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 <linux/netdev_features.h>
44#include <linux/neighbour.h>
45#include <linux/netdevice_xmit.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <uapi/linux/netdev.h>
50#include <linux/hashtable.h>
51#include <linux/rbtree.h>
52#include <net/net_trackers.h>
53#include <net/net_debug.h>
54#include <net/dropreason-core.h>
55#include <net/neighbour_tables.h>
56
57struct netpoll_info;
58struct device;
59struct ethtool_ops;
60struct kernel_hwtstamp_config;
61struct phy_device;
62struct dsa_port;
63struct ip_tunnel_parm_kern;
64struct macsec_context;
65struct macsec_ops;
66struct netdev_config;
67struct netdev_name_node;
68struct sd_flow_limit;
69struct sfp_bus;
70/* 802.11 specific */
71struct wireless_dev;
72/* 802.15.4 specific */
73struct wpan_dev;
74struct mpls_dev;
75/* UDP Tunnel offloads */
76struct udp_tunnel_info;
77struct udp_tunnel_nic_info;
78struct udp_tunnel_nic;
79struct bpf_prog;
80struct xdp_buff;
81struct xdp_frame;
82struct xdp_metadata_ops;
83struct xdp_md;
84struct ethtool_netdev_state;
85struct phy_link_topology;
86struct hwtstamp_provider;
87
88typedef u32 xdp_features_t;
89
90void synchronize_net(void);
91void netdev_set_default_ethtool_ops(struct net_device *dev,
92 const struct ethtool_ops *ops);
93void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
94
95/* Backlog congestion levels */
96#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
97#define NET_RX_DROP 1 /* packet dropped */
98
99#define MAX_NEST_DEV 8
100
101/*
102 * Transmit return codes: transmit return codes originate from three different
103 * namespaces:
104 *
105 * - qdisc return codes
106 * - driver transmit return codes
107 * - errno values
108 *
109 * Drivers are allowed to return any one of those in their hard_start_xmit()
110 * function. Real network devices commonly used with qdiscs should only return
111 * the driver transmit return codes though - when qdiscs are used, the actual
112 * transmission happens asynchronously, so the value is not propagated to
113 * higher layers. Virtual network devices transmit synchronously; in this case
114 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
115 * others are propagated to higher layers.
116 */
117
118/* qdisc ->enqueue() return codes. */
119#define NET_XMIT_SUCCESS 0x00
120#define NET_XMIT_DROP 0x01 /* skb dropped */
121#define NET_XMIT_CN 0x02 /* congestion notification */
122#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
123
124/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
125 * indicates that the device will soon be dropping packets, or already drops
126 * some packets of the same priority; prompting us to send less aggressively. */
127#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
128#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
129
130/* Driver transmit return codes */
131#define NETDEV_TX_MASK 0xf0
132
133enum netdev_tx {
134 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
135 NETDEV_TX_OK = 0x00, /* driver took care of packet */
136 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
137};
138typedef enum netdev_tx netdev_tx_t;
139
140/*
141 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
142 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
143 */
144static inline bool dev_xmit_complete(int rc)
145{
146 /*
147 * Positive cases with an skb consumed by a driver:
148 * - successful transmission (rc == NETDEV_TX_OK)
149 * - error while transmitting (rc < 0)
150 * - error while queueing to a different device (rc & NET_XMIT_MASK)
151 */
152 if (likely(rc < NET_XMIT_MASK))
153 return true;
154
155 return false;
156}
157
158/*
159 * Compute the worst-case header length according to the protocols
160 * used.
161 */
162
163#if defined(CONFIG_HYPERV_NET)
164# define LL_MAX_HEADER 128
165#elif defined(CONFIG_WLAN)
166# if defined(CONFIG_MAC80211_MESH)
167# define LL_MAX_HEADER 128
168# else
169# define LL_MAX_HEADER 96
170# endif
171#else
172# define LL_MAX_HEADER 32
173#endif
174
175#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
176 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
177#define MAX_HEADER LL_MAX_HEADER
178#else
179#define MAX_HEADER (LL_MAX_HEADER + 48)
180#endif
181
182/*
183 * Old network device statistics. Fields are native words
184 * (unsigned long) so they can be read and written atomically.
185 */
186
187#define NET_DEV_STAT(FIELD) \
188 union { \
189 unsigned long FIELD; \
190 atomic_long_t __##FIELD; \
191 }
192
193struct net_device_stats {
194 NET_DEV_STAT(rx_packets);
195 NET_DEV_STAT(tx_packets);
196 NET_DEV_STAT(rx_bytes);
197 NET_DEV_STAT(tx_bytes);
198 NET_DEV_STAT(rx_errors);
199 NET_DEV_STAT(tx_errors);
200 NET_DEV_STAT(rx_dropped);
201 NET_DEV_STAT(tx_dropped);
202 NET_DEV_STAT(multicast);
203 NET_DEV_STAT(collisions);
204 NET_DEV_STAT(rx_length_errors);
205 NET_DEV_STAT(rx_over_errors);
206 NET_DEV_STAT(rx_crc_errors);
207 NET_DEV_STAT(rx_frame_errors);
208 NET_DEV_STAT(rx_fifo_errors);
209 NET_DEV_STAT(rx_missed_errors);
210 NET_DEV_STAT(tx_aborted_errors);
211 NET_DEV_STAT(tx_carrier_errors);
212 NET_DEV_STAT(tx_fifo_errors);
213 NET_DEV_STAT(tx_heartbeat_errors);
214 NET_DEV_STAT(tx_window_errors);
215 NET_DEV_STAT(rx_compressed);
216 NET_DEV_STAT(tx_compressed);
217};
218#undef NET_DEV_STAT
219
220/* per-cpu stats, allocated on demand.
221 * Try to fit them in a single cache line, for dev_get_stats() sake.
222 */
223struct net_device_core_stats {
224 unsigned long rx_dropped;
225 unsigned long tx_dropped;
226 unsigned long rx_nohandler;
227 unsigned long rx_otherhost_dropped;
228} __aligned(4 * sizeof(unsigned long));
229
230#include <linux/cache.h>
231#include <linux/skbuff.h>
232
233struct neighbour;
234struct neigh_parms;
235struct sk_buff;
236
237struct netdev_hw_addr {
238 struct list_head list;
239 struct rb_node node;
240 unsigned char addr[MAX_ADDR_LEN];
241 unsigned char type;
242#define NETDEV_HW_ADDR_T_LAN 1
243#define NETDEV_HW_ADDR_T_SAN 2
244#define NETDEV_HW_ADDR_T_UNICAST 3
245#define NETDEV_HW_ADDR_T_MULTICAST 4
246 bool global_use;
247 int sync_cnt;
248 int refcount;
249 int synced;
250 struct rcu_head rcu_head;
251};
252
253struct netdev_hw_addr_list {
254 struct list_head list;
255 int count;
256
257 /* Auxiliary tree for faster lookup on addition and deletion */
258 struct rb_root tree;
259};
260
261#define netdev_hw_addr_list_count(l) ((l)->count)
262#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
263#define netdev_hw_addr_list_for_each(ha, l) \
264 list_for_each_entry(ha, &(l)->list, list)
265
266#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
267#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
268#define netdev_for_each_uc_addr(ha, dev) \
269 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
270#define netdev_for_each_synced_uc_addr(_ha, _dev) \
271 netdev_for_each_uc_addr((_ha), (_dev)) \
272 if ((_ha)->sync_cnt)
273
274#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
275#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
276#define netdev_for_each_mc_addr(ha, dev) \
277 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
278#define netdev_for_each_synced_mc_addr(_ha, _dev) \
279 netdev_for_each_mc_addr((_ha), (_dev)) \
280 if ((_ha)->sync_cnt)
281
282struct hh_cache {
283 unsigned int hh_len;
284 seqlock_t hh_lock;
285
286 /* cached hardware header; allow for machine alignment needs. */
287#define HH_DATA_MOD 16
288#define HH_DATA_OFF(__len) \
289 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
290#define HH_DATA_ALIGN(__len) \
291 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
292 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
293};
294
295/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
296 * Alternative is:
297 * dev->hard_header_len ? (dev->hard_header_len +
298 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
299 *
300 * We could use other alignment values, but we must maintain the
301 * relationship HH alignment <= LL alignment.
302 */
303#define LL_RESERVED_SPACE(dev) \
304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
306#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
309
310struct header_ops {
311 int (*create) (struct sk_buff *skb, struct net_device *dev,
312 unsigned short type, const void *daddr,
313 const void *saddr, unsigned int len);
314 int (*parse)(const struct sk_buff *skb,
315 const struct net_device *dev,
316 unsigned char *haddr);
317 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
318 void (*cache_update)(struct hh_cache *hh,
319 const struct net_device *dev,
320 const unsigned char *haddr);
321 bool (*validate)(const char *ll_header, unsigned int len);
322 __be16 (*parse_protocol)(const struct sk_buff *skb);
323};
324
325/* These flag bits are private to the generic network queueing
326 * layer; they may not be explicitly referenced by any other
327 * code.
328 */
329
330enum netdev_state_t {
331 __LINK_STATE_START,
332 __LINK_STATE_PRESENT,
333 __LINK_STATE_NOCARRIER,
334 __LINK_STATE_LINKWATCH_PENDING,
335 __LINK_STATE_DORMANT,
336 __LINK_STATE_TESTING,
337};
338
339struct gro_list {
340 struct list_head list;
341 int count;
342};
343
344/*
345 * size of gro hash buckets, must be <= the number of bits in
346 * gro_node::bitmask
347 */
348#define GRO_HASH_BUCKETS 8
349
350/**
351 * struct gro_node - structure to support Generic Receive Offload
352 * @bitmask: bitmask to indicate used buckets in @hash
353 * @hash: hashtable of pending aggregated skbs, separated by flows
354 * @rx_list: list of pending ``GRO_NORMAL`` skbs
355 * @rx_count: cached current length of @rx_list
356 * @cached_napi_id: napi_struct::napi_id cached for hotpath, 0 for standalone
357 */
358struct gro_node {
359 unsigned long bitmask;
360 struct gro_list hash[GRO_HASH_BUCKETS];
361 struct list_head rx_list;
362 u32 rx_count;
363 u32 cached_napi_id;
364};
365
366/*
367 * Structure for per-NAPI config
368 */
369struct napi_config {
370 u64 gro_flush_timeout;
371 u64 irq_suspend_timeout;
372 u32 defer_hard_irqs;
373 cpumask_t affinity_mask;
374 u8 threaded;
375 unsigned int napi_id;
376};
377
378/*
379 * Structure for NAPI scheduling similar to tasklet but with weighting
380 */
381struct napi_struct {
382 /* This field should be first or softnet_data.backlog needs tweaks. */
383 unsigned long state;
384 /* The poll_list must only be managed by the entity which
385 * changes the state of the NAPI_STATE_SCHED bit. This means
386 * whoever atomically sets that bit can add this napi_struct
387 * to the per-CPU poll_list, and whoever clears that bit
388 * can remove from the list right before clearing the bit.
389 */
390 struct list_head poll_list;
391
392 int weight;
393 u32 defer_hard_irqs_count;
394 int (*poll)(struct napi_struct *, int);
395#ifdef CONFIG_NETPOLL
396 /* CPU actively polling if netpoll is configured */
397 int poll_owner;
398#endif
399 /* CPU on which NAPI has been scheduled for processing */
400 int list_owner;
401 struct net_device *dev;
402 struct sk_buff *skb;
403 struct gro_node gro;
404 struct hrtimer timer;
405 /* all fields past this point are write-protected by netdev_lock */
406 struct task_struct *thread;
407 unsigned long gro_flush_timeout;
408 unsigned long irq_suspend_timeout;
409 u32 defer_hard_irqs;
410 /* control-path-only fields follow */
411 u32 napi_id;
412 struct list_head dev_list;
413 struct hlist_node napi_hash_node;
414 int irq;
415 struct irq_affinity_notify notify;
416 int napi_rmap_idx;
417 int index;
418 struct napi_config *config;
419};
420
421enum {
422 NAPI_STATE_SCHED, /* Poll is scheduled */
423 NAPI_STATE_MISSED, /* reschedule a napi */
424 NAPI_STATE_DISABLE, /* Disable pending */
425 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
426 NAPI_STATE_LISTED, /* NAPI added to system lists */
427 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
428 NAPI_STATE_IN_BUSY_POLL, /* Do not rearm NAPI interrupt */
429 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
430 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
431 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
432 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */
433 NAPI_STATE_THREADED_BUSY_POLL, /* The threaded NAPI poller will busy poll */
434};
435
436enum {
437 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
438 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
439 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
440 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
441 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
442 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
443 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
444 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
445 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
446 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
447 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER),
448 NAPIF_STATE_THREADED_BUSY_POLL = BIT(NAPI_STATE_THREADED_BUSY_POLL),
449};
450
451enum gro_result {
452 GRO_MERGED,
453 GRO_MERGED_FREE,
454 GRO_HELD,
455 GRO_NORMAL,
456 GRO_CONSUMED,
457};
458typedef enum gro_result gro_result_t;
459
460/*
461 * enum rx_handler_result - Possible return values for rx_handlers.
462 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
463 * further.
464 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
465 * case skb->dev was changed by rx_handler.
466 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
467 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
468 *
469 * rx_handlers are functions called from inside __netif_receive_skb(), to do
470 * special processing of the skb, prior to delivery to protocol handlers.
471 *
472 * Currently, a net_device can only have a single rx_handler registered. Trying
473 * to register a second rx_handler will return -EBUSY.
474 *
475 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
476 * To unregister a rx_handler on a net_device, use
477 * netdev_rx_handler_unregister().
478 *
479 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
480 * do with the skb.
481 *
482 * If the rx_handler consumed the skb in some way, it should return
483 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
484 * the skb to be delivered in some other way.
485 *
486 * If the rx_handler changed skb->dev, to divert the skb to another
487 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
488 * new device will be called if it exists.
489 *
490 * If the rx_handler decides the skb should be ignored, it should return
491 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
492 * are registered on exact device (ptype->dev == skb->dev).
493 *
494 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
495 * delivered, it should return RX_HANDLER_PASS.
496 *
497 * A device without a registered rx_handler will behave as if rx_handler
498 * returned RX_HANDLER_PASS.
499 */
500
501enum rx_handler_result {
502 RX_HANDLER_CONSUMED,
503 RX_HANDLER_ANOTHER,
504 RX_HANDLER_EXACT,
505 RX_HANDLER_PASS,
506};
507typedef enum rx_handler_result rx_handler_result_t;
508typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
509
510void __napi_schedule(struct napi_struct *n);
511void __napi_schedule_irqoff(struct napi_struct *n);
512
513static inline bool napi_disable_pending(struct napi_struct *n)
514{
515 return test_bit(NAPI_STATE_DISABLE, &n->state);
516}
517
518static inline bool napi_prefer_busy_poll(struct napi_struct *n)
519{
520 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
521}
522
523/**
524 * napi_is_scheduled - test if NAPI is scheduled
525 * @n: NAPI context
526 *
527 * This check is "best-effort". With no locking implemented,
528 * a NAPI can be scheduled or terminate right after this check
529 * and produce not precise results.
530 *
531 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
532 * should not be used normally and napi_schedule should be
533 * used instead.
534 *
535 * Use only if the driver really needs to check if a NAPI
536 * is scheduled for example in the context of delayed timer
537 * that can be skipped if a NAPI is already scheduled.
538 *
539 * Return: True if NAPI is scheduled, False otherwise.
540 */
541static inline bool napi_is_scheduled(struct napi_struct *n)
542{
543 return test_bit(NAPI_STATE_SCHED, &n->state);
544}
545
546bool napi_schedule_prep(struct napi_struct *n);
547
548/**
549 * napi_schedule - schedule NAPI poll
550 * @n: NAPI context
551 *
552 * Schedule NAPI poll routine to be called if it is not already
553 * running.
554 * Return: true if we schedule a NAPI or false if not.
555 * Refer to napi_schedule_prep() for additional reason on why
556 * a NAPI might not be scheduled.
557 */
558static inline bool napi_schedule(struct napi_struct *n)
559{
560 if (napi_schedule_prep(n)) {
561 __napi_schedule(n);
562 return true;
563 }
564
565 return false;
566}
567
568/**
569 * napi_schedule_irqoff - schedule NAPI poll
570 * @n: NAPI context
571 *
572 * Variant of napi_schedule(), assuming hard irqs are masked.
573 */
574static inline void napi_schedule_irqoff(struct napi_struct *n)
575{
576 if (napi_schedule_prep(n))
577 __napi_schedule_irqoff(n);
578}
579
580/**
581 * napi_complete_done - NAPI processing complete
582 * @n: NAPI context
583 * @work_done: number of packets processed
584 *
585 * Mark NAPI processing as complete. Should only be called if poll budget
586 * has not been completely consumed.
587 * Prefer over napi_complete().
588 * Return: false if device should avoid rearming interrupts.
589 */
590bool napi_complete_done(struct napi_struct *n, int work_done);
591
592static inline bool napi_complete(struct napi_struct *n)
593{
594 return napi_complete_done(n, 0);
595}
596
597void netif_threaded_enable(struct net_device *dev);
598int dev_set_threaded(struct net_device *dev,
599 enum netdev_napi_threaded threaded);
600
601void napi_disable(struct napi_struct *n);
602void napi_disable_locked(struct napi_struct *n);
603
604void napi_enable(struct napi_struct *n);
605void napi_enable_locked(struct napi_struct *n);
606
607/**
608 * napi_synchronize - wait until NAPI is not running
609 * @n: NAPI context
610 *
611 * Wait until NAPI is done being scheduled on this context.
612 * Waits till any outstanding processing completes but
613 * does not disable future activations.
614 */
615static inline void napi_synchronize(const struct napi_struct *n)
616{
617 if (IS_ENABLED(CONFIG_SMP))
618 while (test_bit(NAPI_STATE_SCHED, &n->state))
619 msleep(1);
620 else
621 barrier();
622}
623
624/**
625 * napi_if_scheduled_mark_missed - if napi is running, set the
626 * NAPIF_STATE_MISSED
627 * @n: NAPI context
628 *
629 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
630 * NAPI is scheduled.
631 **/
632static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
633{
634 unsigned long val, new;
635
636 val = READ_ONCE(n->state);
637 do {
638 if (val & NAPIF_STATE_DISABLE)
639 return true;
640
641 if (!(val & NAPIF_STATE_SCHED))
642 return false;
643
644 new = val | NAPIF_STATE_MISSED;
645 } while (!try_cmpxchg(&n->state, &val, new));
646
647 return true;
648}
649
650enum netdev_queue_state_t {
651 __QUEUE_STATE_DRV_XOFF,
652 __QUEUE_STATE_STACK_XOFF,
653 __QUEUE_STATE_FROZEN,
654};
655
656#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
657#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
658#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
659
660#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
661#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
662 QUEUE_STATE_FROZEN)
663#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
664 QUEUE_STATE_FROZEN)
665
666/*
667 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
668 * netif_tx_* functions below are used to manipulate this flag. The
669 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
670 * queue independently. The netif_xmit_*stopped functions below are called
671 * to check if the queue has been stopped by the driver or stack (either
672 * of the XOFF bits are set in the state). Drivers should not need to call
673 * netif_xmit*stopped functions, they should only be using netif_tx_*.
674 */
675
676struct netdev_queue {
677/*
678 * read-mostly part
679 */
680 struct net_device *dev;
681 netdevice_tracker dev_tracker;
682
683 struct Qdisc __rcu *qdisc;
684 struct Qdisc __rcu *qdisc_sleeping;
685#ifdef CONFIG_SYSFS
686 struct kobject kobj;
687 const struct attribute_group **groups;
688#endif
689 unsigned long tx_maxrate;
690 /*
691 * Number of TX timeouts for this queue
692 * (/sys/class/net/DEV/Q/trans_timeout)
693 */
694 atomic_long_t trans_timeout;
695
696 /* Subordinate device that the queue has been assigned to */
697 struct net_device *sb_dev;
698#ifdef CONFIG_XDP_SOCKETS
699 /* "ops protected", see comment about net_device::lock */
700 struct xsk_buff_pool *pool;
701#endif
702
703/*
704 * write-mostly part
705 */
706#ifdef CONFIG_BQL
707 struct dql dql;
708#endif
709 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
710 int xmit_lock_owner;
711 /*
712 * Time (in jiffies) of last Tx
713 */
714 unsigned long trans_start;
715
716 unsigned long state;
717
718/*
719 * slow- / control-path part
720 */
721 /* NAPI instance for the queue
722 * "ops protected", see comment about net_device::lock
723 */
724 struct napi_struct *napi;
725
726#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
727 int numa_node;
728#endif
729} ____cacheline_aligned_in_smp;
730
731extern int sysctl_fb_tunnels_only_for_init_net;
732extern int sysctl_devconf_inherit_init_net;
733
734/*
735 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
736 * == 1 : For initns only
737 * == 2 : For none.
738 */
739static inline bool net_has_fallback_tunnels(const struct net *net)
740{
741#if IS_ENABLED(CONFIG_SYSCTL)
742 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
743
744 return !fb_tunnels_only_for_init_net ||
745 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
746#else
747 return true;
748#endif
749}
750
751static inline int net_inherit_devconf(void)
752{
753#if IS_ENABLED(CONFIG_SYSCTL)
754 return READ_ONCE(sysctl_devconf_inherit_init_net);
755#else
756 return 0;
757#endif
758}
759
760static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
761{
762#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
763 return q->numa_node;
764#else
765 return NUMA_NO_NODE;
766#endif
767}
768
769static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
770{
771#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
772 q->numa_node = node;
773#endif
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
781/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
782enum xps_map_type {
783 XPS_CPUS = 0,
784 XPS_RXQS,
785 XPS_MAPS_MAX,
786};
787
788#ifdef CONFIG_XPS
789/*
790 * This structure holds an XPS map which can be of variable length. The
791 * map is an array of queues.
792 */
793struct xps_map {
794 unsigned int len;
795 unsigned int alloc_len;
796 struct rcu_head rcu;
797 u16 queues[];
798};
799#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
800#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
801 - sizeof(struct xps_map)) / sizeof(u16))
802
803/*
804 * This structure holds all XPS maps for device. Maps are indexed by CPU.
805 *
806 * We keep track of the number of cpus/rxqs used when the struct is allocated,
807 * in nr_ids. This will help not accessing out-of-bound memory.
808 *
809 * We keep track of the number of traffic classes used when the struct is
810 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
811 * not crossing its upper bound, as the original dev->num_tc can be updated in
812 * the meantime.
813 */
814struct xps_dev_maps {
815 struct rcu_head rcu;
816 unsigned int nr_ids;
817 s16 num_tc;
818 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
819};
820
821#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
822 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
823
824#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
825 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
826
827#endif /* CONFIG_XPS */
828
829#define TC_MAX_QUEUE 16
830#define TC_BITMASK 15
831/* HW offloaded queuing disciplines txq count and offset maps */
832struct netdev_tc_txq {
833 u16 count;
834 u16 offset;
835};
836
837#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
838/*
839 * This structure is to hold information about the device
840 * configured to run FCoE protocol stack.
841 */
842struct netdev_fcoe_hbainfo {
843 char manufacturer[64];
844 char serial_number[64];
845 char hardware_version[64];
846 char driver_version[64];
847 char optionrom_version[64];
848 char firmware_version[64];
849 char model[256];
850 char model_description[256];
851};
852#endif
853
854#define MAX_PHYS_ITEM_ID_LEN 32
855
856/* This structure holds a unique identifier to identify some
857 * physical item (port for example) used by a netdevice.
858 */
859struct netdev_phys_item_id {
860 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
861 unsigned char id_len;
862};
863
864static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
865 struct netdev_phys_item_id *b)
866{
867 return a->id_len == b->id_len &&
868 memcmp(a->id, b->id, a->id_len) == 0;
869}
870
871typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
872 struct sk_buff *skb,
873 struct net_device *sb_dev);
874
875enum net_device_path_type {
876 DEV_PATH_ETHERNET = 0,
877 DEV_PATH_VLAN,
878 DEV_PATH_BRIDGE,
879 DEV_PATH_PPPOE,
880 DEV_PATH_DSA,
881 DEV_PATH_MTK_WDMA,
882 DEV_PATH_TUN,
883};
884
885struct net_device_path {
886 enum net_device_path_type type;
887 const struct net_device *dev;
888 union {
889 struct {
890 u16 id;
891 __be16 proto;
892 u8 h_dest[ETH_ALEN];
893 } encap;
894 struct {
895 union {
896 struct in_addr src_v4;
897 struct in6_addr src_v6;
898 };
899 union {
900 struct in_addr dst_v4;
901 struct in6_addr dst_v6;
902 };
903
904 u8 l3_proto;
905 } tun;
906 struct {
907 enum {
908 DEV_PATH_BR_VLAN_KEEP,
909 DEV_PATH_BR_VLAN_TAG,
910 DEV_PATH_BR_VLAN_UNTAG,
911 DEV_PATH_BR_VLAN_UNTAG_HW,
912 } vlan_mode;
913 u16 vlan_id;
914 __be16 vlan_proto;
915 } bridge;
916 struct {
917 int port;
918 u16 proto;
919 } dsa;
920 struct {
921 u8 wdma_idx;
922 u8 queue;
923 u16 wcid;
924 u8 bss;
925 u8 amsdu;
926 } mtk_wdma;
927 };
928};
929
930#define NET_DEVICE_PATH_STACK_MAX 5
931#define NET_DEVICE_PATH_VLAN_MAX 2
932
933struct net_device_path_stack {
934 int num_paths;
935 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
936};
937
938struct net_device_path_ctx {
939 const struct net_device *dev;
940 u8 daddr[ETH_ALEN];
941
942 int num_vlans;
943 struct {
944 u16 id;
945 __be16 proto;
946 } vlan[NET_DEVICE_PATH_VLAN_MAX];
947};
948
949enum tc_setup_type {
950 TC_QUERY_CAPS,
951 TC_SETUP_QDISC_MQPRIO,
952 TC_SETUP_CLSU32,
953 TC_SETUP_CLSFLOWER,
954 TC_SETUP_CLSMATCHALL,
955 TC_SETUP_CLSBPF,
956 TC_SETUP_BLOCK,
957 TC_SETUP_QDISC_CBS,
958 TC_SETUP_QDISC_RED,
959 TC_SETUP_QDISC_PRIO,
960 TC_SETUP_QDISC_MQ,
961 TC_SETUP_QDISC_ETF,
962 TC_SETUP_ROOT_QDISC,
963 TC_SETUP_QDISC_GRED,
964 TC_SETUP_QDISC_TAPRIO,
965 TC_SETUP_FT,
966 TC_SETUP_QDISC_ETS,
967 TC_SETUP_QDISC_TBF,
968 TC_SETUP_QDISC_FIFO,
969 TC_SETUP_QDISC_HTB,
970 TC_SETUP_ACT,
971};
972
973/* These structures hold the attributes of bpf state that are being passed
974 * to the netdevice through the bpf op.
975 */
976enum bpf_netdev_command {
977 /* Set or clear a bpf program used in the earliest stages of packet
978 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
979 * is responsible for calling bpf_prog_put on any old progs that are
980 * stored. In case of error, the callee need not release the new prog
981 * reference, but on success it takes ownership and must bpf_prog_put
982 * when it is no longer used.
983 */
984 XDP_SETUP_PROG,
985 XDP_SETUP_PROG_HW,
986 /* BPF program for offload callbacks, invoked at program load time. */
987 BPF_OFFLOAD_MAP_ALLOC,
988 BPF_OFFLOAD_MAP_FREE,
989 XDP_SETUP_XSK_POOL,
990};
991
992struct bpf_prog_offload_ops;
993struct netlink_ext_ack;
994struct xdp_umem;
995struct xdp_dev_bulk_queue;
996struct bpf_xdp_link;
997
998enum bpf_xdp_mode {
999 XDP_MODE_SKB = 0,
1000 XDP_MODE_DRV = 1,
1001 XDP_MODE_HW = 2,
1002 __MAX_XDP_MODE
1003};
1004
1005struct bpf_xdp_entity {
1006 struct bpf_prog *prog;
1007 struct bpf_xdp_link *link;
1008};
1009
1010struct netdev_bpf {
1011 enum bpf_netdev_command command;
1012 union {
1013 /* XDP_SETUP_PROG */
1014 struct {
1015 u32 flags;
1016 struct bpf_prog *prog;
1017 struct netlink_ext_ack *extack;
1018 };
1019 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1020 struct {
1021 struct bpf_offloaded_map *offmap;
1022 };
1023 /* XDP_SETUP_XSK_POOL */
1024 struct {
1025 struct xsk_buff_pool *pool;
1026 u16 queue_id;
1027 } xsk;
1028 };
1029};
1030
1031/* Flags for ndo_xsk_wakeup. */
1032#define XDP_WAKEUP_RX (1 << 0)
1033#define XDP_WAKEUP_TX (1 << 1)
1034
1035#ifdef CONFIG_XFRM_OFFLOAD
1036struct xfrmdev_ops {
1037 int (*xdo_dev_state_add)(struct net_device *dev,
1038 struct xfrm_state *x,
1039 struct netlink_ext_ack *extack);
1040 void (*xdo_dev_state_delete)(struct net_device *dev,
1041 struct xfrm_state *x);
1042 void (*xdo_dev_state_free)(struct net_device *dev,
1043 struct xfrm_state *x);
1044 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1045 struct xfrm_state *x);
1046 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1047 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
1048 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1049 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1050 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1051};
1052#endif
1053
1054struct dev_ifalias {
1055 struct rcu_head rcuhead;
1056 char ifalias[];
1057};
1058
1059struct devlink;
1060struct tlsdev_ops;
1061
1062struct netdev_net_notifier {
1063 struct list_head list;
1064 struct notifier_block *nb;
1065};
1066
1067/*
1068 * This structure defines the management hooks for network devices.
1069 * The following hooks can be defined; unless noted otherwise, they are
1070 * optional and can be filled with a null pointer.
1071 *
1072 * int (*ndo_init)(struct net_device *dev);
1073 * This function is called once when a network device is registered.
1074 * The network device can use this for any late stage initialization
1075 * or semantic validation. It can fail with an error code which will
1076 * be propagated back to register_netdev.
1077 *
1078 * void (*ndo_uninit)(struct net_device *dev);
1079 * This function is called when device is unregistered or when registration
1080 * fails. It is not called if init fails.
1081 *
1082 * int (*ndo_open)(struct net_device *dev);
1083 * This function is called when a network device transitions to the up
1084 * state.
1085 *
1086 * int (*ndo_stop)(struct net_device *dev);
1087 * This function is called when a network device transitions to the down
1088 * state.
1089 *
1090 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1091 * struct net_device *dev);
1092 * Called when a packet needs to be transmitted.
1093 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1094 * the queue before that can happen; it's for obsolete devices and weird
1095 * corner cases, but the stack really does a non-trivial amount
1096 * of useless work if you return NETDEV_TX_BUSY.
1097 * Required; cannot be NULL.
1098 *
1099 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1100 * struct net_device *dev
1101 * netdev_features_t features);
1102 * Called by core transmit path to determine if device is capable of
1103 * performing offload operations on a given packet. This is to give
1104 * the device an opportunity to implement any restrictions that cannot
1105 * be otherwise expressed by feature flags. The check is called with
1106 * the set of features that the stack has calculated and it returns
1107 * those the driver believes to be appropriate.
1108 *
1109 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1110 * struct net_device *sb_dev);
1111 * Called to decide which queue to use when device supports multiple
1112 * transmit queues.
1113 *
1114 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1115 * This function is called to allow device receiver to make
1116 * changes to configuration when multicast or promiscuous is enabled.
1117 *
1118 * void (*ndo_set_rx_mode)(struct net_device *dev);
1119 * This function is called device changes address list filtering.
1120 * If driver handles unicast address filtering, it should set
1121 * IFF_UNICAST_FLT in its priv_flags.
1122 * Cannot sleep, called with netif_addr_lock_bh held.
1123 * Deprecated in favor of ndo_set_rx_mode_async.
1124 *
1125 * void (*ndo_set_rx_mode_async)(struct net_device *dev,
1126 * struct netdev_hw_addr_list *uc,
1127 * struct netdev_hw_addr_list *mc);
1128 * Async version of ndo_set_rx_mode which runs in process context
1129 * with rtnl_lock and netdev_lock_ops(dev) held. The uc/mc parameters
1130 * are snapshots of the address lists - iterate with
1131 * netdev_hw_addr_list_for_each(ha, uc).
1132 *
1133 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1134 * This function is called when the Media Access Control address
1135 * needs to be changed. If this interface is not defined, the
1136 * MAC address can not be changed.
1137 *
1138 * int (*ndo_validate_addr)(struct net_device *dev);
1139 * Test if Media Access Control address is valid for the device.
1140 *
1141 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1142 * Old-style ioctl entry point. This is used internally by the
1143 * ieee802154 subsystem but is no longer called by the device
1144 * ioctl handler.
1145 *
1146 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1147 * Used by the bonding driver for its device specific ioctls:
1148 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1149 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1150 *
1151 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1152 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1153 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1154 *
1155 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1156 * Used to set network devices bus interface parameters. This interface
1157 * is retained for legacy reasons; new devices should use the bus
1158 * interface (PCI) for low level management.
1159 *
1160 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1161 * Called when a user wants to change the Maximum Transfer Unit
1162 * of a device.
1163 *
1164 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1165 * Callback used when the transmitter has not made any progress
1166 * for dev->watchdog ticks.
1167 *
1168 * void (*ndo_get_stats64)(struct net_device *dev,
1169 * struct rtnl_link_stats64 *storage);
1170 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1171 * Called when a user wants to get the network device usage
1172 * statistics. Drivers must do one of the following:
1173 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1174 * rtnl_link_stats64 structure passed by the caller.
1175 * 2. Define @ndo_get_stats to update a net_device_stats structure
1176 * (which should normally be dev->stats) and return a pointer to
1177 * it. The structure may be changed asynchronously only if each
1178 * field is written atomically.
1179 * 3. Update dev->stats asynchronously and atomically, and define
1180 * neither operation.
1181 *
1182 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1183 * Return true if this device supports offload stats of this attr_id.
1184 *
1185 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1186 * void *attr_data)
1187 * Get statistics for offload operations by attr_id. Write it into the
1188 * attr_data pointer.
1189 *
1190 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1191 * If device supports VLAN filtering this function is called when a
1192 * VLAN id is registered.
1193 *
1194 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1195 * If device supports VLAN filtering this function is called when a
1196 * VLAN id is unregistered.
1197 *
1198 * void (*ndo_poll_controller)(struct net_device *dev);
1199 *
1200 * SR-IOV management functions.
1201 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1202 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1203 * u8 qos, __be16 proto);
1204 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1205 * int max_tx_rate);
1206 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1207 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1208 * int (*ndo_get_vf_config)(struct net_device *dev,
1209 * int vf, struct ifla_vf_info *ivf);
1210 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1211 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1212 * struct nlattr *port[]);
1213 *
1214 * Enable or disable the VF ability to query its RSS Redirection Table and
1215 * Hash Key. This is needed since on some devices VF share this information
1216 * with PF and querying it may introduce a theoretical security risk.
1217 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1218 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1219 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1220 * void *type_data);
1221 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1222 * This is always called from the stack with the rtnl lock held and netif
1223 * tx queues stopped. This allows the netdevice to perform queue
1224 * management safely.
1225 *
1226 * Fiber Channel over Ethernet (FCoE) offload functions.
1227 * int (*ndo_fcoe_enable)(struct net_device *dev);
1228 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1229 * so the underlying device can perform whatever needed configuration or
1230 * initialization to support acceleration of FCoE traffic.
1231 *
1232 * int (*ndo_fcoe_disable)(struct net_device *dev);
1233 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1234 * so the underlying device can perform whatever needed clean-ups to
1235 * stop supporting acceleration of FCoE traffic.
1236 *
1237 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1238 * struct scatterlist *sgl, unsigned int sgc);
1239 * Called when the FCoE Initiator wants to initialize an I/O that
1240 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1241 * perform necessary setup and returns 1 to indicate the device is set up
1242 * successfully to perform DDP on this I/O, otherwise this returns 0.
1243 *
1244 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1245 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1246 * indicated by the FC exchange id 'xid', so the underlying device can
1247 * clean up and reuse resources for later DDP requests.
1248 *
1249 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1250 * struct scatterlist *sgl, unsigned int sgc);
1251 * Called when the FCoE Target wants to initialize an I/O that
1252 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1253 * perform necessary setup and returns 1 to indicate the device is set up
1254 * successfully to perform DDP on this I/O, otherwise this returns 0.
1255 *
1256 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1257 * struct netdev_fcoe_hbainfo *hbainfo);
1258 * Called when the FCoE Protocol stack wants information on the underlying
1259 * device. This information is utilized by the FCoE protocol stack to
1260 * register attributes with Fiber Channel management service as per the
1261 * FC-GS Fabric Device Management Information(FDMI) specification.
1262 *
1263 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1264 * Called when the underlying device wants to override default World Wide
1265 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1266 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1267 * protocol stack to use.
1268 *
1269 * RFS acceleration.
1270 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1271 * u16 rxq_index, u32 flow_id);
1272 * Set hardware filter for RFS. rxq_index is the target queue index;
1273 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1274 * Return the filter ID on success, or a negative error code.
1275 *
1276 * Slave management functions (for bridge, bonding, etc).
1277 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1278 * Called to make another netdev an underling.
1279 *
1280 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1281 * Called to release previously enslaved netdev.
1282 *
1283 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1284 * struct sk_buff *skb,
1285 * bool all_slaves);
1286 * Get the xmit slave of master device. If all_slaves is true, function
1287 * assume all the slaves can transmit.
1288 *
1289 * Feature/offload setting functions.
1290 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1291 * netdev_features_t features);
1292 * Adjusts the requested feature flags according to device-specific
1293 * constraints, and returns the resulting flags. Must not modify
1294 * the device state.
1295 *
1296 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1297 * Called to update device configuration to new features. Passed
1298 * feature set might be less than what was returned by ndo_fix_features()).
1299 * Must return >0 or -errno if it changed dev->features itself.
1300 *
1301 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1302 * struct net_device *dev,
1303 * const unsigned char *addr, u16 vid, u16 flags,
1304 * bool *notified, struct netlink_ext_ack *extack);
1305 * Adds an FDB entry to dev for addr.
1306 * Callee shall set *notified to true if it sent any appropriate
1307 * notification(s). Otherwise core will send a generic one.
1308 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1309 * struct net_device *dev,
1310 * const unsigned char *addr, u16 vid
1311 * bool *notified, struct netlink_ext_ack *extack);
1312 * Deletes the FDB entry from dev corresponding to addr.
1313 * Callee shall set *notified to true if it sent any appropriate
1314 * notification(s). Otherwise core will send a generic one.
1315 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1316 * struct netlink_ext_ack *extack);
1317 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1318 * struct net_device *dev, struct net_device *filter_dev,
1319 * int *idx)
1320 * Used to add FDB entries to dump requests. Implementers should add
1321 * entries to skb and update idx with the number of entries.
1322 *
1323 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1324 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1325 * Adds an MDB entry to dev.
1326 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1327 * struct netlink_ext_ack *extack);
1328 * Deletes the MDB entry from dev.
1329 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1330 * struct netlink_ext_ack *extack);
1331 * Bulk deletes MDB entries from dev.
1332 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1333 * struct netlink_callback *cb);
1334 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1335 * callback is used by core rtnetlink code.
1336 *
1337 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1338 * u16 flags, struct netlink_ext_ack *extack)
1339 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1340 * struct net_device *dev, u32 filter_mask,
1341 * int nlflags)
1342 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1343 * u16 flags);
1344 *
1345 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1346 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1347 * which do not represent real hardware may define this to allow their
1348 * userspace components to manage their virtual carrier state. Devices
1349 * that determine carrier state from physical hardware properties (eg
1350 * network cables) or protocol-dependent mechanisms (eg
1351 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1352 *
1353 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1354 * struct netdev_phys_item_id *ppid);
1355 * Called to get ID of physical port of this device. If driver does
1356 * not implement this, it is assumed that the hw is not able to have
1357 * multiple net devices on single physical port.
1358 *
1359 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1360 * struct netdev_phys_item_id *ppid)
1361 * Called to get the parent ID of the physical port of this device.
1362 *
1363 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1364 * struct net_device *dev)
1365 * Called by upper layer devices to accelerate switching or other
1366 * station functionality into hardware. 'pdev is the lowerdev
1367 * to use for the offload and 'dev' is the net device that will
1368 * back the offload. Returns a pointer to the private structure
1369 * the upper layer will maintain.
1370 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1371 * Called by upper layer device to delete the station created
1372 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1373 * the station and priv is the structure returned by the add
1374 * operation.
1375 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1376 * int queue_index, u32 maxrate);
1377 * Called when a user wants to set a max-rate limitation of specific
1378 * TX queue.
1379 * int (*ndo_get_iflink)(const struct net_device *dev);
1380 * Called to get the iflink value of this device.
1381 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1382 * This function is used to get egress tunnel information for given skb.
1383 * This is useful for retrieving outer tunnel header parameters while
1384 * sampling packet.
1385 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1386 * This function is used to specify the headroom that the skb must
1387 * consider when allocation skb during packet reception. Setting
1388 * appropriate rx headroom value allows avoiding skb head copy on
1389 * forward. Setting a negative value resets the rx headroom to the
1390 * default value.
1391 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1392 * This function is used to set or query state related to XDP on the
1393 * netdevice and manage BPF offload. See definition of
1394 * enum bpf_netdev_command for details.
1395 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1396 * u32 flags);
1397 * This function is used to submit @n XDP packets for transmit on a
1398 * netdevice. Returns number of frames successfully transmitted, frames
1399 * that got dropped are freed/returned via xdp_return_frame().
1400 * Returns negative number, means general error invoking ndo, meaning
1401 * no frames were xmit'ed and core-caller will free all frames.
1402 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1403 * struct xdp_buff *xdp);
1404 * Get the xmit slave of master device based on the xdp_buff.
1405 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1406 * This function is used to wake up the softirq, ksoftirqd or kthread
1407 * responsible for sending and/or receiving packets on a specific
1408 * queue id bound to an AF_XDP socket. The flags field specifies if
1409 * only RX, only Tx, or both should be woken up using the flags
1410 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1411 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1412 * int cmd);
1413 * Add, change, delete or get information on an IPv4 tunnel.
1414 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1415 * If a device is paired with a peer device, return the peer instance.
1416 * The caller must be under RCU read context.
1417 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1418 * Get the forwarding path to reach the real device from the HW destination address
1419 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1420 * const struct skb_shared_hwtstamps *hwtstamps,
1421 * bool cycles);
1422 * Get hardware timestamp based on normal/adjustable time or free running
1423 * cycle counter. This function is required if physical clock supports a
1424 * free running cycle counter.
1425 *
1426 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1427 * struct kernel_hwtstamp_config *kernel_config);
1428 * Get the currently configured hardware timestamping parameters for the
1429 * NIC device.
1430 *
1431 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1432 * struct kernel_hwtstamp_config *kernel_config,
1433 * struct netlink_ext_ack *extack);
1434 * Change the hardware timestamping parameters for NIC device.
1435 */
1436struct net_device_ops {
1437 int (*ndo_init)(struct net_device *dev);
1438 void (*ndo_uninit)(struct net_device *dev);
1439 int (*ndo_open)(struct net_device *dev);
1440 int (*ndo_stop)(struct net_device *dev);
1441 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1442 struct net_device *dev);
1443 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1444 struct net_device *dev,
1445 netdev_features_t features);
1446 u16 (*ndo_select_queue)(struct net_device *dev,
1447 struct sk_buff *skb,
1448 struct net_device *sb_dev);
1449 void (*ndo_change_rx_flags)(struct net_device *dev,
1450 int flags);
1451 void (*ndo_set_rx_mode)(struct net_device *dev);
1452 void (*ndo_set_rx_mode_async)(
1453 struct net_device *dev,
1454 struct netdev_hw_addr_list *uc,
1455 struct netdev_hw_addr_list *mc);
1456 int (*ndo_set_mac_address)(struct net_device *dev,
1457 void *addr);
1458 int (*ndo_validate_addr)(struct net_device *dev);
1459 int (*ndo_do_ioctl)(struct net_device *dev,
1460 struct ifreq *ifr, int cmd);
1461 int (*ndo_eth_ioctl)(struct net_device *dev,
1462 struct ifreq *ifr, int cmd);
1463 int (*ndo_siocbond)(struct net_device *dev,
1464 struct ifreq *ifr, int cmd);
1465 int (*ndo_siocwandev)(struct net_device *dev,
1466 struct if_settings *ifs);
1467 int (*ndo_siocdevprivate)(struct net_device *dev,
1468 struct ifreq *ifr,
1469 void __user *data, int cmd);
1470 int (*ndo_set_config)(struct net_device *dev,
1471 struct ifmap *map);
1472 int (*ndo_change_mtu)(struct net_device *dev,
1473 int new_mtu);
1474 int (*ndo_neigh_setup)(struct net_device *dev,
1475 struct neigh_parms *);
1476 void (*ndo_tx_timeout) (struct net_device *dev,
1477 unsigned int txqueue);
1478
1479 void (*ndo_get_stats64)(struct net_device *dev,
1480 struct rtnl_link_stats64 *storage);
1481 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1482 int (*ndo_get_offload_stats)(int attr_id,
1483 const struct net_device *dev,
1484 void *attr_data);
1485 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1486
1487 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1488 __be16 proto, u16 vid);
1489 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1490 __be16 proto, u16 vid);
1491#ifdef CONFIG_NET_POLL_CONTROLLER
1492 void (*ndo_poll_controller)(struct net_device *dev);
1493 int (*ndo_netpoll_setup)(struct net_device *dev);
1494 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1495#endif
1496 int (*ndo_set_vf_mac)(struct net_device *dev,
1497 int queue, u8 *mac);
1498 int (*ndo_set_vf_vlan)(struct net_device *dev,
1499 int queue, u16 vlan,
1500 u8 qos, __be16 proto);
1501 int (*ndo_set_vf_rate)(struct net_device *dev,
1502 int vf, int min_tx_rate,
1503 int max_tx_rate);
1504 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1505 int vf, bool setting);
1506 int (*ndo_set_vf_trust)(struct net_device *dev,
1507 int vf, bool setting);
1508 int (*ndo_get_vf_config)(struct net_device *dev,
1509 int vf,
1510 struct ifla_vf_info *ivf);
1511 int (*ndo_set_vf_link_state)(struct net_device *dev,
1512 int vf, int link_state);
1513 int (*ndo_get_vf_stats)(struct net_device *dev,
1514 int vf,
1515 struct ifla_vf_stats
1516 *vf_stats);
1517 int (*ndo_set_vf_port)(struct net_device *dev,
1518 int vf,
1519 struct nlattr *port[]);
1520 int (*ndo_get_vf_port)(struct net_device *dev,
1521 int vf, struct sk_buff *skb);
1522 int (*ndo_get_vf_guid)(struct net_device *dev,
1523 int vf,
1524 struct ifla_vf_guid *node_guid,
1525 struct ifla_vf_guid *port_guid);
1526 int (*ndo_set_vf_guid)(struct net_device *dev,
1527 int vf, u64 guid,
1528 int guid_type);
1529 int (*ndo_set_vf_rss_query_en)(
1530 struct net_device *dev,
1531 int vf, bool setting);
1532 int (*ndo_setup_tc)(struct net_device *dev,
1533 enum tc_setup_type type,
1534 void *type_data);
1535#if IS_ENABLED(CONFIG_FCOE)
1536 int (*ndo_fcoe_enable)(struct net_device *dev);
1537 int (*ndo_fcoe_disable)(struct net_device *dev);
1538 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1539 u16 xid,
1540 struct scatterlist *sgl,
1541 unsigned int sgc);
1542 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1543 u16 xid);
1544 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1545 u16 xid,
1546 struct scatterlist *sgl,
1547 unsigned int sgc);
1548 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1549 struct netdev_fcoe_hbainfo *hbainfo);
1550#endif
1551
1552#if IS_ENABLED(CONFIG_LIBFCOE)
1553#define NETDEV_FCOE_WWNN 0
1554#define NETDEV_FCOE_WWPN 1
1555 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1556 u64 *wwn, int type);
1557#endif
1558
1559#ifdef CONFIG_RFS_ACCEL
1560 int (*ndo_rx_flow_steer)(struct net_device *dev,
1561 const struct sk_buff *skb,
1562 u16 rxq_index,
1563 u32 flow_id);
1564#endif
1565 int (*ndo_add_slave)(struct net_device *dev,
1566 struct net_device *slave_dev,
1567 struct netlink_ext_ack *extack);
1568 int (*ndo_del_slave)(struct net_device *dev,
1569 struct net_device *slave_dev);
1570 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1571 struct sk_buff *skb,
1572 bool all_slaves);
1573 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1574 struct sock *sk);
1575 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1576 netdev_features_t features);
1577 int (*ndo_set_features)(struct net_device *dev,
1578 netdev_features_t features);
1579 int (*ndo_neigh_construct)(struct net_device *dev,
1580 struct neighbour *n);
1581 void (*ndo_neigh_destroy)(struct net_device *dev,
1582 struct neighbour *n);
1583
1584 int (*ndo_fdb_add)(struct ndmsg *ndm,
1585 struct nlattr *tb[],
1586 struct net_device *dev,
1587 const unsigned char *addr,
1588 u16 vid,
1589 u16 flags,
1590 bool *notified,
1591 struct netlink_ext_ack *extack);
1592 int (*ndo_fdb_del)(struct ndmsg *ndm,
1593 struct nlattr *tb[],
1594 struct net_device *dev,
1595 const unsigned char *addr,
1596 u16 vid,
1597 bool *notified,
1598 struct netlink_ext_ack *extack);
1599 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1600 struct net_device *dev,
1601 struct netlink_ext_ack *extack);
1602 int (*ndo_fdb_dump)(struct sk_buff *skb,
1603 struct netlink_callback *cb,
1604 struct net_device *dev,
1605 struct net_device *filter_dev,
1606 int *idx);
1607 int (*ndo_fdb_get)(struct sk_buff *skb,
1608 struct nlattr *tb[],
1609 struct net_device *dev,
1610 const unsigned char *addr,
1611 u16 vid, u32 portid, u32 seq,
1612 struct netlink_ext_ack *extack);
1613 int (*ndo_mdb_add)(struct net_device *dev,
1614 struct nlattr *tb[],
1615 u16 nlmsg_flags,
1616 struct netlink_ext_ack *extack);
1617 int (*ndo_mdb_del)(struct net_device *dev,
1618 struct nlattr *tb[],
1619 struct netlink_ext_ack *extack);
1620 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1621 struct nlattr *tb[],
1622 struct netlink_ext_ack *extack);
1623 int (*ndo_mdb_dump)(struct net_device *dev,
1624 struct sk_buff *skb,
1625 struct netlink_callback *cb);
1626 int (*ndo_mdb_get)(struct net_device *dev,
1627 struct nlattr *tb[], u32 portid,
1628 u32 seq,
1629 struct netlink_ext_ack *extack);
1630 int (*ndo_bridge_setlink)(struct net_device *dev,
1631 struct nlmsghdr *nlh,
1632 u16 flags,
1633 struct netlink_ext_ack *extack);
1634 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1635 u32 pid, u32 seq,
1636 struct net_device *dev,
1637 u32 filter_mask,
1638 int nlflags);
1639 int (*ndo_bridge_dellink)(struct net_device *dev,
1640 struct nlmsghdr *nlh,
1641 u16 flags);
1642 int (*ndo_change_carrier)(struct net_device *dev,
1643 bool new_carrier);
1644 int (*ndo_get_phys_port_id)(struct net_device *dev,
1645 struct netdev_phys_item_id *ppid);
1646 int (*ndo_get_port_parent_id)(struct net_device *dev,
1647 struct netdev_phys_item_id *ppid);
1648 int (*ndo_get_phys_port_name)(struct net_device *dev,
1649 char *name, size_t len);
1650 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1651 struct net_device *dev);
1652 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1653 void *priv);
1654
1655 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1656 int queue_index,
1657 u32 maxrate);
1658 int (*ndo_get_iflink)(const struct net_device *dev);
1659 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1660 struct sk_buff *skb);
1661 void (*ndo_set_rx_headroom)(struct net_device *dev,
1662 int needed_headroom);
1663 int (*ndo_bpf)(struct net_device *dev,
1664 struct netdev_bpf *bpf);
1665 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1666 struct xdp_frame **xdp,
1667 u32 flags);
1668 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1669 struct xdp_buff *xdp);
1670 int (*ndo_xsk_wakeup)(struct net_device *dev,
1671 u32 queue_id, u32 flags);
1672 int (*ndo_tunnel_ctl)(struct net_device *dev,
1673 struct ip_tunnel_parm_kern *p,
1674 int cmd);
1675 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1676 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1677 struct net_device_path *path);
1678 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1679 const struct skb_shared_hwtstamps *hwtstamps,
1680 bool cycles);
1681 int (*ndo_hwtstamp_get)(struct net_device *dev,
1682 struct kernel_hwtstamp_config *kernel_config);
1683 int (*ndo_hwtstamp_set)(struct net_device *dev,
1684 struct kernel_hwtstamp_config *kernel_config,
1685 struct netlink_ext_ack *extack);
1686
1687#if IS_ENABLED(CONFIG_NET_SHAPER)
1688 /**
1689 * @net_shaper_ops: Device shaping offload operations
1690 * see include/net/net_shapers.h
1691 */
1692 const struct net_shaper_ops *net_shaper_ops;
1693#endif
1694};
1695
1696/**
1697 * enum netdev_priv_flags - &struct net_device priv_flags
1698 *
1699 * These are the &struct net_device, they are only set internally
1700 * by drivers and used in the kernel. These flags are invisible to
1701 * userspace; this means that the order of these flags can change
1702 * during any kernel release.
1703 *
1704 * You should add bitfield booleans after either net_device::priv_flags
1705 * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1706 *
1707 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1708 * @IFF_EBRIDGE: Ethernet bridging device
1709 * @IFF_BONDING: bonding master or slave
1710 * @IFF_ISATAP: ISATAP interface (RFC4214)
1711 * @IFF_WAN_HDLC: WAN HDLC device
1712 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1713 * release skb->dst
1714 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1715 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1716 * @IFF_MACVLAN_PORT: device used as macvlan port
1717 * @IFF_BRIDGE_PORT: device used as bridge port
1718 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1719 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1720 * @IFF_UNICAST_FLT: Supports unicast filtering
1721 * @IFF_TEAM_PORT: device used as team port
1722 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1723 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1724 * change when it's running
1725 * @IFF_MACVLAN: Macvlan device
1726 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1727 * underlying stacked devices
1728 * @IFF_L3MDEV_MASTER: device is an L3 master device
1729 * @IFF_NO_QUEUE: device can run without qdisc attached
1730 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1731 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1732 * @IFF_TEAM: device is a team device
1733 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1734 * entity (i.e. the master device for bridged veth)
1735 * @IFF_MACSEC: device is a MACsec device
1736 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1737 * @IFF_FAILOVER: device is a failover master device
1738 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1739 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1740 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1741 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1742 * skb_headlen(skb) == 0 (data starts from frag0)
1743 */
1744enum netdev_priv_flags {
1745 IFF_802_1Q_VLAN = 1<<0,
1746 IFF_EBRIDGE = 1<<1,
1747 IFF_BONDING = 1<<2,
1748 IFF_ISATAP = 1<<3,
1749 IFF_WAN_HDLC = 1<<4,
1750 IFF_XMIT_DST_RELEASE = 1<<5,
1751 IFF_DONT_BRIDGE = 1<<6,
1752 IFF_DISABLE_NETPOLL = 1<<7,
1753 IFF_MACVLAN_PORT = 1<<8,
1754 IFF_BRIDGE_PORT = 1<<9,
1755 IFF_OVS_DATAPATH = 1<<10,
1756 IFF_TX_SKB_SHARING = 1<<11,
1757 IFF_UNICAST_FLT = 1<<12,
1758 IFF_TEAM_PORT = 1<<13,
1759 IFF_SUPP_NOFCS = 1<<14,
1760 IFF_LIVE_ADDR_CHANGE = 1<<15,
1761 IFF_MACVLAN = 1<<16,
1762 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1763 IFF_L3MDEV_MASTER = 1<<18,
1764 IFF_NO_QUEUE = 1<<19,
1765 IFF_OPENVSWITCH = 1<<20,
1766 IFF_L3MDEV_SLAVE = 1<<21,
1767 IFF_TEAM = 1<<22,
1768 IFF_PHONY_HEADROOM = 1<<24,
1769 IFF_MACSEC = 1<<25,
1770 IFF_NO_RX_HANDLER = 1<<26,
1771 IFF_FAILOVER = 1<<27,
1772 IFF_FAILOVER_SLAVE = 1<<28,
1773 IFF_L3MDEV_RX_HANDLER = 1<<29,
1774 IFF_NO_ADDRCONF = BIT_ULL(30),
1775 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1776};
1777
1778/* Specifies the type of the struct net_device::ml_priv pointer */
1779enum netdev_ml_priv_type {
1780 ML_PRIV_NONE,
1781 ML_PRIV_CAN,
1782};
1783
1784enum netdev_stat_type {
1785 NETDEV_PCPU_STAT_NONE,
1786 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1787 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1788 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1789};
1790
1791enum netdev_reg_state {
1792 NETREG_UNINITIALIZED = 0,
1793 NETREG_REGISTERED, /* completed register_netdevice */
1794 NETREG_UNREGISTERING, /* called unregister_netdevice */
1795 NETREG_UNREGISTERED, /* completed unregister todo */
1796 NETREG_RELEASED, /* called free_netdev */
1797 NETREG_DUMMY, /* dummy device for NAPI poll */
1798};
1799
1800/**
1801 * struct net_device - The DEVICE structure.
1802 *
1803 * Actually, this whole structure is a big mistake. It mixes I/O
1804 * data with strictly "high-level" data, and it has to know about
1805 * almost every data structure used in the INET module.
1806 *
1807 * @priv_flags: flags invisible to userspace defined as bits, see
1808 * enum netdev_priv_flags for the definitions
1809 * @lltx: device supports lockless Tx. Deprecated for real HW
1810 * drivers. Mainly used by logical interfaces, such as
1811 * bonding and tunnels
1812 * @netmem_tx: device support netmem_tx.
1813 *
1814 * @name: This is the first field of the "visible" part of this structure
1815 * (i.e. as seen by users in the "Space.c" file). It is the name
1816 * of the interface.
1817 *
1818 * @name_node: Name hashlist node
1819 * @ifalias: SNMP alias
1820 * @mem_end: Shared memory end
1821 * @mem_start: Shared memory start
1822 * @base_addr: Device I/O address
1823 * @irq: Device IRQ number
1824 *
1825 * @state: Generic network queuing layer state, see netdev_state_t
1826 * @dev_list: The global list of network devices
1827 * @napi_list: List entry used for polling NAPI devices
1828 * @unreg_list: List entry when we are unregistering the
1829 * device; see the function unregister_netdev
1830 * @close_list: List entry used when we are closing the device
1831 * @ptype_all: Device-specific packet handlers for all protocols
1832 * @ptype_specific: Device-specific, protocol-specific packet handlers
1833 *
1834 * @adj_list: Directly linked devices, like slaves for bonding
1835 * @features: Currently active device features
1836 * @hw_features: User-changeable features
1837 *
1838 * @wanted_features: User-requested features
1839 * @vlan_features: Mask of features inheritable by VLAN devices
1840 *
1841 * @hw_enc_features: Mask of features inherited by encapsulating devices
1842 * This field indicates what encapsulation
1843 * offloads the hardware is capable of doing,
1844 * and drivers will need to set them appropriately.
1845 *
1846 * @mpls_features: Mask of features inheritable by MPLS
1847 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1848 * @mangleid_features: Mask of features requiring MANGLEID, will be
1849 * disabled together with the latter.
1850 *
1851 * @ifindex: interface index
1852 * @group: The group the device belongs to
1853 *
1854 * @stats: Statistics struct, which was left as a legacy, use
1855 * rtnl_link_stats64 instead
1856 *
1857 * @core_stats: core networking counters,
1858 * do not use this in drivers
1859 * @carrier_up_count: Number of times the carrier has been up
1860 * @carrier_down_count: Number of times the carrier has been down
1861 *
1862 * @wireless_handlers: List of functions to handle Wireless Extensions,
1863 * instead of ioctl,
1864 * see <net/iw_handler.h> for details.
1865 *
1866 * @netdev_ops: Includes several pointers to callbacks,
1867 * if one wants to override the ndo_*() functions
1868 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1869 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1870 * @ethtool_ops: Management operations
1871 * @l3mdev_ops: Layer 3 master device operations
1872 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1873 * discovery handling. Necessary for e.g. 6LoWPAN.
1874 * @xfrmdev_ops: Transformation offload operations
1875 * @tlsdev_ops: Transport Layer Security offload operations
1876 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1877 * of Layer 2 headers.
1878 *
1879 * @flags: Interface flags (a la BSD)
1880 * @xdp_features: XDP capability supported by the device
1881 * @gflags: Global flags ( kept as legacy )
1882 * @priv_len: Size of the ->priv flexible array
1883 * @priv: Flexible array containing private data
1884 * @operstate: RFC2863 operstate
1885 * @link_mode: Mapping policy to operstate
1886 * @if_port: Selectable AUI, TP, ...
1887 * @dma: DMA channel
1888 * @mtu: Interface MTU value
1889 * @min_mtu: Interface Minimum MTU value
1890 * @max_mtu: Interface Maximum MTU value
1891 * @type: Interface hardware type
1892 * @hard_header_len: Maximum hardware header length.
1893 * @min_header_len: Minimum hardware header length
1894 *
1895 * @needed_headroom: Extra headroom the hardware may need, but not in all
1896 * cases can this be guaranteed
1897 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1898 * cases can this be guaranteed. Some cases also use
1899 * LL_MAX_HEADER instead to allocate the skb
1900 *
1901 * interface address info:
1902 *
1903 * @perm_addr: Permanent hw address
1904 * @addr_assign_type: Hw address assignment type
1905 * @addr_len: Hardware address length
1906 * @upper_level: Maximum depth level of upper devices.
1907 * @lower_level: Maximum depth level of lower devices.
1908 * @threaded: napi threaded state.
1909 * @neigh_priv_len: Used in neigh_alloc()
1910 * @dev_id: Used to differentiate devices that share
1911 * the same link layer address
1912 * @dev_port: Used to differentiate devices that share
1913 * the same function
1914 * @addr_list_lock: XXX: need comments on this one
1915 * @name_assign_type: network interface name assignment type
1916 * @uc_promisc: Counter that indicates promiscuous mode
1917 * has been enabled due to the need to listen to
1918 * additional unicast addresses in a device that
1919 * does not implement ndo_set_rx_mode()
1920 * @rx_mode_node: List entry for rx_mode work processing
1921 * @rx_mode_tracker: Refcount tracker for rx_mode work
1922 * @rx_mode_addr_cache: Recycled snapshot entries for rx_mode work
1923 * @uc: unicast mac addresses
1924 * @mc: multicast mac addresses
1925 * @dev_addrs: list of device hw addresses
1926 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1927 * @promiscuity: Number of times the NIC is told to work in
1928 * promiscuous mode; if it becomes 0 the NIC will
1929 * exit promiscuous mode
1930 * @allmulti: Counter, enables or disables allmulticast mode
1931 *
1932 * @vlan_info: VLAN info
1933 * @dsa_ptr: dsa specific data
1934 * @tipc_ptr: TIPC specific data
1935 * @atalk_ptr: AppleTalk link
1936 * @ip_ptr: IPv4 specific data
1937 * @ip6_ptr: IPv6 specific data
1938 * @ax25_ptr: AX.25 specific data
1939 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1940 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1941 * device struct
1942 * @mpls_ptr: mpls_dev struct pointer
1943 * @mctp_ptr: MCTP specific data
1944 * @psp_dev: PSP crypto device registered for this netdev
1945 *
1946 * @dev_addr: Hw address (before bcast,
1947 * because most packets are unicast)
1948 *
1949 * @_rx: Array of RX queues
1950 * @num_rx_queues: Number of RX queues
1951 * allocated at register_netdev() time
1952 * @real_num_rx_queues: Number of RX queues currently active in device
1953 * @xdp_prog: XDP sockets filter program pointer
1954 *
1955 * @rx_handler: handler for received packets
1956 * @rx_handler_data: XXX: need comments on this one
1957 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1958 * @ingress_queue: XXX: need comments on this one
1959 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1960 * @broadcast: hw bcast address
1961 *
1962 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1963 * indexed by RX queue number. Assigned by driver.
1964 * This must only be set if the ndo_rx_flow_steer
1965 * operation is defined
1966 * @index_hlist: Device index hash chain
1967 *
1968 * @_tx: Array of TX queues
1969 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1970 * @real_num_tx_queues: Number of TX queues currently active in device
1971 * @qdisc: Root qdisc from userspace point of view
1972 * @tx_queue_len: Max frames per queue allowed
1973 * @tx_global_lock: XXX: need comments on this one
1974 * @xdp_bulkq: XDP device bulk queue
1975 * @xps_maps: all CPUs/RXQs maps for XPS device
1976 *
1977 * @xps_maps: XXX: need comments on this one
1978 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1979 * @nf_hooks_egress: netfilter hooks executed for egress packets
1980 * @qdisc_hash: qdisc hash table
1981 * @watchdog_timeo: Represents the timeout that is used by
1982 * the watchdog (see dev_watchdog())
1983 * @watchdog_timer: List of timers
1984 *
1985 * @proto_down_reason: reason a netdev interface is held down
1986 * @pcpu_refcnt: Number of references to this device
1987 * @dev_refcnt: Number of references to this device
1988 * @refcnt_tracker: Tracker directory for tracked references to this device
1989 * @todo_list: Delayed register/unregister
1990 * @link_watch_list: XXX: need comments on this one
1991 *
1992 * @reg_state: Register/unregister state machine
1993 * @dismantle: Device is going to be freed
1994 * @needs_free_netdev: Should unregister perform free_netdev?
1995 * @priv_destructor: Called from unregister
1996 * @npinfo: XXX: need comments on this one
1997 * @nd_net: Network namespace this network device is inside
1998 * protected by @lock
1999 *
2000 * @ml_priv: Mid-layer private
2001 * @ml_priv_type: Mid-layer private type
2002 *
2003 * @pcpu_stat_type: Type of device statistics which the core should
2004 * allocate/free: none, lstats, tstats, dstats. none
2005 * means the driver is handling statistics allocation/
2006 * freeing internally.
2007 * @lstats: Loopback statistics: packets, bytes
2008 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
2009 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
2010 *
2011 * @garp_port: GARP
2012 * @mrp_port: MRP
2013 *
2014 * @dm_private: Drop monitor private
2015 *
2016 * @dev: Class/net/name entry
2017 * @sysfs_groups: Space for optional device, statistics and wireless
2018 * sysfs groups
2019 *
2020 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2021 * @rtnl_link_ops: Rtnl_link_ops
2022 * @stat_ops: Optional ops for queue-aware statistics
2023 * @queue_mgmt_ops: Optional ops for queue management
2024 *
2025 * @gso_max_size: Maximum size of generic segmentation offload
2026 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2027 * @gso_max_segs: Maximum number of segments that can be passed to the
2028 * NIC for GSO
2029 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2030 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2031 * for IPv4.
2032 *
2033 * @dcbnl_ops: Data Center Bridging netlink ops
2034 * @num_tc: Number of traffic classes in the net device
2035 * @tc_to_txq: XXX: need comments on this one
2036 * @prio_tc_map: XXX: need comments on this one
2037 *
2038 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2039 *
2040 * @priomap: XXX: need comments on this one
2041 * @link_topo: Physical link topology tracking attached PHYs
2042 * @phydev: Physical device may attach itself
2043 * for hardware timestamping
2044 * @sfp_bus: attached &struct sfp_bus structure.
2045 *
2046 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2047 *
2048 * @proto_down: protocol port state information can be sent to the
2049 * switch driver and used to set the phys state of the
2050 * switch port.
2051 *
2052 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ
2053 * affinity. Set by netif_enable_irq_affinity(), then
2054 * the driver must create a persistent napi by
2055 * netif_napi_add_config() and finally bind the napi to
2056 * IRQ (via netif_napi_set_irq()).
2057 *
2058 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap.
2059 * Set by calling netif_enable_cpu_rmap().
2060 *
2061 * @see_all_hwtstamp_requests: device wants to see calls to
2062 * ndo_hwtstamp_set() for all timestamp requests
2063 * regardless of source, even if those aren't
2064 * HWTSTAMP_SOURCE_NETDEV
2065 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN
2066 * @netns_immutable: interface can't change network namespaces
2067 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes
2068 *
2069 * @net_notifier_list: List of per-net netdev notifier block
2070 * that follow this device when it is moved
2071 * to another network namespace.
2072 *
2073 * @macsec_ops: MACsec offloading ops
2074 *
2075 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2076 * offload capabilities of the device
2077 * @udp_tunnel_nic: UDP tunnel offload state
2078 * @ethtool: ethtool related state
2079 * @xdp_state: stores info on attached XDP BPF programs
2080 *
2081 * @nested_level: Used as a parameter of spin_lock_nested() of
2082 * dev->addr_list_lock.
2083 * @unlink_list: As netif_addr_lock() can be called recursively,
2084 * keep a list of interfaces to be deleted.
2085 * @gro_max_size: Maximum size of aggregated packet in generic
2086 * receive offload (GRO)
2087 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2088 * receive offload (GRO), for IPv4.
2089 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2090 * zero copy driver
2091 *
2092 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2093 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2094 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2095 * @dev_registered_tracker: tracker for reference held while
2096 * registered
2097 * @offload_xstats_l3: L3 HW stats for this netdevice.
2098 *
2099 * @devlink_port: Pointer to related devlink port structure.
2100 * Assigned by a driver before netdev registration using
2101 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2102 * during the time netdevice is registered.
2103 *
2104 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2105 * where the clock is recovered.
2106 *
2107 * @max_pacing_offload_horizon: max EDT offload horizon in nsec.
2108 * @napi_config: An array of napi_config structures containing per-NAPI
2109 * settings.
2110 * @num_napi_configs: number of allocated NAPI config structs,
2111 * always >= max(num_rx_queues, num_tx_queues).
2112 * @gro_flush_timeout: timeout for GRO layer in NAPI
2113 * @napi_defer_hard_irqs: If not zero, provides a counter that would
2114 * allow to avoid NIC hard IRQ, on busy queues.
2115 *
2116 * @neighbours: List heads pointing to this device's neighbours'
2117 * dev_list, one per address-family.
2118 * @hwprov: Tracks which PTP performs hardware packet time stamping.
2119 *
2120 * FIXME: cleanup struct net_device such that network protocol info
2121 * moves out.
2122 */
2123
2124struct net_device {
2125 /* Cacheline organization can be found documented in
2126 * Documentation/networking/net_cachelines/net_device.rst.
2127 * Please update the document when adding new fields.
2128 */
2129
2130 /* TX read-mostly hotpath */
2131 __cacheline_group_begin(net_device_read_tx);
2132 struct_group(priv_flags_fast,
2133 unsigned long priv_flags:32;
2134 unsigned long lltx:1;
2135 unsigned long netmem_tx:1;
2136 );
2137 const struct net_device_ops *netdev_ops;
2138 const struct header_ops *header_ops;
2139 struct netdev_queue *_tx;
2140 netdev_features_t gso_partial_features;
2141 unsigned int real_num_tx_queues;
2142 unsigned int gso_max_size;
2143 unsigned int gso_ipv4_max_size;
2144 u16 gso_max_segs;
2145 s16 num_tc;
2146 /* Note : dev->mtu is often read without holding a lock.
2147 * Writers usually hold RTNL.
2148 * It is recommended to use READ_ONCE() to annotate the reads,
2149 * and to use WRITE_ONCE() to annotate the writes.
2150 */
2151 unsigned int mtu;
2152 unsigned short needed_headroom;
2153 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2154#ifdef CONFIG_XPS
2155 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2156#endif
2157#ifdef CONFIG_NETFILTER_EGRESS
2158 struct nf_hook_entries __rcu *nf_hooks_egress;
2159#endif
2160#ifdef CONFIG_NET_XGRESS
2161 struct bpf_mprog_entry __rcu *tcx_egress;
2162#endif
2163 __cacheline_group_end(net_device_read_tx);
2164
2165 /* TXRX read-mostly hotpath */
2166 __cacheline_group_begin(net_device_read_txrx);
2167 union {
2168 struct pcpu_lstats __percpu *lstats;
2169 struct pcpu_sw_netstats __percpu *tstats;
2170 struct pcpu_dstats __percpu *dstats;
2171 };
2172 unsigned long state;
2173 unsigned int flags;
2174 unsigned short hard_header_len;
2175 enum netdev_stat_type pcpu_stat_type:8;
2176 netdev_features_t features;
2177 struct inet6_dev __rcu *ip6_ptr;
2178 __cacheline_group_end(net_device_read_txrx);
2179
2180 /* RX read-mostly hotpath */
2181 __cacheline_group_begin(net_device_read_rx);
2182 struct bpf_prog __rcu *xdp_prog;
2183 struct list_head ptype_specific;
2184 int ifindex;
2185 unsigned int real_num_rx_queues;
2186 struct netdev_rx_queue *_rx;
2187 unsigned int gro_max_size;
2188 unsigned int gro_ipv4_max_size;
2189 rx_handler_func_t __rcu *rx_handler;
2190 void __rcu *rx_handler_data;
2191 possible_net_t nd_net;
2192#ifdef CONFIG_NETPOLL
2193 struct netpoll_info __rcu *npinfo;
2194#endif
2195#ifdef CONFIG_NET_XGRESS
2196 struct bpf_mprog_entry __rcu *tcx_ingress;
2197#endif
2198 __cacheline_group_end(net_device_read_rx);
2199
2200 char name[IFNAMSIZ];
2201 struct netdev_name_node *name_node;
2202 struct dev_ifalias __rcu *ifalias;
2203 /*
2204 * I/O specific fields
2205 * FIXME: Merge these and struct ifmap into one
2206 */
2207 unsigned long mem_end;
2208 unsigned long mem_start;
2209 unsigned long base_addr;
2210
2211 /*
2212 * Some hardware also needs these fields (state,dev_list,
2213 * napi_list,unreg_list,close_list) but they are not
2214 * part of the usual set specified in Space.c.
2215 */
2216
2217
2218 struct list_head dev_list;
2219 struct list_head napi_list;
2220 struct list_head unreg_list;
2221 struct list_head close_list;
2222 struct list_head ptype_all;
2223
2224 struct {
2225 struct list_head upper;
2226 struct list_head lower;
2227 } adj_list;
2228
2229 /* Read-mostly cache-line for fast-path access */
2230 xdp_features_t xdp_features;
2231 const struct xdp_metadata_ops *xdp_metadata_ops;
2232 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2233 unsigned short gflags;
2234
2235 unsigned short needed_tailroom;
2236
2237 netdev_features_t hw_features;
2238 netdev_features_t wanted_features;
2239 netdev_features_t vlan_features;
2240 netdev_features_t hw_enc_features;
2241 netdev_features_t mpls_features;
2242 netdev_features_t mangleid_features;
2243
2244 unsigned int min_mtu;
2245 unsigned int max_mtu;
2246 unsigned short type;
2247 unsigned char min_header_len;
2248 unsigned char name_assign_type;
2249
2250 int group;
2251
2252 struct net_device_stats stats; /* not used by modern drivers */
2253
2254 struct net_device_core_stats __percpu *core_stats;
2255
2256 /* Stats to monitor link on/off, flapping */
2257 atomic_t carrier_up_count;
2258 atomic_t carrier_down_count;
2259
2260#ifdef CONFIG_WIRELESS_EXT
2261 const struct iw_handler_def *wireless_handlers;
2262#endif
2263 const struct ethtool_ops *ethtool_ops;
2264#ifdef CONFIG_NET_L3_MASTER_DEV
2265 const struct l3mdev_ops *l3mdev_ops;
2266#endif
2267#if IS_ENABLED(CONFIG_IPV6)
2268 const struct ndisc_ops *ndisc_ops;
2269#endif
2270
2271#ifdef CONFIG_XFRM_OFFLOAD
2272 const struct xfrmdev_ops *xfrmdev_ops;
2273#endif
2274
2275#if IS_ENABLED(CONFIG_TLS_DEVICE)
2276 const struct tlsdev_ops *tlsdev_ops;
2277#endif
2278
2279 unsigned int operstate;
2280 unsigned char link_mode;
2281
2282 unsigned char if_port;
2283 unsigned char dma;
2284
2285 /* Interface address info. */
2286 unsigned char perm_addr[MAX_ADDR_LEN];
2287 unsigned char addr_assign_type;
2288 unsigned char addr_len;
2289 unsigned char upper_level;
2290 unsigned char lower_level;
2291 u8 threaded;
2292
2293 unsigned short neigh_priv_len;
2294 unsigned short dev_id;
2295 unsigned short dev_port;
2296 int irq;
2297 u32 priv_len;
2298
2299 spinlock_t addr_list_lock;
2300
2301 struct netdev_hw_addr_list uc;
2302 struct netdev_hw_addr_list mc;
2303 struct netdev_hw_addr_list dev_addrs;
2304
2305#ifdef CONFIG_SYSFS
2306 struct kset *queues_kset;
2307#endif
2308#ifdef CONFIG_LOCKDEP
2309 struct list_head unlink_list;
2310#endif
2311 unsigned int promiscuity;
2312 unsigned int allmulti;
2313 bool uc_promisc;
2314 struct list_head rx_mode_node;
2315 netdevice_tracker rx_mode_tracker;
2316 struct netdev_hw_addr_list rx_mode_addr_cache;
2317#ifdef CONFIG_LOCKDEP
2318 unsigned char nested_level;
2319#endif
2320
2321
2322 /* Protocol-specific pointers */
2323 struct in_device __rcu *ip_ptr;
2324 /** @fib_nh_head: nexthops associated with this netdev */
2325 struct hlist_head fib_nh_head;
2326
2327#if IS_ENABLED(CONFIG_VLAN_8021Q)
2328 struct vlan_info __rcu *vlan_info;
2329#endif
2330#if IS_ENABLED(CONFIG_NET_DSA)
2331 struct dsa_port *dsa_ptr;
2332#endif
2333#if IS_ENABLED(CONFIG_TIPC)
2334 struct tipc_bearer __rcu *tipc_ptr;
2335#endif
2336#if IS_ENABLED(CONFIG_ATALK)
2337 void *atalk_ptr;
2338#endif
2339#if IS_ENABLED(CONFIG_CFG80211)
2340 struct wireless_dev *ieee80211_ptr;
2341#endif
2342#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2343 struct wpan_dev *ieee802154_ptr;
2344#endif
2345#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2346 struct mpls_dev __rcu *mpls_ptr;
2347#endif
2348#if IS_ENABLED(CONFIG_MCTP)
2349 struct mctp_dev __rcu *mctp_ptr;
2350#endif
2351#if IS_ENABLED(CONFIG_INET_PSP)
2352 struct psp_dev __rcu *psp_dev;
2353#endif
2354
2355/*
2356 * Cache lines mostly used on receive path (including eth_type_trans())
2357 */
2358 /* Interface address info used in eth_type_trans() */
2359 const unsigned char *dev_addr;
2360
2361 unsigned int num_rx_queues;
2362#define GRO_LEGACY_MAX_SIZE 65536u
2363/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2364 * and shinfo->gso_segs is a 16bit field.
2365 */
2366#define GRO_MAX_SIZE (8 * 65535u)
2367 unsigned int xdp_zc_max_segs;
2368 struct netdev_queue __rcu *ingress_queue;
2369#ifdef CONFIG_NETFILTER_INGRESS
2370 struct nf_hook_entries __rcu *nf_hooks_ingress;
2371#endif
2372
2373 unsigned char broadcast[MAX_ADDR_LEN];
2374#ifdef CONFIG_RFS_ACCEL
2375 struct cpu_rmap *rx_cpu_rmap;
2376#endif
2377 struct hlist_node index_hlist;
2378
2379/*
2380 * Cache lines mostly used on transmit path
2381 */
2382 unsigned int num_tx_queues;
2383 struct Qdisc __rcu *qdisc;
2384 unsigned int tx_queue_len;
2385 spinlock_t tx_global_lock;
2386
2387 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2388
2389#ifdef CONFIG_NET_SCHED
2390 DECLARE_HASHTABLE (qdisc_hash, 4);
2391#endif
2392 /* These may be needed for future network-power-down code. */
2393 struct timer_list watchdog_timer;
2394 int watchdog_timeo;
2395
2396 u32 proto_down_reason;
2397
2398 struct list_head todo_list;
2399
2400#ifdef CONFIG_PCPU_DEV_REFCNT
2401 int __percpu *pcpu_refcnt;
2402#else
2403 refcount_t dev_refcnt;
2404#endif
2405 struct ref_tracker_dir refcnt_tracker;
2406
2407 struct list_head link_watch_list;
2408
2409 u8 reg_state;
2410
2411 bool dismantle;
2412
2413 /** @moving_ns: device is changing netns, protected by @lock */
2414 bool moving_ns;
2415 /** @rtnl_link_initializing: Device being created, suppress events */
2416 bool rtnl_link_initializing;
2417
2418 bool needs_free_netdev;
2419 void (*priv_destructor)(struct net_device *dev);
2420
2421 /* mid-layer private */
2422 void *ml_priv;
2423 enum netdev_ml_priv_type ml_priv_type;
2424
2425#if IS_ENABLED(CONFIG_GARP)
2426 struct garp_port __rcu *garp_port;
2427#endif
2428#if IS_ENABLED(CONFIG_MRP)
2429 struct mrp_port __rcu *mrp_port;
2430#endif
2431#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2432 struct dm_hw_stat_delta __rcu *dm_private;
2433#endif
2434 struct device dev;
2435 const struct attribute_group *sysfs_groups[5];
2436 const struct attribute_group *sysfs_rx_queue_group;
2437
2438 const struct rtnl_link_ops *rtnl_link_ops;
2439
2440 const struct netdev_stat_ops *stat_ops;
2441
2442 const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2443
2444 /* for setting kernel sock attribute on TCP connection setup */
2445#define GSO_MAX_SEGS 65535u
2446#define GSO_LEGACY_MAX_SIZE 65536u
2447/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2448 * and shinfo->gso_segs is a 16bit field.
2449 */
2450#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2451
2452#define TSO_LEGACY_MAX_SIZE 65536
2453#define TSO_MAX_SIZE UINT_MAX
2454 unsigned int tso_max_size;
2455#define TSO_MAX_SEGS U16_MAX
2456 u16 tso_max_segs;
2457
2458#ifdef CONFIG_DCB
2459 const struct dcbnl_rtnl_ops *dcbnl_ops;
2460#endif
2461 u8 prio_tc_map[TC_BITMASK + 1];
2462
2463#if IS_ENABLED(CONFIG_FCOE)
2464 unsigned int fcoe_ddp_xid;
2465#endif
2466#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2467 struct netprio_map __rcu *priomap;
2468#endif
2469 struct phy_link_topology *link_topo;
2470 struct phy_device *phydev;
2471 struct sfp_bus *sfp_bus;
2472 struct lock_class_key *qdisc_tx_busylock;
2473 bool proto_down;
2474 bool irq_affinity_auto;
2475 bool rx_cpu_rmap_auto;
2476
2477 /* priv_flags_slow, ungrouped to save space */
2478 unsigned long see_all_hwtstamp_requests:1;
2479 unsigned long change_proto_down:1;
2480 unsigned long netns_immutable:1;
2481 unsigned long fcoe_mtu:1;
2482
2483 struct list_head net_notifier_list;
2484
2485#if IS_ENABLED(CONFIG_MACSEC)
2486 /* MACsec management functions */
2487 const struct macsec_ops *macsec_ops;
2488#endif
2489 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2490 struct udp_tunnel_nic *udp_tunnel_nic;
2491
2492 /** @cfg: net_device queue-related configuration */
2493 struct netdev_config *cfg;
2494 /**
2495 * @cfg_pending: same as @cfg but when device is being actively
2496 * reconfigured includes any changes to the configuration
2497 * requested by the user, but which may or may not be rejected.
2498 */
2499 struct netdev_config *cfg_pending;
2500 struct ethtool_netdev_state *ethtool;
2501
2502 /* protected by rtnl_lock */
2503 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2504
2505 u8 dev_addr_shadow[MAX_ADDR_LEN];
2506 netdevice_tracker linkwatch_dev_tracker;
2507 netdevice_tracker watchdog_dev_tracker;
2508 netdevice_tracker dev_registered_tracker;
2509 struct rtnl_hw_stats64 *offload_xstats_l3;
2510
2511 struct devlink_port *devlink_port;
2512
2513#if IS_ENABLED(CONFIG_DPLL)
2514 struct dpll_pin __rcu *dpll_pin;
2515#endif
2516#if IS_ENABLED(CONFIG_PAGE_POOL)
2517 /** @page_pools: page pools created for this netdevice */
2518 struct hlist_head page_pools;
2519#endif
2520
2521 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2522 struct dim_irq_moder *irq_moder;
2523
2524 u64 max_pacing_offload_horizon;
2525 struct napi_config *napi_config;
2526 u32 num_napi_configs;
2527 u32 napi_defer_hard_irqs;
2528 unsigned long gro_flush_timeout;
2529
2530 /**
2531 * @up: copy of @state's IFF_UP, but safe to read with just @lock.
2532 * May report false negatives while the device is being opened
2533 * or closed (@lock does not protect .ndo_open, or .ndo_close).
2534 */
2535 bool up;
2536
2537 /**
2538 * @request_ops_lock: request the core to run all @netdev_ops and
2539 * @ethtool_ops under the @lock.
2540 */
2541 bool request_ops_lock;
2542
2543 /**
2544 * @lock: netdev-scope lock, protects a small selection of fields.
2545 * Should always be taken using netdev_lock() / netdev_unlock() helpers.
2546 * Drivers are free to use it for other protection.
2547 *
2548 * For the drivers that implement shaper or queue API, the scope
2549 * of this lock is expanded to cover most ndo/queue/ethtool/sysfs
2550 * operations. Drivers may opt-in to this behavior by setting
2551 * @request_ops_lock.
2552 *
2553 * @lock protection mixes with rtnl_lock in multiple ways, fields are
2554 * either:
2555 *
2556 * - simply protected by the instance @lock;
2557 *
2558 * - double protected - writers hold both locks, readers hold either;
2559 *
2560 * - ops protected - protected by the lock held around the NDOs
2561 * and other callbacks, that is the instance lock on devices for
2562 * which netdev_need_ops_lock() returns true, otherwise by rtnl_lock;
2563 *
2564 * - double ops protected - always protected by rtnl_lock but for
2565 * devices for which netdev_need_ops_lock() returns true - also
2566 * the instance lock.
2567 *
2568 * Simply protects:
2569 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list,
2570 * @net_shaper_hierarchy, @reg_state, @threaded
2571 *
2572 * Double protects:
2573 * @up, @moving_ns, @nd_net, @xdp_features
2574 *
2575 * Double ops protects:
2576 * @real_num_rx_queues, @real_num_tx_queues
2577 *
2578 * Also protects some fields in:
2579 * struct napi_struct, struct netdev_queue, struct netdev_rx_queue
2580 *
2581 * Ordering:
2582 *
2583 * - take after rtnl_lock
2584 *
2585 * - for the case of netdev queue leasing, the netdev-scope lock is
2586 * taken for both the virtual and the physical device; to prevent
2587 * deadlocks, the virtual device's lock must always be acquired
2588 * before the physical device's (see netdev_nl_queue_create_doit)
2589 */
2590 struct mutex lock;
2591
2592#if IS_ENABLED(CONFIG_NET_SHAPER)
2593 /**
2594 * @net_shaper_hierarchy: data tracking the current shaper status
2595 * see include/net/net_shapers.h
2596 */
2597 struct net_shaper_hierarchy *net_shaper_hierarchy;
2598#endif
2599
2600 struct hlist_head neighbours[NEIGH_NR_TABLES];
2601
2602 struct hwtstamp_provider __rcu *hwprov;
2603
2604 u8 priv[] ____cacheline_aligned
2605 __counted_by(priv_len);
2606} ____cacheline_aligned;
2607#define to_net_dev(d) container_of(d, struct net_device, dev)
2608
2609/*
2610 * Driver should use this to assign devlink port instance to a netdevice
2611 * before it registers the netdevice. Therefore devlink_port is static
2612 * during the netdev lifetime after it is registered.
2613 */
2614#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2615({ \
2616 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2617 ((dev)->devlink_port = (port)); \
2618})
2619
2620static inline bool netif_elide_gro(const struct net_device *dev)
2621{
2622 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2623 return true;
2624 return false;
2625}
2626
2627#define NETDEV_ALIGN 32
2628
2629static inline
2630int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2631{
2632 return dev->prio_tc_map[prio & TC_BITMASK];
2633}
2634
2635static inline
2636int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2637{
2638 if (tc >= dev->num_tc)
2639 return -EINVAL;
2640
2641 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2642 return 0;
2643}
2644
2645int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2646void netdev_reset_tc(struct net_device *dev);
2647int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2648int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2649
2650static inline
2651int netdev_get_num_tc(struct net_device *dev)
2652{
2653 return dev->num_tc;
2654}
2655
2656static inline void net_prefetch(void *p)
2657{
2658 prefetch(p);
2659#if L1_CACHE_BYTES < 128
2660 prefetch((u8 *)p + L1_CACHE_BYTES);
2661#endif
2662}
2663
2664static inline void net_prefetchw(void *p)
2665{
2666 prefetchw(p);
2667#if L1_CACHE_BYTES < 128
2668 prefetchw((u8 *)p + L1_CACHE_BYTES);
2669#endif
2670}
2671
2672void netdev_unbind_sb_channel(struct net_device *dev,
2673 struct net_device *sb_dev);
2674int netdev_bind_sb_channel_queue(struct net_device *dev,
2675 struct net_device *sb_dev,
2676 u8 tc, u16 count, u16 offset);
2677int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2678static inline int netdev_get_sb_channel(struct net_device *dev)
2679{
2680 return max_t(int, -dev->num_tc, 0);
2681}
2682
2683static inline
2684struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2685 unsigned int index)
2686{
2687 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2688 return &dev->_tx[index];
2689}
2690
2691static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2692 const struct sk_buff *skb)
2693{
2694 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2695}
2696
2697static inline void netdev_for_each_tx_queue(struct net_device *dev,
2698 void (*f)(struct net_device *,
2699 struct netdev_queue *,
2700 void *),
2701 void *arg)
2702{
2703 unsigned int i;
2704
2705 for (i = 0; i < dev->num_tx_queues; i++)
2706 f(dev, &dev->_tx[i], arg);
2707}
2708
2709u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2710 struct net_device *sb_dev);
2711struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2712 struct sk_buff *skb,
2713 struct net_device *sb_dev);
2714
2715/* returns the headroom that the master device needs to take in account
2716 * when forwarding to this dev
2717 */
2718static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2719{
2720 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2721}
2722
2723static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2724{
2725 if (dev->netdev_ops->ndo_set_rx_headroom)
2726 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2727}
2728
2729/* set the device rx headroom to the dev's default */
2730static inline void netdev_reset_rx_headroom(struct net_device *dev)
2731{
2732 netdev_set_rx_headroom(dev, -1);
2733}
2734
2735static inline void *netdev_get_ml_priv(struct net_device *dev,
2736 enum netdev_ml_priv_type type)
2737{
2738 if (dev->ml_priv_type != type)
2739 return NULL;
2740
2741 return dev->ml_priv;
2742}
2743
2744static inline void netdev_set_ml_priv(struct net_device *dev,
2745 void *ml_priv,
2746 enum netdev_ml_priv_type type)
2747{
2748 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2749 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2750 dev->ml_priv_type, type);
2751 WARN(!dev->ml_priv_type && dev->ml_priv,
2752 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2753
2754 dev->ml_priv = ml_priv;
2755 dev->ml_priv_type = type;
2756}
2757
2758/*
2759 * Net namespace inlines
2760 */
2761static inline
2762struct net *dev_net(const struct net_device *dev)
2763{
2764 return read_pnet(&dev->nd_net);
2765}
2766
2767static inline
2768struct net *dev_net_rcu(const struct net_device *dev)
2769{
2770 return read_pnet_rcu(&dev->nd_net);
2771}
2772
2773static inline
2774void dev_net_set(struct net_device *dev, struct net *net)
2775{
2776 write_pnet(&dev->nd_net, net);
2777}
2778
2779/**
2780 * netdev_priv - access network device private data
2781 * @dev: network device
2782 *
2783 * Get network device private data
2784 */
2785static inline void *netdev_priv(const struct net_device *dev)
2786{
2787 return (void *)dev->priv;
2788}
2789
2790/**
2791 * netdev_from_priv() - get network device from priv
2792 * @priv: network device private data
2793 *
2794 * Returns: net_device to which @priv belongs
2795 */
2796static inline struct net_device *netdev_from_priv(const void *priv)
2797{
2798 return container_of(priv, struct net_device, priv);
2799}
2800
2801/* Set the sysfs physical device reference for the network logical device
2802 * if set prior to registration will cause a symlink during initialization.
2803 */
2804#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2805
2806/* Set the sysfs device type for the network logical device to allow
2807 * fine-grained identification of different network device types. For
2808 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2809 */
2810#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2811
2812void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2813 enum netdev_queue_type type,
2814 struct napi_struct *napi);
2815
2816static inline void netdev_lock(struct net_device *dev)
2817{
2818 mutex_lock(&dev->lock);
2819}
2820
2821static inline void netdev_unlock(struct net_device *dev)
2822{
2823 mutex_unlock(&dev->lock);
2824}
2825/* Additional netdev_lock()-related helpers are in net/netdev_lock.h */
2826
2827void netif_napi_set_irq_locked(struct napi_struct *napi, int irq);
2828
2829static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2830{
2831 netdev_lock(napi->dev);
2832 netif_napi_set_irq_locked(napi, irq);
2833 netdev_unlock(napi->dev);
2834}
2835
2836/* Default NAPI poll() weight
2837 * Device drivers are strongly advised to not use bigger value
2838 */
2839#define NAPI_POLL_WEIGHT 64
2840
2841void netif_napi_add_weight_locked(struct net_device *dev,
2842 struct napi_struct *napi,
2843 int (*poll)(struct napi_struct *, int),
2844 int weight);
2845
2846static inline void
2847netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2848 int (*poll)(struct napi_struct *, int), int weight)
2849{
2850 netdev_lock(dev);
2851 netif_napi_add_weight_locked(dev, napi, poll, weight);
2852 netdev_unlock(dev);
2853}
2854
2855/**
2856 * netif_napi_add() - initialize a NAPI context
2857 * @dev: network device
2858 * @napi: NAPI context
2859 * @poll: polling function
2860 *
2861 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2862 * *any* of the other NAPI-related functions.
2863 */
2864static inline void
2865netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2866 int (*poll)(struct napi_struct *, int))
2867{
2868 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2869}
2870
2871static inline void
2872netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi,
2873 int (*poll)(struct napi_struct *, int))
2874{
2875 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT);
2876}
2877
2878static inline void
2879netif_napi_add_tx_weight(struct net_device *dev,
2880 struct napi_struct *napi,
2881 int (*poll)(struct napi_struct *, int),
2882 int weight)
2883{
2884 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2885 netif_napi_add_weight(dev, napi, poll, weight);
2886}
2887
2888static inline void
2889netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi,
2890 int (*poll)(struct napi_struct *, int), int index)
2891{
2892 napi->index = index;
2893 napi->config = &dev->napi_config[index];
2894 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT);
2895}
2896
2897/**
2898 * netif_napi_add_config - initialize a NAPI context with persistent config
2899 * @dev: network device
2900 * @napi: NAPI context
2901 * @poll: polling function
2902 * @index: the NAPI index
2903 */
2904static inline void
2905netif_napi_add_config(struct net_device *dev, struct napi_struct *napi,
2906 int (*poll)(struct napi_struct *, int), int index)
2907{
2908 netdev_lock(dev);
2909 netif_napi_add_config_locked(dev, napi, poll, index);
2910 netdev_unlock(dev);
2911}
2912
2913/**
2914 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2915 * @dev: network device
2916 * @napi: NAPI context
2917 * @poll: polling function
2918 *
2919 * This variant of netif_napi_add() should be used from drivers using NAPI
2920 * to exclusively poll a TX queue.
2921 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2922 */
2923static inline void netif_napi_add_tx(struct net_device *dev,
2924 struct napi_struct *napi,
2925 int (*poll)(struct napi_struct *, int))
2926{
2927 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2928}
2929
2930void __netif_napi_del_locked(struct napi_struct *napi);
2931
2932/**
2933 * __netif_napi_del - remove a NAPI context
2934 * @napi: NAPI context
2935 *
2936 * Warning: caller must observe RCU grace period before freeing memory
2937 * containing @napi. Drivers might want to call this helper to combine
2938 * all the needed RCU grace periods into a single one.
2939 */
2940static inline void __netif_napi_del(struct napi_struct *napi)
2941{
2942 netdev_lock(napi->dev);
2943 __netif_napi_del_locked(napi);
2944 netdev_unlock(napi->dev);
2945}
2946
2947static inline void netif_napi_del_locked(struct napi_struct *napi)
2948{
2949 __netif_napi_del_locked(napi);
2950 synchronize_net();
2951}
2952
2953/**
2954 * netif_napi_del - remove a NAPI context
2955 * @napi: NAPI context
2956 *
2957 * netif_napi_del() removes a NAPI context from the network device NAPI list
2958 */
2959static inline void netif_napi_del(struct napi_struct *napi)
2960{
2961 __netif_napi_del(napi);
2962 synchronize_net();
2963}
2964
2965int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs);
2966void netif_set_affinity_auto(struct net_device *dev);
2967
2968struct packet_type {
2969 __be16 type; /* This is really htons(ether_type). */
2970 bool ignore_outgoing;
2971 struct net_device *dev; /* NULL is wildcarded here */
2972 netdevice_tracker dev_tracker;
2973 int (*func) (struct sk_buff *,
2974 struct net_device *,
2975 struct packet_type *,
2976 struct net_device *);
2977 void (*list_func) (struct list_head *,
2978 struct packet_type *,
2979 struct net_device *);
2980 bool (*id_match)(struct packet_type *ptype,
2981 struct sock *sk);
2982 struct net *af_packet_net;
2983 void *af_packet_priv;
2984 struct list_head list;
2985};
2986
2987struct offload_callbacks {
2988 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2989 netdev_features_t features);
2990 struct sk_buff *(*gro_receive)(struct list_head *head,
2991 struct sk_buff *skb);
2992 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2993};
2994
2995struct packet_offload {
2996 __be16 type; /* This is really htons(ether_type). */
2997 u16 priority;
2998 struct offload_callbacks callbacks;
2999 struct list_head list;
3000};
3001
3002/* often modified stats are per-CPU, other are shared (netdev->stats) */
3003struct pcpu_sw_netstats {
3004 u64_stats_t rx_packets;
3005 u64_stats_t rx_bytes;
3006 u64_stats_t tx_packets;
3007 u64_stats_t tx_bytes;
3008 struct u64_stats_sync syncp;
3009} __aligned(4 * sizeof(u64));
3010
3011struct pcpu_dstats {
3012 u64_stats_t rx_packets;
3013 u64_stats_t rx_bytes;
3014 u64_stats_t tx_packets;
3015 u64_stats_t tx_bytes;
3016 u64_stats_t rx_drops;
3017 u64_stats_t tx_drops;
3018 struct u64_stats_sync syncp;
3019} __aligned(8 * sizeof(u64));
3020
3021struct pcpu_lstats {
3022 u64_stats_t packets;
3023 u64_stats_t bytes;
3024 struct u64_stats_sync syncp;
3025} __aligned(2 * sizeof(u64));
3026
3027void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
3028
3029static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
3030{
3031 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
3032
3033 u64_stats_update_begin(&tstats->syncp);
3034 u64_stats_add(&tstats->rx_bytes, len);
3035 u64_stats_inc(&tstats->rx_packets);
3036 u64_stats_update_end(&tstats->syncp);
3037}
3038
3039static inline void dev_sw_netstats_tx_add(struct net_device *dev,
3040 unsigned int packets,
3041 unsigned int len)
3042{
3043 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
3044
3045 u64_stats_update_begin(&tstats->syncp);
3046 u64_stats_add(&tstats->tx_bytes, len);
3047 u64_stats_add(&tstats->tx_packets, packets);
3048 u64_stats_update_end(&tstats->syncp);
3049}
3050
3051static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
3052{
3053 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
3054
3055 u64_stats_update_begin(&lstats->syncp);
3056 u64_stats_add(&lstats->bytes, len);
3057 u64_stats_inc(&lstats->packets);
3058 u64_stats_update_end(&lstats->syncp);
3059}
3060
3061static inline void dev_dstats_rx_add(struct net_device *dev,
3062 unsigned int len)
3063{
3064 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
3065
3066 u64_stats_update_begin(&dstats->syncp);
3067 u64_stats_inc(&dstats->rx_packets);
3068 u64_stats_add(&dstats->rx_bytes, len);
3069 u64_stats_update_end(&dstats->syncp);
3070}
3071
3072static inline void dev_dstats_rx_dropped(struct net_device *dev)
3073{
3074 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
3075
3076 u64_stats_update_begin(&dstats->syncp);
3077 u64_stats_inc(&dstats->rx_drops);
3078 u64_stats_update_end(&dstats->syncp);
3079}
3080
3081static inline void dev_dstats_rx_dropped_add(struct net_device *dev,
3082 unsigned int packets)
3083{
3084 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
3085
3086 u64_stats_update_begin(&dstats->syncp);
3087 u64_stats_add(&dstats->rx_drops, packets);
3088 u64_stats_update_end(&dstats->syncp);
3089}
3090
3091static inline void dev_dstats_tx_add(struct net_device *dev,
3092 unsigned int len)
3093{
3094 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
3095
3096 u64_stats_update_begin(&dstats->syncp);
3097 u64_stats_inc(&dstats->tx_packets);
3098 u64_stats_add(&dstats->tx_bytes, len);
3099 u64_stats_update_end(&dstats->syncp);
3100}
3101
3102static inline void dev_dstats_tx_dropped(struct net_device *dev)
3103{
3104 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
3105
3106 u64_stats_update_begin(&dstats->syncp);
3107 u64_stats_inc(&dstats->tx_drops);
3108 u64_stats_update_end(&dstats->syncp);
3109}
3110
3111#define __netdev_alloc_pcpu_stats(type, gfp) \
3112({ \
3113 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
3114 if (pcpu_stats) { \
3115 int __cpu; \
3116 for_each_possible_cpu(__cpu) { \
3117 typeof(type) *stat; \
3118 stat = per_cpu_ptr(pcpu_stats, __cpu); \
3119 u64_stats_init(&stat->syncp); \
3120 } \
3121 } \
3122 pcpu_stats; \
3123})
3124
3125#define netdev_alloc_pcpu_stats(type) \
3126 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
3127
3128#define devm_netdev_alloc_pcpu_stats(dev, type) \
3129({ \
3130 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
3131 if (pcpu_stats) { \
3132 int __cpu; \
3133 for_each_possible_cpu(__cpu) { \
3134 typeof(type) *stat; \
3135 stat = per_cpu_ptr(pcpu_stats, __cpu); \
3136 u64_stats_init(&stat->syncp); \
3137 } \
3138 } \
3139 pcpu_stats; \
3140})
3141
3142enum netdev_lag_tx_type {
3143 NETDEV_LAG_TX_TYPE_UNKNOWN,
3144 NETDEV_LAG_TX_TYPE_RANDOM,
3145 NETDEV_LAG_TX_TYPE_BROADCAST,
3146 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
3147 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
3148 NETDEV_LAG_TX_TYPE_HASH,
3149};
3150
3151enum netdev_lag_hash {
3152 NETDEV_LAG_HASH_NONE,
3153 NETDEV_LAG_HASH_L2,
3154 NETDEV_LAG_HASH_L34,
3155 NETDEV_LAG_HASH_L23,
3156 NETDEV_LAG_HASH_E23,
3157 NETDEV_LAG_HASH_E34,
3158 NETDEV_LAG_HASH_VLAN_SRCMAC,
3159 NETDEV_LAG_HASH_UNKNOWN,
3160};
3161
3162struct netdev_lag_upper_info {
3163 enum netdev_lag_tx_type tx_type;
3164 enum netdev_lag_hash hash_type;
3165};
3166
3167struct netdev_lag_lower_state_info {
3168 u8 link_up : 1,
3169 tx_enabled : 1;
3170};
3171
3172#include <linux/notifier.h>
3173
3174/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
3175 * and the rtnetlink notification exclusion list in rtnetlink_event() when
3176 * adding new types.
3177 */
3178enum netdev_cmd {
3179 NETDEV_UP = 1, /* For now you can't veto a device up/down */
3180 NETDEV_DOWN,
3181 NETDEV_REBOOT, /* Tell a protocol stack a network interface
3182 detected a hardware crash and restarted
3183 - we can use this eg to kick tcp sessions
3184 once done */
3185 NETDEV_CHANGE, /* Notify device state change */
3186 NETDEV_REGISTER,
3187 NETDEV_UNREGISTER,
3188 NETDEV_CHANGEMTU, /* notify after mtu change happened */
3189 NETDEV_CHANGEADDR, /* notify after the address change */
3190 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
3191 NETDEV_GOING_DOWN,
3192 NETDEV_CHANGENAME,
3193 NETDEV_FEAT_CHANGE,
3194 NETDEV_BONDING_FAILOVER,
3195 NETDEV_PRE_UP,
3196 NETDEV_PRE_TYPE_CHANGE,
3197 NETDEV_POST_TYPE_CHANGE,
3198 NETDEV_POST_INIT,
3199 NETDEV_PRE_UNINIT,
3200 NETDEV_RELEASE,
3201 NETDEV_NOTIFY_PEERS,
3202 NETDEV_JOIN,
3203 NETDEV_CHANGEUPPER,
3204 NETDEV_RESEND_IGMP,
3205 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
3206 NETDEV_CHANGEINFODATA,
3207 NETDEV_BONDING_INFO,
3208 NETDEV_PRECHANGEUPPER,
3209 NETDEV_CHANGELOWERSTATE,
3210 NETDEV_UDP_TUNNEL_PUSH_INFO,
3211 NETDEV_UDP_TUNNEL_DROP_INFO,
3212 NETDEV_CHANGE_TX_QUEUE_LEN,
3213 NETDEV_CVLAN_FILTER_PUSH_INFO,
3214 NETDEV_CVLAN_FILTER_DROP_INFO,
3215 NETDEV_SVLAN_FILTER_PUSH_INFO,
3216 NETDEV_SVLAN_FILTER_DROP_INFO,
3217 NETDEV_OFFLOAD_XSTATS_ENABLE,
3218 NETDEV_OFFLOAD_XSTATS_DISABLE,
3219 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
3220 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
3221 NETDEV_XDP_FEAT_CHANGE,
3222};
3223const char *netdev_cmd_to_name(enum netdev_cmd cmd);
3224
3225int register_netdevice_notifier(struct notifier_block *nb);
3226int unregister_netdevice_notifier(struct notifier_block *nb);
3227int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
3228int unregister_netdevice_notifier_net(struct net *net,
3229 struct notifier_block *nb);
3230int register_netdevice_notifier_dev_net(struct net_device *dev,
3231 struct notifier_block *nb,
3232 struct netdev_net_notifier *nn);
3233int unregister_netdevice_notifier_dev_net(struct net_device *dev,
3234 struct notifier_block *nb,
3235 struct netdev_net_notifier *nn);
3236
3237struct netdev_notifier_info {
3238 struct net_device *dev;
3239 struct netlink_ext_ack *extack;
3240};
3241
3242struct netdev_notifier_info_ext {
3243 struct netdev_notifier_info info; /* must be first */
3244 union {
3245 u32 mtu;
3246 } ext;
3247};
3248
3249struct netdev_notifier_change_info {
3250 struct netdev_notifier_info info; /* must be first */
3251 unsigned int flags_changed;
3252};
3253
3254struct netdev_notifier_changeupper_info {
3255 struct netdev_notifier_info info; /* must be first */
3256 struct net_device *upper_dev; /* new upper dev */
3257 bool master; /* is upper dev master */
3258 bool linking; /* is the notification for link or unlink */
3259 void *upper_info; /* upper dev info */
3260};
3261
3262struct netdev_notifier_changelowerstate_info {
3263 struct netdev_notifier_info info; /* must be first */
3264 void *lower_state_info; /* is lower dev state */
3265};
3266
3267struct netdev_notifier_pre_changeaddr_info {
3268 struct netdev_notifier_info info; /* must be first */
3269 const unsigned char *dev_addr;
3270};
3271
3272enum netdev_offload_xstats_type {
3273 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3274};
3275
3276struct netdev_notifier_offload_xstats_info {
3277 struct netdev_notifier_info info; /* must be first */
3278 enum netdev_offload_xstats_type type;
3279
3280 union {
3281 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3282 struct netdev_notifier_offload_xstats_rd *report_delta;
3283 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3284 struct netdev_notifier_offload_xstats_ru *report_used;
3285 };
3286};
3287
3288int netdev_offload_xstats_enable(struct net_device *dev,
3289 enum netdev_offload_xstats_type type,
3290 struct netlink_ext_ack *extack);
3291int netdev_offload_xstats_disable(struct net_device *dev,
3292 enum netdev_offload_xstats_type type);
3293bool netdev_offload_xstats_enabled(const struct net_device *dev,
3294 enum netdev_offload_xstats_type type);
3295int netdev_offload_xstats_get(struct net_device *dev,
3296 enum netdev_offload_xstats_type type,
3297 struct rtnl_hw_stats64 *stats, bool *used,
3298 struct netlink_ext_ack *extack);
3299void
3300netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3301 const struct rtnl_hw_stats64 *stats);
3302void
3303netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3304void netdev_offload_xstats_push_delta(struct net_device *dev,
3305 enum netdev_offload_xstats_type type,
3306 const struct rtnl_hw_stats64 *stats);
3307
3308static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3309 struct net_device *dev)
3310{
3311 info->dev = dev;
3312 info->extack = NULL;
3313}
3314
3315static inline struct net_device *
3316netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3317{
3318 return info->dev;
3319}
3320
3321static inline struct netlink_ext_ack *
3322netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3323{
3324 return info->extack;
3325}
3326
3327int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3328int call_netdevice_notifiers_info(unsigned long val,
3329 struct netdev_notifier_info *info);
3330
3331#define for_each_netdev(net, d) \
3332 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3333#define for_each_netdev_reverse(net, d) \
3334 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3335#define for_each_netdev_rcu(net, d) \
3336 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3337#define for_each_netdev_safe(net, d, n) \
3338 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3339#define for_each_netdev_continue(net, d) \
3340 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3341#define for_each_netdev_continue_reverse(net, d) \
3342 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3343 dev_list)
3344#define for_each_netdev_continue_rcu(net, d) \
3345 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3346#define for_each_netdev_in_bond_rcu(bond, slave) \
3347 for_each_netdev_rcu(dev_net_rcu(bond), slave) \
3348 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3349#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3350
3351#define for_each_netdev_dump(net, d, ifindex) \
3352 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \
3353 ULONG_MAX, XA_PRESENT)); ifindex++)
3354
3355static inline struct net_device *next_net_device(struct net_device *dev)
3356{
3357 struct list_head *lh;
3358 struct net *net;
3359
3360 net = dev_net(dev);
3361 lh = dev->dev_list.next;
3362 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3363}
3364
3365static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3366{
3367 struct list_head *lh;
3368 struct net *net;
3369
3370 net = dev_net(dev);
3371 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3372 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3373}
3374
3375static inline struct net_device *first_net_device(struct net *net)
3376{
3377 return list_empty(&net->dev_base_head) ? NULL :
3378 net_device_entry(net->dev_base_head.next);
3379}
3380
3381int netdev_boot_setup_check(struct net_device *dev);
3382struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type,
3383 const char *hwaddr);
3384struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3385 const char *hwaddr);
3386struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3387void dev_add_pack(struct packet_type *pt);
3388void dev_remove_pack(struct packet_type *pt);
3389void __dev_remove_pack(struct packet_type *pt);
3390void dev_add_offload(struct packet_offload *po);
3391void dev_remove_offload(struct packet_offload *po);
3392
3393int dev_get_iflink(const struct net_device *dev);
3394int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3395int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3396 struct net_device_path_stack *stack);
3397struct net_device *dev_get_by_name(struct net *net, const char *name);
3398struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3399struct net_device *__dev_get_by_name(struct net *net, const char *name);
3400bool netdev_name_in_use(struct net *net, const char *name);
3401int dev_alloc_name(struct net_device *dev, const char *name);
3402int netif_open(struct net_device *dev, struct netlink_ext_ack *extack);
3403int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3404void netif_close(struct net_device *dev);
3405void dev_close(struct net_device *dev);
3406void netif_close_many(struct list_head *head, bool unlink);
3407void netif_disable_lro(struct net_device *dev);
3408void dev_disable_lro(struct net_device *dev);
3409int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3410u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3411 struct net_device *sb_dev);
3412
3413int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3414int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3415
3416static inline int dev_queue_xmit(struct sk_buff *skb)
3417{
3418 return __dev_queue_xmit(skb, NULL);
3419}
3420
3421static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3422 struct net_device *sb_dev)
3423{
3424 return __dev_queue_xmit(skb, sb_dev);
3425}
3426
3427static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3428{
3429 int ret;
3430
3431 ret = __dev_direct_xmit(skb, queue_id);
3432 if (!dev_xmit_complete(ret))
3433 kfree_skb(skb);
3434 return ret;
3435}
3436
3437int register_netdevice(struct net_device *dev);
3438void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3439void unregister_netdevice_many(struct list_head *head);
3440bool unregister_netdevice_queued(const struct net_device *dev);
3441
3442static inline void unregister_netdevice(struct net_device *dev)
3443{
3444 unregister_netdevice_queue(dev, NULL);
3445}
3446
3447int netdev_refcnt_read(const struct net_device *dev);
3448void free_netdev(struct net_device *dev);
3449
3450struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3451 struct sk_buff *skb,
3452 bool all_slaves);
3453struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3454 struct sock *sk);
3455struct net_device *dev_get_by_index(struct net *net, int ifindex);
3456struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3457struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3458 netdevice_tracker *tracker, gfp_t gfp);
3459struct net_device *netdev_get_by_index_lock(struct net *net, int ifindex);
3460struct net_device *netdev_get_by_name(struct net *net, const char *name,
3461 netdevice_tracker *tracker, gfp_t gfp);
3462struct net_device *netdev_get_by_flags_rcu(struct net *net, netdevice_tracker *tracker,
3463 unsigned short flags, unsigned short mask);
3464struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3465void netdev_copy_name(struct net_device *dev, char *name);
3466
3467static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3468 unsigned short type,
3469 const void *daddr, const void *saddr,
3470 unsigned int len)
3471{
3472 if (!dev->header_ops || !dev->header_ops->create)
3473 return 0;
3474
3475 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3476}
3477
3478static inline int dev_parse_header(const struct sk_buff *skb,
3479 unsigned char *haddr)
3480{
3481 const struct net_device *dev = skb->dev;
3482
3483 if (!dev->header_ops || !dev->header_ops->parse)
3484 return 0;
3485 return dev->header_ops->parse(skb, dev, haddr);
3486}
3487
3488static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3489{
3490 const struct net_device *dev = skb->dev;
3491
3492 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3493 return 0;
3494 return dev->header_ops->parse_protocol(skb);
3495}
3496
3497/* ll_header must have at least hard_header_len allocated */
3498static inline bool dev_validate_header(const struct net_device *dev,
3499 char *ll_header, int len)
3500{
3501 if (likely(len >= dev->hard_header_len))
3502 return true;
3503 if (len < dev->min_header_len)
3504 return false;
3505
3506 if (capable(CAP_SYS_RAWIO)) {
3507 memset(ll_header + len, 0, dev->hard_header_len - len);
3508 return true;
3509 }
3510
3511 if (dev->header_ops && dev->header_ops->validate)
3512 return dev->header_ops->validate(ll_header, len);
3513
3514 return false;
3515}
3516
3517static inline bool dev_has_header(const struct net_device *dev)
3518{
3519 return dev->header_ops && dev->header_ops->create;
3520}
3521
3522struct numa_drop_counters {
3523 atomic_t drops0 ____cacheline_aligned_in_smp;
3524 atomic_t drops1 ____cacheline_aligned_in_smp;
3525};
3526
3527static inline int numa_drop_read(const struct numa_drop_counters *ndc)
3528{
3529 return atomic_read(&ndc->drops0) + atomic_read(&ndc->drops1);
3530}
3531
3532static inline void numa_drop_add(struct numa_drop_counters *ndc, int val)
3533{
3534 int n = numa_node_id() % 2;
3535
3536 if (n)
3537 atomic_add(val, &ndc->drops1);
3538 else
3539 atomic_add(val, &ndc->drops0);
3540}
3541
3542static inline void numa_drop_reset(struct numa_drop_counters *ndc)
3543{
3544 atomic_set(&ndc->drops0, 0);
3545 atomic_set(&ndc->drops1, 0);
3546}
3547
3548/*
3549 * Incoming packets are placed on per-CPU queues
3550 */
3551struct softnet_data {
3552 struct list_head poll_list;
3553 struct sk_buff_head process_queue;
3554 local_lock_t process_queue_bh_lock;
3555
3556 /* stats */
3557 unsigned int processed;
3558 unsigned int time_squeeze;
3559#ifdef CONFIG_RPS
3560 struct softnet_data *rps_ipi_list;
3561#endif
3562
3563 unsigned int received_rps;
3564 bool in_net_rx_action;
3565 bool in_napi_threaded_poll;
3566
3567#ifdef CONFIG_NET_FLOW_LIMIT
3568 struct sd_flow_limit __rcu *flow_limit;
3569#endif
3570 struct Qdisc *output_queue;
3571 struct Qdisc **output_queue_tailp;
3572 struct sk_buff *completion_queue;
3573#ifdef CONFIG_XFRM_OFFLOAD
3574 struct sk_buff_head xfrm_backlog;
3575#endif
3576 /* written and read only by owning cpu: */
3577 struct netdev_xmit xmit;
3578#ifdef CONFIG_RPS
3579 /* input_queue_head should be written by cpu owning this struct,
3580 * and only read by other cpus. Worth using a cache line.
3581 */
3582 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3583
3584 /* Elements below can be accessed between CPUs for RPS/RFS */
3585 call_single_data_t csd ____cacheline_aligned_in_smp;
3586 struct softnet_data *rps_ipi_next;
3587 unsigned int cpu;
3588
3589 /* We force a cacheline alignment from here, to hold together
3590 * input_queue_tail, input_pkt_queue and backlog.state.
3591 * We add holes so that backlog.state is the last field
3592 * of this cache line.
3593 */
3594 long pad[3] ____cacheline_aligned_in_smp;
3595 unsigned int input_queue_tail;
3596#endif
3597 struct sk_buff_head input_pkt_queue;
3598
3599 struct napi_struct backlog;
3600
3601 struct numa_drop_counters drop_counters;
3602
3603 int defer_ipi_scheduled ____cacheline_aligned_in_smp;
3604 call_single_data_t defer_csd;
3605};
3606
3607DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3608
3609struct page_pool_bh {
3610 struct page_pool *pool;
3611 local_lock_t bh_lock;
3612};
3613DECLARE_PER_CPU(struct page_pool_bh, system_page_pool);
3614
3615#define XMIT_RECURSION_LIMIT 8
3616
3617#ifndef CONFIG_PREEMPT_RT
3618static inline int dev_recursion_level(void)
3619{
3620 return this_cpu_read(softnet_data.xmit.recursion);
3621}
3622
3623static inline bool dev_xmit_recursion(void)
3624{
3625 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3626 XMIT_RECURSION_LIMIT);
3627}
3628
3629static inline void dev_xmit_recursion_inc(void)
3630{
3631 __this_cpu_inc(softnet_data.xmit.recursion);
3632}
3633
3634static inline void dev_xmit_recursion_dec(void)
3635{
3636 __this_cpu_dec(softnet_data.xmit.recursion);
3637}
3638#else
3639static inline int dev_recursion_level(void)
3640{
3641 return current->net_xmit.recursion;
3642}
3643
3644static inline bool dev_xmit_recursion(void)
3645{
3646 return unlikely(current->net_xmit.recursion > XMIT_RECURSION_LIMIT);
3647}
3648
3649static inline void dev_xmit_recursion_inc(void)
3650{
3651 current->net_xmit.recursion++;
3652}
3653
3654static inline void dev_xmit_recursion_dec(void)
3655{
3656 current->net_xmit.recursion--;
3657}
3658#endif
3659
3660void __netif_schedule(struct Qdisc *q);
3661void netif_schedule_queue(struct netdev_queue *txq);
3662
3663static inline void netif_tx_schedule_all(struct net_device *dev)
3664{
3665 unsigned int i;
3666
3667 for (i = 0; i < dev->num_tx_queues; i++)
3668 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3669}
3670
3671static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3672{
3673 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3674}
3675
3676/**
3677 * netif_start_queue - allow transmit
3678 * @dev: network device
3679 *
3680 * Allow upper layers to call the device hard_start_xmit routine.
3681 */
3682static inline void netif_start_queue(struct net_device *dev)
3683{
3684 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3685}
3686
3687static inline void netif_tx_start_all_queues(struct net_device *dev)
3688{
3689 unsigned int i;
3690
3691 for (i = 0; i < dev->num_tx_queues; i++) {
3692 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3693 netif_tx_start_queue(txq);
3694 }
3695}
3696
3697void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3698
3699/**
3700 * netif_wake_queue - restart transmit
3701 * @dev: network device
3702 *
3703 * Allow upper layers to call the device hard_start_xmit routine.
3704 * Used for flow control when transmit resources are available.
3705 */
3706static inline void netif_wake_queue(struct net_device *dev)
3707{
3708 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3709}
3710
3711static inline void netif_tx_wake_all_queues(struct net_device *dev)
3712{
3713 unsigned int i;
3714
3715 for (i = 0; i < dev->num_tx_queues; i++) {
3716 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3717 netif_tx_wake_queue(txq);
3718 }
3719}
3720
3721static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3722{
3723 /* Paired with READ_ONCE() from dev_watchdog() */
3724 WRITE_ONCE(dev_queue->trans_start, jiffies);
3725
3726 /* This barrier is paired with smp_mb() from dev_watchdog() */
3727 smp_mb__before_atomic();
3728
3729 /* Must be an atomic op see netif_txq_try_stop() */
3730 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3731}
3732
3733/**
3734 * netif_stop_queue - stop transmitted packets
3735 * @dev: network device
3736 *
3737 * Stop upper layers calling the device hard_start_xmit routine.
3738 * Used for flow control when transmit resources are unavailable.
3739 */
3740static inline void netif_stop_queue(struct net_device *dev)
3741{
3742 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3743}
3744
3745void netif_tx_stop_all_queues(struct net_device *dev);
3746
3747static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3748{
3749 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3750}
3751
3752/**
3753 * netif_queue_stopped - test if transmit queue is flowblocked
3754 * @dev: network device
3755 *
3756 * Test if transmit queue on device is currently unable to send.
3757 */
3758static inline bool netif_queue_stopped(const struct net_device *dev)
3759{
3760 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3761}
3762
3763static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3764{
3765 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3766}
3767
3768static inline bool
3769netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3770{
3771 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3772}
3773
3774static inline bool
3775netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3776{
3777 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3778}
3779
3780/**
3781 * netdev_queue_set_dql_min_limit - set dql minimum limit
3782 * @dev_queue: pointer to transmit queue
3783 * @min_limit: dql minimum limit
3784 *
3785 * Forces xmit_more() to return true until the minimum threshold
3786 * defined by @min_limit is reached (or until the tx queue is
3787 * empty). Warning: to be use with care, misuse will impact the
3788 * latency.
3789 */
3790static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3791 unsigned int min_limit)
3792{
3793#ifdef CONFIG_BQL
3794 dev_queue->dql.min_limit = min_limit;
3795#endif
3796}
3797
3798static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3799{
3800#ifdef CONFIG_BQL
3801 /* Non-BQL migrated drivers will return 0, too. */
3802 return dql_avail(&txq->dql);
3803#else
3804 return 0;
3805#endif
3806}
3807
3808/**
3809 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3810 * @dev_queue: pointer to transmit queue
3811 *
3812 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3813 * to give appropriate hint to the CPU.
3814 */
3815static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3816{
3817#ifdef CONFIG_BQL
3818 prefetchw(&dev_queue->dql.num_queued);
3819#endif
3820}
3821
3822/**
3823 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3824 * @dev_queue: pointer to transmit queue
3825 *
3826 * BQL enabled drivers might use this helper in their TX completion path,
3827 * to give appropriate hint to the CPU.
3828 */
3829static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3830{
3831#ifdef CONFIG_BQL
3832 prefetchw(&dev_queue->dql.limit);
3833#endif
3834}
3835
3836/**
3837 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3838 * @dev_queue: network device queue
3839 * @bytes: number of bytes queued to the device queue
3840 *
3841 * Report the number of bytes queued for sending/completion to the network
3842 * device hardware queue. @bytes should be a good approximation and should
3843 * exactly match netdev_completed_queue() @bytes.
3844 * This is typically called once per packet, from ndo_start_xmit().
3845 */
3846static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3847 unsigned int bytes)
3848{
3849#ifdef CONFIG_BQL
3850 dql_queued(&dev_queue->dql, bytes);
3851
3852 if (likely(dql_avail(&dev_queue->dql) >= 0))
3853 return;
3854
3855 /* Paired with READ_ONCE() from dev_watchdog() */
3856 WRITE_ONCE(dev_queue->trans_start, jiffies);
3857
3858 /* This barrier is paired with smp_mb() from dev_watchdog() */
3859 smp_mb__before_atomic();
3860
3861 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3862
3863 /*
3864 * The XOFF flag must be set before checking the dql_avail below,
3865 * because in netdev_tx_completed_queue we update the dql_completed
3866 * before checking the XOFF flag.
3867 */
3868 smp_mb__after_atomic();
3869
3870 /* check again in case another CPU has just made room avail */
3871 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3872 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3873#endif
3874}
3875
3876/* Variant of netdev_tx_sent_queue() for drivers that are aware
3877 * that they should not test BQL status themselves.
3878 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3879 * skb of a batch.
3880 * Returns true if the doorbell must be used to kick the NIC.
3881 */
3882static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3883 unsigned int bytes,
3884 bool xmit_more)
3885{
3886 if (xmit_more) {
3887#ifdef CONFIG_BQL
3888 dql_queued(&dev_queue->dql, bytes);
3889#endif
3890 return netif_tx_queue_stopped(dev_queue);
3891 }
3892 netdev_tx_sent_queue(dev_queue, bytes);
3893 return true;
3894}
3895
3896/**
3897 * netdev_sent_queue - report the number of bytes queued to hardware
3898 * @dev: network device
3899 * @bytes: number of bytes queued to the hardware device queue
3900 *
3901 * Report the number of bytes queued for sending/completion to the network
3902 * device hardware queue#0. @bytes should be a good approximation and should
3903 * exactly match netdev_completed_queue() @bytes.
3904 * This is typically called once per packet, from ndo_start_xmit().
3905 */
3906static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3907{
3908 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3909}
3910
3911static inline bool __netdev_sent_queue(struct net_device *dev,
3912 unsigned int bytes,
3913 bool xmit_more)
3914{
3915 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3916 xmit_more);
3917}
3918
3919/**
3920 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3921 * @dev_queue: network device queue
3922 * @pkts: number of packets (currently ignored)
3923 * @bytes: number of bytes dequeued from the device queue
3924 *
3925 * Must be called at most once per TX completion round (and not per
3926 * individual packet), so that BQL can adjust its limits appropriately.
3927 */
3928static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3929 unsigned int pkts, unsigned int bytes)
3930{
3931#ifdef CONFIG_BQL
3932 if (unlikely(!bytes))
3933 return;
3934
3935 dql_completed(&dev_queue->dql, bytes);
3936
3937 /*
3938 * Without the memory barrier there is a small possibility that
3939 * netdev_tx_sent_queue will miss the update and cause the queue to
3940 * be stopped forever
3941 */
3942 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3943
3944 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3945 return;
3946
3947 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3948 netif_schedule_queue(dev_queue);
3949#endif
3950}
3951
3952/**
3953 * netdev_completed_queue - report bytes and packets completed by device
3954 * @dev: network device
3955 * @pkts: actual number of packets sent over the medium
3956 * @bytes: actual number of bytes sent over the medium
3957 *
3958 * Report the number of bytes and packets transmitted by the network device
3959 * hardware queue over the physical medium, @bytes must exactly match the
3960 * @bytes amount passed to netdev_sent_queue()
3961 */
3962static inline void netdev_completed_queue(struct net_device *dev,
3963 unsigned int pkts, unsigned int bytes)
3964{
3965 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3966}
3967
3968static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3969{
3970#ifdef CONFIG_BQL
3971 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3972 dql_reset(&q->dql);
3973#endif
3974}
3975
3976/**
3977 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3978 * @dev: network device
3979 * @qid: stack index of the queue to reset
3980 */
3981static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3982 u32 qid)
3983{
3984 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3985}
3986
3987/**
3988 * netdev_reset_queue - reset the packets and bytes count of a network device
3989 * @dev_queue: network device
3990 *
3991 * Reset the bytes and packet count of a network device and clear the
3992 * software flow control OFF bit for this network device
3993 */
3994static inline void netdev_reset_queue(struct net_device *dev_queue)
3995{
3996 netdev_tx_reset_subqueue(dev_queue, 0);
3997}
3998
3999/**
4000 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
4001 * @dev: network device
4002 * @queue_index: given tx queue index
4003 *
4004 * Returns 0 if given tx queue index >= number of device tx queues,
4005 * otherwise returns the originally passed tx queue index.
4006 */
4007static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
4008{
4009 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
4010 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
4011 dev->name, queue_index,
4012 dev->real_num_tx_queues);
4013 return 0;
4014 }
4015
4016 return queue_index;
4017}
4018
4019/**
4020 * netif_running - test if up
4021 * @dev: network device
4022 *
4023 * Test if the device has been brought up.
4024 */
4025static inline bool netif_running(const struct net_device *dev)
4026{
4027 return test_bit(__LINK_STATE_START, &dev->state);
4028}
4029
4030/*
4031 * Routines to manage the subqueues on a device. We only need start,
4032 * stop, and a check if it's stopped. All other device management is
4033 * done at the overall netdevice level.
4034 * Also test the device if we're multiqueue.
4035 */
4036
4037/**
4038 * netif_start_subqueue - allow sending packets on subqueue
4039 * @dev: network device
4040 * @queue_index: sub queue index
4041 *
4042 * Start individual transmit queue of a device with multiple transmit queues.
4043 */
4044static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
4045{
4046 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
4047
4048 netif_tx_start_queue(txq);
4049}
4050
4051/**
4052 * netif_stop_subqueue - stop sending packets on subqueue
4053 * @dev: network device
4054 * @queue_index: sub queue index
4055 *
4056 * Stop individual transmit queue of a device with multiple transmit queues.
4057 */
4058static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
4059{
4060 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
4061 netif_tx_stop_queue(txq);
4062}
4063
4064/**
4065 * __netif_subqueue_stopped - test status of subqueue
4066 * @dev: network device
4067 * @queue_index: sub queue index
4068 *
4069 * Check individual transmit queue of a device with multiple transmit queues.
4070 */
4071static inline bool __netif_subqueue_stopped(const struct net_device *dev,
4072 u16 queue_index)
4073{
4074 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
4075
4076 return netif_tx_queue_stopped(txq);
4077}
4078
4079/**
4080 * netif_subqueue_stopped - test status of subqueue
4081 * @dev: network device
4082 * @skb: sub queue buffer pointer
4083 *
4084 * Check individual transmit queue of a device with multiple transmit queues.
4085 */
4086static inline bool netif_subqueue_stopped(const struct net_device *dev,
4087 struct sk_buff *skb)
4088{
4089 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
4090}
4091
4092/**
4093 * netif_wake_subqueue - allow sending packets on subqueue
4094 * @dev: network device
4095 * @queue_index: sub queue index
4096 *
4097 * Resume individual transmit queue of a device with multiple transmit queues.
4098 */
4099static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
4100{
4101 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
4102
4103 netif_tx_wake_queue(txq);
4104}
4105
4106#ifdef CONFIG_XPS
4107int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
4108 u16 index);
4109int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
4110 u16 index, enum xps_map_type type);
4111
4112/**
4113 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
4114 * @j: CPU/Rx queue index
4115 * @mask: bitmask of all cpus/rx queues
4116 * @nr_bits: number of bits in the bitmask
4117 *
4118 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
4119 */
4120static inline bool netif_attr_test_mask(unsigned long j,
4121 const unsigned long *mask,
4122 unsigned int nr_bits)
4123{
4124 cpu_max_bits_warn(j, nr_bits);
4125 return test_bit(j, mask);
4126}
4127
4128/**
4129 * netif_attr_test_online - Test for online CPU/Rx queue
4130 * @j: CPU/Rx queue index
4131 * @online_mask: bitmask for CPUs/Rx queues that are online
4132 * @nr_bits: number of bits in the bitmask
4133 *
4134 * Returns: true if a CPU/Rx queue is online.
4135 */
4136static inline bool netif_attr_test_online(unsigned long j,
4137 const unsigned long *online_mask,
4138 unsigned int nr_bits)
4139{
4140 cpu_max_bits_warn(j, nr_bits);
4141
4142 if (online_mask)
4143 return test_bit(j, online_mask);
4144
4145 return (j < nr_bits);
4146}
4147
4148/**
4149 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
4150 * @n: CPU/Rx queue index
4151 * @srcp: the cpumask/Rx queue mask pointer
4152 * @nr_bits: number of bits in the bitmask
4153 *
4154 * Returns: next (after n) CPU/Rx queue index in the mask;
4155 * >= nr_bits if no further CPUs/Rx queues set.
4156 */
4157static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
4158 unsigned int nr_bits)
4159{
4160 /* -1 is a legal arg here. */
4161 if (n != -1)
4162 cpu_max_bits_warn(n, nr_bits);
4163
4164 if (srcp)
4165 return find_next_bit(srcp, nr_bits, n + 1);
4166
4167 return n + 1;
4168}
4169
4170/**
4171 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
4172 * @n: CPU/Rx queue index
4173 * @src1p: the first CPUs/Rx queues mask pointer
4174 * @src2p: the second CPUs/Rx queues mask pointer
4175 * @nr_bits: number of bits in the bitmask
4176 *
4177 * Returns: next (after n) CPU/Rx queue index set in both masks;
4178 * >= nr_bits if no further CPUs/Rx queues set in both.
4179 */
4180static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
4181 const unsigned long *src2p,
4182 unsigned int nr_bits)
4183{
4184 /* -1 is a legal arg here. */
4185 if (n != -1)
4186 cpu_max_bits_warn(n, nr_bits);
4187
4188 if (src1p && src2p)
4189 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
4190 else if (src1p)
4191 return find_next_bit(src1p, nr_bits, n + 1);
4192 else if (src2p)
4193 return find_next_bit(src2p, nr_bits, n + 1);
4194
4195 return n + 1;
4196}
4197#else
4198static inline int netif_set_xps_queue(struct net_device *dev,
4199 const struct cpumask *mask,
4200 u16 index)
4201{
4202 return 0;
4203}
4204
4205static inline int __netif_set_xps_queue(struct net_device *dev,
4206 const unsigned long *mask,
4207 u16 index, enum xps_map_type type)
4208{
4209 return 0;
4210}
4211#endif
4212
4213/**
4214 * netif_is_multiqueue - test if device has multiple transmit queues
4215 * @dev: network device
4216 *
4217 * Check if device has multiple transmit queues
4218 */
4219static inline bool netif_is_multiqueue(const struct net_device *dev)
4220{
4221 return dev->num_tx_queues > 1;
4222}
4223
4224int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
4225int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
4226int netif_set_real_num_queues(struct net_device *dev,
4227 unsigned int txq, unsigned int rxq);
4228
4229int netif_get_num_default_rss_queues(void);
4230
4231void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
4232void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
4233
4234/*
4235 * It is not allowed to call kfree_skb() or consume_skb() from hardware
4236 * interrupt context or with hardware interrupts being disabled.
4237 * (in_hardirq() || irqs_disabled())
4238 *
4239 * We provide four helpers that can be used in following contexts :
4240 *
4241 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
4242 * replacing kfree_skb(skb)
4243 *
4244 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
4245 * Typically used in place of consume_skb(skb) in TX completion path
4246 *
4247 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
4248 * replacing kfree_skb(skb)
4249 *
4250 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
4251 * and consumed a packet. Used in place of consume_skb(skb)
4252 */
4253static inline void dev_kfree_skb_irq(struct sk_buff *skb)
4254{
4255 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
4256}
4257
4258static inline void dev_consume_skb_irq(struct sk_buff *skb)
4259{
4260 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
4261}
4262
4263static inline void dev_kfree_skb_any(struct sk_buff *skb)
4264{
4265 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
4266}
4267
4268static inline void dev_consume_skb_any(struct sk_buff *skb)
4269{
4270 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
4271}
4272
4273u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
4274 const struct bpf_prog *xdp_prog);
4275void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog);
4276int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb);
4277int netif_rx(struct sk_buff *skb);
4278int __netif_rx(struct sk_buff *skb);
4279
4280int netif_receive_skb(struct sk_buff *skb);
4281int netif_receive_skb_core(struct sk_buff *skb);
4282void netif_receive_skb_list_internal(struct list_head *head);
4283void netif_receive_skb_list(struct list_head *head);
4284gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb);
4285
4286static inline gro_result_t napi_gro_receive(struct napi_struct *napi,
4287 struct sk_buff *skb)
4288{
4289 return gro_receive_skb(&napi->gro, skb);
4290}
4291
4292struct sk_buff *napi_get_frags(struct napi_struct *napi);
4293gro_result_t napi_gro_frags(struct napi_struct *napi);
4294
4295static inline void napi_free_frags(struct napi_struct *napi)
4296{
4297 kfree_skb(napi->skb);
4298 napi->skb = NULL;
4299}
4300
4301bool netdev_is_rx_handler_busy(struct net_device *dev);
4302int netdev_rx_handler_register(struct net_device *dev,
4303 rx_handler_func_t *rx_handler,
4304 void *rx_handler_data);
4305void netdev_rx_handler_unregister(struct net_device *dev);
4306
4307bool dev_valid_name(const char *name);
4308static inline bool is_socket_ioctl_cmd(unsigned int cmd)
4309{
4310 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4311}
4312int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4313int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4314int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4315 void __user *data, bool *need_copyout);
4316int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4317int dev_eth_ioctl(struct net_device *dev,
4318 struct ifreq *ifr, unsigned int cmd);
4319int generic_hwtstamp_get_lower(struct net_device *dev,
4320 struct kernel_hwtstamp_config *kernel_cfg);
4321int generic_hwtstamp_set_lower(struct net_device *dev,
4322 struct kernel_hwtstamp_config *kernel_cfg,
4323 struct netlink_ext_ack *extack);
4324int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4325unsigned int netif_get_flags(const struct net_device *dev);
4326int __dev_change_flags(struct net_device *dev, unsigned int flags,
4327 struct netlink_ext_ack *extack);
4328int netif_change_flags(struct net_device *dev, unsigned int flags,
4329 struct netlink_ext_ack *extack);
4330int dev_change_flags(struct net_device *dev, unsigned int flags,
4331 struct netlink_ext_ack *extack);
4332int netif_set_alias(struct net_device *dev, const char *alias, size_t len);
4333int dev_set_alias(struct net_device *, const char *, size_t);
4334int dev_get_alias(const struct net_device *, char *, size_t);
4335int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4336 const char *pat, int new_ifindex,
4337 struct netlink_ext_ack *extack);
4338int dev_change_net_namespace(struct net_device *dev, struct net *net,
4339 const char *pat);
4340int __netif_set_mtu(struct net_device *dev, int new_mtu);
4341int netif_set_mtu(struct net_device *dev, int new_mtu);
4342int dev_set_mtu(struct net_device *, int);
4343int netif_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4344 struct netlink_ext_ack *extack);
4345int netif_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss,
4346 struct netlink_ext_ack *extack);
4347int dev_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss,
4348 struct netlink_ext_ack *extack);
4349int dev_set_mac_address_user(struct net_device *dev, struct sockaddr_storage *ss,
4350 struct netlink_ext_ack *extack);
4351int netif_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4352int netif_get_port_parent_id(struct net_device *dev,
4353 struct netdev_phys_item_id *ppid, bool recurse);
4354bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4355
4356struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4357struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4358 struct netdev_queue *txq, int *ret);
4359
4360int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4361u8 dev_xdp_prog_count(struct net_device *dev);
4362int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
4363int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
4364u8 dev_xdp_sb_prog_count(struct net_device *dev);
4365u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4366
4367u32 dev_get_min_mp_channel_count(const struct net_device *dev);
4368
4369int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4370int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4371int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4372bool is_skb_forwardable(const struct net_device *dev,
4373 const struct sk_buff *skb);
4374
4375static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4376 const struct sk_buff *skb,
4377 const bool check_mtu)
4378{
4379 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4380 unsigned int len;
4381
4382 if (!(dev->flags & IFF_UP))
4383 return false;
4384
4385 if (!check_mtu)
4386 return true;
4387
4388 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4389 if (skb->len <= len)
4390 return true;
4391
4392 /* if TSO is enabled, we don't care about the length as the packet
4393 * could be forwarded without being segmented before
4394 */
4395 if (skb_is_gso(skb))
4396 return true;
4397
4398 return false;
4399}
4400
4401void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4402
4403#define DEV_CORE_STATS_INC(FIELD) \
4404static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4405{ \
4406 netdev_core_stats_inc(dev, \
4407 offsetof(struct net_device_core_stats, FIELD)); \
4408}
4409DEV_CORE_STATS_INC(rx_dropped)
4410DEV_CORE_STATS_INC(tx_dropped)
4411DEV_CORE_STATS_INC(rx_nohandler)
4412DEV_CORE_STATS_INC(rx_otherhost_dropped)
4413#undef DEV_CORE_STATS_INC
4414
4415static __always_inline int ____dev_forward_skb(struct net_device *dev,
4416 struct sk_buff *skb,
4417 const bool check_mtu)
4418{
4419 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4420 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4421 dev_core_stats_rx_dropped_inc(dev);
4422 kfree_skb(skb);
4423 return NET_RX_DROP;
4424 }
4425
4426 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4427 skb->priority = 0;
4428 return 0;
4429}
4430
4431bool dev_nit_active_rcu(const struct net_device *dev);
4432static inline bool dev_nit_active(const struct net_device *dev)
4433{
4434 bool ret;
4435
4436 rcu_read_lock();
4437 ret = dev_nit_active_rcu(dev);
4438 rcu_read_unlock();
4439 return ret;
4440}
4441
4442void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4443
4444static inline void __dev_put(struct net_device *dev)
4445{
4446 if (dev) {
4447#ifdef CONFIG_PCPU_DEV_REFCNT
4448 this_cpu_dec(*dev->pcpu_refcnt);
4449#else
4450 refcount_dec(&dev->dev_refcnt);
4451#endif
4452 }
4453}
4454
4455static inline void __dev_hold(struct net_device *dev)
4456{
4457 if (dev) {
4458#ifdef CONFIG_PCPU_DEV_REFCNT
4459 this_cpu_inc(*dev->pcpu_refcnt);
4460#else
4461 refcount_inc(&dev->dev_refcnt);
4462#endif
4463 }
4464}
4465
4466static inline void __netdev_tracker_alloc(struct net_device *dev,
4467 netdevice_tracker *tracker,
4468 gfp_t gfp)
4469{
4470#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4471 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4472#endif
4473}
4474
4475/* netdev_tracker_alloc() can upgrade a prior untracked reference
4476 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4477 */
4478static inline void netdev_tracker_alloc(struct net_device *dev,
4479 netdevice_tracker *tracker, gfp_t gfp)
4480{
4481#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4482 refcount_dec(&dev->refcnt_tracker.no_tracker);
4483 __netdev_tracker_alloc(dev, tracker, gfp);
4484#endif
4485}
4486
4487static inline void netdev_tracker_free(struct net_device *dev,
4488 netdevice_tracker *tracker)
4489{
4490#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4491 ref_tracker_free(&dev->refcnt_tracker, tracker);
4492#endif
4493}
4494
4495static inline void netdev_hold(struct net_device *dev,
4496 netdevice_tracker *tracker, gfp_t gfp)
4497{
4498 if (dev) {
4499 __dev_hold(dev);
4500 __netdev_tracker_alloc(dev, tracker, gfp);
4501 }
4502}
4503
4504static inline void netdev_put(struct net_device *dev,
4505 netdevice_tracker *tracker)
4506{
4507 if (dev) {
4508 netdev_tracker_free(dev, tracker);
4509 __dev_put(dev);
4510 }
4511}
4512
4513/**
4514 * dev_hold - get reference to device
4515 * @dev: network device
4516 *
4517 * Hold reference to device to keep it from being freed.
4518 * Try using netdev_hold() instead.
4519 */
4520static inline void dev_hold(struct net_device *dev)
4521{
4522 netdev_hold(dev, NULL, GFP_ATOMIC);
4523}
4524
4525/**
4526 * dev_put - release reference to device
4527 * @dev: network device
4528 *
4529 * Release reference to device to allow it to be freed.
4530 * Try using netdev_put() instead.
4531 */
4532static inline void dev_put(struct net_device *dev)
4533{
4534 netdev_put(dev, NULL);
4535}
4536
4537DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4538
4539static inline void netdev_ref_replace(struct net_device *odev,
4540 struct net_device *ndev,
4541 netdevice_tracker *tracker,
4542 gfp_t gfp)
4543{
4544 if (odev)
4545 netdev_tracker_free(odev, tracker);
4546
4547 __dev_hold(ndev);
4548 __dev_put(odev);
4549
4550 if (ndev)
4551 __netdev_tracker_alloc(ndev, tracker, gfp);
4552}
4553
4554/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4555 * and _off may be called from IRQ context, but it is caller
4556 * who is responsible for serialization of these calls.
4557 *
4558 * The name carrier is inappropriate, these functions should really be
4559 * called netif_lowerlayer_*() because they represent the state of any
4560 * kind of lower layer not just hardware media.
4561 */
4562void linkwatch_fire_event(struct net_device *dev);
4563
4564/**
4565 * linkwatch_sync_dev - sync linkwatch for the given device
4566 * @dev: network device to sync linkwatch for
4567 *
4568 * Sync linkwatch for the given device, removing it from the
4569 * pending work list (if queued).
4570 */
4571void linkwatch_sync_dev(struct net_device *dev);
4572void __linkwatch_sync_dev(struct net_device *dev);
4573
4574/**
4575 * netif_carrier_ok - test if carrier present
4576 * @dev: network device
4577 *
4578 * Check if carrier is present on device
4579 */
4580static inline bool netif_carrier_ok(const struct net_device *dev)
4581{
4582 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4583}
4584
4585unsigned long dev_trans_start(struct net_device *dev);
4586
4587void netdev_watchdog_up(struct net_device *dev);
4588
4589void netif_carrier_on(struct net_device *dev);
4590void netif_carrier_off(struct net_device *dev);
4591void netif_carrier_event(struct net_device *dev);
4592
4593/**
4594 * netif_dormant_on - mark device as dormant.
4595 * @dev: network device
4596 *
4597 * Mark device as dormant (as per RFC2863).
4598 *
4599 * The dormant state indicates that the relevant interface is not
4600 * actually in a condition to pass packets (i.e., it is not 'up') but is
4601 * in a "pending" state, waiting for some external event. For "on-
4602 * demand" interfaces, this new state identifies the situation where the
4603 * interface is waiting for events to place it in the up state.
4604 */
4605static inline void netif_dormant_on(struct net_device *dev)
4606{
4607 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4608 linkwatch_fire_event(dev);
4609}
4610
4611/**
4612 * netif_dormant_off - set device as not dormant.
4613 * @dev: network device
4614 *
4615 * Device is not in dormant state.
4616 */
4617static inline void netif_dormant_off(struct net_device *dev)
4618{
4619 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4620 linkwatch_fire_event(dev);
4621}
4622
4623/**
4624 * netif_dormant - test if device is dormant
4625 * @dev: network device
4626 *
4627 * Check if device is dormant.
4628 */
4629static inline bool netif_dormant(const struct net_device *dev)
4630{
4631 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4632}
4633
4634
4635/**
4636 * netif_testing_on - mark device as under test.
4637 * @dev: network device
4638 *
4639 * Mark device as under test (as per RFC2863).
4640 *
4641 * The testing state indicates that some test(s) must be performed on
4642 * the interface. After completion, of the test, the interface state
4643 * will change to up, dormant, or down, as appropriate.
4644 */
4645static inline void netif_testing_on(struct net_device *dev)
4646{
4647 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4648 linkwatch_fire_event(dev);
4649}
4650
4651/**
4652 * netif_testing_off - set device as not under test.
4653 * @dev: network device
4654 *
4655 * Device is not in testing state.
4656 */
4657static inline void netif_testing_off(struct net_device *dev)
4658{
4659 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4660 linkwatch_fire_event(dev);
4661}
4662
4663/**
4664 * netif_testing - test if device is under test
4665 * @dev: network device
4666 *
4667 * Check if device is under test
4668 */
4669static inline bool netif_testing(const struct net_device *dev)
4670{
4671 return test_bit(__LINK_STATE_TESTING, &dev->state);
4672}
4673
4674
4675/**
4676 * netif_oper_up - test if device is operational
4677 * @dev: network device
4678 *
4679 * Check if carrier is operational
4680 */
4681static inline bool netif_oper_up(const struct net_device *dev)
4682{
4683 unsigned int operstate = READ_ONCE(dev->operstate);
4684
4685 return operstate == IF_OPER_UP ||
4686 operstate == IF_OPER_UNKNOWN /* backward compat */;
4687}
4688
4689/**
4690 * netif_device_present - is device available or removed
4691 * @dev: network device
4692 *
4693 * Check if device has not been removed from system.
4694 */
4695static inline bool netif_device_present(const struct net_device *dev)
4696{
4697 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4698}
4699
4700void netif_device_detach(struct net_device *dev);
4701
4702void netif_device_attach(struct net_device *dev);
4703
4704/*
4705 * Network interface message level settings
4706 */
4707
4708enum {
4709 NETIF_MSG_DRV_BIT,
4710 NETIF_MSG_PROBE_BIT,
4711 NETIF_MSG_LINK_BIT,
4712 NETIF_MSG_TIMER_BIT,
4713 NETIF_MSG_IFDOWN_BIT,
4714 NETIF_MSG_IFUP_BIT,
4715 NETIF_MSG_RX_ERR_BIT,
4716 NETIF_MSG_TX_ERR_BIT,
4717 NETIF_MSG_TX_QUEUED_BIT,
4718 NETIF_MSG_INTR_BIT,
4719 NETIF_MSG_TX_DONE_BIT,
4720 NETIF_MSG_RX_STATUS_BIT,
4721 NETIF_MSG_PKTDATA_BIT,
4722 NETIF_MSG_HW_BIT,
4723 NETIF_MSG_WOL_BIT,
4724
4725 /* When you add a new bit above, update netif_msg_class_names array
4726 * in net/ethtool/common.c
4727 */
4728 NETIF_MSG_CLASS_COUNT,
4729};
4730/* Both ethtool_ops interface and internal driver implementation use u32 */
4731static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4732
4733#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4734#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4735
4736#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4737#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4738#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4739#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4740#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4741#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4742#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4743#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4744#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4745#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4746#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4747#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4748#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4749#define NETIF_MSG_HW __NETIF_MSG(HW)
4750#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4751
4752#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4753#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4754#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4755#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4756#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4757#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4758#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4759#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4760#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4761#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4762#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4763#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4764#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4765#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4766#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4767
4768static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4769{
4770 /* use default */
4771 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4772 return default_msg_enable_bits;
4773 if (debug_value == 0) /* no output */
4774 return 0;
4775 /* set low N bits */
4776 return (1U << debug_value) - 1;
4777}
4778
4779static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4780{
4781 spin_lock(&txq->_xmit_lock);
4782 /* Pairs with READ_ONCE() in netif_tx_owned() */
4783 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4784}
4785
4786static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4787{
4788 __acquire(&txq->_xmit_lock);
4789 return true;
4790}
4791
4792static inline void __netif_tx_release(struct netdev_queue *txq)
4793{
4794 __release(&txq->_xmit_lock);
4795}
4796
4797static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4798{
4799 spin_lock_bh(&txq->_xmit_lock);
4800 /* Pairs with READ_ONCE() in netif_tx_owned() */
4801 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4802}
4803
4804static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4805{
4806 bool ok = spin_trylock(&txq->_xmit_lock);
4807
4808 if (likely(ok)) {
4809 /* Pairs with READ_ONCE() in netif_tx_owned() */
4810 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4811 }
4812 return ok;
4813}
4814
4815static inline void __netif_tx_unlock(struct netdev_queue *txq)
4816{
4817 /* Pairs with READ_ONCE() in netif_tx_owned() */
4818 WRITE_ONCE(txq->xmit_lock_owner, -1);
4819 spin_unlock(&txq->_xmit_lock);
4820}
4821
4822static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4823{
4824 /* Pairs with READ_ONCE() in netif_tx_owned() */
4825 WRITE_ONCE(txq->xmit_lock_owner, -1);
4826 spin_unlock_bh(&txq->_xmit_lock);
4827}
4828
4829/*
4830 * txq->trans_start can be read locklessly from dev_watchdog()
4831 */
4832static inline void txq_trans_update(const struct net_device *dev,
4833 struct netdev_queue *txq)
4834{
4835 if (!dev->lltx)
4836 WRITE_ONCE(txq->trans_start, jiffies);
4837}
4838
4839static inline void txq_trans_cond_update(struct netdev_queue *txq)
4840{
4841 unsigned long now = jiffies;
4842
4843 if (READ_ONCE(txq->trans_start) != now)
4844 WRITE_ONCE(txq->trans_start, now);
4845}
4846
4847/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4848static inline void netif_trans_update(struct net_device *dev)
4849{
4850 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4851
4852 txq_trans_cond_update(txq);
4853}
4854
4855/**
4856 * netif_tx_lock - grab network device transmit lock
4857 * @dev: network device
4858 *
4859 * Get network device transmit lock
4860 */
4861void netif_tx_lock(struct net_device *dev);
4862
4863static inline void netif_tx_lock_bh(struct net_device *dev)
4864{
4865 local_bh_disable();
4866 netif_tx_lock(dev);
4867}
4868
4869void netif_tx_unlock(struct net_device *dev);
4870
4871static inline void netif_tx_unlock_bh(struct net_device *dev)
4872{
4873 netif_tx_unlock(dev);
4874 local_bh_enable();
4875}
4876
4877#define HARD_TX_LOCK(dev, txq, cpu) { \
4878 if (!(dev)->lltx) { \
4879 __netif_tx_lock(txq, cpu); \
4880 } else { \
4881 __netif_tx_acquire(txq); \
4882 } \
4883}
4884
4885#define HARD_TX_TRYLOCK(dev, txq) \
4886 (!(dev)->lltx ? \
4887 __netif_tx_trylock(txq) : \
4888 __netif_tx_acquire(txq))
4889
4890#define HARD_TX_UNLOCK(dev, txq) { \
4891 if (!(dev)->lltx) { \
4892 __netif_tx_unlock(txq); \
4893 } else { \
4894 __netif_tx_release(txq); \
4895 } \
4896}
4897
4898static inline void netif_tx_disable(struct net_device *dev)
4899{
4900 unsigned int i;
4901 int cpu;
4902
4903 local_bh_disable();
4904 cpu = smp_processor_id();
4905 spin_lock(&dev->tx_global_lock);
4906 for (i = 0; i < dev->num_tx_queues; i++) {
4907 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4908
4909 __netif_tx_lock(txq, cpu);
4910 netif_tx_stop_queue(txq);
4911 __netif_tx_unlock(txq);
4912 }
4913 spin_unlock(&dev->tx_global_lock);
4914 local_bh_enable();
4915}
4916
4917#ifndef CONFIG_PREEMPT_RT
4918static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu)
4919{
4920 /* Other cpus might concurrently change txq->xmit_lock_owner
4921 * to -1 or to their cpu id, but not to our id.
4922 */
4923 return READ_ONCE(txq->xmit_lock_owner) == cpu;
4924}
4925
4926#else
4927static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu)
4928{
4929 return rt_mutex_owner(&txq->_xmit_lock.lock) == current;
4930}
4931
4932#endif
4933
4934static inline void netif_addr_lock(struct net_device *dev)
4935{
4936 unsigned char nest_level = 0;
4937
4938#ifdef CONFIG_LOCKDEP
4939 nest_level = dev->nested_level;
4940#endif
4941 spin_lock_nested(&dev->addr_list_lock, nest_level);
4942}
4943
4944static inline void netif_addr_lock_bh(struct net_device *dev)
4945{
4946 unsigned char nest_level = 0;
4947
4948#ifdef CONFIG_LOCKDEP
4949 nest_level = dev->nested_level;
4950#endif
4951 local_bh_disable();
4952 spin_lock_nested(&dev->addr_list_lock, nest_level);
4953}
4954
4955static inline void netif_addr_unlock(struct net_device *dev)
4956{
4957 spin_unlock(&dev->addr_list_lock);
4958}
4959
4960static inline void netif_addr_unlock_bh(struct net_device *dev)
4961{
4962 spin_unlock_bh(&dev->addr_list_lock);
4963}
4964
4965/*
4966 * dev_addrs walker. Should be used only for read access. Call with
4967 * rcu_read_lock held.
4968 */
4969#define for_each_dev_addr(dev, ha) \
4970 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4971
4972/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4973
4974void ether_setup(struct net_device *dev);
4975
4976/* Allocate dummy net_device */
4977struct net_device *alloc_netdev_dummy(int sizeof_priv);
4978
4979/* Support for loadable net-drivers */
4980struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4981 unsigned char name_assign_type,
4982 void (*setup)(struct net_device *),
4983 unsigned int txqs, unsigned int rxqs);
4984#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4985 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4986
4987#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4988 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4989 count)
4990
4991int register_netdev(struct net_device *dev);
4992void unregister_netdev(struct net_device *dev);
4993
4994int devm_register_netdev(struct device *dev, struct net_device *ndev);
4995
4996/* General hardware address lists handling functions */
4997int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4998 struct netdev_hw_addr_list *from_list, int addr_len);
4999int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list,
5000 struct netdev_hw_addr_list *from_list,
5001 int addr_len);
5002void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
5003 struct netdev_hw_addr_list *from_list, int addr_len);
5004int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
5005 struct net_device *dev,
5006 int (*sync)(struct net_device *, const unsigned char *),
5007 int (*unsync)(struct net_device *,
5008 const unsigned char *));
5009int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
5010 struct net_device *dev,
5011 int (*sync)(struct net_device *,
5012 const unsigned char *, int),
5013 int (*unsync)(struct net_device *,
5014 const unsigned char *, int));
5015void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
5016 struct net_device *dev,
5017 int (*unsync)(struct net_device *,
5018 const unsigned char *, int));
5019void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
5020 struct net_device *dev,
5021 int (*unsync)(struct net_device *,
5022 const unsigned char *));
5023void __hw_addr_init(struct netdev_hw_addr_list *list);
5024void __hw_addr_flush(struct netdev_hw_addr_list *list);
5025int __hw_addr_list_snapshot(struct netdev_hw_addr_list *snap,
5026 const struct netdev_hw_addr_list *list,
5027 int addr_len, struct netdev_hw_addr_list *cache);
5028void __hw_addr_list_reconcile(struct netdev_hw_addr_list *real_list,
5029 struct netdev_hw_addr_list *work,
5030 struct netdev_hw_addr_list *ref, int addr_len,
5031 struct netdev_hw_addr_list *cache);
5032
5033/* Functions used for device addresses handling */
5034void dev_addr_mod(struct net_device *dev, unsigned int offset,
5035 const void *addr, size_t len);
5036
5037static inline void
5038__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
5039{
5040 dev_addr_mod(dev, 0, addr, len);
5041}
5042
5043static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
5044{
5045 __dev_addr_set(dev, addr, dev->addr_len);
5046}
5047
5048int dev_addr_add(struct net_device *dev, const unsigned char *addr,
5049 unsigned char addr_type);
5050int dev_addr_del(struct net_device *dev, const unsigned char *addr,
5051 unsigned char addr_type);
5052
5053/* Functions used for unicast addresses handling */
5054int dev_uc_add(struct net_device *dev, const unsigned char *addr);
5055int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
5056int dev_uc_del(struct net_device *dev, const unsigned char *addr);
5057int dev_uc_sync(struct net_device *to, struct net_device *from);
5058int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
5059void dev_uc_unsync(struct net_device *to, struct net_device *from);
5060void dev_uc_flush(struct net_device *dev);
5061void dev_uc_init(struct net_device *dev);
5062
5063/**
5064 * __dev_uc_sync - Synchronize device's unicast list
5065 * @dev: device to sync
5066 * @sync: function to call if address should be added
5067 * @unsync: function to call if address should be removed
5068 *
5069 * Add newly added addresses to the interface, and release
5070 * addresses that have been deleted.
5071 */
5072static inline int __dev_uc_sync(struct net_device *dev,
5073 int (*sync)(struct net_device *,
5074 const unsigned char *),
5075 int (*unsync)(struct net_device *,
5076 const unsigned char *))
5077{
5078 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
5079}
5080
5081/**
5082 * __dev_uc_unsync - Remove synchronized addresses from device
5083 * @dev: device to sync
5084 * @unsync: function to call if address should be removed
5085 *
5086 * Remove all addresses that were added to the device by dev_uc_sync().
5087 */
5088static inline void __dev_uc_unsync(struct net_device *dev,
5089 int (*unsync)(struct net_device *,
5090 const unsigned char *))
5091{
5092 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
5093}
5094
5095/* Functions used for multicast addresses handling */
5096int dev_mc_add(struct net_device *dev, const unsigned char *addr);
5097int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
5098int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
5099int dev_mc_del(struct net_device *dev, const unsigned char *addr);
5100int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
5101int dev_mc_sync(struct net_device *to, struct net_device *from);
5102int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
5103void dev_mc_unsync(struct net_device *to, struct net_device *from);
5104void dev_mc_flush(struct net_device *dev);
5105void dev_mc_init(struct net_device *dev);
5106
5107/**
5108 * __dev_mc_sync - Synchronize device's multicast list
5109 * @dev: device to sync
5110 * @sync: function to call if address should be added
5111 * @unsync: function to call if address should be removed
5112 *
5113 * Add newly added addresses to the interface, and release
5114 * addresses that have been deleted.
5115 */
5116static inline int __dev_mc_sync(struct net_device *dev,
5117 int (*sync)(struct net_device *,
5118 const unsigned char *),
5119 int (*unsync)(struct net_device *,
5120 const unsigned char *))
5121{
5122 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
5123}
5124
5125/**
5126 * __dev_mc_unsync - Remove synchronized addresses from device
5127 * @dev: device to sync
5128 * @unsync: function to call if address should be removed
5129 *
5130 * Remove all addresses that were added to the device by dev_mc_sync().
5131 */
5132static inline void __dev_mc_unsync(struct net_device *dev,
5133 int (*unsync)(struct net_device *,
5134 const unsigned char *))
5135{
5136 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
5137}
5138
5139/* Functions used for secondary unicast and multicast support */
5140void dev_set_rx_mode(struct net_device *dev);
5141int netif_set_promiscuity(struct net_device *dev, int inc);
5142int dev_set_promiscuity(struct net_device *dev, int inc);
5143int netif_set_allmulti(struct net_device *dev, int inc, bool notify);
5144int dev_set_allmulti(struct net_device *dev, int inc);
5145void netif_state_change(struct net_device *dev);
5146void netdev_state_change(struct net_device *dev);
5147void __netdev_notify_peers(struct net_device *dev);
5148void netdev_notify_peers(struct net_device *dev);
5149void netdev_features_change(struct net_device *dev);
5150/* Load a device via the kmod */
5151void dev_load(struct net *net, const char *name);
5152struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
5153 struct rtnl_link_stats64 *storage);
5154void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
5155 const struct net_device_stats *netdev_stats);
5156void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
5157 const struct pcpu_sw_netstats __percpu *netstats);
5158void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
5159
5160enum {
5161 NESTED_SYNC_IMM_BIT,
5162 NESTED_SYNC_TODO_BIT,
5163};
5164
5165#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
5166#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
5167
5168#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
5169#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
5170
5171struct netdev_nested_priv {
5172 unsigned char flags;
5173 void *data;
5174};
5175
5176bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
5177struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
5178 struct list_head **iter);
5179
5180/* iterate through upper list, must be called under RCU read lock */
5181#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
5182 for (iter = &(dev)->adj_list.upper, \
5183 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
5184 updev; \
5185 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
5186
5187int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
5188 int (*fn)(struct net_device *upper_dev,
5189 struct netdev_nested_priv *priv),
5190 struct netdev_nested_priv *priv);
5191
5192bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
5193 struct net_device *upper_dev);
5194
5195bool netdev_has_any_upper_dev(struct net_device *dev);
5196
5197void *netdev_lower_get_next_private(struct net_device *dev,
5198 struct list_head **iter);
5199void *netdev_lower_get_next_private_rcu(struct net_device *dev,
5200 struct list_head **iter);
5201
5202#define netdev_for_each_lower_private(dev, priv, iter) \
5203 for (iter = (dev)->adj_list.lower.next, \
5204 priv = netdev_lower_get_next_private(dev, &(iter)); \
5205 priv; \
5206 priv = netdev_lower_get_next_private(dev, &(iter)))
5207
5208#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
5209 for (iter = &(dev)->adj_list.lower, \
5210 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
5211 priv; \
5212 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
5213
5214void *netdev_lower_get_next(struct net_device *dev,
5215 struct list_head **iter);
5216
5217#define netdev_for_each_lower_dev(dev, ldev, iter) \
5218 for (iter = (dev)->adj_list.lower.next, \
5219 ldev = netdev_lower_get_next(dev, &(iter)); \
5220 ldev; \
5221 ldev = netdev_lower_get_next(dev, &(iter)))
5222
5223struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
5224 struct list_head **iter);
5225int netdev_walk_all_lower_dev(struct net_device *dev,
5226 int (*fn)(struct net_device *lower_dev,
5227 struct netdev_nested_priv *priv),
5228 struct netdev_nested_priv *priv);
5229int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
5230 int (*fn)(struct net_device *lower_dev,
5231 struct netdev_nested_priv *priv),
5232 struct netdev_nested_priv *priv);
5233
5234void *netdev_adjacent_get_private(struct list_head *adj_list);
5235void *netdev_lower_get_first_private_rcu(struct net_device *dev);
5236struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
5237struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
5238int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
5239 struct netlink_ext_ack *extack);
5240int netdev_master_upper_dev_link(struct net_device *dev,
5241 struct net_device *upper_dev,
5242 void *upper_priv, void *upper_info,
5243 struct netlink_ext_ack *extack);
5244void netdev_upper_dev_unlink(struct net_device *dev,
5245 struct net_device *upper_dev);
5246int netdev_adjacent_change_prepare(struct net_device *old_dev,
5247 struct net_device *new_dev,
5248 struct net_device *dev,
5249 struct netlink_ext_ack *extack);
5250void netdev_adjacent_change_commit(struct net_device *old_dev,
5251 struct net_device *new_dev,
5252 struct net_device *dev);
5253void netdev_adjacent_change_abort(struct net_device *old_dev,
5254 struct net_device *new_dev,
5255 struct net_device *dev);
5256void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
5257void *netdev_lower_dev_get_private(struct net_device *dev,
5258 struct net_device *lower_dev);
5259void netdev_lower_state_changed(struct net_device *lower_dev,
5260 void *lower_state_info);
5261
5262#define NETDEV_RSS_KEY_LEN 256
5263extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
5264void netdev_rss_key_fill(void *buffer, size_t len);
5265
5266int skb_checksum_help(struct sk_buff *skb);
5267int skb_crc32c_csum_help(struct sk_buff *skb);
5268int skb_csum_hwoffload_help(struct sk_buff *skb,
5269 const netdev_features_t features);
5270
5271struct netdev_bonding_info {
5272 ifslave slave;
5273 ifbond master;
5274};
5275
5276struct netdev_notifier_bonding_info {
5277 struct netdev_notifier_info info; /* must be first */
5278 struct netdev_bonding_info bonding_info;
5279};
5280
5281void netdev_bonding_info_change(struct net_device *dev,
5282 struct netdev_bonding_info *bonding_info);
5283
5284#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
5285void ethtool_notify(struct net_device *dev, unsigned int cmd);
5286#else
5287static inline void ethtool_notify(struct net_device *dev, unsigned int cmd)
5288{
5289}
5290#endif
5291
5292__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
5293
5294static inline bool can_checksum_protocol(netdev_features_t features,
5295 __be16 protocol)
5296{
5297 if (protocol == htons(ETH_P_FCOE))
5298 return !!(features & NETIF_F_FCOE_CRC);
5299
5300 /* Assume this is an IP checksum (not SCTP CRC) */
5301
5302 if (features & NETIF_F_HW_CSUM) {
5303 /* Can checksum everything */
5304 return true;
5305 }
5306
5307 switch (protocol) {
5308 case htons(ETH_P_IP):
5309 return !!(features & NETIF_F_IP_CSUM);
5310 case htons(ETH_P_IPV6):
5311 return !!(features & NETIF_F_IPV6_CSUM);
5312 default:
5313 return false;
5314 }
5315}
5316
5317#ifdef CONFIG_BUG
5318void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
5319#else
5320static inline void netdev_rx_csum_fault(struct net_device *dev,
5321 struct sk_buff *skb)
5322{
5323}
5324#endif
5325/* rx skb timestamps */
5326void net_enable_timestamp(void);
5327void net_disable_timestamp(void);
5328
5329static inline ktime_t netdev_get_tstamp(struct net_device *dev,
5330 const struct skb_shared_hwtstamps *hwtstamps,
5331 bool cycles)
5332{
5333 const struct net_device_ops *ops = dev->netdev_ops;
5334
5335 if (ops->ndo_get_tstamp)
5336 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
5337
5338 return hwtstamps->hwtstamp;
5339}
5340
5341#ifndef CONFIG_PREEMPT_RT
5342static inline void netdev_xmit_set_more(bool more)
5343{
5344 __this_cpu_write(softnet_data.xmit.more, more);
5345}
5346
5347static inline bool netdev_xmit_more(void)
5348{
5349 return __this_cpu_read(softnet_data.xmit.more);
5350}
5351#else
5352static inline void netdev_xmit_set_more(bool more)
5353{
5354 current->net_xmit.more = more;
5355}
5356
5357static inline bool netdev_xmit_more(void)
5358{
5359 return current->net_xmit.more;
5360}
5361#endif
5362
5363static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
5364 struct sk_buff *skb, struct net_device *dev,
5365 bool more)
5366{
5367 netdev_xmit_set_more(more);
5368 return ops->ndo_start_xmit(skb, dev);
5369}
5370
5371static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
5372 struct netdev_queue *txq, bool more)
5373{
5374 const struct net_device_ops *ops = dev->netdev_ops;
5375 netdev_tx_t rc;
5376
5377 rc = __netdev_start_xmit(ops, skb, dev, more);
5378 if (rc == NETDEV_TX_OK)
5379 txq_trans_update(dev, txq);
5380
5381 return rc;
5382}
5383
5384int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5385 const struct ns_common *ns);
5386void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5387 const struct ns_common *ns);
5388
5389extern const struct kobj_ns_type_operations net_ns_type_operations;
5390
5391const char *netdev_drivername(const struct net_device *dev);
5392
5393static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5394 netdev_features_t f2)
5395{
5396 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5397 if (f1 & NETIF_F_HW_CSUM)
5398 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5399 else
5400 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5401 }
5402
5403 return f1 & f2;
5404}
5405
5406static inline netdev_features_t netdev_get_wanted_features(
5407 struct net_device *dev)
5408{
5409 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5410}
5411netdev_features_t netdev_increment_features(netdev_features_t all,
5412 netdev_features_t one, netdev_features_t mask);
5413
5414/* Allow TSO being used on stacked device :
5415 * Performing the GSO segmentation before last device
5416 * is a performance improvement.
5417 */
5418static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5419 netdev_features_t mask)
5420{
5421 return netdev_increment_features(features, NETIF_F_ALL_TSO |
5422 NETIF_F_ALL_FOR_ALL, mask);
5423}
5424
5425int __netdev_update_features(struct net_device *dev);
5426void netdev_update_features(struct net_device *dev);
5427void netdev_change_features(struct net_device *dev);
5428void netdev_compute_master_upper_features(struct net_device *dev, bool update_header);
5429
5430void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5431 struct net_device *dev);
5432
5433netdev_features_t passthru_features_check(struct sk_buff *skb,
5434 struct net_device *dev,
5435 netdev_features_t features);
5436netdev_features_t netif_skb_features(struct sk_buff *skb);
5437void skb_warn_bad_offload(const struct sk_buff *skb);
5438
5439static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5440{
5441 netdev_features_t feature;
5442
5443 if (gso_type & (SKB_GSO_TCP_FIXEDID | SKB_GSO_TCP_FIXEDID_INNER))
5444 gso_type |= __SKB_GSO_TCP_FIXEDID;
5445
5446 feature = ((netdev_features_t)gso_type << NETIF_F_GSO_SHIFT) & NETIF_F_GSO_MASK;
5447
5448 /* check flags correspondence */
5449 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5450 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5451 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5452 BUILD_BUG_ON(__SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5453 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5454 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5455 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5456 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5457 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5458 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5459 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5460 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5461 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5462 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5463 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5464 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5465 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5466 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5467 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5468 BUILD_BUG_ON(SKB_GSO_TCP_ACCECN !=
5469 (NETIF_F_GSO_ACCECN >> NETIF_F_GSO_SHIFT));
5470
5471 return (features & feature) == feature;
5472}
5473
5474static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5475{
5476 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5477 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5478}
5479
5480static inline bool netif_needs_gso(struct sk_buff *skb,
5481 netdev_features_t features)
5482{
5483 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5484 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5485 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5486}
5487
5488void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5489void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5490void netif_inherit_tso_max(struct net_device *to,
5491 const struct net_device *from);
5492
5493static inline unsigned int
5494netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5495{
5496 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5497 return skb->protocol == htons(ETH_P_IPV6) ?
5498 READ_ONCE(dev->gro_max_size) :
5499 READ_ONCE(dev->gro_ipv4_max_size);
5500}
5501
5502static inline unsigned int
5503netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5504{
5505 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5506 return skb->protocol == htons(ETH_P_IPV6) ?
5507 READ_ONCE(dev->gso_max_size) :
5508 READ_ONCE(dev->gso_ipv4_max_size);
5509}
5510
5511static inline bool netif_is_macsec(const struct net_device *dev)
5512{
5513 return dev->priv_flags & IFF_MACSEC;
5514}
5515
5516static inline bool netif_is_macvlan(const struct net_device *dev)
5517{
5518 return dev->priv_flags & IFF_MACVLAN;
5519}
5520
5521static inline bool netif_is_macvlan_port(const struct net_device *dev)
5522{
5523 return dev->priv_flags & IFF_MACVLAN_PORT;
5524}
5525
5526static inline bool netif_is_bond_master(const struct net_device *dev)
5527{
5528 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5529}
5530
5531static inline bool netif_is_bond_slave(const struct net_device *dev)
5532{
5533 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5534}
5535
5536static inline bool netif_supports_nofcs(struct net_device *dev)
5537{
5538 return dev->priv_flags & IFF_SUPP_NOFCS;
5539}
5540
5541static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5542{
5543 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5544}
5545
5546static inline bool netif_is_l3_master(const struct net_device *dev)
5547{
5548 return dev->priv_flags & IFF_L3MDEV_MASTER;
5549}
5550
5551static inline bool netif_is_l3_slave(const struct net_device *dev)
5552{
5553 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5554}
5555
5556static inline int dev_sdif(const struct net_device *dev)
5557{
5558#ifdef CONFIG_NET_L3_MASTER_DEV
5559 if (netif_is_l3_slave(dev))
5560 return dev->ifindex;
5561#endif
5562 return 0;
5563}
5564
5565static inline bool netif_is_bridge_master(const struct net_device *dev)
5566{
5567 return dev->priv_flags & IFF_EBRIDGE;
5568}
5569
5570static inline bool netif_is_bridge_port(const struct net_device *dev)
5571{
5572 return dev->priv_flags & IFF_BRIDGE_PORT;
5573}
5574
5575static inline bool netif_is_ovs_master(const struct net_device *dev)
5576{
5577 return dev->priv_flags & IFF_OPENVSWITCH;
5578}
5579
5580static inline bool netif_is_ovs_port(const struct net_device *dev)
5581{
5582 return dev->priv_flags & IFF_OVS_DATAPATH;
5583}
5584
5585static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5586{
5587 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5588}
5589
5590static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5591{
5592 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5593}
5594
5595static inline bool netif_is_team_master(const struct net_device *dev)
5596{
5597 return dev->priv_flags & IFF_TEAM;
5598}
5599
5600static inline bool netif_is_team_port(const struct net_device *dev)
5601{
5602 return dev->priv_flags & IFF_TEAM_PORT;
5603}
5604
5605static inline bool netif_is_lag_master(const struct net_device *dev)
5606{
5607 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5608}
5609
5610static inline bool netif_is_lag_port(const struct net_device *dev)
5611{
5612 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5613}
5614
5615bool netif_is_rxfh_configured(const struct net_device *dev);
5616
5617static inline bool netif_is_failover(const struct net_device *dev)
5618{
5619 return dev->priv_flags & IFF_FAILOVER;
5620}
5621
5622static inline bool netif_is_failover_slave(const struct net_device *dev)
5623{
5624 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5625}
5626
5627/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5628static inline void netif_keep_dst(struct net_device *dev)
5629{
5630 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5631}
5632
5633/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5634static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5635{
5636 /* TODO: reserve and use an additional IFF bit, if we get more users */
5637 return netif_is_macsec(dev);
5638}
5639
5640extern struct pernet_operations __net_initdata loopback_net_ops;
5641
5642/* Logging, debugging and troubleshooting/diagnostic helpers. */
5643
5644/* netdev_printk helpers, similar to dev_printk */
5645
5646static inline const char *netdev_name(const struct net_device *dev)
5647{
5648 if (!dev->name[0] || strchr(dev->name, '%'))
5649 return "(unnamed net_device)";
5650 return dev->name;
5651}
5652
5653static inline const char *netdev_reg_state(const struct net_device *dev)
5654{
5655 u8 reg_state = READ_ONCE(dev->reg_state);
5656
5657 switch (reg_state) {
5658 case NETREG_UNINITIALIZED: return " (uninitialized)";
5659 case NETREG_REGISTERED: return "";
5660 case NETREG_UNREGISTERING: return " (unregistering)";
5661 case NETREG_UNREGISTERED: return " (unregistered)";
5662 case NETREG_RELEASED: return " (released)";
5663 case NETREG_DUMMY: return " (dummy)";
5664 }
5665
5666 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5667 return " (unknown)";
5668}
5669
5670#define MODULE_ALIAS_NETDEV(device) \
5671 MODULE_ALIAS("netdev-" device)
5672
5673/*
5674 * netdev_WARN() acts like dev_printk(), but with the key difference
5675 * of using a WARN/WARN_ON to get the message out, including the
5676 * file/line information and a backtrace.
5677 */
5678#define netdev_WARN(dev, format, args...) \
5679 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5680 netdev_reg_state(dev), ##args)
5681
5682#define netdev_WARN_ONCE(dev, format, args...) \
5683 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5684 netdev_reg_state(dev), ##args)
5685
5686/*
5687 * The list of packet types we will receive (as opposed to discard)
5688 * and the routines to invoke.
5689 *
5690 * Why 16. Because with 16 the only overlap we get on a hash of the
5691 * low nibble of the protocol value is RARP/SNAP/X.25.
5692 *
5693 * 0800 IP
5694 * 0001 802.3
5695 * 0002 AX.25
5696 * 0004 802.2
5697 * 8035 RARP
5698 * 0005 SNAP
5699 * 0805 X.25
5700 * 0806 ARP
5701 * 8137 IPX
5702 * 0009 Localtalk
5703 * 86DD IPv6
5704 */
5705#define PTYPE_HASH_SIZE (16)
5706#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5707
5708extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5709
5710extern struct net_device *blackhole_netdev;
5711
5712/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5713#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5714#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5715 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5716#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5717
5718#endif /* _LINUX_NETDEVICE_H */