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 <uapi/linux/netdev.h>
51#include <linux/hashtable.h>
52#include <linux/rbtree.h>
53#include <net/net_trackers.h>
54#include <net/net_debug.h>
55#include <net/dropreason-core.h>
56
57struct netpoll_info;
58struct device;
59struct ethtool_ops;
60struct phy_device;
61struct dsa_port;
62struct ip_tunnel_parm;
63struct macsec_context;
64struct macsec_ops;
65struct netdev_name_node;
66struct sd_flow_limit;
67struct sfp_bus;
68/* 802.11 specific */
69struct wireless_dev;
70/* 802.15.4 specific */
71struct wpan_dev;
72struct mpls_dev;
73/* UDP Tunnel offloads */
74struct udp_tunnel_info;
75struct udp_tunnel_nic_info;
76struct udp_tunnel_nic;
77struct bpf_prog;
78struct xdp_buff;
79struct xdp_md;
80
81void synchronize_net(void);
82void netdev_set_default_ethtool_ops(struct net_device *dev,
83 const struct ethtool_ops *ops);
84void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
85
86/* Backlog congestion levels */
87#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
88#define NET_RX_DROP 1 /* packet dropped */
89
90#define MAX_NEST_DEV 8
91
92/*
93 * Transmit return codes: transmit return codes originate from three different
94 * namespaces:
95 *
96 * - qdisc return codes
97 * - driver transmit return codes
98 * - errno values
99 *
100 * Drivers are allowed to return any one of those in their hard_start_xmit()
101 * function. Real network devices commonly used with qdiscs should only return
102 * the driver transmit return codes though - when qdiscs are used, the actual
103 * transmission happens asynchronously, so the value is not propagated to
104 * higher layers. Virtual network devices transmit synchronously; in this case
105 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
106 * others are propagated to higher layers.
107 */
108
109/* qdisc ->enqueue() return codes. */
110#define NET_XMIT_SUCCESS 0x00
111#define NET_XMIT_DROP 0x01 /* skb dropped */
112#define NET_XMIT_CN 0x02 /* congestion notification */
113#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
114
115/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
116 * indicates that the device will soon be dropping packets, or already drops
117 * some packets of the same priority; prompting us to send less aggressively. */
118#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
119#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
120
121/* Driver transmit return codes */
122#define NETDEV_TX_MASK 0xf0
123
124enum netdev_tx {
125 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
126 NETDEV_TX_OK = 0x00, /* driver took care of packet */
127 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
128};
129typedef enum netdev_tx netdev_tx_t;
130
131/*
132 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
133 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
134 */
135static inline bool dev_xmit_complete(int rc)
136{
137 /*
138 * Positive cases with an skb consumed by a driver:
139 * - successful transmission (rc == NETDEV_TX_OK)
140 * - error while transmitting (rc < 0)
141 * - error while queueing to a different device (rc & NET_XMIT_MASK)
142 */
143 if (likely(rc < NET_XMIT_MASK))
144 return true;
145
146 return false;
147}
148
149/*
150 * Compute the worst-case header length according to the protocols
151 * used.
152 */
153
154#if defined(CONFIG_HYPERV_NET)
155# define LL_MAX_HEADER 128
156#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
157# if defined(CONFIG_MAC80211_MESH)
158# define LL_MAX_HEADER 128
159# else
160# define LL_MAX_HEADER 96
161# endif
162#else
163# define LL_MAX_HEADER 32
164#endif
165
166#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
167 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
168#define MAX_HEADER LL_MAX_HEADER
169#else
170#define MAX_HEADER (LL_MAX_HEADER + 48)
171#endif
172
173/*
174 * Old network device statistics. Fields are native words
175 * (unsigned long) so they can be read and written atomically.
176 */
177
178#define NET_DEV_STAT(FIELD) \
179 union { \
180 unsigned long FIELD; \
181 atomic_long_t __##FIELD; \
182 }
183
184struct net_device_stats {
185 NET_DEV_STAT(rx_packets);
186 NET_DEV_STAT(tx_packets);
187 NET_DEV_STAT(rx_bytes);
188 NET_DEV_STAT(tx_bytes);
189 NET_DEV_STAT(rx_errors);
190 NET_DEV_STAT(tx_errors);
191 NET_DEV_STAT(rx_dropped);
192 NET_DEV_STAT(tx_dropped);
193 NET_DEV_STAT(multicast);
194 NET_DEV_STAT(collisions);
195 NET_DEV_STAT(rx_length_errors);
196 NET_DEV_STAT(rx_over_errors);
197 NET_DEV_STAT(rx_crc_errors);
198 NET_DEV_STAT(rx_frame_errors);
199 NET_DEV_STAT(rx_fifo_errors);
200 NET_DEV_STAT(rx_missed_errors);
201 NET_DEV_STAT(tx_aborted_errors);
202 NET_DEV_STAT(tx_carrier_errors);
203 NET_DEV_STAT(tx_fifo_errors);
204 NET_DEV_STAT(tx_heartbeat_errors);
205 NET_DEV_STAT(tx_window_errors);
206 NET_DEV_STAT(rx_compressed);
207 NET_DEV_STAT(tx_compressed);
208};
209#undef NET_DEV_STAT
210
211/* per-cpu stats, allocated on demand.
212 * Try to fit them in a single cache line, for dev_get_stats() sake.
213 */
214struct net_device_core_stats {
215 unsigned long rx_dropped;
216 unsigned long tx_dropped;
217 unsigned long rx_nohandler;
218 unsigned long rx_otherhost_dropped;
219} __aligned(4 * sizeof(unsigned long));
220
221#include <linux/cache.h>
222#include <linux/skbuff.h>
223
224#ifdef CONFIG_RPS
225#include <linux/static_key.h>
226extern struct static_key_false rps_needed;
227extern struct static_key_false rfs_needed;
228#endif
229
230struct neighbour;
231struct neigh_parms;
232struct sk_buff;
233
234struct netdev_hw_addr {
235 struct list_head list;
236 struct rb_node node;
237 unsigned char addr[MAX_ADDR_LEN];
238 unsigned char type;
239#define NETDEV_HW_ADDR_T_LAN 1
240#define NETDEV_HW_ADDR_T_SAN 2
241#define NETDEV_HW_ADDR_T_UNICAST 3
242#define NETDEV_HW_ADDR_T_MULTICAST 4
243 bool global_use;
244 int sync_cnt;
245 int refcount;
246 int synced;
247 struct rcu_head rcu_head;
248};
249
250struct netdev_hw_addr_list {
251 struct list_head list;
252 int count;
253
254 /* Auxiliary tree for faster lookup on addition and deletion */
255 struct rb_root tree;
256};
257
258#define netdev_hw_addr_list_count(l) ((l)->count)
259#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
260#define netdev_hw_addr_list_for_each(ha, l) \
261 list_for_each_entry(ha, &(l)->list, list)
262
263#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
264#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
265#define netdev_for_each_uc_addr(ha, dev) \
266 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
267#define netdev_for_each_synced_uc_addr(_ha, _dev) \
268 netdev_for_each_uc_addr((_ha), (_dev)) \
269 if ((_ha)->sync_cnt)
270
271#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
272#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
273#define netdev_for_each_mc_addr(ha, dev) \
274 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
275#define netdev_for_each_synced_mc_addr(_ha, _dev) \
276 netdev_for_each_mc_addr((_ha), (_dev)) \
277 if ((_ha)->sync_cnt)
278
279struct hh_cache {
280 unsigned int hh_len;
281 seqlock_t hh_lock;
282
283 /* cached hardware header; allow for machine alignment needs. */
284#define HH_DATA_MOD 16
285#define HH_DATA_OFF(__len) \
286 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
287#define HH_DATA_ALIGN(__len) \
288 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
289 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
290};
291
292/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
293 * Alternative is:
294 * dev->hard_header_len ? (dev->hard_header_len +
295 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
296 *
297 * We could use other alignment values, but we must maintain the
298 * relationship HH alignment <= LL alignment.
299 */
300#define LL_RESERVED_SPACE(dev) \
301 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
302 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
303#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
306
307struct header_ops {
308 int (*create) (struct sk_buff *skb, struct net_device *dev,
309 unsigned short type, const void *daddr,
310 const void *saddr, unsigned int len);
311 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
312 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
313 void (*cache_update)(struct hh_cache *hh,
314 const struct net_device *dev,
315 const unsigned char *haddr);
316 bool (*validate)(const char *ll_header, unsigned int len);
317 __be16 (*parse_protocol)(const struct sk_buff *skb);
318};
319
320/* These flag bits are private to the generic network queueing
321 * layer; they may not be explicitly referenced by any other
322 * code.
323 */
324
325enum netdev_state_t {
326 __LINK_STATE_START,
327 __LINK_STATE_PRESENT,
328 __LINK_STATE_NOCARRIER,
329 __LINK_STATE_LINKWATCH_PENDING,
330 __LINK_STATE_DORMANT,
331 __LINK_STATE_TESTING,
332};
333
334struct gro_list {
335 struct list_head list;
336 int count;
337};
338
339/*
340 * size of gro hash buckets, must less than bit number of
341 * napi_struct::gro_bitmask
342 */
343#define GRO_HASH_BUCKETS 8
344
345/*
346 * Structure for NAPI scheduling similar to tasklet but with weighting
347 */
348struct napi_struct {
349 /* The poll_list must only be managed by the entity which
350 * changes the state of the NAPI_STATE_SCHED bit. This means
351 * whoever atomically sets that bit can add this napi_struct
352 * to the per-CPU poll_list, and whoever clears that bit
353 * can remove from the list right before clearing the bit.
354 */
355 struct list_head poll_list;
356
357 unsigned long state;
358 int weight;
359 int defer_hard_irqs_count;
360 unsigned long gro_bitmask;
361 int (*poll)(struct napi_struct *, int);
362#ifdef CONFIG_NETPOLL
363 /* CPU actively polling if netpoll is configured */
364 int poll_owner;
365#endif
366 /* CPU on which NAPI has been scheduled for processing */
367 int list_owner;
368 struct net_device *dev;
369 struct gro_list gro_hash[GRO_HASH_BUCKETS];
370 struct sk_buff *skb;
371 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
372 int rx_count; /* length of rx_list */
373 unsigned int napi_id;
374 struct hrtimer timer;
375 struct task_struct *thread;
376 /* control-path-only fields follow */
377 struct list_head dev_list;
378 struct hlist_node napi_hash_node;
379};
380
381enum {
382 NAPI_STATE_SCHED, /* Poll is scheduled */
383 NAPI_STATE_MISSED, /* reschedule a napi */
384 NAPI_STATE_DISABLE, /* Disable pending */
385 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
386 NAPI_STATE_LISTED, /* NAPI added to system lists */
387 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
388 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
389 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
390 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
391 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
392};
393
394enum {
395 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
396 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
397 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
398 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
399 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
400 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
401 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
402 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
403 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
404 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
405};
406
407enum gro_result {
408 GRO_MERGED,
409 GRO_MERGED_FREE,
410 GRO_HELD,
411 GRO_NORMAL,
412 GRO_CONSUMED,
413};
414typedef enum gro_result gro_result_t;
415
416/*
417 * enum rx_handler_result - Possible return values for rx_handlers.
418 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
419 * further.
420 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
421 * case skb->dev was changed by rx_handler.
422 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
423 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
424 *
425 * rx_handlers are functions called from inside __netif_receive_skb(), to do
426 * special processing of the skb, prior to delivery to protocol handlers.
427 *
428 * Currently, a net_device can only have a single rx_handler registered. Trying
429 * to register a second rx_handler will return -EBUSY.
430 *
431 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
432 * To unregister a rx_handler on a net_device, use
433 * netdev_rx_handler_unregister().
434 *
435 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
436 * do with the skb.
437 *
438 * If the rx_handler consumed the skb in some way, it should return
439 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
440 * the skb to be delivered in some other way.
441 *
442 * If the rx_handler changed skb->dev, to divert the skb to another
443 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
444 * new device will be called if it exists.
445 *
446 * If the rx_handler decides the skb should be ignored, it should return
447 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
448 * are registered on exact device (ptype->dev == skb->dev).
449 *
450 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
451 * delivered, it should return RX_HANDLER_PASS.
452 *
453 * A device without a registered rx_handler will behave as if rx_handler
454 * returned RX_HANDLER_PASS.
455 */
456
457enum rx_handler_result {
458 RX_HANDLER_CONSUMED,
459 RX_HANDLER_ANOTHER,
460 RX_HANDLER_EXACT,
461 RX_HANDLER_PASS,
462};
463typedef enum rx_handler_result rx_handler_result_t;
464typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
465
466void __napi_schedule(struct napi_struct *n);
467void __napi_schedule_irqoff(struct napi_struct *n);
468
469static inline bool napi_disable_pending(struct napi_struct *n)
470{
471 return test_bit(NAPI_STATE_DISABLE, &n->state);
472}
473
474static inline bool napi_prefer_busy_poll(struct napi_struct *n)
475{
476 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
477}
478
479bool napi_schedule_prep(struct napi_struct *n);
480
481/**
482 * napi_schedule - schedule NAPI poll
483 * @n: NAPI context
484 *
485 * Schedule NAPI poll routine to be called if it is not already
486 * running.
487 */
488static inline void napi_schedule(struct napi_struct *n)
489{
490 if (napi_schedule_prep(n))
491 __napi_schedule(n);
492}
493
494/**
495 * napi_schedule_irqoff - schedule NAPI poll
496 * @n: NAPI context
497 *
498 * Variant of napi_schedule(), assuming hard irqs are masked.
499 */
500static inline void napi_schedule_irqoff(struct napi_struct *n)
501{
502 if (napi_schedule_prep(n))
503 __napi_schedule_irqoff(n);
504}
505
506/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
507static inline bool napi_reschedule(struct napi_struct *napi)
508{
509 if (napi_schedule_prep(napi)) {
510 __napi_schedule(napi);
511 return true;
512 }
513 return false;
514}
515
516/**
517 * napi_complete_done - NAPI processing complete
518 * @n: NAPI context
519 * @work_done: number of packets processed
520 *
521 * Mark NAPI processing as complete. Should only be called if poll budget
522 * has not been completely consumed.
523 * Prefer over napi_complete().
524 * Return false if device should avoid rearming interrupts.
525 */
526bool napi_complete_done(struct napi_struct *n, int work_done);
527
528static inline bool napi_complete(struct napi_struct *n)
529{
530 return napi_complete_done(n, 0);
531}
532
533int dev_set_threaded(struct net_device *dev, bool threaded);
534
535/**
536 * napi_disable - prevent NAPI from scheduling
537 * @n: NAPI context
538 *
539 * Stop NAPI from being scheduled on this context.
540 * Waits till any outstanding processing completes.
541 */
542void napi_disable(struct napi_struct *n);
543
544void napi_enable(struct napi_struct *n);
545
546/**
547 * napi_synchronize - wait until NAPI is not running
548 * @n: NAPI context
549 *
550 * Wait until NAPI is done being scheduled on this context.
551 * Waits till any outstanding processing completes but
552 * does not disable future activations.
553 */
554static inline void napi_synchronize(const struct napi_struct *n)
555{
556 if (IS_ENABLED(CONFIG_SMP))
557 while (test_bit(NAPI_STATE_SCHED, &n->state))
558 msleep(1);
559 else
560 barrier();
561}
562
563/**
564 * napi_if_scheduled_mark_missed - if napi is running, set the
565 * NAPIF_STATE_MISSED
566 * @n: NAPI context
567 *
568 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
569 * NAPI is scheduled.
570 **/
571static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
572{
573 unsigned long val, new;
574
575 val = READ_ONCE(n->state);
576 do {
577 if (val & NAPIF_STATE_DISABLE)
578 return true;
579
580 if (!(val & NAPIF_STATE_SCHED))
581 return false;
582
583 new = val | NAPIF_STATE_MISSED;
584 } while (!try_cmpxchg(&n->state, &val, new));
585
586 return true;
587}
588
589enum netdev_queue_state_t {
590 __QUEUE_STATE_DRV_XOFF,
591 __QUEUE_STATE_STACK_XOFF,
592 __QUEUE_STATE_FROZEN,
593};
594
595#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
596#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
597#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
598
599#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
600#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
601 QUEUE_STATE_FROZEN)
602#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
603 QUEUE_STATE_FROZEN)
604
605/*
606 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
607 * netif_tx_* functions below are used to manipulate this flag. The
608 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
609 * queue independently. The netif_xmit_*stopped functions below are called
610 * to check if the queue has been stopped by the driver or stack (either
611 * of the XOFF bits are set in the state). Drivers should not need to call
612 * netif_xmit*stopped functions, they should only be using netif_tx_*.
613 */
614
615struct netdev_queue {
616/*
617 * read-mostly part
618 */
619 struct net_device *dev;
620 netdevice_tracker dev_tracker;
621
622 struct Qdisc __rcu *qdisc;
623 struct Qdisc *qdisc_sleeping;
624#ifdef CONFIG_SYSFS
625 struct kobject kobj;
626#endif
627#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
628 int numa_node;
629#endif
630 unsigned long tx_maxrate;
631 /*
632 * Number of TX timeouts for this queue
633 * (/sys/class/net/DEV/Q/trans_timeout)
634 */
635 atomic_long_t trans_timeout;
636
637 /* Subordinate device that the queue has been assigned to */
638 struct net_device *sb_dev;
639#ifdef CONFIG_XDP_SOCKETS
640 struct xsk_buff_pool *pool;
641#endif
642/*
643 * write-mostly part
644 */
645 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
646 int xmit_lock_owner;
647 /*
648 * Time (in jiffies) of last Tx
649 */
650 unsigned long trans_start;
651
652 unsigned long state;
653
654#ifdef CONFIG_BQL
655 struct dql dql;
656#endif
657} ____cacheline_aligned_in_smp;
658
659extern int sysctl_fb_tunnels_only_for_init_net;
660extern int sysctl_devconf_inherit_init_net;
661
662/*
663 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
664 * == 1 : For initns only
665 * == 2 : For none.
666 */
667static inline bool net_has_fallback_tunnels(const struct net *net)
668{
669#if IS_ENABLED(CONFIG_SYSCTL)
670 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
671
672 return !fb_tunnels_only_for_init_net ||
673 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
674#else
675 return true;
676#endif
677}
678
679static inline int net_inherit_devconf(void)
680{
681#if IS_ENABLED(CONFIG_SYSCTL)
682 return READ_ONCE(sysctl_devconf_inherit_init_net);
683#else
684 return 0;
685#endif
686}
687
688static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
689{
690#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
691 return q->numa_node;
692#else
693 return NUMA_NO_NODE;
694#endif
695}
696
697static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
698{
699#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
700 q->numa_node = node;
701#endif
702}
703
704#ifdef CONFIG_RPS
705/*
706 * This structure holds an RPS map which can be of variable length. The
707 * map is an array of CPUs.
708 */
709struct rps_map {
710 unsigned int len;
711 struct rcu_head rcu;
712 u16 cpus[];
713};
714#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
715
716/*
717 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
718 * tail pointer for that CPU's input queue at the time of last enqueue, and
719 * a hardware filter index.
720 */
721struct rps_dev_flow {
722 u16 cpu;
723 u16 filter;
724 unsigned int last_qtail;
725};
726#define RPS_NO_FILTER 0xffff
727
728/*
729 * The rps_dev_flow_table structure contains a table of flow mappings.
730 */
731struct rps_dev_flow_table {
732 unsigned int mask;
733 struct rcu_head rcu;
734 struct rps_dev_flow flows[];
735};
736#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
737 ((_num) * sizeof(struct rps_dev_flow)))
738
739/*
740 * The rps_sock_flow_table contains mappings of flows to the last CPU
741 * on which they were processed by the application (set in recvmsg).
742 * Each entry is a 32bit value. Upper part is the high-order bits
743 * of flow hash, lower part is CPU number.
744 * rps_cpu_mask is used to partition the space, depending on number of
745 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
746 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
747 * meaning we use 32-6=26 bits for the hash.
748 */
749struct rps_sock_flow_table {
750 u32 mask;
751
752 u32 ents[] ____cacheline_aligned_in_smp;
753};
754#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
755
756#define RPS_NO_CPU 0xffff
757
758extern u32 rps_cpu_mask;
759extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
760
761static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
762 u32 hash)
763{
764 if (table && hash) {
765 unsigned int index = hash & table->mask;
766 u32 val = hash & ~rps_cpu_mask;
767
768 /* We only give a hint, preemption can change CPU under us */
769 val |= raw_smp_processor_id();
770
771 if (table->ents[index] != val)
772 table->ents[index] = val;
773 }
774}
775
776#ifdef CONFIG_RFS_ACCEL
777bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
778 u16 filter_id);
779#endif
780#endif /* CONFIG_RPS */
781
782/* This structure contains an instance of an RX queue. */
783struct netdev_rx_queue {
784 struct xdp_rxq_info xdp_rxq;
785#ifdef CONFIG_RPS
786 struct rps_map __rcu *rps_map;
787 struct rps_dev_flow_table __rcu *rps_flow_table;
788#endif
789 struct kobject kobj;
790 struct net_device *dev;
791 netdevice_tracker dev_tracker;
792
793#ifdef CONFIG_XDP_SOCKETS
794 struct xsk_buff_pool *pool;
795#endif
796} ____cacheline_aligned_in_smp;
797
798/*
799 * RX queue sysfs structures and functions.
800 */
801struct rx_queue_attribute {
802 struct attribute attr;
803 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
804 ssize_t (*store)(struct netdev_rx_queue *queue,
805 const char *buf, size_t len);
806};
807
808/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
809enum xps_map_type {
810 XPS_CPUS = 0,
811 XPS_RXQS,
812 XPS_MAPS_MAX,
813};
814
815#ifdef CONFIG_XPS
816/*
817 * This structure holds an XPS map which can be of variable length. The
818 * map is an array of queues.
819 */
820struct xps_map {
821 unsigned int len;
822 unsigned int alloc_len;
823 struct rcu_head rcu;
824 u16 queues[];
825};
826#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
827#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
828 - sizeof(struct xps_map)) / sizeof(u16))
829
830/*
831 * This structure holds all XPS maps for device. Maps are indexed by CPU.
832 *
833 * We keep track of the number of cpus/rxqs used when the struct is allocated,
834 * in nr_ids. This will help not accessing out-of-bound memory.
835 *
836 * We keep track of the number of traffic classes used when the struct is
837 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
838 * not crossing its upper bound, as the original dev->num_tc can be updated in
839 * the meantime.
840 */
841struct xps_dev_maps {
842 struct rcu_head rcu;
843 unsigned int nr_ids;
844 s16 num_tc;
845 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
846};
847
848#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
849 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
850
851#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
852 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
853
854#endif /* CONFIG_XPS */
855
856#define TC_MAX_QUEUE 16
857#define TC_BITMASK 15
858/* HW offloaded queuing disciplines txq count and offset maps */
859struct netdev_tc_txq {
860 u16 count;
861 u16 offset;
862};
863
864#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
865/*
866 * This structure is to hold information about the device
867 * configured to run FCoE protocol stack.
868 */
869struct netdev_fcoe_hbainfo {
870 char manufacturer[64];
871 char serial_number[64];
872 char hardware_version[64];
873 char driver_version[64];
874 char optionrom_version[64];
875 char firmware_version[64];
876 char model[256];
877 char model_description[256];
878};
879#endif
880
881#define MAX_PHYS_ITEM_ID_LEN 32
882
883/* This structure holds a unique identifier to identify some
884 * physical item (port for example) used by a netdevice.
885 */
886struct netdev_phys_item_id {
887 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
888 unsigned char id_len;
889};
890
891static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
892 struct netdev_phys_item_id *b)
893{
894 return a->id_len == b->id_len &&
895 memcmp(a->id, b->id, a->id_len) == 0;
896}
897
898typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
899 struct sk_buff *skb,
900 struct net_device *sb_dev);
901
902enum net_device_path_type {
903 DEV_PATH_ETHERNET = 0,
904 DEV_PATH_VLAN,
905 DEV_PATH_BRIDGE,
906 DEV_PATH_PPPOE,
907 DEV_PATH_DSA,
908 DEV_PATH_MTK_WDMA,
909};
910
911struct net_device_path {
912 enum net_device_path_type type;
913 const struct net_device *dev;
914 union {
915 struct {
916 u16 id;
917 __be16 proto;
918 u8 h_dest[ETH_ALEN];
919 } encap;
920 struct {
921 enum {
922 DEV_PATH_BR_VLAN_KEEP,
923 DEV_PATH_BR_VLAN_TAG,
924 DEV_PATH_BR_VLAN_UNTAG,
925 DEV_PATH_BR_VLAN_UNTAG_HW,
926 } vlan_mode;
927 u16 vlan_id;
928 __be16 vlan_proto;
929 } bridge;
930 struct {
931 int port;
932 u16 proto;
933 } dsa;
934 struct {
935 u8 wdma_idx;
936 u8 queue;
937 u16 wcid;
938 u8 bss;
939 } mtk_wdma;
940 };
941};
942
943#define NET_DEVICE_PATH_STACK_MAX 5
944#define NET_DEVICE_PATH_VLAN_MAX 2
945
946struct net_device_path_stack {
947 int num_paths;
948 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
949};
950
951struct net_device_path_ctx {
952 const struct net_device *dev;
953 u8 daddr[ETH_ALEN];
954
955 int num_vlans;
956 struct {
957 u16 id;
958 __be16 proto;
959 } vlan[NET_DEVICE_PATH_VLAN_MAX];
960};
961
962enum tc_setup_type {
963 TC_QUERY_CAPS,
964 TC_SETUP_QDISC_MQPRIO,
965 TC_SETUP_CLSU32,
966 TC_SETUP_CLSFLOWER,
967 TC_SETUP_CLSMATCHALL,
968 TC_SETUP_CLSBPF,
969 TC_SETUP_BLOCK,
970 TC_SETUP_QDISC_CBS,
971 TC_SETUP_QDISC_RED,
972 TC_SETUP_QDISC_PRIO,
973 TC_SETUP_QDISC_MQ,
974 TC_SETUP_QDISC_ETF,
975 TC_SETUP_ROOT_QDISC,
976 TC_SETUP_QDISC_GRED,
977 TC_SETUP_QDISC_TAPRIO,
978 TC_SETUP_FT,
979 TC_SETUP_QDISC_ETS,
980 TC_SETUP_QDISC_TBF,
981 TC_SETUP_QDISC_FIFO,
982 TC_SETUP_QDISC_HTB,
983 TC_SETUP_ACT,
984};
985
986/* These structures hold the attributes of bpf state that are being passed
987 * to the netdevice through the bpf op.
988 */
989enum bpf_netdev_command {
990 /* Set or clear a bpf program used in the earliest stages of packet
991 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
992 * is responsible for calling bpf_prog_put on any old progs that are
993 * stored. In case of error, the callee need not release the new prog
994 * reference, but on success it takes ownership and must bpf_prog_put
995 * when it is no longer used.
996 */
997 XDP_SETUP_PROG,
998 XDP_SETUP_PROG_HW,
999 /* BPF program for offload callbacks, invoked at program load time. */
1000 BPF_OFFLOAD_MAP_ALLOC,
1001 BPF_OFFLOAD_MAP_FREE,
1002 XDP_SETUP_XSK_POOL,
1003};
1004
1005struct bpf_prog_offload_ops;
1006struct netlink_ext_ack;
1007struct xdp_umem;
1008struct xdp_dev_bulk_queue;
1009struct bpf_xdp_link;
1010
1011enum bpf_xdp_mode {
1012 XDP_MODE_SKB = 0,
1013 XDP_MODE_DRV = 1,
1014 XDP_MODE_HW = 2,
1015 __MAX_XDP_MODE
1016};
1017
1018struct bpf_xdp_entity {
1019 struct bpf_prog *prog;
1020 struct bpf_xdp_link *link;
1021};
1022
1023struct netdev_bpf {
1024 enum bpf_netdev_command command;
1025 union {
1026 /* XDP_SETUP_PROG */
1027 struct {
1028 u32 flags;
1029 struct bpf_prog *prog;
1030 struct netlink_ext_ack *extack;
1031 };
1032 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1033 struct {
1034 struct bpf_offloaded_map *offmap;
1035 };
1036 /* XDP_SETUP_XSK_POOL */
1037 struct {
1038 struct xsk_buff_pool *pool;
1039 u16 queue_id;
1040 } xsk;
1041 };
1042};
1043
1044/* Flags for ndo_xsk_wakeup. */
1045#define XDP_WAKEUP_RX (1 << 0)
1046#define XDP_WAKEUP_TX (1 << 1)
1047
1048#ifdef CONFIG_XFRM_OFFLOAD
1049struct xfrmdev_ops {
1050 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1051 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1052 void (*xdo_dev_state_free) (struct xfrm_state *x);
1053 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1054 struct xfrm_state *x);
1055 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1056 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1057 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1058 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1059 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1060};
1061#endif
1062
1063struct dev_ifalias {
1064 struct rcu_head rcuhead;
1065 char ifalias[];
1066};
1067
1068struct devlink;
1069struct tlsdev_ops;
1070
1071struct netdev_net_notifier {
1072 struct list_head list;
1073 struct notifier_block *nb;
1074};
1075
1076/*
1077 * This structure defines the management hooks for network devices.
1078 * The following hooks can be defined; unless noted otherwise, they are
1079 * optional and can be filled with a null pointer.
1080 *
1081 * int (*ndo_init)(struct net_device *dev);
1082 * This function is called once when a network device is registered.
1083 * The network device can use this for any late stage initialization
1084 * or semantic validation. It can fail with an error code which will
1085 * be propagated back to register_netdev.
1086 *
1087 * void (*ndo_uninit)(struct net_device *dev);
1088 * This function is called when device is unregistered or when registration
1089 * fails. It is not called if init fails.
1090 *
1091 * int (*ndo_open)(struct net_device *dev);
1092 * This function is called when a network device transitions to the up
1093 * state.
1094 *
1095 * int (*ndo_stop)(struct net_device *dev);
1096 * This function is called when a network device transitions to the down
1097 * state.
1098 *
1099 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1100 * struct net_device *dev);
1101 * Called when a packet needs to be transmitted.
1102 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1103 * the queue before that can happen; it's for obsolete devices and weird
1104 * corner cases, but the stack really does a non-trivial amount
1105 * of useless work if you return NETDEV_TX_BUSY.
1106 * Required; cannot be NULL.
1107 *
1108 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1109 * struct net_device *dev
1110 * netdev_features_t features);
1111 * Called by core transmit path to determine if device is capable of
1112 * performing offload operations on a given packet. This is to give
1113 * the device an opportunity to implement any restrictions that cannot
1114 * be otherwise expressed by feature flags. The check is called with
1115 * the set of features that the stack has calculated and it returns
1116 * those the driver believes to be appropriate.
1117 *
1118 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1119 * struct net_device *sb_dev);
1120 * Called to decide which queue to use when device supports multiple
1121 * transmit queues.
1122 *
1123 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1124 * This function is called to allow device receiver to make
1125 * changes to configuration when multicast or promiscuous is enabled.
1126 *
1127 * void (*ndo_set_rx_mode)(struct net_device *dev);
1128 * This function is called device changes address list filtering.
1129 * If driver handles unicast address filtering, it should set
1130 * IFF_UNICAST_FLT in its priv_flags.
1131 *
1132 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1133 * This function is called when the Media Access Control address
1134 * needs to be changed. If this interface is not defined, the
1135 * MAC address can not be changed.
1136 *
1137 * int (*ndo_validate_addr)(struct net_device *dev);
1138 * Test if Media Access Control address is valid for the device.
1139 *
1140 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1141 * Old-style ioctl entry point. This is used internally by the
1142 * appletalk and ieee802154 subsystems but is no longer called by
1143 * the device ioctl handler.
1144 *
1145 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1146 * Used by the bonding driver for its device specific ioctls:
1147 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1148 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1149 *
1150 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1151 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1152 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1153 *
1154 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1155 * Used to set network devices bus interface parameters. This interface
1156 * is retained for legacy reasons; new devices should use the bus
1157 * interface (PCI) for low level management.
1158 *
1159 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1160 * Called when a user wants to change the Maximum Transfer Unit
1161 * of a device.
1162 *
1163 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1164 * Callback used when the transmitter has not made any progress
1165 * for dev->watchdog ticks.
1166 *
1167 * void (*ndo_get_stats64)(struct net_device *dev,
1168 * struct rtnl_link_stats64 *storage);
1169 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1170 * Called when a user wants to get the network device usage
1171 * statistics. Drivers must do one of the following:
1172 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1173 * rtnl_link_stats64 structure passed by the caller.
1174 * 2. Define @ndo_get_stats to update a net_device_stats structure
1175 * (which should normally be dev->stats) and return a pointer to
1176 * it. The structure may be changed asynchronously only if each
1177 * field is written atomically.
1178 * 3. Update dev->stats asynchronously and atomically, and define
1179 * neither operation.
1180 *
1181 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1182 * Return true if this device supports offload stats of this attr_id.
1183 *
1184 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1185 * void *attr_data)
1186 * Get statistics for offload operations by attr_id. Write it into the
1187 * attr_data pointer.
1188 *
1189 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1190 * If device supports VLAN filtering this function is called when a
1191 * VLAN id is registered.
1192 *
1193 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1194 * If device supports VLAN filtering this function is called when a
1195 * VLAN id is unregistered.
1196 *
1197 * void (*ndo_poll_controller)(struct net_device *dev);
1198 *
1199 * SR-IOV management functions.
1200 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1201 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1202 * u8 qos, __be16 proto);
1203 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1204 * int max_tx_rate);
1205 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1206 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1207 * int (*ndo_get_vf_config)(struct net_device *dev,
1208 * int vf, struct ifla_vf_info *ivf);
1209 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1210 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1211 * struct nlattr *port[]);
1212 *
1213 * Enable or disable the VF ability to query its RSS Redirection Table and
1214 * Hash Key. This is needed since on some devices VF share this information
1215 * with PF and querying it may introduce a theoretical security risk.
1216 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1217 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1218 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1219 * void *type_data);
1220 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1221 * This is always called from the stack with the rtnl lock held and netif
1222 * tx queues stopped. This allows the netdevice to perform queue
1223 * management safely.
1224 *
1225 * Fiber Channel over Ethernet (FCoE) offload functions.
1226 * int (*ndo_fcoe_enable)(struct net_device *dev);
1227 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1228 * so the underlying device can perform whatever needed configuration or
1229 * initialization to support acceleration of FCoE traffic.
1230 *
1231 * int (*ndo_fcoe_disable)(struct net_device *dev);
1232 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1233 * so the underlying device can perform whatever needed clean-ups to
1234 * stop supporting acceleration of FCoE traffic.
1235 *
1236 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1237 * struct scatterlist *sgl, unsigned int sgc);
1238 * Called when the FCoE Initiator wants to initialize an I/O that
1239 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1240 * perform necessary setup and returns 1 to indicate the device is set up
1241 * successfully to perform DDP on this I/O, otherwise this returns 0.
1242 *
1243 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1244 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1245 * indicated by the FC exchange id 'xid', so the underlying device can
1246 * clean up and reuse resources for later DDP requests.
1247 *
1248 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1249 * struct scatterlist *sgl, unsigned int sgc);
1250 * Called when the FCoE Target wants to initialize an I/O that
1251 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1252 * perform necessary setup and returns 1 to indicate the device is set up
1253 * successfully to perform DDP on this I/O, otherwise this returns 0.
1254 *
1255 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1256 * struct netdev_fcoe_hbainfo *hbainfo);
1257 * Called when the FCoE Protocol stack wants information on the underlying
1258 * device. This information is utilized by the FCoE protocol stack to
1259 * register attributes with Fiber Channel management service as per the
1260 * FC-GS Fabric Device Management Information(FDMI) specification.
1261 *
1262 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1263 * Called when the underlying device wants to override default World Wide
1264 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1265 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1266 * protocol stack to use.
1267 *
1268 * RFS acceleration.
1269 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1270 * u16 rxq_index, u32 flow_id);
1271 * Set hardware filter for RFS. rxq_index is the target queue index;
1272 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1273 * Return the filter ID on success, or a negative error code.
1274 *
1275 * Slave management functions (for bridge, bonding, etc).
1276 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1277 * Called to make another netdev an underling.
1278 *
1279 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1280 * Called to release previously enslaved netdev.
1281 *
1282 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1283 * struct sk_buff *skb,
1284 * bool all_slaves);
1285 * Get the xmit slave of master device. If all_slaves is true, function
1286 * assume all the slaves can transmit.
1287 *
1288 * Feature/offload setting functions.
1289 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1290 * netdev_features_t features);
1291 * Adjusts the requested feature flags according to device-specific
1292 * constraints, and returns the resulting flags. Must not modify
1293 * the device state.
1294 *
1295 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1296 * Called to update device configuration to new features. Passed
1297 * feature set might be less than what was returned by ndo_fix_features()).
1298 * Must return >0 or -errno if it changed dev->features itself.
1299 *
1300 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1301 * struct net_device *dev,
1302 * const unsigned char *addr, u16 vid, u16 flags,
1303 * struct netlink_ext_ack *extack);
1304 * Adds an FDB entry to dev for addr.
1305 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1306 * struct net_device *dev,
1307 * const unsigned char *addr, u16 vid)
1308 * Deletes the FDB entry from dev coresponding to addr.
1309 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1310 * struct net_device *dev,
1311 * u16 vid,
1312 * struct netlink_ext_ack *extack);
1313 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1314 * struct net_device *dev, struct net_device *filter_dev,
1315 * int *idx)
1316 * Used to add FDB entries to dump requests. Implementers should add
1317 * entries to skb and update idx with the number of entries.
1318 *
1319 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1320 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1321 * Adds an MDB entry to dev.
1322 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1323 * struct netlink_ext_ack *extack);
1324 * Deletes the MDB entry from dev.
1325 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1326 * struct netlink_callback *cb);
1327 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1328 * callback is used by core rtnetlink code.
1329 *
1330 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1331 * u16 flags, struct netlink_ext_ack *extack)
1332 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1333 * struct net_device *dev, u32 filter_mask,
1334 * int nlflags)
1335 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1336 * u16 flags);
1337 *
1338 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1339 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1340 * which do not represent real hardware may define this to allow their
1341 * userspace components to manage their virtual carrier state. Devices
1342 * that determine carrier state from physical hardware properties (eg
1343 * network cables) or protocol-dependent mechanisms (eg
1344 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1345 *
1346 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1347 * struct netdev_phys_item_id *ppid);
1348 * Called to get ID of physical port of this device. If driver does
1349 * not implement this, it is assumed that the hw is not able to have
1350 * multiple net devices on single physical port.
1351 *
1352 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1353 * struct netdev_phys_item_id *ppid)
1354 * Called to get the parent ID of the physical port of this device.
1355 *
1356 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1357 * struct net_device *dev)
1358 * Called by upper layer devices to accelerate switching or other
1359 * station functionality into hardware. 'pdev is the lowerdev
1360 * to use for the offload and 'dev' is the net device that will
1361 * back the offload. Returns a pointer to the private structure
1362 * the upper layer will maintain.
1363 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1364 * Called by upper layer device to delete the station created
1365 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1366 * the station and priv is the structure returned by the add
1367 * operation.
1368 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1369 * int queue_index, u32 maxrate);
1370 * Called when a user wants to set a max-rate limitation of specific
1371 * TX queue.
1372 * int (*ndo_get_iflink)(const struct net_device *dev);
1373 * Called to get the iflink value of this device.
1374 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1375 * This function is used to get egress tunnel information for given skb.
1376 * This is useful for retrieving outer tunnel header parameters while
1377 * sampling packet.
1378 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1379 * This function is used to specify the headroom that the skb must
1380 * consider when allocation skb during packet reception. Setting
1381 * appropriate rx headroom value allows avoiding skb head copy on
1382 * forward. Setting a negative value resets the rx headroom to the
1383 * default value.
1384 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1385 * This function is used to set or query state related to XDP on the
1386 * netdevice and manage BPF offload. See definition of
1387 * enum bpf_netdev_command for details.
1388 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1389 * u32 flags);
1390 * This function is used to submit @n XDP packets for transmit on a
1391 * netdevice. Returns number of frames successfully transmitted, frames
1392 * that got dropped are freed/returned via xdp_return_frame().
1393 * Returns negative number, means general error invoking ndo, meaning
1394 * no frames were xmit'ed and core-caller will free all frames.
1395 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1396 * struct xdp_buff *xdp);
1397 * Get the xmit slave of master device based on the xdp_buff.
1398 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1399 * This function is used to wake up the softirq, ksoftirqd or kthread
1400 * responsible for sending and/or receiving packets on a specific
1401 * queue id bound to an AF_XDP socket. The flags field specifies if
1402 * only RX, only Tx, or both should be woken up using the flags
1403 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1404 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1405 * int cmd);
1406 * Add, change, delete or get information on an IPv4 tunnel.
1407 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1408 * If a device is paired with a peer device, return the peer instance.
1409 * The caller must be under RCU read context.
1410 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1411 * Get the forwarding path to reach the real device from the HW destination address
1412 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1413 * const struct skb_shared_hwtstamps *hwtstamps,
1414 * bool cycles);
1415 * Get hardware timestamp based on normal/adjustable time or free running
1416 * cycle counter. This function is required if physical clock supports a
1417 * free running cycle counter.
1418 */
1419struct net_device_ops {
1420 int (*ndo_init)(struct net_device *dev);
1421 void (*ndo_uninit)(struct net_device *dev);
1422 int (*ndo_open)(struct net_device *dev);
1423 int (*ndo_stop)(struct net_device *dev);
1424 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1425 struct net_device *dev);
1426 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1427 struct net_device *dev,
1428 netdev_features_t features);
1429 u16 (*ndo_select_queue)(struct net_device *dev,
1430 struct sk_buff *skb,
1431 struct net_device *sb_dev);
1432 void (*ndo_change_rx_flags)(struct net_device *dev,
1433 int flags);
1434 void (*ndo_set_rx_mode)(struct net_device *dev);
1435 int (*ndo_set_mac_address)(struct net_device *dev,
1436 void *addr);
1437 int (*ndo_validate_addr)(struct net_device *dev);
1438 int (*ndo_do_ioctl)(struct net_device *dev,
1439 struct ifreq *ifr, int cmd);
1440 int (*ndo_eth_ioctl)(struct net_device *dev,
1441 struct ifreq *ifr, int cmd);
1442 int (*ndo_siocbond)(struct net_device *dev,
1443 struct ifreq *ifr, int cmd);
1444 int (*ndo_siocwandev)(struct net_device *dev,
1445 struct if_settings *ifs);
1446 int (*ndo_siocdevprivate)(struct net_device *dev,
1447 struct ifreq *ifr,
1448 void __user *data, int cmd);
1449 int (*ndo_set_config)(struct net_device *dev,
1450 struct ifmap *map);
1451 int (*ndo_change_mtu)(struct net_device *dev,
1452 int new_mtu);
1453 int (*ndo_neigh_setup)(struct net_device *dev,
1454 struct neigh_parms *);
1455 void (*ndo_tx_timeout) (struct net_device *dev,
1456 unsigned int txqueue);
1457
1458 void (*ndo_get_stats64)(struct net_device *dev,
1459 struct rtnl_link_stats64 *storage);
1460 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1461 int (*ndo_get_offload_stats)(int attr_id,
1462 const struct net_device *dev,
1463 void *attr_data);
1464 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1465
1466 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1467 __be16 proto, u16 vid);
1468 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1469 __be16 proto, u16 vid);
1470#ifdef CONFIG_NET_POLL_CONTROLLER
1471 void (*ndo_poll_controller)(struct net_device *dev);
1472 int (*ndo_netpoll_setup)(struct net_device *dev,
1473 struct netpoll_info *info);
1474 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1475#endif
1476 int (*ndo_set_vf_mac)(struct net_device *dev,
1477 int queue, u8 *mac);
1478 int (*ndo_set_vf_vlan)(struct net_device *dev,
1479 int queue, u16 vlan,
1480 u8 qos, __be16 proto);
1481 int (*ndo_set_vf_rate)(struct net_device *dev,
1482 int vf, int min_tx_rate,
1483 int max_tx_rate);
1484 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1485 int vf, bool setting);
1486 int (*ndo_set_vf_trust)(struct net_device *dev,
1487 int vf, bool setting);
1488 int (*ndo_get_vf_config)(struct net_device *dev,
1489 int vf,
1490 struct ifla_vf_info *ivf);
1491 int (*ndo_set_vf_link_state)(struct net_device *dev,
1492 int vf, int link_state);
1493 int (*ndo_get_vf_stats)(struct net_device *dev,
1494 int vf,
1495 struct ifla_vf_stats
1496 *vf_stats);
1497 int (*ndo_set_vf_port)(struct net_device *dev,
1498 int vf,
1499 struct nlattr *port[]);
1500 int (*ndo_get_vf_port)(struct net_device *dev,
1501 int vf, struct sk_buff *skb);
1502 int (*ndo_get_vf_guid)(struct net_device *dev,
1503 int vf,
1504 struct ifla_vf_guid *node_guid,
1505 struct ifla_vf_guid *port_guid);
1506 int (*ndo_set_vf_guid)(struct net_device *dev,
1507 int vf, u64 guid,
1508 int guid_type);
1509 int (*ndo_set_vf_rss_query_en)(
1510 struct net_device *dev,
1511 int vf, bool setting);
1512 int (*ndo_setup_tc)(struct net_device *dev,
1513 enum tc_setup_type type,
1514 void *type_data);
1515#if IS_ENABLED(CONFIG_FCOE)
1516 int (*ndo_fcoe_enable)(struct net_device *dev);
1517 int (*ndo_fcoe_disable)(struct net_device *dev);
1518 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1519 u16 xid,
1520 struct scatterlist *sgl,
1521 unsigned int sgc);
1522 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1523 u16 xid);
1524 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1525 u16 xid,
1526 struct scatterlist *sgl,
1527 unsigned int sgc);
1528 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1529 struct netdev_fcoe_hbainfo *hbainfo);
1530#endif
1531
1532#if IS_ENABLED(CONFIG_LIBFCOE)
1533#define NETDEV_FCOE_WWNN 0
1534#define NETDEV_FCOE_WWPN 1
1535 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1536 u64 *wwn, int type);
1537#endif
1538
1539#ifdef CONFIG_RFS_ACCEL
1540 int (*ndo_rx_flow_steer)(struct net_device *dev,
1541 const struct sk_buff *skb,
1542 u16 rxq_index,
1543 u32 flow_id);
1544#endif
1545 int (*ndo_add_slave)(struct net_device *dev,
1546 struct net_device *slave_dev,
1547 struct netlink_ext_ack *extack);
1548 int (*ndo_del_slave)(struct net_device *dev,
1549 struct net_device *slave_dev);
1550 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1551 struct sk_buff *skb,
1552 bool all_slaves);
1553 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1554 struct sock *sk);
1555 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1556 netdev_features_t features);
1557 int (*ndo_set_features)(struct net_device *dev,
1558 netdev_features_t features);
1559 int (*ndo_neigh_construct)(struct net_device *dev,
1560 struct neighbour *n);
1561 void (*ndo_neigh_destroy)(struct net_device *dev,
1562 struct neighbour *n);
1563
1564 int (*ndo_fdb_add)(struct ndmsg *ndm,
1565 struct nlattr *tb[],
1566 struct net_device *dev,
1567 const unsigned char *addr,
1568 u16 vid,
1569 u16 flags,
1570 struct netlink_ext_ack *extack);
1571 int (*ndo_fdb_del)(struct ndmsg *ndm,
1572 struct nlattr *tb[],
1573 struct net_device *dev,
1574 const unsigned char *addr,
1575 u16 vid, struct netlink_ext_ack *extack);
1576 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1577 struct nlattr *tb[],
1578 struct net_device *dev,
1579 u16 vid,
1580 struct netlink_ext_ack *extack);
1581 int (*ndo_fdb_dump)(struct sk_buff *skb,
1582 struct netlink_callback *cb,
1583 struct net_device *dev,
1584 struct net_device *filter_dev,
1585 int *idx);
1586 int (*ndo_fdb_get)(struct sk_buff *skb,
1587 struct nlattr *tb[],
1588 struct net_device *dev,
1589 const unsigned char *addr,
1590 u16 vid, u32 portid, u32 seq,
1591 struct netlink_ext_ack *extack);
1592 int (*ndo_mdb_add)(struct net_device *dev,
1593 struct nlattr *tb[],
1594 u16 nlmsg_flags,
1595 struct netlink_ext_ack *extack);
1596 int (*ndo_mdb_del)(struct net_device *dev,
1597 struct nlattr *tb[],
1598 struct netlink_ext_ack *extack);
1599 int (*ndo_mdb_dump)(struct net_device *dev,
1600 struct sk_buff *skb,
1601 struct netlink_callback *cb);
1602 int (*ndo_bridge_setlink)(struct net_device *dev,
1603 struct nlmsghdr *nlh,
1604 u16 flags,
1605 struct netlink_ext_ack *extack);
1606 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1607 u32 pid, u32 seq,
1608 struct net_device *dev,
1609 u32 filter_mask,
1610 int nlflags);
1611 int (*ndo_bridge_dellink)(struct net_device *dev,
1612 struct nlmsghdr *nlh,
1613 u16 flags);
1614 int (*ndo_change_carrier)(struct net_device *dev,
1615 bool new_carrier);
1616 int (*ndo_get_phys_port_id)(struct net_device *dev,
1617 struct netdev_phys_item_id *ppid);
1618 int (*ndo_get_port_parent_id)(struct net_device *dev,
1619 struct netdev_phys_item_id *ppid);
1620 int (*ndo_get_phys_port_name)(struct net_device *dev,
1621 char *name, size_t len);
1622 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1623 struct net_device *dev);
1624 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1625 void *priv);
1626
1627 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1628 int queue_index,
1629 u32 maxrate);
1630 int (*ndo_get_iflink)(const struct net_device *dev);
1631 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1632 struct sk_buff *skb);
1633 void (*ndo_set_rx_headroom)(struct net_device *dev,
1634 int needed_headroom);
1635 int (*ndo_bpf)(struct net_device *dev,
1636 struct netdev_bpf *bpf);
1637 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1638 struct xdp_frame **xdp,
1639 u32 flags);
1640 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1641 struct xdp_buff *xdp);
1642 int (*ndo_xsk_wakeup)(struct net_device *dev,
1643 u32 queue_id, u32 flags);
1644 int (*ndo_tunnel_ctl)(struct net_device *dev,
1645 struct ip_tunnel_parm *p, int cmd);
1646 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1647 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1648 struct net_device_path *path);
1649 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1650 const struct skb_shared_hwtstamps *hwtstamps,
1651 bool cycles);
1652};
1653
1654struct xdp_metadata_ops {
1655 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp);
1656 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash,
1657 enum xdp_rss_hash_type *rss_type);
1658};
1659
1660/**
1661 * enum netdev_priv_flags - &struct net_device priv_flags
1662 *
1663 * These are the &struct net_device, they are only set internally
1664 * by drivers and used in the kernel. These flags are invisible to
1665 * userspace; this means that the order of these flags can change
1666 * during any kernel release.
1667 *
1668 * You should have a pretty good reason to be extending these flags.
1669 *
1670 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1671 * @IFF_EBRIDGE: Ethernet bridging device
1672 * @IFF_BONDING: bonding master or slave
1673 * @IFF_ISATAP: ISATAP interface (RFC4214)
1674 * @IFF_WAN_HDLC: WAN HDLC device
1675 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1676 * release skb->dst
1677 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1678 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1679 * @IFF_MACVLAN_PORT: device used as macvlan port
1680 * @IFF_BRIDGE_PORT: device used as bridge port
1681 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1682 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1683 * @IFF_UNICAST_FLT: Supports unicast filtering
1684 * @IFF_TEAM_PORT: device used as team port
1685 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1686 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1687 * change when it's running
1688 * @IFF_MACVLAN: Macvlan device
1689 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1690 * underlying stacked devices
1691 * @IFF_L3MDEV_MASTER: device is an L3 master device
1692 * @IFF_NO_QUEUE: device can run without qdisc attached
1693 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1694 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1695 * @IFF_TEAM: device is a team device
1696 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1697 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1698 * entity (i.e. the master device for bridged veth)
1699 * @IFF_MACSEC: device is a MACsec device
1700 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1701 * @IFF_FAILOVER: device is a failover master device
1702 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1703 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1704 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1705 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1706 * skb_headlen(skb) == 0 (data starts from frag0)
1707 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1708 */
1709enum netdev_priv_flags {
1710 IFF_802_1Q_VLAN = 1<<0,
1711 IFF_EBRIDGE = 1<<1,
1712 IFF_BONDING = 1<<2,
1713 IFF_ISATAP = 1<<3,
1714 IFF_WAN_HDLC = 1<<4,
1715 IFF_XMIT_DST_RELEASE = 1<<5,
1716 IFF_DONT_BRIDGE = 1<<6,
1717 IFF_DISABLE_NETPOLL = 1<<7,
1718 IFF_MACVLAN_PORT = 1<<8,
1719 IFF_BRIDGE_PORT = 1<<9,
1720 IFF_OVS_DATAPATH = 1<<10,
1721 IFF_TX_SKB_SHARING = 1<<11,
1722 IFF_UNICAST_FLT = 1<<12,
1723 IFF_TEAM_PORT = 1<<13,
1724 IFF_SUPP_NOFCS = 1<<14,
1725 IFF_LIVE_ADDR_CHANGE = 1<<15,
1726 IFF_MACVLAN = 1<<16,
1727 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1728 IFF_L3MDEV_MASTER = 1<<18,
1729 IFF_NO_QUEUE = 1<<19,
1730 IFF_OPENVSWITCH = 1<<20,
1731 IFF_L3MDEV_SLAVE = 1<<21,
1732 IFF_TEAM = 1<<22,
1733 IFF_RXFH_CONFIGURED = 1<<23,
1734 IFF_PHONY_HEADROOM = 1<<24,
1735 IFF_MACSEC = 1<<25,
1736 IFF_NO_RX_HANDLER = 1<<26,
1737 IFF_FAILOVER = 1<<27,
1738 IFF_FAILOVER_SLAVE = 1<<28,
1739 IFF_L3MDEV_RX_HANDLER = 1<<29,
1740 IFF_NO_ADDRCONF = BIT_ULL(30),
1741 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1742 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1743};
1744
1745#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1746#define IFF_EBRIDGE IFF_EBRIDGE
1747#define IFF_BONDING IFF_BONDING
1748#define IFF_ISATAP IFF_ISATAP
1749#define IFF_WAN_HDLC IFF_WAN_HDLC
1750#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1751#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1752#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1753#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1754#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1755#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1756#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1757#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1758#define IFF_TEAM_PORT IFF_TEAM_PORT
1759#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1760#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1761#define IFF_MACVLAN IFF_MACVLAN
1762#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1763#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1764#define IFF_NO_QUEUE IFF_NO_QUEUE
1765#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1766#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1767#define IFF_TEAM IFF_TEAM
1768#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1769#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1770#define IFF_MACSEC IFF_MACSEC
1771#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1772#define IFF_FAILOVER IFF_FAILOVER
1773#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1774#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1775#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1776
1777/* Specifies the type of the struct net_device::ml_priv pointer */
1778enum netdev_ml_priv_type {
1779 ML_PRIV_NONE,
1780 ML_PRIV_CAN,
1781};
1782
1783/**
1784 * struct net_device - The DEVICE structure.
1785 *
1786 * Actually, this whole structure is a big mistake. It mixes I/O
1787 * data with strictly "high-level" data, and it has to know about
1788 * almost every data structure used in the INET module.
1789 *
1790 * @name: This is the first field of the "visible" part of this structure
1791 * (i.e. as seen by users in the "Space.c" file). It is the name
1792 * of the interface.
1793 *
1794 * @name_node: Name hashlist node
1795 * @ifalias: SNMP alias
1796 * @mem_end: Shared memory end
1797 * @mem_start: Shared memory start
1798 * @base_addr: Device I/O address
1799 * @irq: Device IRQ number
1800 *
1801 * @state: Generic network queuing layer state, see netdev_state_t
1802 * @dev_list: The global list of network devices
1803 * @napi_list: List entry used for polling NAPI devices
1804 * @unreg_list: List entry when we are unregistering the
1805 * device; see the function unregister_netdev
1806 * @close_list: List entry used when we are closing the device
1807 * @ptype_all: Device-specific packet handlers for all protocols
1808 * @ptype_specific: Device-specific, protocol-specific packet handlers
1809 *
1810 * @adj_list: Directly linked devices, like slaves for bonding
1811 * @features: Currently active device features
1812 * @hw_features: User-changeable features
1813 *
1814 * @wanted_features: User-requested features
1815 * @vlan_features: Mask of features inheritable by VLAN devices
1816 *
1817 * @hw_enc_features: Mask of features inherited by encapsulating devices
1818 * This field indicates what encapsulation
1819 * offloads the hardware is capable of doing,
1820 * and drivers will need to set them appropriately.
1821 *
1822 * @mpls_features: Mask of features inheritable by MPLS
1823 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1824 *
1825 * @ifindex: interface index
1826 * @group: The group the device belongs to
1827 *
1828 * @stats: Statistics struct, which was left as a legacy, use
1829 * rtnl_link_stats64 instead
1830 *
1831 * @core_stats: core networking counters,
1832 * do not use this in drivers
1833 * @carrier_up_count: Number of times the carrier has been up
1834 * @carrier_down_count: Number of times the carrier has been down
1835 *
1836 * @wireless_handlers: List of functions to handle Wireless Extensions,
1837 * instead of ioctl,
1838 * see <net/iw_handler.h> for details.
1839 * @wireless_data: Instance data managed by the core of wireless extensions
1840 *
1841 * @netdev_ops: Includes several pointers to callbacks,
1842 * if one wants to override the ndo_*() functions
1843 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1844 * @ethtool_ops: Management operations
1845 * @l3mdev_ops: Layer 3 master device operations
1846 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1847 * discovery handling. Necessary for e.g. 6LoWPAN.
1848 * @xfrmdev_ops: Transformation offload operations
1849 * @tlsdev_ops: Transport Layer Security offload operations
1850 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1851 * of Layer 2 headers.
1852 *
1853 * @flags: Interface flags (a la BSD)
1854 * @xdp_features: XDP capability supported by the device
1855 * @priv_flags: Like 'flags' but invisible to userspace,
1856 * see if.h for the definitions
1857 * @gflags: Global flags ( kept as legacy )
1858 * @padded: How much padding added by alloc_netdev()
1859 * @operstate: RFC2863 operstate
1860 * @link_mode: Mapping policy to operstate
1861 * @if_port: Selectable AUI, TP, ...
1862 * @dma: DMA channel
1863 * @mtu: Interface MTU value
1864 * @min_mtu: Interface Minimum MTU value
1865 * @max_mtu: Interface Maximum MTU value
1866 * @type: Interface hardware type
1867 * @hard_header_len: Maximum hardware header length.
1868 * @min_header_len: Minimum hardware header length
1869 *
1870 * @needed_headroom: Extra headroom the hardware may need, but not in all
1871 * cases can this be guaranteed
1872 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1873 * cases can this be guaranteed. Some cases also use
1874 * LL_MAX_HEADER instead to allocate the skb
1875 *
1876 * interface address info:
1877 *
1878 * @perm_addr: Permanent hw address
1879 * @addr_assign_type: Hw address assignment type
1880 * @addr_len: Hardware address length
1881 * @upper_level: Maximum depth level of upper devices.
1882 * @lower_level: Maximum depth level of lower devices.
1883 * @neigh_priv_len: Used in neigh_alloc()
1884 * @dev_id: Used to differentiate devices that share
1885 * the same link layer address
1886 * @dev_port: Used to differentiate devices that share
1887 * the same function
1888 * @addr_list_lock: XXX: need comments on this one
1889 * @name_assign_type: network interface name assignment type
1890 * @uc_promisc: Counter that indicates promiscuous mode
1891 * has been enabled due to the need to listen to
1892 * additional unicast addresses in a device that
1893 * does not implement ndo_set_rx_mode()
1894 * @uc: unicast mac addresses
1895 * @mc: multicast mac addresses
1896 * @dev_addrs: list of device hw addresses
1897 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1898 * @promiscuity: Number of times the NIC is told to work in
1899 * promiscuous mode; if it becomes 0 the NIC will
1900 * exit promiscuous mode
1901 * @allmulti: Counter, enables or disables allmulticast mode
1902 *
1903 * @vlan_info: VLAN info
1904 * @dsa_ptr: dsa specific data
1905 * @tipc_ptr: TIPC specific data
1906 * @atalk_ptr: AppleTalk link
1907 * @ip_ptr: IPv4 specific data
1908 * @ip6_ptr: IPv6 specific data
1909 * @ax25_ptr: AX.25 specific data
1910 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1911 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1912 * device struct
1913 * @mpls_ptr: mpls_dev struct pointer
1914 * @mctp_ptr: MCTP specific data
1915 *
1916 * @dev_addr: Hw address (before bcast,
1917 * because most packets are unicast)
1918 *
1919 * @_rx: Array of RX queues
1920 * @num_rx_queues: Number of RX queues
1921 * allocated at register_netdev() time
1922 * @real_num_rx_queues: Number of RX queues currently active in device
1923 * @xdp_prog: XDP sockets filter program pointer
1924 * @gro_flush_timeout: timeout for GRO layer in NAPI
1925 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1926 * allow to avoid NIC hard IRQ, on busy queues.
1927 *
1928 * @rx_handler: handler for received packets
1929 * @rx_handler_data: XXX: need comments on this one
1930 * @miniq_ingress: ingress/clsact qdisc specific data for
1931 * ingress processing
1932 * @ingress_queue: XXX: need comments on this one
1933 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1934 * @broadcast: hw bcast address
1935 *
1936 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1937 * indexed by RX queue number. Assigned by driver.
1938 * This must only be set if the ndo_rx_flow_steer
1939 * operation is defined
1940 * @index_hlist: Device index hash chain
1941 *
1942 * @_tx: Array of TX queues
1943 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1944 * @real_num_tx_queues: Number of TX queues currently active in device
1945 * @qdisc: Root qdisc from userspace point of view
1946 * @tx_queue_len: Max frames per queue allowed
1947 * @tx_global_lock: XXX: need comments on this one
1948 * @xdp_bulkq: XDP device bulk queue
1949 * @xps_maps: all CPUs/RXQs maps for XPS device
1950 *
1951 * @xps_maps: XXX: need comments on this one
1952 * @miniq_egress: clsact qdisc specific data for
1953 * egress processing
1954 * @nf_hooks_egress: netfilter hooks executed for egress packets
1955 * @qdisc_hash: qdisc hash table
1956 * @watchdog_timeo: Represents the timeout that is used by
1957 * the watchdog (see dev_watchdog())
1958 * @watchdog_timer: List of timers
1959 *
1960 * @proto_down_reason: reason a netdev interface is held down
1961 * @pcpu_refcnt: Number of references to this device
1962 * @dev_refcnt: Number of references to this device
1963 * @refcnt_tracker: Tracker directory for tracked references to this device
1964 * @todo_list: Delayed register/unregister
1965 * @link_watch_list: XXX: need comments on this one
1966 *
1967 * @reg_state: Register/unregister state machine
1968 * @dismantle: Device is going to be freed
1969 * @rtnl_link_state: This enum represents the phases of creating
1970 * a new link
1971 *
1972 * @needs_free_netdev: Should unregister perform free_netdev?
1973 * @priv_destructor: Called from unregister
1974 * @npinfo: XXX: need comments on this one
1975 * @nd_net: Network namespace this network device is inside
1976 *
1977 * @ml_priv: Mid-layer private
1978 * @ml_priv_type: Mid-layer private type
1979 * @lstats: Loopback statistics
1980 * @tstats: Tunnel statistics
1981 * @dstats: Dummy statistics
1982 * @vstats: Virtual ethernet statistics
1983 *
1984 * @garp_port: GARP
1985 * @mrp_port: MRP
1986 *
1987 * @dm_private: Drop monitor private
1988 *
1989 * @dev: Class/net/name entry
1990 * @sysfs_groups: Space for optional device, statistics and wireless
1991 * sysfs groups
1992 *
1993 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1994 * @rtnl_link_ops: Rtnl_link_ops
1995 *
1996 * @gso_max_size: Maximum size of generic segmentation offload
1997 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1998 * @gso_max_segs: Maximum number of segments that can be passed to the
1999 * NIC for GSO
2000 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2001 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2002 * for IPv4.
2003 *
2004 * @dcbnl_ops: Data Center Bridging netlink ops
2005 * @num_tc: Number of traffic classes in the net device
2006 * @tc_to_txq: XXX: need comments on this one
2007 * @prio_tc_map: XXX: need comments on this one
2008 *
2009 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2010 *
2011 * @priomap: XXX: need comments on this one
2012 * @phydev: Physical device may attach itself
2013 * for hardware timestamping
2014 * @sfp_bus: attached &struct sfp_bus structure.
2015 *
2016 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2017 *
2018 * @proto_down: protocol port state information can be sent to the
2019 * switch driver and used to set the phys state of the
2020 * switch port.
2021 *
2022 * @wol_enabled: Wake-on-LAN is enabled
2023 *
2024 * @threaded: napi threaded mode is enabled
2025 *
2026 * @net_notifier_list: List of per-net netdev notifier block
2027 * that follow this device when it is moved
2028 * to another network namespace.
2029 *
2030 * @macsec_ops: MACsec offloading ops
2031 *
2032 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2033 * offload capabilities of the device
2034 * @udp_tunnel_nic: UDP tunnel offload state
2035 * @xdp_state: stores info on attached XDP BPF programs
2036 *
2037 * @nested_level: Used as a parameter of spin_lock_nested() of
2038 * dev->addr_list_lock.
2039 * @unlink_list: As netif_addr_lock() can be called recursively,
2040 * keep a list of interfaces to be deleted.
2041 * @gro_max_size: Maximum size of aggregated packet in generic
2042 * receive offload (GRO)
2043 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2044 * receive offload (GRO), for IPv4.
2045 *
2046 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2047 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2048 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2049 * @dev_registered_tracker: tracker for reference held while
2050 * registered
2051 * @offload_xstats_l3: L3 HW stats for this netdevice.
2052 *
2053 * @devlink_port: Pointer to related devlink port structure.
2054 * Assigned by a driver before netdev registration using
2055 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2056 * during the time netdevice is registered.
2057 *
2058 * FIXME: cleanup struct net_device such that network protocol info
2059 * moves out.
2060 */
2061
2062struct net_device {
2063 char name[IFNAMSIZ];
2064 struct netdev_name_node *name_node;
2065 struct dev_ifalias __rcu *ifalias;
2066 /*
2067 * I/O specific fields
2068 * FIXME: Merge these and struct ifmap into one
2069 */
2070 unsigned long mem_end;
2071 unsigned long mem_start;
2072 unsigned long base_addr;
2073
2074 /*
2075 * Some hardware also needs these fields (state,dev_list,
2076 * napi_list,unreg_list,close_list) but they are not
2077 * part of the usual set specified in Space.c.
2078 */
2079
2080 unsigned long state;
2081
2082 struct list_head dev_list;
2083 struct list_head napi_list;
2084 struct list_head unreg_list;
2085 struct list_head close_list;
2086 struct list_head ptype_all;
2087 struct list_head ptype_specific;
2088
2089 struct {
2090 struct list_head upper;
2091 struct list_head lower;
2092 } adj_list;
2093
2094 /* Read-mostly cache-line for fast-path access */
2095 unsigned int flags;
2096 xdp_features_t xdp_features;
2097 unsigned long long priv_flags;
2098 const struct net_device_ops *netdev_ops;
2099 const struct xdp_metadata_ops *xdp_metadata_ops;
2100 int ifindex;
2101 unsigned short gflags;
2102 unsigned short hard_header_len;
2103
2104 /* Note : dev->mtu is often read without holding a lock.
2105 * Writers usually hold RTNL.
2106 * It is recommended to use READ_ONCE() to annotate the reads,
2107 * and to use WRITE_ONCE() to annotate the writes.
2108 */
2109 unsigned int mtu;
2110 unsigned short needed_headroom;
2111 unsigned short needed_tailroom;
2112
2113 netdev_features_t features;
2114 netdev_features_t hw_features;
2115 netdev_features_t wanted_features;
2116 netdev_features_t vlan_features;
2117 netdev_features_t hw_enc_features;
2118 netdev_features_t mpls_features;
2119 netdev_features_t gso_partial_features;
2120
2121 unsigned int min_mtu;
2122 unsigned int max_mtu;
2123 unsigned short type;
2124 unsigned char min_header_len;
2125 unsigned char name_assign_type;
2126
2127 int group;
2128
2129 struct net_device_stats stats; /* not used by modern drivers */
2130
2131 struct net_device_core_stats __percpu *core_stats;
2132
2133 /* Stats to monitor link on/off, flapping */
2134 atomic_t carrier_up_count;
2135 atomic_t carrier_down_count;
2136
2137#ifdef CONFIG_WIRELESS_EXT
2138 const struct iw_handler_def *wireless_handlers;
2139 struct iw_public_data *wireless_data;
2140#endif
2141 const struct ethtool_ops *ethtool_ops;
2142#ifdef CONFIG_NET_L3_MASTER_DEV
2143 const struct l3mdev_ops *l3mdev_ops;
2144#endif
2145#if IS_ENABLED(CONFIG_IPV6)
2146 const struct ndisc_ops *ndisc_ops;
2147#endif
2148
2149#ifdef CONFIG_XFRM_OFFLOAD
2150 const struct xfrmdev_ops *xfrmdev_ops;
2151#endif
2152
2153#if IS_ENABLED(CONFIG_TLS_DEVICE)
2154 const struct tlsdev_ops *tlsdev_ops;
2155#endif
2156
2157 const struct header_ops *header_ops;
2158
2159 unsigned char operstate;
2160 unsigned char link_mode;
2161
2162 unsigned char if_port;
2163 unsigned char dma;
2164
2165 /* Interface address info. */
2166 unsigned char perm_addr[MAX_ADDR_LEN];
2167 unsigned char addr_assign_type;
2168 unsigned char addr_len;
2169 unsigned char upper_level;
2170 unsigned char lower_level;
2171
2172 unsigned short neigh_priv_len;
2173 unsigned short dev_id;
2174 unsigned short dev_port;
2175 unsigned short padded;
2176
2177 spinlock_t addr_list_lock;
2178 int irq;
2179
2180 struct netdev_hw_addr_list uc;
2181 struct netdev_hw_addr_list mc;
2182 struct netdev_hw_addr_list dev_addrs;
2183
2184#ifdef CONFIG_SYSFS
2185 struct kset *queues_kset;
2186#endif
2187#ifdef CONFIG_LOCKDEP
2188 struct list_head unlink_list;
2189#endif
2190 unsigned int promiscuity;
2191 unsigned int allmulti;
2192 bool uc_promisc;
2193#ifdef CONFIG_LOCKDEP
2194 unsigned char nested_level;
2195#endif
2196
2197
2198 /* Protocol-specific pointers */
2199
2200 struct in_device __rcu *ip_ptr;
2201 struct inet6_dev __rcu *ip6_ptr;
2202#if IS_ENABLED(CONFIG_VLAN_8021Q)
2203 struct vlan_info __rcu *vlan_info;
2204#endif
2205#if IS_ENABLED(CONFIG_NET_DSA)
2206 struct dsa_port *dsa_ptr;
2207#endif
2208#if IS_ENABLED(CONFIG_TIPC)
2209 struct tipc_bearer __rcu *tipc_ptr;
2210#endif
2211#if IS_ENABLED(CONFIG_ATALK)
2212 void *atalk_ptr;
2213#endif
2214#if IS_ENABLED(CONFIG_AX25)
2215 void *ax25_ptr;
2216#endif
2217#if IS_ENABLED(CONFIG_CFG80211)
2218 struct wireless_dev *ieee80211_ptr;
2219#endif
2220#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2221 struct wpan_dev *ieee802154_ptr;
2222#endif
2223#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2224 struct mpls_dev __rcu *mpls_ptr;
2225#endif
2226#if IS_ENABLED(CONFIG_MCTP)
2227 struct mctp_dev __rcu *mctp_ptr;
2228#endif
2229
2230/*
2231 * Cache lines mostly used on receive path (including eth_type_trans())
2232 */
2233 /* Interface address info used in eth_type_trans() */
2234 const unsigned char *dev_addr;
2235
2236 struct netdev_rx_queue *_rx;
2237 unsigned int num_rx_queues;
2238 unsigned int real_num_rx_queues;
2239
2240 struct bpf_prog __rcu *xdp_prog;
2241 unsigned long gro_flush_timeout;
2242 int napi_defer_hard_irqs;
2243#define GRO_LEGACY_MAX_SIZE 65536u
2244/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2245 * and shinfo->gso_segs is a 16bit field.
2246 */
2247#define GRO_MAX_SIZE (8 * 65535u)
2248 unsigned int gro_max_size;
2249 unsigned int gro_ipv4_max_size;
2250 rx_handler_func_t __rcu *rx_handler;
2251 void __rcu *rx_handler_data;
2252
2253#ifdef CONFIG_NET_CLS_ACT
2254 struct mini_Qdisc __rcu *miniq_ingress;
2255#endif
2256 struct netdev_queue __rcu *ingress_queue;
2257#ifdef CONFIG_NETFILTER_INGRESS
2258 struct nf_hook_entries __rcu *nf_hooks_ingress;
2259#endif
2260
2261 unsigned char broadcast[MAX_ADDR_LEN];
2262#ifdef CONFIG_RFS_ACCEL
2263 struct cpu_rmap *rx_cpu_rmap;
2264#endif
2265 struct hlist_node index_hlist;
2266
2267/*
2268 * Cache lines mostly used on transmit path
2269 */
2270 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2271 unsigned int num_tx_queues;
2272 unsigned int real_num_tx_queues;
2273 struct Qdisc __rcu *qdisc;
2274 unsigned int tx_queue_len;
2275 spinlock_t tx_global_lock;
2276
2277 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2278
2279#ifdef CONFIG_XPS
2280 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2281#endif
2282#ifdef CONFIG_NET_CLS_ACT
2283 struct mini_Qdisc __rcu *miniq_egress;
2284#endif
2285#ifdef CONFIG_NETFILTER_EGRESS
2286 struct nf_hook_entries __rcu *nf_hooks_egress;
2287#endif
2288
2289#ifdef CONFIG_NET_SCHED
2290 DECLARE_HASHTABLE (qdisc_hash, 4);
2291#endif
2292 /* These may be needed for future network-power-down code. */
2293 struct timer_list watchdog_timer;
2294 int watchdog_timeo;
2295
2296 u32 proto_down_reason;
2297
2298 struct list_head todo_list;
2299
2300#ifdef CONFIG_PCPU_DEV_REFCNT
2301 int __percpu *pcpu_refcnt;
2302#else
2303 refcount_t dev_refcnt;
2304#endif
2305 struct ref_tracker_dir refcnt_tracker;
2306
2307 struct list_head link_watch_list;
2308
2309 enum { NETREG_UNINITIALIZED=0,
2310 NETREG_REGISTERED, /* completed register_netdevice */
2311 NETREG_UNREGISTERING, /* called unregister_netdevice */
2312 NETREG_UNREGISTERED, /* completed unregister todo */
2313 NETREG_RELEASED, /* called free_netdev */
2314 NETREG_DUMMY, /* dummy device for NAPI poll */
2315 } reg_state:8;
2316
2317 bool dismantle;
2318
2319 enum {
2320 RTNL_LINK_INITIALIZED,
2321 RTNL_LINK_INITIALIZING,
2322 } rtnl_link_state:16;
2323
2324 bool needs_free_netdev;
2325 void (*priv_destructor)(struct net_device *dev);
2326
2327#ifdef CONFIG_NETPOLL
2328 struct netpoll_info __rcu *npinfo;
2329#endif
2330
2331 possible_net_t nd_net;
2332
2333 /* mid-layer private */
2334 void *ml_priv;
2335 enum netdev_ml_priv_type ml_priv_type;
2336
2337 union {
2338 struct pcpu_lstats __percpu *lstats;
2339 struct pcpu_sw_netstats __percpu *tstats;
2340 struct pcpu_dstats __percpu *dstats;
2341 };
2342
2343#if IS_ENABLED(CONFIG_GARP)
2344 struct garp_port __rcu *garp_port;
2345#endif
2346#if IS_ENABLED(CONFIG_MRP)
2347 struct mrp_port __rcu *mrp_port;
2348#endif
2349#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2350 struct dm_hw_stat_delta __rcu *dm_private;
2351#endif
2352 struct device dev;
2353 const struct attribute_group *sysfs_groups[4];
2354 const struct attribute_group *sysfs_rx_queue_group;
2355
2356 const struct rtnl_link_ops *rtnl_link_ops;
2357
2358 /* for setting kernel sock attribute on TCP connection setup */
2359#define GSO_MAX_SEGS 65535u
2360#define GSO_LEGACY_MAX_SIZE 65536u
2361/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2362 * and shinfo->gso_segs is a 16bit field.
2363 */
2364#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2365
2366 unsigned int gso_max_size;
2367#define TSO_LEGACY_MAX_SIZE 65536
2368#define TSO_MAX_SIZE UINT_MAX
2369 unsigned int tso_max_size;
2370 u16 gso_max_segs;
2371#define TSO_MAX_SEGS U16_MAX
2372 u16 tso_max_segs;
2373 unsigned int gso_ipv4_max_size;
2374
2375#ifdef CONFIG_DCB
2376 const struct dcbnl_rtnl_ops *dcbnl_ops;
2377#endif
2378 s16 num_tc;
2379 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2380 u8 prio_tc_map[TC_BITMASK + 1];
2381
2382#if IS_ENABLED(CONFIG_FCOE)
2383 unsigned int fcoe_ddp_xid;
2384#endif
2385#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2386 struct netprio_map __rcu *priomap;
2387#endif
2388 struct phy_device *phydev;
2389 struct sfp_bus *sfp_bus;
2390 struct lock_class_key *qdisc_tx_busylock;
2391 bool proto_down;
2392 unsigned wol_enabled:1;
2393 unsigned threaded:1;
2394
2395 struct list_head net_notifier_list;
2396
2397#if IS_ENABLED(CONFIG_MACSEC)
2398 /* MACsec management functions */
2399 const struct macsec_ops *macsec_ops;
2400#endif
2401 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2402 struct udp_tunnel_nic *udp_tunnel_nic;
2403
2404 /* protected by rtnl_lock */
2405 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2406
2407 u8 dev_addr_shadow[MAX_ADDR_LEN];
2408 netdevice_tracker linkwatch_dev_tracker;
2409 netdevice_tracker watchdog_dev_tracker;
2410 netdevice_tracker dev_registered_tracker;
2411 struct rtnl_hw_stats64 *offload_xstats_l3;
2412
2413 struct devlink_port *devlink_port;
2414};
2415#define to_net_dev(d) container_of(d, struct net_device, dev)
2416
2417/*
2418 * Driver should use this to assign devlink port instance to a netdevice
2419 * before it registers the netdevice. Therefore devlink_port is static
2420 * during the netdev lifetime after it is registered.
2421 */
2422#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2423({ \
2424 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2425 ((dev)->devlink_port = (port)); \
2426})
2427
2428static inline bool netif_elide_gro(const struct net_device *dev)
2429{
2430 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2431 return true;
2432 return false;
2433}
2434
2435#define NETDEV_ALIGN 32
2436
2437static inline
2438int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2439{
2440 return dev->prio_tc_map[prio & TC_BITMASK];
2441}
2442
2443static inline
2444int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2445{
2446 if (tc >= dev->num_tc)
2447 return -EINVAL;
2448
2449 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2450 return 0;
2451}
2452
2453int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2454void netdev_reset_tc(struct net_device *dev);
2455int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2456int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2457
2458static inline
2459int netdev_get_num_tc(struct net_device *dev)
2460{
2461 return dev->num_tc;
2462}
2463
2464static inline void net_prefetch(void *p)
2465{
2466 prefetch(p);
2467#if L1_CACHE_BYTES < 128
2468 prefetch((u8 *)p + L1_CACHE_BYTES);
2469#endif
2470}
2471
2472static inline void net_prefetchw(void *p)
2473{
2474 prefetchw(p);
2475#if L1_CACHE_BYTES < 128
2476 prefetchw((u8 *)p + L1_CACHE_BYTES);
2477#endif
2478}
2479
2480void netdev_unbind_sb_channel(struct net_device *dev,
2481 struct net_device *sb_dev);
2482int netdev_bind_sb_channel_queue(struct net_device *dev,
2483 struct net_device *sb_dev,
2484 u8 tc, u16 count, u16 offset);
2485int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2486static inline int netdev_get_sb_channel(struct net_device *dev)
2487{
2488 return max_t(int, -dev->num_tc, 0);
2489}
2490
2491static inline
2492struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2493 unsigned int index)
2494{
2495 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2496 return &dev->_tx[index];
2497}
2498
2499static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2500 const struct sk_buff *skb)
2501{
2502 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2503}
2504
2505static inline void netdev_for_each_tx_queue(struct net_device *dev,
2506 void (*f)(struct net_device *,
2507 struct netdev_queue *,
2508 void *),
2509 void *arg)
2510{
2511 unsigned int i;
2512
2513 for (i = 0; i < dev->num_tx_queues; i++)
2514 f(dev, &dev->_tx[i], arg);
2515}
2516
2517#define netdev_lockdep_set_classes(dev) \
2518{ \
2519 static struct lock_class_key qdisc_tx_busylock_key; \
2520 static struct lock_class_key qdisc_xmit_lock_key; \
2521 static struct lock_class_key dev_addr_list_lock_key; \
2522 unsigned int i; \
2523 \
2524 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2525 lockdep_set_class(&(dev)->addr_list_lock, \
2526 &dev_addr_list_lock_key); \
2527 for (i = 0; i < (dev)->num_tx_queues; i++) \
2528 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2529 &qdisc_xmit_lock_key); \
2530}
2531
2532u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2533 struct net_device *sb_dev);
2534struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2535 struct sk_buff *skb,
2536 struct net_device *sb_dev);
2537
2538/* returns the headroom that the master device needs to take in account
2539 * when forwarding to this dev
2540 */
2541static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2542{
2543 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2544}
2545
2546static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2547{
2548 if (dev->netdev_ops->ndo_set_rx_headroom)
2549 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2550}
2551
2552/* set the device rx headroom to the dev's default */
2553static inline void netdev_reset_rx_headroom(struct net_device *dev)
2554{
2555 netdev_set_rx_headroom(dev, -1);
2556}
2557
2558static inline void *netdev_get_ml_priv(struct net_device *dev,
2559 enum netdev_ml_priv_type type)
2560{
2561 if (dev->ml_priv_type != type)
2562 return NULL;
2563
2564 return dev->ml_priv;
2565}
2566
2567static inline void netdev_set_ml_priv(struct net_device *dev,
2568 void *ml_priv,
2569 enum netdev_ml_priv_type type)
2570{
2571 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2572 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2573 dev->ml_priv_type, type);
2574 WARN(!dev->ml_priv_type && dev->ml_priv,
2575 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2576
2577 dev->ml_priv = ml_priv;
2578 dev->ml_priv_type = type;
2579}
2580
2581/*
2582 * Net namespace inlines
2583 */
2584static inline
2585struct net *dev_net(const struct net_device *dev)
2586{
2587 return read_pnet(&dev->nd_net);
2588}
2589
2590static inline
2591void dev_net_set(struct net_device *dev, struct net *net)
2592{
2593 write_pnet(&dev->nd_net, net);
2594}
2595
2596/**
2597 * netdev_priv - access network device private data
2598 * @dev: network device
2599 *
2600 * Get network device private data
2601 */
2602static inline void *netdev_priv(const struct net_device *dev)
2603{
2604 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2605}
2606
2607/* Set the sysfs physical device reference for the network logical device
2608 * if set prior to registration will cause a symlink during initialization.
2609 */
2610#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2611
2612/* Set the sysfs device type for the network logical device to allow
2613 * fine-grained identification of different network device types. For
2614 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2615 */
2616#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2617
2618/* Default NAPI poll() weight
2619 * Device drivers are strongly advised to not use bigger value
2620 */
2621#define NAPI_POLL_WEIGHT 64
2622
2623void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2624 int (*poll)(struct napi_struct *, int), int weight);
2625
2626/**
2627 * netif_napi_add() - initialize a NAPI context
2628 * @dev: network device
2629 * @napi: NAPI context
2630 * @poll: polling function
2631 *
2632 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2633 * *any* of the other NAPI-related functions.
2634 */
2635static inline void
2636netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2637 int (*poll)(struct napi_struct *, int))
2638{
2639 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2640}
2641
2642static inline void
2643netif_napi_add_tx_weight(struct net_device *dev,
2644 struct napi_struct *napi,
2645 int (*poll)(struct napi_struct *, int),
2646 int weight)
2647{
2648 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2649 netif_napi_add_weight(dev, napi, poll, weight);
2650}
2651
2652/**
2653 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2654 * @dev: network device
2655 * @napi: NAPI context
2656 * @poll: polling function
2657 *
2658 * This variant of netif_napi_add() should be used from drivers using NAPI
2659 * to exclusively poll a TX queue.
2660 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2661 */
2662static inline void netif_napi_add_tx(struct net_device *dev,
2663 struct napi_struct *napi,
2664 int (*poll)(struct napi_struct *, int))
2665{
2666 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2667}
2668
2669/**
2670 * __netif_napi_del - remove a NAPI context
2671 * @napi: NAPI context
2672 *
2673 * Warning: caller must observe RCU grace period before freeing memory
2674 * containing @napi. Drivers might want to call this helper to combine
2675 * all the needed RCU grace periods into a single one.
2676 */
2677void __netif_napi_del(struct napi_struct *napi);
2678
2679/**
2680 * netif_napi_del - remove a NAPI context
2681 * @napi: NAPI context
2682 *
2683 * netif_napi_del() removes a NAPI context from the network device NAPI list
2684 */
2685static inline void netif_napi_del(struct napi_struct *napi)
2686{
2687 __netif_napi_del(napi);
2688 synchronize_net();
2689}
2690
2691struct packet_type {
2692 __be16 type; /* This is really htons(ether_type). */
2693 bool ignore_outgoing;
2694 struct net_device *dev; /* NULL is wildcarded here */
2695 netdevice_tracker dev_tracker;
2696 int (*func) (struct sk_buff *,
2697 struct net_device *,
2698 struct packet_type *,
2699 struct net_device *);
2700 void (*list_func) (struct list_head *,
2701 struct packet_type *,
2702 struct net_device *);
2703 bool (*id_match)(struct packet_type *ptype,
2704 struct sock *sk);
2705 struct net *af_packet_net;
2706 void *af_packet_priv;
2707 struct list_head list;
2708};
2709
2710struct offload_callbacks {
2711 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2712 netdev_features_t features);
2713 struct sk_buff *(*gro_receive)(struct list_head *head,
2714 struct sk_buff *skb);
2715 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2716};
2717
2718struct packet_offload {
2719 __be16 type; /* This is really htons(ether_type). */
2720 u16 priority;
2721 struct offload_callbacks callbacks;
2722 struct list_head list;
2723};
2724
2725/* often modified stats are per-CPU, other are shared (netdev->stats) */
2726struct pcpu_sw_netstats {
2727 u64_stats_t rx_packets;
2728 u64_stats_t rx_bytes;
2729 u64_stats_t tx_packets;
2730 u64_stats_t tx_bytes;
2731 struct u64_stats_sync syncp;
2732} __aligned(4 * sizeof(u64));
2733
2734struct pcpu_lstats {
2735 u64_stats_t packets;
2736 u64_stats_t bytes;
2737 struct u64_stats_sync syncp;
2738} __aligned(2 * sizeof(u64));
2739
2740void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2741
2742static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2743{
2744 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2745
2746 u64_stats_update_begin(&tstats->syncp);
2747 u64_stats_add(&tstats->rx_bytes, len);
2748 u64_stats_inc(&tstats->rx_packets);
2749 u64_stats_update_end(&tstats->syncp);
2750}
2751
2752static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2753 unsigned int packets,
2754 unsigned int len)
2755{
2756 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2757
2758 u64_stats_update_begin(&tstats->syncp);
2759 u64_stats_add(&tstats->tx_bytes, len);
2760 u64_stats_add(&tstats->tx_packets, packets);
2761 u64_stats_update_end(&tstats->syncp);
2762}
2763
2764static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2765{
2766 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2767
2768 u64_stats_update_begin(&lstats->syncp);
2769 u64_stats_add(&lstats->bytes, len);
2770 u64_stats_inc(&lstats->packets);
2771 u64_stats_update_end(&lstats->syncp);
2772}
2773
2774#define __netdev_alloc_pcpu_stats(type, gfp) \
2775({ \
2776 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2777 if (pcpu_stats) { \
2778 int __cpu; \
2779 for_each_possible_cpu(__cpu) { \
2780 typeof(type) *stat; \
2781 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2782 u64_stats_init(&stat->syncp); \
2783 } \
2784 } \
2785 pcpu_stats; \
2786})
2787
2788#define netdev_alloc_pcpu_stats(type) \
2789 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2790
2791#define devm_netdev_alloc_pcpu_stats(dev, type) \
2792({ \
2793 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2794 if (pcpu_stats) { \
2795 int __cpu; \
2796 for_each_possible_cpu(__cpu) { \
2797 typeof(type) *stat; \
2798 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2799 u64_stats_init(&stat->syncp); \
2800 } \
2801 } \
2802 pcpu_stats; \
2803})
2804
2805enum netdev_lag_tx_type {
2806 NETDEV_LAG_TX_TYPE_UNKNOWN,
2807 NETDEV_LAG_TX_TYPE_RANDOM,
2808 NETDEV_LAG_TX_TYPE_BROADCAST,
2809 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2810 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2811 NETDEV_LAG_TX_TYPE_HASH,
2812};
2813
2814enum netdev_lag_hash {
2815 NETDEV_LAG_HASH_NONE,
2816 NETDEV_LAG_HASH_L2,
2817 NETDEV_LAG_HASH_L34,
2818 NETDEV_LAG_HASH_L23,
2819 NETDEV_LAG_HASH_E23,
2820 NETDEV_LAG_HASH_E34,
2821 NETDEV_LAG_HASH_VLAN_SRCMAC,
2822 NETDEV_LAG_HASH_UNKNOWN,
2823};
2824
2825struct netdev_lag_upper_info {
2826 enum netdev_lag_tx_type tx_type;
2827 enum netdev_lag_hash hash_type;
2828};
2829
2830struct netdev_lag_lower_state_info {
2831 u8 link_up : 1,
2832 tx_enabled : 1;
2833};
2834
2835#include <linux/notifier.h>
2836
2837/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2838 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2839 * adding new types.
2840 */
2841enum netdev_cmd {
2842 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2843 NETDEV_DOWN,
2844 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2845 detected a hardware crash and restarted
2846 - we can use this eg to kick tcp sessions
2847 once done */
2848 NETDEV_CHANGE, /* Notify device state change */
2849 NETDEV_REGISTER,
2850 NETDEV_UNREGISTER,
2851 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2852 NETDEV_CHANGEADDR, /* notify after the address change */
2853 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2854 NETDEV_GOING_DOWN,
2855 NETDEV_CHANGENAME,
2856 NETDEV_FEAT_CHANGE,
2857 NETDEV_BONDING_FAILOVER,
2858 NETDEV_PRE_UP,
2859 NETDEV_PRE_TYPE_CHANGE,
2860 NETDEV_POST_TYPE_CHANGE,
2861 NETDEV_POST_INIT,
2862 NETDEV_PRE_UNINIT,
2863 NETDEV_RELEASE,
2864 NETDEV_NOTIFY_PEERS,
2865 NETDEV_JOIN,
2866 NETDEV_CHANGEUPPER,
2867 NETDEV_RESEND_IGMP,
2868 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2869 NETDEV_CHANGEINFODATA,
2870 NETDEV_BONDING_INFO,
2871 NETDEV_PRECHANGEUPPER,
2872 NETDEV_CHANGELOWERSTATE,
2873 NETDEV_UDP_TUNNEL_PUSH_INFO,
2874 NETDEV_UDP_TUNNEL_DROP_INFO,
2875 NETDEV_CHANGE_TX_QUEUE_LEN,
2876 NETDEV_CVLAN_FILTER_PUSH_INFO,
2877 NETDEV_CVLAN_FILTER_DROP_INFO,
2878 NETDEV_SVLAN_FILTER_PUSH_INFO,
2879 NETDEV_SVLAN_FILTER_DROP_INFO,
2880 NETDEV_OFFLOAD_XSTATS_ENABLE,
2881 NETDEV_OFFLOAD_XSTATS_DISABLE,
2882 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2883 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2884 NETDEV_XDP_FEAT_CHANGE,
2885};
2886const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2887
2888int register_netdevice_notifier(struct notifier_block *nb);
2889int unregister_netdevice_notifier(struct notifier_block *nb);
2890int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2891int unregister_netdevice_notifier_net(struct net *net,
2892 struct notifier_block *nb);
2893int register_netdevice_notifier_dev_net(struct net_device *dev,
2894 struct notifier_block *nb,
2895 struct netdev_net_notifier *nn);
2896int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2897 struct notifier_block *nb,
2898 struct netdev_net_notifier *nn);
2899
2900struct netdev_notifier_info {
2901 struct net_device *dev;
2902 struct netlink_ext_ack *extack;
2903};
2904
2905struct netdev_notifier_info_ext {
2906 struct netdev_notifier_info info; /* must be first */
2907 union {
2908 u32 mtu;
2909 } ext;
2910};
2911
2912struct netdev_notifier_change_info {
2913 struct netdev_notifier_info info; /* must be first */
2914 unsigned int flags_changed;
2915};
2916
2917struct netdev_notifier_changeupper_info {
2918 struct netdev_notifier_info info; /* must be first */
2919 struct net_device *upper_dev; /* new upper dev */
2920 bool master; /* is upper dev master */
2921 bool linking; /* is the notification for link or unlink */
2922 void *upper_info; /* upper dev info */
2923};
2924
2925struct netdev_notifier_changelowerstate_info {
2926 struct netdev_notifier_info info; /* must be first */
2927 void *lower_state_info; /* is lower dev state */
2928};
2929
2930struct netdev_notifier_pre_changeaddr_info {
2931 struct netdev_notifier_info info; /* must be first */
2932 const unsigned char *dev_addr;
2933};
2934
2935enum netdev_offload_xstats_type {
2936 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2937};
2938
2939struct netdev_notifier_offload_xstats_info {
2940 struct netdev_notifier_info info; /* must be first */
2941 enum netdev_offload_xstats_type type;
2942
2943 union {
2944 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2945 struct netdev_notifier_offload_xstats_rd *report_delta;
2946 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2947 struct netdev_notifier_offload_xstats_ru *report_used;
2948 };
2949};
2950
2951int netdev_offload_xstats_enable(struct net_device *dev,
2952 enum netdev_offload_xstats_type type,
2953 struct netlink_ext_ack *extack);
2954int netdev_offload_xstats_disable(struct net_device *dev,
2955 enum netdev_offload_xstats_type type);
2956bool netdev_offload_xstats_enabled(const struct net_device *dev,
2957 enum netdev_offload_xstats_type type);
2958int netdev_offload_xstats_get(struct net_device *dev,
2959 enum netdev_offload_xstats_type type,
2960 struct rtnl_hw_stats64 *stats, bool *used,
2961 struct netlink_ext_ack *extack);
2962void
2963netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2964 const struct rtnl_hw_stats64 *stats);
2965void
2966netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2967void netdev_offload_xstats_push_delta(struct net_device *dev,
2968 enum netdev_offload_xstats_type type,
2969 const struct rtnl_hw_stats64 *stats);
2970
2971static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2972 struct net_device *dev)
2973{
2974 info->dev = dev;
2975 info->extack = NULL;
2976}
2977
2978static inline struct net_device *
2979netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2980{
2981 return info->dev;
2982}
2983
2984static inline struct netlink_ext_ack *
2985netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2986{
2987 return info->extack;
2988}
2989
2990int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2991int call_netdevice_notifiers_info(unsigned long val,
2992 struct netdev_notifier_info *info);
2993
2994extern rwlock_t dev_base_lock; /* Device list lock */
2995
2996#define for_each_netdev(net, d) \
2997 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2998#define for_each_netdev_reverse(net, d) \
2999 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3000#define for_each_netdev_rcu(net, d) \
3001 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3002#define for_each_netdev_safe(net, d, n) \
3003 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3004#define for_each_netdev_continue(net, d) \
3005 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3006#define for_each_netdev_continue_reverse(net, d) \
3007 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3008 dev_list)
3009#define for_each_netdev_continue_rcu(net, d) \
3010 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3011#define for_each_netdev_in_bond_rcu(bond, slave) \
3012 for_each_netdev_rcu(&init_net, slave) \
3013 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3014#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3015
3016static inline struct net_device *next_net_device(struct net_device *dev)
3017{
3018 struct list_head *lh;
3019 struct net *net;
3020
3021 net = dev_net(dev);
3022 lh = dev->dev_list.next;
3023 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3024}
3025
3026static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3027{
3028 struct list_head *lh;
3029 struct net *net;
3030
3031 net = dev_net(dev);
3032 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3033 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3034}
3035
3036static inline struct net_device *first_net_device(struct net *net)
3037{
3038 return list_empty(&net->dev_base_head) ? NULL :
3039 net_device_entry(net->dev_base_head.next);
3040}
3041
3042static inline struct net_device *first_net_device_rcu(struct net *net)
3043{
3044 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3045
3046 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3047}
3048
3049int netdev_boot_setup_check(struct net_device *dev);
3050struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3051 const char *hwaddr);
3052struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3053void dev_add_pack(struct packet_type *pt);
3054void dev_remove_pack(struct packet_type *pt);
3055void __dev_remove_pack(struct packet_type *pt);
3056void dev_add_offload(struct packet_offload *po);
3057void dev_remove_offload(struct packet_offload *po);
3058
3059int dev_get_iflink(const struct net_device *dev);
3060int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3061int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3062 struct net_device_path_stack *stack);
3063struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3064 unsigned short mask);
3065struct net_device *dev_get_by_name(struct net *net, const char *name);
3066struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3067struct net_device *__dev_get_by_name(struct net *net, const char *name);
3068bool netdev_name_in_use(struct net *net, const char *name);
3069int dev_alloc_name(struct net_device *dev, const char *name);
3070int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3071void dev_close(struct net_device *dev);
3072void dev_close_many(struct list_head *head, bool unlink);
3073void dev_disable_lro(struct net_device *dev);
3074int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3075u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3076 struct net_device *sb_dev);
3077u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3078 struct net_device *sb_dev);
3079
3080int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3081int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3082
3083static inline int dev_queue_xmit(struct sk_buff *skb)
3084{
3085 return __dev_queue_xmit(skb, NULL);
3086}
3087
3088static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3089 struct net_device *sb_dev)
3090{
3091 return __dev_queue_xmit(skb, sb_dev);
3092}
3093
3094static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3095{
3096 int ret;
3097
3098 ret = __dev_direct_xmit(skb, queue_id);
3099 if (!dev_xmit_complete(ret))
3100 kfree_skb(skb);
3101 return ret;
3102}
3103
3104int register_netdevice(struct net_device *dev);
3105void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3106void unregister_netdevice_many(struct list_head *head);
3107static inline void unregister_netdevice(struct net_device *dev)
3108{
3109 unregister_netdevice_queue(dev, NULL);
3110}
3111
3112int netdev_refcnt_read(const struct net_device *dev);
3113void free_netdev(struct net_device *dev);
3114void netdev_freemem(struct net_device *dev);
3115int init_dummy_netdev(struct net_device *dev);
3116
3117struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3118 struct sk_buff *skb,
3119 bool all_slaves);
3120struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3121 struct sock *sk);
3122struct net_device *dev_get_by_index(struct net *net, int ifindex);
3123struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3124struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3125struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3126int dev_restart(struct net_device *dev);
3127
3128
3129static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3130 unsigned short type,
3131 const void *daddr, const void *saddr,
3132 unsigned int len)
3133{
3134 if (!dev->header_ops || !dev->header_ops->create)
3135 return 0;
3136
3137 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3138}
3139
3140static inline int dev_parse_header(const struct sk_buff *skb,
3141 unsigned char *haddr)
3142{
3143 const struct net_device *dev = skb->dev;
3144
3145 if (!dev->header_ops || !dev->header_ops->parse)
3146 return 0;
3147 return dev->header_ops->parse(skb, haddr);
3148}
3149
3150static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3151{
3152 const struct net_device *dev = skb->dev;
3153
3154 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3155 return 0;
3156 return dev->header_ops->parse_protocol(skb);
3157}
3158
3159/* ll_header must have at least hard_header_len allocated */
3160static inline bool dev_validate_header(const struct net_device *dev,
3161 char *ll_header, int len)
3162{
3163 if (likely(len >= dev->hard_header_len))
3164 return true;
3165 if (len < dev->min_header_len)
3166 return false;
3167
3168 if (capable(CAP_SYS_RAWIO)) {
3169 memset(ll_header + len, 0, dev->hard_header_len - len);
3170 return true;
3171 }
3172
3173 if (dev->header_ops && dev->header_ops->validate)
3174 return dev->header_ops->validate(ll_header, len);
3175
3176 return false;
3177}
3178
3179static inline bool dev_has_header(const struct net_device *dev)
3180{
3181 return dev->header_ops && dev->header_ops->create;
3182}
3183
3184/*
3185 * Incoming packets are placed on per-CPU queues
3186 */
3187struct softnet_data {
3188 struct list_head poll_list;
3189 struct sk_buff_head process_queue;
3190
3191 /* stats */
3192 unsigned int processed;
3193 unsigned int time_squeeze;
3194#ifdef CONFIG_RPS
3195 struct softnet_data *rps_ipi_list;
3196#endif
3197
3198 bool in_net_rx_action;
3199 bool in_napi_threaded_poll;
3200
3201#ifdef CONFIG_NET_FLOW_LIMIT
3202 struct sd_flow_limit __rcu *flow_limit;
3203#endif
3204 struct Qdisc *output_queue;
3205 struct Qdisc **output_queue_tailp;
3206 struct sk_buff *completion_queue;
3207#ifdef CONFIG_XFRM_OFFLOAD
3208 struct sk_buff_head xfrm_backlog;
3209#endif
3210 /* written and read only by owning cpu: */
3211 struct {
3212 u16 recursion;
3213 u8 more;
3214#ifdef CONFIG_NET_EGRESS
3215 u8 skip_txqueue;
3216#endif
3217 } xmit;
3218#ifdef CONFIG_RPS
3219 /* input_queue_head should be written by cpu owning this struct,
3220 * and only read by other cpus. Worth using a cache line.
3221 */
3222 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3223
3224 /* Elements below can be accessed between CPUs for RPS/RFS */
3225 call_single_data_t csd ____cacheline_aligned_in_smp;
3226 struct softnet_data *rps_ipi_next;
3227 unsigned int cpu;
3228 unsigned int input_queue_tail;
3229#endif
3230 unsigned int received_rps;
3231 unsigned int dropped;
3232 struct sk_buff_head input_pkt_queue;
3233 struct napi_struct backlog;
3234
3235 /* Another possibly contended cache line */
3236 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3237 int defer_count;
3238 int defer_ipi_scheduled;
3239 struct sk_buff *defer_list;
3240 call_single_data_t defer_csd;
3241};
3242
3243static inline void input_queue_head_incr(struct softnet_data *sd)
3244{
3245#ifdef CONFIG_RPS
3246 sd->input_queue_head++;
3247#endif
3248}
3249
3250static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3251 unsigned int *qtail)
3252{
3253#ifdef CONFIG_RPS
3254 *qtail = ++sd->input_queue_tail;
3255#endif
3256}
3257
3258DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3259
3260static inline int dev_recursion_level(void)
3261{
3262 return this_cpu_read(softnet_data.xmit.recursion);
3263}
3264
3265#define XMIT_RECURSION_LIMIT 8
3266static inline bool dev_xmit_recursion(void)
3267{
3268 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3269 XMIT_RECURSION_LIMIT);
3270}
3271
3272static inline void dev_xmit_recursion_inc(void)
3273{
3274 __this_cpu_inc(softnet_data.xmit.recursion);
3275}
3276
3277static inline void dev_xmit_recursion_dec(void)
3278{
3279 __this_cpu_dec(softnet_data.xmit.recursion);
3280}
3281
3282void __netif_schedule(struct Qdisc *q);
3283void netif_schedule_queue(struct netdev_queue *txq);
3284
3285static inline void netif_tx_schedule_all(struct net_device *dev)
3286{
3287 unsigned int i;
3288
3289 for (i = 0; i < dev->num_tx_queues; i++)
3290 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3291}
3292
3293static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3294{
3295 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3296}
3297
3298/**
3299 * netif_start_queue - allow transmit
3300 * @dev: network device
3301 *
3302 * Allow upper layers to call the device hard_start_xmit routine.
3303 */
3304static inline void netif_start_queue(struct net_device *dev)
3305{
3306 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3307}
3308
3309static inline void netif_tx_start_all_queues(struct net_device *dev)
3310{
3311 unsigned int i;
3312
3313 for (i = 0; i < dev->num_tx_queues; i++) {
3314 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3315 netif_tx_start_queue(txq);
3316 }
3317}
3318
3319void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3320
3321/**
3322 * netif_wake_queue - restart transmit
3323 * @dev: network device
3324 *
3325 * Allow upper layers to call the device hard_start_xmit routine.
3326 * Used for flow control when transmit resources are available.
3327 */
3328static inline void netif_wake_queue(struct net_device *dev)
3329{
3330 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3331}
3332
3333static inline void netif_tx_wake_all_queues(struct net_device *dev)
3334{
3335 unsigned int i;
3336
3337 for (i = 0; i < dev->num_tx_queues; i++) {
3338 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3339 netif_tx_wake_queue(txq);
3340 }
3341}
3342
3343static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3344{
3345 /* Must be an atomic op see netif_txq_try_stop() */
3346 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3347}
3348
3349/**
3350 * netif_stop_queue - stop transmitted packets
3351 * @dev: network device
3352 *
3353 * Stop upper layers calling the device hard_start_xmit routine.
3354 * Used for flow control when transmit resources are unavailable.
3355 */
3356static inline void netif_stop_queue(struct net_device *dev)
3357{
3358 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3359}
3360
3361void netif_tx_stop_all_queues(struct net_device *dev);
3362
3363static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3364{
3365 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3366}
3367
3368/**
3369 * netif_queue_stopped - test if transmit queue is flowblocked
3370 * @dev: network device
3371 *
3372 * Test if transmit queue on device is currently unable to send.
3373 */
3374static inline bool netif_queue_stopped(const struct net_device *dev)
3375{
3376 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3377}
3378
3379static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3380{
3381 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3382}
3383
3384static inline bool
3385netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3386{
3387 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3388}
3389
3390static inline bool
3391netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3392{
3393 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3394}
3395
3396/**
3397 * netdev_queue_set_dql_min_limit - set dql minimum limit
3398 * @dev_queue: pointer to transmit queue
3399 * @min_limit: dql minimum limit
3400 *
3401 * Forces xmit_more() to return true until the minimum threshold
3402 * defined by @min_limit is reached (or until the tx queue is
3403 * empty). Warning: to be use with care, misuse will impact the
3404 * latency.
3405 */
3406static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3407 unsigned int min_limit)
3408{
3409#ifdef CONFIG_BQL
3410 dev_queue->dql.min_limit = min_limit;
3411#endif
3412}
3413
3414/**
3415 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3416 * @dev_queue: pointer to transmit queue
3417 *
3418 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3419 * to give appropriate hint to the CPU.
3420 */
3421static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3422{
3423#ifdef CONFIG_BQL
3424 prefetchw(&dev_queue->dql.num_queued);
3425#endif
3426}
3427
3428/**
3429 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3430 * @dev_queue: pointer to transmit queue
3431 *
3432 * BQL enabled drivers might use this helper in their TX completion path,
3433 * to give appropriate hint to the CPU.
3434 */
3435static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3436{
3437#ifdef CONFIG_BQL
3438 prefetchw(&dev_queue->dql.limit);
3439#endif
3440}
3441
3442/**
3443 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3444 * @dev_queue: network device queue
3445 * @bytes: number of bytes queued to the device queue
3446 *
3447 * Report the number of bytes queued for sending/completion to the network
3448 * device hardware queue. @bytes should be a good approximation and should
3449 * exactly match netdev_completed_queue() @bytes.
3450 * This is typically called once per packet, from ndo_start_xmit().
3451 */
3452static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3453 unsigned int bytes)
3454{
3455#ifdef CONFIG_BQL
3456 dql_queued(&dev_queue->dql, bytes);
3457
3458 if (likely(dql_avail(&dev_queue->dql) >= 0))
3459 return;
3460
3461 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3462
3463 /*
3464 * The XOFF flag must be set before checking the dql_avail below,
3465 * because in netdev_tx_completed_queue we update the dql_completed
3466 * before checking the XOFF flag.
3467 */
3468 smp_mb();
3469
3470 /* check again in case another CPU has just made room avail */
3471 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3472 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3473#endif
3474}
3475
3476/* Variant of netdev_tx_sent_queue() for drivers that are aware
3477 * that they should not test BQL status themselves.
3478 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3479 * skb of a batch.
3480 * Returns true if the doorbell must be used to kick the NIC.
3481 */
3482static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3483 unsigned int bytes,
3484 bool xmit_more)
3485{
3486 if (xmit_more) {
3487#ifdef CONFIG_BQL
3488 dql_queued(&dev_queue->dql, bytes);
3489#endif
3490 return netif_tx_queue_stopped(dev_queue);
3491 }
3492 netdev_tx_sent_queue(dev_queue, bytes);
3493 return true;
3494}
3495
3496/**
3497 * netdev_sent_queue - report the number of bytes queued to hardware
3498 * @dev: network device
3499 * @bytes: number of bytes queued to the hardware device queue
3500 *
3501 * Report the number of bytes queued for sending/completion to the network
3502 * device hardware queue#0. @bytes should be a good approximation and should
3503 * exactly match netdev_completed_queue() @bytes.
3504 * This is typically called once per packet, from ndo_start_xmit().
3505 */
3506static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3507{
3508 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3509}
3510
3511static inline bool __netdev_sent_queue(struct net_device *dev,
3512 unsigned int bytes,
3513 bool xmit_more)
3514{
3515 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3516 xmit_more);
3517}
3518
3519/**
3520 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3521 * @dev_queue: network device queue
3522 * @pkts: number of packets (currently ignored)
3523 * @bytes: number of bytes dequeued from the device queue
3524 *
3525 * Must be called at most once per TX completion round (and not per
3526 * individual packet), so that BQL can adjust its limits appropriately.
3527 */
3528static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3529 unsigned int pkts, unsigned int bytes)
3530{
3531#ifdef CONFIG_BQL
3532 if (unlikely(!bytes))
3533 return;
3534
3535 dql_completed(&dev_queue->dql, bytes);
3536
3537 /*
3538 * Without the memory barrier there is a small possiblity that
3539 * netdev_tx_sent_queue will miss the update and cause the queue to
3540 * be stopped forever
3541 */
3542 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3543
3544 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3545 return;
3546
3547 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3548 netif_schedule_queue(dev_queue);
3549#endif
3550}
3551
3552/**
3553 * netdev_completed_queue - report bytes and packets completed by device
3554 * @dev: network device
3555 * @pkts: actual number of packets sent over the medium
3556 * @bytes: actual number of bytes sent over the medium
3557 *
3558 * Report the number of bytes and packets transmitted by the network device
3559 * hardware queue over the physical medium, @bytes must exactly match the
3560 * @bytes amount passed to netdev_sent_queue()
3561 */
3562static inline void netdev_completed_queue(struct net_device *dev,
3563 unsigned int pkts, unsigned int bytes)
3564{
3565 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3566}
3567
3568static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3569{
3570#ifdef CONFIG_BQL
3571 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3572 dql_reset(&q->dql);
3573#endif
3574}
3575
3576/**
3577 * netdev_reset_queue - reset the packets and bytes count of a network device
3578 * @dev_queue: network device
3579 *
3580 * Reset the bytes and packet count of a network device and clear the
3581 * software flow control OFF bit for this network device
3582 */
3583static inline void netdev_reset_queue(struct net_device *dev_queue)
3584{
3585 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3586}
3587
3588/**
3589 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3590 * @dev: network device
3591 * @queue_index: given tx queue index
3592 *
3593 * Returns 0 if given tx queue index >= number of device tx queues,
3594 * otherwise returns the originally passed tx queue index.
3595 */
3596static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3597{
3598 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3599 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3600 dev->name, queue_index,
3601 dev->real_num_tx_queues);
3602 return 0;
3603 }
3604
3605 return queue_index;
3606}
3607
3608/**
3609 * netif_running - test if up
3610 * @dev: network device
3611 *
3612 * Test if the device has been brought up.
3613 */
3614static inline bool netif_running(const struct net_device *dev)
3615{
3616 return test_bit(__LINK_STATE_START, &dev->state);
3617}
3618
3619/*
3620 * Routines to manage the subqueues on a device. We only need start,
3621 * stop, and a check if it's stopped. All other device management is
3622 * done at the overall netdevice level.
3623 * Also test the device if we're multiqueue.
3624 */
3625
3626/**
3627 * netif_start_subqueue - allow sending packets on subqueue
3628 * @dev: network device
3629 * @queue_index: sub queue index
3630 *
3631 * Start individual transmit queue of a device with multiple transmit queues.
3632 */
3633static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3634{
3635 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3636
3637 netif_tx_start_queue(txq);
3638}
3639
3640/**
3641 * netif_stop_subqueue - stop sending packets on subqueue
3642 * @dev: network device
3643 * @queue_index: sub queue index
3644 *
3645 * Stop individual transmit queue of a device with multiple transmit queues.
3646 */
3647static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3648{
3649 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3650 netif_tx_stop_queue(txq);
3651}
3652
3653/**
3654 * __netif_subqueue_stopped - test status of subqueue
3655 * @dev: network device
3656 * @queue_index: sub queue index
3657 *
3658 * Check individual transmit queue of a device with multiple transmit queues.
3659 */
3660static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3661 u16 queue_index)
3662{
3663 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3664
3665 return netif_tx_queue_stopped(txq);
3666}
3667
3668/**
3669 * netif_subqueue_stopped - test status of subqueue
3670 * @dev: network device
3671 * @skb: sub queue buffer pointer
3672 *
3673 * Check individual transmit queue of a device with multiple transmit queues.
3674 */
3675static inline bool netif_subqueue_stopped(const struct net_device *dev,
3676 struct sk_buff *skb)
3677{
3678 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3679}
3680
3681/**
3682 * netif_wake_subqueue - allow sending packets on subqueue
3683 * @dev: network device
3684 * @queue_index: sub queue index
3685 *
3686 * Resume individual transmit queue of a device with multiple transmit queues.
3687 */
3688static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3689{
3690 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3691
3692 netif_tx_wake_queue(txq);
3693}
3694
3695#ifdef CONFIG_XPS
3696int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3697 u16 index);
3698int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3699 u16 index, enum xps_map_type type);
3700
3701/**
3702 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3703 * @j: CPU/Rx queue index
3704 * @mask: bitmask of all cpus/rx queues
3705 * @nr_bits: number of bits in the bitmask
3706 *
3707 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3708 */
3709static inline bool netif_attr_test_mask(unsigned long j,
3710 const unsigned long *mask,
3711 unsigned int nr_bits)
3712{
3713 cpu_max_bits_warn(j, nr_bits);
3714 return test_bit(j, mask);
3715}
3716
3717/**
3718 * netif_attr_test_online - Test for online CPU/Rx queue
3719 * @j: CPU/Rx queue index
3720 * @online_mask: bitmask for CPUs/Rx queues that are online
3721 * @nr_bits: number of bits in the bitmask
3722 *
3723 * Returns true if a CPU/Rx queue is online.
3724 */
3725static inline bool netif_attr_test_online(unsigned long j,
3726 const unsigned long *online_mask,
3727 unsigned int nr_bits)
3728{
3729 cpu_max_bits_warn(j, nr_bits);
3730
3731 if (online_mask)
3732 return test_bit(j, online_mask);
3733
3734 return (j < nr_bits);
3735}
3736
3737/**
3738 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3739 * @n: CPU/Rx queue index
3740 * @srcp: the cpumask/Rx queue mask pointer
3741 * @nr_bits: number of bits in the bitmask
3742 *
3743 * Returns >= nr_bits if no further CPUs/Rx queues set.
3744 */
3745static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3746 unsigned int nr_bits)
3747{
3748 /* -1 is a legal arg here. */
3749 if (n != -1)
3750 cpu_max_bits_warn(n, nr_bits);
3751
3752 if (srcp)
3753 return find_next_bit(srcp, nr_bits, n + 1);
3754
3755 return n + 1;
3756}
3757
3758/**
3759 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3760 * @n: CPU/Rx queue index
3761 * @src1p: the first CPUs/Rx queues mask pointer
3762 * @src2p: the second CPUs/Rx queues mask pointer
3763 * @nr_bits: number of bits in the bitmask
3764 *
3765 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3766 */
3767static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3768 const unsigned long *src2p,
3769 unsigned int nr_bits)
3770{
3771 /* -1 is a legal arg here. */
3772 if (n != -1)
3773 cpu_max_bits_warn(n, nr_bits);
3774
3775 if (src1p && src2p)
3776 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3777 else if (src1p)
3778 return find_next_bit(src1p, nr_bits, n + 1);
3779 else if (src2p)
3780 return find_next_bit(src2p, nr_bits, n + 1);
3781
3782 return n + 1;
3783}
3784#else
3785static inline int netif_set_xps_queue(struct net_device *dev,
3786 const struct cpumask *mask,
3787 u16 index)
3788{
3789 return 0;
3790}
3791
3792static inline int __netif_set_xps_queue(struct net_device *dev,
3793 const unsigned long *mask,
3794 u16 index, enum xps_map_type type)
3795{
3796 return 0;
3797}
3798#endif
3799
3800/**
3801 * netif_is_multiqueue - test if device has multiple transmit queues
3802 * @dev: network device
3803 *
3804 * Check if device has multiple transmit queues
3805 */
3806static inline bool netif_is_multiqueue(const struct net_device *dev)
3807{
3808 return dev->num_tx_queues > 1;
3809}
3810
3811int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3812
3813#ifdef CONFIG_SYSFS
3814int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3815#else
3816static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3817 unsigned int rxqs)
3818{
3819 dev->real_num_rx_queues = rxqs;
3820 return 0;
3821}
3822#endif
3823int netif_set_real_num_queues(struct net_device *dev,
3824 unsigned int txq, unsigned int rxq);
3825
3826static inline struct netdev_rx_queue *
3827__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3828{
3829 return dev->_rx + rxq;
3830}
3831
3832#ifdef CONFIG_SYSFS
3833static inline unsigned int get_netdev_rx_queue_index(
3834 struct netdev_rx_queue *queue)
3835{
3836 struct net_device *dev = queue->dev;
3837 int index = queue - dev->_rx;
3838
3839 BUG_ON(index >= dev->num_rx_queues);
3840 return index;
3841}
3842#endif
3843
3844int netif_get_num_default_rss_queues(void);
3845
3846void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3847void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3848
3849/*
3850 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3851 * interrupt context or with hardware interrupts being disabled.
3852 * (in_hardirq() || irqs_disabled())
3853 *
3854 * We provide four helpers that can be used in following contexts :
3855 *
3856 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3857 * replacing kfree_skb(skb)
3858 *
3859 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3860 * Typically used in place of consume_skb(skb) in TX completion path
3861 *
3862 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3863 * replacing kfree_skb(skb)
3864 *
3865 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3866 * and consumed a packet. Used in place of consume_skb(skb)
3867 */
3868static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3869{
3870 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3871}
3872
3873static inline void dev_consume_skb_irq(struct sk_buff *skb)
3874{
3875 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3876}
3877
3878static inline void dev_kfree_skb_any(struct sk_buff *skb)
3879{
3880 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3881}
3882
3883static inline void dev_consume_skb_any(struct sk_buff *skb)
3884{
3885 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3886}
3887
3888u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3889 struct bpf_prog *xdp_prog);
3890void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3891int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3892int netif_rx(struct sk_buff *skb);
3893int __netif_rx(struct sk_buff *skb);
3894
3895int netif_receive_skb(struct sk_buff *skb);
3896int netif_receive_skb_core(struct sk_buff *skb);
3897void netif_receive_skb_list_internal(struct list_head *head);
3898void netif_receive_skb_list(struct list_head *head);
3899gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3900void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3901struct sk_buff *napi_get_frags(struct napi_struct *napi);
3902void napi_get_frags_check(struct napi_struct *napi);
3903gro_result_t napi_gro_frags(struct napi_struct *napi);
3904struct packet_offload *gro_find_receive_by_type(__be16 type);
3905struct packet_offload *gro_find_complete_by_type(__be16 type);
3906
3907static inline void napi_free_frags(struct napi_struct *napi)
3908{
3909 kfree_skb(napi->skb);
3910 napi->skb = NULL;
3911}
3912
3913bool netdev_is_rx_handler_busy(struct net_device *dev);
3914int netdev_rx_handler_register(struct net_device *dev,
3915 rx_handler_func_t *rx_handler,
3916 void *rx_handler_data);
3917void netdev_rx_handler_unregister(struct net_device *dev);
3918
3919bool dev_valid_name(const char *name);
3920static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3921{
3922 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3923}
3924int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3925int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3926int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3927 void __user *data, bool *need_copyout);
3928int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3929int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3930unsigned int dev_get_flags(const struct net_device *);
3931int __dev_change_flags(struct net_device *dev, unsigned int flags,
3932 struct netlink_ext_ack *extack);
3933int dev_change_flags(struct net_device *dev, unsigned int flags,
3934 struct netlink_ext_ack *extack);
3935int dev_set_alias(struct net_device *, const char *, size_t);
3936int dev_get_alias(const struct net_device *, char *, size_t);
3937int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3938 const char *pat, int new_ifindex);
3939static inline
3940int dev_change_net_namespace(struct net_device *dev, struct net *net,
3941 const char *pat)
3942{
3943 return __dev_change_net_namespace(dev, net, pat, 0);
3944}
3945int __dev_set_mtu(struct net_device *, int);
3946int dev_set_mtu(struct net_device *, int);
3947int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3948 struct netlink_ext_ack *extack);
3949int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3950 struct netlink_ext_ack *extack);
3951int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3952 struct netlink_ext_ack *extack);
3953int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3954int dev_get_port_parent_id(struct net_device *dev,
3955 struct netdev_phys_item_id *ppid, bool recurse);
3956bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3957struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3958struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3959 struct netdev_queue *txq, int *ret);
3960
3961int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3962u8 dev_xdp_prog_count(struct net_device *dev);
3963u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3964
3965int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3966int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3967int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3968bool is_skb_forwardable(const struct net_device *dev,
3969 const struct sk_buff *skb);
3970
3971static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3972 const struct sk_buff *skb,
3973 const bool check_mtu)
3974{
3975 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3976 unsigned int len;
3977
3978 if (!(dev->flags & IFF_UP))
3979 return false;
3980
3981 if (!check_mtu)
3982 return true;
3983
3984 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3985 if (skb->len <= len)
3986 return true;
3987
3988 /* if TSO is enabled, we don't care about the length as the packet
3989 * could be forwarded without being segmented before
3990 */
3991 if (skb_is_gso(skb))
3992 return true;
3993
3994 return false;
3995}
3996
3997struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3998
3999static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
4000{
4001 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
4002 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
4003
4004 if (likely(p))
4005 return p;
4006
4007 return netdev_core_stats_alloc(dev);
4008}
4009
4010#define DEV_CORE_STATS_INC(FIELD) \
4011static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4012{ \
4013 struct net_device_core_stats __percpu *p; \
4014 \
4015 p = dev_core_stats(dev); \
4016 if (p) \
4017 this_cpu_inc(p->FIELD); \
4018}
4019DEV_CORE_STATS_INC(rx_dropped)
4020DEV_CORE_STATS_INC(tx_dropped)
4021DEV_CORE_STATS_INC(rx_nohandler)
4022DEV_CORE_STATS_INC(rx_otherhost_dropped)
4023
4024static __always_inline int ____dev_forward_skb(struct net_device *dev,
4025 struct sk_buff *skb,
4026 const bool check_mtu)
4027{
4028 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4029 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4030 dev_core_stats_rx_dropped_inc(dev);
4031 kfree_skb(skb);
4032 return NET_RX_DROP;
4033 }
4034
4035 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4036 skb->priority = 0;
4037 return 0;
4038}
4039
4040bool dev_nit_active(struct net_device *dev);
4041void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4042
4043static inline void __dev_put(struct net_device *dev)
4044{
4045 if (dev) {
4046#ifdef CONFIG_PCPU_DEV_REFCNT
4047 this_cpu_dec(*dev->pcpu_refcnt);
4048#else
4049 refcount_dec(&dev->dev_refcnt);
4050#endif
4051 }
4052}
4053
4054static inline void __dev_hold(struct net_device *dev)
4055{
4056 if (dev) {
4057#ifdef CONFIG_PCPU_DEV_REFCNT
4058 this_cpu_inc(*dev->pcpu_refcnt);
4059#else
4060 refcount_inc(&dev->dev_refcnt);
4061#endif
4062 }
4063}
4064
4065static inline void __netdev_tracker_alloc(struct net_device *dev,
4066 netdevice_tracker *tracker,
4067 gfp_t gfp)
4068{
4069#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4070 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4071#endif
4072}
4073
4074/* netdev_tracker_alloc() can upgrade a prior untracked reference
4075 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4076 */
4077static inline void netdev_tracker_alloc(struct net_device *dev,
4078 netdevice_tracker *tracker, gfp_t gfp)
4079{
4080#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4081 refcount_dec(&dev->refcnt_tracker.no_tracker);
4082 __netdev_tracker_alloc(dev, tracker, gfp);
4083#endif
4084}
4085
4086static inline void netdev_tracker_free(struct net_device *dev,
4087 netdevice_tracker *tracker)
4088{
4089#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4090 ref_tracker_free(&dev->refcnt_tracker, tracker);
4091#endif
4092}
4093
4094static inline void netdev_hold(struct net_device *dev,
4095 netdevice_tracker *tracker, gfp_t gfp)
4096{
4097 if (dev) {
4098 __dev_hold(dev);
4099 __netdev_tracker_alloc(dev, tracker, gfp);
4100 }
4101}
4102
4103static inline void netdev_put(struct net_device *dev,
4104 netdevice_tracker *tracker)
4105{
4106 if (dev) {
4107 netdev_tracker_free(dev, tracker);
4108 __dev_put(dev);
4109 }
4110}
4111
4112/**
4113 * dev_hold - get reference to device
4114 * @dev: network device
4115 *
4116 * Hold reference to device to keep it from being freed.
4117 * Try using netdev_hold() instead.
4118 */
4119static inline void dev_hold(struct net_device *dev)
4120{
4121 netdev_hold(dev, NULL, GFP_ATOMIC);
4122}
4123
4124/**
4125 * dev_put - release reference to device
4126 * @dev: network device
4127 *
4128 * Release reference to device to allow it to be freed.
4129 * Try using netdev_put() instead.
4130 */
4131static inline void dev_put(struct net_device *dev)
4132{
4133 netdev_put(dev, NULL);
4134}
4135
4136static inline void netdev_ref_replace(struct net_device *odev,
4137 struct net_device *ndev,
4138 netdevice_tracker *tracker,
4139 gfp_t gfp)
4140{
4141 if (odev)
4142 netdev_tracker_free(odev, tracker);
4143
4144 __dev_hold(ndev);
4145 __dev_put(odev);
4146
4147 if (ndev)
4148 __netdev_tracker_alloc(ndev, tracker, gfp);
4149}
4150
4151/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4152 * and _off may be called from IRQ context, but it is caller
4153 * who is responsible for serialization of these calls.
4154 *
4155 * The name carrier is inappropriate, these functions should really be
4156 * called netif_lowerlayer_*() because they represent the state of any
4157 * kind of lower layer not just hardware media.
4158 */
4159void linkwatch_fire_event(struct net_device *dev);
4160
4161/**
4162 * netif_carrier_ok - test if carrier present
4163 * @dev: network device
4164 *
4165 * Check if carrier is present on device
4166 */
4167static inline bool netif_carrier_ok(const struct net_device *dev)
4168{
4169 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4170}
4171
4172unsigned long dev_trans_start(struct net_device *dev);
4173
4174void __netdev_watchdog_up(struct net_device *dev);
4175
4176void netif_carrier_on(struct net_device *dev);
4177void netif_carrier_off(struct net_device *dev);
4178void netif_carrier_event(struct net_device *dev);
4179
4180/**
4181 * netif_dormant_on - mark device as dormant.
4182 * @dev: network device
4183 *
4184 * Mark device as dormant (as per RFC2863).
4185 *
4186 * The dormant state indicates that the relevant interface is not
4187 * actually in a condition to pass packets (i.e., it is not 'up') but is
4188 * in a "pending" state, waiting for some external event. For "on-
4189 * demand" interfaces, this new state identifies the situation where the
4190 * interface is waiting for events to place it in the up state.
4191 */
4192static inline void netif_dormant_on(struct net_device *dev)
4193{
4194 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4195 linkwatch_fire_event(dev);
4196}
4197
4198/**
4199 * netif_dormant_off - set device as not dormant.
4200 * @dev: network device
4201 *
4202 * Device is not in dormant state.
4203 */
4204static inline void netif_dormant_off(struct net_device *dev)
4205{
4206 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4207 linkwatch_fire_event(dev);
4208}
4209
4210/**
4211 * netif_dormant - test if device is dormant
4212 * @dev: network device
4213 *
4214 * Check if device is dormant.
4215 */
4216static inline bool netif_dormant(const struct net_device *dev)
4217{
4218 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4219}
4220
4221
4222/**
4223 * netif_testing_on - mark device as under test.
4224 * @dev: network device
4225 *
4226 * Mark device as under test (as per RFC2863).
4227 *
4228 * The testing state indicates that some test(s) must be performed on
4229 * the interface. After completion, of the test, the interface state
4230 * will change to up, dormant, or down, as appropriate.
4231 */
4232static inline void netif_testing_on(struct net_device *dev)
4233{
4234 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4235 linkwatch_fire_event(dev);
4236}
4237
4238/**
4239 * netif_testing_off - set device as not under test.
4240 * @dev: network device
4241 *
4242 * Device is not in testing state.
4243 */
4244static inline void netif_testing_off(struct net_device *dev)
4245{
4246 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4247 linkwatch_fire_event(dev);
4248}
4249
4250/**
4251 * netif_testing - test if device is under test
4252 * @dev: network device
4253 *
4254 * Check if device is under test
4255 */
4256static inline bool netif_testing(const struct net_device *dev)
4257{
4258 return test_bit(__LINK_STATE_TESTING, &dev->state);
4259}
4260
4261
4262/**
4263 * netif_oper_up - test if device is operational
4264 * @dev: network device
4265 *
4266 * Check if carrier is operational
4267 */
4268static inline bool netif_oper_up(const struct net_device *dev)
4269{
4270 return (dev->operstate == IF_OPER_UP ||
4271 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4272}
4273
4274/**
4275 * netif_device_present - is device available or removed
4276 * @dev: network device
4277 *
4278 * Check if device has not been removed from system.
4279 */
4280static inline bool netif_device_present(const struct net_device *dev)
4281{
4282 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4283}
4284
4285void netif_device_detach(struct net_device *dev);
4286
4287void netif_device_attach(struct net_device *dev);
4288
4289/*
4290 * Network interface message level settings
4291 */
4292
4293enum {
4294 NETIF_MSG_DRV_BIT,
4295 NETIF_MSG_PROBE_BIT,
4296 NETIF_MSG_LINK_BIT,
4297 NETIF_MSG_TIMER_BIT,
4298 NETIF_MSG_IFDOWN_BIT,
4299 NETIF_MSG_IFUP_BIT,
4300 NETIF_MSG_RX_ERR_BIT,
4301 NETIF_MSG_TX_ERR_BIT,
4302 NETIF_MSG_TX_QUEUED_BIT,
4303 NETIF_MSG_INTR_BIT,
4304 NETIF_MSG_TX_DONE_BIT,
4305 NETIF_MSG_RX_STATUS_BIT,
4306 NETIF_MSG_PKTDATA_BIT,
4307 NETIF_MSG_HW_BIT,
4308 NETIF_MSG_WOL_BIT,
4309
4310 /* When you add a new bit above, update netif_msg_class_names array
4311 * in net/ethtool/common.c
4312 */
4313 NETIF_MSG_CLASS_COUNT,
4314};
4315/* Both ethtool_ops interface and internal driver implementation use u32 */
4316static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4317
4318#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4319#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4320
4321#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4322#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4323#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4324#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4325#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4326#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4327#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4328#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4329#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4330#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4331#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4332#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4333#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4334#define NETIF_MSG_HW __NETIF_MSG(HW)
4335#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4336
4337#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4338#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4339#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4340#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4341#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4342#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4343#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4344#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4345#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4346#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4347#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4348#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4349#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4350#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4351#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4352
4353static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4354{
4355 /* use default */
4356 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4357 return default_msg_enable_bits;
4358 if (debug_value == 0) /* no output */
4359 return 0;
4360 /* set low N bits */
4361 return (1U << debug_value) - 1;
4362}
4363
4364static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4365{
4366 spin_lock(&txq->_xmit_lock);
4367 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4368 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4369}
4370
4371static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4372{
4373 __acquire(&txq->_xmit_lock);
4374 return true;
4375}
4376
4377static inline void __netif_tx_release(struct netdev_queue *txq)
4378{
4379 __release(&txq->_xmit_lock);
4380}
4381
4382static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4383{
4384 spin_lock_bh(&txq->_xmit_lock);
4385 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4386 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4387}
4388
4389static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4390{
4391 bool ok = spin_trylock(&txq->_xmit_lock);
4392
4393 if (likely(ok)) {
4394 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4395 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4396 }
4397 return ok;
4398}
4399
4400static inline void __netif_tx_unlock(struct netdev_queue *txq)
4401{
4402 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4403 WRITE_ONCE(txq->xmit_lock_owner, -1);
4404 spin_unlock(&txq->_xmit_lock);
4405}
4406
4407static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4408{
4409 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4410 WRITE_ONCE(txq->xmit_lock_owner, -1);
4411 spin_unlock_bh(&txq->_xmit_lock);
4412}
4413
4414/*
4415 * txq->trans_start can be read locklessly from dev_watchdog()
4416 */
4417static inline void txq_trans_update(struct netdev_queue *txq)
4418{
4419 if (txq->xmit_lock_owner != -1)
4420 WRITE_ONCE(txq->trans_start, jiffies);
4421}
4422
4423static inline void txq_trans_cond_update(struct netdev_queue *txq)
4424{
4425 unsigned long now = jiffies;
4426
4427 if (READ_ONCE(txq->trans_start) != now)
4428 WRITE_ONCE(txq->trans_start, now);
4429}
4430
4431/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4432static inline void netif_trans_update(struct net_device *dev)
4433{
4434 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4435
4436 txq_trans_cond_update(txq);
4437}
4438
4439/**
4440 * netif_tx_lock - grab network device transmit lock
4441 * @dev: network device
4442 *
4443 * Get network device transmit lock
4444 */
4445void netif_tx_lock(struct net_device *dev);
4446
4447static inline void netif_tx_lock_bh(struct net_device *dev)
4448{
4449 local_bh_disable();
4450 netif_tx_lock(dev);
4451}
4452
4453void netif_tx_unlock(struct net_device *dev);
4454
4455static inline void netif_tx_unlock_bh(struct net_device *dev)
4456{
4457 netif_tx_unlock(dev);
4458 local_bh_enable();
4459}
4460
4461#define HARD_TX_LOCK(dev, txq, cpu) { \
4462 if ((dev->features & NETIF_F_LLTX) == 0) { \
4463 __netif_tx_lock(txq, cpu); \
4464 } else { \
4465 __netif_tx_acquire(txq); \
4466 } \
4467}
4468
4469#define HARD_TX_TRYLOCK(dev, txq) \
4470 (((dev->features & NETIF_F_LLTX) == 0) ? \
4471 __netif_tx_trylock(txq) : \
4472 __netif_tx_acquire(txq))
4473
4474#define HARD_TX_UNLOCK(dev, txq) { \
4475 if ((dev->features & NETIF_F_LLTX) == 0) { \
4476 __netif_tx_unlock(txq); \
4477 } else { \
4478 __netif_tx_release(txq); \
4479 } \
4480}
4481
4482static inline void netif_tx_disable(struct net_device *dev)
4483{
4484 unsigned int i;
4485 int cpu;
4486
4487 local_bh_disable();
4488 cpu = smp_processor_id();
4489 spin_lock(&dev->tx_global_lock);
4490 for (i = 0; i < dev->num_tx_queues; i++) {
4491 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4492
4493 __netif_tx_lock(txq, cpu);
4494 netif_tx_stop_queue(txq);
4495 __netif_tx_unlock(txq);
4496 }
4497 spin_unlock(&dev->tx_global_lock);
4498 local_bh_enable();
4499}
4500
4501static inline void netif_addr_lock(struct net_device *dev)
4502{
4503 unsigned char nest_level = 0;
4504
4505#ifdef CONFIG_LOCKDEP
4506 nest_level = dev->nested_level;
4507#endif
4508 spin_lock_nested(&dev->addr_list_lock, nest_level);
4509}
4510
4511static inline void netif_addr_lock_bh(struct net_device *dev)
4512{
4513 unsigned char nest_level = 0;
4514
4515#ifdef CONFIG_LOCKDEP
4516 nest_level = dev->nested_level;
4517#endif
4518 local_bh_disable();
4519 spin_lock_nested(&dev->addr_list_lock, nest_level);
4520}
4521
4522static inline void netif_addr_unlock(struct net_device *dev)
4523{
4524 spin_unlock(&dev->addr_list_lock);
4525}
4526
4527static inline void netif_addr_unlock_bh(struct net_device *dev)
4528{
4529 spin_unlock_bh(&dev->addr_list_lock);
4530}
4531
4532/*
4533 * dev_addrs walker. Should be used only for read access. Call with
4534 * rcu_read_lock held.
4535 */
4536#define for_each_dev_addr(dev, ha) \
4537 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4538
4539/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4540
4541void ether_setup(struct net_device *dev);
4542
4543/* Support for loadable net-drivers */
4544struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4545 unsigned char name_assign_type,
4546 void (*setup)(struct net_device *),
4547 unsigned int txqs, unsigned int rxqs);
4548#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4549 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4550
4551#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4552 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4553 count)
4554
4555int register_netdev(struct net_device *dev);
4556void unregister_netdev(struct net_device *dev);
4557
4558int devm_register_netdev(struct device *dev, struct net_device *ndev);
4559
4560/* General hardware address lists handling functions */
4561int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4562 struct netdev_hw_addr_list *from_list, int addr_len);
4563void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4564 struct netdev_hw_addr_list *from_list, int addr_len);
4565int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4566 struct net_device *dev,
4567 int (*sync)(struct net_device *, const unsigned char *),
4568 int (*unsync)(struct net_device *,
4569 const unsigned char *));
4570int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4571 struct net_device *dev,
4572 int (*sync)(struct net_device *,
4573 const unsigned char *, int),
4574 int (*unsync)(struct net_device *,
4575 const unsigned char *, int));
4576void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4577 struct net_device *dev,
4578 int (*unsync)(struct net_device *,
4579 const unsigned char *, int));
4580void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4581 struct net_device *dev,
4582 int (*unsync)(struct net_device *,
4583 const unsigned char *));
4584void __hw_addr_init(struct netdev_hw_addr_list *list);
4585
4586/* Functions used for device addresses handling */
4587void dev_addr_mod(struct net_device *dev, unsigned int offset,
4588 const void *addr, size_t len);
4589
4590static inline void
4591__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4592{
4593 dev_addr_mod(dev, 0, addr, len);
4594}
4595
4596static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4597{
4598 __dev_addr_set(dev, addr, dev->addr_len);
4599}
4600
4601int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4602 unsigned char addr_type);
4603int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4604 unsigned char addr_type);
4605
4606/* Functions used for unicast addresses handling */
4607int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4608int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4609int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4610int dev_uc_sync(struct net_device *to, struct net_device *from);
4611int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4612void dev_uc_unsync(struct net_device *to, struct net_device *from);
4613void dev_uc_flush(struct net_device *dev);
4614void dev_uc_init(struct net_device *dev);
4615
4616/**
4617 * __dev_uc_sync - Synchonize device's unicast list
4618 * @dev: device to sync
4619 * @sync: function to call if address should be added
4620 * @unsync: function to call if address should be removed
4621 *
4622 * Add newly added addresses to the interface, and release
4623 * addresses that have been deleted.
4624 */
4625static inline int __dev_uc_sync(struct net_device *dev,
4626 int (*sync)(struct net_device *,
4627 const unsigned char *),
4628 int (*unsync)(struct net_device *,
4629 const unsigned char *))
4630{
4631 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4632}
4633
4634/**
4635 * __dev_uc_unsync - Remove synchronized addresses from device
4636 * @dev: device to sync
4637 * @unsync: function to call if address should be removed
4638 *
4639 * Remove all addresses that were added to the device by dev_uc_sync().
4640 */
4641static inline void __dev_uc_unsync(struct net_device *dev,
4642 int (*unsync)(struct net_device *,
4643 const unsigned char *))
4644{
4645 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4646}
4647
4648/* Functions used for multicast addresses handling */
4649int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4650int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4651int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4652int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4653int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4654int dev_mc_sync(struct net_device *to, struct net_device *from);
4655int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4656void dev_mc_unsync(struct net_device *to, struct net_device *from);
4657void dev_mc_flush(struct net_device *dev);
4658void dev_mc_init(struct net_device *dev);
4659
4660/**
4661 * __dev_mc_sync - Synchonize device's multicast list
4662 * @dev: device to sync
4663 * @sync: function to call if address should be added
4664 * @unsync: function to call if address should be removed
4665 *
4666 * Add newly added addresses to the interface, and release
4667 * addresses that have been deleted.
4668 */
4669static inline int __dev_mc_sync(struct net_device *dev,
4670 int (*sync)(struct net_device *,
4671 const unsigned char *),
4672 int (*unsync)(struct net_device *,
4673 const unsigned char *))
4674{
4675 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4676}
4677
4678/**
4679 * __dev_mc_unsync - Remove synchronized addresses from device
4680 * @dev: device to sync
4681 * @unsync: function to call if address should be removed
4682 *
4683 * Remove all addresses that were added to the device by dev_mc_sync().
4684 */
4685static inline void __dev_mc_unsync(struct net_device *dev,
4686 int (*unsync)(struct net_device *,
4687 const unsigned char *))
4688{
4689 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4690}
4691
4692/* Functions used for secondary unicast and multicast support */
4693void dev_set_rx_mode(struct net_device *dev);
4694int dev_set_promiscuity(struct net_device *dev, int inc);
4695int dev_set_allmulti(struct net_device *dev, int inc);
4696void netdev_state_change(struct net_device *dev);
4697void __netdev_notify_peers(struct net_device *dev);
4698void netdev_notify_peers(struct net_device *dev);
4699void netdev_features_change(struct net_device *dev);
4700/* Load a device via the kmod */
4701void dev_load(struct net *net, const char *name);
4702struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4703 struct rtnl_link_stats64 *storage);
4704void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4705 const struct net_device_stats *netdev_stats);
4706void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4707 const struct pcpu_sw_netstats __percpu *netstats);
4708void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4709
4710extern int netdev_max_backlog;
4711extern int dev_rx_weight;
4712extern int dev_tx_weight;
4713extern int gro_normal_batch;
4714
4715enum {
4716 NESTED_SYNC_IMM_BIT,
4717 NESTED_SYNC_TODO_BIT,
4718};
4719
4720#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4721#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4722
4723#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4724#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4725
4726struct netdev_nested_priv {
4727 unsigned char flags;
4728 void *data;
4729};
4730
4731bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4732struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4733 struct list_head **iter);
4734
4735/* iterate through upper list, must be called under RCU read lock */
4736#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4737 for (iter = &(dev)->adj_list.upper, \
4738 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4739 updev; \
4740 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4741
4742int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4743 int (*fn)(struct net_device *upper_dev,
4744 struct netdev_nested_priv *priv),
4745 struct netdev_nested_priv *priv);
4746
4747bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4748 struct net_device *upper_dev);
4749
4750bool netdev_has_any_upper_dev(struct net_device *dev);
4751
4752void *netdev_lower_get_next_private(struct net_device *dev,
4753 struct list_head **iter);
4754void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4755 struct list_head **iter);
4756
4757#define netdev_for_each_lower_private(dev, priv, iter) \
4758 for (iter = (dev)->adj_list.lower.next, \
4759 priv = netdev_lower_get_next_private(dev, &(iter)); \
4760 priv; \
4761 priv = netdev_lower_get_next_private(dev, &(iter)))
4762
4763#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4764 for (iter = &(dev)->adj_list.lower, \
4765 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4766 priv; \
4767 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4768
4769void *netdev_lower_get_next(struct net_device *dev,
4770 struct list_head **iter);
4771
4772#define netdev_for_each_lower_dev(dev, ldev, iter) \
4773 for (iter = (dev)->adj_list.lower.next, \
4774 ldev = netdev_lower_get_next(dev, &(iter)); \
4775 ldev; \
4776 ldev = netdev_lower_get_next(dev, &(iter)))
4777
4778struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4779 struct list_head **iter);
4780int netdev_walk_all_lower_dev(struct net_device *dev,
4781 int (*fn)(struct net_device *lower_dev,
4782 struct netdev_nested_priv *priv),
4783 struct netdev_nested_priv *priv);
4784int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4785 int (*fn)(struct net_device *lower_dev,
4786 struct netdev_nested_priv *priv),
4787 struct netdev_nested_priv *priv);
4788
4789void *netdev_adjacent_get_private(struct list_head *adj_list);
4790void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4791struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4792struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4793int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4794 struct netlink_ext_ack *extack);
4795int netdev_master_upper_dev_link(struct net_device *dev,
4796 struct net_device *upper_dev,
4797 void *upper_priv, void *upper_info,
4798 struct netlink_ext_ack *extack);
4799void netdev_upper_dev_unlink(struct net_device *dev,
4800 struct net_device *upper_dev);
4801int netdev_adjacent_change_prepare(struct net_device *old_dev,
4802 struct net_device *new_dev,
4803 struct net_device *dev,
4804 struct netlink_ext_ack *extack);
4805void netdev_adjacent_change_commit(struct net_device *old_dev,
4806 struct net_device *new_dev,
4807 struct net_device *dev);
4808void netdev_adjacent_change_abort(struct net_device *old_dev,
4809 struct net_device *new_dev,
4810 struct net_device *dev);
4811void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4812void *netdev_lower_dev_get_private(struct net_device *dev,
4813 struct net_device *lower_dev);
4814void netdev_lower_state_changed(struct net_device *lower_dev,
4815 void *lower_state_info);
4816
4817/* RSS keys are 40 or 52 bytes long */
4818#define NETDEV_RSS_KEY_LEN 52
4819extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4820void netdev_rss_key_fill(void *buffer, size_t len);
4821
4822int skb_checksum_help(struct sk_buff *skb);
4823int skb_crc32c_csum_help(struct sk_buff *skb);
4824int skb_csum_hwoffload_help(struct sk_buff *skb,
4825 const netdev_features_t features);
4826
4827struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4828 netdev_features_t features, bool tx_path);
4829struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4830 netdev_features_t features, __be16 type);
4831struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4832 netdev_features_t features);
4833
4834struct netdev_bonding_info {
4835 ifslave slave;
4836 ifbond master;
4837};
4838
4839struct netdev_notifier_bonding_info {
4840 struct netdev_notifier_info info; /* must be first */
4841 struct netdev_bonding_info bonding_info;
4842};
4843
4844void netdev_bonding_info_change(struct net_device *dev,
4845 struct netdev_bonding_info *bonding_info);
4846
4847#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4848void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4849#else
4850static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4851 const void *data)
4852{
4853}
4854#endif
4855
4856static inline
4857struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4858{
4859 return __skb_gso_segment(skb, features, true);
4860}
4861__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4862
4863static inline bool can_checksum_protocol(netdev_features_t features,
4864 __be16 protocol)
4865{
4866 if (protocol == htons(ETH_P_FCOE))
4867 return !!(features & NETIF_F_FCOE_CRC);
4868
4869 /* Assume this is an IP checksum (not SCTP CRC) */
4870
4871 if (features & NETIF_F_HW_CSUM) {
4872 /* Can checksum everything */
4873 return true;
4874 }
4875
4876 switch (protocol) {
4877 case htons(ETH_P_IP):
4878 return !!(features & NETIF_F_IP_CSUM);
4879 case htons(ETH_P_IPV6):
4880 return !!(features & NETIF_F_IPV6_CSUM);
4881 default:
4882 return false;
4883 }
4884}
4885
4886#ifdef CONFIG_BUG
4887void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4888#else
4889static inline void netdev_rx_csum_fault(struct net_device *dev,
4890 struct sk_buff *skb)
4891{
4892}
4893#endif
4894/* rx skb timestamps */
4895void net_enable_timestamp(void);
4896void net_disable_timestamp(void);
4897
4898static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4899 const struct skb_shared_hwtstamps *hwtstamps,
4900 bool cycles)
4901{
4902 const struct net_device_ops *ops = dev->netdev_ops;
4903
4904 if (ops->ndo_get_tstamp)
4905 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4906
4907 return hwtstamps->hwtstamp;
4908}
4909
4910static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4911 struct sk_buff *skb, struct net_device *dev,
4912 bool more)
4913{
4914 __this_cpu_write(softnet_data.xmit.more, more);
4915 return ops->ndo_start_xmit(skb, dev);
4916}
4917
4918static inline bool netdev_xmit_more(void)
4919{
4920 return __this_cpu_read(softnet_data.xmit.more);
4921}
4922
4923static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4924 struct netdev_queue *txq, bool more)
4925{
4926 const struct net_device_ops *ops = dev->netdev_ops;
4927 netdev_tx_t rc;
4928
4929 rc = __netdev_start_xmit(ops, skb, dev, more);
4930 if (rc == NETDEV_TX_OK)
4931 txq_trans_update(txq);
4932
4933 return rc;
4934}
4935
4936int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4937 const void *ns);
4938void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4939 const void *ns);
4940
4941extern const struct kobj_ns_type_operations net_ns_type_operations;
4942
4943const char *netdev_drivername(const struct net_device *dev);
4944
4945static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4946 netdev_features_t f2)
4947{
4948 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4949 if (f1 & NETIF_F_HW_CSUM)
4950 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4951 else
4952 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4953 }
4954
4955 return f1 & f2;
4956}
4957
4958static inline netdev_features_t netdev_get_wanted_features(
4959 struct net_device *dev)
4960{
4961 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4962}
4963netdev_features_t netdev_increment_features(netdev_features_t all,
4964 netdev_features_t one, netdev_features_t mask);
4965
4966/* Allow TSO being used on stacked device :
4967 * Performing the GSO segmentation before last device
4968 * is a performance improvement.
4969 */
4970static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4971 netdev_features_t mask)
4972{
4973 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4974}
4975
4976int __netdev_update_features(struct net_device *dev);
4977void netdev_update_features(struct net_device *dev);
4978void netdev_change_features(struct net_device *dev);
4979
4980void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4981 struct net_device *dev);
4982
4983netdev_features_t passthru_features_check(struct sk_buff *skb,
4984 struct net_device *dev,
4985 netdev_features_t features);
4986netdev_features_t netif_skb_features(struct sk_buff *skb);
4987
4988static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4989{
4990 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4991
4992 /* check flags correspondence */
4993 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4994 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4995 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4996 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4997 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4998 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4999 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5000 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5001 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5002 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5003 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5004 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5005 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5006 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5007 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5008 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5009 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5010 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5011 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5012
5013 return (features & feature) == feature;
5014}
5015
5016static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5017{
5018 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5019 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5020}
5021
5022static inline bool netif_needs_gso(struct sk_buff *skb,
5023 netdev_features_t features)
5024{
5025 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5026 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5027 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5028}
5029
5030void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5031void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5032void netif_inherit_tso_max(struct net_device *to,
5033 const struct net_device *from);
5034
5035static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5036 int pulled_hlen, u16 mac_offset,
5037 int mac_len)
5038{
5039 skb->protocol = protocol;
5040 skb->encapsulation = 1;
5041 skb_push(skb, pulled_hlen);
5042 skb_reset_transport_header(skb);
5043 skb->mac_header = mac_offset;
5044 skb->network_header = skb->mac_header + mac_len;
5045 skb->mac_len = mac_len;
5046}
5047
5048static inline bool netif_is_macsec(const struct net_device *dev)
5049{
5050 return dev->priv_flags & IFF_MACSEC;
5051}
5052
5053static inline bool netif_is_macvlan(const struct net_device *dev)
5054{
5055 return dev->priv_flags & IFF_MACVLAN;
5056}
5057
5058static inline bool netif_is_macvlan_port(const struct net_device *dev)
5059{
5060 return dev->priv_flags & IFF_MACVLAN_PORT;
5061}
5062
5063static inline bool netif_is_bond_master(const struct net_device *dev)
5064{
5065 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5066}
5067
5068static inline bool netif_is_bond_slave(const struct net_device *dev)
5069{
5070 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5071}
5072
5073static inline bool netif_supports_nofcs(struct net_device *dev)
5074{
5075 return dev->priv_flags & IFF_SUPP_NOFCS;
5076}
5077
5078static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5079{
5080 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5081}
5082
5083static inline bool netif_is_l3_master(const struct net_device *dev)
5084{
5085 return dev->priv_flags & IFF_L3MDEV_MASTER;
5086}
5087
5088static inline bool netif_is_l3_slave(const struct net_device *dev)
5089{
5090 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5091}
5092
5093static inline bool netif_is_bridge_master(const struct net_device *dev)
5094{
5095 return dev->priv_flags & IFF_EBRIDGE;
5096}
5097
5098static inline bool netif_is_bridge_port(const struct net_device *dev)
5099{
5100 return dev->priv_flags & IFF_BRIDGE_PORT;
5101}
5102
5103static inline bool netif_is_ovs_master(const struct net_device *dev)
5104{
5105 return dev->priv_flags & IFF_OPENVSWITCH;
5106}
5107
5108static inline bool netif_is_ovs_port(const struct net_device *dev)
5109{
5110 return dev->priv_flags & IFF_OVS_DATAPATH;
5111}
5112
5113static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5114{
5115 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5116}
5117
5118static inline bool netif_is_team_master(const struct net_device *dev)
5119{
5120 return dev->priv_flags & IFF_TEAM;
5121}
5122
5123static inline bool netif_is_team_port(const struct net_device *dev)
5124{
5125 return dev->priv_flags & IFF_TEAM_PORT;
5126}
5127
5128static inline bool netif_is_lag_master(const struct net_device *dev)
5129{
5130 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5131}
5132
5133static inline bool netif_is_lag_port(const struct net_device *dev)
5134{
5135 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5136}
5137
5138static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5139{
5140 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5141}
5142
5143static inline bool netif_is_failover(const struct net_device *dev)
5144{
5145 return dev->priv_flags & IFF_FAILOVER;
5146}
5147
5148static inline bool netif_is_failover_slave(const struct net_device *dev)
5149{
5150 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5151}
5152
5153/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5154static inline void netif_keep_dst(struct net_device *dev)
5155{
5156 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5157}
5158
5159/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5160static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5161{
5162 /* TODO: reserve and use an additional IFF bit, if we get more users */
5163 return netif_is_macsec(dev);
5164}
5165
5166extern struct pernet_operations __net_initdata loopback_net_ops;
5167
5168/* Logging, debugging and troubleshooting/diagnostic helpers. */
5169
5170/* netdev_printk helpers, similar to dev_printk */
5171
5172static inline const char *netdev_name(const struct net_device *dev)
5173{
5174 if (!dev->name[0] || strchr(dev->name, '%'))
5175 return "(unnamed net_device)";
5176 return dev->name;
5177}
5178
5179static inline const char *netdev_reg_state(const struct net_device *dev)
5180{
5181 switch (dev->reg_state) {
5182 case NETREG_UNINITIALIZED: return " (uninitialized)";
5183 case NETREG_REGISTERED: return "";
5184 case NETREG_UNREGISTERING: return " (unregistering)";
5185 case NETREG_UNREGISTERED: return " (unregistered)";
5186 case NETREG_RELEASED: return " (released)";
5187 case NETREG_DUMMY: return " (dummy)";
5188 }
5189
5190 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5191 return " (unknown)";
5192}
5193
5194#define MODULE_ALIAS_NETDEV(device) \
5195 MODULE_ALIAS("netdev-" device)
5196
5197/*
5198 * netdev_WARN() acts like dev_printk(), but with the key difference
5199 * of using a WARN/WARN_ON to get the message out, including the
5200 * file/line information and a backtrace.
5201 */
5202#define netdev_WARN(dev, format, args...) \
5203 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5204 netdev_reg_state(dev), ##args)
5205
5206#define netdev_WARN_ONCE(dev, format, args...) \
5207 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5208 netdev_reg_state(dev), ##args)
5209
5210/*
5211 * The list of packet types we will receive (as opposed to discard)
5212 * and the routines to invoke.
5213 *
5214 * Why 16. Because with 16 the only overlap we get on a hash of the
5215 * low nibble of the protocol value is RARP/SNAP/X.25.
5216 *
5217 * 0800 IP
5218 * 0001 802.3
5219 * 0002 AX.25
5220 * 0004 802.2
5221 * 8035 RARP
5222 * 0005 SNAP
5223 * 0805 X.25
5224 * 0806 ARP
5225 * 8137 IPX
5226 * 0009 Localtalk
5227 * 86DD IPv6
5228 */
5229#define PTYPE_HASH_SIZE (16)
5230#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5231
5232extern struct list_head ptype_all __read_mostly;
5233extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5234
5235extern struct net_device *blackhole_netdev;
5236
5237/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5238#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5239#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5240 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5241
5242#endif /* _LINUX_NETDEVICE_H */