projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Linux 6.10-rc4
[linux-2.6-block.git] / include / linux / netdevice.h
... / ...
CommitLineData
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
44#include <linux/netdev_features.h>
45#include <linux/neighbour.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
56struct netpoll_info;
57struct device;
58struct ethtool_ops;
59struct kernel_hwtstamp_config;
60struct phy_device;
61struct dsa_port;
62struct ip_tunnel_parm_kern;
63struct macsec_context;
64struct macsec_ops;
65struct netdev_name_node;
66struct sd_flow_limit;
67struct sfp_bus;
68/* 802.11 specific */
69struct wireless_dev;
70/* 802.15.4 specific */
71struct wpan_dev;
72struct mpls_dev;
73/* UDP Tunnel offloads */
74struct udp_tunnel_info;
75struct udp_tunnel_nic_info;
76struct udp_tunnel_nic;
77struct bpf_prog;
78struct xdp_buff;
79struct xdp_frame;
80struct xdp_metadata_ops;
81struct xdp_md;
82
83typedef u32 xdp_features_t;
84
85void synchronize_net(void);
86void netdev_set_default_ethtool_ops(struct net_device *dev,
87 const struct ethtool_ops *ops);
88void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
89
90/* Backlog congestion levels */
91#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
92#define NET_RX_DROP 1 /* packet dropped */
93
94#define MAX_NEST_DEV 8
95
96/*
97 * Transmit return codes: transmit return codes originate from three different
98 * namespaces:
99 *
100 * - qdisc return codes
101 * - driver transmit return codes
102 * - errno values
103 *
104 * Drivers are allowed to return any one of those in their hard_start_xmit()
105 * function. Real network devices commonly used with qdiscs should only return
106 * the driver transmit return codes though - when qdiscs are used, the actual
107 * transmission happens asynchronously, so the value is not propagated to
108 * higher layers. Virtual network devices transmit synchronously; in this case
109 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
110 * others are propagated to higher layers.
111 */
112
113/* qdisc ->enqueue() return codes. */
114#define NET_XMIT_SUCCESS 0x00
115#define NET_XMIT_DROP 0x01 /* skb dropped */
116#define NET_XMIT_CN 0x02 /* congestion notification */
117#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
118
119/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
120 * indicates that the device will soon be dropping packets, or already drops
121 * some packets of the same priority; prompting us to send less aggressively. */
122#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
123#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
124
125/* Driver transmit return codes */
126#define NETDEV_TX_MASK 0xf0
127
128enum netdev_tx {
129 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
130 NETDEV_TX_OK = 0x00, /* driver took care of packet */
131 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
132};
133typedef enum netdev_tx netdev_tx_t;
134
135/*
136 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
137 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
138 */
139static inline bool dev_xmit_complete(int rc)
140{
141 /*
142 * Positive cases with an skb consumed by a driver:
143 * - successful transmission (rc == NETDEV_TX_OK)
144 * - error while transmitting (rc < 0)
145 * - error while queueing to a different device (rc & NET_XMIT_MASK)
146 */
147 if (likely(rc < NET_XMIT_MASK))
148 return true;
149
150 return false;
151}
152
153/*
154 * Compute the worst-case header length according to the protocols
155 * used.
156 */
157
158#if defined(CONFIG_HYPERV_NET)
159# define LL_MAX_HEADER 128
160#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
161# if defined(CONFIG_MAC80211_MESH)
162# define LL_MAX_HEADER 128
163# else
164# define LL_MAX_HEADER 96
165# endif
166#else
167# define LL_MAX_HEADER 32
168#endif
169
170#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
171 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
172#define MAX_HEADER LL_MAX_HEADER
173#else
174#define MAX_HEADER (LL_MAX_HEADER + 48)
175#endif
176
177/*
178 * Old network device statistics. Fields are native words
179 * (unsigned long) so they can be read and written atomically.
180 */
181
182#define NET_DEV_STAT(FIELD) \
183 union { \
184 unsigned long FIELD; \
185 atomic_long_t __##FIELD; \
186 }
187
188struct net_device_stats {
189 NET_DEV_STAT(rx_packets);
190 NET_DEV_STAT(tx_packets);
191 NET_DEV_STAT(rx_bytes);
192 NET_DEV_STAT(tx_bytes);
193 NET_DEV_STAT(rx_errors);
194 NET_DEV_STAT(tx_errors);
195 NET_DEV_STAT(rx_dropped);
196 NET_DEV_STAT(tx_dropped);
197 NET_DEV_STAT(multicast);
198 NET_DEV_STAT(collisions);
199 NET_DEV_STAT(rx_length_errors);
200 NET_DEV_STAT(rx_over_errors);
201 NET_DEV_STAT(rx_crc_errors);
202 NET_DEV_STAT(rx_frame_errors);
203 NET_DEV_STAT(rx_fifo_errors);
204 NET_DEV_STAT(rx_missed_errors);
205 NET_DEV_STAT(tx_aborted_errors);
206 NET_DEV_STAT(tx_carrier_errors);
207 NET_DEV_STAT(tx_fifo_errors);
208 NET_DEV_STAT(tx_heartbeat_errors);
209 NET_DEV_STAT(tx_window_errors);
210 NET_DEV_STAT(rx_compressed);
211 NET_DEV_STAT(tx_compressed);
212};
213#undef NET_DEV_STAT
214
215/* per-cpu stats, allocated on demand.
216 * Try to fit them in a single cache line, for dev_get_stats() sake.
217 */
218struct net_device_core_stats {
219 unsigned long rx_dropped;
220 unsigned long tx_dropped;
221 unsigned long rx_nohandler;
222 unsigned long rx_otherhost_dropped;
223} __aligned(4 * sizeof(unsigned long));
224
225#include <linux/cache.h>
226#include <linux/skbuff.h>
227
228struct neighbour;
229struct neigh_parms;
230struct sk_buff;
231
232struct netdev_hw_addr {
233 struct list_head list;
234 struct rb_node node;
235 unsigned char addr[MAX_ADDR_LEN];
236 unsigned char type;
237#define NETDEV_HW_ADDR_T_LAN 1
238#define NETDEV_HW_ADDR_T_SAN 2
239#define NETDEV_HW_ADDR_T_UNICAST 3
240#define NETDEV_HW_ADDR_T_MULTICAST 4
241 bool global_use;
242 int sync_cnt;
243 int refcount;
244 int synced;
245 struct rcu_head rcu_head;
246};
247
248struct netdev_hw_addr_list {
249 struct list_head list;
250 int count;
251
252 /* Auxiliary tree for faster lookup on addition and deletion */
253 struct rb_root tree;
254};
255
256#define netdev_hw_addr_list_count(l) ((l)->count)
257#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
258#define netdev_hw_addr_list_for_each(ha, l) \
259 list_for_each_entry(ha, &(l)->list, list)
260
261#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
262#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
263#define netdev_for_each_uc_addr(ha, dev) \
264 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
265#define netdev_for_each_synced_uc_addr(_ha, _dev) \
266 netdev_for_each_uc_addr((_ha), (_dev)) \
267 if ((_ha)->sync_cnt)
268
269#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
270#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
271#define netdev_for_each_mc_addr(ha, dev) \
272 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
273#define netdev_for_each_synced_mc_addr(_ha, _dev) \
274 netdev_for_each_mc_addr((_ha), (_dev)) \
275 if ((_ha)->sync_cnt)
276
277struct hh_cache {
278 unsigned int hh_len;
279 seqlock_t hh_lock;
280
281 /* cached hardware header; allow for machine alignment needs. */
282#define HH_DATA_MOD 16
283#define HH_DATA_OFF(__len) \
284 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
285#define HH_DATA_ALIGN(__len) \
286 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
287 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
288};
289
290/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
291 * Alternative is:
292 * dev->hard_header_len ? (dev->hard_header_len +
293 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
294 *
295 * We could use other alignment values, but we must maintain the
296 * relationship HH alignment <= LL alignment.
297 */
298#define LL_RESERVED_SPACE(dev) \
299 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
300 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
301#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
302 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
303 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
304
305struct header_ops {
306 int (*create) (struct sk_buff *skb, struct net_device *dev,
307 unsigned short type, const void *daddr,
308 const void *saddr, unsigned int len);
309 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
310 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
311 void (*cache_update)(struct hh_cache *hh,
312 const struct net_device *dev,
313 const unsigned char *haddr);
314 bool (*validate)(const char *ll_header, unsigned int len);
315 __be16 (*parse_protocol)(const struct sk_buff *skb);
316};
317
318/* These flag bits are private to the generic network queueing
319 * layer; they may not be explicitly referenced by any other
320 * code.
321 */
322
323enum netdev_state_t {
324 __LINK_STATE_START,
325 __LINK_STATE_PRESENT,
326 __LINK_STATE_NOCARRIER,
327 __LINK_STATE_LINKWATCH_PENDING,
328 __LINK_STATE_DORMANT,
329 __LINK_STATE_TESTING,
330};
331
332struct gro_list {
333 struct list_head list;
334 int count;
335};
336
337/*
338 * size of gro hash buckets, must less than bit number of
339 * napi_struct::gro_bitmask
340 */
341#define GRO_HASH_BUCKETS 8
342
343/*
344 * Structure for NAPI scheduling similar to tasklet but with weighting
345 */
346struct napi_struct {
347 /* The poll_list must only be managed by the entity which
348 * changes the state of the NAPI_STATE_SCHED bit. This means
349 * whoever atomically sets that bit can add this napi_struct
350 * to the per-CPU poll_list, and whoever clears that bit
351 * can remove from the list right before clearing the bit.
352 */
353 struct list_head poll_list;
354
355 unsigned long state;
356 int weight;
357 int defer_hard_irqs_count;
358 unsigned long gro_bitmask;
359 int (*poll)(struct napi_struct *, int);
360#ifdef CONFIG_NETPOLL
361 /* CPU actively polling if netpoll is configured */
362 int poll_owner;
363#endif
364 /* CPU on which NAPI has been scheduled for processing */
365 int list_owner;
366 struct net_device *dev;
367 struct gro_list gro_hash[GRO_HASH_BUCKETS];
368 struct sk_buff *skb;
369 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
370 int rx_count; /* length of rx_list */
371 unsigned int napi_id;
372 struct hrtimer timer;
373 struct task_struct *thread;
374 /* control-path-only fields follow */
375 struct list_head dev_list;
376 struct hlist_node napi_hash_node;
377 int irq;
378};
379
380enum {
381 NAPI_STATE_SCHED, /* Poll is scheduled */
382 NAPI_STATE_MISSED, /* reschedule a napi */
383 NAPI_STATE_DISABLE, /* Disable pending */
384 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
385 NAPI_STATE_LISTED, /* NAPI added to system lists */
386 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
387 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
388 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
389 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
390 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
391};
392
393enum {
394 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
395 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
396 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
397 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
398 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
399 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
400 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
401 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
402 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
403 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
404};
405
406enum gro_result {
407 GRO_MERGED,
408 GRO_MERGED_FREE,
409 GRO_HELD,
410 GRO_NORMAL,
411 GRO_CONSUMED,
412};
413typedef enum gro_result gro_result_t;
414
415/*
416 * enum rx_handler_result - Possible return values for rx_handlers.
417 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
418 * further.
419 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
420 * case skb->dev was changed by rx_handler.
421 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
422 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
423 *
424 * rx_handlers are functions called from inside __netif_receive_skb(), to do
425 * special processing of the skb, prior to delivery to protocol handlers.
426 *
427 * Currently, a net_device can only have a single rx_handler registered. Trying
428 * to register a second rx_handler will return -EBUSY.
429 *
430 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
431 * To unregister a rx_handler on a net_device, use
432 * netdev_rx_handler_unregister().
433 *
434 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
435 * do with the skb.
436 *
437 * If the rx_handler consumed the skb in some way, it should return
438 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
439 * the skb to be delivered in some other way.
440 *
441 * If the rx_handler changed skb->dev, to divert the skb to another
442 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
443 * new device will be called if it exists.
444 *
445 * If the rx_handler decides the skb should be ignored, it should return
446 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
447 * are registered on exact device (ptype->dev == skb->dev).
448 *
449 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
450 * delivered, it should return RX_HANDLER_PASS.
451 *
452 * A device without a registered rx_handler will behave as if rx_handler
453 * returned RX_HANDLER_PASS.
454 */
455
456enum rx_handler_result {
457 RX_HANDLER_CONSUMED,
458 RX_HANDLER_ANOTHER,
459 RX_HANDLER_EXACT,
460 RX_HANDLER_PASS,
461};
462typedef enum rx_handler_result rx_handler_result_t;
463typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
464
465void __napi_schedule(struct napi_struct *n);
466void __napi_schedule_irqoff(struct napi_struct *n);
467
468static inline bool napi_disable_pending(struct napi_struct *n)
469{
470 return test_bit(NAPI_STATE_DISABLE, &n->state);
471}
472
473static inline bool napi_prefer_busy_poll(struct napi_struct *n)
474{
475 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
476}
477
478/**
479 * napi_is_scheduled - test if NAPI is scheduled
480 * @n: NAPI context
481 *
482 * This check is "best-effort". With no locking implemented,
483 * a NAPI can be scheduled or terminate right after this check
484 * and produce not precise results.
485 *
486 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
487 * should not be used normally and napi_schedule should be
488 * used instead.
489 *
490 * Use only if the driver really needs to check if a NAPI
491 * is scheduled for example in the context of delayed timer
492 * that can be skipped if a NAPI is already scheduled.
493 *
494 * Return True if NAPI is scheduled, False otherwise.
495 */
496static inline bool napi_is_scheduled(struct napi_struct *n)
497{
498 return test_bit(NAPI_STATE_SCHED, &n->state);
499}
500
501bool napi_schedule_prep(struct napi_struct *n);
502
503/**
504 * napi_schedule - schedule NAPI poll
505 * @n: NAPI context
506 *
507 * Schedule NAPI poll routine to be called if it is not already
508 * running.
509 * Return true if we schedule a NAPI or false if not.
510 * Refer to napi_schedule_prep() for additional reason on why
511 * a NAPI might not be scheduled.
512 */
513static inline bool napi_schedule(struct napi_struct *n)
514{
515 if (napi_schedule_prep(n)) {
516 __napi_schedule(n);
517 return true;
518 }
519
520 return false;
521}
522
523/**
524 * napi_schedule_irqoff - schedule NAPI poll
525 * @n: NAPI context
526 *
527 * Variant of napi_schedule(), assuming hard irqs are masked.
528 */
529static inline void napi_schedule_irqoff(struct napi_struct *n)
530{
531 if (napi_schedule_prep(n))
532 __napi_schedule_irqoff(n);
533}
534
535/**
536 * napi_complete_done - NAPI processing complete
537 * @n: NAPI context
538 * @work_done: number of packets processed
539 *
540 * Mark NAPI processing as complete. Should only be called if poll budget
541 * has not been completely consumed.
542 * Prefer over napi_complete().
543 * Return false if device should avoid rearming interrupts.
544 */
545bool napi_complete_done(struct napi_struct *n, int work_done);
546
547static inline bool napi_complete(struct napi_struct *n)
548{
549 return napi_complete_done(n, 0);
550}
551
552int dev_set_threaded(struct net_device *dev, bool threaded);
553
554/**
555 * napi_disable - prevent NAPI from scheduling
556 * @n: NAPI context
557 *
558 * Stop NAPI from being scheduled on this context.
559 * Waits till any outstanding processing completes.
560 */
561void napi_disable(struct napi_struct *n);
562
563void napi_enable(struct napi_struct *n);
564
565/**
566 * napi_synchronize - wait until NAPI is not running
567 * @n: NAPI context
568 *
569 * Wait until NAPI is done being scheduled on this context.
570 * Waits till any outstanding processing completes but
571 * does not disable future activations.
572 */
573static inline void napi_synchronize(const struct napi_struct *n)
574{
575 if (IS_ENABLED(CONFIG_SMP))
576 while (test_bit(NAPI_STATE_SCHED, &n->state))
577 msleep(1);
578 else
579 barrier();
580}
581
582/**
583 * napi_if_scheduled_mark_missed - if napi is running, set the
584 * NAPIF_STATE_MISSED
585 * @n: NAPI context
586 *
587 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
588 * NAPI is scheduled.
589 **/
590static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
591{
592 unsigned long val, new;
593
594 val = READ_ONCE(n->state);
595 do {
596 if (val & NAPIF_STATE_DISABLE)
597 return true;
598
599 if (!(val & NAPIF_STATE_SCHED))
600 return false;
601
602 new = val | NAPIF_STATE_MISSED;
603 } while (!try_cmpxchg(&n->state, &val, new));
604
605 return true;
606}
607
608enum netdev_queue_state_t {
609 __QUEUE_STATE_DRV_XOFF,
610 __QUEUE_STATE_STACK_XOFF,
611 __QUEUE_STATE_FROZEN,
612};
613
614#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
615#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
616#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
617
618#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
619#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
620 QUEUE_STATE_FROZEN)
621#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
622 QUEUE_STATE_FROZEN)
623
624/*
625 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
626 * netif_tx_* functions below are used to manipulate this flag. The
627 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
628 * queue independently. The netif_xmit_*stopped functions below are called
629 * to check if the queue has been stopped by the driver or stack (either
630 * of the XOFF bits are set in the state). Drivers should not need to call
631 * netif_xmit*stopped functions, they should only be using netif_tx_*.
632 */
633
634struct netdev_queue {
635/*
636 * read-mostly part
637 */
638 struct net_device *dev;
639 netdevice_tracker dev_tracker;
640
641 struct Qdisc __rcu *qdisc;
642 struct Qdisc __rcu *qdisc_sleeping;
643#ifdef CONFIG_SYSFS
644 struct kobject kobj;
645#endif
646#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
647 int numa_node;
648#endif
649 unsigned long tx_maxrate;
650 /*
651 * Number of TX timeouts for this queue
652 * (/sys/class/net/DEV/Q/trans_timeout)
653 */
654 atomic_long_t trans_timeout;
655
656 /* Subordinate device that the queue has been assigned to */
657 struct net_device *sb_dev;
658#ifdef CONFIG_XDP_SOCKETS
659 struct xsk_buff_pool *pool;
660#endif
661 /* NAPI instance for the queue
662 * Readers and writers must hold RTNL
663 */
664 struct napi_struct *napi;
665/*
666 * write-mostly part
667 */
668 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
669 int xmit_lock_owner;
670 /*
671 * Time (in jiffies) of last Tx
672 */
673 unsigned long trans_start;
674
675 unsigned long state;
676
677#ifdef CONFIG_BQL
678 struct dql dql;
679#endif
680} ____cacheline_aligned_in_smp;
681
682extern int sysctl_fb_tunnels_only_for_init_net;
683extern int sysctl_devconf_inherit_init_net;
684
685/*
686 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
687 * == 1 : For initns only
688 * == 2 : For none.
689 */
690static inline bool net_has_fallback_tunnels(const struct net *net)
691{
692#if IS_ENABLED(CONFIG_SYSCTL)
693 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
694
695 return !fb_tunnels_only_for_init_net ||
696 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
697#else
698 return true;
699#endif
700}
701
702static inline int net_inherit_devconf(void)
703{
704#if IS_ENABLED(CONFIG_SYSCTL)
705 return READ_ONCE(sysctl_devconf_inherit_init_net);
706#else
707 return 0;
708#endif
709}
710
711static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
712{
713#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
714 return q->numa_node;
715#else
716 return NUMA_NO_NODE;
717#endif
718}
719
720static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
721{
722#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
723 q->numa_node = node;
724#endif
725}
726
727#ifdef CONFIG_RFS_ACCEL
728bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
729 u16 filter_id);
730#endif
731
732/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
733enum xps_map_type {
734 XPS_CPUS = 0,
735 XPS_RXQS,
736 XPS_MAPS_MAX,
737};
738
739#ifdef CONFIG_XPS
740/*
741 * This structure holds an XPS map which can be of variable length. The
742 * map is an array of queues.
743 */
744struct xps_map {
745 unsigned int len;
746 unsigned int alloc_len;
747 struct rcu_head rcu;
748 u16 queues[];
749};
750#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
751#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
752 - sizeof(struct xps_map)) / sizeof(u16))
753
754/*
755 * This structure holds all XPS maps for device. Maps are indexed by CPU.
756 *
757 * We keep track of the number of cpus/rxqs used when the struct is allocated,
758 * in nr_ids. This will help not accessing out-of-bound memory.
759 *
760 * We keep track of the number of traffic classes used when the struct is
761 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
762 * not crossing its upper bound, as the original dev->num_tc can be updated in
763 * the meantime.
764 */
765struct xps_dev_maps {
766 struct rcu_head rcu;
767 unsigned int nr_ids;
768 s16 num_tc;
769 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
770};
771
772#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
773 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
774
775#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
776 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
777
778#endif /* CONFIG_XPS */
779
780#define TC_MAX_QUEUE 16
781#define TC_BITMASK 15
782/* HW offloaded queuing disciplines txq count and offset maps */
783struct netdev_tc_txq {
784 u16 count;
785 u16 offset;
786};
787
788#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
789/*
790 * This structure is to hold information about the device
791 * configured to run FCoE protocol stack.
792 */
793struct netdev_fcoe_hbainfo {
794 char manufacturer[64];
795 char serial_number[64];
796 char hardware_version[64];
797 char driver_version[64];
798 char optionrom_version[64];
799 char firmware_version[64];
800 char model[256];
801 char model_description[256];
802};
803#endif
804
805#define MAX_PHYS_ITEM_ID_LEN 32
806
807/* This structure holds a unique identifier to identify some
808 * physical item (port for example) used by a netdevice.
809 */
810struct netdev_phys_item_id {
811 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
812 unsigned char id_len;
813};
814
815static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
816 struct netdev_phys_item_id *b)
817{
818 return a->id_len == b->id_len &&
819 memcmp(a->id, b->id, a->id_len) == 0;
820}
821
822typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
823 struct sk_buff *skb,
824 struct net_device *sb_dev);
825
826enum net_device_path_type {
827 DEV_PATH_ETHERNET = 0,
828 DEV_PATH_VLAN,
829 DEV_PATH_BRIDGE,
830 DEV_PATH_PPPOE,
831 DEV_PATH_DSA,
832 DEV_PATH_MTK_WDMA,
833};
834
835struct net_device_path {
836 enum net_device_path_type type;
837 const struct net_device *dev;
838 union {
839 struct {
840 u16 id;
841 __be16 proto;
842 u8 h_dest[ETH_ALEN];
843 } encap;
844 struct {
845 enum {
846 DEV_PATH_BR_VLAN_KEEP,
847 DEV_PATH_BR_VLAN_TAG,
848 DEV_PATH_BR_VLAN_UNTAG,
849 DEV_PATH_BR_VLAN_UNTAG_HW,
850 } vlan_mode;
851 u16 vlan_id;
852 __be16 vlan_proto;
853 } bridge;
854 struct {
855 int port;
856 u16 proto;
857 } dsa;
858 struct {
859 u8 wdma_idx;
860 u8 queue;
861 u16 wcid;
862 u8 bss;
863 u8 amsdu;
864 } mtk_wdma;
865 };
866};
867
868#define NET_DEVICE_PATH_STACK_MAX 5
869#define NET_DEVICE_PATH_VLAN_MAX 2
870
871struct net_device_path_stack {
872 int num_paths;
873 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
874};
875
876struct net_device_path_ctx {
877 const struct net_device *dev;
878 u8 daddr[ETH_ALEN];
879
880 int num_vlans;
881 struct {
882 u16 id;
883 __be16 proto;
884 } vlan[NET_DEVICE_PATH_VLAN_MAX];
885};
886
887enum tc_setup_type {
888 TC_QUERY_CAPS,
889 TC_SETUP_QDISC_MQPRIO,
890 TC_SETUP_CLSU32,
891 TC_SETUP_CLSFLOWER,
892 TC_SETUP_CLSMATCHALL,
893 TC_SETUP_CLSBPF,
894 TC_SETUP_BLOCK,
895 TC_SETUP_QDISC_CBS,
896 TC_SETUP_QDISC_RED,
897 TC_SETUP_QDISC_PRIO,
898 TC_SETUP_QDISC_MQ,
899 TC_SETUP_QDISC_ETF,
900 TC_SETUP_ROOT_QDISC,
901 TC_SETUP_QDISC_GRED,
902 TC_SETUP_QDISC_TAPRIO,
903 TC_SETUP_FT,
904 TC_SETUP_QDISC_ETS,
905 TC_SETUP_QDISC_TBF,
906 TC_SETUP_QDISC_FIFO,
907 TC_SETUP_QDISC_HTB,
908 TC_SETUP_ACT,
909};
910
911/* These structures hold the attributes of bpf state that are being passed
912 * to the netdevice through the bpf op.
913 */
914enum bpf_netdev_command {
915 /* Set or clear a bpf program used in the earliest stages of packet
916 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
917 * is responsible for calling bpf_prog_put on any old progs that are
918 * stored. In case of error, the callee need not release the new prog
919 * reference, but on success it takes ownership and must bpf_prog_put
920 * when it is no longer used.
921 */
922 XDP_SETUP_PROG,
923 XDP_SETUP_PROG_HW,
924 /* BPF program for offload callbacks, invoked at program load time. */
925 BPF_OFFLOAD_MAP_ALLOC,
926 BPF_OFFLOAD_MAP_FREE,
927 XDP_SETUP_XSK_POOL,
928};
929
930struct bpf_prog_offload_ops;
931struct netlink_ext_ack;
932struct xdp_umem;
933struct xdp_dev_bulk_queue;
934struct bpf_xdp_link;
935
936enum bpf_xdp_mode {
937 XDP_MODE_SKB = 0,
938 XDP_MODE_DRV = 1,
939 XDP_MODE_HW = 2,
940 __MAX_XDP_MODE
941};
942
943struct bpf_xdp_entity {
944 struct bpf_prog *prog;
945 struct bpf_xdp_link *link;
946};
947
948struct netdev_bpf {
949 enum bpf_netdev_command command;
950 union {
951 /* XDP_SETUP_PROG */
952 struct {
953 u32 flags;
954 struct bpf_prog *prog;
955 struct netlink_ext_ack *extack;
956 };
957 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
958 struct {
959 struct bpf_offloaded_map *offmap;
960 };
961 /* XDP_SETUP_XSK_POOL */
962 struct {
963 struct xsk_buff_pool *pool;
964 u16 queue_id;
965 } xsk;
966 };
967};
968
969/* Flags for ndo_xsk_wakeup. */
970#define XDP_WAKEUP_RX (1 << 0)
971#define XDP_WAKEUP_TX (1 << 1)
972
973#ifdef CONFIG_XFRM_OFFLOAD
974struct xfrmdev_ops {
975 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
976 void (*xdo_dev_state_delete) (struct xfrm_state *x);
977 void (*xdo_dev_state_free) (struct xfrm_state *x);
978 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
979 struct xfrm_state *x);
980 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
981 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
982 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
983 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
984 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
985};
986#endif
987
988struct dev_ifalias {
989 struct rcu_head rcuhead;
990 char ifalias[];
991};
992
993struct devlink;
994struct tlsdev_ops;
995
996struct netdev_net_notifier {
997 struct list_head list;
998 struct notifier_block *nb;
999};
1000
1001/*
1002 * This structure defines the management hooks for network devices.
1003 * The following hooks can be defined; unless noted otherwise, they are
1004 * optional and can be filled with a null pointer.
1005 *
1006 * int (*ndo_init)(struct net_device *dev);
1007 * This function is called once when a network device is registered.
1008 * The network device can use this for any late stage initialization
1009 * or semantic validation. It can fail with an error code which will
1010 * be propagated back to register_netdev.
1011 *
1012 * void (*ndo_uninit)(struct net_device *dev);
1013 * This function is called when device is unregistered or when registration
1014 * fails. It is not called if init fails.
1015 *
1016 * int (*ndo_open)(struct net_device *dev);
1017 * This function is called when a network device transitions to the up
1018 * state.
1019 *
1020 * int (*ndo_stop)(struct net_device *dev);
1021 * This function is called when a network device transitions to the down
1022 * state.
1023 *
1024 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1025 * struct net_device *dev);
1026 * Called when a packet needs to be transmitted.
1027 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1028 * the queue before that can happen; it's for obsolete devices and weird
1029 * corner cases, but the stack really does a non-trivial amount
1030 * of useless work if you return NETDEV_TX_BUSY.
1031 * Required; cannot be NULL.
1032 *
1033 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1034 * struct net_device *dev
1035 * netdev_features_t features);
1036 * Called by core transmit path to determine if device is capable of
1037 * performing offload operations on a given packet. This is to give
1038 * the device an opportunity to implement any restrictions that cannot
1039 * be otherwise expressed by feature flags. The check is called with
1040 * the set of features that the stack has calculated and it returns
1041 * those the driver believes to be appropriate.
1042 *
1043 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1044 * struct net_device *sb_dev);
1045 * Called to decide which queue to use when device supports multiple
1046 * transmit queues.
1047 *
1048 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1049 * This function is called to allow device receiver to make
1050 * changes to configuration when multicast or promiscuous is enabled.
1051 *
1052 * void (*ndo_set_rx_mode)(struct net_device *dev);
1053 * This function is called device changes address list filtering.
1054 * If driver handles unicast address filtering, it should set
1055 * IFF_UNICAST_FLT in its priv_flags.
1056 *
1057 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1058 * This function is called when the Media Access Control address
1059 * needs to be changed. If this interface is not defined, the
1060 * MAC address can not be changed.
1061 *
1062 * int (*ndo_validate_addr)(struct net_device *dev);
1063 * Test if Media Access Control address is valid for the device.
1064 *
1065 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1066 * Old-style ioctl entry point. This is used internally by the
1067 * appletalk and ieee802154 subsystems but is no longer called by
1068 * the device ioctl handler.
1069 *
1070 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1071 * Used by the bonding driver for its device specific ioctls:
1072 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1073 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1074 *
1075 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1076 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1077 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1078 *
1079 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1080 * Used to set network devices bus interface parameters. This interface
1081 * is retained for legacy reasons; new devices should use the bus
1082 * interface (PCI) for low level management.
1083 *
1084 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1085 * Called when a user wants to change the Maximum Transfer Unit
1086 * of a device.
1087 *
1088 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1089 * Callback used when the transmitter has not made any progress
1090 * for dev->watchdog ticks.
1091 *
1092 * void (*ndo_get_stats64)(struct net_device *dev,
1093 * struct rtnl_link_stats64 *storage);
1094 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1095 * Called when a user wants to get the network device usage
1096 * statistics. Drivers must do one of the following:
1097 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1098 * rtnl_link_stats64 structure passed by the caller.
1099 * 2. Define @ndo_get_stats to update a net_device_stats structure
1100 * (which should normally be dev->stats) and return a pointer to
1101 * it. The structure may be changed asynchronously only if each
1102 * field is written atomically.
1103 * 3. Update dev->stats asynchronously and atomically, and define
1104 * neither operation.
1105 *
1106 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1107 * Return true if this device supports offload stats of this attr_id.
1108 *
1109 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1110 * void *attr_data)
1111 * Get statistics for offload operations by attr_id. Write it into the
1112 * attr_data pointer.
1113 *
1114 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1115 * If device supports VLAN filtering this function is called when a
1116 * VLAN id is registered.
1117 *
1118 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1119 * If device supports VLAN filtering this function is called when a
1120 * VLAN id is unregistered.
1121 *
1122 * void (*ndo_poll_controller)(struct net_device *dev);
1123 *
1124 * SR-IOV management functions.
1125 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1126 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1127 * u8 qos, __be16 proto);
1128 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1129 * int max_tx_rate);
1130 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1131 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1132 * int (*ndo_get_vf_config)(struct net_device *dev,
1133 * int vf, struct ifla_vf_info *ivf);
1134 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1135 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1136 * struct nlattr *port[]);
1137 *
1138 * Enable or disable the VF ability to query its RSS Redirection Table and
1139 * Hash Key. This is needed since on some devices VF share this information
1140 * with PF and querying it may introduce a theoretical security risk.
1141 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1142 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1143 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1144 * void *type_data);
1145 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1146 * This is always called from the stack with the rtnl lock held and netif
1147 * tx queues stopped. This allows the netdevice to perform queue
1148 * management safely.
1149 *
1150 * Fiber Channel over Ethernet (FCoE) offload functions.
1151 * int (*ndo_fcoe_enable)(struct net_device *dev);
1152 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1153 * so the underlying device can perform whatever needed configuration or
1154 * initialization to support acceleration of FCoE traffic.
1155 *
1156 * int (*ndo_fcoe_disable)(struct net_device *dev);
1157 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1158 * so the underlying device can perform whatever needed clean-ups to
1159 * stop supporting acceleration of FCoE traffic.
1160 *
1161 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1162 * struct scatterlist *sgl, unsigned int sgc);
1163 * Called when the FCoE Initiator wants to initialize an I/O that
1164 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1165 * perform necessary setup and returns 1 to indicate the device is set up
1166 * successfully to perform DDP on this I/O, otherwise this returns 0.
1167 *
1168 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1169 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1170 * indicated by the FC exchange id 'xid', so the underlying device can
1171 * clean up and reuse resources for later DDP requests.
1172 *
1173 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1174 * struct scatterlist *sgl, unsigned int sgc);
1175 * Called when the FCoE Target wants to initialize an I/O that
1176 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1177 * perform necessary setup and returns 1 to indicate the device is set up
1178 * successfully to perform DDP on this I/O, otherwise this returns 0.
1179 *
1180 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1181 * struct netdev_fcoe_hbainfo *hbainfo);
1182 * Called when the FCoE Protocol stack wants information on the underlying
1183 * device. This information is utilized by the FCoE protocol stack to
1184 * register attributes with Fiber Channel management service as per the
1185 * FC-GS Fabric Device Management Information(FDMI) specification.
1186 *
1187 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1188 * Called when the underlying device wants to override default World Wide
1189 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1190 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1191 * protocol stack to use.
1192 *
1193 * RFS acceleration.
1194 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1195 * u16 rxq_index, u32 flow_id);
1196 * Set hardware filter for RFS. rxq_index is the target queue index;
1197 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1198 * Return the filter ID on success, or a negative error code.
1199 *
1200 * Slave management functions (for bridge, bonding, etc).
1201 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1202 * Called to make another netdev an underling.
1203 *
1204 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1205 * Called to release previously enslaved netdev.
1206 *
1207 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1208 * struct sk_buff *skb,
1209 * bool all_slaves);
1210 * Get the xmit slave of master device. If all_slaves is true, function
1211 * assume all the slaves can transmit.
1212 *
1213 * Feature/offload setting functions.
1214 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1215 * netdev_features_t features);
1216 * Adjusts the requested feature flags according to device-specific
1217 * constraints, and returns the resulting flags. Must not modify
1218 * the device state.
1219 *
1220 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1221 * Called to update device configuration to new features. Passed
1222 * feature set might be less than what was returned by ndo_fix_features()).
1223 * Must return >0 or -errno if it changed dev->features itself.
1224 *
1225 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1226 * struct net_device *dev,
1227 * const unsigned char *addr, u16 vid, u16 flags,
1228 * struct netlink_ext_ack *extack);
1229 * Adds an FDB entry to dev for addr.
1230 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1231 * struct net_device *dev,
1232 * const unsigned char *addr, u16 vid)
1233 * Deletes the FDB entry from dev coresponding to addr.
1234 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1235 * struct netlink_ext_ack *extack);
1236 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1237 * struct net_device *dev, struct net_device *filter_dev,
1238 * int *idx)
1239 * Used to add FDB entries to dump requests. Implementers should add
1240 * entries to skb and update idx with the number of entries.
1241 *
1242 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1243 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1244 * Adds an MDB entry to dev.
1245 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1246 * struct netlink_ext_ack *extack);
1247 * Deletes the MDB entry from dev.
1248 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1249 * struct netlink_ext_ack *extack);
1250 * Bulk deletes MDB entries from dev.
1251 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1252 * struct netlink_callback *cb);
1253 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1254 * callback is used by core rtnetlink code.
1255 *
1256 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1257 * u16 flags, struct netlink_ext_ack *extack)
1258 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1259 * struct net_device *dev, u32 filter_mask,
1260 * int nlflags)
1261 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1262 * u16 flags);
1263 *
1264 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1265 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1266 * which do not represent real hardware may define this to allow their
1267 * userspace components to manage their virtual carrier state. Devices
1268 * that determine carrier state from physical hardware properties (eg
1269 * network cables) or protocol-dependent mechanisms (eg
1270 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1271 *
1272 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1273 * struct netdev_phys_item_id *ppid);
1274 * Called to get ID of physical port of this device. If driver does
1275 * not implement this, it is assumed that the hw is not able to have
1276 * multiple net devices on single physical port.
1277 *
1278 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1279 * struct netdev_phys_item_id *ppid)
1280 * Called to get the parent ID of the physical port of this device.
1281 *
1282 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1283 * struct net_device *dev)
1284 * Called by upper layer devices to accelerate switching or other
1285 * station functionality into hardware. 'pdev is the lowerdev
1286 * to use for the offload and 'dev' is the net device that will
1287 * back the offload. Returns a pointer to the private structure
1288 * the upper layer will maintain.
1289 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1290 * Called by upper layer device to delete the station created
1291 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1292 * the station and priv is the structure returned by the add
1293 * operation.
1294 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1295 * int queue_index, u32 maxrate);
1296 * Called when a user wants to set a max-rate limitation of specific
1297 * TX queue.
1298 * int (*ndo_get_iflink)(const struct net_device *dev);
1299 * Called to get the iflink value of this device.
1300 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1301 * This function is used to get egress tunnel information for given skb.
1302 * This is useful for retrieving outer tunnel header parameters while
1303 * sampling packet.
1304 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1305 * This function is used to specify the headroom that the skb must
1306 * consider when allocation skb during packet reception. Setting
1307 * appropriate rx headroom value allows avoiding skb head copy on
1308 * forward. Setting a negative value resets the rx headroom to the
1309 * default value.
1310 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1311 * This function is used to set or query state related to XDP on the
1312 * netdevice and manage BPF offload. See definition of
1313 * enum bpf_netdev_command for details.
1314 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1315 * u32 flags);
1316 * This function is used to submit @n XDP packets for transmit on a
1317 * netdevice. Returns number of frames successfully transmitted, frames
1318 * that got dropped are freed/returned via xdp_return_frame().
1319 * Returns negative number, means general error invoking ndo, meaning
1320 * no frames were xmit'ed and core-caller will free all frames.
1321 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1322 * struct xdp_buff *xdp);
1323 * Get the xmit slave of master device based on the xdp_buff.
1324 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1325 * This function is used to wake up the softirq, ksoftirqd or kthread
1326 * responsible for sending and/or receiving packets on a specific
1327 * queue id bound to an AF_XDP socket. The flags field specifies if
1328 * only RX, only Tx, or both should be woken up using the flags
1329 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1330 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1331 * int cmd);
1332 * Add, change, delete or get information on an IPv4 tunnel.
1333 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1334 * If a device is paired with a peer device, return the peer instance.
1335 * The caller must be under RCU read context.
1336 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1337 * Get the forwarding path to reach the real device from the HW destination address
1338 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1339 * const struct skb_shared_hwtstamps *hwtstamps,
1340 * bool cycles);
1341 * Get hardware timestamp based on normal/adjustable time or free running
1342 * cycle counter. This function is required if physical clock supports a
1343 * free running cycle counter.
1344 *
1345 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1346 * struct kernel_hwtstamp_config *kernel_config);
1347 * Get the currently configured hardware timestamping parameters for the
1348 * NIC device.
1349 *
1350 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1351 * struct kernel_hwtstamp_config *kernel_config,
1352 * struct netlink_ext_ack *extack);
1353 * Change the hardware timestamping parameters for NIC device.
1354 */
1355struct net_device_ops {
1356 int (*ndo_init)(struct net_device *dev);
1357 void (*ndo_uninit)(struct net_device *dev);
1358 int (*ndo_open)(struct net_device *dev);
1359 int (*ndo_stop)(struct net_device *dev);
1360 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1361 struct net_device *dev);
1362 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1363 struct net_device *dev,
1364 netdev_features_t features);
1365 u16 (*ndo_select_queue)(struct net_device *dev,
1366 struct sk_buff *skb,
1367 struct net_device *sb_dev);
1368 void (*ndo_change_rx_flags)(struct net_device *dev,
1369 int flags);
1370 void (*ndo_set_rx_mode)(struct net_device *dev);
1371 int (*ndo_set_mac_address)(struct net_device *dev,
1372 void *addr);
1373 int (*ndo_validate_addr)(struct net_device *dev);
1374 int (*ndo_do_ioctl)(struct net_device *dev,
1375 struct ifreq *ifr, int cmd);
1376 int (*ndo_eth_ioctl)(struct net_device *dev,
1377 struct ifreq *ifr, int cmd);
1378 int (*ndo_siocbond)(struct net_device *dev,
1379 struct ifreq *ifr, int cmd);
1380 int (*ndo_siocwandev)(struct net_device *dev,
1381 struct if_settings *ifs);
1382 int (*ndo_siocdevprivate)(struct net_device *dev,
1383 struct ifreq *ifr,
1384 void __user *data, int cmd);
1385 int (*ndo_set_config)(struct net_device *dev,
1386 struct ifmap *map);
1387 int (*ndo_change_mtu)(struct net_device *dev,
1388 int new_mtu);
1389 int (*ndo_neigh_setup)(struct net_device *dev,
1390 struct neigh_parms *);
1391 void (*ndo_tx_timeout) (struct net_device *dev,
1392 unsigned int txqueue);
1393
1394 void (*ndo_get_stats64)(struct net_device *dev,
1395 struct rtnl_link_stats64 *storage);
1396 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1397 int (*ndo_get_offload_stats)(int attr_id,
1398 const struct net_device *dev,
1399 void *attr_data);
1400 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1401
1402 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1403 __be16 proto, u16 vid);
1404 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1405 __be16 proto, u16 vid);
1406#ifdef CONFIG_NET_POLL_CONTROLLER
1407 void (*ndo_poll_controller)(struct net_device *dev);
1408 int (*ndo_netpoll_setup)(struct net_device *dev,
1409 struct netpoll_info *info);
1410 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1411#endif
1412 int (*ndo_set_vf_mac)(struct net_device *dev,
1413 int queue, u8 *mac);
1414 int (*ndo_set_vf_vlan)(struct net_device *dev,
1415 int queue, u16 vlan,
1416 u8 qos, __be16 proto);
1417 int (*ndo_set_vf_rate)(struct net_device *dev,
1418 int vf, int min_tx_rate,
1419 int max_tx_rate);
1420 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1421 int vf, bool setting);
1422 int (*ndo_set_vf_trust)(struct net_device *dev,
1423 int vf, bool setting);
1424 int (*ndo_get_vf_config)(struct net_device *dev,
1425 int vf,
1426 struct ifla_vf_info *ivf);
1427 int (*ndo_set_vf_link_state)(struct net_device *dev,
1428 int vf, int link_state);
1429 int (*ndo_get_vf_stats)(struct net_device *dev,
1430 int vf,
1431 struct ifla_vf_stats
1432 *vf_stats);
1433 int (*ndo_set_vf_port)(struct net_device *dev,
1434 int vf,
1435 struct nlattr *port[]);
1436 int (*ndo_get_vf_port)(struct net_device *dev,
1437 int vf, struct sk_buff *skb);
1438 int (*ndo_get_vf_guid)(struct net_device *dev,
1439 int vf,
1440 struct ifla_vf_guid *node_guid,
1441 struct ifla_vf_guid *port_guid);
1442 int (*ndo_set_vf_guid)(struct net_device *dev,
1443 int vf, u64 guid,
1444 int guid_type);
1445 int (*ndo_set_vf_rss_query_en)(
1446 struct net_device *dev,
1447 int vf, bool setting);
1448 int (*ndo_setup_tc)(struct net_device *dev,
1449 enum tc_setup_type type,
1450 void *type_data);
1451#if IS_ENABLED(CONFIG_FCOE)
1452 int (*ndo_fcoe_enable)(struct net_device *dev);
1453 int (*ndo_fcoe_disable)(struct net_device *dev);
1454 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1455 u16 xid,
1456 struct scatterlist *sgl,
1457 unsigned int sgc);
1458 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1459 u16 xid);
1460 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1461 u16 xid,
1462 struct scatterlist *sgl,
1463 unsigned int sgc);
1464 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1465 struct netdev_fcoe_hbainfo *hbainfo);
1466#endif
1467
1468#if IS_ENABLED(CONFIG_LIBFCOE)
1469#define NETDEV_FCOE_WWNN 0
1470#define NETDEV_FCOE_WWPN 1
1471 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1472 u64 *wwn, int type);
1473#endif
1474
1475#ifdef CONFIG_RFS_ACCEL
1476 int (*ndo_rx_flow_steer)(struct net_device *dev,
1477 const struct sk_buff *skb,
1478 u16 rxq_index,
1479 u32 flow_id);
1480#endif
1481 int (*ndo_add_slave)(struct net_device *dev,
1482 struct net_device *slave_dev,
1483 struct netlink_ext_ack *extack);
1484 int (*ndo_del_slave)(struct net_device *dev,
1485 struct net_device *slave_dev);
1486 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1487 struct sk_buff *skb,
1488 bool all_slaves);
1489 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1490 struct sock *sk);
1491 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1492 netdev_features_t features);
1493 int (*ndo_set_features)(struct net_device *dev,
1494 netdev_features_t features);
1495 int (*ndo_neigh_construct)(struct net_device *dev,
1496 struct neighbour *n);
1497 void (*ndo_neigh_destroy)(struct net_device *dev,
1498 struct neighbour *n);
1499
1500 int (*ndo_fdb_add)(struct ndmsg *ndm,
1501 struct nlattr *tb[],
1502 struct net_device *dev,
1503 const unsigned char *addr,
1504 u16 vid,
1505 u16 flags,
1506 struct netlink_ext_ack *extack);
1507 int (*ndo_fdb_del)(struct ndmsg *ndm,
1508 struct nlattr *tb[],
1509 struct net_device *dev,
1510 const unsigned char *addr,
1511 u16 vid, struct netlink_ext_ack *extack);
1512 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1513 struct net_device *dev,
1514 struct netlink_ext_ack *extack);
1515 int (*ndo_fdb_dump)(struct sk_buff *skb,
1516 struct netlink_callback *cb,
1517 struct net_device *dev,
1518 struct net_device *filter_dev,
1519 int *idx);
1520 int (*ndo_fdb_get)(struct sk_buff *skb,
1521 struct nlattr *tb[],
1522 struct net_device *dev,
1523 const unsigned char *addr,
1524 u16 vid, u32 portid, u32 seq,
1525 struct netlink_ext_ack *extack);
1526 int (*ndo_mdb_add)(struct net_device *dev,
1527 struct nlattr *tb[],
1528 u16 nlmsg_flags,
1529 struct netlink_ext_ack *extack);
1530 int (*ndo_mdb_del)(struct net_device *dev,
1531 struct nlattr *tb[],
1532 struct netlink_ext_ack *extack);
1533 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1534 struct nlattr *tb[],
1535 struct netlink_ext_ack *extack);
1536 int (*ndo_mdb_dump)(struct net_device *dev,
1537 struct sk_buff *skb,
1538 struct netlink_callback *cb);
1539 int (*ndo_mdb_get)(struct net_device *dev,
1540 struct nlattr *tb[], u32 portid,
1541 u32 seq,
1542 struct netlink_ext_ack *extack);
1543 int (*ndo_bridge_setlink)(struct net_device *dev,
1544 struct nlmsghdr *nlh,
1545 u16 flags,
1546 struct netlink_ext_ack *extack);
1547 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1548 u32 pid, u32 seq,
1549 struct net_device *dev,
1550 u32 filter_mask,
1551 int nlflags);
1552 int (*ndo_bridge_dellink)(struct net_device *dev,
1553 struct nlmsghdr *nlh,
1554 u16 flags);
1555 int (*ndo_change_carrier)(struct net_device *dev,
1556 bool new_carrier);
1557 int (*ndo_get_phys_port_id)(struct net_device *dev,
1558 struct netdev_phys_item_id *ppid);
1559 int (*ndo_get_port_parent_id)(struct net_device *dev,
1560 struct netdev_phys_item_id *ppid);
1561 int (*ndo_get_phys_port_name)(struct net_device *dev,
1562 char *name, size_t len);
1563 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1564 struct net_device *dev);
1565 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1566 void *priv);
1567
1568 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1569 int queue_index,
1570 u32 maxrate);
1571 int (*ndo_get_iflink)(const struct net_device *dev);
1572 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1573 struct sk_buff *skb);
1574 void (*ndo_set_rx_headroom)(struct net_device *dev,
1575 int needed_headroom);
1576 int (*ndo_bpf)(struct net_device *dev,
1577 struct netdev_bpf *bpf);
1578 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1579 struct xdp_frame **xdp,
1580 u32 flags);
1581 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1582 struct xdp_buff *xdp);
1583 int (*ndo_xsk_wakeup)(struct net_device *dev,
1584 u32 queue_id, u32 flags);
1585 int (*ndo_tunnel_ctl)(struct net_device *dev,
1586 struct ip_tunnel_parm_kern *p,
1587 int cmd);
1588 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1589 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1590 struct net_device_path *path);
1591 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1592 const struct skb_shared_hwtstamps *hwtstamps,
1593 bool cycles);
1594 int (*ndo_hwtstamp_get)(struct net_device *dev,
1595 struct kernel_hwtstamp_config *kernel_config);
1596 int (*ndo_hwtstamp_set)(struct net_device *dev,
1597 struct kernel_hwtstamp_config *kernel_config,
1598 struct netlink_ext_ack *extack);
1599};
1600
1601/**
1602 * enum netdev_priv_flags - &struct net_device priv_flags
1603 *
1604 * These are the &struct net_device, they are only set internally
1605 * by drivers and used in the kernel. These flags are invisible to
1606 * userspace; this means that the order of these flags can change
1607 * during any kernel release.
1608 *
1609 * You should have a pretty good reason to be extending these flags.
1610 *
1611 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1612 * @IFF_EBRIDGE: Ethernet bridging device
1613 * @IFF_BONDING: bonding master or slave
1614 * @IFF_ISATAP: ISATAP interface (RFC4214)
1615 * @IFF_WAN_HDLC: WAN HDLC device
1616 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1617 * release skb->dst
1618 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1619 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1620 * @IFF_MACVLAN_PORT: device used as macvlan port
1621 * @IFF_BRIDGE_PORT: device used as bridge port
1622 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1623 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1624 * @IFF_UNICAST_FLT: Supports unicast filtering
1625 * @IFF_TEAM_PORT: device used as team port
1626 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1627 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1628 * change when it's running
1629 * @IFF_MACVLAN: Macvlan device
1630 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1631 * underlying stacked devices
1632 * @IFF_L3MDEV_MASTER: device is an L3 master device
1633 * @IFF_NO_QUEUE: device can run without qdisc attached
1634 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1635 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1636 * @IFF_TEAM: device is a team device
1637 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1638 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1639 * entity (i.e. the master device for bridged veth)
1640 * @IFF_MACSEC: device is a MACsec device
1641 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1642 * @IFF_FAILOVER: device is a failover master device
1643 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1644 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1645 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1646 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1647 * skb_headlen(skb) == 0 (data starts from frag0)
1648 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1649 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1650 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1651 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1652 */
1653enum netdev_priv_flags {
1654 IFF_802_1Q_VLAN = 1<<0,
1655 IFF_EBRIDGE = 1<<1,
1656 IFF_BONDING = 1<<2,
1657 IFF_ISATAP = 1<<3,
1658 IFF_WAN_HDLC = 1<<4,
1659 IFF_XMIT_DST_RELEASE = 1<<5,
1660 IFF_DONT_BRIDGE = 1<<6,
1661 IFF_DISABLE_NETPOLL = 1<<7,
1662 IFF_MACVLAN_PORT = 1<<8,
1663 IFF_BRIDGE_PORT = 1<<9,
1664 IFF_OVS_DATAPATH = 1<<10,
1665 IFF_TX_SKB_SHARING = 1<<11,
1666 IFF_UNICAST_FLT = 1<<12,
1667 IFF_TEAM_PORT = 1<<13,
1668 IFF_SUPP_NOFCS = 1<<14,
1669 IFF_LIVE_ADDR_CHANGE = 1<<15,
1670 IFF_MACVLAN = 1<<16,
1671 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1672 IFF_L3MDEV_MASTER = 1<<18,
1673 IFF_NO_QUEUE = 1<<19,
1674 IFF_OPENVSWITCH = 1<<20,
1675 IFF_L3MDEV_SLAVE = 1<<21,
1676 IFF_TEAM = 1<<22,
1677 IFF_RXFH_CONFIGURED = 1<<23,
1678 IFF_PHONY_HEADROOM = 1<<24,
1679 IFF_MACSEC = 1<<25,
1680 IFF_NO_RX_HANDLER = 1<<26,
1681 IFF_FAILOVER = 1<<27,
1682 IFF_FAILOVER_SLAVE = 1<<28,
1683 IFF_L3MDEV_RX_HANDLER = 1<<29,
1684 IFF_NO_ADDRCONF = BIT_ULL(30),
1685 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1686 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1687 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1688};
1689
1690#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1691#define IFF_EBRIDGE IFF_EBRIDGE
1692#define IFF_BONDING IFF_BONDING
1693#define IFF_ISATAP IFF_ISATAP
1694#define IFF_WAN_HDLC IFF_WAN_HDLC
1695#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1696#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1697#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1698#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1699#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1700#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1701#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1702#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1703#define IFF_TEAM_PORT IFF_TEAM_PORT
1704#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1705#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1706#define IFF_MACVLAN IFF_MACVLAN
1707#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1708#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1709#define IFF_NO_QUEUE IFF_NO_QUEUE
1710#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1711#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1712#define IFF_TEAM IFF_TEAM
1713#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1714#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1715#define IFF_MACSEC IFF_MACSEC
1716#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1717#define IFF_FAILOVER IFF_FAILOVER
1718#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1719#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1720#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1721
1722/* Specifies the type of the struct net_device::ml_priv pointer */
1723enum netdev_ml_priv_type {
1724 ML_PRIV_NONE,
1725 ML_PRIV_CAN,
1726};
1727
1728enum netdev_stat_type {
1729 NETDEV_PCPU_STAT_NONE,
1730 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1731 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1732 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1733};
1734
1735enum netdev_reg_state {
1736 NETREG_UNINITIALIZED = 0,
1737 NETREG_REGISTERED, /* completed register_netdevice */
1738 NETREG_UNREGISTERING, /* called unregister_netdevice */
1739 NETREG_UNREGISTERED, /* completed unregister todo */
1740 NETREG_RELEASED, /* called free_netdev */
1741 NETREG_DUMMY, /* dummy device for NAPI poll */
1742};
1743
1744/**
1745 * struct net_device - The DEVICE structure.
1746 *
1747 * Actually, this whole structure is a big mistake. It mixes I/O
1748 * data with strictly "high-level" data, and it has to know about
1749 * almost every data structure used in the INET module.
1750 *
1751 * @name: This is the first field of the "visible" part of this structure
1752 * (i.e. as seen by users in the "Space.c" file). It is the name
1753 * of the interface.
1754 *
1755 * @name_node: Name hashlist node
1756 * @ifalias: SNMP alias
1757 * @mem_end: Shared memory end
1758 * @mem_start: Shared memory start
1759 * @base_addr: Device I/O address
1760 * @irq: Device IRQ number
1761 *
1762 * @state: Generic network queuing layer state, see netdev_state_t
1763 * @dev_list: The global list of network devices
1764 * @napi_list: List entry used for polling NAPI devices
1765 * @unreg_list: List entry when we are unregistering the
1766 * device; see the function unregister_netdev
1767 * @close_list: List entry used when we are closing the device
1768 * @ptype_all: Device-specific packet handlers for all protocols
1769 * @ptype_specific: Device-specific, protocol-specific packet handlers
1770 *
1771 * @adj_list: Directly linked devices, like slaves for bonding
1772 * @features: Currently active device features
1773 * @hw_features: User-changeable features
1774 *
1775 * @wanted_features: User-requested features
1776 * @vlan_features: Mask of features inheritable by VLAN devices
1777 *
1778 * @hw_enc_features: Mask of features inherited by encapsulating devices
1779 * This field indicates what encapsulation
1780 * offloads the hardware is capable of doing,
1781 * and drivers will need to set them appropriately.
1782 *
1783 * @mpls_features: Mask of features inheritable by MPLS
1784 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1785 *
1786 * @ifindex: interface index
1787 * @group: The group the device belongs to
1788 *
1789 * @stats: Statistics struct, which was left as a legacy, use
1790 * rtnl_link_stats64 instead
1791 *
1792 * @core_stats: core networking counters,
1793 * do not use this in drivers
1794 * @carrier_up_count: Number of times the carrier has been up
1795 * @carrier_down_count: Number of times the carrier has been down
1796 *
1797 * @wireless_handlers: List of functions to handle Wireless Extensions,
1798 * instead of ioctl,
1799 * see <net/iw_handler.h> for details.
1800 * @wireless_data: Instance data managed by the core of wireless extensions
1801 *
1802 * @netdev_ops: Includes several pointers to callbacks,
1803 * if one wants to override the ndo_*() functions
1804 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1805 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1806 * @ethtool_ops: Management operations
1807 * @l3mdev_ops: Layer 3 master device operations
1808 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1809 * discovery handling. Necessary for e.g. 6LoWPAN.
1810 * @xfrmdev_ops: Transformation offload operations
1811 * @tlsdev_ops: Transport Layer Security offload operations
1812 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1813 * of Layer 2 headers.
1814 *
1815 * @flags: Interface flags (a la BSD)
1816 * @xdp_features: XDP capability supported by the device
1817 * @priv_flags: Like 'flags' but invisible to userspace,
1818 * see if.h for the definitions
1819 * @gflags: Global flags ( kept as legacy )
1820 * @padded: How much padding added by alloc_netdev()
1821 * @operstate: RFC2863 operstate
1822 * @link_mode: Mapping policy to operstate
1823 * @if_port: Selectable AUI, TP, ...
1824 * @dma: DMA channel
1825 * @mtu: Interface MTU value
1826 * @min_mtu: Interface Minimum MTU value
1827 * @max_mtu: Interface Maximum MTU value
1828 * @type: Interface hardware type
1829 * @hard_header_len: Maximum hardware header length.
1830 * @min_header_len: Minimum hardware header length
1831 *
1832 * @needed_headroom: Extra headroom the hardware may need, but not in all
1833 * cases can this be guaranteed
1834 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1835 * cases can this be guaranteed. Some cases also use
1836 * LL_MAX_HEADER instead to allocate the skb
1837 *
1838 * interface address info:
1839 *
1840 * @perm_addr: Permanent hw address
1841 * @addr_assign_type: Hw address assignment type
1842 * @addr_len: Hardware address length
1843 * @upper_level: Maximum depth level of upper devices.
1844 * @lower_level: Maximum depth level of lower devices.
1845 * @neigh_priv_len: Used in neigh_alloc()
1846 * @dev_id: Used to differentiate devices that share
1847 * the same link layer address
1848 * @dev_port: Used to differentiate devices that share
1849 * the same function
1850 * @addr_list_lock: XXX: need comments on this one
1851 * @name_assign_type: network interface name assignment type
1852 * @uc_promisc: Counter that indicates promiscuous mode
1853 * has been enabled due to the need to listen to
1854 * additional unicast addresses in a device that
1855 * does not implement ndo_set_rx_mode()
1856 * @uc: unicast mac addresses
1857 * @mc: multicast mac addresses
1858 * @dev_addrs: list of device hw addresses
1859 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1860 * @promiscuity: Number of times the NIC is told to work in
1861 * promiscuous mode; if it becomes 0 the NIC will
1862 * exit promiscuous mode
1863 * @allmulti: Counter, enables or disables allmulticast mode
1864 *
1865 * @vlan_info: VLAN info
1866 * @dsa_ptr: dsa specific data
1867 * @tipc_ptr: TIPC specific data
1868 * @atalk_ptr: AppleTalk link
1869 * @ip_ptr: IPv4 specific data
1870 * @ip6_ptr: IPv6 specific data
1871 * @ax25_ptr: AX.25 specific data
1872 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1873 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1874 * device struct
1875 * @mpls_ptr: mpls_dev struct pointer
1876 * @mctp_ptr: MCTP specific data
1877 *
1878 * @dev_addr: Hw address (before bcast,
1879 * because most packets are unicast)
1880 *
1881 * @_rx: Array of RX queues
1882 * @num_rx_queues: Number of RX queues
1883 * allocated at register_netdev() time
1884 * @real_num_rx_queues: Number of RX queues currently active in device
1885 * @xdp_prog: XDP sockets filter program pointer
1886 * @gro_flush_timeout: timeout for GRO layer in NAPI
1887 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1888 * allow to avoid NIC hard IRQ, on busy queues.
1889 *
1890 * @rx_handler: handler for received packets
1891 * @rx_handler_data: XXX: need comments on this one
1892 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1893 * @ingress_queue: XXX: need comments on this one
1894 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1895 * @broadcast: hw bcast address
1896 *
1897 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1898 * indexed by RX queue number. Assigned by driver.
1899 * This must only be set if the ndo_rx_flow_steer
1900 * operation is defined
1901 * @index_hlist: Device index hash chain
1902 *
1903 * @_tx: Array of TX queues
1904 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1905 * @real_num_tx_queues: Number of TX queues currently active in device
1906 * @qdisc: Root qdisc from userspace point of view
1907 * @tx_queue_len: Max frames per queue allowed
1908 * @tx_global_lock: XXX: need comments on this one
1909 * @xdp_bulkq: XDP device bulk queue
1910 * @xps_maps: all CPUs/RXQs maps for XPS device
1911 *
1912 * @xps_maps: XXX: need comments on this one
1913 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1914 * @nf_hooks_egress: netfilter hooks executed for egress packets
1915 * @qdisc_hash: qdisc hash table
1916 * @watchdog_timeo: Represents the timeout that is used by
1917 * the watchdog (see dev_watchdog())
1918 * @watchdog_timer: List of timers
1919 *
1920 * @proto_down_reason: reason a netdev interface is held down
1921 * @pcpu_refcnt: Number of references to this device
1922 * @dev_refcnt: Number of references to this device
1923 * @refcnt_tracker: Tracker directory for tracked references to this device
1924 * @todo_list: Delayed register/unregister
1925 * @link_watch_list: XXX: need comments on this one
1926 *
1927 * @reg_state: Register/unregister state machine
1928 * @dismantle: Device is going to be freed
1929 * @rtnl_link_state: This enum represents the phases of creating
1930 * a new link
1931 *
1932 * @needs_free_netdev: Should unregister perform free_netdev?
1933 * @priv_destructor: Called from unregister
1934 * @npinfo: XXX: need comments on this one
1935 * @nd_net: Network namespace this network device is inside
1936 *
1937 * @ml_priv: Mid-layer private
1938 * @ml_priv_type: Mid-layer private type
1939 *
1940 * @pcpu_stat_type: Type of device statistics which the core should
1941 * allocate/free: none, lstats, tstats, dstats. none
1942 * means the driver is handling statistics allocation/
1943 * freeing internally.
1944 * @lstats: Loopback statistics: packets, bytes
1945 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
1946 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
1947 *
1948 * @garp_port: GARP
1949 * @mrp_port: MRP
1950 *
1951 * @dm_private: Drop monitor private
1952 *
1953 * @dev: Class/net/name entry
1954 * @sysfs_groups: Space for optional device, statistics and wireless
1955 * sysfs groups
1956 *
1957 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1958 * @rtnl_link_ops: Rtnl_link_ops
1959 * @stat_ops: Optional ops for queue-aware statistics
1960 * @queue_mgmt_ops: Optional ops for queue management
1961 *
1962 * @gso_max_size: Maximum size of generic segmentation offload
1963 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1964 * @gso_max_segs: Maximum number of segments that can be passed to the
1965 * NIC for GSO
1966 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1967 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1968 * for IPv4.
1969 *
1970 * @dcbnl_ops: Data Center Bridging netlink ops
1971 * @num_tc: Number of traffic classes in the net device
1972 * @tc_to_txq: XXX: need comments on this one
1973 * @prio_tc_map: XXX: need comments on this one
1974 *
1975 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1976 *
1977 * @priomap: XXX: need comments on this one
1978 * @phydev: Physical device may attach itself
1979 * for hardware timestamping
1980 * @sfp_bus: attached &struct sfp_bus structure.
1981 *
1982 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1983 *
1984 * @proto_down: protocol port state information can be sent to the
1985 * switch driver and used to set the phys state of the
1986 * switch port.
1987 *
1988 * @wol_enabled: Wake-on-LAN is enabled
1989 *
1990 * @threaded: napi threaded mode is enabled
1991 *
1992 * @net_notifier_list: List of per-net netdev notifier block
1993 * that follow this device when it is moved
1994 * to another network namespace.
1995 *
1996 * @macsec_ops: MACsec offloading ops
1997 *
1998 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1999 * offload capabilities of the device
2000 * @udp_tunnel_nic: UDP tunnel offload state
2001 * @xdp_state: stores info on attached XDP BPF programs
2002 *
2003 * @nested_level: Used as a parameter of spin_lock_nested() of
2004 * dev->addr_list_lock.
2005 * @unlink_list: As netif_addr_lock() can be called recursively,
2006 * keep a list of interfaces to be deleted.
2007 * @gro_max_size: Maximum size of aggregated packet in generic
2008 * receive offload (GRO)
2009 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2010 * receive offload (GRO), for IPv4.
2011 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2012 * zero copy driver
2013 *
2014 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2015 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2016 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2017 * @dev_registered_tracker: tracker for reference held while
2018 * registered
2019 * @offload_xstats_l3: L3 HW stats for this netdevice.
2020 *
2021 * @devlink_port: Pointer to related devlink port structure.
2022 * Assigned by a driver before netdev registration using
2023 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2024 * during the time netdevice is registered.
2025 *
2026 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2027 * where the clock is recovered.
2028 *
2029 * FIXME: cleanup struct net_device such that network protocol info
2030 * moves out.
2031 */
2032
2033struct net_device {
2034 /* Cacheline organization can be found documented in
2035 * Documentation/networking/net_cachelines/net_device.rst.
2036 * Please update the document when adding new fields.
2037 */
2038
2039 /* TX read-mostly hotpath */
2040 __cacheline_group_begin(net_device_read_tx);
2041 unsigned long long priv_flags;
2042 const struct net_device_ops *netdev_ops;
2043 const struct header_ops *header_ops;
2044 struct netdev_queue *_tx;
2045 netdev_features_t gso_partial_features;
2046 unsigned int real_num_tx_queues;
2047 unsigned int gso_max_size;
2048 unsigned int gso_ipv4_max_size;
2049 u16 gso_max_segs;
2050 s16 num_tc;
2051 /* Note : dev->mtu is often read without holding a lock.
2052 * Writers usually hold RTNL.
2053 * It is recommended to use READ_ONCE() to annotate the reads,
2054 * and to use WRITE_ONCE() to annotate the writes.
2055 */
2056 unsigned int mtu;
2057 unsigned short needed_headroom;
2058 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2059#ifdef CONFIG_XPS
2060 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2061#endif
2062#ifdef CONFIG_NETFILTER_EGRESS
2063 struct nf_hook_entries __rcu *nf_hooks_egress;
2064#endif
2065#ifdef CONFIG_NET_XGRESS
2066 struct bpf_mprog_entry __rcu *tcx_egress;
2067#endif
2068 __cacheline_group_end(net_device_read_tx);
2069
2070 /* TXRX read-mostly hotpath */
2071 __cacheline_group_begin(net_device_read_txrx);
2072 union {
2073 struct pcpu_lstats __percpu *lstats;
2074 struct pcpu_sw_netstats __percpu *tstats;
2075 struct pcpu_dstats __percpu *dstats;
2076 };
2077 unsigned long state;
2078 unsigned int flags;
2079 unsigned short hard_header_len;
2080 netdev_features_t features;
2081 struct inet6_dev __rcu *ip6_ptr;
2082 __cacheline_group_end(net_device_read_txrx);
2083
2084 /* RX read-mostly hotpath */
2085 __cacheline_group_begin(net_device_read_rx);
2086 struct bpf_prog __rcu *xdp_prog;
2087 struct list_head ptype_specific;
2088 int ifindex;
2089 unsigned int real_num_rx_queues;
2090 struct netdev_rx_queue *_rx;
2091 unsigned long gro_flush_timeout;
2092 int napi_defer_hard_irqs;
2093 unsigned int gro_max_size;
2094 unsigned int gro_ipv4_max_size;
2095 rx_handler_func_t __rcu *rx_handler;
2096 void __rcu *rx_handler_data;
2097 possible_net_t nd_net;
2098#ifdef CONFIG_NETPOLL
2099 struct netpoll_info __rcu *npinfo;
2100#endif
2101#ifdef CONFIG_NET_XGRESS
2102 struct bpf_mprog_entry __rcu *tcx_ingress;
2103#endif
2104 __cacheline_group_end(net_device_read_rx);
2105
2106 char name[IFNAMSIZ];
2107 struct netdev_name_node *name_node;
2108 struct dev_ifalias __rcu *ifalias;
2109 /*
2110 * I/O specific fields
2111 * FIXME: Merge these and struct ifmap into one
2112 */
2113 unsigned long mem_end;
2114 unsigned long mem_start;
2115 unsigned long base_addr;
2116
2117 /*
2118 * Some hardware also needs these fields (state,dev_list,
2119 * napi_list,unreg_list,close_list) but they are not
2120 * part of the usual set specified in Space.c.
2121 */
2122
2123
2124 struct list_head dev_list;
2125 struct list_head napi_list;
2126 struct list_head unreg_list;
2127 struct list_head close_list;
2128 struct list_head ptype_all;
2129
2130 struct {
2131 struct list_head upper;
2132 struct list_head lower;
2133 } adj_list;
2134
2135 /* Read-mostly cache-line for fast-path access */
2136 xdp_features_t xdp_features;
2137 const struct xdp_metadata_ops *xdp_metadata_ops;
2138 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2139 unsigned short gflags;
2140
2141 unsigned short needed_tailroom;
2142
2143 netdev_features_t hw_features;
2144 netdev_features_t wanted_features;
2145 netdev_features_t vlan_features;
2146 netdev_features_t hw_enc_features;
2147 netdev_features_t mpls_features;
2148
2149 unsigned int min_mtu;
2150 unsigned int max_mtu;
2151 unsigned short type;
2152 unsigned char min_header_len;
2153 unsigned char name_assign_type;
2154
2155 int group;
2156
2157 struct net_device_stats stats; /* not used by modern drivers */
2158
2159 struct net_device_core_stats __percpu *core_stats;
2160
2161 /* Stats to monitor link on/off, flapping */
2162 atomic_t carrier_up_count;
2163 atomic_t carrier_down_count;
2164
2165#ifdef CONFIG_WIRELESS_EXT
2166 const struct iw_handler_def *wireless_handlers;
2167 struct iw_public_data *wireless_data;
2168#endif
2169 const struct ethtool_ops *ethtool_ops;
2170#ifdef CONFIG_NET_L3_MASTER_DEV
2171 const struct l3mdev_ops *l3mdev_ops;
2172#endif
2173#if IS_ENABLED(CONFIG_IPV6)
2174 const struct ndisc_ops *ndisc_ops;
2175#endif
2176
2177#ifdef CONFIG_XFRM_OFFLOAD
2178 const struct xfrmdev_ops *xfrmdev_ops;
2179#endif
2180
2181#if IS_ENABLED(CONFIG_TLS_DEVICE)
2182 const struct tlsdev_ops *tlsdev_ops;
2183#endif
2184
2185 unsigned int operstate;
2186 unsigned char link_mode;
2187
2188 unsigned char if_port;
2189 unsigned char dma;
2190
2191 /* Interface address info. */
2192 unsigned char perm_addr[MAX_ADDR_LEN];
2193 unsigned char addr_assign_type;
2194 unsigned char addr_len;
2195 unsigned char upper_level;
2196 unsigned char lower_level;
2197
2198 unsigned short neigh_priv_len;
2199 unsigned short dev_id;
2200 unsigned short dev_port;
2201 unsigned short padded;
2202
2203 spinlock_t addr_list_lock;
2204 int irq;
2205
2206 struct netdev_hw_addr_list uc;
2207 struct netdev_hw_addr_list mc;
2208 struct netdev_hw_addr_list dev_addrs;
2209
2210#ifdef CONFIG_SYSFS
2211 struct kset *queues_kset;
2212#endif
2213#ifdef CONFIG_LOCKDEP
2214 struct list_head unlink_list;
2215#endif
2216 unsigned int promiscuity;
2217 unsigned int allmulti;
2218 bool uc_promisc;
2219#ifdef CONFIG_LOCKDEP
2220 unsigned char nested_level;
2221#endif
2222
2223
2224 /* Protocol-specific pointers */
2225 struct in_device __rcu *ip_ptr;
2226#if IS_ENABLED(CONFIG_VLAN_8021Q)
2227 struct vlan_info __rcu *vlan_info;
2228#endif
2229#if IS_ENABLED(CONFIG_NET_DSA)
2230 struct dsa_port *dsa_ptr;
2231#endif
2232#if IS_ENABLED(CONFIG_TIPC)
2233 struct tipc_bearer __rcu *tipc_ptr;
2234#endif
2235#if IS_ENABLED(CONFIG_ATALK)
2236 void *atalk_ptr;
2237#endif
2238#if IS_ENABLED(CONFIG_AX25)
2239 void *ax25_ptr;
2240#endif
2241#if IS_ENABLED(CONFIG_CFG80211)
2242 struct wireless_dev *ieee80211_ptr;
2243#endif
2244#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2245 struct wpan_dev *ieee802154_ptr;
2246#endif
2247#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2248 struct mpls_dev __rcu *mpls_ptr;
2249#endif
2250#if IS_ENABLED(CONFIG_MCTP)
2251 struct mctp_dev __rcu *mctp_ptr;
2252#endif
2253
2254/*
2255 * Cache lines mostly used on receive path (including eth_type_trans())
2256 */
2257 /* Interface address info used in eth_type_trans() */
2258 const unsigned char *dev_addr;
2259
2260 unsigned int num_rx_queues;
2261#define GRO_LEGACY_MAX_SIZE 65536u
2262/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2263 * and shinfo->gso_segs is a 16bit field.
2264 */
2265#define GRO_MAX_SIZE (8 * 65535u)
2266 unsigned int xdp_zc_max_segs;
2267 struct netdev_queue __rcu *ingress_queue;
2268#ifdef CONFIG_NETFILTER_INGRESS
2269 struct nf_hook_entries __rcu *nf_hooks_ingress;
2270#endif
2271
2272 unsigned char broadcast[MAX_ADDR_LEN];
2273#ifdef CONFIG_RFS_ACCEL
2274 struct cpu_rmap *rx_cpu_rmap;
2275#endif
2276 struct hlist_node index_hlist;
2277
2278/*
2279 * Cache lines mostly used on transmit path
2280 */
2281 unsigned int num_tx_queues;
2282 struct Qdisc __rcu *qdisc;
2283 unsigned int tx_queue_len;
2284 spinlock_t tx_global_lock;
2285
2286 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2287
2288#ifdef CONFIG_NET_SCHED
2289 DECLARE_HASHTABLE (qdisc_hash, 4);
2290#endif
2291 /* These may be needed for future network-power-down code. */
2292 struct timer_list watchdog_timer;
2293 int watchdog_timeo;
2294
2295 u32 proto_down_reason;
2296
2297 struct list_head todo_list;
2298
2299#ifdef CONFIG_PCPU_DEV_REFCNT
2300 int __percpu *pcpu_refcnt;
2301#else
2302 refcount_t dev_refcnt;
2303#endif
2304 struct ref_tracker_dir refcnt_tracker;
2305
2306 struct list_head link_watch_list;
2307
2308 u8 reg_state;
2309
2310 bool dismantle;
2311
2312 enum {
2313 RTNL_LINK_INITIALIZED,
2314 RTNL_LINK_INITIALIZING,
2315 } rtnl_link_state:16;
2316
2317 bool needs_free_netdev;
2318 void (*priv_destructor)(struct net_device *dev);
2319
2320 /* mid-layer private */
2321 void *ml_priv;
2322 enum netdev_ml_priv_type ml_priv_type;
2323
2324 enum netdev_stat_type pcpu_stat_type:8;
2325
2326#if IS_ENABLED(CONFIG_GARP)
2327 struct garp_port __rcu *garp_port;
2328#endif
2329#if IS_ENABLED(CONFIG_MRP)
2330 struct mrp_port __rcu *mrp_port;
2331#endif
2332#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2333 struct dm_hw_stat_delta __rcu *dm_private;
2334#endif
2335 struct device dev;
2336 const struct attribute_group *sysfs_groups[4];
2337 const struct attribute_group *sysfs_rx_queue_group;
2338
2339 const struct rtnl_link_ops *rtnl_link_ops;
2340
2341 const struct netdev_stat_ops *stat_ops;
2342
2343 const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2344
2345 /* for setting kernel sock attribute on TCP connection setup */
2346#define GSO_MAX_SEGS 65535u
2347#define GSO_LEGACY_MAX_SIZE 65536u
2348/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2349 * and shinfo->gso_segs is a 16bit field.
2350 */
2351#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2352
2353#define TSO_LEGACY_MAX_SIZE 65536
2354#define TSO_MAX_SIZE UINT_MAX
2355 unsigned int tso_max_size;
2356#define TSO_MAX_SEGS U16_MAX
2357 u16 tso_max_segs;
2358
2359#ifdef CONFIG_DCB
2360 const struct dcbnl_rtnl_ops *dcbnl_ops;
2361#endif
2362 u8 prio_tc_map[TC_BITMASK + 1];
2363
2364#if IS_ENABLED(CONFIG_FCOE)
2365 unsigned int fcoe_ddp_xid;
2366#endif
2367#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2368 struct netprio_map __rcu *priomap;
2369#endif
2370 struct phy_device *phydev;
2371 struct sfp_bus *sfp_bus;
2372 struct lock_class_key *qdisc_tx_busylock;
2373 bool proto_down;
2374 bool threaded;
2375 unsigned wol_enabled:1;
2376
2377 struct list_head net_notifier_list;
2378
2379#if IS_ENABLED(CONFIG_MACSEC)
2380 /* MACsec management functions */
2381 const struct macsec_ops *macsec_ops;
2382#endif
2383 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2384 struct udp_tunnel_nic *udp_tunnel_nic;
2385
2386 /* protected by rtnl_lock */
2387 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2388
2389 u8 dev_addr_shadow[MAX_ADDR_LEN];
2390 netdevice_tracker linkwatch_dev_tracker;
2391 netdevice_tracker watchdog_dev_tracker;
2392 netdevice_tracker dev_registered_tracker;
2393 struct rtnl_hw_stats64 *offload_xstats_l3;
2394
2395 struct devlink_port *devlink_port;
2396
2397#if IS_ENABLED(CONFIG_DPLL)
2398 struct dpll_pin __rcu *dpll_pin;
2399#endif
2400#if IS_ENABLED(CONFIG_PAGE_POOL)
2401 /** @page_pools: page pools created for this netdevice */
2402 struct hlist_head page_pools;
2403#endif
2404};
2405#define to_net_dev(d) container_of(d, struct net_device, dev)
2406
2407/*
2408 * Driver should use this to assign devlink port instance to a netdevice
2409 * before it registers the netdevice. Therefore devlink_port is static
2410 * during the netdev lifetime after it is registered.
2411 */
2412#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2413({ \
2414 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2415 ((dev)->devlink_port = (port)); \
2416})
2417
2418static inline bool netif_elide_gro(const struct net_device *dev)
2419{
2420 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2421 return true;
2422 return false;
2423}
2424
2425#define NETDEV_ALIGN 32
2426
2427static inline
2428int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2429{
2430 return dev->prio_tc_map[prio & TC_BITMASK];
2431}
2432
2433static inline
2434int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2435{
2436 if (tc >= dev->num_tc)
2437 return -EINVAL;
2438
2439 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2440 return 0;
2441}
2442
2443int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2444void netdev_reset_tc(struct net_device *dev);
2445int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2446int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2447
2448static inline
2449int netdev_get_num_tc(struct net_device *dev)
2450{
2451 return dev->num_tc;
2452}
2453
2454static inline void net_prefetch(void *p)
2455{
2456 prefetch(p);
2457#if L1_CACHE_BYTES < 128
2458 prefetch((u8 *)p + L1_CACHE_BYTES);
2459#endif
2460}
2461
2462static inline void net_prefetchw(void *p)
2463{
2464 prefetchw(p);
2465#if L1_CACHE_BYTES < 128
2466 prefetchw((u8 *)p + L1_CACHE_BYTES);
2467#endif
2468}
2469
2470void netdev_unbind_sb_channel(struct net_device *dev,
2471 struct net_device *sb_dev);
2472int netdev_bind_sb_channel_queue(struct net_device *dev,
2473 struct net_device *sb_dev,
2474 u8 tc, u16 count, u16 offset);
2475int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2476static inline int netdev_get_sb_channel(struct net_device *dev)
2477{
2478 return max_t(int, -dev->num_tc, 0);
2479}
2480
2481static inline
2482struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2483 unsigned int index)
2484{
2485 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2486 return &dev->_tx[index];
2487}
2488
2489static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2490 const struct sk_buff *skb)
2491{
2492 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2493}
2494
2495static inline void netdev_for_each_tx_queue(struct net_device *dev,
2496 void (*f)(struct net_device *,
2497 struct netdev_queue *,
2498 void *),
2499 void *arg)
2500{
2501 unsigned int i;
2502
2503 for (i = 0; i < dev->num_tx_queues; i++)
2504 f(dev, &dev->_tx[i], arg);
2505}
2506
2507#define netdev_lockdep_set_classes(dev) \
2508{ \
2509 static struct lock_class_key qdisc_tx_busylock_key; \
2510 static struct lock_class_key qdisc_xmit_lock_key; \
2511 static struct lock_class_key dev_addr_list_lock_key; \
2512 unsigned int i; \
2513 \
2514 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2515 lockdep_set_class(&(dev)->addr_list_lock, \
2516 &dev_addr_list_lock_key); \
2517 for (i = 0; i < (dev)->num_tx_queues; i++) \
2518 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2519 &qdisc_xmit_lock_key); \
2520}
2521
2522u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2523 struct net_device *sb_dev);
2524struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2525 struct sk_buff *skb,
2526 struct net_device *sb_dev);
2527
2528/* returns the headroom that the master device needs to take in account
2529 * when forwarding to this dev
2530 */
2531static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2532{
2533 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2534}
2535
2536static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2537{
2538 if (dev->netdev_ops->ndo_set_rx_headroom)
2539 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2540}
2541
2542/* set the device rx headroom to the dev's default */
2543static inline void netdev_reset_rx_headroom(struct net_device *dev)
2544{
2545 netdev_set_rx_headroom(dev, -1);
2546}
2547
2548static inline void *netdev_get_ml_priv(struct net_device *dev,
2549 enum netdev_ml_priv_type type)
2550{
2551 if (dev->ml_priv_type != type)
2552 return NULL;
2553
2554 return dev->ml_priv;
2555}
2556
2557static inline void netdev_set_ml_priv(struct net_device *dev,
2558 void *ml_priv,
2559 enum netdev_ml_priv_type type)
2560{
2561 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2562 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2563 dev->ml_priv_type, type);
2564 WARN(!dev->ml_priv_type && dev->ml_priv,
2565 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2566
2567 dev->ml_priv = ml_priv;
2568 dev->ml_priv_type = type;
2569}
2570
2571/*
2572 * Net namespace inlines
2573 */
2574static inline
2575struct net *dev_net(const struct net_device *dev)
2576{
2577 return read_pnet(&dev->nd_net);
2578}
2579
2580static inline
2581void dev_net_set(struct net_device *dev, struct net *net)
2582{
2583 write_pnet(&dev->nd_net, net);
2584}
2585
2586/**
2587 * netdev_priv - access network device private data
2588 * @dev: network device
2589 *
2590 * Get network device private data
2591 */
2592static inline void *netdev_priv(const struct net_device *dev)
2593{
2594 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2595}
2596
2597/* Set the sysfs physical device reference for the network logical device
2598 * if set prior to registration will cause a symlink during initialization.
2599 */
2600#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2601
2602/* Set the sysfs device type for the network logical device to allow
2603 * fine-grained identification of different network device types. For
2604 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2605 */
2606#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2607
2608void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2609 enum netdev_queue_type type,
2610 struct napi_struct *napi);
2611
2612static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2613{
2614 napi->irq = irq;
2615}
2616
2617/* Default NAPI poll() weight
2618 * Device drivers are strongly advised to not use bigger value
2619 */
2620#define NAPI_POLL_WEIGHT 64
2621
2622void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2623 int (*poll)(struct napi_struct *, int), int weight);
2624
2625/**
2626 * netif_napi_add() - initialize a NAPI context
2627 * @dev: network device
2628 * @napi: NAPI context
2629 * @poll: polling function
2630 *
2631 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2632 * *any* of the other NAPI-related functions.
2633 */
2634static inline void
2635netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2636 int (*poll)(struct napi_struct *, int))
2637{
2638 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2639}
2640
2641static inline void
2642netif_napi_add_tx_weight(struct net_device *dev,
2643 struct napi_struct *napi,
2644 int (*poll)(struct napi_struct *, int),
2645 int weight)
2646{
2647 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2648 netif_napi_add_weight(dev, napi, poll, weight);
2649}
2650
2651/**
2652 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2653 * @dev: network device
2654 * @napi: NAPI context
2655 * @poll: polling function
2656 *
2657 * This variant of netif_napi_add() should be used from drivers using NAPI
2658 * to exclusively poll a TX queue.
2659 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2660 */
2661static inline void netif_napi_add_tx(struct net_device *dev,
2662 struct napi_struct *napi,
2663 int (*poll)(struct napi_struct *, int))
2664{
2665 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2666}
2667
2668/**
2669 * __netif_napi_del - remove a NAPI context
2670 * @napi: NAPI context
2671 *
2672 * Warning: caller must observe RCU grace period before freeing memory
2673 * containing @napi. Drivers might want to call this helper to combine
2674 * all the needed RCU grace periods into a single one.
2675 */
2676void __netif_napi_del(struct napi_struct *napi);
2677
2678/**
2679 * netif_napi_del - remove a NAPI context
2680 * @napi: NAPI context
2681 *
2682 * netif_napi_del() removes a NAPI context from the network device NAPI list
2683 */
2684static inline void netif_napi_del(struct napi_struct *napi)
2685{
2686 __netif_napi_del(napi);
2687 synchronize_net();
2688}
2689
2690struct packet_type {
2691 __be16 type; /* This is really htons(ether_type). */
2692 bool ignore_outgoing;
2693 struct net_device *dev; /* NULL is wildcarded here */
2694 netdevice_tracker dev_tracker;
2695 int (*func) (struct sk_buff *,
2696 struct net_device *,
2697 struct packet_type *,
2698 struct net_device *);
2699 void (*list_func) (struct list_head *,
2700 struct packet_type *,
2701 struct net_device *);
2702 bool (*id_match)(struct packet_type *ptype,
2703 struct sock *sk);
2704 struct net *af_packet_net;
2705 void *af_packet_priv;
2706 struct list_head list;
2707};
2708
2709struct offload_callbacks {
2710 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2711 netdev_features_t features);
2712 struct sk_buff *(*gro_receive)(struct list_head *head,
2713 struct sk_buff *skb);
2714 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2715};
2716
2717struct packet_offload {
2718 __be16 type; /* This is really htons(ether_type). */
2719 u16 priority;
2720 struct offload_callbacks callbacks;
2721 struct list_head list;
2722};
2723
2724/* often modified stats are per-CPU, other are shared (netdev->stats) */
2725struct pcpu_sw_netstats {
2726 u64_stats_t rx_packets;
2727 u64_stats_t rx_bytes;
2728 u64_stats_t tx_packets;
2729 u64_stats_t tx_bytes;
2730 struct u64_stats_sync syncp;
2731} __aligned(4 * sizeof(u64));
2732
2733struct pcpu_dstats {
2734 u64 rx_packets;
2735 u64 rx_bytes;
2736 u64 rx_drops;
2737 u64 tx_packets;
2738 u64 tx_bytes;
2739 u64 tx_drops;
2740 struct u64_stats_sync syncp;
2741} __aligned(8 * sizeof(u64));
2742
2743struct pcpu_lstats {
2744 u64_stats_t packets;
2745 u64_stats_t bytes;
2746 struct u64_stats_sync syncp;
2747} __aligned(2 * sizeof(u64));
2748
2749void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2750
2751static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2752{
2753 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2754
2755 u64_stats_update_begin(&tstats->syncp);
2756 u64_stats_add(&tstats->rx_bytes, len);
2757 u64_stats_inc(&tstats->rx_packets);
2758 u64_stats_update_end(&tstats->syncp);
2759}
2760
2761static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2762 unsigned int packets,
2763 unsigned int len)
2764{
2765 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2766
2767 u64_stats_update_begin(&tstats->syncp);
2768 u64_stats_add(&tstats->tx_bytes, len);
2769 u64_stats_add(&tstats->tx_packets, packets);
2770 u64_stats_update_end(&tstats->syncp);
2771}
2772
2773static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2774{
2775 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2776
2777 u64_stats_update_begin(&lstats->syncp);
2778 u64_stats_add(&lstats->bytes, len);
2779 u64_stats_inc(&lstats->packets);
2780 u64_stats_update_end(&lstats->syncp);
2781}
2782
2783#define __netdev_alloc_pcpu_stats(type, gfp) \
2784({ \
2785 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2786 if (pcpu_stats) { \
2787 int __cpu; \
2788 for_each_possible_cpu(__cpu) { \
2789 typeof(type) *stat; \
2790 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2791 u64_stats_init(&stat->syncp); \
2792 } \
2793 } \
2794 pcpu_stats; \
2795})
2796
2797#define netdev_alloc_pcpu_stats(type) \
2798 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2799
2800#define devm_netdev_alloc_pcpu_stats(dev, type) \
2801({ \
2802 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2803 if (pcpu_stats) { \
2804 int __cpu; \
2805 for_each_possible_cpu(__cpu) { \
2806 typeof(type) *stat; \
2807 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2808 u64_stats_init(&stat->syncp); \
2809 } \
2810 } \
2811 pcpu_stats; \
2812})
2813
2814enum netdev_lag_tx_type {
2815 NETDEV_LAG_TX_TYPE_UNKNOWN,
2816 NETDEV_LAG_TX_TYPE_RANDOM,
2817 NETDEV_LAG_TX_TYPE_BROADCAST,
2818 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2819 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2820 NETDEV_LAG_TX_TYPE_HASH,
2821};
2822
2823enum netdev_lag_hash {
2824 NETDEV_LAG_HASH_NONE,
2825 NETDEV_LAG_HASH_L2,
2826 NETDEV_LAG_HASH_L34,
2827 NETDEV_LAG_HASH_L23,
2828 NETDEV_LAG_HASH_E23,
2829 NETDEV_LAG_HASH_E34,
2830 NETDEV_LAG_HASH_VLAN_SRCMAC,
2831 NETDEV_LAG_HASH_UNKNOWN,
2832};
2833
2834struct netdev_lag_upper_info {
2835 enum netdev_lag_tx_type tx_type;
2836 enum netdev_lag_hash hash_type;
2837};
2838
2839struct netdev_lag_lower_state_info {
2840 u8 link_up : 1,
2841 tx_enabled : 1;
2842};
2843
2844#include <linux/notifier.h>
2845
2846/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2847 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2848 * adding new types.
2849 */
2850enum netdev_cmd {
2851 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2852 NETDEV_DOWN,
2853 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2854 detected a hardware crash and restarted
2855 - we can use this eg to kick tcp sessions
2856 once done */
2857 NETDEV_CHANGE, /* Notify device state change */
2858 NETDEV_REGISTER,
2859 NETDEV_UNREGISTER,
2860 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2861 NETDEV_CHANGEADDR, /* notify after the address change */
2862 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2863 NETDEV_GOING_DOWN,
2864 NETDEV_CHANGENAME,
2865 NETDEV_FEAT_CHANGE,
2866 NETDEV_BONDING_FAILOVER,
2867 NETDEV_PRE_UP,
2868 NETDEV_PRE_TYPE_CHANGE,
2869 NETDEV_POST_TYPE_CHANGE,
2870 NETDEV_POST_INIT,
2871 NETDEV_PRE_UNINIT,
2872 NETDEV_RELEASE,
2873 NETDEV_NOTIFY_PEERS,
2874 NETDEV_JOIN,
2875 NETDEV_CHANGEUPPER,
2876 NETDEV_RESEND_IGMP,
2877 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2878 NETDEV_CHANGEINFODATA,
2879 NETDEV_BONDING_INFO,
2880 NETDEV_PRECHANGEUPPER,
2881 NETDEV_CHANGELOWERSTATE,
2882 NETDEV_UDP_TUNNEL_PUSH_INFO,
2883 NETDEV_UDP_TUNNEL_DROP_INFO,
2884 NETDEV_CHANGE_TX_QUEUE_LEN,
2885 NETDEV_CVLAN_FILTER_PUSH_INFO,
2886 NETDEV_CVLAN_FILTER_DROP_INFO,
2887 NETDEV_SVLAN_FILTER_PUSH_INFO,
2888 NETDEV_SVLAN_FILTER_DROP_INFO,
2889 NETDEV_OFFLOAD_XSTATS_ENABLE,
2890 NETDEV_OFFLOAD_XSTATS_DISABLE,
2891 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2892 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2893 NETDEV_XDP_FEAT_CHANGE,
2894};
2895const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2896
2897int register_netdevice_notifier(struct notifier_block *nb);
2898int unregister_netdevice_notifier(struct notifier_block *nb);
2899int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2900int unregister_netdevice_notifier_net(struct net *net,
2901 struct notifier_block *nb);
2902int register_netdevice_notifier_dev_net(struct net_device *dev,
2903 struct notifier_block *nb,
2904 struct netdev_net_notifier *nn);
2905int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2906 struct notifier_block *nb,
2907 struct netdev_net_notifier *nn);
2908
2909struct netdev_notifier_info {
2910 struct net_device *dev;
2911 struct netlink_ext_ack *extack;
2912};
2913
2914struct netdev_notifier_info_ext {
2915 struct netdev_notifier_info info; /* must be first */
2916 union {
2917 u32 mtu;
2918 } ext;
2919};
2920
2921struct netdev_notifier_change_info {
2922 struct netdev_notifier_info info; /* must be first */
2923 unsigned int flags_changed;
2924};
2925
2926struct netdev_notifier_changeupper_info {
2927 struct netdev_notifier_info info; /* must be first */
2928 struct net_device *upper_dev; /* new upper dev */
2929 bool master; /* is upper dev master */
2930 bool linking; /* is the notification for link or unlink */
2931 void *upper_info; /* upper dev info */
2932};
2933
2934struct netdev_notifier_changelowerstate_info {
2935 struct netdev_notifier_info info; /* must be first */
2936 void *lower_state_info; /* is lower dev state */
2937};
2938
2939struct netdev_notifier_pre_changeaddr_info {
2940 struct netdev_notifier_info info; /* must be first */
2941 const unsigned char *dev_addr;
2942};
2943
2944enum netdev_offload_xstats_type {
2945 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2946};
2947
2948struct netdev_notifier_offload_xstats_info {
2949 struct netdev_notifier_info info; /* must be first */
2950 enum netdev_offload_xstats_type type;
2951
2952 union {
2953 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2954 struct netdev_notifier_offload_xstats_rd *report_delta;
2955 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2956 struct netdev_notifier_offload_xstats_ru *report_used;
2957 };
2958};
2959
2960int netdev_offload_xstats_enable(struct net_device *dev,
2961 enum netdev_offload_xstats_type type,
2962 struct netlink_ext_ack *extack);
2963int netdev_offload_xstats_disable(struct net_device *dev,
2964 enum netdev_offload_xstats_type type);
2965bool netdev_offload_xstats_enabled(const struct net_device *dev,
2966 enum netdev_offload_xstats_type type);
2967int netdev_offload_xstats_get(struct net_device *dev,
2968 enum netdev_offload_xstats_type type,
2969 struct rtnl_hw_stats64 *stats, bool *used,
2970 struct netlink_ext_ack *extack);
2971void
2972netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2973 const struct rtnl_hw_stats64 *stats);
2974void
2975netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2976void netdev_offload_xstats_push_delta(struct net_device *dev,
2977 enum netdev_offload_xstats_type type,
2978 const struct rtnl_hw_stats64 *stats);
2979
2980static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2981 struct net_device *dev)
2982{
2983 info->dev = dev;
2984 info->extack = NULL;
2985}
2986
2987static inline struct net_device *
2988netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2989{
2990 return info->dev;
2991}
2992
2993static inline struct netlink_ext_ack *
2994netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2995{
2996 return info->extack;
2997}
2998
2999int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3000int call_netdevice_notifiers_info(unsigned long val,
3001 struct netdev_notifier_info *info);
3002
3003#define for_each_netdev(net, d) \
3004 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3005#define for_each_netdev_reverse(net, d) \
3006 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3007#define for_each_netdev_rcu(net, d) \
3008 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3009#define for_each_netdev_safe(net, d, n) \
3010 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3011#define for_each_netdev_continue(net, d) \
3012 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3013#define for_each_netdev_continue_reverse(net, d) \
3014 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3015 dev_list)
3016#define for_each_netdev_continue_rcu(net, d) \
3017 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3018#define for_each_netdev_in_bond_rcu(bond, slave) \
3019 for_each_netdev_rcu(&init_net, slave) \
3020 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3021#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3022
3023#define for_each_netdev_dump(net, d, ifindex) \
3024 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3025
3026static inline struct net_device *next_net_device(struct net_device *dev)
3027{
3028 struct list_head *lh;
3029 struct net *net;
3030
3031 net = dev_net(dev);
3032 lh = dev->dev_list.next;
3033 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3034}
3035
3036static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3037{
3038 struct list_head *lh;
3039 struct net *net;
3040
3041 net = dev_net(dev);
3042 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3043 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3044}
3045
3046static inline struct net_device *first_net_device(struct net *net)
3047{
3048 return list_empty(&net->dev_base_head) ? NULL :
3049 net_device_entry(net->dev_base_head.next);
3050}
3051
3052static inline struct net_device *first_net_device_rcu(struct net *net)
3053{
3054 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3055
3056 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3057}
3058
3059int netdev_boot_setup_check(struct net_device *dev);
3060struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3061 const char *hwaddr);
3062struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3063void dev_add_pack(struct packet_type *pt);
3064void dev_remove_pack(struct packet_type *pt);
3065void __dev_remove_pack(struct packet_type *pt);
3066void dev_add_offload(struct packet_offload *po);
3067void dev_remove_offload(struct packet_offload *po);
3068
3069int dev_get_iflink(const struct net_device *dev);
3070int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3071int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3072 struct net_device_path_stack *stack);
3073struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3074 unsigned short mask);
3075struct net_device *dev_get_by_name(struct net *net, const char *name);
3076struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3077struct net_device *__dev_get_by_name(struct net *net, const char *name);
3078bool netdev_name_in_use(struct net *net, const char *name);
3079int dev_alloc_name(struct net_device *dev, const char *name);
3080int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3081void dev_close(struct net_device *dev);
3082void dev_close_many(struct list_head *head, bool unlink);
3083void dev_disable_lro(struct net_device *dev);
3084int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3085u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3086 struct net_device *sb_dev);
3087u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3088 struct net_device *sb_dev);
3089
3090int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3091int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3092
3093static inline int dev_queue_xmit(struct sk_buff *skb)
3094{
3095 return __dev_queue_xmit(skb, NULL);
3096}
3097
3098static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3099 struct net_device *sb_dev)
3100{
3101 return __dev_queue_xmit(skb, sb_dev);
3102}
3103
3104static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3105{
3106 int ret;
3107
3108 ret = __dev_direct_xmit(skb, queue_id);
3109 if (!dev_xmit_complete(ret))
3110 kfree_skb(skb);
3111 return ret;
3112}
3113
3114int register_netdevice(struct net_device *dev);
3115void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3116void unregister_netdevice_many(struct list_head *head);
3117static inline void unregister_netdevice(struct net_device *dev)
3118{
3119 unregister_netdevice_queue(dev, NULL);
3120}
3121
3122int netdev_refcnt_read(const struct net_device *dev);
3123void free_netdev(struct net_device *dev);
3124void netdev_freemem(struct net_device *dev);
3125void init_dummy_netdev(struct net_device *dev);
3126
3127struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3128 struct sk_buff *skb,
3129 bool all_slaves);
3130struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3131 struct sock *sk);
3132struct net_device *dev_get_by_index(struct net *net, int ifindex);
3133struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3134struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3135 netdevice_tracker *tracker, gfp_t gfp);
3136struct net_device *netdev_get_by_name(struct net *net, const char *name,
3137 netdevice_tracker *tracker, gfp_t gfp);
3138struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3139struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3140void netdev_copy_name(struct net_device *dev, char *name);
3141
3142static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3143 unsigned short type,
3144 const void *daddr, const void *saddr,
3145 unsigned int len)
3146{
3147 if (!dev->header_ops || !dev->header_ops->create)
3148 return 0;
3149
3150 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3151}
3152
3153static inline int dev_parse_header(const struct sk_buff *skb,
3154 unsigned char *haddr)
3155{
3156 const struct net_device *dev = skb->dev;
3157
3158 if (!dev->header_ops || !dev->header_ops->parse)
3159 return 0;
3160 return dev->header_ops->parse(skb, haddr);
3161}
3162
3163static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3164{
3165 const struct net_device *dev = skb->dev;
3166
3167 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3168 return 0;
3169 return dev->header_ops->parse_protocol(skb);
3170}
3171
3172/* ll_header must have at least hard_header_len allocated */
3173static inline bool dev_validate_header(const struct net_device *dev,
3174 char *ll_header, int len)
3175{
3176 if (likely(len >= dev->hard_header_len))
3177 return true;
3178 if (len < dev->min_header_len)
3179 return false;
3180
3181 if (capable(CAP_SYS_RAWIO)) {
3182 memset(ll_header + len, 0, dev->hard_header_len - len);
3183 return true;
3184 }
3185
3186 if (dev->header_ops && dev->header_ops->validate)
3187 return dev->header_ops->validate(ll_header, len);
3188
3189 return false;
3190}
3191
3192static inline bool dev_has_header(const struct net_device *dev)
3193{
3194 return dev->header_ops && dev->header_ops->create;
3195}
3196
3197/*
3198 * Incoming packets are placed on per-CPU queues
3199 */
3200struct softnet_data {
3201 struct list_head poll_list;
3202 struct sk_buff_head process_queue;
3203
3204 /* stats */
3205 unsigned int processed;
3206 unsigned int time_squeeze;
3207#ifdef CONFIG_RPS
3208 struct softnet_data *rps_ipi_list;
3209#endif
3210
3211 unsigned int received_rps;
3212 bool in_net_rx_action;
3213 bool in_napi_threaded_poll;
3214
3215#ifdef CONFIG_NET_FLOW_LIMIT
3216 struct sd_flow_limit __rcu *flow_limit;
3217#endif
3218 struct Qdisc *output_queue;
3219 struct Qdisc **output_queue_tailp;
3220 struct sk_buff *completion_queue;
3221#ifdef CONFIG_XFRM_OFFLOAD
3222 struct sk_buff_head xfrm_backlog;
3223#endif
3224 /* written and read only by owning cpu: */
3225 struct {
3226 u16 recursion;
3227 u8 more;
3228#ifdef CONFIG_NET_EGRESS
3229 u8 skip_txqueue;
3230#endif
3231 } xmit;
3232#ifdef CONFIG_RPS
3233 /* input_queue_head should be written by cpu owning this struct,
3234 * and only read by other cpus. Worth using a cache line.
3235 */
3236 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3237
3238 /* Elements below can be accessed between CPUs for RPS/RFS */
3239 call_single_data_t csd ____cacheline_aligned_in_smp;
3240 struct softnet_data *rps_ipi_next;
3241 unsigned int cpu;
3242 unsigned int input_queue_tail;
3243#endif
3244 struct sk_buff_head input_pkt_queue;
3245 struct napi_struct backlog;
3246
3247 atomic_t dropped ____cacheline_aligned_in_smp;
3248
3249 /* Another possibly contended cache line */
3250 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3251 int defer_count;
3252 int defer_ipi_scheduled;
3253 struct sk_buff *defer_list;
3254 call_single_data_t defer_csd;
3255};
3256
3257DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3258
3259static inline int dev_recursion_level(void)
3260{
3261 return this_cpu_read(softnet_data.xmit.recursion);
3262}
3263
3264void __netif_schedule(struct Qdisc *q);
3265void netif_schedule_queue(struct netdev_queue *txq);
3266
3267static inline void netif_tx_schedule_all(struct net_device *dev)
3268{
3269 unsigned int i;
3270
3271 for (i = 0; i < dev->num_tx_queues; i++)
3272 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3273}
3274
3275static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3276{
3277 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3278}
3279
3280/**
3281 * netif_start_queue - allow transmit
3282 * @dev: network device
3283 *
3284 * Allow upper layers to call the device hard_start_xmit routine.
3285 */
3286static inline void netif_start_queue(struct net_device *dev)
3287{
3288 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3289}
3290
3291static inline void netif_tx_start_all_queues(struct net_device *dev)
3292{
3293 unsigned int i;
3294
3295 for (i = 0; i < dev->num_tx_queues; i++) {
3296 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3297 netif_tx_start_queue(txq);
3298 }
3299}
3300
3301void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3302
3303/**
3304 * netif_wake_queue - restart transmit
3305 * @dev: network device
3306 *
3307 * Allow upper layers to call the device hard_start_xmit routine.
3308 * Used for flow control when transmit resources are available.
3309 */
3310static inline void netif_wake_queue(struct net_device *dev)
3311{
3312 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3313}
3314
3315static inline void netif_tx_wake_all_queues(struct net_device *dev)
3316{
3317 unsigned int i;
3318
3319 for (i = 0; i < dev->num_tx_queues; i++) {
3320 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3321 netif_tx_wake_queue(txq);
3322 }
3323}
3324
3325static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3326{
3327 /* Must be an atomic op see netif_txq_try_stop() */
3328 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3329}
3330
3331/**
3332 * netif_stop_queue - stop transmitted packets
3333 * @dev: network device
3334 *
3335 * Stop upper layers calling the device hard_start_xmit routine.
3336 * Used for flow control when transmit resources are unavailable.
3337 */
3338static inline void netif_stop_queue(struct net_device *dev)
3339{
3340 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3341}
3342
3343void netif_tx_stop_all_queues(struct net_device *dev);
3344
3345static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3346{
3347 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3348}
3349
3350/**
3351 * netif_queue_stopped - test if transmit queue is flowblocked
3352 * @dev: network device
3353 *
3354 * Test if transmit queue on device is currently unable to send.
3355 */
3356static inline bool netif_queue_stopped(const struct net_device *dev)
3357{
3358 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3359}
3360
3361static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3362{
3363 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3364}
3365
3366static inline bool
3367netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3368{
3369 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3370}
3371
3372static inline bool
3373netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3374{
3375 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3376}
3377
3378/**
3379 * netdev_queue_set_dql_min_limit - set dql minimum limit
3380 * @dev_queue: pointer to transmit queue
3381 * @min_limit: dql minimum limit
3382 *
3383 * Forces xmit_more() to return true until the minimum threshold
3384 * defined by @min_limit is reached (or until the tx queue is
3385 * empty). Warning: to be use with care, misuse will impact the
3386 * latency.
3387 */
3388static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3389 unsigned int min_limit)
3390{
3391#ifdef CONFIG_BQL
3392 dev_queue->dql.min_limit = min_limit;
3393#endif
3394}
3395
3396static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3397{
3398#ifdef CONFIG_BQL
3399 /* Non-BQL migrated drivers will return 0, too. */
3400 return dql_avail(&txq->dql);
3401#else
3402 return 0;
3403#endif
3404}
3405
3406/**
3407 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3408 * @dev_queue: pointer to transmit queue
3409 *
3410 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3411 * to give appropriate hint to the CPU.
3412 */
3413static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3414{
3415#ifdef CONFIG_BQL
3416 prefetchw(&dev_queue->dql.num_queued);
3417#endif
3418}
3419
3420/**
3421 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3422 * @dev_queue: pointer to transmit queue
3423 *
3424 * BQL enabled drivers might use this helper in their TX completion path,
3425 * to give appropriate hint to the CPU.
3426 */
3427static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3428{
3429#ifdef CONFIG_BQL
3430 prefetchw(&dev_queue->dql.limit);
3431#endif
3432}
3433
3434/**
3435 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3436 * @dev_queue: network device queue
3437 * @bytes: number of bytes queued to the device queue
3438 *
3439 * Report the number of bytes queued for sending/completion to the network
3440 * device hardware queue. @bytes should be a good approximation and should
3441 * exactly match netdev_completed_queue() @bytes.
3442 * This is typically called once per packet, from ndo_start_xmit().
3443 */
3444static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3445 unsigned int bytes)
3446{
3447#ifdef CONFIG_BQL
3448 dql_queued(&dev_queue->dql, bytes);
3449
3450 if (likely(dql_avail(&dev_queue->dql) >= 0))
3451 return;
3452
3453 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3454
3455 /*
3456 * The XOFF flag must be set before checking the dql_avail below,
3457 * because in netdev_tx_completed_queue we update the dql_completed
3458 * before checking the XOFF flag.
3459 */
3460 smp_mb();
3461
3462 /* check again in case another CPU has just made room avail */
3463 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3464 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3465#endif
3466}
3467
3468/* Variant of netdev_tx_sent_queue() for drivers that are aware
3469 * that they should not test BQL status themselves.
3470 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3471 * skb of a batch.
3472 * Returns true if the doorbell must be used to kick the NIC.
3473 */
3474static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3475 unsigned int bytes,
3476 bool xmit_more)
3477{
3478 if (xmit_more) {
3479#ifdef CONFIG_BQL
3480 dql_queued(&dev_queue->dql, bytes);
3481#endif
3482 return netif_tx_queue_stopped(dev_queue);
3483 }
3484 netdev_tx_sent_queue(dev_queue, bytes);
3485 return true;
3486}
3487
3488/**
3489 * netdev_sent_queue - report the number of bytes queued to hardware
3490 * @dev: network device
3491 * @bytes: number of bytes queued to the hardware device queue
3492 *
3493 * Report the number of bytes queued for sending/completion to the network
3494 * device hardware queue#0. @bytes should be a good approximation and should
3495 * exactly match netdev_completed_queue() @bytes.
3496 * This is typically called once per packet, from ndo_start_xmit().
3497 */
3498static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3499{
3500 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3501}
3502
3503static inline bool __netdev_sent_queue(struct net_device *dev,
3504 unsigned int bytes,
3505 bool xmit_more)
3506{
3507 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3508 xmit_more);
3509}
3510
3511/**
3512 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3513 * @dev_queue: network device queue
3514 * @pkts: number of packets (currently ignored)
3515 * @bytes: number of bytes dequeued from the device queue
3516 *
3517 * Must be called at most once per TX completion round (and not per
3518 * individual packet), so that BQL can adjust its limits appropriately.
3519 */
3520static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3521 unsigned int pkts, unsigned int bytes)
3522{
3523#ifdef CONFIG_BQL
3524 if (unlikely(!bytes))
3525 return;
3526
3527 dql_completed(&dev_queue->dql, bytes);
3528
3529 /*
3530 * Without the memory barrier there is a small possiblity that
3531 * netdev_tx_sent_queue will miss the update and cause the queue to
3532 * be stopped forever
3533 */
3534 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3535
3536 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3537 return;
3538
3539 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3540 netif_schedule_queue(dev_queue);
3541#endif
3542}
3543
3544/**
3545 * netdev_completed_queue - report bytes and packets completed by device
3546 * @dev: network device
3547 * @pkts: actual number of packets sent over the medium
3548 * @bytes: actual number of bytes sent over the medium
3549 *
3550 * Report the number of bytes and packets transmitted by the network device
3551 * hardware queue over the physical medium, @bytes must exactly match the
3552 * @bytes amount passed to netdev_sent_queue()
3553 */
3554static inline void netdev_completed_queue(struct net_device *dev,
3555 unsigned int pkts, unsigned int bytes)
3556{
3557 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3558}
3559
3560static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3561{
3562#ifdef CONFIG_BQL
3563 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3564 dql_reset(&q->dql);
3565#endif
3566}
3567
3568/**
3569 * netdev_reset_queue - reset the packets and bytes count of a network device
3570 * @dev_queue: network device
3571 *
3572 * Reset the bytes and packet count of a network device and clear the
3573 * software flow control OFF bit for this network device
3574 */
3575static inline void netdev_reset_queue(struct net_device *dev_queue)
3576{
3577 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3578}
3579
3580/**
3581 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3582 * @dev: network device
3583 * @queue_index: given tx queue index
3584 *
3585 * Returns 0 if given tx queue index >= number of device tx queues,
3586 * otherwise returns the originally passed tx queue index.
3587 */
3588static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3589{
3590 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3591 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3592 dev->name, queue_index,
3593 dev->real_num_tx_queues);
3594 return 0;
3595 }
3596
3597 return queue_index;
3598}
3599
3600/**
3601 * netif_running - test if up
3602 * @dev: network device
3603 *
3604 * Test if the device has been brought up.
3605 */
3606static inline bool netif_running(const struct net_device *dev)
3607{
3608 return test_bit(__LINK_STATE_START, &dev->state);
3609}
3610
3611/*
3612 * Routines to manage the subqueues on a device. We only need start,
3613 * stop, and a check if it's stopped. All other device management is
3614 * done at the overall netdevice level.
3615 * Also test the device if we're multiqueue.
3616 */
3617
3618/**
3619 * netif_start_subqueue - allow sending packets on subqueue
3620 * @dev: network device
3621 * @queue_index: sub queue index
3622 *
3623 * Start individual transmit queue of a device with multiple transmit queues.
3624 */
3625static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3626{
3627 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3628
3629 netif_tx_start_queue(txq);
3630}
3631
3632/**
3633 * netif_stop_subqueue - stop sending packets on subqueue
3634 * @dev: network device
3635 * @queue_index: sub queue index
3636 *
3637 * Stop individual transmit queue of a device with multiple transmit queues.
3638 */
3639static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3640{
3641 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3642 netif_tx_stop_queue(txq);
3643}
3644
3645/**
3646 * __netif_subqueue_stopped - test status of subqueue
3647 * @dev: network device
3648 * @queue_index: sub queue index
3649 *
3650 * Check individual transmit queue of a device with multiple transmit queues.
3651 */
3652static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3653 u16 queue_index)
3654{
3655 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3656
3657 return netif_tx_queue_stopped(txq);
3658}
3659
3660/**
3661 * netif_subqueue_stopped - test status of subqueue
3662 * @dev: network device
3663 * @skb: sub queue buffer pointer
3664 *
3665 * Check individual transmit queue of a device with multiple transmit queues.
3666 */
3667static inline bool netif_subqueue_stopped(const struct net_device *dev,
3668 struct sk_buff *skb)
3669{
3670 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3671}
3672
3673/**
3674 * netif_wake_subqueue - allow sending packets on subqueue
3675 * @dev: network device
3676 * @queue_index: sub queue index
3677 *
3678 * Resume individual transmit queue of a device with multiple transmit queues.
3679 */
3680static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3681{
3682 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3683
3684 netif_tx_wake_queue(txq);
3685}
3686
3687#ifdef CONFIG_XPS
3688int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3689 u16 index);
3690int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3691 u16 index, enum xps_map_type type);
3692
3693/**
3694 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3695 * @j: CPU/Rx queue index
3696 * @mask: bitmask of all cpus/rx queues
3697 * @nr_bits: number of bits in the bitmask
3698 *
3699 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3700 */
3701static inline bool netif_attr_test_mask(unsigned long j,
3702 const unsigned long *mask,
3703 unsigned int nr_bits)
3704{
3705 cpu_max_bits_warn(j, nr_bits);
3706 return test_bit(j, mask);
3707}
3708
3709/**
3710 * netif_attr_test_online - Test for online CPU/Rx queue
3711 * @j: CPU/Rx queue index
3712 * @online_mask: bitmask for CPUs/Rx queues that are online
3713 * @nr_bits: number of bits in the bitmask
3714 *
3715 * Returns true if a CPU/Rx queue is online.
3716 */
3717static inline bool netif_attr_test_online(unsigned long j,
3718 const unsigned long *online_mask,
3719 unsigned int nr_bits)
3720{
3721 cpu_max_bits_warn(j, nr_bits);
3722
3723 if (online_mask)
3724 return test_bit(j, online_mask);
3725
3726 return (j < nr_bits);
3727}
3728
3729/**
3730 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3731 * @n: CPU/Rx queue index
3732 * @srcp: the cpumask/Rx queue mask pointer
3733 * @nr_bits: number of bits in the bitmask
3734 *
3735 * Returns >= nr_bits if no further CPUs/Rx queues set.
3736 */
3737static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3738 unsigned int nr_bits)
3739{
3740 /* -1 is a legal arg here. */
3741 if (n != -1)
3742 cpu_max_bits_warn(n, nr_bits);
3743
3744 if (srcp)
3745 return find_next_bit(srcp, nr_bits, n + 1);
3746
3747 return n + 1;
3748}
3749
3750/**
3751 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3752 * @n: CPU/Rx queue index
3753 * @src1p: the first CPUs/Rx queues mask pointer
3754 * @src2p: the second CPUs/Rx queues mask pointer
3755 * @nr_bits: number of bits in the bitmask
3756 *
3757 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3758 */
3759static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3760 const unsigned long *src2p,
3761 unsigned int nr_bits)
3762{
3763 /* -1 is a legal arg here. */
3764 if (n != -1)
3765 cpu_max_bits_warn(n, nr_bits);
3766
3767 if (src1p && src2p)
3768 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3769 else if (src1p)
3770 return find_next_bit(src1p, nr_bits, n + 1);
3771 else if (src2p)
3772 return find_next_bit(src2p, nr_bits, n + 1);
3773
3774 return n + 1;
3775}
3776#else
3777static inline int netif_set_xps_queue(struct net_device *dev,
3778 const struct cpumask *mask,
3779 u16 index)
3780{
3781 return 0;
3782}
3783
3784static inline int __netif_set_xps_queue(struct net_device *dev,
3785 const unsigned long *mask,
3786 u16 index, enum xps_map_type type)
3787{
3788 return 0;
3789}
3790#endif
3791
3792/**
3793 * netif_is_multiqueue - test if device has multiple transmit queues
3794 * @dev: network device
3795 *
3796 * Check if device has multiple transmit queues
3797 */
3798static inline bool netif_is_multiqueue(const struct net_device *dev)
3799{
3800 return dev->num_tx_queues > 1;
3801}
3802
3803int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3804
3805#ifdef CONFIG_SYSFS
3806int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3807#else
3808static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3809 unsigned int rxqs)
3810{
3811 dev->real_num_rx_queues = rxqs;
3812 return 0;
3813}
3814#endif
3815int netif_set_real_num_queues(struct net_device *dev,
3816 unsigned int txq, unsigned int rxq);
3817
3818int netif_get_num_default_rss_queues(void);
3819
3820void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3821void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3822
3823/*
3824 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3825 * interrupt context or with hardware interrupts being disabled.
3826 * (in_hardirq() || irqs_disabled())
3827 *
3828 * We provide four helpers that can be used in following contexts :
3829 *
3830 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3831 * replacing kfree_skb(skb)
3832 *
3833 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3834 * Typically used in place of consume_skb(skb) in TX completion path
3835 *
3836 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3837 * replacing kfree_skb(skb)
3838 *
3839 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3840 * and consumed a packet. Used in place of consume_skb(skb)
3841 */
3842static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3843{
3844 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3845}
3846
3847static inline void dev_consume_skb_irq(struct sk_buff *skb)
3848{
3849 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3850}
3851
3852static inline void dev_kfree_skb_any(struct sk_buff *skb)
3853{
3854 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3855}
3856
3857static inline void dev_consume_skb_any(struct sk_buff *skb)
3858{
3859 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3860}
3861
3862u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3863 struct bpf_prog *xdp_prog);
3864void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3865int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3866int netif_rx(struct sk_buff *skb);
3867int __netif_rx(struct sk_buff *skb);
3868
3869int netif_receive_skb(struct sk_buff *skb);
3870int netif_receive_skb_core(struct sk_buff *skb);
3871void netif_receive_skb_list_internal(struct list_head *head);
3872void netif_receive_skb_list(struct list_head *head);
3873gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3874void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3875struct sk_buff *napi_get_frags(struct napi_struct *napi);
3876void napi_get_frags_check(struct napi_struct *napi);
3877gro_result_t napi_gro_frags(struct napi_struct *napi);
3878
3879static inline void napi_free_frags(struct napi_struct *napi)
3880{
3881 kfree_skb(napi->skb);
3882 napi->skb = NULL;
3883}
3884
3885bool netdev_is_rx_handler_busy(struct net_device *dev);
3886int netdev_rx_handler_register(struct net_device *dev,
3887 rx_handler_func_t *rx_handler,
3888 void *rx_handler_data);
3889void netdev_rx_handler_unregister(struct net_device *dev);
3890
3891bool dev_valid_name(const char *name);
3892static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3893{
3894 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3895}
3896int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3897int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3898int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3899 void __user *data, bool *need_copyout);
3900int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3901int generic_hwtstamp_get_lower(struct net_device *dev,
3902 struct kernel_hwtstamp_config *kernel_cfg);
3903int generic_hwtstamp_set_lower(struct net_device *dev,
3904 struct kernel_hwtstamp_config *kernel_cfg,
3905 struct netlink_ext_ack *extack);
3906int dev_set_hwtstamp_phylib(struct net_device *dev,
3907 struct kernel_hwtstamp_config *cfg,
3908 struct netlink_ext_ack *extack);
3909int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3910unsigned int dev_get_flags(const struct net_device *);
3911int __dev_change_flags(struct net_device *dev, unsigned int flags,
3912 struct netlink_ext_ack *extack);
3913int dev_change_flags(struct net_device *dev, unsigned int flags,
3914 struct netlink_ext_ack *extack);
3915int dev_set_alias(struct net_device *, const char *, size_t);
3916int dev_get_alias(const struct net_device *, char *, size_t);
3917int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3918 const char *pat, int new_ifindex);
3919static inline
3920int dev_change_net_namespace(struct net_device *dev, struct net *net,
3921 const char *pat)
3922{
3923 return __dev_change_net_namespace(dev, net, pat, 0);
3924}
3925int __dev_set_mtu(struct net_device *, int);
3926int dev_set_mtu(struct net_device *, int);
3927int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3928 struct netlink_ext_ack *extack);
3929int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3930 struct netlink_ext_ack *extack);
3931int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3932 struct netlink_ext_ack *extack);
3933int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3934int dev_get_port_parent_id(struct net_device *dev,
3935 struct netdev_phys_item_id *ppid, bool recurse);
3936bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3937
3938struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3939struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3940 struct netdev_queue *txq, int *ret);
3941
3942int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3943u8 dev_xdp_prog_count(struct net_device *dev);
3944u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3945
3946int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3947int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3948int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3949bool is_skb_forwardable(const struct net_device *dev,
3950 const struct sk_buff *skb);
3951
3952static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3953 const struct sk_buff *skb,
3954 const bool check_mtu)
3955{
3956 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3957 unsigned int len;
3958
3959 if (!(dev->flags & IFF_UP))
3960 return false;
3961
3962 if (!check_mtu)
3963 return true;
3964
3965 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3966 if (skb->len <= len)
3967 return true;
3968
3969 /* if TSO is enabled, we don't care about the length as the packet
3970 * could be forwarded without being segmented before
3971 */
3972 if (skb_is_gso(skb))
3973 return true;
3974
3975 return false;
3976}
3977
3978void netdev_core_stats_inc(struct net_device *dev, u32 offset);
3979
3980#define DEV_CORE_STATS_INC(FIELD) \
3981static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3982{ \
3983 netdev_core_stats_inc(dev, \
3984 offsetof(struct net_device_core_stats, FIELD)); \
3985}
3986DEV_CORE_STATS_INC(rx_dropped)
3987DEV_CORE_STATS_INC(tx_dropped)
3988DEV_CORE_STATS_INC(rx_nohandler)
3989DEV_CORE_STATS_INC(rx_otherhost_dropped)
3990#undef DEV_CORE_STATS_INC
3991
3992static __always_inline int ____dev_forward_skb(struct net_device *dev,
3993 struct sk_buff *skb,
3994 const bool check_mtu)
3995{
3996 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3997 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3998 dev_core_stats_rx_dropped_inc(dev);
3999 kfree_skb(skb);
4000 return NET_RX_DROP;
4001 }
4002
4003 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4004 skb->priority = 0;
4005 return 0;
4006}
4007
4008bool dev_nit_active(struct net_device *dev);
4009void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4010
4011static inline void __dev_put(struct net_device *dev)
4012{
4013 if (dev) {
4014#ifdef CONFIG_PCPU_DEV_REFCNT
4015 this_cpu_dec(*dev->pcpu_refcnt);
4016#else
4017 refcount_dec(&dev->dev_refcnt);
4018#endif
4019 }
4020}
4021
4022static inline void __dev_hold(struct net_device *dev)
4023{
4024 if (dev) {
4025#ifdef CONFIG_PCPU_DEV_REFCNT
4026 this_cpu_inc(*dev->pcpu_refcnt);
4027#else
4028 refcount_inc(&dev->dev_refcnt);
4029#endif
4030 }
4031}
4032
4033static inline void __netdev_tracker_alloc(struct net_device *dev,
4034 netdevice_tracker *tracker,
4035 gfp_t gfp)
4036{
4037#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4038 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4039#endif
4040}
4041
4042/* netdev_tracker_alloc() can upgrade a prior untracked reference
4043 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4044 */
4045static inline void netdev_tracker_alloc(struct net_device *dev,
4046 netdevice_tracker *tracker, gfp_t gfp)
4047{
4048#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4049 refcount_dec(&dev->refcnt_tracker.no_tracker);
4050 __netdev_tracker_alloc(dev, tracker, gfp);
4051#endif
4052}
4053
4054static inline void netdev_tracker_free(struct net_device *dev,
4055 netdevice_tracker *tracker)
4056{
4057#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4058 ref_tracker_free(&dev->refcnt_tracker, tracker);
4059#endif
4060}
4061
4062static inline void netdev_hold(struct net_device *dev,
4063 netdevice_tracker *tracker, gfp_t gfp)
4064{
4065 if (dev) {
4066 __dev_hold(dev);
4067 __netdev_tracker_alloc(dev, tracker, gfp);
4068 }
4069}
4070
4071static inline void netdev_put(struct net_device *dev,
4072 netdevice_tracker *tracker)
4073{
4074 if (dev) {
4075 netdev_tracker_free(dev, tracker);
4076 __dev_put(dev);
4077 }
4078}
4079
4080/**
4081 * dev_hold - get reference to device
4082 * @dev: network device
4083 *
4084 * Hold reference to device to keep it from being freed.
4085 * Try using netdev_hold() instead.
4086 */
4087static inline void dev_hold(struct net_device *dev)
4088{
4089 netdev_hold(dev, NULL, GFP_ATOMIC);
4090}
4091
4092/**
4093 * dev_put - release reference to device
4094 * @dev: network device
4095 *
4096 * Release reference to device to allow it to be freed.
4097 * Try using netdev_put() instead.
4098 */
4099static inline void dev_put(struct net_device *dev)
4100{
4101 netdev_put(dev, NULL);
4102}
4103
4104DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4105
4106static inline void netdev_ref_replace(struct net_device *odev,
4107 struct net_device *ndev,
4108 netdevice_tracker *tracker,
4109 gfp_t gfp)
4110{
4111 if (odev)
4112 netdev_tracker_free(odev, tracker);
4113
4114 __dev_hold(ndev);
4115 __dev_put(odev);
4116
4117 if (ndev)
4118 __netdev_tracker_alloc(ndev, tracker, gfp);
4119}
4120
4121/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4122 * and _off may be called from IRQ context, but it is caller
4123 * who is responsible for serialization of these calls.
4124 *
4125 * The name carrier is inappropriate, these functions should really be
4126 * called netif_lowerlayer_*() because they represent the state of any
4127 * kind of lower layer not just hardware media.
4128 */
4129void linkwatch_fire_event(struct net_device *dev);
4130
4131/**
4132 * linkwatch_sync_dev - sync linkwatch for the given device
4133 * @dev: network device to sync linkwatch for
4134 *
4135 * Sync linkwatch for the given device, removing it from the
4136 * pending work list (if queued).
4137 */
4138void linkwatch_sync_dev(struct net_device *dev);
4139
4140/**
4141 * netif_carrier_ok - test if carrier present
4142 * @dev: network device
4143 *
4144 * Check if carrier is present on device
4145 */
4146static inline bool netif_carrier_ok(const struct net_device *dev)
4147{
4148 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4149}
4150
4151unsigned long dev_trans_start(struct net_device *dev);
4152
4153void __netdev_watchdog_up(struct net_device *dev);
4154
4155void netif_carrier_on(struct net_device *dev);
4156void netif_carrier_off(struct net_device *dev);
4157void netif_carrier_event(struct net_device *dev);
4158
4159/**
4160 * netif_dormant_on - mark device as dormant.
4161 * @dev: network device
4162 *
4163 * Mark device as dormant (as per RFC2863).
4164 *
4165 * The dormant state indicates that the relevant interface is not
4166 * actually in a condition to pass packets (i.e., it is not 'up') but is
4167 * in a "pending" state, waiting for some external event. For "on-
4168 * demand" interfaces, this new state identifies the situation where the
4169 * interface is waiting for events to place it in the up state.
4170 */
4171static inline void netif_dormant_on(struct net_device *dev)
4172{
4173 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4174 linkwatch_fire_event(dev);
4175}
4176
4177/**
4178 * netif_dormant_off - set device as not dormant.
4179 * @dev: network device
4180 *
4181 * Device is not in dormant state.
4182 */
4183static inline void netif_dormant_off(struct net_device *dev)
4184{
4185 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4186 linkwatch_fire_event(dev);
4187}
4188
4189/**
4190 * netif_dormant - test if device is dormant
4191 * @dev: network device
4192 *
4193 * Check if device is dormant.
4194 */
4195static inline bool netif_dormant(const struct net_device *dev)
4196{
4197 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4198}
4199
4200
4201/**
4202 * netif_testing_on - mark device as under test.
4203 * @dev: network device
4204 *
4205 * Mark device as under test (as per RFC2863).
4206 *
4207 * The testing state indicates that some test(s) must be performed on
4208 * the interface. After completion, of the test, the interface state
4209 * will change to up, dormant, or down, as appropriate.
4210 */
4211static inline void netif_testing_on(struct net_device *dev)
4212{
4213 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4214 linkwatch_fire_event(dev);
4215}
4216
4217/**
4218 * netif_testing_off - set device as not under test.
4219 * @dev: network device
4220 *
4221 * Device is not in testing state.
4222 */
4223static inline void netif_testing_off(struct net_device *dev)
4224{
4225 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4226 linkwatch_fire_event(dev);
4227}
4228
4229/**
4230 * netif_testing - test if device is under test
4231 * @dev: network device
4232 *
4233 * Check if device is under test
4234 */
4235static inline bool netif_testing(const struct net_device *dev)
4236{
4237 return test_bit(__LINK_STATE_TESTING, &dev->state);
4238}
4239
4240
4241/**
4242 * netif_oper_up - test if device is operational
4243 * @dev: network device
4244 *
4245 * Check if carrier is operational
4246 */
4247static inline bool netif_oper_up(const struct net_device *dev)
4248{
4249 unsigned int operstate = READ_ONCE(dev->operstate);
4250
4251 return operstate == IF_OPER_UP ||
4252 operstate == IF_OPER_UNKNOWN /* backward compat */;
4253}
4254
4255/**
4256 * netif_device_present - is device available or removed
4257 * @dev: network device
4258 *
4259 * Check if device has not been removed from system.
4260 */
4261static inline bool netif_device_present(const struct net_device *dev)
4262{
4263 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4264}
4265
4266void netif_device_detach(struct net_device *dev);
4267
4268void netif_device_attach(struct net_device *dev);
4269
4270/*
4271 * Network interface message level settings
4272 */
4273
4274enum {
4275 NETIF_MSG_DRV_BIT,
4276 NETIF_MSG_PROBE_BIT,
4277 NETIF_MSG_LINK_BIT,
4278 NETIF_MSG_TIMER_BIT,
4279 NETIF_MSG_IFDOWN_BIT,
4280 NETIF_MSG_IFUP_BIT,
4281 NETIF_MSG_RX_ERR_BIT,
4282 NETIF_MSG_TX_ERR_BIT,
4283 NETIF_MSG_TX_QUEUED_BIT,
4284 NETIF_MSG_INTR_BIT,
4285 NETIF_MSG_TX_DONE_BIT,
4286 NETIF_MSG_RX_STATUS_BIT,
4287 NETIF_MSG_PKTDATA_BIT,
4288 NETIF_MSG_HW_BIT,
4289 NETIF_MSG_WOL_BIT,
4290
4291 /* When you add a new bit above, update netif_msg_class_names array
4292 * in net/ethtool/common.c
4293 */
4294 NETIF_MSG_CLASS_COUNT,
4295};
4296/* Both ethtool_ops interface and internal driver implementation use u32 */
4297static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4298
4299#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4300#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4301
4302#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4303#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4304#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4305#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4306#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4307#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4308#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4309#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4310#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4311#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4312#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4313#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4314#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4315#define NETIF_MSG_HW __NETIF_MSG(HW)
4316#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4317
4318#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4319#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4320#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4321#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4322#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4323#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4324#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4325#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4326#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4327#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4328#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4329#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4330#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4331#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4332#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4333
4334static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4335{
4336 /* use default */
4337 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4338 return default_msg_enable_bits;
4339 if (debug_value == 0) /* no output */
4340 return 0;
4341 /* set low N bits */
4342 return (1U << debug_value) - 1;
4343}
4344
4345static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4346{
4347 spin_lock(&txq->_xmit_lock);
4348 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4349 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4350}
4351
4352static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4353{
4354 __acquire(&txq->_xmit_lock);
4355 return true;
4356}
4357
4358static inline void __netif_tx_release(struct netdev_queue *txq)
4359{
4360 __release(&txq->_xmit_lock);
4361}
4362
4363static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4364{
4365 spin_lock_bh(&txq->_xmit_lock);
4366 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4367 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4368}
4369
4370static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4371{
4372 bool ok = spin_trylock(&txq->_xmit_lock);
4373
4374 if (likely(ok)) {
4375 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4376 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4377 }
4378 return ok;
4379}
4380
4381static inline void __netif_tx_unlock(struct netdev_queue *txq)
4382{
4383 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4384 WRITE_ONCE(txq->xmit_lock_owner, -1);
4385 spin_unlock(&txq->_xmit_lock);
4386}
4387
4388static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4389{
4390 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4391 WRITE_ONCE(txq->xmit_lock_owner, -1);
4392 spin_unlock_bh(&txq->_xmit_lock);
4393}
4394
4395/*
4396 * txq->trans_start can be read locklessly from dev_watchdog()
4397 */
4398static inline void txq_trans_update(struct netdev_queue *txq)
4399{
4400 if (txq->xmit_lock_owner != -1)
4401 WRITE_ONCE(txq->trans_start, jiffies);
4402}
4403
4404static inline void txq_trans_cond_update(struct netdev_queue *txq)
4405{
4406 unsigned long now = jiffies;
4407
4408 if (READ_ONCE(txq->trans_start) != now)
4409 WRITE_ONCE(txq->trans_start, now);
4410}
4411
4412/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4413static inline void netif_trans_update(struct net_device *dev)
4414{
4415 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4416
4417 txq_trans_cond_update(txq);
4418}
4419
4420/**
4421 * netif_tx_lock - grab network device transmit lock
4422 * @dev: network device
4423 *
4424 * Get network device transmit lock
4425 */
4426void netif_tx_lock(struct net_device *dev);
4427
4428static inline void netif_tx_lock_bh(struct net_device *dev)
4429{
4430 local_bh_disable();
4431 netif_tx_lock(dev);
4432}
4433
4434void netif_tx_unlock(struct net_device *dev);
4435
4436static inline void netif_tx_unlock_bh(struct net_device *dev)
4437{
4438 netif_tx_unlock(dev);
4439 local_bh_enable();
4440}
4441
4442#define HARD_TX_LOCK(dev, txq, cpu) { \
4443 if ((dev->features & NETIF_F_LLTX) == 0) { \
4444 __netif_tx_lock(txq, cpu); \
4445 } else { \
4446 __netif_tx_acquire(txq); \
4447 } \
4448}
4449
4450#define HARD_TX_TRYLOCK(dev, txq) \
4451 (((dev->features & NETIF_F_LLTX) == 0) ? \
4452 __netif_tx_trylock(txq) : \
4453 __netif_tx_acquire(txq))
4454
4455#define HARD_TX_UNLOCK(dev, txq) { \
4456 if ((dev->features & NETIF_F_LLTX) == 0) { \
4457 __netif_tx_unlock(txq); \
4458 } else { \
4459 __netif_tx_release(txq); \
4460 } \
4461}
4462
4463static inline void netif_tx_disable(struct net_device *dev)
4464{
4465 unsigned int i;
4466 int cpu;
4467
4468 local_bh_disable();
4469 cpu = smp_processor_id();
4470 spin_lock(&dev->tx_global_lock);
4471 for (i = 0; i < dev->num_tx_queues; i++) {
4472 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4473
4474 __netif_tx_lock(txq, cpu);
4475 netif_tx_stop_queue(txq);
4476 __netif_tx_unlock(txq);
4477 }
4478 spin_unlock(&dev->tx_global_lock);
4479 local_bh_enable();
4480}
4481
4482static inline void netif_addr_lock(struct net_device *dev)
4483{
4484 unsigned char nest_level = 0;
4485
4486#ifdef CONFIG_LOCKDEP
4487 nest_level = dev->nested_level;
4488#endif
4489 spin_lock_nested(&dev->addr_list_lock, nest_level);
4490}
4491
4492static inline void netif_addr_lock_bh(struct net_device *dev)
4493{
4494 unsigned char nest_level = 0;
4495
4496#ifdef CONFIG_LOCKDEP
4497 nest_level = dev->nested_level;
4498#endif
4499 local_bh_disable();
4500 spin_lock_nested(&dev->addr_list_lock, nest_level);
4501}
4502
4503static inline void netif_addr_unlock(struct net_device *dev)
4504{
4505 spin_unlock(&dev->addr_list_lock);
4506}
4507
4508static inline void netif_addr_unlock_bh(struct net_device *dev)
4509{
4510 spin_unlock_bh(&dev->addr_list_lock);
4511}
4512
4513/*
4514 * dev_addrs walker. Should be used only for read access. Call with
4515 * rcu_read_lock held.
4516 */
4517#define for_each_dev_addr(dev, ha) \
4518 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4519
4520/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4521
4522void ether_setup(struct net_device *dev);
4523
4524/* Allocate dummy net_device */
4525struct net_device *alloc_netdev_dummy(int sizeof_priv);
4526
4527/* Support for loadable net-drivers */
4528struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4529 unsigned char name_assign_type,
4530 void (*setup)(struct net_device *),
4531 unsigned int txqs, unsigned int rxqs);
4532#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4533 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4534
4535#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4536 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4537 count)
4538
4539int register_netdev(struct net_device *dev);
4540void unregister_netdev(struct net_device *dev);
4541
4542int devm_register_netdev(struct device *dev, struct net_device *ndev);
4543
4544/* General hardware address lists handling functions */
4545int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4546 struct netdev_hw_addr_list *from_list, int addr_len);
4547void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4548 struct netdev_hw_addr_list *from_list, int addr_len);
4549int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4550 struct net_device *dev,
4551 int (*sync)(struct net_device *, const unsigned char *),
4552 int (*unsync)(struct net_device *,
4553 const unsigned char *));
4554int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4555 struct net_device *dev,
4556 int (*sync)(struct net_device *,
4557 const unsigned char *, int),
4558 int (*unsync)(struct net_device *,
4559 const unsigned char *, int));
4560void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4561 struct net_device *dev,
4562 int (*unsync)(struct net_device *,
4563 const unsigned char *, int));
4564void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4565 struct net_device *dev,
4566 int (*unsync)(struct net_device *,
4567 const unsigned char *));
4568void __hw_addr_init(struct netdev_hw_addr_list *list);
4569
4570/* Functions used for device addresses handling */
4571void dev_addr_mod(struct net_device *dev, unsigned int offset,
4572 const void *addr, size_t len);
4573
4574static inline void
4575__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4576{
4577 dev_addr_mod(dev, 0, addr, len);
4578}
4579
4580static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4581{
4582 __dev_addr_set(dev, addr, dev->addr_len);
4583}
4584
4585int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4586 unsigned char addr_type);
4587int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4588 unsigned char addr_type);
4589
4590/* Functions used for unicast addresses handling */
4591int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4592int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4593int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4594int dev_uc_sync(struct net_device *to, struct net_device *from);
4595int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4596void dev_uc_unsync(struct net_device *to, struct net_device *from);
4597void dev_uc_flush(struct net_device *dev);
4598void dev_uc_init(struct net_device *dev);
4599
4600/**
4601 * __dev_uc_sync - Synchonize device's unicast list
4602 * @dev: device to sync
4603 * @sync: function to call if address should be added
4604 * @unsync: function to call if address should be removed
4605 *
4606 * Add newly added addresses to the interface, and release
4607 * addresses that have been deleted.
4608 */
4609static inline int __dev_uc_sync(struct net_device *dev,
4610 int (*sync)(struct net_device *,
4611 const unsigned char *),
4612 int (*unsync)(struct net_device *,
4613 const unsigned char *))
4614{
4615 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4616}
4617
4618/**
4619 * __dev_uc_unsync - Remove synchronized addresses from device
4620 * @dev: device to sync
4621 * @unsync: function to call if address should be removed
4622 *
4623 * Remove all addresses that were added to the device by dev_uc_sync().
4624 */
4625static inline void __dev_uc_unsync(struct net_device *dev,
4626 int (*unsync)(struct net_device *,
4627 const unsigned char *))
4628{
4629 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4630}
4631
4632/* Functions used for multicast addresses handling */
4633int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4634int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4635int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4636int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4637int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4638int dev_mc_sync(struct net_device *to, struct net_device *from);
4639int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4640void dev_mc_unsync(struct net_device *to, struct net_device *from);
4641void dev_mc_flush(struct net_device *dev);
4642void dev_mc_init(struct net_device *dev);
4643
4644/**
4645 * __dev_mc_sync - Synchonize device's multicast list
4646 * @dev: device to sync
4647 * @sync: function to call if address should be added
4648 * @unsync: function to call if address should be removed
4649 *
4650 * Add newly added addresses to the interface, and release
4651 * addresses that have been deleted.
4652 */
4653static inline int __dev_mc_sync(struct net_device *dev,
4654 int (*sync)(struct net_device *,
4655 const unsigned char *),
4656 int (*unsync)(struct net_device *,
4657 const unsigned char *))
4658{
4659 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4660}
4661
4662/**
4663 * __dev_mc_unsync - Remove synchronized addresses from device
4664 * @dev: device to sync
4665 * @unsync: function to call if address should be removed
4666 *
4667 * Remove all addresses that were added to the device by dev_mc_sync().
4668 */
4669static inline void __dev_mc_unsync(struct net_device *dev,
4670 int (*unsync)(struct net_device *,
4671 const unsigned char *))
4672{
4673 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4674}
4675
4676/* Functions used for secondary unicast and multicast support */
4677void dev_set_rx_mode(struct net_device *dev);
4678int dev_set_promiscuity(struct net_device *dev, int inc);
4679int dev_set_allmulti(struct net_device *dev, int inc);
4680void netdev_state_change(struct net_device *dev);
4681void __netdev_notify_peers(struct net_device *dev);
4682void netdev_notify_peers(struct net_device *dev);
4683void netdev_features_change(struct net_device *dev);
4684/* Load a device via the kmod */
4685void dev_load(struct net *net, const char *name);
4686struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4687 struct rtnl_link_stats64 *storage);
4688void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4689 const struct net_device_stats *netdev_stats);
4690void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4691 const struct pcpu_sw_netstats __percpu *netstats);
4692void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4693
4694enum {
4695 NESTED_SYNC_IMM_BIT,
4696 NESTED_SYNC_TODO_BIT,
4697};
4698
4699#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4700#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4701
4702#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4703#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4704
4705struct netdev_nested_priv {
4706 unsigned char flags;
4707 void *data;
4708};
4709
4710bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4711struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4712 struct list_head **iter);
4713
4714/* iterate through upper list, must be called under RCU read lock */
4715#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4716 for (iter = &(dev)->adj_list.upper, \
4717 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4718 updev; \
4719 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4720
4721int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4722 int (*fn)(struct net_device *upper_dev,
4723 struct netdev_nested_priv *priv),
4724 struct netdev_nested_priv *priv);
4725
4726bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4727 struct net_device *upper_dev);
4728
4729bool netdev_has_any_upper_dev(struct net_device *dev);
4730
4731void *netdev_lower_get_next_private(struct net_device *dev,
4732 struct list_head **iter);
4733void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4734 struct list_head **iter);
4735
4736#define netdev_for_each_lower_private(dev, priv, iter) \
4737 for (iter = (dev)->adj_list.lower.next, \
4738 priv = netdev_lower_get_next_private(dev, &(iter)); \
4739 priv; \
4740 priv = netdev_lower_get_next_private(dev, &(iter)))
4741
4742#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4743 for (iter = &(dev)->adj_list.lower, \
4744 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4745 priv; \
4746 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4747
4748void *netdev_lower_get_next(struct net_device *dev,
4749 struct list_head **iter);
4750
4751#define netdev_for_each_lower_dev(dev, ldev, iter) \
4752 for (iter = (dev)->adj_list.lower.next, \
4753 ldev = netdev_lower_get_next(dev, &(iter)); \
4754 ldev; \
4755 ldev = netdev_lower_get_next(dev, &(iter)))
4756
4757struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4758 struct list_head **iter);
4759int netdev_walk_all_lower_dev(struct net_device *dev,
4760 int (*fn)(struct net_device *lower_dev,
4761 struct netdev_nested_priv *priv),
4762 struct netdev_nested_priv *priv);
4763int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4764 int (*fn)(struct net_device *lower_dev,
4765 struct netdev_nested_priv *priv),
4766 struct netdev_nested_priv *priv);
4767
4768void *netdev_adjacent_get_private(struct list_head *adj_list);
4769void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4770struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4771struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4772int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4773 struct netlink_ext_ack *extack);
4774int netdev_master_upper_dev_link(struct net_device *dev,
4775 struct net_device *upper_dev,
4776 void *upper_priv, void *upper_info,
4777 struct netlink_ext_ack *extack);
4778void netdev_upper_dev_unlink(struct net_device *dev,
4779 struct net_device *upper_dev);
4780int netdev_adjacent_change_prepare(struct net_device *old_dev,
4781 struct net_device *new_dev,
4782 struct net_device *dev,
4783 struct netlink_ext_ack *extack);
4784void netdev_adjacent_change_commit(struct net_device *old_dev,
4785 struct net_device *new_dev,
4786 struct net_device *dev);
4787void netdev_adjacent_change_abort(struct net_device *old_dev,
4788 struct net_device *new_dev,
4789 struct net_device *dev);
4790void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4791void *netdev_lower_dev_get_private(struct net_device *dev,
4792 struct net_device *lower_dev);
4793void netdev_lower_state_changed(struct net_device *lower_dev,
4794 void *lower_state_info);
4795
4796/* RSS keys are 40 or 52 bytes long */
4797#define NETDEV_RSS_KEY_LEN 52
4798extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4799void netdev_rss_key_fill(void *buffer, size_t len);
4800
4801int skb_checksum_help(struct sk_buff *skb);
4802int skb_crc32c_csum_help(struct sk_buff *skb);
4803int skb_csum_hwoffload_help(struct sk_buff *skb,
4804 const netdev_features_t features);
4805
4806struct netdev_bonding_info {
4807 ifslave slave;
4808 ifbond master;
4809};
4810
4811struct netdev_notifier_bonding_info {
4812 struct netdev_notifier_info info; /* must be first */
4813 struct netdev_bonding_info bonding_info;
4814};
4815
4816void netdev_bonding_info_change(struct net_device *dev,
4817 struct netdev_bonding_info *bonding_info);
4818
4819#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4820void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4821#else
4822static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4823 const void *data)
4824{
4825}
4826#endif
4827
4828__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4829
4830static inline bool can_checksum_protocol(netdev_features_t features,
4831 __be16 protocol)
4832{
4833 if (protocol == htons(ETH_P_FCOE))
4834 return !!(features & NETIF_F_FCOE_CRC);
4835
4836 /* Assume this is an IP checksum (not SCTP CRC) */
4837
4838 if (features & NETIF_F_HW_CSUM) {
4839 /* Can checksum everything */
4840 return true;
4841 }
4842
4843 switch (protocol) {
4844 case htons(ETH_P_IP):
4845 return !!(features & NETIF_F_IP_CSUM);
4846 case htons(ETH_P_IPV6):
4847 return !!(features & NETIF_F_IPV6_CSUM);
4848 default:
4849 return false;
4850 }
4851}
4852
4853#ifdef CONFIG_BUG
4854void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4855#else
4856static inline void netdev_rx_csum_fault(struct net_device *dev,
4857 struct sk_buff *skb)
4858{
4859}
4860#endif
4861/* rx skb timestamps */
4862void net_enable_timestamp(void);
4863void net_disable_timestamp(void);
4864
4865static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4866 const struct skb_shared_hwtstamps *hwtstamps,
4867 bool cycles)
4868{
4869 const struct net_device_ops *ops = dev->netdev_ops;
4870
4871 if (ops->ndo_get_tstamp)
4872 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4873
4874 return hwtstamps->hwtstamp;
4875}
4876
4877static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4878 struct sk_buff *skb, struct net_device *dev,
4879 bool more)
4880{
4881 __this_cpu_write(softnet_data.xmit.more, more);
4882 return ops->ndo_start_xmit(skb, dev);
4883}
4884
4885static inline bool netdev_xmit_more(void)
4886{
4887 return __this_cpu_read(softnet_data.xmit.more);
4888}
4889
4890static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4891 struct netdev_queue *txq, bool more)
4892{
4893 const struct net_device_ops *ops = dev->netdev_ops;
4894 netdev_tx_t rc;
4895
4896 rc = __netdev_start_xmit(ops, skb, dev, more);
4897 if (rc == NETDEV_TX_OK)
4898 txq_trans_update(txq);
4899
4900 return rc;
4901}
4902
4903int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4904 const void *ns);
4905void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4906 const void *ns);
4907
4908extern const struct kobj_ns_type_operations net_ns_type_operations;
4909
4910const char *netdev_drivername(const struct net_device *dev);
4911
4912static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4913 netdev_features_t f2)
4914{
4915 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4916 if (f1 & NETIF_F_HW_CSUM)
4917 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4918 else
4919 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4920 }
4921
4922 return f1 & f2;
4923}
4924
4925static inline netdev_features_t netdev_get_wanted_features(
4926 struct net_device *dev)
4927{
4928 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4929}
4930netdev_features_t netdev_increment_features(netdev_features_t all,
4931 netdev_features_t one, netdev_features_t mask);
4932
4933/* Allow TSO being used on stacked device :
4934 * Performing the GSO segmentation before last device
4935 * is a performance improvement.
4936 */
4937static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4938 netdev_features_t mask)
4939{
4940 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4941}
4942
4943int __netdev_update_features(struct net_device *dev);
4944void netdev_update_features(struct net_device *dev);
4945void netdev_change_features(struct net_device *dev);
4946
4947void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4948 struct net_device *dev);
4949
4950netdev_features_t passthru_features_check(struct sk_buff *skb,
4951 struct net_device *dev,
4952 netdev_features_t features);
4953netdev_features_t netif_skb_features(struct sk_buff *skb);
4954void skb_warn_bad_offload(const struct sk_buff *skb);
4955
4956static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4957{
4958 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4959
4960 /* check flags correspondence */
4961 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4962 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4963 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4964 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4965 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4966 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4967 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4968 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4969 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4970 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4971 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4972 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4973 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4974 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4975 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4976 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4977 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4978 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4979 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4980
4981 return (features & feature) == feature;
4982}
4983
4984static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4985{
4986 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4987 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4988}
4989
4990static inline bool netif_needs_gso(struct sk_buff *skb,
4991 netdev_features_t features)
4992{
4993 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4994 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4995 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4996}
4997
4998void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4999void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5000void netif_inherit_tso_max(struct net_device *to,
5001 const struct net_device *from);
5002
5003static inline bool netif_is_macsec(const struct net_device *dev)
5004{
5005 return dev->priv_flags & IFF_MACSEC;
5006}
5007
5008static inline bool netif_is_macvlan(const struct net_device *dev)
5009{
5010 return dev->priv_flags & IFF_MACVLAN;
5011}
5012
5013static inline bool netif_is_macvlan_port(const struct net_device *dev)
5014{
5015 return dev->priv_flags & IFF_MACVLAN_PORT;
5016}
5017
5018static inline bool netif_is_bond_master(const struct net_device *dev)
5019{
5020 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5021}
5022
5023static inline bool netif_is_bond_slave(const struct net_device *dev)
5024{
5025 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5026}
5027
5028static inline bool netif_supports_nofcs(struct net_device *dev)
5029{
5030 return dev->priv_flags & IFF_SUPP_NOFCS;
5031}
5032
5033static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5034{
5035 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5036}
5037
5038static inline bool netif_is_l3_master(const struct net_device *dev)
5039{
5040 return dev->priv_flags & IFF_L3MDEV_MASTER;
5041}
5042
5043static inline bool netif_is_l3_slave(const struct net_device *dev)
5044{
5045 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5046}
5047
5048static inline int dev_sdif(const struct net_device *dev)
5049{
5050#ifdef CONFIG_NET_L3_MASTER_DEV
5051 if (netif_is_l3_slave(dev))
5052 return dev->ifindex;
5053#endif
5054 return 0;
5055}
5056
5057static inline bool netif_is_bridge_master(const struct net_device *dev)
5058{
5059 return dev->priv_flags & IFF_EBRIDGE;
5060}
5061
5062static inline bool netif_is_bridge_port(const struct net_device *dev)
5063{
5064 return dev->priv_flags & IFF_BRIDGE_PORT;
5065}
5066
5067static inline bool netif_is_ovs_master(const struct net_device *dev)
5068{
5069 return dev->priv_flags & IFF_OPENVSWITCH;
5070}
5071
5072static inline bool netif_is_ovs_port(const struct net_device *dev)
5073{
5074 return dev->priv_flags & IFF_OVS_DATAPATH;
5075}
5076
5077static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5078{
5079 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5080}
5081
5082static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5083{
5084 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5085}
5086
5087static inline bool netif_is_team_master(const struct net_device *dev)
5088{
5089 return dev->priv_flags & IFF_TEAM;
5090}
5091
5092static inline bool netif_is_team_port(const struct net_device *dev)
5093{
5094 return dev->priv_flags & IFF_TEAM_PORT;
5095}
5096
5097static inline bool netif_is_lag_master(const struct net_device *dev)
5098{
5099 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5100}
5101
5102static inline bool netif_is_lag_port(const struct net_device *dev)
5103{
5104 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5105}
5106
5107static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5108{
5109 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5110}
5111
5112static inline bool netif_is_failover(const struct net_device *dev)
5113{
5114 return dev->priv_flags & IFF_FAILOVER;
5115}
5116
5117static inline bool netif_is_failover_slave(const struct net_device *dev)
5118{
5119 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5120}
5121
5122/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5123static inline void netif_keep_dst(struct net_device *dev)
5124{
5125 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5126}
5127
5128/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5129static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5130{
5131 /* TODO: reserve and use an additional IFF bit, if we get more users */
5132 return netif_is_macsec(dev);
5133}
5134
5135extern struct pernet_operations __net_initdata loopback_net_ops;
5136
5137/* Logging, debugging and troubleshooting/diagnostic helpers. */
5138
5139/* netdev_printk helpers, similar to dev_printk */
5140
5141static inline const char *netdev_name(const struct net_device *dev)
5142{
5143 if (!dev->name[0] || strchr(dev->name, '%'))
5144 return "(unnamed net_device)";
5145 return dev->name;
5146}
5147
5148static inline const char *netdev_reg_state(const struct net_device *dev)
5149{
5150 u8 reg_state = READ_ONCE(dev->reg_state);
5151
5152 switch (reg_state) {
5153 case NETREG_UNINITIALIZED: return " (uninitialized)";
5154 case NETREG_REGISTERED: return "";
5155 case NETREG_UNREGISTERING: return " (unregistering)";
5156 case NETREG_UNREGISTERED: return " (unregistered)";
5157 case NETREG_RELEASED: return " (released)";
5158 case NETREG_DUMMY: return " (dummy)";
5159 }
5160
5161 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5162 return " (unknown)";
5163}
5164
5165#define MODULE_ALIAS_NETDEV(device) \
5166 MODULE_ALIAS("netdev-" device)
5167
5168/*
5169 * netdev_WARN() acts like dev_printk(), but with the key difference
5170 * of using a WARN/WARN_ON to get the message out, including the
5171 * file/line information and a backtrace.
5172 */
5173#define netdev_WARN(dev, format, args...) \
5174 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5175 netdev_reg_state(dev), ##args)
5176
5177#define netdev_WARN_ONCE(dev, format, args...) \
5178 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5179 netdev_reg_state(dev), ##args)
5180
5181/*
5182 * The list of packet types we will receive (as opposed to discard)
5183 * and the routines to invoke.
5184 *
5185 * Why 16. Because with 16 the only overlap we get on a hash of the
5186 * low nibble of the protocol value is RARP/SNAP/X.25.
5187 *
5188 * 0800 IP
5189 * 0001 802.3
5190 * 0002 AX.25
5191 * 0004 802.2
5192 * 8035 RARP
5193 * 0005 SNAP
5194 * 0805 X.25
5195 * 0806 ARP
5196 * 8137 IPX
5197 * 0009 Localtalk
5198 * 86DD IPv6
5199 */
5200#define PTYPE_HASH_SIZE (16)
5201#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5202
5203extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5204
5205extern struct net_device *blackhole_netdev;
5206
5207/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5208#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5209#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5210 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5211#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5212
5213#endif /* _LINUX_NETDEVICE_H */
Linux 6.x block layer and io_uring tree(s)