <= len)
3835 return true;
3836
3837 /* if TSO is enabled, we don't care about the length as the packet
3838 * could be forwarded without being segmented before
3839 */
3840 if (skb_is_gso(skb))
3841 return true;
3842
3843 return false;
3844}
3845
3846struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3847
3848static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3849{
3850 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3851 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3852
3853 if (likely(p))
3854 return p;
3855
3856 return netdev_core_stats_alloc(dev);
3857}
3858
3859#define DEV_CORE_STATS_INC(FIELD) \
3860static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3861{ \
3862 struct net_device_core_stats __percpu *p; \
3863 \
3864 p = dev_core_stats(dev); \
3865 if (p) \
3866 this_cpu_inc(p->FIELD); \
3867}
3868DEV_CORE_STATS_INC(rx_dropped)
3869DEV_CORE_STATS_INC(tx_dropped)
3870DEV_CORE_STATS_INC(rx_nohandler)
3871
3872static __always_inline int ____dev_forward_skb(struct net_device *dev,
3873 struct sk_buff *skb,
3874 const bool check_mtu)
3875{
3876 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3877 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3878 dev_core_stats_rx_dropped_inc(dev);
3879 kfree_skb(skb);
3880 return NET_RX_DROP;
3881 }
3882
3883 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3884 skb->priority = 0;
3885 return 0;
3886}
3887
3888bool dev_nit_active(struct net_device *dev);
3889void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3890
3891extern int netdev_budget;
3892extern unsigned int netdev_budget_usecs;
3893
3894/* Called by rtnetlink.c:rtnl_unlock() */
3895void netdev_run_todo(void);
3896
3897static inline void __dev_put(struct net_device *dev)
3898{
3899 if (dev) {
3900#ifdef CONFIG_PCPU_DEV_REFCNT
3901 this_cpu_dec(*dev->pcpu_refcnt);
3902#else
3903 refcount_dec(&dev->dev_refcnt);
3904#endif
3905 }
3906}
3907
3908static inline void __dev_hold(struct net_device *dev)
3909{
3910 if (dev) {
3911#ifdef CONFIG_PCPU_DEV_REFCNT
3912 this_cpu_inc(*dev->pcpu_refcnt);
3913#else
3914 refcount_inc(&dev->dev_refcnt);
3915#endif
3916 }
3917}
3918
3919static inline void __netdev_tracker_alloc(struct net_device *dev,
3920 netdevice_tracker *tracker,
3921 gfp_t gfp)
3922{
3923#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3924 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
3925#endif
3926}
3927
3928/* netdev_tracker_alloc() can upgrade a prior untracked reference
3929 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
3930 */
3931static inline void netdev_tracker_alloc(struct net_device *dev,
3932 netdevice_tracker *tracker, gfp_t gfp)
3933{
3934#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3935 refcount_dec(&dev->refcnt_tracker.no_tracker);
3936 __netdev_tracker_alloc(dev, tracker, gfp);
3937#endif
3938}
3939
3940static inline void netdev_tracker_free(struct net_device *dev,
3941 netdevice_tracker *tracker)
3942{
3943#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3944 ref_tracker_free(&dev->refcnt_tracker, tracker);
3945#endif
3946}
3947
3948static inline void dev_hold_track(struct net_device *dev,
3949 netdevice_tracker *tracker, gfp_t gfp)
3950{
3951 if (dev) {
3952 __dev_hold(dev);
3953 __netdev_tracker_alloc(dev, tracker, gfp);
3954 }
3955}
3956
3957static inline void dev_put_track(struct net_device *dev,
3958 netdevice_tracker *tracker)
3959{
3960 if (dev) {
3961 netdev_tracker_free(dev, tracker);
3962 __dev_put(dev);
3963 }
3964}
3965
3966/**
3967 * dev_hold - get reference to device
3968 * @dev: network device
3969 *
3970 * Hold reference to device to keep it from being freed.
3971 * Try using dev_hold_track() instead.
3972 */
3973static inline void dev_hold(struct net_device *dev)
3974{
3975 dev_hold_track(dev, NULL, GFP_ATOMIC);
3976}
3977
3978/**
3979 * dev_put - release reference to device
3980 * @dev: network device
3981 *
3982 * Release reference to device to allow it to be freed.
3983 * Try using dev_put_track() instead.
3984 */
3985static inline void dev_put(struct net_device *dev)
3986{
3987 dev_put_track(dev, NULL);
3988}
3989
3990static inline void dev_replace_track(struct net_device *odev,
3991 struct net_device *ndev,
3992 netdevice_tracker *tracker,
3993 gfp_t gfp)
3994{
3995 if (odev)
3996 netdev_tracker_free(odev, tracker);
3997
3998 __dev_hold(ndev);
3999 __dev_put(odev);
4000
4001 if (ndev)
4002 __netdev_tracker_alloc(ndev, tracker, gfp);
4003}
4004
4005/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4006 * and _off may be called from IRQ context, but it is caller
4007 * who is responsible for serialization of these calls.
4008 *
4009 * The name carrier is inappropriate, these functions should really be
4010 * called netif_lowerlayer_*() because they represent the state of any
4011 * kind of lower layer not just hardware media.
4012 */
4013
4014void linkwatch_init_dev(struct net_device *dev);
4015void linkwatch_fire_event(struct net_device *dev);
4016void linkwatch_forget_dev(struct net_device *dev);
4017
4018/**
4019 * netif_carrier_ok - test if carrier present
4020 * @dev: network device
4021 *
4022 * Check if carrier is present on device
4023 */
4024static inline bool netif_carrier_ok(const struct net_device *dev)
4025{
4026 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4027}
4028
4029unsigned long dev_trans_start(struct net_device *dev);
4030
4031void __netdev_watchdog_up(struct net_device *dev);
4032
4033void netif_carrier_on(struct net_device *dev);
4034void netif_carrier_off(struct net_device *dev);
4035void netif_carrier_event(struct net_device *dev);
4036
4037/**
4038 * netif_dormant_on - mark device as dormant.
4039 * @dev: network device
4040 *
4041 * Mark device as dormant (as per RFC2863).
4042 *
4043 * The dormant state indicates that the relevant interface is not
4044 * actually in a condition to pass packets (i.e., it is not 'up') but is
4045 * in a "pending" state, waiting for some external event. For "on-
4046 * demand" interfaces, this new state identifies the situation where the
4047 * interface is waiting for events to place it in the up state.
4048 */
4049static inline void netif_dormant_on(struct net_device *dev)
4050{
4051 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4052 linkwatch_fire_event(dev);
4053}
4054
4055/**
4056 * netif_dormant_off - set device as not dormant.
4057 * @dev: network device
4058 *
4059 * Device is not in dormant state.
4060 */
4061static inline void netif_dormant_off(struct net_device *dev)
4062{
4063 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4064 linkwatch_fire_event(dev);
4065}
4066
4067/**
4068 * netif_dormant - test if device is dormant
4069 * @dev: network device
4070 *
4071 * Check if device is dormant.
4072 */
4073static inline bool netif_dormant(const struct net_device *dev)
4074{
4075 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4076}
4077
4078
4079/**
4080 * netif_testing_on - mark device as under test.
4081 * @dev: network device
4082 *
4083 * Mark device as under test (as per RFC2863).
4084 *
4085 * The testing state indicates that some test(s) must be performed on
4086 * the interface. After completion, of the test, the interface state
4087 * will change to up, dormant, or down, as appropriate.
4088 */
4089static inline void netif_testing_on(struct net_device *dev)
4090{
4091 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4092 linkwatch_fire_event(dev);
4093}
4094
4095/**
4096 * netif_testing_off - set device as not under test.
4097 * @dev: network device
4098 *
4099 * Device is not in testing state.
4100 */
4101static inline void netif_testing_off(struct net_device *dev)
4102{
4103 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4104 linkwatch_fire_event(dev);
4105}
4106
4107/**
4108 * netif_testing - test if device is under test
4109 * @dev: network device
4110 *
4111 * Check if device is under test
4112 */
4113static inline bool netif_testing(const struct net_device *dev)
4114{
4115 return test_bit(__LINK_STATE_TESTING, &dev->state);
4116}
4117
4118
4119/**
4120 * netif_oper_up - test if device is operational
4121 * @dev: network device
4122 *
4123 * Check if carrier is operational
4124 */
4125static inline bool netif_oper_up(const struct net_device *dev)
4126{
4127 return (dev->operstate == IF_OPER_UP ||
4128 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4129}
4130
4131/**
4132 * netif_device_present - is device available or removed
4133 * @dev: network device
4134 *
4135 * Check if device has not been removed from system.
4136 */
4137static inline bool netif_device_present(const struct net_device *dev)
4138{
4139 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4140}
4141
4142void netif_device_detach(struct net_device *dev);
4143
4144void netif_device_attach(struct net_device *dev);
4145
4146/*
4147 * Network interface message level settings
4148 */
4149
4150enum {
4151 NETIF_MSG_DRV_BIT,
4152 NETIF_MSG_PROBE_BIT,
4153 NETIF_MSG_LINK_BIT,
4154 NETIF_MSG_TIMER_BIT,
4155 NETIF_MSG_IFDOWN_BIT,
4156 NETIF_MSG_IFUP_BIT,
4157 NETIF_MSG_RX_ERR_BIT,
4158 NETIF_MSG_TX_ERR_BIT,
4159 NETIF_MSG_TX_QUEUED_BIT,
4160 NETIF_MSG_INTR_BIT,
4161 NETIF_MSG_TX_DONE_BIT,
4162 NETIF_MSG_RX_STATUS_BIT,
4163 NETIF_MSG_PKTDATA_BIT,
4164 NETIF_MSG_HW_BIT,
4165 NETIF_MSG_WOL_BIT,
4166
4167 /* When you add a new bit above, update netif_msg_class_names array
4168 * in net/ethtool/common.c
4169 */
4170 NETIF_MSG_CLASS_COUNT,
4171};
4172/* Both ethtool_ops interface and internal driver implementation use u32 */
4173static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4174
4175#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4176#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4177
4178#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4179#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4180#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4181#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4182#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4183#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4184#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4185#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4186#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4187#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4188#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4189#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4190#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4191#define NETIF_MSG_HW __NETIF_MSG(HW)
4192#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4193
4194#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4195#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4196#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4197#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4198#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4199#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4200#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4201#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4202#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4203#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4204#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4205#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4206#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4207#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4208#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4209
4210static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4211{
4212 /* use default */
4213 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4214 return default_msg_enable_bits;
4215 if (debug_value == 0) /* no output */
4216 return 0;
4217 /* set low N bits */
4218 return (1U << debug_value) - 1;
4219}
4220
4221static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4222{
4223 spin_lock(&txq->_xmit_lock);
4224 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4225 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4226}
4227
4228static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4229{
4230 __acquire(&txq->_xmit_lock);
4231 return true;
4232}
4233
4234static inline void __netif_tx_release(struct netdev_queue *txq)
4235{
4236 __release(&txq->_xmit_lock);
4237}
4238
4239static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4240{
4241 spin_lock_bh(&txq->_xmit_lock);
4242 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4243 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4244}
4245
4246static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4247{
4248 bool ok = spin_trylock(&txq->_xmit_lock);
4249
4250 if (likely(ok)) {
4251 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4252 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4253 }
4254 return ok;
4255}
4256
4257static inline void __netif_tx_unlock(struct netdev_queue *txq)
4258{
4259 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4260 WRITE_ONCE(txq->xmit_lock_owner, -1);
4261 spin_unlock(&txq->_xmit_lock);
4262}
4263
4264static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4265{
4266 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4267 WRITE_ONCE(txq->xmit_lock_owner, -1);
4268 spin_unlock_bh(&txq->_xmit_lock);
4269}
4270
4271/*
4272 * txq->trans_start can be read locklessly from dev_watchdog()
4273 */
4274static inline void txq_trans_update(struct netdev_queue *txq)
4275{
4276 if (txq->xmit_lock_owner != -1)
4277 WRITE_ONCE(txq->trans_start, jiffies);
4278}
4279
4280static inline void txq_trans_cond_update(struct netdev_queue *txq)
4281{
4282 unsigned long now = jiffies;
4283
4284 if (READ_ONCE(txq->trans_start) != now)
4285 WRITE_ONCE(txq->trans_start, now);
4286}
4287
4288/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4289static inline void netif_trans_update(struct net_device *dev)
4290{
4291 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4292
4293 txq_trans_cond_update(txq);
4294}
4295
4296/**
4297 * netif_tx_lock - grab network device transmit lock
4298 * @dev: network device
4299 *
4300 * Get network device transmit lock
4301 */
4302void netif_tx_lock(struct net_device *dev);
4303
4304static inline void netif_tx_lock_bh(struct net_device *dev)
4305{
4306 local_bh_disable();
4307 netif_tx_lock(dev);
4308}
4309
4310void netif_tx_unlock(struct net_device *dev);
4311
4312static inline void netif_tx_unlock_bh(struct net_device *dev)
4313{
4314 netif_tx_unlock(dev);
4315 local_bh_enable();
4316}
4317
4318#define HARD_TX_LOCK(dev, txq, cpu) { \
4319 if ((dev->features & NETIF_F_LLTX) == 0) { \
4320 __netif_tx_lock(txq, cpu); \
4321 } else { \
4322 __netif_tx_acquire(txq); \
4323 } \
4324}
4325
4326#define HARD_TX_TRYLOCK(dev, txq) \
4327 (((dev->features & NETIF_F_LLTX) == 0) ? \
4328 __netif_tx_trylock(txq) : \
4329 __netif_tx_acquire(txq))
4330
4331#define HARD_TX_UNLOCK(dev, txq) { \
4332 if ((dev->features & NETIF_F_LLTX) == 0) { \
4333 __netif_tx_unlock(txq); \
4334 } else { \
4335 __netif_tx_release(txq); \
4336 } \
4337}
4338
4339static inline void netif_tx_disable(struct net_device *dev)
4340{
4341 unsigned int i;
4342 int cpu;
4343
4344 local_bh_disable();
4345 cpu = smp_processor_id();
4346 spin_lock(&dev->tx_global_lock);
4347 for (i = 0; i < dev->num_tx_queues; i++) {
4348 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4349
4350 __netif_tx_lock(txq, cpu);
4351 netif_tx_stop_queue(txq);
4352 __netif_tx_unlock(txq);
4353 }
4354 spin_unlock(&dev->tx_global_lock);
4355 local_bh_enable();
4356}
4357
4358static inline void netif_addr_lock(struct net_device *dev)
4359{
4360 unsigned char nest_level = 0;
4361
4362#ifdef CONFIG_LOCKDEP
4363 nest_level = dev->nested_level;
4364#endif
4365 spin_lock_nested(&dev->addr_list_lock, nest_level);
4366}
4367
4368static inline void netif_addr_lock_bh(struct net_device *dev)
4369{
4370 unsigned char nest_level = 0;
4371
4372#ifdef CONFIG_LOCKDEP
4373 nest_level = dev->nested_level;
4374#endif
4375 local_bh_disable();
4376 spin_lock_nested(&dev->addr_list_lock, nest_level);
4377}
4378
4379static inline void netif_addr_unlock(struct net_device *dev)
4380{
4381 spin_unlock(&dev->addr_list_lock);
4382}
4383
4384static inline void netif_addr_unlock_bh(struct net_device *dev)
4385{
4386 spin_unlock_bh(&dev->addr_list_lock);
4387}
4388
4389/*
4390 * dev_addrs walker. Should be used only for read access. Call with
4391 * rcu_read_lock held.
4392 */
4393#define for_each_dev_addr(dev, ha) \
4394 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4395
4396/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4397
4398void ether_setup(struct net_device *dev);
4399
4400/* Support for loadable net-drivers */
4401struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4402 unsigned char name_assign_type,
4403 void (*setup)(struct net_device *),
4404 unsigned int txqs, unsigned int rxqs);
4405#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4406 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4407
4408#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4409 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4410 count)
4411
4412int register_netdev(struct net_device *dev);
4413void unregister_netdev(struct net_device *dev);
4414
4415int devm_register_netdev(struct device *dev, struct net_device *ndev);
4416
4417/* General hardware address lists handling functions */
4418int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4419 struct netdev_hw_addr_list *from_list, int addr_len);
4420void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4421 struct netdev_hw_addr_list *from_list, int addr_len);
4422int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4423 struct net_device *dev,
4424 int (*sync)(struct net_device *, const unsigned char *),
4425 int (*unsync)(struct net_device *,
4426 const unsigned char *));
4427int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4428 struct net_device *dev,
4429 int (*sync)(struct net_device *,
4430 const unsigned char *, int),
4431 int (*unsync)(struct net_device *,
4432 const unsigned char *, int));
4433void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4434 struct net_device *dev,
4435 int (*unsync)(struct net_device *,
4436 const unsigned char *, int));
4437void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4438 struct net_device *dev,
4439 int (*unsync)(struct net_device *,
4440 const unsigned char *));
4441void __hw_addr_init(struct netdev_hw_addr_list *list);
4442
4443/* Functions used for device addresses handling */
4444void dev_addr_mod(struct net_device *dev, unsigned int offset,
4445 const void *addr, size_t len);
4446
4447static inline void
4448__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4449{
4450 dev_addr_mod(dev, 0, addr, len);
4451}
4452
4453static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4454{
4455 __dev_addr_set(dev, addr, dev->addr_len);
4456}
4457
4458int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4459 unsigned char addr_type);
4460int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4461 unsigned char addr_type);
4462void dev_addr_flush(struct net_device *dev);
4463int dev_addr_init(struct net_device *dev);
4464void dev_addr_check(struct net_device *dev);
4465
4466/* Functions used for unicast addresses handling */
4467int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4468int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4469int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4470int dev_uc_sync(struct net_device *to, struct net_device *from);
4471int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4472void dev_uc_unsync(struct net_device *to, struct net_device *from);
4473void dev_uc_flush(struct net_device *dev);
4474void dev_uc_init(struct net_device *dev);
4475
4476/**
4477 * __dev_uc_sync - Synchonize device's unicast list
4478 * @dev: device to sync
4479 * @sync: function to call if address should be added
4480 * @unsync: function to call if address should be removed
4481 *
4482 * Add newly added addresses to the interface, and release
4483 * addresses that have been deleted.
4484 */
4485static inline int __dev_uc_sync(struct net_device *dev,
4486 int (*sync)(struct net_device *,
4487 const unsigned char *),
4488 int (*unsync)(struct net_device *,
4489 const unsigned char *))
4490{
4491 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4492}
4493
4494/**
4495 * __dev_uc_unsync - Remove synchronized addresses from device
4496 * @dev: device to sync
4497 * @unsync: function to call if address should be removed
4498 *
4499 * Remove all addresses that were added to the device by dev_uc_sync().
4500 */
4501static inline void __dev_uc_unsync(struct net_device *dev,
4502 int (*unsync)(struct net_device *,
4503 const unsigned char *))
4504{
4505 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4506}
4507
4508/* Functions used for multicast addresses handling */
4509int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4510int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4511int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4512int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4513int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4514int dev_mc_sync(struct net_device *to, struct net_device *from);
4515int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4516void dev_mc_unsync(struct net_device *to, struct net_device *from);
4517void dev_mc_flush(struct net_device *dev);
4518void dev_mc_init(struct net_device *dev);
4519
4520/**
4521 * __dev_mc_sync - Synchonize device's multicast list
4522 * @dev: device to sync
4523 * @sync: function to call if address should be added
4524 * @unsync: function to call if address should be removed
4525 *
4526 * Add newly added addresses to the interface, and release
4527 * addresses that have been deleted.
4528 */
4529static inline int __dev_mc_sync(struct net_device *dev,
4530 int (*sync)(struct net_device *,
4531 const unsigned char *),
4532 int (*unsync)(struct net_device *,
4533 const unsigned char *))
4534{
4535 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4536}
4537
4538/**
4539 * __dev_mc_unsync - Remove synchronized addresses from device
4540 * @dev: device to sync
4541 * @unsync: function to call if address should be removed
4542 *
4543 * Remove all addresses that were added to the device by dev_mc_sync().
4544 */
4545static inline void __dev_mc_unsync(struct net_device *dev,
4546 int (*unsync)(struct net_device *,
4547 const unsigned char *))
4548{
4549 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4550}
4551
4552/* Functions used for secondary unicast and multicast support */
4553void dev_set_rx_mode(struct net_device *dev);
4554void __dev_set_rx_mode(struct net_device *dev);
4555int dev_set_promiscuity(struct net_device *dev, int inc);
4556int dev_set_allmulti(struct net_device *dev, int inc);
4557void netdev_state_change(struct net_device *dev);
4558void __netdev_notify_peers(struct net_device *dev);
4559void netdev_notify_peers(struct net_device *dev);
4560void netdev_features_change(struct net_device *dev);
4561/* Load a device via the kmod */
4562void dev_load(struct net *net, const char *name);
4563struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4564 struct rtnl_link_stats64 *storage);
4565void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4566 const struct net_device_stats *netdev_stats);
4567void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4568 const struct pcpu_sw_netstats __percpu *netstats);
4569void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4570
4571extern int netdev_max_backlog;
4572extern int netdev_tstamp_prequeue;
4573extern int netdev_unregister_timeout_secs;
4574extern int weight_p;
4575extern int dev_weight_rx_bias;
4576extern int dev_weight_tx_bias;
4577extern int dev_rx_weight;
4578extern int dev_tx_weight;
4579extern int gro_normal_batch;
4580
4581enum {
4582 NESTED_SYNC_IMM_BIT,
4583 NESTED_SYNC_TODO_BIT,
4584};
4585
4586#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4587#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4588
4589#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4590#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4591
4592struct netdev_nested_priv {
4593 unsigned char flags;
4594 void *data;
4595};
4596
4597bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4598struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4599 struct list_head **iter);
4600
4601/* iterate through upper list, must be called under RCU read lock */
4602#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4603 for (iter = &(dev)->adj_list.upper, \
4604 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4605 updev; \
4606 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4607
4608int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4609 int (*fn)(struct net_device *upper_dev,
4610 struct netdev_nested_priv *priv),
4611 struct netdev_nested_priv *priv);
4612
4613bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4614 struct net_device *upper_dev);
4615
4616bool netdev_has_any_upper_dev(struct net_device *dev);
4617
4618void *netdev_lower_get_next_private(struct net_device *dev,
4619 struct list_head **iter);
4620void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4621 struct list_head **iter);
4622
4623#define netdev_for_each_lower_private(dev, priv, iter) \
4624 for (iter = (dev)->adj_list.lower.next, \
4625 priv = netdev_lower_get_next_private(dev, &(iter)); \
4626 priv; \
4627 priv = netdev_lower_get_next_private(dev, &(iter)))
4628
4629#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4630 for (iter = &(dev)->adj_list.lower, \
4631 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4632 priv; \
4633 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4634
4635void *netdev_lower_get_next(struct net_device *dev,
4636 struct list_head **iter);
4637
4638#define netdev_for_each_lower_dev(dev, ldev, iter) \
4639 for (iter = (dev)->adj_list.lower.next, \
4640 ldev = netdev_lower_get_next(dev, &(iter)); \
4641 ldev; \
4642 ldev = netdev_lower_get_next(dev, &(iter)))
4643
4644struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4645 struct list_head **iter);
4646int netdev_walk_all_lower_dev(struct net_device *dev,
4647 int (*fn)(struct net_device *lower_dev,
4648 struct netdev_nested_priv *priv),
4649 struct netdev_nested_priv *priv);
4650int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4651 int (*fn)(struct net_device *lower_dev,
4652 struct netdev_nested_priv *priv),
4653 struct netdev_nested_priv *priv);
4654
4655void *netdev_adjacent_get_private(struct list_head *adj_list);
4656void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4657struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4658struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4659int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4660 struct netlink_ext_ack *extack);
4661int netdev_master_upper_dev_link(struct net_device *dev,
4662 struct net_device *upper_dev,
4663 void *upper_priv, void *upper_info,
4664 struct netlink_ext_ack *extack);
4665void netdev_upper_dev_unlink(struct net_device *dev,
4666 struct net_device *upper_dev);
4667int netdev_adjacent_change_prepare(struct net_device *old_dev,
4668 struct net_device *new_dev,
4669 struct net_device *dev,
4670 struct netlink_ext_ack *extack);
4671void netdev_adjacent_change_commit(struct net_device *old_dev,
4672 struct net_device *new_dev,
4673 struct net_device *dev);
4674void netdev_adjacent_change_abort(struct net_device *old_dev,
4675 struct net_device *new_dev,
4676 struct net_device *dev);
4677void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4678void *netdev_lower_dev_get_private(struct net_device *dev,
4679 struct net_device *lower_dev);
4680void netdev_lower_state_changed(struct net_device *lower_dev,
4681 void *lower_state_info);
4682
4683/* RSS keys are 40 or 52 bytes long */
4684#define NETDEV_RSS_KEY_LEN 52
4685extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4686void netdev_rss_key_fill(void *buffer, size_t len);
4687
4688int skb_checksum_help(struct sk_buff *skb);
4689int skb_crc32c_csum_help(struct sk_buff *skb);
4690int skb_csum_hwoffload_help(struct sk_buff *skb,
4691 const netdev_features_t features);
4692
4693struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4694 netdev_features_t features, bool tx_path);
4695struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4696 netdev_features_t features, __be16 type);
4697struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4698 netdev_features_t features);
4699
4700struct netdev_bonding_info {
4701 ifslave slave;
4702 ifbond master;
4703};
4704
4705struct netdev_notifier_bonding_info {
4706 struct netdev_notifier_info info; /* must be first */
4707 struct netdev_bonding_info bonding_info;
4708};
4709
4710void netdev_bonding_info_change(struct net_device *dev,
4711 struct netdev_bonding_info *bonding_info);
4712
4713#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4714void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4715#else
4716static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4717 const void *data)
4718{
4719}
4720#endif
4721
4722static inline
4723struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4724{
4725 return __skb_gso_segment(skb, features, true);
4726}
4727__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4728
4729static inline bool can_checksum_protocol(netdev_features_t features,
4730 __be16 protocol)
4731{
4732 if (protocol == htons(ETH_P_FCOE))
4733 return !!(features & NETIF_F_FCOE_CRC);
4734
4735 /* Assume this is an IP checksum (not SCTP CRC) */
4736
4737 if (features & NETIF_F_HW_CSUM) {
4738 /* Can checksum everything */
4739 return true;
4740 }
4741
4742 switch (protocol) {
4743 case htons(ETH_P_IP):
4744 return !!(features & NETIF_F_IP_CSUM);
4745 case htons(ETH_P_IPV6):
4746 return !!(features & NETIF_F_IPV6_CSUM);
4747 default:
4748 return false;
4749 }
4750}
4751
4752#ifdef CONFIG_BUG
4753void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4754#else
4755static inline void netdev_rx_csum_fault(struct net_device *dev,
4756 struct sk_buff *skb)
4757{
4758}
4759#endif
4760/* rx skb timestamps */
4761void net_enable_timestamp(void);
4762void net_disable_timestamp(void);
4763
4764#ifdef CONFIG_PROC_FS
4765int __init dev_proc_init(void);
4766#else
4767#define dev_proc_init() 0
4768#endif
4769
4770static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4771 struct sk_buff *skb, struct net_device *dev,
4772 bool more)
4773{
4774 __this_cpu_write(softnet_data.xmit.more, more);
4775 return ops->ndo_start_xmit(skb, dev);
4776}
4777
4778static inline bool netdev_xmit_more(void)
4779{
4780 return __this_cpu_read(softnet_data.xmit.more);
4781}
4782
4783static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4784 struct netdev_queue *txq, bool more)
4785{
4786 const struct net_device_ops *ops = dev->netdev_ops;
4787 netdev_tx_t rc;
4788
4789 rc = __netdev_start_xmit(ops, skb, dev, more);
4790 if (rc == NETDEV_TX_OK)
4791 txq_trans_update(txq);
4792
4793 return rc;
4794}
4795
4796int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4797 const void *ns);
4798void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4799 const void *ns);
4800
4801extern const struct kobj_ns_type_operations net_ns_type_operations;
4802
4803const char *netdev_drivername(const struct net_device *dev);
4804
4805void linkwatch_run_queue(void);
4806
4807static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4808 netdev_features_t f2)
4809{
4810 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4811 if (f1 & NETIF_F_HW_CSUM)
4812 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4813 else
4814 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4815 }
4816
4817 return f1 & f2;
4818}
4819
4820static inline netdev_features_t netdev_get_wanted_features(
4821 struct net_device *dev)
4822{
4823 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4824}
4825netdev_features_t netdev_increment_features(netdev_features_t all,
4826 netdev_features_t one, netdev_features_t mask);
4827
4828/* Allow TSO being used on stacked device :
4829 * Performing the GSO segmentation before last device
4830 * is a performance improvement.
4831 */
4832static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4833 netdev_features_t mask)
4834{
4835 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4836}
4837
4838int __netdev_update_features(struct net_device *dev);
4839void netdev_update_features(struct net_device *dev);
4840void netdev_change_features(struct net_device *dev);
4841
4842void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4843 struct net_device *dev);
4844
4845netdev_features_t passthru_features_check(struct sk_buff *skb,
4846 struct net_device *dev,
4847 netdev_features_t features);
4848netdev_features_t netif_skb_features(struct sk_buff *skb);
4849
4850static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4851{
4852 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4853
4854 /* check flags correspondence */
4855 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4856 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4857 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4858 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4859 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4860 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4861 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4862 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4863 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4864 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4865 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4866 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4867 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4868 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4869 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4870 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4871 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4872 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4873 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4874
4875 return (features & feature) == feature;
4876}
4877
4878static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4879{
4880 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4881 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4882}
4883
4884static inline bool netif_needs_gso(struct sk_buff *skb,
4885 netdev_features_t features)
4886{
4887 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4888 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4889 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4890}
4891
4892static inline void netif_set_gso_max_size(struct net_device *dev,
4893 unsigned int size)
4894{
4895 /* dev->gso_max_size is read locklessly from sk_setup_caps() */
4896 WRITE_ONCE(dev->gso_max_size, size);
4897}
4898
4899static inline void netif_set_gso_max_segs(struct net_device *dev,
4900 unsigned int segs)
4901{
4902 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */
4903 WRITE_ONCE(dev->gso_max_segs, segs);
4904}
4905
4906static inline void netif_set_gro_max_size(struct net_device *dev,
4907 unsigned int size)
4908{
4909 /* This pairs with the READ_ONCE() in skb_gro_receive() */
4910 WRITE_ONCE(dev->gro_max_size, size);
4911}
4912
4913static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4914 int pulled_hlen, u16 mac_offset,
4915 int mac_len)
4916{
4917 skb->protocol = protocol;
4918 skb->encapsulation = 1;
4919 skb_push(skb, pulled_hlen);
4920 skb_reset_transport_header(skb);
4921 skb->mac_header = mac_offset;
4922 skb->network_header = skb->mac_header + mac_len;
4923 skb->mac_len = mac_len;
4924}
4925
4926static inline bool netif_is_macsec(const struct net_device *dev)
4927{
4928 return dev->priv_flags & IFF_MACSEC;
4929}
4930
4931static inline bool netif_is_macvlan(const struct net_device *dev)
4932{
4933 return dev->priv_flags & IFF_MACVLAN;
4934}
4935
4936static inline bool netif_is_macvlan_port(const struct net_device *dev)
4937{
4938 return dev->priv_flags & IFF_MACVLAN_PORT;
4939}
4940
4941static inline bool netif_is_bond_master(const struct net_device *dev)
4942{
4943 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4944}
4945
4946static inline bool netif_is_bond_slave(const struct net_device *dev)
4947{
4948 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4949}
4950
4951static inline bool netif_supports_nofcs(struct net_device *dev)
4952{
4953 return dev->priv_flags & IFF_SUPP_NOFCS;
4954}
4955
4956static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4957{
4958 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4959}
4960
4961static inline bool netif_is_l3_master(const struct net_device *dev)
4962{
4963 return dev->priv_flags & IFF_L3MDEV_MASTER;
4964}
4965
4966static inline bool netif_is_l3_slave(const struct net_device *dev)
4967{
4968 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4969}
4970
4971static inline bool netif_is_bridge_master(const struct net_device *dev)
4972{
4973 return dev->priv_flags & IFF_EBRIDGE;
4974}
4975
4976static inline bool netif_is_bridge_port(const struct net_device *dev)
4977{
4978 return dev->priv_flags & IFF_BRIDGE_PORT;
4979}
4980
4981static inline bool netif_is_ovs_master(const struct net_device *dev)
4982{
4983 return dev->priv_flags & IFF_OPENVSWITCH;
4984}
4985
4986static inline bool netif_is_ovs_port(const struct net_device *dev)
4987{
4988 return dev->priv_flags & IFF_OVS_DATAPATH;
4989}
4990
4991static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4992{
4993 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4994}
4995
4996static inline bool netif_is_team_master(const struct net_device *dev)
4997{
4998 return dev->priv_flags & IFF_TEAM;
4999}
5000
5001static inline bool netif_is_team_port(const struct net_device *dev)
5002{
5003 return dev->priv_flags & IFF_TEAM_PORT;
5004}
5005
5006static inline bool netif_is_lag_master(const struct net_device *dev)
5007{
5008 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5009}
5010
5011static inline bool netif_is_lag_port(const struct net_device *dev)
5012{
5013 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5014}
5015
5016static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5017{
5018 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5019}
5020
5021static inline bool netif_is_failover(const struct net_device *dev)
5022{
5023 return dev->priv_flags & IFF_FAILOVER;
5024}
5025
5026static inline bool netif_is_failover_slave(const struct net_device *dev)
5027{
5028 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5029}
5030
5031/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5032static inline void netif_keep_dst(struct net_device *dev)
5033{
5034 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5035}
5036
5037/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5038static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5039{
5040 /* TODO: reserve and use an additional IFF bit, if we get more users */
5041 return netif_is_macsec(dev);
5042}
5043
5044extern struct pernet_operations __net_initdata loopback_net_ops;
5045
5046/* Logging, debugging and troubleshooting/diagnostic helpers. */
5047
5048/* netdev_printk helpers, similar to dev_printk */
5049
5050static inline const char *netdev_name(const struct net_device *dev)
5051{
5052 if (!dev->name[0] || strchr(dev->name, '%'))
5053 return "(unnamed net_device)";
5054 return dev->name;
5055}
5056
5057static inline bool netdev_unregistering(const struct net_device *dev)
5058{
5059 return dev->reg_state == NETREG_UNREGISTERING;
5060}
5061
5062static inline const char *netdev_reg_state(const struct net_device *dev)
5063{
5064 switch (dev->reg_state) {
5065 case NETREG_UNINITIALIZED: return " (uninitialized)";
5066 case NETREG_REGISTERED: return "";
5067 case NETREG_UNREGISTERING: return " (unregistering)";
5068 case NETREG_UNREGISTERED: return " (unregistered)";
5069 case NETREG_RELEASED: return " (released)";
5070 case NETREG_DUMMY: return " (dummy)";
5071 }
5072
5073 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5074 return " (unknown)";
5075}
5076
5077__printf(3, 4) __cold
5078void netdev_printk(const char *level, const struct net_device *dev,
5079 const char *format, ...);
5080__printf(2, 3) __cold
5081void netdev_emerg(const struct net_device *dev, const char *format, ...);
5082__printf(2, 3) __cold
5083void netdev_alert(const struct net_device *dev, const char *format, ...);
5084__printf(2, 3) __cold
5085void netdev_crit(const struct net_device *dev, const char *format, ...);
5086__printf(2, 3) __cold
5087void netdev_err(const struct net_device *dev, const char *format, ...);
5088__printf(2, 3) __cold
5089void netdev_warn(const struct net_device *dev, const char *format, ...);
5090__printf(2, 3) __cold
5091void netdev_notice(const struct net_device *dev, const char *format, ...);
5092__printf(2, 3) __cold
5093void netdev_info(const struct net_device *dev, const char *format, ...);
5094
5095#define netdev_level_once(level, dev, fmt, ...) \
5096do { \
5097 static bool __section(".data.once") __print_once; \
5098 \
5099 if (!__print_once) { \
5100 __print_once = true; \
5101 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5102 } \
5103} while (0)
5104
5105#define netdev_emerg_once(dev, fmt, ...) \
5106 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5107#define netdev_alert_once(dev, fmt, ...) \
5108 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5109#define netdev_crit_once(dev, fmt, ...) \
5110 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5111#define netdev_err_once(dev, fmt, ...) \
5112 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5113#define netdev_warn_once(dev, fmt, ...) \
5114 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5115#define netdev_notice_once(dev, fmt, ...) \
5116 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5117#define netdev_info_once(dev, fmt, ...) \
5118 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5119
5120#define MODULE_ALIAS_NETDEV(device) \
5121 MODULE_ALIAS("netdev-" device)
5122
5123#if defined(CONFIG_DYNAMIC_DEBUG) || \
5124 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5125#define netdev_dbg(__dev, format, args...) \
5126do { \
5127 dynamic_netdev_dbg(__dev, format, ##args); \
5128} while (0)
5129#elif defined(DEBUG)
5130#define netdev_dbg(__dev, format, args...) \
5131 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5132#else
5133#define netdev_dbg(__dev, format, args...) \
5134({ \
5135 if (0) \
5136 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5137})
5138#endif
5139
5140#if defined(VERBOSE_DEBUG)
5141#define netdev_vdbg netdev_dbg
5142#else
5143
5144#define netdev_vdbg(dev, format, args...) \
5145({ \
5146 if (0) \
5147 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5148 0; \
5149})
5150#endif
5151
5152/*
5153 * netdev_WARN() acts like dev_printk(), but with the key difference
5154 * of using a WARN/WARN_ON to get the message out, including the
5155 * file/line information and a backtrace.
5156 */
5157#define netdev_WARN(dev, format, args...) \
5158 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5159 netdev_reg_state(dev), ##args)
5160
5161#define netdev_WARN_ONCE(dev, format, args...) \
5162 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5163 netdev_reg_state(dev), ##args)
5164
5165/* netif printk helpers, similar to netdev_printk */
5166
5167#define netif_printk(priv, type, level, dev, fmt, args...) \
5168do { \
5169 if (netif_msg_##type(priv)) \
5170 netdev_printk(level, (dev), fmt, ##args); \
5171} while (0)
5172
5173#define netif_level(level, priv, type, dev, fmt, args...) \
5174do { \
5175 if (netif_msg_##type(priv)) \
5176 netdev_##level(dev, fmt, ##args); \
5177} while (0)
5178
5179#define netif_emerg(priv, type, dev, fmt, args...) \
5180 netif_level(emerg, priv, type, dev, fmt, ##args)
5181#define netif_alert(priv, type, dev, fmt, args...) \
5182 netif_level(alert, priv, type, dev, fmt, ##args)
5183#define netif_crit(priv, type, dev, fmt, args...) \
5184 netif_level(crit, priv, type, dev, fmt, ##args)
5185#define netif_err(priv, type, dev, fmt, args...) \
5186 netif_level(err, priv, type, dev, fmt, ##args)
5187#define netif_warn(priv, type, dev, fmt, args...) \
5188 netif_level(warn, priv, type, dev, fmt, ##args)
5189#define netif_notice(priv, type, dev, fmt, args...) \
5190 netif_level(notice, priv, type, dev, fmt, ##args)
5191#define netif_info(priv, type, dev, fmt, args...) \
5192 netif_level(info, priv, type, dev, fmt, ##args)
5193
5194#if defined(CONFIG_DYNAMIC_DEBUG) || \
5195 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5196#define netif_dbg(priv, type, netdev, format, args...) \
5197do { \
5198 if (netif_msg_##type(priv)) \
5199 dynamic_netdev_dbg(netdev, format, ##args); \
5200} while (0)
5201#elif defined(DEBUG)
5202#define netif_dbg(priv, type, dev, format, args...) \
5203 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5204#else
5205#define netif_dbg(priv, type, dev, format, args...) \
5206({ \
5207 if (0) \
5208 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5209 0; \
5210})
5211#endif
5212
5213/* if @cond then downgrade to debug, else print at @level */
5214#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5215 do { \
5216 if (cond) \
5217 netif_dbg(priv, type, netdev, fmt, ##args); \
5218 else \
5219 netif_ ## level(priv, type, netdev, fmt, ##args); \
5220 } while (0)
5221
5222#if defined(VERBOSE_DEBUG)
5223#define netif_vdbg netif_dbg
5224#else
5225#define netif_vdbg(priv, type, dev, format, args...) \
5226({ \
5227 if (0) \
5228 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5229 0; \
5230})
5231#endif
5232
5233/*
5234 * The list of packet types we will receive (as opposed to discard)
5235 * and the routines to invoke.
5236 *
5237 * Why 16. Because with 16 the only overlap we get on a hash of the
5238 * low nibble of the protocol value is RARP/SNAP/X.25.
5239 *
5240 * 0800 IP
5241 * 0001 802.3
5242 * 0002 AX.25
5243 * 0004 802.2
5244 * 8035 RARP
5245 * 0005 SNAP
5246 * 0805 X.25
5247 * 0806 ARP
5248 * 8137 IPX
5249 * 0009 Localtalk
5250 * 86DD IPv6
5251 */
5252#define PTYPE_HASH_SIZE (16)
5253#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5254
5255extern struct list_head ptype_all __read_mostly;
5256extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5257
5258extern struct net_device *blackhole_netdev;
5259
5260#endif /* _LINUX_NETDEVICE_H */