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#include <net/net_debug.h>
54
55struct netpoll_info;
56struct device;
57struct ethtool_ops;
58struct phy_device;
59struct dsa_port;
60struct ip_tunnel_parm;
61struct macsec_context;
62struct macsec_ops;
63struct netdev_name_node;
64struct sd_flow_limit;
65struct sfp_bus;
66/* 802.11 specific */
67struct wireless_dev;
68/* 802.15.4 specific */
69struct wpan_dev;
70struct mpls_dev;
71/* UDP Tunnel offloads */
72struct udp_tunnel_info;
73struct udp_tunnel_nic_info;
74struct udp_tunnel_nic;
75struct bpf_prog;
76struct xdp_buff;
77
78void synchronize_net(void);
79void netdev_set_default_ethtool_ops(struct net_device *dev,
80 const struct ethtool_ops *ops);
81
82/* Backlog congestion levels */
83#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
84#define NET_RX_DROP 1 /* packet dropped */
85
86#define MAX_NEST_DEV 8
87
88/*
89 * Transmit return codes: transmit return codes originate from three different
90 * namespaces:
91 *
92 * - qdisc return codes
93 * - driver transmit return codes
94 * - errno values
95 *
96 * Drivers are allowed to return any one of those in their hard_start_xmit()
97 * function. Real network devices commonly used with qdiscs should only return
98 * the driver transmit return codes though - when qdiscs are used, the actual
99 * transmission happens asynchronously, so the value is not propagated to
100 * higher layers. Virtual network devices transmit synchronously; in this case
101 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
102 * others are propagated to higher layers.
103 */
104
105/* qdisc ->enqueue() return codes. */
106#define NET_XMIT_SUCCESS 0x00
107#define NET_XMIT_DROP 0x01 /* skb dropped */
108#define NET_XMIT_CN 0x02 /* congestion notification */
109#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
110
111/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
112 * indicates that the device will soon be dropping packets, or already drops
113 * some packets of the same priority; prompting us to send less aggressively. */
114#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
115#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
116
117/* Driver transmit return codes */
118#define NETDEV_TX_MASK 0xf0
119
120enum netdev_tx {
121 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
122 NETDEV_TX_OK = 0x00, /* driver took care of packet */
123 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
124};
125typedef enum netdev_tx netdev_tx_t;
126
127/*
128 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
129 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
130 */
131static inline bool dev_xmit_complete(int rc)
132{
133 /*
134 * Positive cases with an skb consumed by a driver:
135 * - successful transmission (rc == NETDEV_TX_OK)
136 * - error while transmitting (rc < 0)
137 * - error while queueing to a different device (rc & NET_XMIT_MASK)
138 */
139 if (likely(rc < NET_XMIT_MASK))
140 return true;
141
142 return false;
143}
144
145/*
146 * Compute the worst-case header length according to the protocols
147 * used.
148 */
149
150#if defined(CONFIG_HYPERV_NET)
151# define LL_MAX_HEADER 128
152#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
153# if defined(CONFIG_MAC80211_MESH)
154# define LL_MAX_HEADER 128
155# else
156# define LL_MAX_HEADER 96
157# endif
158#else
159# define LL_MAX_HEADER 32
160#endif
161
162#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
163 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
164#define MAX_HEADER LL_MAX_HEADER
165#else
166#define MAX_HEADER (LL_MAX_HEADER + 48)
167#endif
168
169/*
170 * Old network device statistics. Fields are native words
171 * (unsigned long) so they can be read and written atomically.
172 */
173
174struct net_device_stats {
175 unsigned long rx_packets;
176 unsigned long tx_packets;
177 unsigned long rx_bytes;
178 unsigned long tx_bytes;
179 unsigned long rx_errors;
180 unsigned long tx_errors;
181 unsigned long rx_dropped;
182 unsigned long tx_dropped;
183 unsigned long multicast;
184 unsigned long collisions;
185 unsigned long rx_length_errors;
186 unsigned long rx_over_errors;
187 unsigned long rx_crc_errors;
188 unsigned long rx_frame_errors;
189 unsigned long rx_fifo_errors;
190 unsigned long rx_missed_errors;
191 unsigned long tx_aborted_errors;
192 unsigned long tx_carrier_errors;
193 unsigned long tx_fifo_errors;
194 unsigned long tx_heartbeat_errors;
195 unsigned long tx_window_errors;
196 unsigned long rx_compressed;
197 unsigned long tx_compressed;
198};
199
200/* per-cpu stats, allocated on demand.
201 * Try to fit them in a single cache line, for dev_get_stats() sake.
202 */
203struct net_device_core_stats {
204 unsigned long rx_dropped;
205 unsigned long tx_dropped;
206 unsigned long rx_nohandler;
207 unsigned long rx_otherhost_dropped;
208} __aligned(4 * sizeof(unsigned long));
209
210#include <linux/cache.h>
211#include <linux/skbuff.h>
212
213#ifdef CONFIG_RPS
214#include <linux/static_key.h>
215extern struct static_key_false rps_needed;
216extern struct static_key_false rfs_needed;
217#endif
218
219struct neighbour;
220struct neigh_parms;
221struct sk_buff;
222
223struct netdev_hw_addr {
224 struct list_head list;
225 struct rb_node node;
226 unsigned char addr[MAX_ADDR_LEN];
227 unsigned char type;
228#define NETDEV_HW_ADDR_T_LAN 1
229#define NETDEV_HW_ADDR_T_SAN 2
230#define NETDEV_HW_ADDR_T_UNICAST 3
231#define NETDEV_HW_ADDR_T_MULTICAST 4
232 bool global_use;
233 int sync_cnt;
234 int refcount;
235 int synced;
236 struct rcu_head rcu_head;
237};
238
239struct netdev_hw_addr_list {
240 struct list_head list;
241 int count;
242
243 /* Auxiliary tree for faster lookup on addition and deletion */
244 struct rb_root tree;
245};
246
247#define netdev_hw_addr_list_count(l) ((l)->count)
248#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
249#define netdev_hw_addr_list_for_each(ha, l) \
250 list_for_each_entry(ha, &(l)->list, list)
251
252#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
253#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
254#define netdev_for_each_uc_addr(ha, dev) \
255 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
256#define netdev_for_each_synced_uc_addr(_ha, _dev) \
257 netdev_for_each_uc_addr((_ha), (_dev)) \
258 if ((_ha)->sync_cnt)
259
260#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
261#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
262#define netdev_for_each_mc_addr(ha, dev) \
263 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
264#define netdev_for_each_synced_mc_addr(_ha, _dev) \
265 netdev_for_each_mc_addr((_ha), (_dev)) \
266 if ((_ha)->sync_cnt)
267
268struct hh_cache {
269 unsigned int hh_len;
270 seqlock_t hh_lock;
271
272 /* cached hardware header; allow for machine alignment needs. */
273#define HH_DATA_MOD 16
274#define HH_DATA_OFF(__len) \
275 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
276#define HH_DATA_ALIGN(__len) \
277 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
278 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
279};
280
281/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
282 * Alternative is:
283 * dev->hard_header_len ? (dev->hard_header_len +
284 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
285 *
286 * We could use other alignment values, but we must maintain the
287 * relationship HH alignment <= LL alignment.
288 */
289#define LL_RESERVED_SPACE(dev) \
290 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
291#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
292 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
293
294struct header_ops {
295 int (*create) (struct sk_buff *skb, struct net_device *dev,
296 unsigned short type, const void *daddr,
297 const void *saddr, unsigned int len);
298 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
299 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
300 void (*cache_update)(struct hh_cache *hh,
301 const struct net_device *dev,
302 const unsigned char *haddr);
303 bool (*validate)(const char *ll_header, unsigned int len);
304 __be16 (*parse_protocol)(const struct sk_buff *skb);
305};
306
307/* These flag bits are private to the generic network queueing
308 * layer; they may not be explicitly referenced by any other
309 * code.
310 */
311
312enum netdev_state_t {
313 __LINK_STATE_START,
314 __LINK_STATE_PRESENT,
315 __LINK_STATE_NOCARRIER,
316 __LINK_STATE_LINKWATCH_PENDING,
317 __LINK_STATE_DORMANT,
318 __LINK_STATE_TESTING,
319};
320
321struct gro_list {
322 struct list_head list;
323 int count;
324};
325
326/*
327 * size of gro hash buckets, must less than bit number of
328 * napi_struct::gro_bitmask
329 */
330#define GRO_HASH_BUCKETS 8
331
332/*
333 * Structure for NAPI scheduling similar to tasklet but with weighting
334 */
335struct napi_struct {
336 /* The poll_list must only be managed by the entity which
337 * changes the state of the NAPI_STATE_SCHED bit. This means
338 * whoever atomically sets that bit can add this napi_struct
339 * to the per-CPU poll_list, and whoever clears that bit
340 * can remove from the list right before clearing the bit.
341 */
342 struct list_head poll_list;
343
344 unsigned long state;
345 int weight;
346 int defer_hard_irqs_count;
347 unsigned long gro_bitmask;
348 int (*poll)(struct napi_struct *, int);
349#ifdef CONFIG_NETPOLL
350 int poll_owner;
351#endif
352 struct net_device *dev;
353 struct gro_list gro_hash[GRO_HASH_BUCKETS];
354 struct sk_buff *skb;
355 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
356 int rx_count; /* length of rx_list */
357 struct hrtimer timer;
358 struct list_head dev_list;
359 struct hlist_node napi_hash_node;
360 unsigned int napi_id;
361 struct task_struct *thread;
362};
363
364enum {
365 NAPI_STATE_SCHED, /* Poll is scheduled */
366 NAPI_STATE_MISSED, /* reschedule a napi */
367 NAPI_STATE_DISABLE, /* Disable pending */
368 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
369 NAPI_STATE_LISTED, /* NAPI added to system lists */
370 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
371 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
372 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
373 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
374 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
375};
376
377enum {
378 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
379 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
380 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
381 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
382 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
383 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
384 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
385 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
386 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
387 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
388};
389
390enum gro_result {
391 GRO_MERGED,
392 GRO_MERGED_FREE,
393 GRO_HELD,
394 GRO_NORMAL,
395 GRO_CONSUMED,
396};
397typedef enum gro_result gro_result_t;
398
399/*
400 * enum rx_handler_result - Possible return values for rx_handlers.
401 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
402 * further.
403 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
404 * case skb->dev was changed by rx_handler.
405 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
406 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
407 *
408 * rx_handlers are functions called from inside __netif_receive_skb(), to do
409 * special processing of the skb, prior to delivery to protocol handlers.
410 *
411 * Currently, a net_device can only have a single rx_handler registered. Trying
412 * to register a second rx_handler will return -EBUSY.
413 *
414 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
415 * To unregister a rx_handler on a net_device, use
416 * netdev_rx_handler_unregister().
417 *
418 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
419 * do with the skb.
420 *
421 * If the rx_handler consumed the skb in some way, it should return
422 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
423 * the skb to be delivered in some other way.
424 *
425 * If the rx_handler changed skb->dev, to divert the skb to another
426 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
427 * new device will be called if it exists.
428 *
429 * If the rx_handler decides the skb should be ignored, it should return
430 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
431 * are registered on exact device (ptype->dev == skb->dev).
432 *
433 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
434 * delivered, it should return RX_HANDLER_PASS.
435 *
436 * A device without a registered rx_handler will behave as if rx_handler
437 * returned RX_HANDLER_PASS.
438 */
439
440enum rx_handler_result {
441 RX_HANDLER_CONSUMED,
442 RX_HANDLER_ANOTHER,
443 RX_HANDLER_EXACT,
444 RX_HANDLER_PASS,
445};
446typedef enum rx_handler_result rx_handler_result_t;
447typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
448
449void __napi_schedule(struct napi_struct *n);
450void __napi_schedule_irqoff(struct napi_struct *n);
451
452static inline bool napi_disable_pending(struct napi_struct *n)
453{
454 return test_bit(NAPI_STATE_DISABLE, &n->state);
455}
456
457static inline bool napi_prefer_busy_poll(struct napi_struct *n)
458{
459 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
460}
461
462bool napi_schedule_prep(struct napi_struct *n);
463
464/**
465 * napi_schedule - schedule NAPI poll
466 * @n: NAPI context
467 *
468 * Schedule NAPI poll routine to be called if it is not already
469 * running.
470 */
471static inline void napi_schedule(struct napi_struct *n)
472{
473 if (napi_schedule_prep(n))
474 __napi_schedule(n);
475}
476
477/**
478 * napi_schedule_irqoff - schedule NAPI poll
479 * @n: NAPI context
480 *
481 * Variant of napi_schedule(), assuming hard irqs are masked.
482 */
483static inline void napi_schedule_irqoff(struct napi_struct *n)
484{
485 if (napi_schedule_prep(n))
486 __napi_schedule_irqoff(n);
487}
488
489/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
490static inline bool napi_reschedule(struct napi_struct *napi)
491{
492 if (napi_schedule_prep(napi)) {
493 __napi_schedule(napi);
494 return true;
495 }
496 return false;
497}
498
499bool napi_complete_done(struct napi_struct *n, int work_done);
500/**
501 * napi_complete - NAPI processing complete
502 * @n: NAPI context
503 *
504 * Mark NAPI processing as complete.
505 * Consider using napi_complete_done() instead.
506 * Return false if device should avoid rearming interrupts.
507 */
508static inline bool napi_complete(struct napi_struct *n)
509{
510 return napi_complete_done(n, 0);
511}
512
513int dev_set_threaded(struct net_device *dev, bool threaded);
514
515/**
516 * napi_disable - prevent NAPI from scheduling
517 * @n: NAPI context
518 *
519 * Stop NAPI from being scheduled on this context.
520 * Waits till any outstanding processing completes.
521 */
522void napi_disable(struct napi_struct *n);
523
524void napi_enable(struct napi_struct *n);
525
526/**
527 * napi_synchronize - wait until NAPI is not running
528 * @n: NAPI context
529 *
530 * Wait until NAPI is done being scheduled on this context.
531 * Waits till any outstanding processing completes but
532 * does not disable future activations.
533 */
534static inline void napi_synchronize(const struct napi_struct *n)
535{
536 if (IS_ENABLED(CONFIG_SMP))
537 while (test_bit(NAPI_STATE_SCHED, &n->state))
538 msleep(1);
539 else
540 barrier();
541}
542
543/**
544 * napi_if_scheduled_mark_missed - if napi is running, set the
545 * NAPIF_STATE_MISSED
546 * @n: NAPI context
547 *
548 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
549 * NAPI is scheduled.
550 **/
551static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
552{
553 unsigned long val, new;
554
555 val = READ_ONCE(n->state);
556 do {
557 if (val & NAPIF_STATE_DISABLE)
558 return true;
559
560 if (!(val & NAPIF_STATE_SCHED))
561 return false;
562
563 new = val | NAPIF_STATE_MISSED;
564 } while (!try_cmpxchg(&n->state, &val, new));
565
566 return true;
567}
568
569enum netdev_queue_state_t {
570 __QUEUE_STATE_DRV_XOFF,
571 __QUEUE_STATE_STACK_XOFF,
572 __QUEUE_STATE_FROZEN,
573};
574
575#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
576#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
577#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
578
579#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
580#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
581 QUEUE_STATE_FROZEN)
582#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
583 QUEUE_STATE_FROZEN)
584
585/*
586 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
587 * netif_tx_* functions below are used to manipulate this flag. The
588 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
589 * queue independently. The netif_xmit_*stopped functions below are called
590 * to check if the queue has been stopped by the driver or stack (either
591 * of the XOFF bits are set in the state). Drivers should not need to call
592 * netif_xmit*stopped functions, they should only be using netif_tx_*.
593 */
594
595struct netdev_queue {
596/*
597 * read-mostly part
598 */
599 struct net_device *dev;
600 netdevice_tracker dev_tracker;
601
602 struct Qdisc __rcu *qdisc;
603 struct Qdisc *qdisc_sleeping;
604#ifdef CONFIG_SYSFS
605 struct kobject kobj;
606#endif
607#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 int numa_node;
609#endif
610 unsigned long tx_maxrate;
611 /*
612 * Number of TX timeouts for this queue
613 * (/sys/class/net/DEV/Q/trans_timeout)
614 */
615 atomic_long_t trans_timeout;
616
617 /* Subordinate device that the queue has been assigned to */
618 struct net_device *sb_dev;
619#ifdef CONFIG_XDP_SOCKETS
620 struct xsk_buff_pool *pool;
621#endif
622/*
623 * write-mostly part
624 */
625 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
626 int xmit_lock_owner;
627 /*
628 * Time (in jiffies) of last Tx
629 */
630 unsigned long trans_start;
631
632 unsigned long state;
633
634#ifdef CONFIG_BQL
635 struct dql dql;
636#endif
637} ____cacheline_aligned_in_smp;
638
639extern int sysctl_fb_tunnels_only_for_init_net;
640extern int sysctl_devconf_inherit_init_net;
641
642/*
643 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
644 * == 1 : For initns only
645 * == 2 : For none.
646 */
647static inline bool net_has_fallback_tunnels(const struct net *net)
648{
649#if IS_ENABLED(CONFIG_SYSCTL)
650 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
651
652 return !fb_tunnels_only_for_init_net ||
653 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
654#else
655 return true;
656#endif
657}
658
659static inline int net_inherit_devconf(void)
660{
661#if IS_ENABLED(CONFIG_SYSCTL)
662 return READ_ONCE(sysctl_devconf_inherit_init_net);
663#else
664 return 0;
665#endif
666}
667
668static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
669{
670#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
671 return q->numa_node;
672#else
673 return NUMA_NO_NODE;
674#endif
675}
676
677static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
678{
679#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
680 q->numa_node = node;
681#endif
682}
683
684#ifdef CONFIG_RPS
685/*
686 * This structure holds an RPS map which can be of variable length. The
687 * map is an array of CPUs.
688 */
689struct rps_map {
690 unsigned int len;
691 struct rcu_head rcu;
692 u16 cpus[];
693};
694#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
695
696/*
697 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
698 * tail pointer for that CPU's input queue at the time of last enqueue, and
699 * a hardware filter index.
700 */
701struct rps_dev_flow {
702 u16 cpu;
703 u16 filter;
704 unsigned int last_qtail;
705};
706#define RPS_NO_FILTER 0xffff
707
708/*
709 * The rps_dev_flow_table structure contains a table of flow mappings.
710 */
711struct rps_dev_flow_table {
712 unsigned int mask;
713 struct rcu_head rcu;
714 struct rps_dev_flow flows[];
715};
716#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
717 ((_num) * sizeof(struct rps_dev_flow)))
718
719/*
720 * The rps_sock_flow_table contains mappings of flows to the last CPU
721 * on which they were processed by the application (set in recvmsg).
722 * Each entry is a 32bit value. Upper part is the high-order bits
723 * of flow hash, lower part is CPU number.
724 * rps_cpu_mask is used to partition the space, depending on number of
725 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
726 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
727 * meaning we use 32-6=26 bits for the hash.
728 */
729struct rps_sock_flow_table {
730 u32 mask;
731
732 u32 ents[] ____cacheline_aligned_in_smp;
733};
734#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
735
736#define RPS_NO_CPU 0xffff
737
738extern u32 rps_cpu_mask;
739extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
740
741static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
742 u32 hash)
743{
744 if (table && hash) {
745 unsigned int index = hash & table->mask;
746 u32 val = hash & ~rps_cpu_mask;
747
748 /* We only give a hint, preemption can change CPU under us */
749 val |= raw_smp_processor_id();
750
751 if (table->ents[index] != val)
752 table->ents[index] = val;
753 }
754}
755
756#ifdef CONFIG_RFS_ACCEL
757bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
758 u16 filter_id);
759#endif
760#endif /* CONFIG_RPS */
761
762/* This structure contains an instance of an RX queue. */
763struct netdev_rx_queue {
764 struct xdp_rxq_info xdp_rxq;
765#ifdef CONFIG_RPS
766 struct rps_map __rcu *rps_map;
767 struct rps_dev_flow_table __rcu *rps_flow_table;
768#endif
769 struct kobject kobj;
770 struct net_device *dev;
771 netdevice_tracker dev_tracker;
772
773#ifdef CONFIG_XDP_SOCKETS
774 struct xsk_buff_pool *pool;
775#endif
776} ____cacheline_aligned_in_smp;
777
778/*
779 * RX queue sysfs structures and functions.
780 */
781struct rx_queue_attribute {
782 struct attribute attr;
783 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
784 ssize_t (*store)(struct netdev_rx_queue *queue,
785 const char *buf, size_t len);
786};
787
788/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
789enum xps_map_type {
790 XPS_CPUS = 0,
791 XPS_RXQS,
792 XPS_MAPS_MAX,
793};
794
795#ifdef CONFIG_XPS
796/*
797 * This structure holds an XPS map which can be of variable length. The
798 * map is an array of queues.
799 */
800struct xps_map {
801 unsigned int len;
802 unsigned int alloc_len;
803 struct rcu_head rcu;
804 u16 queues[];
805};
806#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
807#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
808 - sizeof(struct xps_map)) / sizeof(u16))
809
810/*
811 * This structure holds all XPS maps for device. Maps are indexed by CPU.
812 *
813 * We keep track of the number of cpus/rxqs used when the struct is allocated,
814 * in nr_ids. This will help not accessing out-of-bound memory.
815 *
816 * We keep track of the number of traffic classes used when the struct is
817 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
818 * not crossing its upper bound, as the original dev->num_tc can be updated in
819 * the meantime.
820 */
821struct xps_dev_maps {
822 struct rcu_head rcu;
823 unsigned int nr_ids;
824 s16 num_tc;
825 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
826};
827
828#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
829 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
830
831#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
832 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
833
834#endif /* CONFIG_XPS */
835
836#define TC_MAX_QUEUE 16
837#define TC_BITMASK 15
838/* HW offloaded queuing disciplines txq count and offset maps */
839struct netdev_tc_txq {
840 u16 count;
841 u16 offset;
842};
843
844#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
845/*
846 * This structure is to hold information about the device
847 * configured to run FCoE protocol stack.
848 */
849struct netdev_fcoe_hbainfo {
850 char manufacturer[64];
851 char serial_number[64];
852 char hardware_version[64];
853 char driver_version[64];
854 char optionrom_version[64];
855 char firmware_version[64];
856 char model[256];
857 char model_description[256];
858};
859#endif
860
861#define MAX_PHYS_ITEM_ID_LEN 32
862
863/* This structure holds a unique identifier to identify some
864 * physical item (port for example) used by a netdevice.
865 */
866struct netdev_phys_item_id {
867 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
868 unsigned char id_len;
869};
870
871static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
872 struct netdev_phys_item_id *b)
873{
874 return a->id_len == b->id_len &&
875 memcmp(a->id, b->id, a->id_len) == 0;
876}
877
878typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
879 struct sk_buff *skb,
880 struct net_device *sb_dev);
881
882enum net_device_path_type {
883 DEV_PATH_ETHERNET = 0,
884 DEV_PATH_VLAN,
885 DEV_PATH_BRIDGE,
886 DEV_PATH_PPPOE,
887 DEV_PATH_DSA,
888 DEV_PATH_MTK_WDMA,
889};
890
891struct net_device_path {
892 enum net_device_path_type type;
893 const struct net_device *dev;
894 union {
895 struct {
896 u16 id;
897 __be16 proto;
898 u8 h_dest[ETH_ALEN];
899 } encap;
900 struct {
901 enum {
902 DEV_PATH_BR_VLAN_KEEP,
903 DEV_PATH_BR_VLAN_TAG,
904 DEV_PATH_BR_VLAN_UNTAG,
905 DEV_PATH_BR_VLAN_UNTAG_HW,
906 } vlan_mode;
907 u16 vlan_id;
908 __be16 vlan_proto;
909 } bridge;
910 struct {
911 int port;
912 u16 proto;
913 } dsa;
914 struct {
915 u8 wdma_idx;
916 u8 queue;
917 u16 wcid;
918 u8 bss;
919 } mtk_wdma;
920 };
921};
922
923#define NET_DEVICE_PATH_STACK_MAX 5
924#define NET_DEVICE_PATH_VLAN_MAX 2
925
926struct net_device_path_stack {
927 int num_paths;
928 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
929};
930
931struct net_device_path_ctx {
932 const struct net_device *dev;
933 u8 daddr[ETH_ALEN];
934
935 int num_vlans;
936 struct {
937 u16 id;
938 __be16 proto;
939 } vlan[NET_DEVICE_PATH_VLAN_MAX];
940};
941
942enum tc_setup_type {
943 TC_QUERY_CAPS,
944 TC_SETUP_QDISC_MQPRIO,
945 TC_SETUP_CLSU32,
946 TC_SETUP_CLSFLOWER,
947 TC_SETUP_CLSMATCHALL,
948 TC_SETUP_CLSBPF,
949 TC_SETUP_BLOCK,
950 TC_SETUP_QDISC_CBS,
951 TC_SETUP_QDISC_RED,
952 TC_SETUP_QDISC_PRIO,
953 TC_SETUP_QDISC_MQ,
954 TC_SETUP_QDISC_ETF,
955 TC_SETUP_ROOT_QDISC,
956 TC_SETUP_QDISC_GRED,
957 TC_SETUP_QDISC_TAPRIO,
958 TC_SETUP_FT,
959 TC_SETUP_QDISC_ETS,
960 TC_SETUP_QDISC_TBF,
961 TC_SETUP_QDISC_FIFO,
962 TC_SETUP_QDISC_HTB,
963 TC_SETUP_ACT,
964};
965
966/* These structures hold the attributes of bpf state that are being passed
967 * to the netdevice through the bpf op.
968 */
969enum bpf_netdev_command {
970 /* Set or clear a bpf program used in the earliest stages of packet
971 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
972 * is responsible for calling bpf_prog_put on any old progs that are
973 * stored. In case of error, the callee need not release the new prog
974 * reference, but on success it takes ownership and must bpf_prog_put
975 * when it is no longer used.
976 */
977 XDP_SETUP_PROG,
978 XDP_SETUP_PROG_HW,
979 /* BPF program for offload callbacks, invoked at program load time. */
980 BPF_OFFLOAD_MAP_ALLOC,
981 BPF_OFFLOAD_MAP_FREE,
982 XDP_SETUP_XSK_POOL,
983};
984
985struct bpf_prog_offload_ops;
986struct netlink_ext_ack;
987struct xdp_umem;
988struct xdp_dev_bulk_queue;
989struct bpf_xdp_link;
990
991enum bpf_xdp_mode {
992 XDP_MODE_SKB = 0,
993 XDP_MODE_DRV = 1,
994 XDP_MODE_HW = 2,
995 __MAX_XDP_MODE
996};
997
998struct bpf_xdp_entity {
999 struct bpf_prog *prog;
1000 struct bpf_xdp_link *link;
1001};
1002
1003struct netdev_bpf {
1004 enum bpf_netdev_command command;
1005 union {
1006 /* XDP_SETUP_PROG */
1007 struct {
1008 u32 flags;
1009 struct bpf_prog *prog;
1010 struct netlink_ext_ack *extack;
1011 };
1012 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1013 struct {
1014 struct bpf_offloaded_map *offmap;
1015 };
1016 /* XDP_SETUP_XSK_POOL */
1017 struct {
1018 struct xsk_buff_pool *pool;
1019 u16 queue_id;
1020 } xsk;
1021 };
1022};
1023
1024/* Flags for ndo_xsk_wakeup. */
1025#define XDP_WAKEUP_RX (1 << 0)
1026#define XDP_WAKEUP_TX (1 << 1)
1027
1028#ifdef CONFIG_XFRM_OFFLOAD
1029struct xfrmdev_ops {
1030 int (*xdo_dev_state_add) (struct xfrm_state *x);
1031 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1032 void (*xdo_dev_state_free) (struct xfrm_state *x);
1033 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1034 struct xfrm_state *x);
1035 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1036};
1037#endif
1038
1039struct dev_ifalias {
1040 struct rcu_head rcuhead;
1041 char ifalias[];
1042};
1043
1044struct devlink;
1045struct tlsdev_ops;
1046
1047struct netdev_net_notifier {
1048 struct list_head list;
1049 struct notifier_block *nb;
1050};
1051
1052/*
1053 * This structure defines the management hooks for network devices.
1054 * The following hooks can be defined; unless noted otherwise, they are
1055 * optional and can be filled with a null pointer.
1056 *
1057 * int (*ndo_init)(struct net_device *dev);
1058 * This function is called once when a network device is registered.
1059 * The network device can use this for any late stage initialization
1060 * or semantic validation. It can fail with an error code which will
1061 * be propagated back to register_netdev.
1062 *
1063 * void (*ndo_uninit)(struct net_device *dev);
1064 * This function is called when device is unregistered or when registration
1065 * fails. It is not called if init fails.
1066 *
1067 * int (*ndo_open)(struct net_device *dev);
1068 * This function is called when a network device transitions to the up
1069 * state.
1070 *
1071 * int (*ndo_stop)(struct net_device *dev);
1072 * This function is called when a network device transitions to the down
1073 * state.
1074 *
1075 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1076 * struct net_device *dev);
1077 * Called when a packet needs to be transmitted.
1078 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1079 * the queue before that can happen; it's for obsolete devices and weird
1080 * corner cases, but the stack really does a non-trivial amount
1081 * of useless work if you return NETDEV_TX_BUSY.
1082 * Required; cannot be NULL.
1083 *
1084 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1085 * struct net_device *dev
1086 * netdev_features_t features);
1087 * Called by core transmit path to determine if device is capable of
1088 * performing offload operations on a given packet. This is to give
1089 * the device an opportunity to implement any restrictions that cannot
1090 * be otherwise expressed by feature flags. The check is called with
1091 * the set of features that the stack has calculated and it returns
1092 * those the driver believes to be appropriate.
1093 *
1094 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1095 * struct net_device *sb_dev);
1096 * Called to decide which queue to use when device supports multiple
1097 * transmit queues.
1098 *
1099 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1100 * This function is called to allow device receiver to make
1101 * changes to configuration when multicast or promiscuous is enabled.
1102 *
1103 * void (*ndo_set_rx_mode)(struct net_device *dev);
1104 * This function is called device changes address list filtering.
1105 * If driver handles unicast address filtering, it should set
1106 * IFF_UNICAST_FLT in its priv_flags.
1107 *
1108 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1109 * This function is called when the Media Access Control address
1110 * needs to be changed. If this interface is not defined, the
1111 * MAC address can not be changed.
1112 *
1113 * int (*ndo_validate_addr)(struct net_device *dev);
1114 * Test if Media Access Control address is valid for the device.
1115 *
1116 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1117 * Old-style ioctl entry point. This is used internally by the
1118 * appletalk and ieee802154 subsystems but is no longer called by
1119 * the device ioctl handler.
1120 *
1121 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1122 * Used by the bonding driver for its device specific ioctls:
1123 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1124 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1125 *
1126 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1127 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1128 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1129 *
1130 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1131 * Used to set network devices bus interface parameters. This interface
1132 * is retained for legacy reasons; new devices should use the bus
1133 * interface (PCI) for low level management.
1134 *
1135 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1136 * Called when a user wants to change the Maximum Transfer Unit
1137 * of a device.
1138 *
1139 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1140 * Callback used when the transmitter has not made any progress
1141 * for dev->watchdog ticks.
1142 *
1143 * void (*ndo_get_stats64)(struct net_device *dev,
1144 * struct rtnl_link_stats64 *storage);
1145 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1146 * Called when a user wants to get the network device usage
1147 * statistics. Drivers must do one of the following:
1148 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1149 * rtnl_link_stats64 structure passed by the caller.
1150 * 2. Define @ndo_get_stats to update a net_device_stats structure
1151 * (which should normally be dev->stats) and return a pointer to
1152 * it. The structure may be changed asynchronously only if each
1153 * field is written atomically.
1154 * 3. Update dev->stats asynchronously and atomically, and define
1155 * neither operation.
1156 *
1157 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1158 * Return true if this device supports offload stats of this attr_id.
1159 *
1160 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1161 * void *attr_data)
1162 * Get statistics for offload operations by attr_id. Write it into the
1163 * attr_data pointer.
1164 *
1165 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1166 * If device supports VLAN filtering this function is called when a
1167 * VLAN id is registered.
1168 *
1169 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1170 * If device supports VLAN filtering this function is called when a
1171 * VLAN id is unregistered.
1172 *
1173 * void (*ndo_poll_controller)(struct net_device *dev);
1174 *
1175 * SR-IOV management functions.
1176 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1177 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1178 * u8 qos, __be16 proto);
1179 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1180 * int max_tx_rate);
1181 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1182 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1183 * int (*ndo_get_vf_config)(struct net_device *dev,
1184 * int vf, struct ifla_vf_info *ivf);
1185 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1186 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1187 * struct nlattr *port[]);
1188 *
1189 * Enable or disable the VF ability to query its RSS Redirection Table and
1190 * Hash Key. This is needed since on some devices VF share this information
1191 * with PF and querying it may introduce a theoretical security risk.
1192 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1193 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1194 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1195 * void *type_data);
1196 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1197 * This is always called from the stack with the rtnl lock held and netif
1198 * tx queues stopped. This allows the netdevice to perform queue
1199 * management safely.
1200 *
1201 * Fiber Channel over Ethernet (FCoE) offload functions.
1202 * int (*ndo_fcoe_enable)(struct net_device *dev);
1203 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1204 * so the underlying device can perform whatever needed configuration or
1205 * initialization to support acceleration of FCoE traffic.
1206 *
1207 * int (*ndo_fcoe_disable)(struct net_device *dev);
1208 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1209 * so the underlying device can perform whatever needed clean-ups to
1210 * stop supporting acceleration of FCoE traffic.
1211 *
1212 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1213 * struct scatterlist *sgl, unsigned int sgc);
1214 * Called when the FCoE Initiator wants to initialize an I/O that
1215 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1216 * perform necessary setup and returns 1 to indicate the device is set up
1217 * successfully to perform DDP on this I/O, otherwise this returns 0.
1218 *
1219 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1220 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1221 * indicated by the FC exchange id 'xid', so the underlying device can
1222 * clean up and reuse resources for later DDP requests.
1223 *
1224 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1225 * struct scatterlist *sgl, unsigned int sgc);
1226 * Called when the FCoE Target wants to initialize an I/O that
1227 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1228 * perform necessary setup and returns 1 to indicate the device is set up
1229 * successfully to perform DDP on this I/O, otherwise this returns 0.
1230 *
1231 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1232 * struct netdev_fcoe_hbainfo *hbainfo);
1233 * Called when the FCoE Protocol stack wants information on the underlying
1234 * device. This information is utilized by the FCoE protocol stack to
1235 * register attributes with Fiber Channel management service as per the
1236 * FC-GS Fabric Device Management Information(FDMI) specification.
1237 *
1238 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1239 * Called when the underlying device wants to override default World Wide
1240 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1241 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1242 * protocol stack to use.
1243 *
1244 * RFS acceleration.
1245 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1246 * u16 rxq_index, u32 flow_id);
1247 * Set hardware filter for RFS. rxq_index is the target queue index;
1248 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1249 * Return the filter ID on success, or a negative error code.
1250 *
1251 * Slave management functions (for bridge, bonding, etc).
1252 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1253 * Called to make another netdev an underling.
1254 *
1255 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1256 * Called to release previously enslaved netdev.
1257 *
1258 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1259 * struct sk_buff *skb,
1260 * bool all_slaves);
1261 * Get the xmit slave of master device. If all_slaves is true, function
1262 * assume all the slaves can transmit.
1263 *
1264 * Feature/offload setting functions.
1265 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1266 * netdev_features_t features);
1267 * Adjusts the requested feature flags according to device-specific
1268 * constraints, and returns the resulting flags. Must not modify
1269 * the device state.
1270 *
1271 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1272 * Called to update device configuration to new features. Passed
1273 * feature set might be less than what was returned by ndo_fix_features()).
1274 * Must return >0 or -errno if it changed dev->features itself.
1275 *
1276 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1277 * struct net_device *dev,
1278 * const unsigned char *addr, u16 vid, u16 flags,
1279 * struct netlink_ext_ack *extack);
1280 * Adds an FDB entry to dev for addr.
1281 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1282 * struct net_device *dev,
1283 * const unsigned char *addr, u16 vid)
1284 * Deletes the FDB entry from dev coresponding to addr.
1285 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1286 * struct net_device *dev,
1287 * u16 vid,
1288 * struct netlink_ext_ack *extack);
1289 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1290 * struct net_device *dev, struct net_device *filter_dev,
1291 * int *idx)
1292 * Used to add FDB entries to dump requests. Implementers should add
1293 * entries to skb and update idx with the number of entries.
1294 *
1295 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1296 * u16 flags, struct netlink_ext_ack *extack)
1297 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1298 * struct net_device *dev, u32 filter_mask,
1299 * int nlflags)
1300 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1301 * u16 flags);
1302 *
1303 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1304 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1305 * which do not represent real hardware may define this to allow their
1306 * userspace components to manage their virtual carrier state. Devices
1307 * that determine carrier state from physical hardware properties (eg
1308 * network cables) or protocol-dependent mechanisms (eg
1309 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1310 *
1311 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1312 * struct netdev_phys_item_id *ppid);
1313 * Called to get ID of physical port of this device. If driver does
1314 * not implement this, it is assumed that the hw is not able to have
1315 * multiple net devices on single physical port.
1316 *
1317 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1318 * struct netdev_phys_item_id *ppid)
1319 * Called to get the parent ID of the physical port of this device.
1320 *
1321 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1322 * struct net_device *dev)
1323 * Called by upper layer devices to accelerate switching or other
1324 * station functionality into hardware. 'pdev is the lowerdev
1325 * to use for the offload and 'dev' is the net device that will
1326 * back the offload. Returns a pointer to the private structure
1327 * the upper layer will maintain.
1328 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1329 * Called by upper layer device to delete the station created
1330 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1331 * the station and priv is the structure returned by the add
1332 * operation.
1333 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1334 * int queue_index, u32 maxrate);
1335 * Called when a user wants to set a max-rate limitation of specific
1336 * TX queue.
1337 * int (*ndo_get_iflink)(const struct net_device *dev);
1338 * Called to get the iflink value of this device.
1339 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1340 * This function is used to get egress tunnel information for given skb.
1341 * This is useful for retrieving outer tunnel header parameters while
1342 * sampling packet.
1343 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1344 * This function is used to specify the headroom that the skb must
1345 * consider when allocation skb during packet reception. Setting
1346 * appropriate rx headroom value allows avoiding skb head copy on
1347 * forward. Setting a negative value resets the rx headroom to the
1348 * default value.
1349 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1350 * This function is used to set or query state related to XDP on the
1351 * netdevice and manage BPF offload. See definition of
1352 * enum bpf_netdev_command for details.
1353 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1354 * u32 flags);
1355 * This function is used to submit @n XDP packets for transmit on a
1356 * netdevice. Returns number of frames successfully transmitted, frames
1357 * that got dropped are freed/returned via xdp_return_frame().
1358 * Returns negative number, means general error invoking ndo, meaning
1359 * no frames were xmit'ed and core-caller will free all frames.
1360 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1361 * struct xdp_buff *xdp);
1362 * Get the xmit slave of master device based on the xdp_buff.
1363 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1364 * This function is used to wake up the softirq, ksoftirqd or kthread
1365 * responsible for sending and/or receiving packets on a specific
1366 * queue id bound to an AF_XDP socket. The flags field specifies if
1367 * only RX, only Tx, or both should be woken up using the flags
1368 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1369 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1370 * Get devlink port instance associated with a given netdev.
1371 * Called with a reference on the netdevice and devlink locks only,
1372 * rtnl_lock is not held.
1373 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1374 * int cmd);
1375 * Add, change, delete or get information on an IPv4 tunnel.
1376 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1377 * If a device is paired with a peer device, return the peer instance.
1378 * The caller must be under RCU read context.
1379 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1380 * Get the forwarding path to reach the real device from the HW destination address
1381 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1382 * const struct skb_shared_hwtstamps *hwtstamps,
1383 * bool cycles);
1384 * Get hardware timestamp based on normal/adjustable time or free running
1385 * cycle counter. This function is required if physical clock supports a
1386 * free running cycle counter.
1387 */
1388struct net_device_ops {
1389 int (*ndo_init)(struct net_device *dev);
1390 void (*ndo_uninit)(struct net_device *dev);
1391 int (*ndo_open)(struct net_device *dev);
1392 int (*ndo_stop)(struct net_device *dev);
1393 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1394 struct net_device *dev);
1395 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1396 struct net_device *dev,
1397 netdev_features_t features);
1398 u16 (*ndo_select_queue)(struct net_device *dev,
1399 struct sk_buff *skb,
1400 struct net_device *sb_dev);
1401 void (*ndo_change_rx_flags)(struct net_device *dev,
1402 int flags);
1403 void (*ndo_set_rx_mode)(struct net_device *dev);
1404 int (*ndo_set_mac_address)(struct net_device *dev,
1405 void *addr);
1406 int (*ndo_validate_addr)(struct net_device *dev);
1407 int (*ndo_do_ioctl)(struct net_device *dev,
1408 struct ifreq *ifr, int cmd);
1409 int (*ndo_eth_ioctl)(struct net_device *dev,
1410 struct ifreq *ifr, int cmd);
1411 int (*ndo_siocbond)(struct net_device *dev,
1412 struct ifreq *ifr, int cmd);
1413 int (*ndo_siocwandev)(struct net_device *dev,
1414 struct if_settings *ifs);
1415 int (*ndo_siocdevprivate)(struct net_device *dev,
1416 struct ifreq *ifr,
1417 void __user *data, int cmd);
1418 int (*ndo_set_config)(struct net_device *dev,
1419 struct ifmap *map);
1420 int (*ndo_change_mtu)(struct net_device *dev,
1421 int new_mtu);
1422 int (*ndo_neigh_setup)(struct net_device *dev,
1423 struct neigh_parms *);
1424 void (*ndo_tx_timeout) (struct net_device *dev,
1425 unsigned int txqueue);
1426
1427 void (*ndo_get_stats64)(struct net_device *dev,
1428 struct rtnl_link_stats64 *storage);
1429 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1430 int (*ndo_get_offload_stats)(int attr_id,
1431 const struct net_device *dev,
1432 void *attr_data);
1433 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1434
1435 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1436 __be16 proto, u16 vid);
1437 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1438 __be16 proto, u16 vid);
1439#ifdef CONFIG_NET_POLL_CONTROLLER
1440 void (*ndo_poll_controller)(struct net_device *dev);
1441 int (*ndo_netpoll_setup)(struct net_device *dev,
1442 struct netpoll_info *info);
1443 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1444#endif
1445 int (*ndo_set_vf_mac)(struct net_device *dev,
1446 int queue, u8 *mac);
1447 int (*ndo_set_vf_vlan)(struct net_device *dev,
1448 int queue, u16 vlan,
1449 u8 qos, __be16 proto);
1450 int (*ndo_set_vf_rate)(struct net_device *dev,
1451 int vf, int min_tx_rate,
1452 int max_tx_rate);
1453 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1454 int vf, bool setting);
1455 int (*ndo_set_vf_trust)(struct net_device *dev,
1456 int vf, bool setting);
1457 int (*ndo_get_vf_config)(struct net_device *dev,
1458 int vf,
1459 struct ifla_vf_info *ivf);
1460 int (*ndo_set_vf_link_state)(struct net_device *dev,
1461 int vf, int link_state);
1462 int (*ndo_get_vf_stats)(struct net_device *dev,
1463 int vf,
1464 struct ifla_vf_stats
1465 *vf_stats);
1466 int (*ndo_set_vf_port)(struct net_device *dev,
1467 int vf,
1468 struct nlattr *port[]);
1469 int (*ndo_get_vf_port)(struct net_device *dev,
1470 int vf, struct sk_buff *skb);
1471 int (*ndo_get_vf_guid)(struct net_device *dev,
1472 int vf,
1473 struct ifla_vf_guid *node_guid,
1474 struct ifla_vf_guid *port_guid);
1475 int (*ndo_set_vf_guid)(struct net_device *dev,
1476 int vf, u64 guid,
1477 int guid_type);
1478 int (*ndo_set_vf_rss_query_en)(
1479 struct net_device *dev,
1480 int vf, bool setting);
1481 int (*ndo_setup_tc)(struct net_device *dev,
1482 enum tc_setup_type type,
1483 void *type_data);
1484#if IS_ENABLED(CONFIG_FCOE)
1485 int (*ndo_fcoe_enable)(struct net_device *dev);
1486 int (*ndo_fcoe_disable)(struct net_device *dev);
1487 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1488 u16 xid,
1489 struct scatterlist *sgl,
1490 unsigned int sgc);
1491 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1492 u16 xid);
1493 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1494 u16 xid,
1495 struct scatterlist *sgl,
1496 unsigned int sgc);
1497 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1498 struct netdev_fcoe_hbainfo *hbainfo);
1499#endif
1500
1501#if IS_ENABLED(CONFIG_LIBFCOE)
1502#define NETDEV_FCOE_WWNN 0
1503#define NETDEV_FCOE_WWPN 1
1504 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1505 u64 *wwn, int type);
1506#endif
1507
1508#ifdef CONFIG_RFS_ACCEL
1509 int (*ndo_rx_flow_steer)(struct net_device *dev,
1510 const struct sk_buff *skb,
1511 u16 rxq_index,
1512 u32 flow_id);
1513#endif
1514 int (*ndo_add_slave)(struct net_device *dev,
1515 struct net_device *slave_dev,
1516 struct netlink_ext_ack *extack);
1517 int (*ndo_del_slave)(struct net_device *dev,
1518 struct net_device *slave_dev);
1519 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1520 struct sk_buff *skb,
1521 bool all_slaves);
1522 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1523 struct sock *sk);
1524 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1525 netdev_features_t features);
1526 int (*ndo_set_features)(struct net_device *dev,
1527 netdev_features_t features);
1528 int (*ndo_neigh_construct)(struct net_device *dev,
1529 struct neighbour *n);
1530 void (*ndo_neigh_destroy)(struct net_device *dev,
1531 struct neighbour *n);
1532
1533 int (*ndo_fdb_add)(struct ndmsg *ndm,
1534 struct nlattr *tb[],
1535 struct net_device *dev,
1536 const unsigned char *addr,
1537 u16 vid,
1538 u16 flags,
1539 struct netlink_ext_ack *extack);
1540 int (*ndo_fdb_del)(struct ndmsg *ndm,
1541 struct nlattr *tb[],
1542 struct net_device *dev,
1543 const unsigned char *addr,
1544 u16 vid, struct netlink_ext_ack *extack);
1545 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1546 struct nlattr *tb[],
1547 struct net_device *dev,
1548 u16 vid,
1549 struct netlink_ext_ack *extack);
1550 int (*ndo_fdb_dump)(struct sk_buff *skb,
1551 struct netlink_callback *cb,
1552 struct net_device *dev,
1553 struct net_device *filter_dev,
1554 int *idx);
1555 int (*ndo_fdb_get)(struct sk_buff *skb,
1556 struct nlattr *tb[],
1557 struct net_device *dev,
1558 const unsigned char *addr,
1559 u16 vid, u32 portid, u32 seq,
1560 struct netlink_ext_ack *extack);
1561 int (*ndo_bridge_setlink)(struct net_device *dev,
1562 struct nlmsghdr *nlh,
1563 u16 flags,
1564 struct netlink_ext_ack *extack);
1565 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1566 u32 pid, u32 seq,
1567 struct net_device *dev,
1568 u32 filter_mask,
1569 int nlflags);
1570 int (*ndo_bridge_dellink)(struct net_device *dev,
1571 struct nlmsghdr *nlh,
1572 u16 flags);
1573 int (*ndo_change_carrier)(struct net_device *dev,
1574 bool new_carrier);
1575 int (*ndo_get_phys_port_id)(struct net_device *dev,
1576 struct netdev_phys_item_id *ppid);
1577 int (*ndo_get_port_parent_id)(struct net_device *dev,
1578 struct netdev_phys_item_id *ppid);
1579 int (*ndo_get_phys_port_name)(struct net_device *dev,
1580 char *name, size_t len);
1581 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1582 struct net_device *dev);
1583 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1584 void *priv);
1585
1586 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1587 int queue_index,
1588 u32 maxrate);
1589 int (*ndo_get_iflink)(const struct net_device *dev);
1590 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1591 struct sk_buff *skb);
1592 void (*ndo_set_rx_headroom)(struct net_device *dev,
1593 int needed_headroom);
1594 int (*ndo_bpf)(struct net_device *dev,
1595 struct netdev_bpf *bpf);
1596 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1597 struct xdp_frame **xdp,
1598 u32 flags);
1599 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1600 struct xdp_buff *xdp);
1601 int (*ndo_xsk_wakeup)(struct net_device *dev,
1602 u32 queue_id, u32 flags);
1603 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1604 int (*ndo_tunnel_ctl)(struct net_device *dev,
1605 struct ip_tunnel_parm *p, int cmd);
1606 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1607 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1608 struct net_device_path *path);
1609 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1610 const struct skb_shared_hwtstamps *hwtstamps,
1611 bool cycles);
1612};
1613
1614/**
1615 * enum netdev_priv_flags - &struct net_device priv_flags
1616 *
1617 * These are the &struct net_device, they are only set internally
1618 * by drivers and used in the kernel. These flags are invisible to
1619 * userspace; this means that the order of these flags can change
1620 * during any kernel release.
1621 *
1622 * You should have a pretty good reason to be extending these flags.
1623 *
1624 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1625 * @IFF_EBRIDGE: Ethernet bridging device
1626 * @IFF_BONDING: bonding master or slave
1627 * @IFF_ISATAP: ISATAP interface (RFC4214)
1628 * @IFF_WAN_HDLC: WAN HDLC device
1629 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1630 * release skb->dst
1631 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1632 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1633 * @IFF_MACVLAN_PORT: device used as macvlan port
1634 * @IFF_BRIDGE_PORT: device used as bridge port
1635 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1636 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1637 * @IFF_UNICAST_FLT: Supports unicast filtering
1638 * @IFF_TEAM_PORT: device used as team port
1639 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1640 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1641 * change when it's running
1642 * @IFF_MACVLAN: Macvlan device
1643 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1644 * underlying stacked devices
1645 * @IFF_L3MDEV_MASTER: device is an L3 master device
1646 * @IFF_NO_QUEUE: device can run without qdisc attached
1647 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1648 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1649 * @IFF_TEAM: device is a team device
1650 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1651 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1652 * entity (i.e. the master device for bridged veth)
1653 * @IFF_MACSEC: device is a MACsec device
1654 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1655 * @IFF_FAILOVER: device is a failover master device
1656 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1657 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1658 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1659 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1660 * skb_headlen(skb) == 0 (data starts from frag0)
1661 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1662 */
1663enum netdev_priv_flags {
1664 IFF_802_1Q_VLAN = 1<<0,
1665 IFF_EBRIDGE = 1<<1,
1666 IFF_BONDING = 1<<2,
1667 IFF_ISATAP = 1<<3,
1668 IFF_WAN_HDLC = 1<<4,
1669 IFF_XMIT_DST_RELEASE = 1<<5,
1670 IFF_DONT_BRIDGE = 1<<6,
1671 IFF_DISABLE_NETPOLL = 1<<7,
1672 IFF_MACVLAN_PORT = 1<<8,
1673 IFF_BRIDGE_PORT = 1<<9,
1674 IFF_OVS_DATAPATH = 1<<10,
1675 IFF_TX_SKB_SHARING = 1<<11,
1676 IFF_UNICAST_FLT = 1<<12,
1677 IFF_TEAM_PORT = 1<<13,
1678 IFF_SUPP_NOFCS = 1<<14,
1679 IFF_LIVE_ADDR_CHANGE = 1<<15,
1680 IFF_MACVLAN = 1<<16,
1681 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1682 IFF_L3MDEV_MASTER = 1<<18,
1683 IFF_NO_QUEUE = 1<<19,
1684 IFF_OPENVSWITCH = 1<<20,
1685 IFF_L3MDEV_SLAVE = 1<<21,
1686 IFF_TEAM = 1<<22,
1687 IFF_RXFH_CONFIGURED = 1<<23,
1688 IFF_PHONY_HEADROOM = 1<<24,
1689 IFF_MACSEC = 1<<25,
1690 IFF_NO_RX_HANDLER = 1<<26,
1691 IFF_FAILOVER = 1<<27,
1692 IFF_FAILOVER_SLAVE = 1<<28,
1693 IFF_L3MDEV_RX_HANDLER = 1<<29,
1694 IFF_LIVE_RENAME_OK = 1<<30,
1695 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1696 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1697};
1698
1699#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1700#define IFF_EBRIDGE IFF_EBRIDGE
1701#define IFF_BONDING IFF_BONDING
1702#define IFF_ISATAP IFF_ISATAP
1703#define IFF_WAN_HDLC IFF_WAN_HDLC
1704#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1705#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1706#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1707#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1708#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1709#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1710#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1711#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1712#define IFF_TEAM_PORT IFF_TEAM_PORT
1713#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1714#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1715#define IFF_MACVLAN IFF_MACVLAN
1716#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1717#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1718#define IFF_NO_QUEUE IFF_NO_QUEUE
1719#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1720#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1721#define IFF_TEAM IFF_TEAM
1722#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1723#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1724#define IFF_MACSEC IFF_MACSEC
1725#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1726#define IFF_FAILOVER IFF_FAILOVER
1727#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1728#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1729#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1730#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1731
1732/* Specifies the type of the struct net_device::ml_priv pointer */
1733enum netdev_ml_priv_type {
1734 ML_PRIV_NONE,
1735 ML_PRIV_CAN,
1736};
1737
1738/**
1739 * struct net_device - The DEVICE structure.
1740 *
1741 * Actually, this whole structure is a big mistake. It mixes I/O
1742 * data with strictly "high-level" data, and it has to know about
1743 * almost every data structure used in the INET module.
1744 *
1745 * @name: This is the first field of the "visible" part of this structure
1746 * (i.e. as seen by users in the "Space.c" file). It is the name
1747 * of the interface.
1748 *
1749 * @name_node: Name hashlist node
1750 * @ifalias: SNMP alias
1751 * @mem_end: Shared memory end
1752 * @mem_start: Shared memory start
1753 * @base_addr: Device I/O address
1754 * @irq: Device IRQ number
1755 *
1756 * @state: Generic network queuing layer state, see netdev_state_t
1757 * @dev_list: The global list of network devices
1758 * @napi_list: List entry used for polling NAPI devices
1759 * @unreg_list: List entry when we are unregistering the
1760 * device; see the function unregister_netdev
1761 * @close_list: List entry used when we are closing the device
1762 * @ptype_all: Device-specific packet handlers for all protocols
1763 * @ptype_specific: Device-specific, protocol-specific packet handlers
1764 *
1765 * @adj_list: Directly linked devices, like slaves for bonding
1766 * @features: Currently active device features
1767 * @hw_features: User-changeable features
1768 *
1769 * @wanted_features: User-requested features
1770 * @vlan_features: Mask of features inheritable by VLAN devices
1771 *
1772 * @hw_enc_features: Mask of features inherited by encapsulating devices
1773 * This field indicates what encapsulation
1774 * offloads the hardware is capable of doing,
1775 * and drivers will need to set them appropriately.
1776 *
1777 * @mpls_features: Mask of features inheritable by MPLS
1778 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1779 *
1780 * @ifindex: interface index
1781 * @group: The group the device belongs to
1782 *
1783 * @stats: Statistics struct, which was left as a legacy, use
1784 * rtnl_link_stats64 instead
1785 *
1786 * @core_stats: core networking counters,
1787 * do not use this in drivers
1788 * @carrier_up_count: Number of times the carrier has been up
1789 * @carrier_down_count: Number of times the carrier has been down
1790 *
1791 * @wireless_handlers: List of functions to handle Wireless Extensions,
1792 * instead of ioctl,
1793 * see <net/iw_handler.h> for details.
1794 * @wireless_data: Instance data managed by the core of wireless extensions
1795 *
1796 * @netdev_ops: Includes several pointers to callbacks,
1797 * if one wants to override the ndo_*() functions
1798 * @ethtool_ops: Management operations
1799 * @l3mdev_ops: Layer 3 master device operations
1800 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1801 * discovery handling. Necessary for e.g. 6LoWPAN.
1802 * @xfrmdev_ops: Transformation offload operations
1803 * @tlsdev_ops: Transport Layer Security offload operations
1804 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1805 * of Layer 2 headers.
1806 *
1807 * @flags: Interface flags (a la BSD)
1808 * @priv_flags: Like 'flags' but invisible to userspace,
1809 * see if.h for the definitions
1810 * @gflags: Global flags ( kept as legacy )
1811 * @padded: How much padding added by alloc_netdev()
1812 * @operstate: RFC2863 operstate
1813 * @link_mode: Mapping policy to operstate
1814 * @if_port: Selectable AUI, TP, ...
1815 * @dma: DMA channel
1816 * @mtu: Interface MTU value
1817 * @min_mtu: Interface Minimum MTU value
1818 * @max_mtu: Interface Maximum MTU value
1819 * @type: Interface hardware type
1820 * @hard_header_len: Maximum hardware header length.
1821 * @min_header_len: Minimum hardware header length
1822 *
1823 * @needed_headroom: Extra headroom the hardware may need, but not in all
1824 * cases can this be guaranteed
1825 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1826 * cases can this be guaranteed. Some cases also use
1827 * LL_MAX_HEADER instead to allocate the skb
1828 *
1829 * interface address info:
1830 *
1831 * @perm_addr: Permanent hw address
1832 * @addr_assign_type: Hw address assignment type
1833 * @addr_len: Hardware address length
1834 * @upper_level: Maximum depth level of upper devices.
1835 * @lower_level: Maximum depth level of lower devices.
1836 * @neigh_priv_len: Used in neigh_alloc()
1837 * @dev_id: Used to differentiate devices that share
1838 * the same link layer address
1839 * @dev_port: Used to differentiate devices that share
1840 * the same function
1841 * @addr_list_lock: XXX: need comments on this one
1842 * @name_assign_type: network interface name assignment type
1843 * @uc_promisc: Counter that indicates promiscuous mode
1844 * has been enabled due to the need to listen to
1845 * additional unicast addresses in a device that
1846 * does not implement ndo_set_rx_mode()
1847 * @uc: unicast mac addresses
1848 * @mc: multicast mac addresses
1849 * @dev_addrs: list of device hw addresses
1850 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1851 * @promiscuity: Number of times the NIC is told to work in
1852 * promiscuous mode; if it becomes 0 the NIC will
1853 * exit promiscuous mode
1854 * @allmulti: Counter, enables or disables allmulticast mode
1855 *
1856 * @vlan_info: VLAN info
1857 * @dsa_ptr: dsa specific data
1858 * @tipc_ptr: TIPC specific data
1859 * @atalk_ptr: AppleTalk link
1860 * @ip_ptr: IPv4 specific data
1861 * @ip6_ptr: IPv6 specific data
1862 * @ax25_ptr: AX.25 specific data
1863 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1864 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1865 * device struct
1866 * @mpls_ptr: mpls_dev struct pointer
1867 * @mctp_ptr: MCTP specific data
1868 *
1869 * @dev_addr: Hw address (before bcast,
1870 * because most packets are unicast)
1871 *
1872 * @_rx: Array of RX queues
1873 * @num_rx_queues: Number of RX queues
1874 * allocated at register_netdev() time
1875 * @real_num_rx_queues: Number of RX queues currently active in device
1876 * @xdp_prog: XDP sockets filter program pointer
1877 * @gro_flush_timeout: timeout for GRO layer in NAPI
1878 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1879 * allow to avoid NIC hard IRQ, on busy queues.
1880 *
1881 * @rx_handler: handler for received packets
1882 * @rx_handler_data: XXX: need comments on this one
1883 * @miniq_ingress: ingress/clsact qdisc specific data for
1884 * ingress processing
1885 * @ingress_queue: XXX: need comments on this one
1886 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1887 * @broadcast: hw bcast address
1888 *
1889 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1890 * indexed by RX queue number. Assigned by driver.
1891 * This must only be set if the ndo_rx_flow_steer
1892 * operation is defined
1893 * @index_hlist: Device index hash chain
1894 *
1895 * @_tx: Array of TX queues
1896 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1897 * @real_num_tx_queues: Number of TX queues currently active in device
1898 * @qdisc: Root qdisc from userspace point of view
1899 * @tx_queue_len: Max frames per queue allowed
1900 * @tx_global_lock: XXX: need comments on this one
1901 * @xdp_bulkq: XDP device bulk queue
1902 * @xps_maps: all CPUs/RXQs maps for XPS device
1903 *
1904 * @xps_maps: XXX: need comments on this one
1905 * @miniq_egress: clsact qdisc specific data for
1906 * egress processing
1907 * @nf_hooks_egress: netfilter hooks executed for egress packets
1908 * @qdisc_hash: qdisc hash table
1909 * @watchdog_timeo: Represents the timeout that is used by
1910 * the watchdog (see dev_watchdog())
1911 * @watchdog_timer: List of timers
1912 *
1913 * @proto_down_reason: reason a netdev interface is held down
1914 * @pcpu_refcnt: Number of references to this device
1915 * @dev_refcnt: Number of references to this device
1916 * @refcnt_tracker: Tracker directory for tracked references to this device
1917 * @todo_list: Delayed register/unregister
1918 * @link_watch_list: XXX: need comments on this one
1919 *
1920 * @reg_state: Register/unregister state machine
1921 * @dismantle: Device is going to be freed
1922 * @rtnl_link_state: This enum represents the phases of creating
1923 * a new link
1924 *
1925 * @needs_free_netdev: Should unregister perform free_netdev?
1926 * @priv_destructor: Called from unregister
1927 * @npinfo: XXX: need comments on this one
1928 * @nd_net: Network namespace this network device is inside
1929 *
1930 * @ml_priv: Mid-layer private
1931 * @ml_priv_type: Mid-layer private type
1932 * @lstats: Loopback statistics
1933 * @tstats: Tunnel statistics
1934 * @dstats: Dummy statistics
1935 * @vstats: Virtual ethernet statistics
1936 *
1937 * @garp_port: GARP
1938 * @mrp_port: MRP
1939 *
1940 * @dm_private: Drop monitor private
1941 *
1942 * @dev: Class/net/name entry
1943 * @sysfs_groups: Space for optional device, statistics and wireless
1944 * sysfs groups
1945 *
1946 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1947 * @rtnl_link_ops: Rtnl_link_ops
1948 *
1949 * @gso_max_size: Maximum size of generic segmentation offload
1950 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1951 * @gso_max_segs: Maximum number of segments that can be passed to the
1952 * NIC for GSO
1953 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1954 *
1955 * @dcbnl_ops: Data Center Bridging netlink ops
1956 * @num_tc: Number of traffic classes in the net device
1957 * @tc_to_txq: XXX: need comments on this one
1958 * @prio_tc_map: XXX: need comments on this one
1959 *
1960 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1961 *
1962 * @priomap: XXX: need comments on this one
1963 * @phydev: Physical device may attach itself
1964 * for hardware timestamping
1965 * @sfp_bus: attached &struct sfp_bus structure.
1966 *
1967 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1968 *
1969 * @proto_down: protocol port state information can be sent to the
1970 * switch driver and used to set the phys state of the
1971 * switch port.
1972 *
1973 * @wol_enabled: Wake-on-LAN is enabled
1974 *
1975 * @threaded: napi threaded mode is enabled
1976 *
1977 * @net_notifier_list: List of per-net netdev notifier block
1978 * that follow this device when it is moved
1979 * to another network namespace.
1980 *
1981 * @macsec_ops: MACsec offloading ops
1982 *
1983 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1984 * offload capabilities of the device
1985 * @udp_tunnel_nic: UDP tunnel offload state
1986 * @xdp_state: stores info on attached XDP BPF programs
1987 *
1988 * @nested_level: Used as a parameter of spin_lock_nested() of
1989 * dev->addr_list_lock.
1990 * @unlink_list: As netif_addr_lock() can be called recursively,
1991 * keep a list of interfaces to be deleted.
1992 * @gro_max_size: Maximum size of aggregated packet in generic
1993 * receive offload (GRO)
1994 *
1995 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
1996 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
1997 * @watchdog_dev_tracker: refcount tracker used by watchdog.
1998 * @dev_registered_tracker: tracker for reference held while
1999 * registered
2000 * @offload_xstats_l3: L3 HW stats for this netdevice.
2001 *
2002 * FIXME: cleanup struct net_device such that network protocol info
2003 * moves out.
2004 */
2005
2006struct net_device {
2007 char name[IFNAMSIZ];
2008 struct netdev_name_node *name_node;
2009 struct dev_ifalias __rcu *ifalias;
2010 /*
2011 * I/O specific fields
2012 * FIXME: Merge these and struct ifmap into one
2013 */
2014 unsigned long mem_end;
2015 unsigned long mem_start;
2016 unsigned long base_addr;
2017
2018 /*
2019 * Some hardware also needs these fields (state,dev_list,
2020 * napi_list,unreg_list,close_list) but they are not
2021 * part of the usual set specified in Space.c.
2022 */
2023
2024 unsigned long state;
2025
2026 struct list_head dev_list;
2027 struct list_head napi_list;
2028 struct list_head unreg_list;
2029 struct list_head close_list;
2030 struct list_head ptype_all;
2031 struct list_head ptype_specific;
2032
2033 struct {
2034 struct list_head upper;
2035 struct list_head lower;
2036 } adj_list;
2037
2038 /* Read-mostly cache-line for fast-path access */
2039 unsigned int flags;
2040 unsigned long long priv_flags;
2041 const struct net_device_ops *netdev_ops;
2042 int ifindex;
2043 unsigned short gflags;
2044 unsigned short hard_header_len;
2045
2046 /* Note : dev->mtu is often read without holding a lock.
2047 * Writers usually hold RTNL.
2048 * It is recommended to use READ_ONCE() to annotate the reads,
2049 * and to use WRITE_ONCE() to annotate the writes.
2050 */
2051 unsigned int mtu;
2052 unsigned short needed_headroom;
2053 unsigned short needed_tailroom;
2054
2055 netdev_features_t features;
2056 netdev_features_t hw_features;
2057 netdev_features_t wanted_features;
2058 netdev_features_t vlan_features;
2059 netdev_features_t hw_enc_features;
2060 netdev_features_t mpls_features;
2061 netdev_features_t gso_partial_features;
2062
2063 unsigned int min_mtu;
2064 unsigned int max_mtu;
2065 unsigned short type;
2066 unsigned char min_header_len;
2067 unsigned char name_assign_type;
2068
2069 int group;
2070
2071 struct net_device_stats stats; /* not used by modern drivers */
2072
2073 struct net_device_core_stats __percpu *core_stats;
2074
2075 /* Stats to monitor link on/off, flapping */
2076 atomic_t carrier_up_count;
2077 atomic_t carrier_down_count;
2078
2079#ifdef CONFIG_WIRELESS_EXT
2080 const struct iw_handler_def *wireless_handlers;
2081 struct iw_public_data *wireless_data;
2082#endif
2083 const struct ethtool_ops *ethtool_ops;
2084#ifdef CONFIG_NET_L3_MASTER_DEV
2085 const struct l3mdev_ops *l3mdev_ops;
2086#endif
2087#if IS_ENABLED(CONFIG_IPV6)
2088 const struct ndisc_ops *ndisc_ops;
2089#endif
2090
2091#ifdef CONFIG_XFRM_OFFLOAD
2092 const struct xfrmdev_ops *xfrmdev_ops;
2093#endif
2094
2095#if IS_ENABLED(CONFIG_TLS_DEVICE)
2096 const struct tlsdev_ops *tlsdev_ops;
2097#endif
2098
2099 const struct header_ops *header_ops;
2100
2101 unsigned char operstate;
2102 unsigned char link_mode;
2103
2104 unsigned char if_port;
2105 unsigned char dma;
2106
2107 /* Interface address info. */
2108 unsigned char perm_addr[MAX_ADDR_LEN];
2109 unsigned char addr_assign_type;
2110 unsigned char addr_len;
2111 unsigned char upper_level;
2112 unsigned char lower_level;
2113
2114 unsigned short neigh_priv_len;
2115 unsigned short dev_id;
2116 unsigned short dev_port;
2117 unsigned short padded;
2118
2119 spinlock_t addr_list_lock;
2120 int irq;
2121
2122 struct netdev_hw_addr_list uc;
2123 struct netdev_hw_addr_list mc;
2124 struct netdev_hw_addr_list dev_addrs;
2125
2126#ifdef CONFIG_SYSFS
2127 struct kset *queues_kset;
2128#endif
2129#ifdef CONFIG_LOCKDEP
2130 struct list_head unlink_list;
2131#endif
2132 unsigned int promiscuity;
2133 unsigned int allmulti;
2134 bool uc_promisc;
2135#ifdef CONFIG_LOCKDEP
2136 unsigned char nested_level;
2137#endif
2138
2139
2140 /* Protocol-specific pointers */
2141
2142 struct in_device __rcu *ip_ptr;
2143 struct inet6_dev __rcu *ip6_ptr;
2144#if IS_ENABLED(CONFIG_VLAN_8021Q)
2145 struct vlan_info __rcu *vlan_info;
2146#endif
2147#if IS_ENABLED(CONFIG_NET_DSA)
2148 struct dsa_port *dsa_ptr;
2149#endif
2150#if IS_ENABLED(CONFIG_TIPC)
2151 struct tipc_bearer __rcu *tipc_ptr;
2152#endif
2153#if IS_ENABLED(CONFIG_ATALK)
2154 void *atalk_ptr;
2155#endif
2156#if IS_ENABLED(CONFIG_AX25)
2157 void *ax25_ptr;
2158#endif
2159#if IS_ENABLED(CONFIG_CFG80211)
2160 struct wireless_dev *ieee80211_ptr;
2161#endif
2162#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2163 struct wpan_dev *ieee802154_ptr;
2164#endif
2165#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2166 struct mpls_dev __rcu *mpls_ptr;
2167#endif
2168#if IS_ENABLED(CONFIG_MCTP)
2169 struct mctp_dev __rcu *mctp_ptr;
2170#endif
2171
2172/*
2173 * Cache lines mostly used on receive path (including eth_type_trans())
2174 */
2175 /* Interface address info used in eth_type_trans() */
2176 const unsigned char *dev_addr;
2177
2178 struct netdev_rx_queue *_rx;
2179 unsigned int num_rx_queues;
2180 unsigned int real_num_rx_queues;
2181
2182 struct bpf_prog __rcu *xdp_prog;
2183 unsigned long gro_flush_timeout;
2184 int napi_defer_hard_irqs;
2185#define GRO_LEGACY_MAX_SIZE 65536u
2186/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2187 * and shinfo->gso_segs is a 16bit field.
2188 */
2189#define GRO_MAX_SIZE (8 * 65535u)
2190 unsigned int gro_max_size;
2191 rx_handler_func_t __rcu *rx_handler;
2192 void __rcu *rx_handler_data;
2193
2194#ifdef CONFIG_NET_CLS_ACT
2195 struct mini_Qdisc __rcu *miniq_ingress;
2196#endif
2197 struct netdev_queue __rcu *ingress_queue;
2198#ifdef CONFIG_NETFILTER_INGRESS
2199 struct nf_hook_entries __rcu *nf_hooks_ingress;
2200#endif
2201
2202 unsigned char broadcast[MAX_ADDR_LEN];
2203#ifdef CONFIG_RFS_ACCEL
2204 struct cpu_rmap *rx_cpu_rmap;
2205#endif
2206 struct hlist_node index_hlist;
2207
2208/*
2209 * Cache lines mostly used on transmit path
2210 */
2211 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2212 unsigned int num_tx_queues;
2213 unsigned int real_num_tx_queues;
2214 struct Qdisc __rcu *qdisc;
2215 unsigned int tx_queue_len;
2216 spinlock_t tx_global_lock;
2217
2218 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2219
2220#ifdef CONFIG_XPS
2221 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2222#endif
2223#ifdef CONFIG_NET_CLS_ACT
2224 struct mini_Qdisc __rcu *miniq_egress;
2225#endif
2226#ifdef CONFIG_NETFILTER_EGRESS
2227 struct nf_hook_entries __rcu *nf_hooks_egress;
2228#endif
2229
2230#ifdef CONFIG_NET_SCHED
2231 DECLARE_HASHTABLE (qdisc_hash, 4);
2232#endif
2233 /* These may be needed for future network-power-down code. */
2234 struct timer_list watchdog_timer;
2235 int watchdog_timeo;
2236
2237 u32 proto_down_reason;
2238
2239 struct list_head todo_list;
2240
2241#ifdef CONFIG_PCPU_DEV_REFCNT
2242 int __percpu *pcpu_refcnt;
2243#else
2244 refcount_t dev_refcnt;
2245#endif
2246 struct ref_tracker_dir refcnt_tracker;
2247
2248 struct list_head link_watch_list;
2249
2250 enum { NETREG_UNINITIALIZED=0,
2251 NETREG_REGISTERED, /* completed register_netdevice */
2252 NETREG_UNREGISTERING, /* called unregister_netdevice */
2253 NETREG_UNREGISTERED, /* completed unregister todo */
2254 NETREG_RELEASED, /* called free_netdev */
2255 NETREG_DUMMY, /* dummy device for NAPI poll */
2256 } reg_state:8;
2257
2258 bool dismantle;
2259
2260 enum {
2261 RTNL_LINK_INITIALIZED,
2262 RTNL_LINK_INITIALIZING,
2263 } rtnl_link_state:16;
2264
2265 bool needs_free_netdev;
2266 void (*priv_destructor)(struct net_device *dev);
2267
2268#ifdef CONFIG_NETPOLL
2269 struct netpoll_info __rcu *npinfo;
2270#endif
2271
2272 possible_net_t nd_net;
2273
2274 /* mid-layer private */
2275 void *ml_priv;
2276 enum netdev_ml_priv_type ml_priv_type;
2277
2278 union {
2279 struct pcpu_lstats __percpu *lstats;
2280 struct pcpu_sw_netstats __percpu *tstats;
2281 struct pcpu_dstats __percpu *dstats;
2282 };
2283
2284#if IS_ENABLED(CONFIG_GARP)
2285 struct garp_port __rcu *garp_port;
2286#endif
2287#if IS_ENABLED(CONFIG_MRP)
2288 struct mrp_port __rcu *mrp_port;
2289#endif
2290#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2291 struct dm_hw_stat_delta __rcu *dm_private;
2292#endif
2293 struct device dev;
2294 const struct attribute_group *sysfs_groups[4];
2295 const struct attribute_group *sysfs_rx_queue_group;
2296
2297 const struct rtnl_link_ops *rtnl_link_ops;
2298
2299 /* for setting kernel sock attribute on TCP connection setup */
2300#define GSO_MAX_SEGS 65535u
2301#define GSO_LEGACY_MAX_SIZE 65536u
2302/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2303 * and shinfo->gso_segs is a 16bit field.
2304 */
2305#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2306
2307 unsigned int gso_max_size;
2308#define TSO_LEGACY_MAX_SIZE 65536
2309#define TSO_MAX_SIZE UINT_MAX
2310 unsigned int tso_max_size;
2311 u16 gso_max_segs;
2312#define TSO_MAX_SEGS U16_MAX
2313 u16 tso_max_segs;
2314
2315#ifdef CONFIG_DCB
2316 const struct dcbnl_rtnl_ops *dcbnl_ops;
2317#endif
2318 s16 num_tc;
2319 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2320 u8 prio_tc_map[TC_BITMASK + 1];
2321
2322#if IS_ENABLED(CONFIG_FCOE)
2323 unsigned int fcoe_ddp_xid;
2324#endif
2325#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2326 struct netprio_map __rcu *priomap;
2327#endif
2328 struct phy_device *phydev;
2329 struct sfp_bus *sfp_bus;
2330 struct lock_class_key *qdisc_tx_busylock;
2331 bool proto_down;
2332 unsigned wol_enabled:1;
2333 unsigned threaded:1;
2334
2335 struct list_head net_notifier_list;
2336
2337#if IS_ENABLED(CONFIG_MACSEC)
2338 /* MACsec management functions */
2339 const struct macsec_ops *macsec_ops;
2340#endif
2341 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2342 struct udp_tunnel_nic *udp_tunnel_nic;
2343
2344 /* protected by rtnl_lock */
2345 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2346
2347 u8 dev_addr_shadow[MAX_ADDR_LEN];
2348 netdevice_tracker linkwatch_dev_tracker;
2349 netdevice_tracker watchdog_dev_tracker;
2350 netdevice_tracker dev_registered_tracker;
2351 struct rtnl_hw_stats64 *offload_xstats_l3;
2352};
2353#define to_net_dev(d) container_of(d, struct net_device, dev)
2354
2355static inline bool netif_elide_gro(const struct net_device *dev)
2356{
2357 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2358 return true;
2359 return false;
2360}
2361
2362#define NETDEV_ALIGN 32
2363
2364static inline
2365int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2366{
2367 return dev->prio_tc_map[prio & TC_BITMASK];
2368}
2369
2370static inline
2371int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2372{
2373 if (tc >= dev->num_tc)
2374 return -EINVAL;
2375
2376 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2377 return 0;
2378}
2379
2380int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2381void netdev_reset_tc(struct net_device *dev);
2382int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2383int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2384
2385static inline
2386int netdev_get_num_tc(struct net_device *dev)
2387{
2388 return dev->num_tc;
2389}
2390
2391static inline void net_prefetch(void *p)
2392{
2393 prefetch(p);
2394#if L1_CACHE_BYTES < 128
2395 prefetch((u8 *)p + L1_CACHE_BYTES);
2396#endif
2397}
2398
2399static inline void net_prefetchw(void *p)
2400{
2401 prefetchw(p);
2402#if L1_CACHE_BYTES < 128
2403 prefetchw((u8 *)p + L1_CACHE_BYTES);
2404#endif
2405}
2406
2407void netdev_unbind_sb_channel(struct net_device *dev,
2408 struct net_device *sb_dev);
2409int netdev_bind_sb_channel_queue(struct net_device *dev,
2410 struct net_device *sb_dev,
2411 u8 tc, u16 count, u16 offset);
2412int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2413static inline int netdev_get_sb_channel(struct net_device *dev)
2414{
2415 return max_t(int, -dev->num_tc, 0);
2416}
2417
2418static inline
2419struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2420 unsigned int index)
2421{
2422 return &dev->_tx[index];
2423}
2424
2425static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2426 const struct sk_buff *skb)
2427{
2428 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2429}
2430
2431static inline void netdev_for_each_tx_queue(struct net_device *dev,
2432 void (*f)(struct net_device *,
2433 struct netdev_queue *,
2434 void *),
2435 void *arg)
2436{
2437 unsigned int i;
2438
2439 for (i = 0; i < dev->num_tx_queues; i++)
2440 f(dev, &dev->_tx[i], arg);
2441}
2442
2443#define netdev_lockdep_set_classes(dev) \
2444{ \
2445 static struct lock_class_key qdisc_tx_busylock_key; \
2446 static struct lock_class_key qdisc_xmit_lock_key; \
2447 static struct lock_class_key dev_addr_list_lock_key; \
2448 unsigned int i; \
2449 \
2450 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2451 lockdep_set_class(&(dev)->addr_list_lock, \
2452 &dev_addr_list_lock_key); \
2453 for (i = 0; i < (dev)->num_tx_queues; i++) \
2454 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2455 &qdisc_xmit_lock_key); \
2456}
2457
2458u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2459 struct net_device *sb_dev);
2460struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2461 struct sk_buff *skb,
2462 struct net_device *sb_dev);
2463
2464/* returns the headroom that the master device needs to take in account
2465 * when forwarding to this dev
2466 */
2467static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2468{
2469 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2470}
2471
2472static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2473{
2474 if (dev->netdev_ops->ndo_set_rx_headroom)
2475 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2476}
2477
2478/* set the device rx headroom to the dev's default */
2479static inline void netdev_reset_rx_headroom(struct net_device *dev)
2480{
2481 netdev_set_rx_headroom(dev, -1);
2482}
2483
2484static inline void *netdev_get_ml_priv(struct net_device *dev,
2485 enum netdev_ml_priv_type type)
2486{
2487 if (dev->ml_priv_type != type)
2488 return NULL;
2489
2490 return dev->ml_priv;
2491}
2492
2493static inline void netdev_set_ml_priv(struct net_device *dev,
2494 void *ml_priv,
2495 enum netdev_ml_priv_type type)
2496{
2497 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2498 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2499 dev->ml_priv_type, type);
2500 WARN(!dev->ml_priv_type && dev->ml_priv,
2501 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2502
2503 dev->ml_priv = ml_priv;
2504 dev->ml_priv_type = type;
2505}
2506
2507/*
2508 * Net namespace inlines
2509 */
2510static inline
2511struct net *dev_net(const struct net_device *dev)
2512{
2513 return read_pnet(&dev->nd_net);
2514}
2515
2516static inline
2517void dev_net_set(struct net_device *dev, struct net *net)
2518{
2519 write_pnet(&dev->nd_net, net);
2520}
2521
2522/**
2523 * netdev_priv - access network device private data
2524 * @dev: network device
2525 *
2526 * Get network device private data
2527 */
2528static inline void *netdev_priv(const struct net_device *dev)
2529{
2530 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2531}
2532
2533/* Set the sysfs physical device reference for the network logical device
2534 * if set prior to registration will cause a symlink during initialization.
2535 */
2536#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2537
2538/* Set the sysfs device type for the network logical device to allow
2539 * fine-grained identification of different network device types. For
2540 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2541 */
2542#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2543
2544/* Default NAPI poll() weight
2545 * Device drivers are strongly advised to not use bigger value
2546 */
2547#define NAPI_POLL_WEIGHT 64
2548
2549void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2550 int (*poll)(struct napi_struct *, int), int weight);
2551
2552/**
2553 * netif_napi_add() - initialize a NAPI context
2554 * @dev: network device
2555 * @napi: NAPI context
2556 * @poll: polling function
2557 *
2558 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2559 * *any* of the other NAPI-related functions.
2560 */
2561static inline void
2562netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2563 int (*poll)(struct napi_struct *, int))
2564{
2565 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2566}
2567
2568static inline void
2569netif_napi_add_tx_weight(struct net_device *dev,
2570 struct napi_struct *napi,
2571 int (*poll)(struct napi_struct *, int),
2572 int weight)
2573{
2574 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2575 netif_napi_add_weight(dev, napi, poll, weight);
2576}
2577
2578/**
2579 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2580 * @dev: network device
2581 * @napi: NAPI context
2582 * @poll: polling function
2583 *
2584 * This variant of netif_napi_add() should be used from drivers using NAPI
2585 * to exclusively poll a TX queue.
2586 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2587 */
2588static inline void netif_napi_add_tx(struct net_device *dev,
2589 struct napi_struct *napi,
2590 int (*poll)(struct napi_struct *, int))
2591{
2592 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2593}
2594
2595/**
2596 * __netif_napi_del - remove a NAPI context
2597 * @napi: NAPI context
2598 *
2599 * Warning: caller must observe RCU grace period before freeing memory
2600 * containing @napi. Drivers might want to call this helper to combine
2601 * all the needed RCU grace periods into a single one.
2602 */
2603void __netif_napi_del(struct napi_struct *napi);
2604
2605/**
2606 * netif_napi_del - remove a NAPI context
2607 * @napi: NAPI context
2608 *
2609 * netif_napi_del() removes a NAPI context from the network device NAPI list
2610 */
2611static inline void netif_napi_del(struct napi_struct *napi)
2612{
2613 __netif_napi_del(napi);
2614 synchronize_net();
2615}
2616
2617struct packet_type {
2618 __be16 type; /* This is really htons(ether_type). */
2619 bool ignore_outgoing;
2620 struct net_device *dev; /* NULL is wildcarded here */
2621 netdevice_tracker dev_tracker;
2622 int (*func) (struct sk_buff *,
2623 struct net_device *,
2624 struct packet_type *,
2625 struct net_device *);
2626 void (*list_func) (struct list_head *,
2627 struct packet_type *,
2628 struct net_device *);
2629 bool (*id_match)(struct packet_type *ptype,
2630 struct sock *sk);
2631 struct net *af_packet_net;
2632 void *af_packet_priv;
2633 struct list_head list;
2634};
2635
2636struct offload_callbacks {
2637 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2638 netdev_features_t features);
2639 struct sk_buff *(*gro_receive)(struct list_head *head,
2640 struct sk_buff *skb);
2641 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2642};
2643
2644struct packet_offload {
2645 __be16 type; /* This is really htons(ether_type). */
2646 u16 priority;
2647 struct offload_callbacks callbacks;
2648 struct list_head list;
2649};
2650
2651/* often modified stats are per-CPU, other are shared (netdev->stats) */
2652struct pcpu_sw_netstats {
2653 u64_stats_t rx_packets;
2654 u64_stats_t rx_bytes;
2655 u64_stats_t tx_packets;
2656 u64_stats_t tx_bytes;
2657 struct u64_stats_sync syncp;
2658} __aligned(4 * sizeof(u64));
2659
2660struct pcpu_lstats {
2661 u64_stats_t packets;
2662 u64_stats_t bytes;
2663 struct u64_stats_sync syncp;
2664} __aligned(2 * sizeof(u64));
2665
2666void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2667
2668static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2669{
2670 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2671
2672 u64_stats_update_begin(&tstats->syncp);
2673 u64_stats_add(&tstats->rx_bytes, len);
2674 u64_stats_inc(&tstats->rx_packets);
2675 u64_stats_update_end(&tstats->syncp);
2676}
2677
2678static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2679 unsigned int packets,
2680 unsigned int len)
2681{
2682 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2683
2684 u64_stats_update_begin(&tstats->syncp);
2685 u64_stats_add(&tstats->tx_bytes, len);
2686 u64_stats_add(&tstats->tx_packets, packets);
2687 u64_stats_update_end(&tstats->syncp);
2688}
2689
2690static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2691{
2692 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2693
2694 u64_stats_update_begin(&lstats->syncp);
2695 u64_stats_add(&lstats->bytes, len);
2696 u64_stats_inc(&lstats->packets);
2697 u64_stats_update_end(&lstats->syncp);
2698}
2699
2700#define __netdev_alloc_pcpu_stats(type, gfp) \
2701({ \
2702 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2703 if (pcpu_stats) { \
2704 int __cpu; \
2705 for_each_possible_cpu(__cpu) { \
2706 typeof(type) *stat; \
2707 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2708 u64_stats_init(&stat->syncp); \
2709 } \
2710 } \
2711 pcpu_stats; \
2712})
2713
2714#define netdev_alloc_pcpu_stats(type) \
2715 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2716
2717#define devm_netdev_alloc_pcpu_stats(dev, type) \
2718({ \
2719 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2720 if (pcpu_stats) { \
2721 int __cpu; \
2722 for_each_possible_cpu(__cpu) { \
2723 typeof(type) *stat; \
2724 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2725 u64_stats_init(&stat->syncp); \
2726 } \
2727 } \
2728 pcpu_stats; \
2729})
2730
2731enum netdev_lag_tx_type {
2732 NETDEV_LAG_TX_TYPE_UNKNOWN,
2733 NETDEV_LAG_TX_TYPE_RANDOM,
2734 NETDEV_LAG_TX_TYPE_BROADCAST,
2735 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2736 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2737 NETDEV_LAG_TX_TYPE_HASH,
2738};
2739
2740enum netdev_lag_hash {
2741 NETDEV_LAG_HASH_NONE,
2742 NETDEV_LAG_HASH_L2,
2743 NETDEV_LAG_HASH_L34,
2744 NETDEV_LAG_HASH_L23,
2745 NETDEV_LAG_HASH_E23,
2746 NETDEV_LAG_HASH_E34,
2747 NETDEV_LAG_HASH_VLAN_SRCMAC,
2748 NETDEV_LAG_HASH_UNKNOWN,
2749};
2750
2751struct netdev_lag_upper_info {
2752 enum netdev_lag_tx_type tx_type;
2753 enum netdev_lag_hash hash_type;
2754};
2755
2756struct netdev_lag_lower_state_info {
2757 u8 link_up : 1,
2758 tx_enabled : 1;
2759};
2760
2761#include <linux/notifier.h>
2762
2763/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2764 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2765 * adding new types.
2766 */
2767enum netdev_cmd {
2768 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2769 NETDEV_DOWN,
2770 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2771 detected a hardware crash and restarted
2772 - we can use this eg to kick tcp sessions
2773 once done */
2774 NETDEV_CHANGE, /* Notify device state change */
2775 NETDEV_REGISTER,
2776 NETDEV_UNREGISTER,
2777 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2778 NETDEV_CHANGEADDR, /* notify after the address change */
2779 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2780 NETDEV_GOING_DOWN,
2781 NETDEV_CHANGENAME,
2782 NETDEV_FEAT_CHANGE,
2783 NETDEV_BONDING_FAILOVER,
2784 NETDEV_PRE_UP,
2785 NETDEV_PRE_TYPE_CHANGE,
2786 NETDEV_POST_TYPE_CHANGE,
2787 NETDEV_POST_INIT,
2788 NETDEV_RELEASE,
2789 NETDEV_NOTIFY_PEERS,
2790 NETDEV_JOIN,
2791 NETDEV_CHANGEUPPER,
2792 NETDEV_RESEND_IGMP,
2793 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2794 NETDEV_CHANGEINFODATA,
2795 NETDEV_BONDING_INFO,
2796 NETDEV_PRECHANGEUPPER,
2797 NETDEV_CHANGELOWERSTATE,
2798 NETDEV_UDP_TUNNEL_PUSH_INFO,
2799 NETDEV_UDP_TUNNEL_DROP_INFO,
2800 NETDEV_CHANGE_TX_QUEUE_LEN,
2801 NETDEV_CVLAN_FILTER_PUSH_INFO,
2802 NETDEV_CVLAN_FILTER_DROP_INFO,
2803 NETDEV_SVLAN_FILTER_PUSH_INFO,
2804 NETDEV_SVLAN_FILTER_DROP_INFO,
2805 NETDEV_OFFLOAD_XSTATS_ENABLE,
2806 NETDEV_OFFLOAD_XSTATS_DISABLE,
2807 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2808 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2809};
2810const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2811
2812int register_netdevice_notifier(struct notifier_block *nb);
2813int unregister_netdevice_notifier(struct notifier_block *nb);
2814int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2815int unregister_netdevice_notifier_net(struct net *net,
2816 struct notifier_block *nb);
2817int register_netdevice_notifier_dev_net(struct net_device *dev,
2818 struct notifier_block *nb,
2819 struct netdev_net_notifier *nn);
2820int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2821 struct notifier_block *nb,
2822 struct netdev_net_notifier *nn);
2823
2824struct netdev_notifier_info {
2825 struct net_device *dev;
2826 struct netlink_ext_ack *extack;
2827};
2828
2829struct netdev_notifier_info_ext {
2830 struct netdev_notifier_info info; /* must be first */
2831 union {
2832 u32 mtu;
2833 } ext;
2834};
2835
2836struct netdev_notifier_change_info {
2837 struct netdev_notifier_info info; /* must be first */
2838 unsigned int flags_changed;
2839};
2840
2841struct netdev_notifier_changeupper_info {
2842 struct netdev_notifier_info info; /* must be first */
2843 struct net_device *upper_dev; /* new upper dev */
2844 bool master; /* is upper dev master */
2845 bool linking; /* is the notification for link or unlink */
2846 void *upper_info; /* upper dev info */
2847};
2848
2849struct netdev_notifier_changelowerstate_info {
2850 struct netdev_notifier_info info; /* must be first */
2851 void *lower_state_info; /* is lower dev state */
2852};
2853
2854struct netdev_notifier_pre_changeaddr_info {
2855 struct netdev_notifier_info info; /* must be first */
2856 const unsigned char *dev_addr;
2857};
2858
2859enum netdev_offload_xstats_type {
2860 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2861};
2862
2863struct netdev_notifier_offload_xstats_info {
2864 struct netdev_notifier_info info; /* must be first */
2865 enum netdev_offload_xstats_type type;
2866
2867 union {
2868 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2869 struct netdev_notifier_offload_xstats_rd *report_delta;
2870 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2871 struct netdev_notifier_offload_xstats_ru *report_used;
2872 };
2873};
2874
2875int netdev_offload_xstats_enable(struct net_device *dev,
2876 enum netdev_offload_xstats_type type,
2877 struct netlink_ext_ack *extack);
2878int netdev_offload_xstats_disable(struct net_device *dev,
2879 enum netdev_offload_xstats_type type);
2880bool netdev_offload_xstats_enabled(const struct net_device *dev,
2881 enum netdev_offload_xstats_type type);
2882int netdev_offload_xstats_get(struct net_device *dev,
2883 enum netdev_offload_xstats_type type,
2884 struct rtnl_hw_stats64 *stats, bool *used,
2885 struct netlink_ext_ack *extack);
2886void
2887netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2888 const struct rtnl_hw_stats64 *stats);
2889void
2890netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2891void netdev_offload_xstats_push_delta(struct net_device *dev,
2892 enum netdev_offload_xstats_type type,
2893 const struct rtnl_hw_stats64 *stats);
2894
2895static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2896 struct net_device *dev)
2897{
2898 info->dev = dev;
2899 info->extack = NULL;
2900}
2901
2902static inline struct net_device *
2903netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2904{
2905 return info->dev;
2906}
2907
2908static inline struct netlink_ext_ack *
2909netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2910{
2911 return info->extack;
2912}
2913
2914int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2915
2916
2917extern rwlock_t dev_base_lock; /* Device list lock */
2918
2919#define for_each_netdev(net, d) \
2920 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2921#define for_each_netdev_reverse(net, d) \
2922 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2923#define for_each_netdev_rcu(net, d) \
2924 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2925#define for_each_netdev_safe(net, d, n) \
2926 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2927#define for_each_netdev_continue(net, d) \
2928 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2929#define for_each_netdev_continue_reverse(net, d) \
2930 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2931 dev_list)
2932#define for_each_netdev_continue_rcu(net, d) \
2933 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2934#define for_each_netdev_in_bond_rcu(bond, slave) \
2935 for_each_netdev_rcu(&init_net, slave) \
2936 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2937#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2938
2939static inline struct net_device *next_net_device(struct net_device *dev)
2940{
2941 struct list_head *lh;
2942 struct net *net;
2943
2944 net = dev_net(dev);
2945 lh = dev->dev_list.next;
2946 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2947}
2948
2949static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2950{
2951 struct list_head *lh;
2952 struct net *net;
2953
2954 net = dev_net(dev);
2955 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2956 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2957}
2958
2959static inline struct net_device *first_net_device(struct net *net)
2960{
2961 return list_empty(&net->dev_base_head) ? NULL :
2962 net_device_entry(net->dev_base_head.next);
2963}
2964
2965static inline struct net_device *first_net_device_rcu(struct net *net)
2966{
2967 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2968
2969 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2970}
2971
2972int netdev_boot_setup_check(struct net_device *dev);
2973struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2974 const char *hwaddr);
2975struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2976void dev_add_pack(struct packet_type *pt);
2977void dev_remove_pack(struct packet_type *pt);
2978void __dev_remove_pack(struct packet_type *pt);
2979void dev_add_offload(struct packet_offload *po);
2980void dev_remove_offload(struct packet_offload *po);
2981
2982int dev_get_iflink(const struct net_device *dev);
2983int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2984int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2985 struct net_device_path_stack *stack);
2986struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2987 unsigned short mask);
2988struct net_device *dev_get_by_name(struct net *net, const char *name);
2989struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2990struct net_device *__dev_get_by_name(struct net *net, const char *name);
2991bool netdev_name_in_use(struct net *net, const char *name);
2992int dev_alloc_name(struct net_device *dev, const char *name);
2993int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2994void dev_close(struct net_device *dev);
2995void dev_close_many(struct list_head *head, bool unlink);
2996void dev_disable_lro(struct net_device *dev);
2997int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2998u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2999 struct net_device *sb_dev);
3000u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3001 struct net_device *sb_dev);
3002
3003int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3004int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3005
3006static inline int dev_queue_xmit(struct sk_buff *skb)
3007{
3008 return __dev_queue_xmit(skb, NULL);
3009}
3010
3011static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3012 struct net_device *sb_dev)
3013{
3014 return __dev_queue_xmit(skb, sb_dev);
3015}
3016
3017static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3018{
3019 int ret;
3020
3021 ret = __dev_direct_xmit(skb, queue_id);
3022 if (!dev_xmit_complete(ret))
3023 kfree_skb(skb);
3024 return ret;
3025}
3026
3027int register_netdevice(struct net_device *dev);
3028void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3029void unregister_netdevice_many(struct list_head *head);
3030static inline void unregister_netdevice(struct net_device *dev)
3031{
3032 unregister_netdevice_queue(dev, NULL);
3033}
3034
3035int netdev_refcnt_read(const struct net_device *dev);
3036void free_netdev(struct net_device *dev);
3037void netdev_freemem(struct net_device *dev);
3038int init_dummy_netdev(struct net_device *dev);
3039
3040struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3041 struct sk_buff *skb,
3042 bool all_slaves);
3043struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3044 struct sock *sk);
3045struct net_device *dev_get_by_index(struct net *net, int ifindex);
3046struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3047struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3048struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3049int dev_restart(struct net_device *dev);
3050
3051
3052static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3053 unsigned short type,
3054 const void *daddr, const void *saddr,
3055 unsigned int len)
3056{
3057 if (!dev->header_ops || !dev->header_ops->create)
3058 return 0;
3059
3060 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3061}
3062
3063static inline int dev_parse_header(const struct sk_buff *skb,
3064 unsigned char *haddr)
3065{
3066 const struct net_device *dev = skb->dev;
3067
3068 if (!dev->header_ops || !dev->header_ops->parse)
3069 return 0;
3070 return dev->header_ops->parse(skb, haddr);
3071}
3072
3073static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3074{
3075 const struct net_device *dev = skb->dev;
3076
3077 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3078 return 0;
3079 return dev->header_ops->parse_protocol(skb);
3080}
3081
3082/* ll_header must have at least hard_header_len allocated */
3083static inline bool dev_validate_header(const struct net_device *dev,
3084 char *ll_header, int len)
3085{
3086 if (likely(len >= dev->hard_header_len))
3087 return true;
3088 if (len < dev->min_header_len)
3089 return false;
3090
3091 if (capable(CAP_SYS_RAWIO)) {
3092 memset(ll_header + len, 0, dev->hard_header_len - len);
3093 return true;
3094 }
3095
3096 if (dev->header_ops && dev->header_ops->validate)
3097 return dev->header_ops->validate(ll_header, len);
3098
3099 return false;
3100}
3101
3102static inline bool dev_has_header(const struct net_device *dev)
3103{
3104 return dev->header_ops && dev->header_ops->create;
3105}
3106
3107/*
3108 * Incoming packets are placed on per-CPU queues
3109 */
3110struct softnet_data {
3111 struct list_head poll_list;
3112 struct sk_buff_head process_queue;
3113
3114 /* stats */
3115 unsigned int processed;
3116 unsigned int time_squeeze;
3117 unsigned int received_rps;
3118#ifdef CONFIG_RPS
3119 struct softnet_data *rps_ipi_list;
3120#endif
3121#ifdef CONFIG_NET_FLOW_LIMIT
3122 struct sd_flow_limit __rcu *flow_limit;
3123#endif
3124 struct Qdisc *output_queue;
3125 struct Qdisc **output_queue_tailp;
3126 struct sk_buff *completion_queue;
3127#ifdef CONFIG_XFRM_OFFLOAD
3128 struct sk_buff_head xfrm_backlog;
3129#endif
3130 /* written and read only by owning cpu: */
3131 struct {
3132 u16 recursion;
3133 u8 more;
3134#ifdef CONFIG_NET_EGRESS
3135 u8 skip_txqueue;
3136#endif
3137 } xmit;
3138#ifdef CONFIG_RPS
3139 /* input_queue_head should be written by cpu owning this struct,
3140 * and only read by other cpus. Worth using a cache line.
3141 */
3142 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3143
3144 /* Elements below can be accessed between CPUs for RPS/RFS */
3145 call_single_data_t csd ____cacheline_aligned_in_smp;
3146 struct softnet_data *rps_ipi_next;
3147 unsigned int cpu;
3148 unsigned int input_queue_tail;
3149#endif
3150 unsigned int dropped;
3151 struct sk_buff_head input_pkt_queue;
3152 struct napi_struct backlog;
3153
3154 /* Another possibly contended cache line */
3155 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3156 int defer_count;
3157 int defer_ipi_scheduled;
3158 struct sk_buff *defer_list;
3159 call_single_data_t defer_csd;
3160};
3161
3162static inline void input_queue_head_incr(struct softnet_data *sd)
3163{
3164#ifdef CONFIG_RPS
3165 sd->input_queue_head++;
3166#endif
3167}
3168
3169static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3170 unsigned int *qtail)
3171{
3172#ifdef CONFIG_RPS
3173 *qtail = ++sd->input_queue_tail;
3174#endif
3175}
3176
3177DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3178
3179static inline int dev_recursion_level(void)
3180{
3181 return this_cpu_read(softnet_data.xmit.recursion);
3182}
3183
3184#define XMIT_RECURSION_LIMIT 8
3185static inline bool dev_xmit_recursion(void)
3186{
3187 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3188 XMIT_RECURSION_LIMIT);
3189}
3190
3191static inline void dev_xmit_recursion_inc(void)
3192{
3193 __this_cpu_inc(softnet_data.xmit.recursion);
3194}
3195
3196static inline void dev_xmit_recursion_dec(void)
3197{
3198 __this_cpu_dec(softnet_data.xmit.recursion);
3199}
3200
3201void __netif_schedule(struct Qdisc *q);
3202void netif_schedule_queue(struct netdev_queue *txq);
3203
3204static inline void netif_tx_schedule_all(struct net_device *dev)
3205{
3206 unsigned int i;
3207
3208 for (i = 0; i < dev->num_tx_queues; i++)
3209 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3210}
3211
3212static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3213{
3214 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3215}
3216
3217/**
3218 * netif_start_queue - allow transmit
3219 * @dev: network device
3220 *
3221 * Allow upper layers to call the device hard_start_xmit routine.
3222 */
3223static inline void netif_start_queue(struct net_device *dev)
3224{
3225 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3226}
3227
3228static inline void netif_tx_start_all_queues(struct net_device *dev)
3229{
3230 unsigned int i;
3231
3232 for (i = 0; i < dev->num_tx_queues; i++) {
3233 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3234 netif_tx_start_queue(txq);
3235 }
3236}
3237
3238void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3239
3240/**
3241 * netif_wake_queue - restart transmit
3242 * @dev: network device
3243 *
3244 * Allow upper layers to call the device hard_start_xmit routine.
3245 * Used for flow control when transmit resources are available.
3246 */
3247static inline void netif_wake_queue(struct net_device *dev)
3248{
3249 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3250}
3251
3252static inline void netif_tx_wake_all_queues(struct net_device *dev)
3253{
3254 unsigned int i;
3255
3256 for (i = 0; i < dev->num_tx_queues; i++) {
3257 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3258 netif_tx_wake_queue(txq);
3259 }
3260}
3261
3262static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3263{
3264 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3265}
3266
3267/**
3268 * netif_stop_queue - stop transmitted packets
3269 * @dev: network device
3270 *
3271 * Stop upper layers calling the device hard_start_xmit routine.
3272 * Used for flow control when transmit resources are unavailable.
3273 */
3274static inline void netif_stop_queue(struct net_device *dev)
3275{
3276 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3277}
3278
3279void netif_tx_stop_all_queues(struct net_device *dev);
3280
3281static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3282{
3283 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3284}
3285
3286/**
3287 * netif_queue_stopped - test if transmit queue is flowblocked
3288 * @dev: network device
3289 *
3290 * Test if transmit queue on device is currently unable to send.
3291 */
3292static inline bool netif_queue_stopped(const struct net_device *dev)
3293{
3294 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3295}
3296
3297static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3298{
3299 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3300}
3301
3302static inline bool
3303netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3304{
3305 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3306}
3307
3308static inline bool
3309netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3310{
3311 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3312}
3313
3314/**
3315 * netdev_queue_set_dql_min_limit - set dql minimum limit
3316 * @dev_queue: pointer to transmit queue
3317 * @min_limit: dql minimum limit
3318 *
3319 * Forces xmit_more() to return true until the minimum threshold
3320 * defined by @min_limit is reached (or until the tx queue is
3321 * empty). Warning: to be use with care, misuse will impact the
3322 * latency.
3323 */
3324static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3325 unsigned int min_limit)
3326{
3327#ifdef CONFIG_BQL
3328 dev_queue->dql.min_limit = min_limit;
3329#endif
3330}
3331
3332/**
3333 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3334 * @dev_queue: pointer to transmit queue
3335 *
3336 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3337 * to give appropriate hint to the CPU.
3338 */
3339static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3340{
3341#ifdef CONFIG_BQL
3342 prefetchw(&dev_queue->dql.num_queued);
3343#endif
3344}
3345
3346/**
3347 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3348 * @dev_queue: pointer to transmit queue
3349 *
3350 * BQL enabled drivers might use this helper in their TX completion path,
3351 * to give appropriate hint to the CPU.
3352 */
3353static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3354{
3355#ifdef CONFIG_BQL
3356 prefetchw(&dev_queue->dql.limit);
3357#endif
3358}
3359
3360/**
3361 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3362 * @dev_queue: network device queue
3363 * @bytes: number of bytes queued to the device queue
3364 *
3365 * Report the number of bytes queued for sending/completion to the network
3366 * device hardware queue. @bytes should be a good approximation and should
3367 * exactly match netdev_completed_queue() @bytes.
3368 * This is typically called once per packet, from ndo_start_xmit().
3369 */
3370static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3371 unsigned int bytes)
3372{
3373#ifdef CONFIG_BQL
3374 dql_queued(&dev_queue->dql, bytes);
3375
3376 if (likely(dql_avail(&dev_queue->dql) >= 0))
3377 return;
3378
3379 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3380
3381 /*
3382 * The XOFF flag must be set before checking the dql_avail below,
3383 * because in netdev_tx_completed_queue we update the dql_completed
3384 * before checking the XOFF flag.
3385 */
3386 smp_mb();
3387
3388 /* check again in case another CPU has just made room avail */
3389 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3390 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3391#endif
3392}
3393
3394/* Variant of netdev_tx_sent_queue() for drivers that are aware
3395 * that they should not test BQL status themselves.
3396 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3397 * skb of a batch.
3398 * Returns true if the doorbell must be used to kick the NIC.
3399 */
3400static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3401 unsigned int bytes,
3402 bool xmit_more)
3403{
3404 if (xmit_more) {
3405#ifdef CONFIG_BQL
3406 dql_queued(&dev_queue->dql, bytes);
3407#endif
3408 return netif_tx_queue_stopped(dev_queue);
3409 }
3410 netdev_tx_sent_queue(dev_queue, bytes);
3411 return true;
3412}
3413
3414/**
3415 * netdev_sent_queue - report the number of bytes queued to hardware
3416 * @dev: network device
3417 * @bytes: number of bytes queued to the hardware device queue
3418 *
3419 * Report the number of bytes queued for sending/completion to the network
3420 * device hardware queue#0. @bytes should be a good approximation and should
3421 * exactly match netdev_completed_queue() @bytes.
3422 * This is typically called once per packet, from ndo_start_xmit().
3423 */
3424static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3425{
3426 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3427}
3428
3429static inline bool __netdev_sent_queue(struct net_device *dev,
3430 unsigned int bytes,
3431 bool xmit_more)
3432{
3433 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3434 xmit_more);
3435}
3436
3437/**
3438 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3439 * @dev_queue: network device queue
3440 * @pkts: number of packets (currently ignored)
3441 * @bytes: number of bytes dequeued from the device queue
3442 *
3443 * Must be called at most once per TX completion round (and not per
3444 * individual packet), so that BQL can adjust its limits appropriately.
3445 */
3446static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3447 unsigned int pkts, unsigned int bytes)
3448{
3449#ifdef CONFIG_BQL
3450 if (unlikely(!bytes))
3451 return;
3452
3453 dql_completed(&dev_queue->dql, bytes);
3454
3455 /*
3456 * Without the memory barrier there is a small possiblity that
3457 * netdev_tx_sent_queue will miss the update and cause the queue to
3458 * be stopped forever
3459 */
3460 smp_mb();
3461
3462 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3463 return;
3464
3465 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3466 netif_schedule_queue(dev_queue);
3467#endif
3468}
3469
3470/**
3471 * netdev_completed_queue - report bytes and packets completed by device
3472 * @dev: network device
3473 * @pkts: actual number of packets sent over the medium
3474 * @bytes: actual number of bytes sent over the medium
3475 *
3476 * Report the number of bytes and packets transmitted by the network device
3477 * hardware queue over the physical medium, @bytes must exactly match the
3478 * @bytes amount passed to netdev_sent_queue()
3479 */
3480static inline void netdev_completed_queue(struct net_device *dev,
3481 unsigned int pkts, unsigned int bytes)
3482{
3483 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3484}
3485
3486static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3487{
3488#ifdef CONFIG_BQL
3489 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3490 dql_reset(&q->dql);
3491#endif
3492}
3493
3494/**
3495 * netdev_reset_queue - reset the packets and bytes count of a network device
3496 * @dev_queue: network device
3497 *
3498 * Reset the bytes and packet count of a network device and clear the
3499 * software flow control OFF bit for this network device
3500 */
3501static inline void netdev_reset_queue(struct net_device *dev_queue)
3502{
3503 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3504}
3505
3506/**
3507 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3508 * @dev: network device
3509 * @queue_index: given tx queue index
3510 *
3511 * Returns 0 if given tx queue index >= number of device tx queues,
3512 * otherwise returns the originally passed tx queue index.
3513 */
3514static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3515{
3516 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3517 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3518 dev->name, queue_index,
3519 dev->real_num_tx_queues);
3520 return 0;
3521 }
3522
3523 return queue_index;
3524}
3525
3526/**
3527 * netif_running - test if up
3528 * @dev: network device
3529 *
3530 * Test if the device has been brought up.
3531 */
3532static inline bool netif_running(const struct net_device *dev)
3533{
3534 return test_bit(__LINK_STATE_START, &dev->state);
3535}
3536
3537/*
3538 * Routines to manage the subqueues on a device. We only need start,
3539 * stop, and a check if it's stopped. All other device management is
3540 * done at the overall netdevice level.
3541 * Also test the device if we're multiqueue.
3542 */
3543
3544/**
3545 * netif_start_subqueue - allow sending packets on subqueue
3546 * @dev: network device
3547 * @queue_index: sub queue index
3548 *
3549 * Start individual transmit queue of a device with multiple transmit queues.
3550 */
3551static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3552{
3553 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3554
3555 netif_tx_start_queue(txq);
3556}
3557
3558/**
3559 * netif_stop_subqueue - stop sending packets on subqueue
3560 * @dev: network device
3561 * @queue_index: sub queue index
3562 *
3563 * Stop individual transmit queue of a device with multiple transmit queues.
3564 */
3565static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3566{
3567 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3568 netif_tx_stop_queue(txq);
3569}
3570
3571/**
3572 * __netif_subqueue_stopped - test status of subqueue
3573 * @dev: network device
3574 * @queue_index: sub queue index
3575 *
3576 * Check individual transmit queue of a device with multiple transmit queues.
3577 */
3578static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3579 u16 queue_index)
3580{
3581 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3582
3583 return netif_tx_queue_stopped(txq);
3584}
3585
3586/**
3587 * netif_subqueue_stopped - test status of subqueue
3588 * @dev: network device
3589 * @skb: sub queue buffer pointer
3590 *
3591 * Check individual transmit queue of a device with multiple transmit queues.
3592 */
3593static inline bool netif_subqueue_stopped(const struct net_device *dev,
3594 struct sk_buff *skb)
3595{
3596 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3597}
3598
3599/**
3600 * netif_wake_subqueue - allow sending packets on subqueue
3601 * @dev: network device
3602 * @queue_index: sub queue index
3603 *
3604 * Resume individual transmit queue of a device with multiple transmit queues.
3605 */
3606static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3607{
3608 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3609
3610 netif_tx_wake_queue(txq);
3611}
3612
3613#ifdef CONFIG_XPS
3614int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3615 u16 index);
3616int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3617 u16 index, enum xps_map_type type);
3618
3619/**
3620 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3621 * @j: CPU/Rx queue index
3622 * @mask: bitmask of all cpus/rx queues
3623 * @nr_bits: number of bits in the bitmask
3624 *
3625 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3626 */
3627static inline bool netif_attr_test_mask(unsigned long j,
3628 const unsigned long *mask,
3629 unsigned int nr_bits)
3630{
3631 cpu_max_bits_warn(j, nr_bits);
3632 return test_bit(j, mask);
3633}
3634
3635/**
3636 * netif_attr_test_online - Test for online CPU/Rx queue
3637 * @j: CPU/Rx queue index
3638 * @online_mask: bitmask for CPUs/Rx queues that are online
3639 * @nr_bits: number of bits in the bitmask
3640 *
3641 * Returns true if a CPU/Rx queue is online.
3642 */
3643static inline bool netif_attr_test_online(unsigned long j,
3644 const unsigned long *online_mask,
3645 unsigned int nr_bits)
3646{
3647 cpu_max_bits_warn(j, nr_bits);
3648
3649 if (online_mask)
3650 return test_bit(j, online_mask);
3651
3652 return (j < nr_bits);
3653}
3654
3655/**
3656 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3657 * @n: CPU/Rx queue index
3658 * @srcp: the cpumask/Rx queue mask pointer
3659 * @nr_bits: number of bits in the bitmask
3660 *
3661 * Returns >= nr_bits if no further CPUs/Rx queues set.
3662 */
3663static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3664 unsigned int nr_bits)
3665{
3666 /* -1 is a legal arg here. */
3667 if (n != -1)
3668 cpu_max_bits_warn(n, nr_bits);
3669
3670 if (srcp)
3671 return find_next_bit(srcp, nr_bits, n + 1);
3672
3673 return n + 1;
3674}
3675
3676/**
3677 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3678 * @n: CPU/Rx queue index
3679 * @src1p: the first CPUs/Rx queues mask pointer
3680 * @src2p: the second CPUs/Rx queues mask pointer
3681 * @nr_bits: number of bits in the bitmask
3682 *
3683 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3684 */
3685static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3686 const unsigned long *src2p,
3687 unsigned int nr_bits)
3688{
3689 /* -1 is a legal arg here. */
3690 if (n != -1)
3691 cpu_max_bits_warn(n, nr_bits);
3692
3693 if (src1p && src2p)
3694 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3695 else if (src1p)
3696 return find_next_bit(src1p, nr_bits, n + 1);
3697 else if (src2p)
3698 return find_next_bit(src2p, nr_bits, n + 1);
3699
3700 return n + 1;
3701}
3702#else
3703static inline int netif_set_xps_queue(struct net_device *dev,
3704 const struct cpumask *mask,
3705 u16 index)
3706{
3707 return 0;
3708}
3709
3710static inline int __netif_set_xps_queue(struct net_device *dev,
3711 const unsigned long *mask,
3712 u16 index, enum xps_map_type type)
3713{
3714 return 0;
3715}
3716#endif
3717
3718/**
3719 * netif_is_multiqueue - test if device has multiple transmit queues
3720 * @dev: network device
3721 *
3722 * Check if device has multiple transmit queues
3723 */
3724static inline bool netif_is_multiqueue(const struct net_device *dev)
3725{
3726 return dev->num_tx_queues > 1;
3727}
3728
3729int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3730
3731#ifdef CONFIG_SYSFS
3732int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3733#else
3734static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3735 unsigned int rxqs)
3736{
3737 dev->real_num_rx_queues = rxqs;
3738 return 0;
3739}
3740#endif
3741int netif_set_real_num_queues(struct net_device *dev,
3742 unsigned int txq, unsigned int rxq);
3743
3744static inline struct netdev_rx_queue *
3745__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3746{
3747 return dev->_rx + rxq;
3748}
3749
3750#ifdef CONFIG_SYSFS
3751static inline unsigned int get_netdev_rx_queue_index(
3752 struct netdev_rx_queue *queue)
3753{
3754 struct net_device *dev = queue->dev;
3755 int index = queue - dev->_rx;
3756
3757 BUG_ON(index >= dev->num_rx_queues);
3758 return index;
3759}
3760#endif
3761
3762int netif_get_num_default_rss_queues(void);
3763
3764enum skb_free_reason {
3765 SKB_REASON_CONSUMED,
3766 SKB_REASON_DROPPED,
3767};
3768
3769void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3770void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3771
3772/*
3773 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3774 * interrupt context or with hardware interrupts being disabled.
3775 * (in_hardirq() || irqs_disabled())
3776 *
3777 * We provide four helpers that can be used in following contexts :
3778 *
3779 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3780 * replacing kfree_skb(skb)
3781 *
3782 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3783 * Typically used in place of consume_skb(skb) in TX completion path
3784 *
3785 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3786 * replacing kfree_skb(skb)
3787 *
3788 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3789 * and consumed a packet. Used in place of consume_skb(skb)
3790 */
3791static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3792{
3793 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3794}
3795
3796static inline void dev_consume_skb_irq(struct sk_buff *skb)
3797{
3798 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3799}
3800
3801static inline void dev_kfree_skb_any(struct sk_buff *skb)
3802{
3803 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3804}
3805
3806static inline void dev_consume_skb_any(struct sk_buff *skb)
3807{
3808 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3809}
3810
3811u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3812 struct bpf_prog *xdp_prog);
3813void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3814int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3815int netif_rx(struct sk_buff *skb);
3816int __netif_rx(struct sk_buff *skb);
3817
3818int netif_receive_skb(struct sk_buff *skb);
3819int netif_receive_skb_core(struct sk_buff *skb);
3820void netif_receive_skb_list_internal(struct list_head *head);
3821void netif_receive_skb_list(struct list_head *head);
3822gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3823void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3824struct sk_buff *napi_get_frags(struct napi_struct *napi);
3825void napi_get_frags_check(struct napi_struct *napi);
3826gro_result_t napi_gro_frags(struct napi_struct *napi);
3827struct packet_offload *gro_find_receive_by_type(__be16 type);
3828struct packet_offload *gro_find_complete_by_type(__be16 type);
3829
3830static inline void napi_free_frags(struct napi_struct *napi)
3831{
3832 kfree_skb(napi->skb);
3833 napi->skb = NULL;
3834}
3835
3836bool netdev_is_rx_handler_busy(struct net_device *dev);
3837int netdev_rx_handler_register(struct net_device *dev,
3838 rx_handler_func_t *rx_handler,
3839 void *rx_handler_data);
3840void netdev_rx_handler_unregister(struct net_device *dev);
3841
3842bool dev_valid_name(const char *name);
3843static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3844{
3845 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3846}
3847int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3848int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3849int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3850 void __user *data, bool *need_copyout);
3851int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3852int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3853unsigned int dev_get_flags(const struct net_device *);
3854int __dev_change_flags(struct net_device *dev, unsigned int flags,
3855 struct netlink_ext_ack *extack);
3856int dev_change_flags(struct net_device *dev, unsigned int flags,
3857 struct netlink_ext_ack *extack);
3858void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3859 unsigned int gchanges);
3860int dev_set_alias(struct net_device *, const char *, size_t);
3861int dev_get_alias(const struct net_device *, char *, size_t);
3862int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3863 const char *pat, int new_ifindex);
3864static inline
3865int dev_change_net_namespace(struct net_device *dev, struct net *net,
3866 const char *pat)
3867{
3868 return __dev_change_net_namespace(dev, net, pat, 0);
3869}
3870int __dev_set_mtu(struct net_device *, int);
3871int dev_set_mtu(struct net_device *, int);
3872int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3873 struct netlink_ext_ack *extack);
3874int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3875 struct netlink_ext_ack *extack);
3876int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3877 struct netlink_ext_ack *extack);
3878int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3879int dev_get_port_parent_id(struct net_device *dev,
3880 struct netdev_phys_item_id *ppid, bool recurse);
3881bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3882struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3883struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3884 struct netdev_queue *txq, int *ret);
3885
3886int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3887u8 dev_xdp_prog_count(struct net_device *dev);
3888u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3889
3890int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3891int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3892int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3893bool is_skb_forwardable(const struct net_device *dev,
3894 const struct sk_buff *skb);
3895
3896static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3897 const struct sk_buff *skb,
3898 const bool check_mtu)
3899{
3900 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3901 unsigned int len;
3902
3903 if (!(dev->flags & IFF_UP))
3904 return false;
3905
3906 if (!check_mtu)
3907 return true;
3908
3909 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3910 if (skb->len <= len)
3911 return true;
3912
3913 /* if TSO is enabled, we don't care about the length as the packet
3914 * could be forwarded without being segmented before
3915 */
3916 if (skb_is_gso(skb))
3917 return true;
3918
3919 return false;
3920}
3921
3922struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3923
3924static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3925{
3926 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3927 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3928
3929 if (likely(p))
3930 return p;
3931
3932 return netdev_core_stats_alloc(dev);
3933}
3934
3935#define DEV_CORE_STATS_INC(FIELD) \
3936static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3937{ \
3938 struct net_device_core_stats __percpu *p; \
3939 \
3940 p = dev_core_stats(dev); \
3941 if (p) \
3942 this_cpu_inc(p->FIELD); \
3943}
3944DEV_CORE_STATS_INC(rx_dropped)
3945DEV_CORE_STATS_INC(tx_dropped)
3946DEV_CORE_STATS_INC(rx_nohandler)
3947DEV_CORE_STATS_INC(rx_otherhost_dropped)
3948
3949static __always_inline int ____dev_forward_skb(struct net_device *dev,
3950 struct sk_buff *skb,
3951 const bool check_mtu)
3952{
3953 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3954 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3955 dev_core_stats_rx_dropped_inc(dev);
3956 kfree_skb(skb);
3957 return NET_RX_DROP;
3958 }
3959
3960 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3961 skb->priority = 0;
3962 return 0;
3963}
3964
3965bool dev_nit_active(struct net_device *dev);
3966void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3967
3968static inline void __dev_put(struct net_device *dev)
3969{
3970 if (dev) {
3971#ifdef CONFIG_PCPU_DEV_REFCNT
3972 this_cpu_dec(*dev->pcpu_refcnt);
3973#else
3974 refcount_dec(&dev->dev_refcnt);
3975#endif
3976 }
3977}
3978
3979static inline void __dev_hold(struct net_device *dev)
3980{
3981 if (dev) {
3982#ifdef CONFIG_PCPU_DEV_REFCNT
3983 this_cpu_inc(*dev->pcpu_refcnt);
3984#else
3985 refcount_inc(&dev->dev_refcnt);
3986#endif
3987 }
3988}
3989
3990static inline void __netdev_tracker_alloc(struct net_device *dev,
3991 netdevice_tracker *tracker,
3992 gfp_t gfp)
3993{
3994#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3995 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
3996#endif
3997}
3998
3999/* netdev_tracker_alloc() can upgrade a prior untracked reference
4000 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4001 */
4002static inline void netdev_tracker_alloc(struct net_device *dev,
4003 netdevice_tracker *tracker, gfp_t gfp)
4004{
4005#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4006 refcount_dec(&dev->refcnt_tracker.no_tracker);
4007 __netdev_tracker_alloc(dev, tracker, gfp);
4008#endif
4009}
4010
4011static inline void netdev_tracker_free(struct net_device *dev,
4012 netdevice_tracker *tracker)
4013{
4014#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4015 ref_tracker_free(&dev->refcnt_tracker, tracker);
4016#endif
4017}
4018
4019static inline void netdev_hold(struct net_device *dev,
4020 netdevice_tracker *tracker, gfp_t gfp)
4021{
4022 if (dev) {
4023 __dev_hold(dev);
4024 __netdev_tracker_alloc(dev, tracker, gfp);
4025 }
4026}
4027
4028static inline void netdev_put(struct net_device *dev,
4029 netdevice_tracker *tracker)
4030{
4031 if (dev) {
4032 netdev_tracker_free(dev, tracker);
4033 __dev_put(dev);
4034 }
4035}
4036
4037/**
4038 * dev_hold - get reference to device
4039 * @dev: network device
4040 *
4041 * Hold reference to device to keep it from being freed.
4042 * Try using netdev_hold() instead.
4043 */
4044static inline void dev_hold(struct net_device *dev)
4045{
4046 netdev_hold(dev, NULL, GFP_ATOMIC);
4047}
4048
4049/**
4050 * dev_put - release reference to device
4051 * @dev: network device
4052 *
4053 * Release reference to device to allow it to be freed.
4054 * Try using netdev_put() instead.
4055 */
4056static inline void dev_put(struct net_device *dev)
4057{
4058 netdev_put(dev, NULL);
4059}
4060
4061static inline void netdev_ref_replace(struct net_device *odev,
4062 struct net_device *ndev,
4063 netdevice_tracker *tracker,
4064 gfp_t gfp)
4065{
4066 if (odev)
4067 netdev_tracker_free(odev, tracker);
4068
4069 __dev_hold(ndev);
4070 __dev_put(odev);
4071
4072 if (ndev)
4073 __netdev_tracker_alloc(ndev, tracker, gfp);
4074}
4075
4076/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4077 * and _off may be called from IRQ context, but it is caller
4078 * who is responsible for serialization of these calls.
4079 *
4080 * The name carrier is inappropriate, these functions should really be
4081 * called netif_lowerlayer_*() because they represent the state of any
4082 * kind of lower layer not just hardware media.
4083 */
4084void linkwatch_fire_event(struct net_device *dev);
4085
4086/**
4087 * netif_carrier_ok - test if carrier present
4088 * @dev: network device
4089 *
4090 * Check if carrier is present on device
4091 */
4092static inline bool netif_carrier_ok(const struct net_device *dev)
4093{
4094 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4095}
4096
4097unsigned long dev_trans_start(struct net_device *dev);
4098
4099void __netdev_watchdog_up(struct net_device *dev);
4100
4101void netif_carrier_on(struct net_device *dev);
4102void netif_carrier_off(struct net_device *dev);
4103void netif_carrier_event(struct net_device *dev);
4104
4105/**
4106 * netif_dormant_on - mark device as dormant.
4107 * @dev: network device
4108 *
4109 * Mark device as dormant (as per RFC2863).
4110 *
4111 * The dormant state indicates that the relevant interface is not
4112 * actually in a condition to pass packets (i.e., it is not 'up') but is
4113 * in a "pending" state, waiting for some external event. For "on-
4114 * demand" interfaces, this new state identifies the situation where the
4115 * interface is waiting for events to place it in the up state.
4116 */
4117static inline void netif_dormant_on(struct net_device *dev)
4118{
4119 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4120 linkwatch_fire_event(dev);
4121}
4122
4123/**
4124 * netif_dormant_off - set device as not dormant.
4125 * @dev: network device
4126 *
4127 * Device is not in dormant state.
4128 */
4129static inline void netif_dormant_off(struct net_device *dev)
4130{
4131 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4132 linkwatch_fire_event(dev);
4133}
4134
4135/**
4136 * netif_dormant - test if device is dormant
4137 * @dev: network device
4138 *
4139 * Check if device is dormant.
4140 */
4141static inline bool netif_dormant(const struct net_device *dev)
4142{
4143 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4144}
4145
4146
4147/**
4148 * netif_testing_on - mark device as under test.
4149 * @dev: network device
4150 *
4151 * Mark device as under test (as per RFC2863).
4152 *
4153 * The testing state indicates that some test(s) must be performed on
4154 * the interface. After completion, of the test, the interface state
4155 * will change to up, dormant, or down, as appropriate.
4156 */
4157static inline void netif_testing_on(struct net_device *dev)
4158{
4159 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4160 linkwatch_fire_event(dev);
4161}
4162
4163/**
4164 * netif_testing_off - set device as not under test.
4165 * @dev: network device
4166 *
4167 * Device is not in testing state.
4168 */
4169static inline void netif_testing_off(struct net_device *dev)
4170{
4171 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4172 linkwatch_fire_event(dev);
4173}
4174
4175/**
4176 * netif_testing - test if device is under test
4177 * @dev: network device
4178 *
4179 * Check if device is under test
4180 */
4181static inline bool netif_testing(const struct net_device *dev)
4182{
4183 return test_bit(__LINK_STATE_TESTING, &dev->state);
4184}
4185
4186
4187/**
4188 * netif_oper_up - test if device is operational
4189 * @dev: network device
4190 *
4191 * Check if carrier is operational
4192 */
4193static inline bool netif_oper_up(const struct net_device *dev)
4194{
4195 return (dev->operstate == IF_OPER_UP ||
4196 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4197}
4198
4199/**
4200 * netif_device_present - is device available or removed
4201 * @dev: network device
4202 *
4203 * Check if device has not been removed from system.
4204 */
4205static inline bool netif_device_present(const struct net_device *dev)
4206{
4207 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4208}
4209
4210void netif_device_detach(struct net_device *dev);
4211
4212void netif_device_attach(struct net_device *dev);
4213
4214/*
4215 * Network interface message level settings
4216 */
4217
4218enum {
4219 NETIF_MSG_DRV_BIT,
4220 NETIF_MSG_PROBE_BIT,
4221 NETIF_MSG_LINK_BIT,
4222 NETIF_MSG_TIMER_BIT,
4223 NETIF_MSG_IFDOWN_BIT,
4224 NETIF_MSG_IFUP_BIT,
4225 NETIF_MSG_RX_ERR_BIT,
4226 NETIF_MSG_TX_ERR_BIT,
4227 NETIF_MSG_TX_QUEUED_BIT,
4228 NETIF_MSG_INTR_BIT,
4229 NETIF_MSG_TX_DONE_BIT,
4230 NETIF_MSG_RX_STATUS_BIT,
4231 NETIF_MSG_PKTDATA_BIT,
4232 NETIF_MSG_HW_BIT,
4233 NETIF_MSG_WOL_BIT,
4234
4235 /* When you add a new bit above, update netif_msg_class_names array
4236 * in net/ethtool/common.c
4237 */
4238 NETIF_MSG_CLASS_COUNT,
4239};
4240/* Both ethtool_ops interface and internal driver implementation use u32 */
4241static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4242
4243#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4244#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4245
4246#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4247#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4248#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4249#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4250#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4251#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4252#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4253#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4254#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4255#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4256#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4257#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4258#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4259#define NETIF_MSG_HW __NETIF_MSG(HW)
4260#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4261
4262#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4263#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4264#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4265#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4266#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4267#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4268#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4269#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4270#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4271#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4272#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4273#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4274#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4275#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4276#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4277
4278static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4279{
4280 /* use default */
4281 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4282 return default_msg_enable_bits;
4283 if (debug_value == 0) /* no output */
4284 return 0;
4285 /* set low N bits */
4286 return (1U << debug_value) - 1;
4287}
4288
4289static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4290{
4291 spin_lock(&txq->_xmit_lock);
4292 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4293 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4294}
4295
4296static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4297{
4298 __acquire(&txq->_xmit_lock);
4299 return true;
4300}
4301
4302static inline void __netif_tx_release(struct netdev_queue *txq)
4303{
4304 __release(&txq->_xmit_lock);
4305}
4306
4307static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4308{
4309 spin_lock_bh(&txq->_xmit_lock);
4310 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4311 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4312}
4313
4314static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4315{
4316 bool ok = spin_trylock(&txq->_xmit_lock);
4317
4318 if (likely(ok)) {
4319 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4320 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4321 }
4322 return ok;
4323}
4324
4325static inline void __netif_tx_unlock(struct netdev_queue *txq)
4326{
4327 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4328 WRITE_ONCE(txq->xmit_lock_owner, -1);
4329 spin_unlock(&txq->_xmit_lock);
4330}
4331
4332static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4333{
4334 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4335 WRITE_ONCE(txq->xmit_lock_owner, -1);
4336 spin_unlock_bh(&txq->_xmit_lock);
4337}
4338
4339/*
4340 * txq->trans_start can be read locklessly from dev_watchdog()
4341 */
4342static inline void txq_trans_update(struct netdev_queue *txq)
4343{
4344 if (txq->xmit_lock_owner != -1)
4345 WRITE_ONCE(txq->trans_start, jiffies);
4346}
4347
4348static inline void txq_trans_cond_update(struct netdev_queue *txq)
4349{
4350 unsigned long now = jiffies;
4351
4352 if (READ_ONCE(txq->trans_start) != now)
4353 WRITE_ONCE(txq->trans_start, now);
4354}
4355
4356/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4357static inline void netif_trans_update(struct net_device *dev)
4358{
4359 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4360
4361 txq_trans_cond_update(txq);
4362}
4363
4364/**
4365 * netif_tx_lock - grab network device transmit lock
4366 * @dev: network device
4367 *
4368 * Get network device transmit lock
4369 */
4370void netif_tx_lock(struct net_device *dev);
4371
4372static inline void netif_tx_lock_bh(struct net_device *dev)
4373{
4374 local_bh_disable();
4375 netif_tx_lock(dev);
4376}
4377
4378void netif_tx_unlock(struct net_device *dev);
4379
4380static inline void netif_tx_unlock_bh(struct net_device *dev)
4381{
4382 netif_tx_unlock(dev);
4383 local_bh_enable();
4384}
4385
4386#define HARD_TX_LOCK(dev, txq, cpu) { \
4387 if ((dev->features & NETIF_F_LLTX) == 0) { \
4388 __netif_tx_lock(txq, cpu); \
4389 } else { \
4390 __netif_tx_acquire(txq); \
4391 } \
4392}
4393
4394#define HARD_TX_TRYLOCK(dev, txq) \
4395 (((dev->features & NETIF_F_LLTX) == 0) ? \
4396 __netif_tx_trylock(txq) : \
4397 __netif_tx_acquire(txq))
4398
4399#define HARD_TX_UNLOCK(dev, txq) { \
4400 if ((dev->features & NETIF_F_LLTX) == 0) { \
4401 __netif_tx_unlock(txq); \
4402 } else { \
4403 __netif_tx_release(txq); \
4404 } \
4405}
4406
4407static inline void netif_tx_disable(struct net_device *dev)
4408{
4409 unsigned int i;
4410 int cpu;
4411
4412 local_bh_disable();
4413 cpu = smp_processor_id();
4414 spin_lock(&dev->tx_global_lock);
4415 for (i = 0; i < dev->num_tx_queues; i++) {
4416 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4417
4418 __netif_tx_lock(txq, cpu);
4419 netif_tx_stop_queue(txq);
4420 __netif_tx_unlock(txq);
4421 }
4422 spin_unlock(&dev->tx_global_lock);
4423 local_bh_enable();
4424}
4425
4426static inline void netif_addr_lock(struct net_device *dev)
4427{
4428 unsigned char nest_level = 0;
4429
4430#ifdef CONFIG_LOCKDEP
4431 nest_level = dev->nested_level;
4432#endif
4433 spin_lock_nested(&dev->addr_list_lock, nest_level);
4434}
4435
4436static inline void netif_addr_lock_bh(struct net_device *dev)
4437{
4438 unsigned char nest_level = 0;
4439
4440#ifdef CONFIG_LOCKDEP
4441 nest_level = dev->nested_level;
4442#endif
4443 local_bh_disable();
4444 spin_lock_nested(&dev->addr_list_lock, nest_level);
4445}
4446
4447static inline void netif_addr_unlock(struct net_device *dev)
4448{
4449 spin_unlock(&dev->addr_list_lock);
4450}
4451
4452static inline void netif_addr_unlock_bh(struct net_device *dev)
4453{
4454 spin_unlock_bh(&dev->addr_list_lock);
4455}
4456
4457/*
4458 * dev_addrs walker. Should be used only for read access. Call with
4459 * rcu_read_lock held.
4460 */
4461#define for_each_dev_addr(dev, ha) \
4462 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4463
4464/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4465
4466void ether_setup(struct net_device *dev);
4467
4468/* Support for loadable net-drivers */
4469struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4470 unsigned char name_assign_type,
4471 void (*setup)(struct net_device *),
4472 unsigned int txqs, unsigned int rxqs);
4473#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4474 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4475
4476#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4477 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4478 count)
4479
4480int register_netdev(struct net_device *dev);
4481void unregister_netdev(struct net_device *dev);
4482
4483int devm_register_netdev(struct device *dev, struct net_device *ndev);
4484
4485/* General hardware address lists handling functions */
4486int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4487 struct netdev_hw_addr_list *from_list, int addr_len);
4488void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4489 struct netdev_hw_addr_list *from_list, int addr_len);
4490int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4491 struct net_device *dev,
4492 int (*sync)(struct net_device *, const unsigned char *),
4493 int (*unsync)(struct net_device *,
4494 const unsigned char *));
4495int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4496 struct net_device *dev,
4497 int (*sync)(struct net_device *,
4498 const unsigned char *, int),
4499 int (*unsync)(struct net_device *,
4500 const unsigned char *, int));
4501void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4502 struct net_device *dev,
4503 int (*unsync)(struct net_device *,
4504 const unsigned char *, int));
4505void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4506 struct net_device *dev,
4507 int (*unsync)(struct net_device *,
4508 const unsigned char *));
4509void __hw_addr_init(struct netdev_hw_addr_list *list);
4510
4511/* Functions used for device addresses handling */
4512void dev_addr_mod(struct net_device *dev, unsigned int offset,
4513 const void *addr, size_t len);
4514
4515static inline void
4516__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4517{
4518 dev_addr_mod(dev, 0, addr, len);
4519}
4520
4521static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4522{
4523 __dev_addr_set(dev, addr, dev->addr_len);
4524}
4525
4526int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4527 unsigned char addr_type);
4528int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4529 unsigned char addr_type);
4530
4531/* Functions used for unicast addresses handling */
4532int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4533int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4534int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4535int dev_uc_sync(struct net_device *to, struct net_device *from);
4536int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4537void dev_uc_unsync(struct net_device *to, struct net_device *from);
4538void dev_uc_flush(struct net_device *dev);
4539void dev_uc_init(struct net_device *dev);
4540
4541/**
4542 * __dev_uc_sync - Synchonize device's unicast list
4543 * @dev: device to sync
4544 * @sync: function to call if address should be added
4545 * @unsync: function to call if address should be removed
4546 *
4547 * Add newly added addresses to the interface, and release
4548 * addresses that have been deleted.
4549 */
4550static inline int __dev_uc_sync(struct net_device *dev,
4551 int (*sync)(struct net_device *,
4552 const unsigned char *),
4553 int (*unsync)(struct net_device *,
4554 const unsigned char *))
4555{
4556 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4557}
4558
4559/**
4560 * __dev_uc_unsync - Remove synchronized addresses from device
4561 * @dev: device to sync
4562 * @unsync: function to call if address should be removed
4563 *
4564 * Remove all addresses that were added to the device by dev_uc_sync().
4565 */
4566static inline void __dev_uc_unsync(struct net_device *dev,
4567 int (*unsync)(struct net_device *,
4568 const unsigned char *))
4569{
4570 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4571}
4572
4573/* Functions used for multicast addresses handling */
4574int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4575int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4576int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4577int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4578int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4579int dev_mc_sync(struct net_device *to, struct net_device *from);
4580int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4581void dev_mc_unsync(struct net_device *to, struct net_device *from);
4582void dev_mc_flush(struct net_device *dev);
4583void dev_mc_init(struct net_device *dev);
4584
4585/**
4586 * __dev_mc_sync - Synchonize device's multicast list
4587 * @dev: device to sync
4588 * @sync: function to call if address should be added
4589 * @unsync: function to call if address should be removed
4590 *
4591 * Add newly added addresses to the interface, and release
4592 * addresses that have been deleted.
4593 */
4594static inline int __dev_mc_sync(struct net_device *dev,
4595 int (*sync)(struct net_device *,
4596 const unsigned char *),
4597 int (*unsync)(struct net_device *,
4598 const unsigned char *))
4599{
4600 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4601}
4602
4603/**
4604 * __dev_mc_unsync - Remove synchronized addresses from device
4605 * @dev: device to sync
4606 * @unsync: function to call if address should be removed
4607 *
4608 * Remove all addresses that were added to the device by dev_mc_sync().
4609 */
4610static inline void __dev_mc_unsync(struct net_device *dev,
4611 int (*unsync)(struct net_device *,
4612 const unsigned char *))
4613{
4614 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4615}
4616
4617/* Functions used for secondary unicast and multicast support */
4618void dev_set_rx_mode(struct net_device *dev);
4619int dev_set_promiscuity(struct net_device *dev, int inc);
4620int dev_set_allmulti(struct net_device *dev, int inc);
4621void netdev_state_change(struct net_device *dev);
4622void __netdev_notify_peers(struct net_device *dev);
4623void netdev_notify_peers(struct net_device *dev);
4624void netdev_features_change(struct net_device *dev);
4625/* Load a device via the kmod */
4626void dev_load(struct net *net, const char *name);
4627struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4628 struct rtnl_link_stats64 *storage);
4629void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4630 const struct net_device_stats *netdev_stats);
4631void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4632 const struct pcpu_sw_netstats __percpu *netstats);
4633void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4634
4635extern int netdev_max_backlog;
4636extern int dev_rx_weight;
4637extern int dev_tx_weight;
4638extern int gro_normal_batch;
4639
4640enum {
4641 NESTED_SYNC_IMM_BIT,
4642 NESTED_SYNC_TODO_BIT,
4643};
4644
4645#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4646#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4647
4648#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4649#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4650
4651struct netdev_nested_priv {
4652 unsigned char flags;
4653 void *data;
4654};
4655
4656bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4657struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4658 struct list_head **iter);
4659
4660/* iterate through upper list, must be called under RCU read lock */
4661#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4662 for (iter = &(dev)->adj_list.upper, \
4663 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4664 updev; \
4665 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4666
4667int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4668 int (*fn)(struct net_device *upper_dev,
4669 struct netdev_nested_priv *priv),
4670 struct netdev_nested_priv *priv);
4671
4672bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4673 struct net_device *upper_dev);
4674
4675bool netdev_has_any_upper_dev(struct net_device *dev);
4676
4677void *netdev_lower_get_next_private(struct net_device *dev,
4678 struct list_head **iter);
4679void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4680 struct list_head **iter);
4681
4682#define netdev_for_each_lower_private(dev, priv, iter) \
4683 for (iter = (dev)->adj_list.lower.next, \
4684 priv = netdev_lower_get_next_private(dev, &(iter)); \
4685 priv; \
4686 priv = netdev_lower_get_next_private(dev, &(iter)))
4687
4688#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4689 for (iter = &(dev)->adj_list.lower, \
4690 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4691 priv; \
4692 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4693
4694void *netdev_lower_get_next(struct net_device *dev,
4695 struct list_head **iter);
4696
4697#define netdev_for_each_lower_dev(dev, ldev, iter) \
4698 for (iter = (dev)->adj_list.lower.next, \
4699 ldev = netdev_lower_get_next(dev, &(iter)); \
4700 ldev; \
4701 ldev = netdev_lower_get_next(dev, &(iter)))
4702
4703struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4704 struct list_head **iter);
4705int netdev_walk_all_lower_dev(struct net_device *dev,
4706 int (*fn)(struct net_device *lower_dev,
4707 struct netdev_nested_priv *priv),
4708 struct netdev_nested_priv *priv);
4709int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4710 int (*fn)(struct net_device *lower_dev,
4711 struct netdev_nested_priv *priv),
4712 struct netdev_nested_priv *priv);
4713
4714void *netdev_adjacent_get_private(struct list_head *adj_list);
4715void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4716struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4717struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4718int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4719 struct netlink_ext_ack *extack);
4720int netdev_master_upper_dev_link(struct net_device *dev,
4721 struct net_device *upper_dev,
4722 void *upper_priv, void *upper_info,
4723 struct netlink_ext_ack *extack);
4724void netdev_upper_dev_unlink(struct net_device *dev,
4725 struct net_device *upper_dev);
4726int netdev_adjacent_change_prepare(struct net_device *old_dev,
4727 struct net_device *new_dev,
4728 struct net_device *dev,
4729 struct netlink_ext_ack *extack);
4730void netdev_adjacent_change_commit(struct net_device *old_dev,
4731 struct net_device *new_dev,
4732 struct net_device *dev);
4733void netdev_adjacent_change_abort(struct net_device *old_dev,
4734 struct net_device *new_dev,
4735 struct net_device *dev);
4736void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4737void *netdev_lower_dev_get_private(struct net_device *dev,
4738 struct net_device *lower_dev);
4739void netdev_lower_state_changed(struct net_device *lower_dev,
4740 void *lower_state_info);
4741
4742/* RSS keys are 40 or 52 bytes long */
4743#define NETDEV_RSS_KEY_LEN 52
4744extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4745void netdev_rss_key_fill(void *buffer, size_t len);
4746
4747int skb_checksum_help(struct sk_buff *skb);
4748int skb_crc32c_csum_help(struct sk_buff *skb);
4749int skb_csum_hwoffload_help(struct sk_buff *skb,
4750 const netdev_features_t features);
4751
4752struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4753 netdev_features_t features, bool tx_path);
4754struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4755 netdev_features_t features, __be16 type);
4756struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4757 netdev_features_t features);
4758
4759struct netdev_bonding_info {
4760 ifslave slave;
4761 ifbond master;
4762};
4763
4764struct netdev_notifier_bonding_info {
4765 struct netdev_notifier_info info; /* must be first */
4766 struct netdev_bonding_info bonding_info;
4767};
4768
4769void netdev_bonding_info_change(struct net_device *dev,
4770 struct netdev_bonding_info *bonding_info);
4771
4772#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4773void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4774#else
4775static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4776 const void *data)
4777{
4778}
4779#endif
4780
4781static inline
4782struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4783{
4784 return __skb_gso_segment(skb, features, true);
4785}
4786__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4787
4788static inline bool can_checksum_protocol(netdev_features_t features,
4789 __be16 protocol)
4790{
4791 if (protocol == htons(ETH_P_FCOE))
4792 return !!(features & NETIF_F_FCOE_CRC);
4793
4794 /* Assume this is an IP checksum (not SCTP CRC) */
4795
4796 if (features & NETIF_F_HW_CSUM) {
4797 /* Can checksum everything */
4798 return true;
4799 }
4800
4801 switch (protocol) {
4802 case htons(ETH_P_IP):
4803 return !!(features & NETIF_F_IP_CSUM);
4804 case htons(ETH_P_IPV6):
4805 return !!(features & NETIF_F_IPV6_CSUM);
4806 default:
4807 return false;
4808 }
4809}
4810
4811#ifdef CONFIG_BUG
4812void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4813#else
4814static inline void netdev_rx_csum_fault(struct net_device *dev,
4815 struct sk_buff *skb)
4816{
4817}
4818#endif
4819/* rx skb timestamps */
4820void net_enable_timestamp(void);
4821void net_disable_timestamp(void);
4822
4823static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4824 const struct skb_shared_hwtstamps *hwtstamps,
4825 bool cycles)
4826{
4827 const struct net_device_ops *ops = dev->netdev_ops;
4828
4829 if (ops->ndo_get_tstamp)
4830 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4831
4832 return hwtstamps->hwtstamp;
4833}
4834
4835static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4836 struct sk_buff *skb, struct net_device *dev,
4837 bool more)
4838{
4839 __this_cpu_write(softnet_data.xmit.more, more);
4840 return ops->ndo_start_xmit(skb, dev);
4841}
4842
4843static inline bool netdev_xmit_more(void)
4844{
4845 return __this_cpu_read(softnet_data.xmit.more);
4846}
4847
4848static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4849 struct netdev_queue *txq, bool more)
4850{
4851 const struct net_device_ops *ops = dev->netdev_ops;
4852 netdev_tx_t rc;
4853
4854 rc = __netdev_start_xmit(ops, skb, dev, more);
4855 if (rc == NETDEV_TX_OK)
4856 txq_trans_update(txq);
4857
4858 return rc;
4859}
4860
4861int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4862 const void *ns);
4863void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4864 const void *ns);
4865
4866extern const struct kobj_ns_type_operations net_ns_type_operations;
4867
4868const char *netdev_drivername(const struct net_device *dev);
4869
4870static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4871 netdev_features_t f2)
4872{
4873 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4874 if (f1 & NETIF_F_HW_CSUM)
4875 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4876 else
4877 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4878 }
4879
4880 return f1 & f2;
4881}
4882
4883static inline netdev_features_t netdev_get_wanted_features(
4884 struct net_device *dev)
4885{
4886 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4887}
4888netdev_features_t netdev_increment_features(netdev_features_t all,
4889 netdev_features_t one, netdev_features_t mask);
4890
4891/* Allow TSO being used on stacked device :
4892 * Performing the GSO segmentation before last device
4893 * is a performance improvement.
4894 */
4895static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4896 netdev_features_t mask)
4897{
4898 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4899}
4900
4901int __netdev_update_features(struct net_device *dev);
4902void netdev_update_features(struct net_device *dev);
4903void netdev_change_features(struct net_device *dev);
4904
4905void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4906 struct net_device *dev);
4907
4908netdev_features_t passthru_features_check(struct sk_buff *skb,
4909 struct net_device *dev,
4910 netdev_features_t features);
4911netdev_features_t netif_skb_features(struct sk_buff *skb);
4912
4913static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4914{
4915 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4916
4917 /* check flags correspondence */
4918 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4919 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4920 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4921 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4922 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4923 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4924 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4925 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4926 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4927 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4928 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4929 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4930 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4931 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4932 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4933 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4934 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4935 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4936 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4937
4938 return (features & feature) == feature;
4939}
4940
4941static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4942{
4943 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4944 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4945}
4946
4947static inline bool netif_needs_gso(struct sk_buff *skb,
4948 netdev_features_t features)
4949{
4950 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4951 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4952 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4953}
4954
4955void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4956void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
4957void netif_inherit_tso_max(struct net_device *to,
4958 const struct net_device *from);
4959
4960static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4961 int pulled_hlen, u16 mac_offset,
4962 int mac_len)
4963{
4964 skb->protocol = protocol;
4965 skb->encapsulation = 1;
4966 skb_push(skb, pulled_hlen);
4967 skb_reset_transport_header(skb);
4968 skb->mac_header = mac_offset;
4969 skb->network_header = skb->mac_header + mac_len;
4970 skb->mac_len = mac_len;
4971}
4972
4973static inline bool netif_is_macsec(const struct net_device *dev)
4974{
4975 return dev->priv_flags & IFF_MACSEC;
4976}
4977
4978static inline bool netif_is_macvlan(const struct net_device *dev)
4979{
4980 return dev->priv_flags & IFF_MACVLAN;
4981}
4982
4983static inline bool netif_is_macvlan_port(const struct net_device *dev)
4984{
4985 return dev->priv_flags & IFF_MACVLAN_PORT;
4986}
4987
4988static inline bool netif_is_bond_master(const struct net_device *dev)
4989{
4990 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4991}
4992
4993static inline bool netif_is_bond_slave(const struct net_device *dev)
4994{
4995 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4996}
4997
4998static inline bool netif_supports_nofcs(struct net_device *dev)
4999{
5000 return dev->priv_flags & IFF_SUPP_NOFCS;
5001}
5002
5003static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5004{
5005 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5006}
5007
5008static inline bool netif_is_l3_master(const struct net_device *dev)
5009{
5010 return dev->priv_flags & IFF_L3MDEV_MASTER;
5011}
5012
5013static inline bool netif_is_l3_slave(const struct net_device *dev)
5014{
5015 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5016}
5017
5018static inline bool netif_is_bridge_master(const struct net_device *dev)
5019{
5020 return dev->priv_flags & IFF_EBRIDGE;
5021}
5022
5023static inline bool netif_is_bridge_port(const struct net_device *dev)
5024{
5025 return dev->priv_flags & IFF_BRIDGE_PORT;
5026}
5027
5028static inline bool netif_is_ovs_master(const struct net_device *dev)
5029{
5030 return dev->priv_flags & IFF_OPENVSWITCH;
5031}
5032
5033static inline bool netif_is_ovs_port(const struct net_device *dev)
5034{
5035 return dev->priv_flags & IFF_OVS_DATAPATH;
5036}
5037
5038static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5039{
5040 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5041}
5042
5043static inline bool netif_is_team_master(const struct net_device *dev)
5044{
5045 return dev->priv_flags & IFF_TEAM;
5046}
5047
5048static inline bool netif_is_team_port(const struct net_device *dev)
5049{
5050 return dev->priv_flags & IFF_TEAM_PORT;
5051}
5052
5053static inline bool netif_is_lag_master(const struct net_device *dev)
5054{
5055 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5056}
5057
5058static inline bool netif_is_lag_port(const struct net_device *dev)
5059{
5060 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5061}
5062
5063static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5064{
5065 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5066}
5067
5068static inline bool netif_is_failover(const struct net_device *dev)
5069{
5070 return dev->priv_flags & IFF_FAILOVER;
5071}
5072
5073static inline bool netif_is_failover_slave(const struct net_device *dev)
5074{
5075 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5076}
5077
5078/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5079static inline void netif_keep_dst(struct net_device *dev)
5080{
5081 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5082}
5083
5084/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5085static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5086{
5087 /* TODO: reserve and use an additional IFF bit, if we get more users */
5088 return netif_is_macsec(dev);
5089}
5090
5091extern struct pernet_operations __net_initdata loopback_net_ops;
5092
5093/* Logging, debugging and troubleshooting/diagnostic helpers. */
5094
5095/* netdev_printk helpers, similar to dev_printk */
5096
5097static inline const char *netdev_name(const struct net_device *dev)
5098{
5099 if (!dev->name[0] || strchr(dev->name, '%'))
5100 return "(unnamed net_device)";
5101 return dev->name;
5102}
5103
5104static inline bool netdev_unregistering(const struct net_device *dev)
5105{
5106 return dev->reg_state == NETREG_UNREGISTERING;
5107}
5108
5109static inline const char *netdev_reg_state(const struct net_device *dev)
5110{
5111 switch (dev->reg_state) {
5112 case NETREG_UNINITIALIZED: return " (uninitialized)";
5113 case NETREG_REGISTERED: return "";
5114 case NETREG_UNREGISTERING: return " (unregistering)";
5115 case NETREG_UNREGISTERED: return " (unregistered)";
5116 case NETREG_RELEASED: return " (released)";
5117 case NETREG_DUMMY: return " (dummy)";
5118 }
5119
5120 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5121 return " (unknown)";
5122}
5123
5124#define MODULE_ALIAS_NETDEV(device) \
5125 MODULE_ALIAS("netdev-" device)
5126
5127/*
5128 * netdev_WARN() acts like dev_printk(), but with the key difference
5129 * of using a WARN/WARN_ON to get the message out, including the
5130 * file/line information and a backtrace.
5131 */
5132#define netdev_WARN(dev, format, args...) \
5133 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5134 netdev_reg_state(dev), ##args)
5135
5136#define netdev_WARN_ONCE(dev, format, args...) \
5137 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5138 netdev_reg_state(dev), ##args)
5139
5140/*
5141 * The list of packet types we will receive (as opposed to discard)
5142 * and the routines to invoke.
5143 *
5144 * Why 16. Because with 16 the only overlap we get on a hash of the
5145 * low nibble of the protocol value is RARP/SNAP/X.25.
5146 *
5147 * 0800 IP
5148 * 0001 802.3
5149 * 0002 AX.25
5150 * 0004 802.2
5151 * 8035 RARP
5152 * 0005 SNAP
5153 * 0805 X.25
5154 * 0806 ARP
5155 * 8137 IPX
5156 * 0009 Localtalk
5157 * 86DD IPv6
5158 */
5159#define PTYPE_HASH_SIZE (16)
5160#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5161
5162extern struct list_head ptype_all __read_mostly;
5163extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5164
5165extern struct net_device *blackhole_netdev;
5166
5167#endif /* _LINUX_NETDEVICE_H */