Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31
32#include <linux/percpu.h>
33#include <linux/rculist.h>
34#include <linux/workqueue.h>
35#include <linux/dynamic_queue_limits.h>
36
37#include <linux/ethtool.h>
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <net/xdp.h>
44
45#include <linux/netdev_features.h>
46#include <linux/neighbour.h>
47#include <uapi/linux/netdevice.h>
48#include <uapi/linux/if_bonding.h>
49#include <uapi/linux/pkt_cls.h>
50#include <linux/hashtable.h>
51
52struct netpoll_info;
53struct device;
54struct phy_device;
55struct dsa_port;
56
57struct sfp_bus;
58/* 802.11 specific */
59struct wireless_dev;
60/* 802.15.4 specific */
61struct wpan_dev;
62struct mpls_dev;
63/* UDP Tunnel offloads */
64struct udp_tunnel_info;
65struct bpf_prog;
66struct xdp_buff;
67
68void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
70
71/* Backlog congestion levels */
72#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73#define NET_RX_DROP 1 /* packet dropped */
74
75/*
76 * Transmit return codes: transmit return codes originate from three different
77 * namespaces:
78 *
79 * - qdisc return codes
80 * - driver transmit return codes
81 * - errno values
82 *
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
87 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
89 * others are propagated to higher layers.
90 */
91
92/* qdisc ->enqueue() return codes. */
93#define NET_XMIT_SUCCESS 0x00
94#define NET_XMIT_DROP 0x01 /* skb dropped */
95#define NET_XMIT_CN 0x02 /* congestion notification */
96#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
97
98/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
101#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
102#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103
104/* Driver transmit return codes */
105#define NETDEV_TX_MASK 0xf0
106
107enum netdev_tx {
108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
111};
112typedef enum netdev_tx netdev_tx_t;
113
114/*
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 */
118static inline bool dev_xmit_complete(int rc)
119{
120 /*
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 */
126 if (likely(rc < NET_XMIT_MASK))
127 return true;
128
129 return false;
130}
131
132/*
133 * Compute the worst-case header length according to the protocols
134 * used.
135 */
136
137#if defined(CONFIG_HYPERV_NET)
138# define LL_MAX_HEADER 128
139#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140# if defined(CONFIG_MAC80211_MESH)
141# define LL_MAX_HEADER 128
142# else
143# define LL_MAX_HEADER 96
144# endif
145#else
146# define LL_MAX_HEADER 32
147#endif
148
149#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
151#define MAX_HEADER LL_MAX_HEADER
152#else
153#define MAX_HEADER (LL_MAX_HEADER + 48)
154#endif
155
156/*
157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
159 */
160
161struct net_device_stats {
162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
171 unsigned long collisions;
172 unsigned long rx_length_errors;
173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
185};
186
187
188#include <linux/cache.h>
189#include <linux/skbuff.h>
190
191#ifdef CONFIG_RPS
192#include <linux/static_key.h>
193extern struct static_key_false rps_needed;
194extern struct static_key_false rfs_needed;
195#endif
196
197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
209#define NETDEV_HW_ADDR_T_MULTICAST 5
210 bool global_use;
211 int sync_cnt;
212 int refcount;
213 int synced;
214 struct rcu_head rcu_head;
215};
216
217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229#define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231
232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234#define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236
237struct hh_cache {
238 unsigned int hh_len;
239 seqlock_t hh_lock;
240
241 /* cached hardware header; allow for machine alignment needs. */
242#define HH_DATA_MOD 16
243#define HH_DATA_OFF(__len) \
244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
245#define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248};
249
250/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
251 * Alternative is:
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 *
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
257 */
258#define LL_RESERVED_SPACE(dev) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262
263struct header_ops {
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
266 const void *saddr, unsigned int len);
267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
272 bool (*validate)(const char *ll_header, unsigned int len);
273 __be16 (*parse_protocol)(const struct sk_buff *skb);
274};
275
276/* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
278 * code.
279 */
280
281enum netdev_state_t {
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
287};
288
289
290/*
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
300int __init netdev_boot_setup(char *str);
301
302struct gro_list {
303 struct list_head list;
304 int count;
305};
306
307/*
308 * size of gro hash buckets, must less than bit number of
309 * napi_struct::gro_bitmask
310 */
311#define GRO_HASH_BUCKETS 8
312
313/*
314 * Structure for NAPI scheduling similar to tasklet but with weighting
315 */
316struct napi_struct {
317 /* The poll_list must only be managed by the entity which
318 * changes the state of the NAPI_STATE_SCHED bit. This means
319 * whoever atomically sets that bit can add this napi_struct
320 * to the per-CPU poll_list, and whoever clears that bit
321 * can remove from the list right before clearing the bit.
322 */
323 struct list_head poll_list;
324
325 unsigned long state;
326 int weight;
327 unsigned long gro_bitmask;
328 int (*poll)(struct napi_struct *, int);
329#ifdef CONFIG_NETPOLL
330 int poll_owner;
331#endif
332 struct net_device *dev;
333 struct gro_list gro_hash[GRO_HASH_BUCKETS];
334 struct sk_buff *skb;
335 struct hrtimer timer;
336 struct list_head dev_list;
337 struct hlist_node napi_hash_node;
338 unsigned int napi_id;
339};
340
341enum {
342 NAPI_STATE_SCHED, /* Poll is scheduled */
343 NAPI_STATE_MISSED, /* reschedule a napi */
344 NAPI_STATE_DISABLE, /* Disable pending */
345 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
346 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
347 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
348 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
349};
350
351enum {
352 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
353 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
354 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
355 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
356 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
357 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
358 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
359};
360
361enum gro_result {
362 GRO_MERGED,
363 GRO_MERGED_FREE,
364 GRO_HELD,
365 GRO_NORMAL,
366 GRO_DROP,
367 GRO_CONSUMED,
368};
369typedef enum gro_result gro_result_t;
370
371/*
372 * enum rx_handler_result - Possible return values for rx_handlers.
373 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
374 * further.
375 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
376 * case skb->dev was changed by rx_handler.
377 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
378 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
379 *
380 * rx_handlers are functions called from inside __netif_receive_skb(), to do
381 * special processing of the skb, prior to delivery to protocol handlers.
382 *
383 * Currently, a net_device can only have a single rx_handler registered. Trying
384 * to register a second rx_handler will return -EBUSY.
385 *
386 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
387 * To unregister a rx_handler on a net_device, use
388 * netdev_rx_handler_unregister().
389 *
390 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
391 * do with the skb.
392 *
393 * If the rx_handler consumed the skb in some way, it should return
394 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
395 * the skb to be delivered in some other way.
396 *
397 * If the rx_handler changed skb->dev, to divert the skb to another
398 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
399 * new device will be called if it exists.
400 *
401 * If the rx_handler decides the skb should be ignored, it should return
402 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
403 * are registered on exact device (ptype->dev == skb->dev).
404 *
405 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
406 * delivered, it should return RX_HANDLER_PASS.
407 *
408 * A device without a registered rx_handler will behave as if rx_handler
409 * returned RX_HANDLER_PASS.
410 */
411
412enum rx_handler_result {
413 RX_HANDLER_CONSUMED,
414 RX_HANDLER_ANOTHER,
415 RX_HANDLER_EXACT,
416 RX_HANDLER_PASS,
417};
418typedef enum rx_handler_result rx_handler_result_t;
419typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
420
421void __napi_schedule(struct napi_struct *n);
422void __napi_schedule_irqoff(struct napi_struct *n);
423
424static inline bool napi_disable_pending(struct napi_struct *n)
425{
426 return test_bit(NAPI_STATE_DISABLE, &n->state);
427}
428
429bool napi_schedule_prep(struct napi_struct *n);
430
431/**
432 * napi_schedule - schedule NAPI poll
433 * @n: NAPI context
434 *
435 * Schedule NAPI poll routine to be called if it is not already
436 * running.
437 */
438static inline void napi_schedule(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule(n);
442}
443
444/**
445 * napi_schedule_irqoff - schedule NAPI poll
446 * @n: NAPI context
447 *
448 * Variant of napi_schedule(), assuming hard irqs are masked.
449 */
450static inline void napi_schedule_irqoff(struct napi_struct *n)
451{
452 if (napi_schedule_prep(n))
453 __napi_schedule_irqoff(n);
454}
455
456/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
457static inline bool napi_reschedule(struct napi_struct *napi)
458{
459 if (napi_schedule_prep(napi)) {
460 __napi_schedule(napi);
461 return true;
462 }
463 return false;
464}
465
466bool napi_complete_done(struct napi_struct *n, int work_done);
467/**
468 * napi_complete - NAPI processing complete
469 * @n: NAPI context
470 *
471 * Mark NAPI processing as complete.
472 * Consider using napi_complete_done() instead.
473 * Return false if device should avoid rearming interrupts.
474 */
475static inline bool napi_complete(struct napi_struct *n)
476{
477 return napi_complete_done(n, 0);
478}
479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: NAPI context
483 *
484 * Warning: caller must observe RCU grace period
485 * before freeing memory containing @napi, if
486 * this function returns true.
487 * Note: core networking stack automatically calls it
488 * from netif_napi_del().
489 * Drivers might want to call this helper to combine all
490 * the needed RCU grace periods into a single one.
491 */
492bool napi_hash_del(struct napi_struct *napi);
493
494/**
495 * napi_disable - prevent NAPI from scheduling
496 * @n: NAPI context
497 *
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
500 */
501void napi_disable(struct napi_struct *n);
502
503/**
504 * napi_enable - enable NAPI scheduling
505 * @n: NAPI context
506 *
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
509 */
510static inline void napi_enable(struct napi_struct *n)
511{
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
513 smp_mb__before_atomic();
514 clear_bit(NAPI_STATE_SCHED, &n->state);
515 clear_bit(NAPI_STATE_NPSVC, &n->state);
516}
517
518/**
519 * napi_synchronize - wait until NAPI is not running
520 * @n: NAPI context
521 *
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
525 */
526static inline void napi_synchronize(const struct napi_struct *n)
527{
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
530 msleep(1);
531 else
532 barrier();
533}
534
535/**
536 * napi_if_scheduled_mark_missed - if napi is running, set the
537 * NAPIF_STATE_MISSED
538 * @n: NAPI context
539 *
540 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
541 * NAPI is scheduled.
542 **/
543static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
544{
545 unsigned long val, new;
546
547 do {
548 val = READ_ONCE(n->state);
549 if (val & NAPIF_STATE_DISABLE)
550 return true;
551
552 if (!(val & NAPIF_STATE_SCHED))
553 return false;
554
555 new = val | NAPIF_STATE_MISSED;
556 } while (cmpxchg(&n->state, val, new) != val);
557
558 return true;
559}
560
561enum netdev_queue_state_t {
562 __QUEUE_STATE_DRV_XOFF,
563 __QUEUE_STATE_STACK_XOFF,
564 __QUEUE_STATE_FROZEN,
565};
566
567#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
568#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
569#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
570
571#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
572#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
573 QUEUE_STATE_FROZEN)
574#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
575 QUEUE_STATE_FROZEN)
576
577/*
578 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
579 * netif_tx_* functions below are used to manipulate this flag. The
580 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
581 * queue independently. The netif_xmit_*stopped functions below are called
582 * to check if the queue has been stopped by the driver or stack (either
583 * of the XOFF bits are set in the state). Drivers should not need to call
584 * netif_xmit*stopped functions, they should only be using netif_tx_*.
585 */
586
587struct netdev_queue {
588/*
589 * read-mostly part
590 */
591 struct net_device *dev;
592 struct Qdisc __rcu *qdisc;
593 struct Qdisc *qdisc_sleeping;
594#ifdef CONFIG_SYSFS
595 struct kobject kobj;
596#endif
597#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 int numa_node;
599#endif
600 unsigned long tx_maxrate;
601 /*
602 * Number of TX timeouts for this queue
603 * (/sys/class/net/DEV/Q/trans_timeout)
604 */
605 unsigned long trans_timeout;
606
607 /* Subordinate device that the queue has been assigned to */
608 struct net_device *sb_dev;
609#ifdef CONFIG_XDP_SOCKETS
610 struct xdp_umem *umem;
611#endif
612/*
613 * write-mostly part
614 */
615 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
616 int xmit_lock_owner;
617 /*
618 * Time (in jiffies) of last Tx
619 */
620 unsigned long trans_start;
621
622 unsigned long state;
623
624#ifdef CONFIG_BQL
625 struct dql dql;
626#endif
627} ____cacheline_aligned_in_smp;
628
629extern int sysctl_fb_tunnels_only_for_init_net;
630extern int sysctl_devconf_inherit_init_net;
631
632static inline bool net_has_fallback_tunnels(const struct net *net)
633{
634 return net == &init_net ||
635 !IS_ENABLED(CONFIG_SYSCTL) ||
636 !sysctl_fb_tunnels_only_for_init_net;
637}
638
639static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
640{
641#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
642 return q->numa_node;
643#else
644 return NUMA_NO_NODE;
645#endif
646}
647
648static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
649{
650#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
651 q->numa_node = node;
652#endif
653}
654
655#ifdef CONFIG_RPS
656/*
657 * This structure holds an RPS map which can be of variable length. The
658 * map is an array of CPUs.
659 */
660struct rps_map {
661 unsigned int len;
662 struct rcu_head rcu;
663 u16 cpus[0];
664};
665#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
666
667/*
668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
669 * tail pointer for that CPU's input queue at the time of last enqueue, and
670 * a hardware filter index.
671 */
672struct rps_dev_flow {
673 u16 cpu;
674 u16 filter;
675 unsigned int last_qtail;
676};
677#define RPS_NO_FILTER 0xffff
678
679/*
680 * The rps_dev_flow_table structure contains a table of flow mappings.
681 */
682struct rps_dev_flow_table {
683 unsigned int mask;
684 struct rcu_head rcu;
685 struct rps_dev_flow flows[0];
686};
687#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
688 ((_num) * sizeof(struct rps_dev_flow)))
689
690/*
691 * The rps_sock_flow_table contains mappings of flows to the last CPU
692 * on which they were processed by the application (set in recvmsg).
693 * Each entry is a 32bit value. Upper part is the high-order bits
694 * of flow hash, lower part is CPU number.
695 * rps_cpu_mask is used to partition the space, depending on number of
696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
698 * meaning we use 32-6=26 bits for the hash.
699 */
700struct rps_sock_flow_table {
701 u32 mask;
702
703 u32 ents[0] ____cacheline_aligned_in_smp;
704};
705#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
706
707#define RPS_NO_CPU 0xffff
708
709extern u32 rps_cpu_mask;
710extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
711
712static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
713 u32 hash)
714{
715 if (table && hash) {
716 unsigned int index = hash & table->mask;
717 u32 val = hash & ~rps_cpu_mask;
718
719 /* We only give a hint, preemption can change CPU under us */
720 val |= raw_smp_processor_id();
721
722 if (table->ents[index] != val)
723 table->ents[index] = val;
724 }
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#endif /* CONFIG_RPS */
732
733/* This structure contains an instance of an RX queue. */
734struct netdev_rx_queue {
735#ifdef CONFIG_RPS
736 struct rps_map __rcu *rps_map;
737 struct rps_dev_flow_table __rcu *rps_flow_table;
738#endif
739 struct kobject kobj;
740 struct net_device *dev;
741 struct xdp_rxq_info xdp_rxq;
742#ifdef CONFIG_XDP_SOCKETS
743 struct xdp_umem *umem;
744#endif
745} ____cacheline_aligned_in_smp;
746
747/*
748 * RX queue sysfs structures and functions.
749 */
750struct rx_queue_attribute {
751 struct attribute attr;
752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
753 ssize_t (*store)(struct netdev_rx_queue *queue,
754 const char *buf, size_t len);
755};
756
757#ifdef CONFIG_XPS
758/*
759 * This structure holds an XPS map which can be of variable length. The
760 * map is an array of queues.
761 */
762struct xps_map {
763 unsigned int len;
764 unsigned int alloc_len;
765 struct rcu_head rcu;
766 u16 queues[0];
767};
768#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
769#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
770 - sizeof(struct xps_map)) / sizeof(u16))
771
772/*
773 * This structure holds all XPS maps for device. Maps are indexed by CPU.
774 */
775struct xps_dev_maps {
776 struct rcu_head rcu;
777 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
778};
779
780#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
781 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
782
783#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
784 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
785
786#endif /* CONFIG_XPS */
787
788#define TC_MAX_QUEUE 16
789#define TC_BITMASK 15
790/* HW offloaded queuing disciplines txq count and offset maps */
791struct netdev_tc_txq {
792 u16 count;
793 u16 offset;
794};
795
796#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
797/*
798 * This structure is to hold information about the device
799 * configured to run FCoE protocol stack.
800 */
801struct netdev_fcoe_hbainfo {
802 char manufacturer[64];
803 char serial_number[64];
804 char hardware_version[64];
805 char driver_version[64];
806 char optionrom_version[64];
807 char firmware_version[64];
808 char model[256];
809 char model_description[256];
810};
811#endif
812
813#define MAX_PHYS_ITEM_ID_LEN 32
814
815/* This structure holds a unique identifier to identify some
816 * physical item (port for example) used by a netdevice.
817 */
818struct netdev_phys_item_id {
819 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
820 unsigned char id_len;
821};
822
823static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
824 struct netdev_phys_item_id *b)
825{
826 return a->id_len == b->id_len &&
827 memcmp(a->id, b->id, a->id_len) == 0;
828}
829
830typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
831 struct sk_buff *skb,
832 struct net_device *sb_dev);
833
834enum tc_setup_type {
835 TC_SETUP_QDISC_MQPRIO,
836 TC_SETUP_CLSU32,
837 TC_SETUP_CLSFLOWER,
838 TC_SETUP_CLSMATCHALL,
839 TC_SETUP_CLSBPF,
840 TC_SETUP_BLOCK,
841 TC_SETUP_QDISC_CBS,
842 TC_SETUP_QDISC_RED,
843 TC_SETUP_QDISC_PRIO,
844 TC_SETUP_QDISC_MQ,
845 TC_SETUP_QDISC_ETF,
846 TC_SETUP_ROOT_QDISC,
847 TC_SETUP_QDISC_GRED,
848};
849
850/* These structures hold the attributes of bpf state that are being passed
851 * to the netdevice through the bpf op.
852 */
853enum bpf_netdev_command {
854 /* Set or clear a bpf program used in the earliest stages of packet
855 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
856 * is responsible for calling bpf_prog_put on any old progs that are
857 * stored. In case of error, the callee need not release the new prog
858 * reference, but on success it takes ownership and must bpf_prog_put
859 * when it is no longer used.
860 */
861 XDP_SETUP_PROG,
862 XDP_SETUP_PROG_HW,
863 XDP_QUERY_PROG,
864 XDP_QUERY_PROG_HW,
865 /* BPF program for offload callbacks, invoked at program load time. */
866 BPF_OFFLOAD_MAP_ALLOC,
867 BPF_OFFLOAD_MAP_FREE,
868 XDP_SETUP_XSK_UMEM,
869};
870
871struct bpf_prog_offload_ops;
872struct netlink_ext_ack;
873struct xdp_umem;
874
875struct netdev_bpf {
876 enum bpf_netdev_command command;
877 union {
878 /* XDP_SETUP_PROG */
879 struct {
880 u32 flags;
881 struct bpf_prog *prog;
882 struct netlink_ext_ack *extack;
883 };
884 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
885 struct {
886 u32 prog_id;
887 /* flags with which program was installed */
888 u32 prog_flags;
889 };
890 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
891 struct {
892 struct bpf_offloaded_map *offmap;
893 };
894 /* XDP_SETUP_XSK_UMEM */
895 struct {
896 struct xdp_umem *umem;
897 u16 queue_id;
898 } xsk;
899 };
900};
901
902#ifdef CONFIG_XFRM_OFFLOAD
903struct xfrmdev_ops {
904 int (*xdo_dev_state_add) (struct xfrm_state *x);
905 void (*xdo_dev_state_delete) (struct xfrm_state *x);
906 void (*xdo_dev_state_free) (struct xfrm_state *x);
907 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
908 struct xfrm_state *x);
909 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
910};
911#endif
912
913struct dev_ifalias {
914 struct rcu_head rcuhead;
915 char ifalias[];
916};
917
918struct devlink;
919struct tlsdev_ops;
920
921/*
922 * This structure defines the management hooks for network devices.
923 * The following hooks can be defined; unless noted otherwise, they are
924 * optional and can be filled with a null pointer.
925 *
926 * int (*ndo_init)(struct net_device *dev);
927 * This function is called once when a network device is registered.
928 * The network device can use this for any late stage initialization
929 * or semantic validation. It can fail with an error code which will
930 * be propagated back to register_netdev.
931 *
932 * void (*ndo_uninit)(struct net_device *dev);
933 * This function is called when device is unregistered or when registration
934 * fails. It is not called if init fails.
935 *
936 * int (*ndo_open)(struct net_device *dev);
937 * This function is called when a network device transitions to the up
938 * state.
939 *
940 * int (*ndo_stop)(struct net_device *dev);
941 * This function is called when a network device transitions to the down
942 * state.
943 *
944 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
945 * struct net_device *dev);
946 * Called when a packet needs to be transmitted.
947 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
948 * the queue before that can happen; it's for obsolete devices and weird
949 * corner cases, but the stack really does a non-trivial amount
950 * of useless work if you return NETDEV_TX_BUSY.
951 * Required; cannot be NULL.
952 *
953 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
954 * struct net_device *dev
955 * netdev_features_t features);
956 * Called by core transmit path to determine if device is capable of
957 * performing offload operations on a given packet. This is to give
958 * the device an opportunity to implement any restrictions that cannot
959 * be otherwise expressed by feature flags. The check is called with
960 * the set of features that the stack has calculated and it returns
961 * those the driver believes to be appropriate.
962 *
963 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
964 * struct net_device *sb_dev);
965 * Called to decide which queue to use when device supports multiple
966 * transmit queues.
967 *
968 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
969 * This function is called to allow device receiver to make
970 * changes to configuration when multicast or promiscuous is enabled.
971 *
972 * void (*ndo_set_rx_mode)(struct net_device *dev);
973 * This function is called device changes address list filtering.
974 * If driver handles unicast address filtering, it should set
975 * IFF_UNICAST_FLT in its priv_flags.
976 *
977 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
978 * This function is called when the Media Access Control address
979 * needs to be changed. If this interface is not defined, the
980 * MAC address can not be changed.
981 *
982 * int (*ndo_validate_addr)(struct net_device *dev);
983 * Test if Media Access Control address is valid for the device.
984 *
985 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
986 * Called when a user requests an ioctl which can't be handled by
987 * the generic interface code. If not defined ioctls return
988 * not supported error code.
989 *
990 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
991 * Used to set network devices bus interface parameters. This interface
992 * is retained for legacy reasons; new devices should use the bus
993 * interface (PCI) for low level management.
994 *
995 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
996 * Called when a user wants to change the Maximum Transfer Unit
997 * of a device.
998 *
999 * void (*ndo_tx_timeout)(struct net_device *dev);
1000 * Callback used when the transmitter has not made any progress
1001 * for dev->watchdog ticks.
1002 *
1003 * void (*ndo_get_stats64)(struct net_device *dev,
1004 * struct rtnl_link_stats64 *storage);
1005 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1006 * Called when a user wants to get the network device usage
1007 * statistics. Drivers must do one of the following:
1008 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1009 * rtnl_link_stats64 structure passed by the caller.
1010 * 2. Define @ndo_get_stats to update a net_device_stats structure
1011 * (which should normally be dev->stats) and return a pointer to
1012 * it. The structure may be changed asynchronously only if each
1013 * field is written atomically.
1014 * 3. Update dev->stats asynchronously and atomically, and define
1015 * neither operation.
1016 *
1017 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1018 * Return true if this device supports offload stats of this attr_id.
1019 *
1020 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1021 * void *attr_data)
1022 * Get statistics for offload operations by attr_id. Write it into the
1023 * attr_data pointer.
1024 *
1025 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1026 * If device supports VLAN filtering this function is called when a
1027 * VLAN id is registered.
1028 *
1029 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1030 * If device supports VLAN filtering this function is called when a
1031 * VLAN id is unregistered.
1032 *
1033 * void (*ndo_poll_controller)(struct net_device *dev);
1034 *
1035 * SR-IOV management functions.
1036 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1037 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1038 * u8 qos, __be16 proto);
1039 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1040 * int max_tx_rate);
1041 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1042 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1043 * int (*ndo_get_vf_config)(struct net_device *dev,
1044 * int vf, struct ifla_vf_info *ivf);
1045 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1046 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1047 * struct nlattr *port[]);
1048 *
1049 * Enable or disable the VF ability to query its RSS Redirection Table and
1050 * Hash Key. This is needed since on some devices VF share this information
1051 * with PF and querying it may introduce a theoretical security risk.
1052 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1053 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1054 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1055 * void *type_data);
1056 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1057 * This is always called from the stack with the rtnl lock held and netif
1058 * tx queues stopped. This allows the netdevice to perform queue
1059 * management safely.
1060 *
1061 * Fiber Channel over Ethernet (FCoE) offload functions.
1062 * int (*ndo_fcoe_enable)(struct net_device *dev);
1063 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1064 * so the underlying device can perform whatever needed configuration or
1065 * initialization to support acceleration of FCoE traffic.
1066 *
1067 * int (*ndo_fcoe_disable)(struct net_device *dev);
1068 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1069 * so the underlying device can perform whatever needed clean-ups to
1070 * stop supporting acceleration of FCoE traffic.
1071 *
1072 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1073 * struct scatterlist *sgl, unsigned int sgc);
1074 * Called when the FCoE Initiator wants to initialize an I/O that
1075 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1076 * perform necessary setup and returns 1 to indicate the device is set up
1077 * successfully to perform DDP on this I/O, otherwise this returns 0.
1078 *
1079 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1080 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1081 * indicated by the FC exchange id 'xid', so the underlying device can
1082 * clean up and reuse resources for later DDP requests.
1083 *
1084 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1085 * struct scatterlist *sgl, unsigned int sgc);
1086 * Called when the FCoE Target wants to initialize an I/O that
1087 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1088 * perform necessary setup and returns 1 to indicate the device is set up
1089 * successfully to perform DDP on this I/O, otherwise this returns 0.
1090 *
1091 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1092 * struct netdev_fcoe_hbainfo *hbainfo);
1093 * Called when the FCoE Protocol stack wants information on the underlying
1094 * device. This information is utilized by the FCoE protocol stack to
1095 * register attributes with Fiber Channel management service as per the
1096 * FC-GS Fabric Device Management Information(FDMI) specification.
1097 *
1098 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1099 * Called when the underlying device wants to override default World Wide
1100 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1101 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1102 * protocol stack to use.
1103 *
1104 * RFS acceleration.
1105 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1106 * u16 rxq_index, u32 flow_id);
1107 * Set hardware filter for RFS. rxq_index is the target queue index;
1108 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1109 * Return the filter ID on success, or a negative error code.
1110 *
1111 * Slave management functions (for bridge, bonding, etc).
1112 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1113 * Called to make another netdev an underling.
1114 *
1115 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1116 * Called to release previously enslaved netdev.
1117 *
1118 * Feature/offload setting functions.
1119 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1120 * netdev_features_t features);
1121 * Adjusts the requested feature flags according to device-specific
1122 * constraints, and returns the resulting flags. Must not modify
1123 * the device state.
1124 *
1125 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1126 * Called to update device configuration to new features. Passed
1127 * feature set might be less than what was returned by ndo_fix_features()).
1128 * Must return >0 or -errno if it changed dev->features itself.
1129 *
1130 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1131 * struct net_device *dev,
1132 * const unsigned char *addr, u16 vid, u16 flags,
1133 * struct netlink_ext_ack *extack);
1134 * Adds an FDB entry to dev for addr.
1135 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1136 * struct net_device *dev,
1137 * const unsigned char *addr, u16 vid)
1138 * Deletes the FDB entry from dev coresponding to addr.
1139 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1140 * struct net_device *dev, struct net_device *filter_dev,
1141 * int *idx)
1142 * Used to add FDB entries to dump requests. Implementers should add
1143 * entries to skb and update idx with the number of entries.
1144 *
1145 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1146 * u16 flags, struct netlink_ext_ack *extack)
1147 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1148 * struct net_device *dev, u32 filter_mask,
1149 * int nlflags)
1150 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1151 * u16 flags);
1152 *
1153 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1154 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1155 * which do not represent real hardware may define this to allow their
1156 * userspace components to manage their virtual carrier state. Devices
1157 * that determine carrier state from physical hardware properties (eg
1158 * network cables) or protocol-dependent mechanisms (eg
1159 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1160 *
1161 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1162 * struct netdev_phys_item_id *ppid);
1163 * Called to get ID of physical port of this device. If driver does
1164 * not implement this, it is assumed that the hw is not able to have
1165 * multiple net devices on single physical port.
1166 *
1167 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1168 * struct netdev_phys_item_id *ppid)
1169 * Called to get the parent ID of the physical port of this device.
1170 *
1171 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1172 * struct udp_tunnel_info *ti);
1173 * Called by UDP tunnel to notify a driver about the UDP port and socket
1174 * address family that a UDP tunnel is listnening to. It is called only
1175 * when a new port starts listening. The operation is protected by the
1176 * RTNL.
1177 *
1178 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1179 * struct udp_tunnel_info *ti);
1180 * Called by UDP tunnel to notify the driver about a UDP port and socket
1181 * address family that the UDP tunnel is not listening to anymore. The
1182 * operation is protected by the RTNL.
1183 *
1184 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1185 * struct net_device *dev)
1186 * Called by upper layer devices to accelerate switching or other
1187 * station functionality into hardware. 'pdev is the lowerdev
1188 * to use for the offload and 'dev' is the net device that will
1189 * back the offload. Returns a pointer to the private structure
1190 * the upper layer will maintain.
1191 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1192 * Called by upper layer device to delete the station created
1193 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1194 * the station and priv is the structure returned by the add
1195 * operation.
1196 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1197 * int queue_index, u32 maxrate);
1198 * Called when a user wants to set a max-rate limitation of specific
1199 * TX queue.
1200 * int (*ndo_get_iflink)(const struct net_device *dev);
1201 * Called to get the iflink value of this device.
1202 * void (*ndo_change_proto_down)(struct net_device *dev,
1203 * bool proto_down);
1204 * This function is used to pass protocol port error state information
1205 * to the switch driver. The switch driver can react to the proto_down
1206 * by doing a phys down on the associated switch port.
1207 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1208 * This function is used to get egress tunnel information for given skb.
1209 * This is useful for retrieving outer tunnel header parameters while
1210 * sampling packet.
1211 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1212 * This function is used to specify the headroom that the skb must
1213 * consider when allocation skb during packet reception. Setting
1214 * appropriate rx headroom value allows avoiding skb head copy on
1215 * forward. Setting a negative value resets the rx headroom to the
1216 * default value.
1217 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1218 * This function is used to set or query state related to XDP on the
1219 * netdevice and manage BPF offload. See definition of
1220 * enum bpf_netdev_command for details.
1221 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1222 * u32 flags);
1223 * This function is used to submit @n XDP packets for transmit on a
1224 * netdevice. Returns number of frames successfully transmitted, frames
1225 * that got dropped are freed/returned via xdp_return_frame().
1226 * Returns negative number, means general error invoking ndo, meaning
1227 * no frames were xmit'ed and core-caller will free all frames.
1228 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1229 * Get devlink port instance associated with a given netdev.
1230 * Called with a reference on the netdevice and devlink locks only,
1231 * rtnl_lock is not held.
1232 */
1233struct net_device_ops {
1234 int (*ndo_init)(struct net_device *dev);
1235 void (*ndo_uninit)(struct net_device *dev);
1236 int (*ndo_open)(struct net_device *dev);
1237 int (*ndo_stop)(struct net_device *dev);
1238 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1239 struct net_device *dev);
1240 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1241 struct net_device *dev,
1242 netdev_features_t features);
1243 u16 (*ndo_select_queue)(struct net_device *dev,
1244 struct sk_buff *skb,
1245 struct net_device *sb_dev);
1246 void (*ndo_change_rx_flags)(struct net_device *dev,
1247 int flags);
1248 void (*ndo_set_rx_mode)(struct net_device *dev);
1249 int (*ndo_set_mac_address)(struct net_device *dev,
1250 void *addr);
1251 int (*ndo_validate_addr)(struct net_device *dev);
1252 int (*ndo_do_ioctl)(struct net_device *dev,
1253 struct ifreq *ifr, int cmd);
1254 int (*ndo_set_config)(struct net_device *dev,
1255 struct ifmap *map);
1256 int (*ndo_change_mtu)(struct net_device *dev,
1257 int new_mtu);
1258 int (*ndo_neigh_setup)(struct net_device *dev,
1259 struct neigh_parms *);
1260 void (*ndo_tx_timeout) (struct net_device *dev);
1261
1262 void (*ndo_get_stats64)(struct net_device *dev,
1263 struct rtnl_link_stats64 *storage);
1264 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1265 int (*ndo_get_offload_stats)(int attr_id,
1266 const struct net_device *dev,
1267 void *attr_data);
1268 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1269
1270 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1271 __be16 proto, u16 vid);
1272 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1273 __be16 proto, u16 vid);
1274#ifdef CONFIG_NET_POLL_CONTROLLER
1275 void (*ndo_poll_controller)(struct net_device *dev);
1276 int (*ndo_netpoll_setup)(struct net_device *dev,
1277 struct netpoll_info *info);
1278 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1279#endif
1280 int (*ndo_set_vf_mac)(struct net_device *dev,
1281 int queue, u8 *mac);
1282 int (*ndo_set_vf_vlan)(struct net_device *dev,
1283 int queue, u16 vlan,
1284 u8 qos, __be16 proto);
1285 int (*ndo_set_vf_rate)(struct net_device *dev,
1286 int vf, int min_tx_rate,
1287 int max_tx_rate);
1288 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1289 int vf, bool setting);
1290 int (*ndo_set_vf_trust)(struct net_device *dev,
1291 int vf, bool setting);
1292 int (*ndo_get_vf_config)(struct net_device *dev,
1293 int vf,
1294 struct ifla_vf_info *ivf);
1295 int (*ndo_set_vf_link_state)(struct net_device *dev,
1296 int vf, int link_state);
1297 int (*ndo_get_vf_stats)(struct net_device *dev,
1298 int vf,
1299 struct ifla_vf_stats
1300 *vf_stats);
1301 int (*ndo_set_vf_port)(struct net_device *dev,
1302 int vf,
1303 struct nlattr *port[]);
1304 int (*ndo_get_vf_port)(struct net_device *dev,
1305 int vf, struct sk_buff *skb);
1306 int (*ndo_set_vf_guid)(struct net_device *dev,
1307 int vf, u64 guid,
1308 int guid_type);
1309 int (*ndo_set_vf_rss_query_en)(
1310 struct net_device *dev,
1311 int vf, bool setting);
1312 int (*ndo_setup_tc)(struct net_device *dev,
1313 enum tc_setup_type type,
1314 void *type_data);
1315#if IS_ENABLED(CONFIG_FCOE)
1316 int (*ndo_fcoe_enable)(struct net_device *dev);
1317 int (*ndo_fcoe_disable)(struct net_device *dev);
1318 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1319 u16 xid,
1320 struct scatterlist *sgl,
1321 unsigned int sgc);
1322 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1323 u16 xid);
1324 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1325 u16 xid,
1326 struct scatterlist *sgl,
1327 unsigned int sgc);
1328 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1329 struct netdev_fcoe_hbainfo *hbainfo);
1330#endif
1331
1332#if IS_ENABLED(CONFIG_LIBFCOE)
1333#define NETDEV_FCOE_WWNN 0
1334#define NETDEV_FCOE_WWPN 1
1335 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1336 u64 *wwn, int type);
1337#endif
1338
1339#ifdef CONFIG_RFS_ACCEL
1340 int (*ndo_rx_flow_steer)(struct net_device *dev,
1341 const struct sk_buff *skb,
1342 u16 rxq_index,
1343 u32 flow_id);
1344#endif
1345 int (*ndo_add_slave)(struct net_device *dev,
1346 struct net_device *slave_dev,
1347 struct netlink_ext_ack *extack);
1348 int (*ndo_del_slave)(struct net_device *dev,
1349 struct net_device *slave_dev);
1350 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1351 netdev_features_t features);
1352 int (*ndo_set_features)(struct net_device *dev,
1353 netdev_features_t features);
1354 int (*ndo_neigh_construct)(struct net_device *dev,
1355 struct neighbour *n);
1356 void (*ndo_neigh_destroy)(struct net_device *dev,
1357 struct neighbour *n);
1358
1359 int (*ndo_fdb_add)(struct ndmsg *ndm,
1360 struct nlattr *tb[],
1361 struct net_device *dev,
1362 const unsigned char *addr,
1363 u16 vid,
1364 u16 flags,
1365 struct netlink_ext_ack *extack);
1366 int (*ndo_fdb_del)(struct ndmsg *ndm,
1367 struct nlattr *tb[],
1368 struct net_device *dev,
1369 const unsigned char *addr,
1370 u16 vid);
1371 int (*ndo_fdb_dump)(struct sk_buff *skb,
1372 struct netlink_callback *cb,
1373 struct net_device *dev,
1374 struct net_device *filter_dev,
1375 int *idx);
1376 int (*ndo_fdb_get)(struct sk_buff *skb,
1377 struct nlattr *tb[],
1378 struct net_device *dev,
1379 const unsigned char *addr,
1380 u16 vid, u32 portid, u32 seq,
1381 struct netlink_ext_ack *extack);
1382 int (*ndo_bridge_setlink)(struct net_device *dev,
1383 struct nlmsghdr *nlh,
1384 u16 flags,
1385 struct netlink_ext_ack *extack);
1386 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1387 u32 pid, u32 seq,
1388 struct net_device *dev,
1389 u32 filter_mask,
1390 int nlflags);
1391 int (*ndo_bridge_dellink)(struct net_device *dev,
1392 struct nlmsghdr *nlh,
1393 u16 flags);
1394 int (*ndo_change_carrier)(struct net_device *dev,
1395 bool new_carrier);
1396 int (*ndo_get_phys_port_id)(struct net_device *dev,
1397 struct netdev_phys_item_id *ppid);
1398 int (*ndo_get_port_parent_id)(struct net_device *dev,
1399 struct netdev_phys_item_id *ppid);
1400 int (*ndo_get_phys_port_name)(struct net_device *dev,
1401 char *name, size_t len);
1402 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1403 struct udp_tunnel_info *ti);
1404 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1405 struct udp_tunnel_info *ti);
1406 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1407 struct net_device *dev);
1408 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1409 void *priv);
1410
1411 int (*ndo_get_lock_subclass)(struct net_device *dev);
1412 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1413 int queue_index,
1414 u32 maxrate);
1415 int (*ndo_get_iflink)(const struct net_device *dev);
1416 int (*ndo_change_proto_down)(struct net_device *dev,
1417 bool proto_down);
1418 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1419 struct sk_buff *skb);
1420 void (*ndo_set_rx_headroom)(struct net_device *dev,
1421 int needed_headroom);
1422 int (*ndo_bpf)(struct net_device *dev,
1423 struct netdev_bpf *bpf);
1424 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1425 struct xdp_frame **xdp,
1426 u32 flags);
1427 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1428 u32 queue_id);
1429 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1430};
1431
1432/**
1433 * enum net_device_priv_flags - &struct net_device priv_flags
1434 *
1435 * These are the &struct net_device, they are only set internally
1436 * by drivers and used in the kernel. These flags are invisible to
1437 * userspace; this means that the order of these flags can change
1438 * during any kernel release.
1439 *
1440 * You should have a pretty good reason to be extending these flags.
1441 *
1442 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1443 * @IFF_EBRIDGE: Ethernet bridging device
1444 * @IFF_BONDING: bonding master or slave
1445 * @IFF_ISATAP: ISATAP interface (RFC4214)
1446 * @IFF_WAN_HDLC: WAN HDLC device
1447 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1448 * release skb->dst
1449 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1450 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1451 * @IFF_MACVLAN_PORT: device used as macvlan port
1452 * @IFF_BRIDGE_PORT: device used as bridge port
1453 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1454 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1455 * @IFF_UNICAST_FLT: Supports unicast filtering
1456 * @IFF_TEAM_PORT: device used as team port
1457 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1458 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1459 * change when it's running
1460 * @IFF_MACVLAN: Macvlan device
1461 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1462 * underlying stacked devices
1463 * @IFF_L3MDEV_MASTER: device is an L3 master device
1464 * @IFF_NO_QUEUE: device can run without qdisc attached
1465 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1466 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1467 * @IFF_TEAM: device is a team device
1468 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1469 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1470 * entity (i.e. the master device for bridged veth)
1471 * @IFF_MACSEC: device is a MACsec device
1472 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1473 * @IFF_FAILOVER: device is a failover master device
1474 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1475 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1476 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1477 */
1478enum netdev_priv_flags {
1479 IFF_802_1Q_VLAN = 1<<0,
1480 IFF_EBRIDGE = 1<<1,
1481 IFF_BONDING = 1<<2,
1482 IFF_ISATAP = 1<<3,
1483 IFF_WAN_HDLC = 1<<4,
1484 IFF_XMIT_DST_RELEASE = 1<<5,
1485 IFF_DONT_BRIDGE = 1<<6,
1486 IFF_DISABLE_NETPOLL = 1<<7,
1487 IFF_MACVLAN_PORT = 1<<8,
1488 IFF_BRIDGE_PORT = 1<<9,
1489 IFF_OVS_DATAPATH = 1<<10,
1490 IFF_TX_SKB_SHARING = 1<<11,
1491 IFF_UNICAST_FLT = 1<<12,
1492 IFF_TEAM_PORT = 1<<13,
1493 IFF_SUPP_NOFCS = 1<<14,
1494 IFF_LIVE_ADDR_CHANGE = 1<<15,
1495 IFF_MACVLAN = 1<<16,
1496 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1497 IFF_L3MDEV_MASTER = 1<<18,
1498 IFF_NO_QUEUE = 1<<19,
1499 IFF_OPENVSWITCH = 1<<20,
1500 IFF_L3MDEV_SLAVE = 1<<21,
1501 IFF_TEAM = 1<<22,
1502 IFF_RXFH_CONFIGURED = 1<<23,
1503 IFF_PHONY_HEADROOM = 1<<24,
1504 IFF_MACSEC = 1<<25,
1505 IFF_NO_RX_HANDLER = 1<<26,
1506 IFF_FAILOVER = 1<<27,
1507 IFF_FAILOVER_SLAVE = 1<<28,
1508 IFF_L3MDEV_RX_HANDLER = 1<<29,
1509 IFF_LIVE_RENAME_OK = 1<<30,
1510};
1511
1512#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1513#define IFF_EBRIDGE IFF_EBRIDGE
1514#define IFF_BONDING IFF_BONDING
1515#define IFF_ISATAP IFF_ISATAP
1516#define IFF_WAN_HDLC IFF_WAN_HDLC
1517#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1518#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1519#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1520#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1521#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1522#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1523#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1524#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1525#define IFF_TEAM_PORT IFF_TEAM_PORT
1526#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1527#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1528#define IFF_MACVLAN IFF_MACVLAN
1529#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1530#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1531#define IFF_NO_QUEUE IFF_NO_QUEUE
1532#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1533#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1534#define IFF_TEAM IFF_TEAM
1535#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1536#define IFF_MACSEC IFF_MACSEC
1537#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1538#define IFF_FAILOVER IFF_FAILOVER
1539#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1540#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1541#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1542
1543/**
1544 * struct net_device - The DEVICE structure.
1545 *
1546 * Actually, this whole structure is a big mistake. It mixes I/O
1547 * data with strictly "high-level" data, and it has to know about
1548 * almost every data structure used in the INET module.
1549 *
1550 * @name: This is the first field of the "visible" part of this structure
1551 * (i.e. as seen by users in the "Space.c" file). It is the name
1552 * of the interface.
1553 *
1554 * @name_hlist: Device name hash chain, please keep it close to name[]
1555 * @ifalias: SNMP alias
1556 * @mem_end: Shared memory end
1557 * @mem_start: Shared memory start
1558 * @base_addr: Device I/O address
1559 * @irq: Device IRQ number
1560 *
1561 * @state: Generic network queuing layer state, see netdev_state_t
1562 * @dev_list: The global list of network devices
1563 * @napi_list: List entry used for polling NAPI devices
1564 * @unreg_list: List entry when we are unregistering the
1565 * device; see the function unregister_netdev
1566 * @close_list: List entry used when we are closing the device
1567 * @ptype_all: Device-specific packet handlers for all protocols
1568 * @ptype_specific: Device-specific, protocol-specific packet handlers
1569 *
1570 * @adj_list: Directly linked devices, like slaves for bonding
1571 * @features: Currently active device features
1572 * @hw_features: User-changeable features
1573 *
1574 * @wanted_features: User-requested features
1575 * @vlan_features: Mask of features inheritable by VLAN devices
1576 *
1577 * @hw_enc_features: Mask of features inherited by encapsulating devices
1578 * This field indicates what encapsulation
1579 * offloads the hardware is capable of doing,
1580 * and drivers will need to set them appropriately.
1581 *
1582 * @mpls_features: Mask of features inheritable by MPLS
1583 *
1584 * @ifindex: interface index
1585 * @group: The group the device belongs to
1586 *
1587 * @stats: Statistics struct, which was left as a legacy, use
1588 * rtnl_link_stats64 instead
1589 *
1590 * @rx_dropped: Dropped packets by core network,
1591 * do not use this in drivers
1592 * @tx_dropped: Dropped packets by core network,
1593 * do not use this in drivers
1594 * @rx_nohandler: nohandler dropped packets by core network on
1595 * inactive devices, do not use this in drivers
1596 * @carrier_up_count: Number of times the carrier has been up
1597 * @carrier_down_count: Number of times the carrier has been down
1598 *
1599 * @wireless_handlers: List of functions to handle Wireless Extensions,
1600 * instead of ioctl,
1601 * see <net/iw_handler.h> for details.
1602 * @wireless_data: Instance data managed by the core of wireless extensions
1603 *
1604 * @netdev_ops: Includes several pointers to callbacks,
1605 * if one wants to override the ndo_*() functions
1606 * @ethtool_ops: Management operations
1607 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1608 * discovery handling. Necessary for e.g. 6LoWPAN.
1609 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1610 * of Layer 2 headers.
1611 *
1612 * @flags: Interface flags (a la BSD)
1613 * @priv_flags: Like 'flags' but invisible to userspace,
1614 * see if.h for the definitions
1615 * @gflags: Global flags ( kept as legacy )
1616 * @padded: How much padding added by alloc_netdev()
1617 * @operstate: RFC2863 operstate
1618 * @link_mode: Mapping policy to operstate
1619 * @if_port: Selectable AUI, TP, ...
1620 * @dma: DMA channel
1621 * @mtu: Interface MTU value
1622 * @min_mtu: Interface Minimum MTU value
1623 * @max_mtu: Interface Maximum MTU value
1624 * @type: Interface hardware type
1625 * @hard_header_len: Maximum hardware header length.
1626 * @min_header_len: Minimum hardware header length
1627 *
1628 * @needed_headroom: Extra headroom the hardware may need, but not in all
1629 * cases can this be guaranteed
1630 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1631 * cases can this be guaranteed. Some cases also use
1632 * LL_MAX_HEADER instead to allocate the skb
1633 *
1634 * interface address info:
1635 *
1636 * @perm_addr: Permanent hw address
1637 * @addr_assign_type: Hw address assignment type
1638 * @addr_len: Hardware address length
1639 * @neigh_priv_len: Used in neigh_alloc()
1640 * @dev_id: Used to differentiate devices that share
1641 * the same link layer address
1642 * @dev_port: Used to differentiate devices that share
1643 * the same function
1644 * @addr_list_lock: XXX: need comments on this one
1645 * @uc_promisc: Counter that indicates promiscuous mode
1646 * has been enabled due to the need to listen to
1647 * additional unicast addresses in a device that
1648 * does not implement ndo_set_rx_mode()
1649 * @uc: unicast mac addresses
1650 * @mc: multicast mac addresses
1651 * @dev_addrs: list of device hw addresses
1652 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1653 * @promiscuity: Number of times the NIC is told to work in
1654 * promiscuous mode; if it becomes 0 the NIC will
1655 * exit promiscuous mode
1656 * @allmulti: Counter, enables or disables allmulticast mode
1657 *
1658 * @vlan_info: VLAN info
1659 * @dsa_ptr: dsa specific data
1660 * @tipc_ptr: TIPC specific data
1661 * @atalk_ptr: AppleTalk link
1662 * @ip_ptr: IPv4 specific data
1663 * @dn_ptr: DECnet specific data
1664 * @ip6_ptr: IPv6 specific data
1665 * @ax25_ptr: AX.25 specific data
1666 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1667 *
1668 * @dev_addr: Hw address (before bcast,
1669 * because most packets are unicast)
1670 *
1671 * @_rx: Array of RX queues
1672 * @num_rx_queues: Number of RX queues
1673 * allocated at register_netdev() time
1674 * @real_num_rx_queues: Number of RX queues currently active in device
1675 *
1676 * @rx_handler: handler for received packets
1677 * @rx_handler_data: XXX: need comments on this one
1678 * @miniq_ingress: ingress/clsact qdisc specific data for
1679 * ingress processing
1680 * @ingress_queue: XXX: need comments on this one
1681 * @broadcast: hw bcast address
1682 *
1683 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1684 * indexed by RX queue number. Assigned by driver.
1685 * This must only be set if the ndo_rx_flow_steer
1686 * operation is defined
1687 * @index_hlist: Device index hash chain
1688 *
1689 * @_tx: Array of TX queues
1690 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1691 * @real_num_tx_queues: Number of TX queues currently active in device
1692 * @qdisc: Root qdisc from userspace point of view
1693 * @tx_queue_len: Max frames per queue allowed
1694 * @tx_global_lock: XXX: need comments on this one
1695 *
1696 * @xps_maps: XXX: need comments on this one
1697 * @miniq_egress: clsact qdisc specific data for
1698 * egress processing
1699 * @watchdog_timeo: Represents the timeout that is used by
1700 * the watchdog (see dev_watchdog())
1701 * @watchdog_timer: List of timers
1702 *
1703 * @pcpu_refcnt: Number of references to this device
1704 * @todo_list: Delayed register/unregister
1705 * @link_watch_list: XXX: need comments on this one
1706 *
1707 * @reg_state: Register/unregister state machine
1708 * @dismantle: Device is going to be freed
1709 * @rtnl_link_state: This enum represents the phases of creating
1710 * a new link
1711 *
1712 * @needs_free_netdev: Should unregister perform free_netdev?
1713 * @priv_destructor: Called from unregister
1714 * @npinfo: XXX: need comments on this one
1715 * @nd_net: Network namespace this network device is inside
1716 *
1717 * @ml_priv: Mid-layer private
1718 * @lstats: Loopback statistics
1719 * @tstats: Tunnel statistics
1720 * @dstats: Dummy statistics
1721 * @vstats: Virtual ethernet statistics
1722 *
1723 * @garp_port: GARP
1724 * @mrp_port: MRP
1725 *
1726 * @dev: Class/net/name entry
1727 * @sysfs_groups: Space for optional device, statistics and wireless
1728 * sysfs groups
1729 *
1730 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1731 * @rtnl_link_ops: Rtnl_link_ops
1732 *
1733 * @gso_max_size: Maximum size of generic segmentation offload
1734 * @gso_max_segs: Maximum number of segments that can be passed to the
1735 * NIC for GSO
1736 *
1737 * @dcbnl_ops: Data Center Bridging netlink ops
1738 * @num_tc: Number of traffic classes in the net device
1739 * @tc_to_txq: XXX: need comments on this one
1740 * @prio_tc_map: XXX: need comments on this one
1741 *
1742 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1743 *
1744 * @priomap: XXX: need comments on this one
1745 * @phydev: Physical device may attach itself
1746 * for hardware timestamping
1747 * @sfp_bus: attached &struct sfp_bus structure.
1748 *
1749 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1750 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1751 *
1752 * @proto_down: protocol port state information can be sent to the
1753 * switch driver and used to set the phys state of the
1754 * switch port.
1755 *
1756 * @wol_enabled: Wake-on-LAN is enabled
1757 *
1758 * FIXME: cleanup struct net_device such that network protocol info
1759 * moves out.
1760 */
1761
1762struct net_device {
1763 char name[IFNAMSIZ];
1764 struct hlist_node name_hlist;
1765 struct dev_ifalias __rcu *ifalias;
1766 /*
1767 * I/O specific fields
1768 * FIXME: Merge these and struct ifmap into one
1769 */
1770 unsigned long mem_end;
1771 unsigned long mem_start;
1772 unsigned long base_addr;
1773 int irq;
1774
1775 /*
1776 * Some hardware also needs these fields (state,dev_list,
1777 * napi_list,unreg_list,close_list) but they are not
1778 * part of the usual set specified in Space.c.
1779 */
1780
1781 unsigned long state;
1782
1783 struct list_head dev_list;
1784 struct list_head napi_list;
1785 struct list_head unreg_list;
1786 struct list_head close_list;
1787 struct list_head ptype_all;
1788 struct list_head ptype_specific;
1789
1790 struct {
1791 struct list_head upper;
1792 struct list_head lower;
1793 } adj_list;
1794
1795 netdev_features_t features;
1796 netdev_features_t hw_features;
1797 netdev_features_t wanted_features;
1798 netdev_features_t vlan_features;
1799 netdev_features_t hw_enc_features;
1800 netdev_features_t mpls_features;
1801 netdev_features_t gso_partial_features;
1802
1803 int ifindex;
1804 int group;
1805
1806 struct net_device_stats stats;
1807
1808 atomic_long_t rx_dropped;
1809 atomic_long_t tx_dropped;
1810 atomic_long_t rx_nohandler;
1811
1812 /* Stats to monitor link on/off, flapping */
1813 atomic_t carrier_up_count;
1814 atomic_t carrier_down_count;
1815
1816#ifdef CONFIG_WIRELESS_EXT
1817 const struct iw_handler_def *wireless_handlers;
1818 struct iw_public_data *wireless_data;
1819#endif
1820 const struct net_device_ops *netdev_ops;
1821 const struct ethtool_ops *ethtool_ops;
1822#ifdef CONFIG_NET_L3_MASTER_DEV
1823 const struct l3mdev_ops *l3mdev_ops;
1824#endif
1825#if IS_ENABLED(CONFIG_IPV6)
1826 const struct ndisc_ops *ndisc_ops;
1827#endif
1828
1829#ifdef CONFIG_XFRM_OFFLOAD
1830 const struct xfrmdev_ops *xfrmdev_ops;
1831#endif
1832
1833#if IS_ENABLED(CONFIG_TLS_DEVICE)
1834 const struct tlsdev_ops *tlsdev_ops;
1835#endif
1836
1837 const struct header_ops *header_ops;
1838
1839 unsigned int flags;
1840 unsigned int priv_flags;
1841
1842 unsigned short gflags;
1843 unsigned short padded;
1844
1845 unsigned char operstate;
1846 unsigned char link_mode;
1847
1848 unsigned char if_port;
1849 unsigned char dma;
1850
1851 unsigned int mtu;
1852 unsigned int min_mtu;
1853 unsigned int max_mtu;
1854 unsigned short type;
1855 unsigned short hard_header_len;
1856 unsigned char min_header_len;
1857
1858 unsigned short needed_headroom;
1859 unsigned short needed_tailroom;
1860
1861 /* Interface address info. */
1862 unsigned char perm_addr[MAX_ADDR_LEN];
1863 unsigned char addr_assign_type;
1864 unsigned char addr_len;
1865 unsigned short neigh_priv_len;
1866 unsigned short dev_id;
1867 unsigned short dev_port;
1868 spinlock_t addr_list_lock;
1869 unsigned char name_assign_type;
1870 bool uc_promisc;
1871 struct netdev_hw_addr_list uc;
1872 struct netdev_hw_addr_list mc;
1873 struct netdev_hw_addr_list dev_addrs;
1874
1875#ifdef CONFIG_SYSFS
1876 struct kset *queues_kset;
1877#endif
1878 unsigned int promiscuity;
1879 unsigned int allmulti;
1880
1881
1882 /* Protocol-specific pointers */
1883
1884#if IS_ENABLED(CONFIG_VLAN_8021Q)
1885 struct vlan_info __rcu *vlan_info;
1886#endif
1887#if IS_ENABLED(CONFIG_NET_DSA)
1888 struct dsa_port *dsa_ptr;
1889#endif
1890#if IS_ENABLED(CONFIG_TIPC)
1891 struct tipc_bearer __rcu *tipc_ptr;
1892#endif
1893#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1894 void *atalk_ptr;
1895#endif
1896 struct in_device __rcu *ip_ptr;
1897#if IS_ENABLED(CONFIG_DECNET)
1898 struct dn_dev __rcu *dn_ptr;
1899#endif
1900 struct inet6_dev __rcu *ip6_ptr;
1901#if IS_ENABLED(CONFIG_AX25)
1902 void *ax25_ptr;
1903#endif
1904 struct wireless_dev *ieee80211_ptr;
1905 struct wpan_dev *ieee802154_ptr;
1906#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1907 struct mpls_dev __rcu *mpls_ptr;
1908#endif
1909
1910/*
1911 * Cache lines mostly used on receive path (including eth_type_trans())
1912 */
1913 /* Interface address info used in eth_type_trans() */
1914 unsigned char *dev_addr;
1915
1916 struct netdev_rx_queue *_rx;
1917 unsigned int num_rx_queues;
1918 unsigned int real_num_rx_queues;
1919
1920 struct bpf_prog __rcu *xdp_prog;
1921 unsigned long gro_flush_timeout;
1922 rx_handler_func_t __rcu *rx_handler;
1923 void __rcu *rx_handler_data;
1924
1925#ifdef CONFIG_NET_CLS_ACT
1926 struct mini_Qdisc __rcu *miniq_ingress;
1927#endif
1928 struct netdev_queue __rcu *ingress_queue;
1929#ifdef CONFIG_NETFILTER_INGRESS
1930 struct nf_hook_entries __rcu *nf_hooks_ingress;
1931#endif
1932
1933 unsigned char broadcast[MAX_ADDR_LEN];
1934#ifdef CONFIG_RFS_ACCEL
1935 struct cpu_rmap *rx_cpu_rmap;
1936#endif
1937 struct hlist_node index_hlist;
1938
1939/*
1940 * Cache lines mostly used on transmit path
1941 */
1942 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1943 unsigned int num_tx_queues;
1944 unsigned int real_num_tx_queues;
1945 struct Qdisc *qdisc;
1946#ifdef CONFIG_NET_SCHED
1947 DECLARE_HASHTABLE (qdisc_hash, 4);
1948#endif
1949 unsigned int tx_queue_len;
1950 spinlock_t tx_global_lock;
1951 int watchdog_timeo;
1952
1953#ifdef CONFIG_XPS
1954 struct xps_dev_maps __rcu *xps_cpus_map;
1955 struct xps_dev_maps __rcu *xps_rxqs_map;
1956#endif
1957#ifdef CONFIG_NET_CLS_ACT
1958 struct mini_Qdisc __rcu *miniq_egress;
1959#endif
1960
1961 /* These may be needed for future network-power-down code. */
1962 struct timer_list watchdog_timer;
1963
1964 int __percpu *pcpu_refcnt;
1965 struct list_head todo_list;
1966
1967 struct list_head link_watch_list;
1968
1969 enum { NETREG_UNINITIALIZED=0,
1970 NETREG_REGISTERED, /* completed register_netdevice */
1971 NETREG_UNREGISTERING, /* called unregister_netdevice */
1972 NETREG_UNREGISTERED, /* completed unregister todo */
1973 NETREG_RELEASED, /* called free_netdev */
1974 NETREG_DUMMY, /* dummy device for NAPI poll */
1975 } reg_state:8;
1976
1977 bool dismantle;
1978
1979 enum {
1980 RTNL_LINK_INITIALIZED,
1981 RTNL_LINK_INITIALIZING,
1982 } rtnl_link_state:16;
1983
1984 bool needs_free_netdev;
1985 void (*priv_destructor)(struct net_device *dev);
1986
1987#ifdef CONFIG_NETPOLL
1988 struct netpoll_info __rcu *npinfo;
1989#endif
1990
1991 possible_net_t nd_net;
1992
1993 /* mid-layer private */
1994 union {
1995 void *ml_priv;
1996 struct pcpu_lstats __percpu *lstats;
1997 struct pcpu_sw_netstats __percpu *tstats;
1998 struct pcpu_dstats __percpu *dstats;
1999 };
2000
2001#if IS_ENABLED(CONFIG_GARP)
2002 struct garp_port __rcu *garp_port;
2003#endif
2004#if IS_ENABLED(CONFIG_MRP)
2005 struct mrp_port __rcu *mrp_port;
2006#endif
2007
2008 struct device dev;
2009 const struct attribute_group *sysfs_groups[4];
2010 const struct attribute_group *sysfs_rx_queue_group;
2011
2012 const struct rtnl_link_ops *rtnl_link_ops;
2013
2014 /* for setting kernel sock attribute on TCP connection setup */
2015#define GSO_MAX_SIZE 65536
2016 unsigned int gso_max_size;
2017#define GSO_MAX_SEGS 65535
2018 u16 gso_max_segs;
2019
2020#ifdef CONFIG_DCB
2021 const struct dcbnl_rtnl_ops *dcbnl_ops;
2022#endif
2023 s16 num_tc;
2024 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2025 u8 prio_tc_map[TC_BITMASK + 1];
2026
2027#if IS_ENABLED(CONFIG_FCOE)
2028 unsigned int fcoe_ddp_xid;
2029#endif
2030#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2031 struct netprio_map __rcu *priomap;
2032#endif
2033 struct phy_device *phydev;
2034 struct sfp_bus *sfp_bus;
2035 struct lock_class_key *qdisc_tx_busylock;
2036 struct lock_class_key *qdisc_running_key;
2037 bool proto_down;
2038 unsigned wol_enabled:1;
2039};
2040#define to_net_dev(d) container_of(d, struct net_device, dev)
2041
2042static inline bool netif_elide_gro(const struct net_device *dev)
2043{
2044 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2045 return true;
2046 return false;
2047}
2048
2049#define NETDEV_ALIGN 32
2050
2051static inline
2052int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2053{
2054 return dev->prio_tc_map[prio & TC_BITMASK];
2055}
2056
2057static inline
2058int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2059{
2060 if (tc >= dev->num_tc)
2061 return -EINVAL;
2062
2063 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2064 return 0;
2065}
2066
2067int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2068void netdev_reset_tc(struct net_device *dev);
2069int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2070int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2071
2072static inline
2073int netdev_get_num_tc(struct net_device *dev)
2074{
2075 return dev->num_tc;
2076}
2077
2078void netdev_unbind_sb_channel(struct net_device *dev,
2079 struct net_device *sb_dev);
2080int netdev_bind_sb_channel_queue(struct net_device *dev,
2081 struct net_device *sb_dev,
2082 u8 tc, u16 count, u16 offset);
2083int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2084static inline int netdev_get_sb_channel(struct net_device *dev)
2085{
2086 return max_t(int, -dev->num_tc, 0);
2087}
2088
2089static inline
2090struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2091 unsigned int index)
2092{
2093 return &dev->_tx[index];
2094}
2095
2096static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2097 const struct sk_buff *skb)
2098{
2099 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2100}
2101
2102static inline void netdev_for_each_tx_queue(struct net_device *dev,
2103 void (*f)(struct net_device *,
2104 struct netdev_queue *,
2105 void *),
2106 void *arg)
2107{
2108 unsigned int i;
2109
2110 for (i = 0; i < dev->num_tx_queues; i++)
2111 f(dev, &dev->_tx[i], arg);
2112}
2113
2114#define netdev_lockdep_set_classes(dev) \
2115{ \
2116 static struct lock_class_key qdisc_tx_busylock_key; \
2117 static struct lock_class_key qdisc_running_key; \
2118 static struct lock_class_key qdisc_xmit_lock_key; \
2119 static struct lock_class_key dev_addr_list_lock_key; \
2120 unsigned int i; \
2121 \
2122 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2123 (dev)->qdisc_running_key = &qdisc_running_key; \
2124 lockdep_set_class(&(dev)->addr_list_lock, \
2125 &dev_addr_list_lock_key); \
2126 for (i = 0; i < (dev)->num_tx_queues; i++) \
2127 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2128 &qdisc_xmit_lock_key); \
2129}
2130
2131u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2132 struct net_device *sb_dev);
2133struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2134 struct sk_buff *skb,
2135 struct net_device *sb_dev);
2136
2137/* returns the headroom that the master device needs to take in account
2138 * when forwarding to this dev
2139 */
2140static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2141{
2142 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2143}
2144
2145static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2146{
2147 if (dev->netdev_ops->ndo_set_rx_headroom)
2148 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2149}
2150
2151/* set the device rx headroom to the dev's default */
2152static inline void netdev_reset_rx_headroom(struct net_device *dev)
2153{
2154 netdev_set_rx_headroom(dev, -1);
2155}
2156
2157/*
2158 * Net namespace inlines
2159 */
2160static inline
2161struct net *dev_net(const struct net_device *dev)
2162{
2163 return read_pnet(&dev->nd_net);
2164}
2165
2166static inline
2167void dev_net_set(struct net_device *dev, struct net *net)
2168{
2169 write_pnet(&dev->nd_net, net);
2170}
2171
2172/**
2173 * netdev_priv - access network device private data
2174 * @dev: network device
2175 *
2176 * Get network device private data
2177 */
2178static inline void *netdev_priv(const struct net_device *dev)
2179{
2180 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2181}
2182
2183/* Set the sysfs physical device reference for the network logical device
2184 * if set prior to registration will cause a symlink during initialization.
2185 */
2186#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2187
2188/* Set the sysfs device type for the network logical device to allow
2189 * fine-grained identification of different network device types. For
2190 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2191 */
2192#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2193
2194/* Default NAPI poll() weight
2195 * Device drivers are strongly advised to not use bigger value
2196 */
2197#define NAPI_POLL_WEIGHT 64
2198
2199/**
2200 * netif_napi_add - initialize a NAPI context
2201 * @dev: network device
2202 * @napi: NAPI context
2203 * @poll: polling function
2204 * @weight: default weight
2205 *
2206 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2207 * *any* of the other NAPI-related functions.
2208 */
2209void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2210 int (*poll)(struct napi_struct *, int), int weight);
2211
2212/**
2213 * netif_tx_napi_add - initialize a NAPI context
2214 * @dev: network device
2215 * @napi: NAPI context
2216 * @poll: polling function
2217 * @weight: default weight
2218 *
2219 * This variant of netif_napi_add() should be used from drivers using NAPI
2220 * to exclusively poll a TX queue.
2221 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2222 */
2223static inline void netif_tx_napi_add(struct net_device *dev,
2224 struct napi_struct *napi,
2225 int (*poll)(struct napi_struct *, int),
2226 int weight)
2227{
2228 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2229 netif_napi_add(dev, napi, poll, weight);
2230}
2231
2232/**
2233 * netif_napi_del - remove a NAPI context
2234 * @napi: NAPI context
2235 *
2236 * netif_napi_del() removes a NAPI context from the network device NAPI list
2237 */
2238void netif_napi_del(struct napi_struct *napi);
2239
2240struct napi_gro_cb {
2241 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2242 void *frag0;
2243
2244 /* Length of frag0. */
2245 unsigned int frag0_len;
2246
2247 /* This indicates where we are processing relative to skb->data. */
2248 int data_offset;
2249
2250 /* This is non-zero if the packet cannot be merged with the new skb. */
2251 u16 flush;
2252
2253 /* Save the IP ID here and check when we get to the transport layer */
2254 u16 flush_id;
2255
2256 /* Number of segments aggregated. */
2257 u16 count;
2258
2259 /* Start offset for remote checksum offload */
2260 u16 gro_remcsum_start;
2261
2262 /* jiffies when first packet was created/queued */
2263 unsigned long age;
2264
2265 /* Used in ipv6_gro_receive() and foo-over-udp */
2266 u16 proto;
2267
2268 /* This is non-zero if the packet may be of the same flow. */
2269 u8 same_flow:1;
2270
2271 /* Used in tunnel GRO receive */
2272 u8 encap_mark:1;
2273
2274 /* GRO checksum is valid */
2275 u8 csum_valid:1;
2276
2277 /* Number of checksums via CHECKSUM_UNNECESSARY */
2278 u8 csum_cnt:3;
2279
2280 /* Free the skb? */
2281 u8 free:2;
2282#define NAPI_GRO_FREE 1
2283#define NAPI_GRO_FREE_STOLEN_HEAD 2
2284
2285 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2286 u8 is_ipv6:1;
2287
2288 /* Used in GRE, set in fou/gue_gro_receive */
2289 u8 is_fou:1;
2290
2291 /* Used to determine if flush_id can be ignored */
2292 u8 is_atomic:1;
2293
2294 /* Number of gro_receive callbacks this packet already went through */
2295 u8 recursion_counter:4;
2296
2297 /* 1 bit hole */
2298
2299 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2300 __wsum csum;
2301
2302 /* used in skb_gro_receive() slow path */
2303 struct sk_buff *last;
2304};
2305
2306#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2307
2308#define GRO_RECURSION_LIMIT 15
2309static inline int gro_recursion_inc_test(struct sk_buff *skb)
2310{
2311 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2312}
2313
2314typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2315static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2316 struct list_head *head,
2317 struct sk_buff *skb)
2318{
2319 if (unlikely(gro_recursion_inc_test(skb))) {
2320 NAPI_GRO_CB(skb)->flush |= 1;
2321 return NULL;
2322 }
2323
2324 return cb(head, skb);
2325}
2326
2327typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2328 struct sk_buff *);
2329static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2330 struct sock *sk,
2331 struct list_head *head,
2332 struct sk_buff *skb)
2333{
2334 if (unlikely(gro_recursion_inc_test(skb))) {
2335 NAPI_GRO_CB(skb)->flush |= 1;
2336 return NULL;
2337 }
2338
2339 return cb(sk, head, skb);
2340}
2341
2342struct packet_type {
2343 __be16 type; /* This is really htons(ether_type). */
2344 bool ignore_outgoing;
2345 struct net_device *dev; /* NULL is wildcarded here */
2346 int (*func) (struct sk_buff *,
2347 struct net_device *,
2348 struct packet_type *,
2349 struct net_device *);
2350 void (*list_func) (struct list_head *,
2351 struct packet_type *,
2352 struct net_device *);
2353 bool (*id_match)(struct packet_type *ptype,
2354 struct sock *sk);
2355 void *af_packet_priv;
2356 struct list_head list;
2357};
2358
2359struct offload_callbacks {
2360 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2361 netdev_features_t features);
2362 struct sk_buff *(*gro_receive)(struct list_head *head,
2363 struct sk_buff *skb);
2364 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2365};
2366
2367struct packet_offload {
2368 __be16 type; /* This is really htons(ether_type). */
2369 u16 priority;
2370 struct offload_callbacks callbacks;
2371 struct list_head list;
2372};
2373
2374/* often modified stats are per-CPU, other are shared (netdev->stats) */
2375struct pcpu_sw_netstats {
2376 u64 rx_packets;
2377 u64 rx_bytes;
2378 u64 tx_packets;
2379 u64 tx_bytes;
2380 struct u64_stats_sync syncp;
2381} __aligned(4 * sizeof(u64));
2382
2383struct pcpu_lstats {
2384 u64 packets;
2385 u64 bytes;
2386 struct u64_stats_sync syncp;
2387} __aligned(2 * sizeof(u64));
2388
2389#define __netdev_alloc_pcpu_stats(type, gfp) \
2390({ \
2391 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2392 if (pcpu_stats) { \
2393 int __cpu; \
2394 for_each_possible_cpu(__cpu) { \
2395 typeof(type) *stat; \
2396 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2397 u64_stats_init(&stat->syncp); \
2398 } \
2399 } \
2400 pcpu_stats; \
2401})
2402
2403#define netdev_alloc_pcpu_stats(type) \
2404 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2405
2406enum netdev_lag_tx_type {
2407 NETDEV_LAG_TX_TYPE_UNKNOWN,
2408 NETDEV_LAG_TX_TYPE_RANDOM,
2409 NETDEV_LAG_TX_TYPE_BROADCAST,
2410 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2411 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2412 NETDEV_LAG_TX_TYPE_HASH,
2413};
2414
2415enum netdev_lag_hash {
2416 NETDEV_LAG_HASH_NONE,
2417 NETDEV_LAG_HASH_L2,
2418 NETDEV_LAG_HASH_L34,
2419 NETDEV_LAG_HASH_L23,
2420 NETDEV_LAG_HASH_E23,
2421 NETDEV_LAG_HASH_E34,
2422 NETDEV_LAG_HASH_UNKNOWN,
2423};
2424
2425struct netdev_lag_upper_info {
2426 enum netdev_lag_tx_type tx_type;
2427 enum netdev_lag_hash hash_type;
2428};
2429
2430struct netdev_lag_lower_state_info {
2431 u8 link_up : 1,
2432 tx_enabled : 1;
2433};
2434
2435#include <linux/notifier.h>
2436
2437/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2438 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2439 * adding new types.
2440 */
2441enum netdev_cmd {
2442 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2443 NETDEV_DOWN,
2444 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2445 detected a hardware crash and restarted
2446 - we can use this eg to kick tcp sessions
2447 once done */
2448 NETDEV_CHANGE, /* Notify device state change */
2449 NETDEV_REGISTER,
2450 NETDEV_UNREGISTER,
2451 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2452 NETDEV_CHANGEADDR, /* notify after the address change */
2453 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2454 NETDEV_GOING_DOWN,
2455 NETDEV_CHANGENAME,
2456 NETDEV_FEAT_CHANGE,
2457 NETDEV_BONDING_FAILOVER,
2458 NETDEV_PRE_UP,
2459 NETDEV_PRE_TYPE_CHANGE,
2460 NETDEV_POST_TYPE_CHANGE,
2461 NETDEV_POST_INIT,
2462 NETDEV_RELEASE,
2463 NETDEV_NOTIFY_PEERS,
2464 NETDEV_JOIN,
2465 NETDEV_CHANGEUPPER,
2466 NETDEV_RESEND_IGMP,
2467 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2468 NETDEV_CHANGEINFODATA,
2469 NETDEV_BONDING_INFO,
2470 NETDEV_PRECHANGEUPPER,
2471 NETDEV_CHANGELOWERSTATE,
2472 NETDEV_UDP_TUNNEL_PUSH_INFO,
2473 NETDEV_UDP_TUNNEL_DROP_INFO,
2474 NETDEV_CHANGE_TX_QUEUE_LEN,
2475 NETDEV_CVLAN_FILTER_PUSH_INFO,
2476 NETDEV_CVLAN_FILTER_DROP_INFO,
2477 NETDEV_SVLAN_FILTER_PUSH_INFO,
2478 NETDEV_SVLAN_FILTER_DROP_INFO,
2479};
2480const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2481
2482int register_netdevice_notifier(struct notifier_block *nb);
2483int unregister_netdevice_notifier(struct notifier_block *nb);
2484
2485struct netdev_notifier_info {
2486 struct net_device *dev;
2487 struct netlink_ext_ack *extack;
2488};
2489
2490struct netdev_notifier_info_ext {
2491 struct netdev_notifier_info info; /* must be first */
2492 union {
2493 u32 mtu;
2494 } ext;
2495};
2496
2497struct netdev_notifier_change_info {
2498 struct netdev_notifier_info info; /* must be first */
2499 unsigned int flags_changed;
2500};
2501
2502struct netdev_notifier_changeupper_info {
2503 struct netdev_notifier_info info; /* must be first */
2504 struct net_device *upper_dev; /* new upper dev */
2505 bool master; /* is upper dev master */
2506 bool linking; /* is the notification for link or unlink */
2507 void *upper_info; /* upper dev info */
2508};
2509
2510struct netdev_notifier_changelowerstate_info {
2511 struct netdev_notifier_info info; /* must be first */
2512 void *lower_state_info; /* is lower dev state */
2513};
2514
2515struct netdev_notifier_pre_changeaddr_info {
2516 struct netdev_notifier_info info; /* must be first */
2517 const unsigned char *dev_addr;
2518};
2519
2520static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2521 struct net_device *dev)
2522{
2523 info->dev = dev;
2524 info->extack = NULL;
2525}
2526
2527static inline struct net_device *
2528netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2529{
2530 return info->dev;
2531}
2532
2533static inline struct netlink_ext_ack *
2534netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2535{
2536 return info->extack;
2537}
2538
2539int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2540
2541
2542extern rwlock_t dev_base_lock; /* Device list lock */
2543
2544#define for_each_netdev(net, d) \
2545 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2546#define for_each_netdev_reverse(net, d) \
2547 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2548#define for_each_netdev_rcu(net, d) \
2549 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2550#define for_each_netdev_safe(net, d, n) \
2551 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2552#define for_each_netdev_continue(net, d) \
2553 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2554#define for_each_netdev_continue_rcu(net, d) \
2555 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2556#define for_each_netdev_in_bond_rcu(bond, slave) \
2557 for_each_netdev_rcu(&init_net, slave) \
2558 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2559#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2560
2561static inline struct net_device *next_net_device(struct net_device *dev)
2562{
2563 struct list_head *lh;
2564 struct net *net;
2565
2566 net = dev_net(dev);
2567 lh = dev->dev_list.next;
2568 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2569}
2570
2571static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2572{
2573 struct list_head *lh;
2574 struct net *net;
2575
2576 net = dev_net(dev);
2577 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2578 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2579}
2580
2581static inline struct net_device *first_net_device(struct net *net)
2582{
2583 return list_empty(&net->dev_base_head) ? NULL :
2584 net_device_entry(net->dev_base_head.next);
2585}
2586
2587static inline struct net_device *first_net_device_rcu(struct net *net)
2588{
2589 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2590
2591 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2592}
2593
2594int netdev_boot_setup_check(struct net_device *dev);
2595unsigned long netdev_boot_base(const char *prefix, int unit);
2596struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2597 const char *hwaddr);
2598struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2599struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2600void dev_add_pack(struct packet_type *pt);
2601void dev_remove_pack(struct packet_type *pt);
2602void __dev_remove_pack(struct packet_type *pt);
2603void dev_add_offload(struct packet_offload *po);
2604void dev_remove_offload(struct packet_offload *po);
2605
2606int dev_get_iflink(const struct net_device *dev);
2607int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2608struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2609 unsigned short mask);
2610struct net_device *dev_get_by_name(struct net *net, const char *name);
2611struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2612struct net_device *__dev_get_by_name(struct net *net, const char *name);
2613int dev_alloc_name(struct net_device *dev, const char *name);
2614int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2615void dev_close(struct net_device *dev);
2616void dev_close_many(struct list_head *head, bool unlink);
2617void dev_disable_lro(struct net_device *dev);
2618int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2619u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2620 struct net_device *sb_dev);
2621u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2622 struct net_device *sb_dev);
2623int dev_queue_xmit(struct sk_buff *skb);
2624int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2625int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2626int register_netdevice(struct net_device *dev);
2627void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2628void unregister_netdevice_many(struct list_head *head);
2629static inline void unregister_netdevice(struct net_device *dev)
2630{
2631 unregister_netdevice_queue(dev, NULL);
2632}
2633
2634int netdev_refcnt_read(const struct net_device *dev);
2635void free_netdev(struct net_device *dev);
2636void netdev_freemem(struct net_device *dev);
2637void synchronize_net(void);
2638int init_dummy_netdev(struct net_device *dev);
2639
2640struct net_device *dev_get_by_index(struct net *net, int ifindex);
2641struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2642struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2643struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2644int netdev_get_name(struct net *net, char *name, int ifindex);
2645int dev_restart(struct net_device *dev);
2646int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2647
2648static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2649{
2650 return NAPI_GRO_CB(skb)->data_offset;
2651}
2652
2653static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2654{
2655 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2656}
2657
2658static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2659{
2660 NAPI_GRO_CB(skb)->data_offset += len;
2661}
2662
2663static inline void *skb_gro_header_fast(struct sk_buff *skb,
2664 unsigned int offset)
2665{
2666 return NAPI_GRO_CB(skb)->frag0 + offset;
2667}
2668
2669static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2670{
2671 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2672}
2673
2674static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2675{
2676 NAPI_GRO_CB(skb)->frag0 = NULL;
2677 NAPI_GRO_CB(skb)->frag0_len = 0;
2678}
2679
2680static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2681 unsigned int offset)
2682{
2683 if (!pskb_may_pull(skb, hlen))
2684 return NULL;
2685
2686 skb_gro_frag0_invalidate(skb);
2687 return skb->data + offset;
2688}
2689
2690static inline void *skb_gro_network_header(struct sk_buff *skb)
2691{
2692 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2693 skb_network_offset(skb);
2694}
2695
2696static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2697 const void *start, unsigned int len)
2698{
2699 if (NAPI_GRO_CB(skb)->csum_valid)
2700 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2701 csum_partial(start, len, 0));
2702}
2703
2704/* GRO checksum functions. These are logical equivalents of the normal
2705 * checksum functions (in skbuff.h) except that they operate on the GRO
2706 * offsets and fields in sk_buff.
2707 */
2708
2709__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2710
2711static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2712{
2713 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2714}
2715
2716static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2717 bool zero_okay,
2718 __sum16 check)
2719{
2720 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2721 skb_checksum_start_offset(skb) <
2722 skb_gro_offset(skb)) &&
2723 !skb_at_gro_remcsum_start(skb) &&
2724 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2725 (!zero_okay || check));
2726}
2727
2728static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2729 __wsum psum)
2730{
2731 if (NAPI_GRO_CB(skb)->csum_valid &&
2732 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2733 return 0;
2734
2735 NAPI_GRO_CB(skb)->csum = psum;
2736
2737 return __skb_gro_checksum_complete(skb);
2738}
2739
2740static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2741{
2742 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2743 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2744 NAPI_GRO_CB(skb)->csum_cnt--;
2745 } else {
2746 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2747 * verified a new top level checksum or an encapsulated one
2748 * during GRO. This saves work if we fallback to normal path.
2749 */
2750 __skb_incr_checksum_unnecessary(skb);
2751 }
2752}
2753
2754#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2755 compute_pseudo) \
2756({ \
2757 __sum16 __ret = 0; \
2758 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2759 __ret = __skb_gro_checksum_validate_complete(skb, \
2760 compute_pseudo(skb, proto)); \
2761 if (!__ret) \
2762 skb_gro_incr_csum_unnecessary(skb); \
2763 __ret; \
2764})
2765
2766#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2767 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2768
2769#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2770 compute_pseudo) \
2771 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2772
2773#define skb_gro_checksum_simple_validate(skb) \
2774 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2775
2776static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2777{
2778 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2779 !NAPI_GRO_CB(skb)->csum_valid);
2780}
2781
2782static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2783 __sum16 check, __wsum pseudo)
2784{
2785 NAPI_GRO_CB(skb)->csum = ~pseudo;
2786 NAPI_GRO_CB(skb)->csum_valid = 1;
2787}
2788
2789#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2790do { \
2791 if (__skb_gro_checksum_convert_check(skb)) \
2792 __skb_gro_checksum_convert(skb, check, \
2793 compute_pseudo(skb, proto)); \
2794} while (0)
2795
2796struct gro_remcsum {
2797 int offset;
2798 __wsum delta;
2799};
2800
2801static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2802{
2803 grc->offset = 0;
2804 grc->delta = 0;
2805}
2806
2807static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2808 unsigned int off, size_t hdrlen,
2809 int start, int offset,
2810 struct gro_remcsum *grc,
2811 bool nopartial)
2812{
2813 __wsum delta;
2814 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2815
2816 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2817
2818 if (!nopartial) {
2819 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2820 return ptr;
2821 }
2822
2823 ptr = skb_gro_header_fast(skb, off);
2824 if (skb_gro_header_hard(skb, off + plen)) {
2825 ptr = skb_gro_header_slow(skb, off + plen, off);
2826 if (!ptr)
2827 return NULL;
2828 }
2829
2830 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2831 start, offset);
2832
2833 /* Adjust skb->csum since we changed the packet */
2834 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2835
2836 grc->offset = off + hdrlen + offset;
2837 grc->delta = delta;
2838
2839 return ptr;
2840}
2841
2842static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2843 struct gro_remcsum *grc)
2844{
2845 void *ptr;
2846 size_t plen = grc->offset + sizeof(u16);
2847
2848 if (!grc->delta)
2849 return;
2850
2851 ptr = skb_gro_header_fast(skb, grc->offset);
2852 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2853 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2854 if (!ptr)
2855 return;
2856 }
2857
2858 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2859}
2860
2861#ifdef CONFIG_XFRM_OFFLOAD
2862static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2863{
2864 if (PTR_ERR(pp) != -EINPROGRESS)
2865 NAPI_GRO_CB(skb)->flush |= flush;
2866}
2867static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2868 struct sk_buff *pp,
2869 int flush,
2870 struct gro_remcsum *grc)
2871{
2872 if (PTR_ERR(pp) != -EINPROGRESS) {
2873 NAPI_GRO_CB(skb)->flush |= flush;
2874 skb_gro_remcsum_cleanup(skb, grc);
2875 skb->remcsum_offload = 0;
2876 }
2877}
2878#else
2879static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2880{
2881 NAPI_GRO_CB(skb)->flush |= flush;
2882}
2883static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2884 struct sk_buff *pp,
2885 int flush,
2886 struct gro_remcsum *grc)
2887{
2888 NAPI_GRO_CB(skb)->flush |= flush;
2889 skb_gro_remcsum_cleanup(skb, grc);
2890 skb->remcsum_offload = 0;
2891}
2892#endif
2893
2894static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2895 unsigned short type,
2896 const void *daddr, const void *saddr,
2897 unsigned int len)
2898{
2899 if (!dev->header_ops || !dev->header_ops->create)
2900 return 0;
2901
2902 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2903}
2904
2905static inline int dev_parse_header(const struct sk_buff *skb,
2906 unsigned char *haddr)
2907{
2908 const struct net_device *dev = skb->dev;
2909
2910 if (!dev->header_ops || !dev->header_ops->parse)
2911 return 0;
2912 return dev->header_ops->parse(skb, haddr);
2913}
2914
2915static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2916{
2917 const struct net_device *dev = skb->dev;
2918
2919 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2920 return 0;
2921 return dev->header_ops->parse_protocol(skb);
2922}
2923
2924/* ll_header must have at least hard_header_len allocated */
2925static inline bool dev_validate_header(const struct net_device *dev,
2926 char *ll_header, int len)
2927{
2928 if (likely(len >= dev->hard_header_len))
2929 return true;
2930 if (len < dev->min_header_len)
2931 return false;
2932
2933 if (capable(CAP_SYS_RAWIO)) {
2934 memset(ll_header + len, 0, dev->hard_header_len - len);
2935 return true;
2936 }
2937
2938 if (dev->header_ops && dev->header_ops->validate)
2939 return dev->header_ops->validate(ll_header, len);
2940
2941 return false;
2942}
2943
2944typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2945 int len, int size);
2946int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2947static inline int unregister_gifconf(unsigned int family)
2948{
2949 return register_gifconf(family, NULL);
2950}
2951
2952#ifdef CONFIG_NET_FLOW_LIMIT
2953#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2954struct sd_flow_limit {
2955 u64 count;
2956 unsigned int num_buckets;
2957 unsigned int history_head;
2958 u16 history[FLOW_LIMIT_HISTORY];
2959 u8 buckets[];
2960};
2961
2962extern int netdev_flow_limit_table_len;
2963#endif /* CONFIG_NET_FLOW_LIMIT */
2964
2965/*
2966 * Incoming packets are placed on per-CPU queues
2967 */
2968struct softnet_data {
2969 struct list_head poll_list;
2970 struct sk_buff_head process_queue;
2971
2972 /* stats */
2973 unsigned int processed;
2974 unsigned int time_squeeze;
2975 unsigned int received_rps;
2976#ifdef CONFIG_RPS
2977 struct softnet_data *rps_ipi_list;
2978#endif
2979#ifdef CONFIG_NET_FLOW_LIMIT
2980 struct sd_flow_limit __rcu *flow_limit;
2981#endif
2982 struct Qdisc *output_queue;
2983 struct Qdisc **output_queue_tailp;
2984 struct sk_buff *completion_queue;
2985#ifdef CONFIG_XFRM_OFFLOAD
2986 struct sk_buff_head xfrm_backlog;
2987#endif
2988 /* written and read only by owning cpu: */
2989 struct {
2990 u16 recursion;
2991 u8 more;
2992 } xmit;
2993#ifdef CONFIG_RPS
2994 /* input_queue_head should be written by cpu owning this struct,
2995 * and only read by other cpus. Worth using a cache line.
2996 */
2997 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2998
2999 /* Elements below can be accessed between CPUs for RPS/RFS */
3000 call_single_data_t csd ____cacheline_aligned_in_smp;
3001 struct softnet_data *rps_ipi_next;
3002 unsigned int cpu;
3003 unsigned int input_queue_tail;
3004#endif
3005 unsigned int dropped;
3006 struct sk_buff_head input_pkt_queue;
3007 struct napi_struct backlog;
3008
3009};
3010
3011static inline void input_queue_head_incr(struct softnet_data *sd)
3012{
3013#ifdef CONFIG_RPS
3014 sd->input_queue_head++;
3015#endif
3016}
3017
3018static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3019 unsigned int *qtail)
3020{
3021#ifdef CONFIG_RPS
3022 *qtail = ++sd->input_queue_tail;
3023#endif
3024}
3025
3026DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3027
3028static inline int dev_recursion_level(void)
3029{
3030 return this_cpu_read(softnet_data.xmit.recursion);
3031}
3032
3033#define XMIT_RECURSION_LIMIT 10
3034static inline bool dev_xmit_recursion(void)
3035{
3036 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3037 XMIT_RECURSION_LIMIT);
3038}
3039
3040static inline void dev_xmit_recursion_inc(void)
3041{
3042 __this_cpu_inc(softnet_data.xmit.recursion);
3043}
3044
3045static inline void dev_xmit_recursion_dec(void)
3046{
3047 __this_cpu_dec(softnet_data.xmit.recursion);
3048}
3049
3050void __netif_schedule(struct Qdisc *q);
3051void netif_schedule_queue(struct netdev_queue *txq);
3052
3053static inline void netif_tx_schedule_all(struct net_device *dev)
3054{
3055 unsigned int i;
3056
3057 for (i = 0; i < dev->num_tx_queues; i++)
3058 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3059}
3060
3061static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3062{
3063 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3064}
3065
3066/**
3067 * netif_start_queue - allow transmit
3068 * @dev: network device
3069 *
3070 * Allow upper layers to call the device hard_start_xmit routine.
3071 */
3072static inline void netif_start_queue(struct net_device *dev)
3073{
3074 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3075}
3076
3077static inline void netif_tx_start_all_queues(struct net_device *dev)
3078{
3079 unsigned int i;
3080
3081 for (i = 0; i < dev->num_tx_queues; i++) {
3082 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3083 netif_tx_start_queue(txq);
3084 }
3085}
3086
3087void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3088
3089/**
3090 * netif_wake_queue - restart transmit
3091 * @dev: network device
3092 *
3093 * Allow upper layers to call the device hard_start_xmit routine.
3094 * Used for flow control when transmit resources are available.
3095 */
3096static inline void netif_wake_queue(struct net_device *dev)
3097{
3098 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3099}
3100
3101static inline void netif_tx_wake_all_queues(struct net_device *dev)
3102{
3103 unsigned int i;
3104
3105 for (i = 0; i < dev->num_tx_queues; i++) {
3106 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3107 netif_tx_wake_queue(txq);
3108 }
3109}
3110
3111static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3112{
3113 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3114}
3115
3116/**
3117 * netif_stop_queue - stop transmitted packets
3118 * @dev: network device
3119 *
3120 * Stop upper layers calling the device hard_start_xmit routine.
3121 * Used for flow control when transmit resources are unavailable.
3122 */
3123static inline void netif_stop_queue(struct net_device *dev)
3124{
3125 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3126}
3127
3128void netif_tx_stop_all_queues(struct net_device *dev);
3129
3130static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3131{
3132 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3133}
3134
3135/**
3136 * netif_queue_stopped - test if transmit queue is flowblocked
3137 * @dev: network device
3138 *
3139 * Test if transmit queue on device is currently unable to send.
3140 */
3141static inline bool netif_queue_stopped(const struct net_device *dev)
3142{
3143 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3144}
3145
3146static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3147{
3148 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3149}
3150
3151static inline bool
3152netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3153{
3154 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3155}
3156
3157static inline bool
3158netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3159{
3160 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3161}
3162
3163/**
3164 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3165 * @dev_queue: pointer to transmit queue
3166 *
3167 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3168 * to give appropriate hint to the CPU.
3169 */
3170static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3171{
3172#ifdef CONFIG_BQL
3173 prefetchw(&dev_queue->dql.num_queued);
3174#endif
3175}
3176
3177/**
3178 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3179 * @dev_queue: pointer to transmit queue
3180 *
3181 * BQL enabled drivers might use this helper in their TX completion path,
3182 * to give appropriate hint to the CPU.
3183 */
3184static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3185{
3186#ifdef CONFIG_BQL
3187 prefetchw(&dev_queue->dql.limit);
3188#endif
3189}
3190
3191static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3192 unsigned int bytes)
3193{
3194#ifdef CONFIG_BQL
3195 dql_queued(&dev_queue->dql, bytes);
3196
3197 if (likely(dql_avail(&dev_queue->dql) >= 0))
3198 return;
3199
3200 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3201
3202 /*
3203 * The XOFF flag must be set before checking the dql_avail below,
3204 * because in netdev_tx_completed_queue we update the dql_completed
3205 * before checking the XOFF flag.
3206 */
3207 smp_mb();
3208
3209 /* check again in case another CPU has just made room avail */
3210 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3211 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3212#endif
3213}
3214
3215/* Variant of netdev_tx_sent_queue() for drivers that are aware
3216 * that they should not test BQL status themselves.
3217 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3218 * skb of a batch.
3219 * Returns true if the doorbell must be used to kick the NIC.
3220 */
3221static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3222 unsigned int bytes,
3223 bool xmit_more)
3224{
3225 if (xmit_more) {
3226#ifdef CONFIG_BQL
3227 dql_queued(&dev_queue->dql, bytes);
3228#endif
3229 return netif_tx_queue_stopped(dev_queue);
3230 }
3231 netdev_tx_sent_queue(dev_queue, bytes);
3232 return true;
3233}
3234
3235/**
3236 * netdev_sent_queue - report the number of bytes queued to hardware
3237 * @dev: network device
3238 * @bytes: number of bytes queued to the hardware device queue
3239 *
3240 * Report the number of bytes queued for sending/completion to the network
3241 * device hardware queue. @bytes should be a good approximation and should
3242 * exactly match netdev_completed_queue() @bytes
3243 */
3244static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3245{
3246 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3247}
3248
3249static inline bool __netdev_sent_queue(struct net_device *dev,
3250 unsigned int bytes,
3251 bool xmit_more)
3252{
3253 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3254 xmit_more);
3255}
3256
3257static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3258 unsigned int pkts, unsigned int bytes)
3259{
3260#ifdef CONFIG_BQL
3261 if (unlikely(!bytes))
3262 return;
3263
3264 dql_completed(&dev_queue->dql, bytes);
3265
3266 /*
3267 * Without the memory barrier there is a small possiblity that
3268 * netdev_tx_sent_queue will miss the update and cause the queue to
3269 * be stopped forever
3270 */
3271 smp_mb();
3272
3273 if (dql_avail(&dev_queue->dql) < 0)
3274 return;
3275
3276 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3277 netif_schedule_queue(dev_queue);
3278#endif
3279}
3280
3281/**
3282 * netdev_completed_queue - report bytes and packets completed by device
3283 * @dev: network device
3284 * @pkts: actual number of packets sent over the medium
3285 * @bytes: actual number of bytes sent over the medium
3286 *
3287 * Report the number of bytes and packets transmitted by the network device
3288 * hardware queue over the physical medium, @bytes must exactly match the
3289 * @bytes amount passed to netdev_sent_queue()
3290 */
3291static inline void netdev_completed_queue(struct net_device *dev,
3292 unsigned int pkts, unsigned int bytes)
3293{
3294 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3295}
3296
3297static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3298{
3299#ifdef CONFIG_BQL
3300 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3301 dql_reset(&q->dql);
3302#endif
3303}
3304
3305/**
3306 * netdev_reset_queue - reset the packets and bytes count of a network device
3307 * @dev_queue: network device
3308 *
3309 * Reset the bytes and packet count of a network device and clear the
3310 * software flow control OFF bit for this network device
3311 */
3312static inline void netdev_reset_queue(struct net_device *dev_queue)
3313{
3314 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3315}
3316
3317/**
3318 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3319 * @dev: network device
3320 * @queue_index: given tx queue index
3321 *
3322 * Returns 0 if given tx queue index >= number of device tx queues,
3323 * otherwise returns the originally passed tx queue index.
3324 */
3325static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3326{
3327 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3328 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3329 dev->name, queue_index,
3330 dev->real_num_tx_queues);
3331 return 0;
3332 }
3333
3334 return queue_index;
3335}
3336
3337/**
3338 * netif_running - test if up
3339 * @dev: network device
3340 *
3341 * Test if the device has been brought up.
3342 */
3343static inline bool netif_running(const struct net_device *dev)
3344{
3345 return test_bit(__LINK_STATE_START, &dev->state);
3346}
3347
3348/*
3349 * Routines to manage the subqueues on a device. We only need start,
3350 * stop, and a check if it's stopped. All other device management is
3351 * done at the overall netdevice level.
3352 * Also test the device if we're multiqueue.
3353 */
3354
3355/**
3356 * netif_start_subqueue - allow sending packets on subqueue
3357 * @dev: network device
3358 * @queue_index: sub queue index
3359 *
3360 * Start individual transmit queue of a device with multiple transmit queues.
3361 */
3362static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3363{
3364 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3365
3366 netif_tx_start_queue(txq);
3367}
3368
3369/**
3370 * netif_stop_subqueue - stop sending packets on subqueue
3371 * @dev: network device
3372 * @queue_index: sub queue index
3373 *
3374 * Stop individual transmit queue of a device with multiple transmit queues.
3375 */
3376static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3377{
3378 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3379 netif_tx_stop_queue(txq);
3380}
3381
3382/**
3383 * netif_subqueue_stopped - test status of subqueue
3384 * @dev: network device
3385 * @queue_index: sub queue index
3386 *
3387 * Check individual transmit queue of a device with multiple transmit queues.
3388 */
3389static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3390 u16 queue_index)
3391{
3392 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3393
3394 return netif_tx_queue_stopped(txq);
3395}
3396
3397static inline bool netif_subqueue_stopped(const struct net_device *dev,
3398 struct sk_buff *skb)
3399{
3400 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3401}
3402
3403/**
3404 * netif_wake_subqueue - allow sending packets on subqueue
3405 * @dev: network device
3406 * @queue_index: sub queue index
3407 *
3408 * Resume individual transmit queue of a device with multiple transmit queues.
3409 */
3410static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3411{
3412 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3413
3414 netif_tx_wake_queue(txq);
3415}
3416
3417#ifdef CONFIG_XPS
3418int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3419 u16 index);
3420int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3421 u16 index, bool is_rxqs_map);
3422
3423/**
3424 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3425 * @j: CPU/Rx queue index
3426 * @mask: bitmask of all cpus/rx queues
3427 * @nr_bits: number of bits in the bitmask
3428 *
3429 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3430 */
3431static inline bool netif_attr_test_mask(unsigned long j,
3432 const unsigned long *mask,
3433 unsigned int nr_bits)
3434{
3435 cpu_max_bits_warn(j, nr_bits);
3436 return test_bit(j, mask);
3437}
3438
3439/**
3440 * netif_attr_test_online - Test for online CPU/Rx queue
3441 * @j: CPU/Rx queue index
3442 * @online_mask: bitmask for CPUs/Rx queues that are online
3443 * @nr_bits: number of bits in the bitmask
3444 *
3445 * Returns true if a CPU/Rx queue is online.
3446 */
3447static inline bool netif_attr_test_online(unsigned long j,
3448 const unsigned long *online_mask,
3449 unsigned int nr_bits)
3450{
3451 cpu_max_bits_warn(j, nr_bits);
3452
3453 if (online_mask)
3454 return test_bit(j, online_mask);
3455
3456 return (j < nr_bits);
3457}
3458
3459/**
3460 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3461 * @n: CPU/Rx queue index
3462 * @srcp: the cpumask/Rx queue mask pointer
3463 * @nr_bits: number of bits in the bitmask
3464 *
3465 * Returns >= nr_bits if no further CPUs/Rx queues set.
3466 */
3467static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3468 unsigned int nr_bits)
3469{
3470 /* -1 is a legal arg here. */
3471 if (n != -1)
3472 cpu_max_bits_warn(n, nr_bits);
3473
3474 if (srcp)
3475 return find_next_bit(srcp, nr_bits, n + 1);
3476
3477 return n + 1;
3478}
3479
3480/**
3481 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3482 * @n: CPU/Rx queue index
3483 * @src1p: the first CPUs/Rx queues mask pointer
3484 * @src2p: the second CPUs/Rx queues mask pointer
3485 * @nr_bits: number of bits in the bitmask
3486 *
3487 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3488 */
3489static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3490 const unsigned long *src2p,
3491 unsigned int nr_bits)
3492{
3493 /* -1 is a legal arg here. */
3494 if (n != -1)
3495 cpu_max_bits_warn(n, nr_bits);
3496
3497 if (src1p && src2p)
3498 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3499 else if (src1p)
3500 return find_next_bit(src1p, nr_bits, n + 1);
3501 else if (src2p)
3502 return find_next_bit(src2p, nr_bits, n + 1);
3503
3504 return n + 1;
3505}
3506#else
3507static inline int netif_set_xps_queue(struct net_device *dev,
3508 const struct cpumask *mask,
3509 u16 index)
3510{
3511 return 0;
3512}
3513
3514static inline int __netif_set_xps_queue(struct net_device *dev,
3515 const unsigned long *mask,
3516 u16 index, bool is_rxqs_map)
3517{
3518 return 0;
3519}
3520#endif
3521
3522/**
3523 * netif_is_multiqueue - test if device has multiple transmit queues
3524 * @dev: network device
3525 *
3526 * Check if device has multiple transmit queues
3527 */
3528static inline bool netif_is_multiqueue(const struct net_device *dev)
3529{
3530 return dev->num_tx_queues > 1;
3531}
3532
3533int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3534
3535#ifdef CONFIG_SYSFS
3536int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3537#else
3538static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3539 unsigned int rxqs)
3540{
3541 dev->real_num_rx_queues = rxqs;
3542 return 0;
3543}
3544#endif
3545
3546static inline struct netdev_rx_queue *
3547__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3548{
3549 return dev->_rx + rxq;
3550}
3551
3552#ifdef CONFIG_SYSFS
3553static inline unsigned int get_netdev_rx_queue_index(
3554 struct netdev_rx_queue *queue)
3555{
3556 struct net_device *dev = queue->dev;
3557 int index = queue - dev->_rx;
3558
3559 BUG_ON(index >= dev->num_rx_queues);
3560 return index;
3561}
3562#endif
3563
3564#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3565int netif_get_num_default_rss_queues(void);
3566
3567enum skb_free_reason {
3568 SKB_REASON_CONSUMED,
3569 SKB_REASON_DROPPED,
3570};
3571
3572void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3573void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3574
3575/*
3576 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3577 * interrupt context or with hardware interrupts being disabled.
3578 * (in_irq() || irqs_disabled())
3579 *
3580 * We provide four helpers that can be used in following contexts :
3581 *
3582 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3583 * replacing kfree_skb(skb)
3584 *
3585 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3586 * Typically used in place of consume_skb(skb) in TX completion path
3587 *
3588 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3589 * replacing kfree_skb(skb)
3590 *
3591 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3592 * and consumed a packet. Used in place of consume_skb(skb)
3593 */
3594static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3595{
3596 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3597}
3598
3599static inline void dev_consume_skb_irq(struct sk_buff *skb)
3600{
3601 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3602}
3603
3604static inline void dev_kfree_skb_any(struct sk_buff *skb)
3605{
3606 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3607}
3608
3609static inline void dev_consume_skb_any(struct sk_buff *skb)
3610{
3611 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3612}
3613
3614void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3615int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3616int netif_rx(struct sk_buff *skb);
3617int netif_rx_ni(struct sk_buff *skb);
3618int netif_receive_skb(struct sk_buff *skb);
3619int netif_receive_skb_core(struct sk_buff *skb);
3620void netif_receive_skb_list(struct list_head *head);
3621gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3622void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3623struct sk_buff *napi_get_frags(struct napi_struct *napi);
3624gro_result_t napi_gro_frags(struct napi_struct *napi);
3625struct packet_offload *gro_find_receive_by_type(__be16 type);
3626struct packet_offload *gro_find_complete_by_type(__be16 type);
3627
3628static inline void napi_free_frags(struct napi_struct *napi)
3629{
3630 kfree_skb(napi->skb);
3631 napi->skb = NULL;
3632}
3633
3634bool netdev_is_rx_handler_busy(struct net_device *dev);
3635int netdev_rx_handler_register(struct net_device *dev,
3636 rx_handler_func_t *rx_handler,
3637 void *rx_handler_data);
3638void netdev_rx_handler_unregister(struct net_device *dev);
3639
3640bool dev_valid_name(const char *name);
3641int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3642 bool *need_copyout);
3643int dev_ifconf(struct net *net, struct ifconf *, int);
3644int dev_ethtool(struct net *net, struct ifreq *);
3645unsigned int dev_get_flags(const struct net_device *);
3646int __dev_change_flags(struct net_device *dev, unsigned int flags,
3647 struct netlink_ext_ack *extack);
3648int dev_change_flags(struct net_device *dev, unsigned int flags,
3649 struct netlink_ext_ack *extack);
3650void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3651 unsigned int gchanges);
3652int dev_change_name(struct net_device *, const char *);
3653int dev_set_alias(struct net_device *, const char *, size_t);
3654int dev_get_alias(const struct net_device *, char *, size_t);
3655int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3656int __dev_set_mtu(struct net_device *, int);
3657int dev_set_mtu_ext(struct net_device *dev, int mtu,
3658 struct netlink_ext_ack *extack);
3659int dev_set_mtu(struct net_device *, int);
3660int dev_change_tx_queue_len(struct net_device *, unsigned long);
3661void dev_set_group(struct net_device *, int);
3662int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3663 struct netlink_ext_ack *extack);
3664int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3665 struct netlink_ext_ack *extack);
3666int dev_change_carrier(struct net_device *, bool new_carrier);
3667int dev_get_phys_port_id(struct net_device *dev,
3668 struct netdev_phys_item_id *ppid);
3669int dev_get_phys_port_name(struct net_device *dev,
3670 char *name, size_t len);
3671int dev_get_port_parent_id(struct net_device *dev,
3672 struct netdev_phys_item_id *ppid, bool recurse);
3673bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3674int dev_change_proto_down(struct net_device *dev, bool proto_down);
3675int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3676struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3677struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3678 struct netdev_queue *txq, int *ret);
3679
3680typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3681int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3682 int fd, u32 flags);
3683u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3684 enum bpf_netdev_command cmd);
3685int xdp_umem_query(struct net_device *dev, u16 queue_id);
3686
3687int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3688int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3689bool is_skb_forwardable(const struct net_device *dev,
3690 const struct sk_buff *skb);
3691
3692static __always_inline int ____dev_forward_skb(struct net_device *dev,
3693 struct sk_buff *skb)
3694{
3695 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3696 unlikely(!is_skb_forwardable(dev, skb))) {
3697 atomic_long_inc(&dev->rx_dropped);
3698 kfree_skb(skb);
3699 return NET_RX_DROP;
3700 }
3701
3702 skb_scrub_packet(skb, true);
3703 skb->priority = 0;
3704 return 0;
3705}
3706
3707bool dev_nit_active(struct net_device *dev);
3708void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3709
3710extern int netdev_budget;
3711extern unsigned int netdev_budget_usecs;
3712
3713/* Called by rtnetlink.c:rtnl_unlock() */
3714void netdev_run_todo(void);
3715
3716/**
3717 * dev_put - release reference to device
3718 * @dev: network device
3719 *
3720 * Release reference to device to allow it to be freed.
3721 */
3722static inline void dev_put(struct net_device *dev)
3723{
3724 this_cpu_dec(*dev->pcpu_refcnt);
3725}
3726
3727/**
3728 * dev_hold - get reference to device
3729 * @dev: network device
3730 *
3731 * Hold reference to device to keep it from being freed.
3732 */
3733static inline void dev_hold(struct net_device *dev)
3734{
3735 this_cpu_inc(*dev->pcpu_refcnt);
3736}
3737
3738/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3739 * and _off may be called from IRQ context, but it is caller
3740 * who is responsible for serialization of these calls.
3741 *
3742 * The name carrier is inappropriate, these functions should really be
3743 * called netif_lowerlayer_*() because they represent the state of any
3744 * kind of lower layer not just hardware media.
3745 */
3746
3747void linkwatch_init_dev(struct net_device *dev);
3748void linkwatch_fire_event(struct net_device *dev);
3749void linkwatch_forget_dev(struct net_device *dev);
3750
3751/**
3752 * netif_carrier_ok - test if carrier present
3753 * @dev: network device
3754 *
3755 * Check if carrier is present on device
3756 */
3757static inline bool netif_carrier_ok(const struct net_device *dev)
3758{
3759 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3760}
3761
3762unsigned long dev_trans_start(struct net_device *dev);
3763
3764void __netdev_watchdog_up(struct net_device *dev);
3765
3766void netif_carrier_on(struct net_device *dev);
3767
3768void netif_carrier_off(struct net_device *dev);
3769
3770/**
3771 * netif_dormant_on - mark device as dormant.
3772 * @dev: network device
3773 *
3774 * Mark device as dormant (as per RFC2863).
3775 *
3776 * The dormant state indicates that the relevant interface is not
3777 * actually in a condition to pass packets (i.e., it is not 'up') but is
3778 * in a "pending" state, waiting for some external event. For "on-
3779 * demand" interfaces, this new state identifies the situation where the
3780 * interface is waiting for events to place it in the up state.
3781 */
3782static inline void netif_dormant_on(struct net_device *dev)
3783{
3784 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3785 linkwatch_fire_event(dev);
3786}
3787
3788/**
3789 * netif_dormant_off - set device as not dormant.
3790 * @dev: network device
3791 *
3792 * Device is not in dormant state.
3793 */
3794static inline void netif_dormant_off(struct net_device *dev)
3795{
3796 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3797 linkwatch_fire_event(dev);
3798}
3799
3800/**
3801 * netif_dormant - test if device is dormant
3802 * @dev: network device
3803 *
3804 * Check if device is dormant.
3805 */
3806static inline bool netif_dormant(const struct net_device *dev)
3807{
3808 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3809}
3810
3811
3812/**
3813 * netif_oper_up - test if device is operational
3814 * @dev: network device
3815 *
3816 * Check if carrier is operational
3817 */
3818static inline bool netif_oper_up(const struct net_device *dev)
3819{
3820 return (dev->operstate == IF_OPER_UP ||
3821 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3822}
3823
3824/**
3825 * netif_device_present - is device available or removed
3826 * @dev: network device
3827 *
3828 * Check if device has not been removed from system.
3829 */
3830static inline bool netif_device_present(struct net_device *dev)
3831{
3832 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3833}
3834
3835void netif_device_detach(struct net_device *dev);
3836
3837void netif_device_attach(struct net_device *dev);
3838
3839/*
3840 * Network interface message level settings
3841 */
3842
3843enum {
3844 NETIF_MSG_DRV = 0x0001,
3845 NETIF_MSG_PROBE = 0x0002,
3846 NETIF_MSG_LINK = 0x0004,
3847 NETIF_MSG_TIMER = 0x0008,
3848 NETIF_MSG_IFDOWN = 0x0010,
3849 NETIF_MSG_IFUP = 0x0020,
3850 NETIF_MSG_RX_ERR = 0x0040,
3851 NETIF_MSG_TX_ERR = 0x0080,
3852 NETIF_MSG_TX_QUEUED = 0x0100,
3853 NETIF_MSG_INTR = 0x0200,
3854 NETIF_MSG_TX_DONE = 0x0400,
3855 NETIF_MSG_RX_STATUS = 0x0800,
3856 NETIF_MSG_PKTDATA = 0x1000,
3857 NETIF_MSG_HW = 0x2000,
3858 NETIF_MSG_WOL = 0x4000,
3859};
3860
3861#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3862#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3863#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3864#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3865#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3866#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3867#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3868#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3869#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3870#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3871#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3872#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3873#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3874#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3875#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3876
3877static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3878{
3879 /* use default */
3880 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3881 return default_msg_enable_bits;
3882 if (debug_value == 0) /* no output */
3883 return 0;
3884 /* set low N bits */
3885 return (1U << debug_value) - 1;
3886}
3887
3888static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3889{
3890 spin_lock(&txq->_xmit_lock);
3891 txq->xmit_lock_owner = cpu;
3892}
3893
3894static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3895{
3896 __acquire(&txq->_xmit_lock);
3897 return true;
3898}
3899
3900static inline void __netif_tx_release(struct netdev_queue *txq)
3901{
3902 __release(&txq->_xmit_lock);
3903}
3904
3905static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3906{
3907 spin_lock_bh(&txq->_xmit_lock);
3908 txq->xmit_lock_owner = smp_processor_id();
3909}
3910
3911static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3912{
3913 bool ok = spin_trylock(&txq->_xmit_lock);
3914 if (likely(ok))
3915 txq->xmit_lock_owner = smp_processor_id();
3916 return ok;
3917}
3918
3919static inline void __netif_tx_unlock(struct netdev_queue *txq)
3920{
3921 txq->xmit_lock_owner = -1;
3922 spin_unlock(&txq->_xmit_lock);
3923}
3924
3925static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3926{
3927 txq->xmit_lock_owner = -1;
3928 spin_unlock_bh(&txq->_xmit_lock);
3929}
3930
3931static inline void txq_trans_update(struct netdev_queue *txq)
3932{
3933 if (txq->xmit_lock_owner != -1)
3934 txq->trans_start = jiffies;
3935}
3936
3937/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3938static inline void netif_trans_update(struct net_device *dev)
3939{
3940 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3941
3942 if (txq->trans_start != jiffies)
3943 txq->trans_start = jiffies;
3944}
3945
3946/**
3947 * netif_tx_lock - grab network device transmit lock
3948 * @dev: network device
3949 *
3950 * Get network device transmit lock
3951 */
3952static inline void netif_tx_lock(struct net_device *dev)
3953{
3954 unsigned int i;
3955 int cpu;
3956
3957 spin_lock(&dev->tx_global_lock);
3958 cpu = smp_processor_id();
3959 for (i = 0; i < dev->num_tx_queues; i++) {
3960 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3961
3962 /* We are the only thread of execution doing a
3963 * freeze, but we have to grab the _xmit_lock in
3964 * order to synchronize with threads which are in
3965 * the ->hard_start_xmit() handler and already
3966 * checked the frozen bit.
3967 */
3968 __netif_tx_lock(txq, cpu);
3969 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3970 __netif_tx_unlock(txq);
3971 }
3972}
3973
3974static inline void netif_tx_lock_bh(struct net_device *dev)
3975{
3976 local_bh_disable();
3977 netif_tx_lock(dev);
3978}
3979
3980static inline void netif_tx_unlock(struct net_device *dev)
3981{
3982 unsigned int i;
3983
3984 for (i = 0; i < dev->num_tx_queues; i++) {
3985 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3986
3987 /* No need to grab the _xmit_lock here. If the
3988 * queue is not stopped for another reason, we
3989 * force a schedule.
3990 */
3991 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3992 netif_schedule_queue(txq);
3993 }
3994 spin_unlock(&dev->tx_global_lock);
3995}
3996
3997static inline void netif_tx_unlock_bh(struct net_device *dev)
3998{
3999 netif_tx_unlock(dev);
4000 local_bh_enable();
4001}
4002
4003#define HARD_TX_LOCK(dev, txq, cpu) { \
4004 if ((dev->features & NETIF_F_LLTX) == 0) { \
4005 __netif_tx_lock(txq, cpu); \
4006 } else { \
4007 __netif_tx_acquire(txq); \
4008 } \
4009}
4010
4011#define HARD_TX_TRYLOCK(dev, txq) \
4012 (((dev->features & NETIF_F_LLTX) == 0) ? \
4013 __netif_tx_trylock(txq) : \
4014 __netif_tx_acquire(txq))
4015
4016#define HARD_TX_UNLOCK(dev, txq) { \
4017 if ((dev->features & NETIF_F_LLTX) == 0) { \
4018 __netif_tx_unlock(txq); \
4019 } else { \
4020 __netif_tx_release(txq); \
4021 } \
4022}
4023
4024static inline void netif_tx_disable(struct net_device *dev)
4025{
4026 unsigned int i;
4027 int cpu;
4028
4029 local_bh_disable();
4030 cpu = smp_processor_id();
4031 for (i = 0; i < dev->num_tx_queues; i++) {
4032 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4033
4034 __netif_tx_lock(txq, cpu);
4035 netif_tx_stop_queue(txq);
4036 __netif_tx_unlock(txq);
4037 }
4038 local_bh_enable();
4039}
4040
4041static inline void netif_addr_lock(struct net_device *dev)
4042{
4043 spin_lock(&dev->addr_list_lock);
4044}
4045
4046static inline void netif_addr_lock_nested(struct net_device *dev)
4047{
4048 int subclass = SINGLE_DEPTH_NESTING;
4049
4050 if (dev->netdev_ops->ndo_get_lock_subclass)
4051 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4052
4053 spin_lock_nested(&dev->addr_list_lock, subclass);
4054}
4055
4056static inline void netif_addr_lock_bh(struct net_device *dev)
4057{
4058 spin_lock_bh(&dev->addr_list_lock);
4059}
4060
4061static inline void netif_addr_unlock(struct net_device *dev)
4062{
4063 spin_unlock(&dev->addr_list_lock);
4064}
4065
4066static inline void netif_addr_unlock_bh(struct net_device *dev)
4067{
4068 spin_unlock_bh(&dev->addr_list_lock);
4069}
4070
4071/*
4072 * dev_addrs walker. Should be used only for read access. Call with
4073 * rcu_read_lock held.
4074 */
4075#define for_each_dev_addr(dev, ha) \
4076 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4077
4078/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4079
4080void ether_setup(struct net_device *dev);
4081
4082/* Support for loadable net-drivers */
4083struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4084 unsigned char name_assign_type,
4085 void (*setup)(struct net_device *),
4086 unsigned int txqs, unsigned int rxqs);
4087int dev_get_valid_name(struct net *net, struct net_device *dev,
4088 const char *name);
4089
4090#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4091 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4092
4093#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4094 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4095 count)
4096
4097int register_netdev(struct net_device *dev);
4098void unregister_netdev(struct net_device *dev);
4099
4100/* General hardware address lists handling functions */
4101int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4102 struct netdev_hw_addr_list *from_list, int addr_len);
4103void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4104 struct netdev_hw_addr_list *from_list, int addr_len);
4105int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4106 struct net_device *dev,
4107 int (*sync)(struct net_device *, const unsigned char *),
4108 int (*unsync)(struct net_device *,
4109 const unsigned char *));
4110int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4111 struct net_device *dev,
4112 int (*sync)(struct net_device *,
4113 const unsigned char *, int),
4114 int (*unsync)(struct net_device *,
4115 const unsigned char *, int));
4116void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4117 struct net_device *dev,
4118 int (*unsync)(struct net_device *,
4119 const unsigned char *, int));
4120void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4121 struct net_device *dev,
4122 int (*unsync)(struct net_device *,
4123 const unsigned char *));
4124void __hw_addr_init(struct netdev_hw_addr_list *list);
4125
4126/* Functions used for device addresses handling */
4127int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4128 unsigned char addr_type);
4129int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4130 unsigned char addr_type);
4131void dev_addr_flush(struct net_device *dev);
4132int dev_addr_init(struct net_device *dev);
4133
4134/* Functions used for unicast addresses handling */
4135int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4136int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4137int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4138int dev_uc_sync(struct net_device *to, struct net_device *from);
4139int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4140void dev_uc_unsync(struct net_device *to, struct net_device *from);
4141void dev_uc_flush(struct net_device *dev);
4142void dev_uc_init(struct net_device *dev);
4143
4144/**
4145 * __dev_uc_sync - Synchonize device's unicast list
4146 * @dev: device to sync
4147 * @sync: function to call if address should be added
4148 * @unsync: function to call if address should be removed
4149 *
4150 * Add newly added addresses to the interface, and release
4151 * addresses that have been deleted.
4152 */
4153static inline int __dev_uc_sync(struct net_device *dev,
4154 int (*sync)(struct net_device *,
4155 const unsigned char *),
4156 int (*unsync)(struct net_device *,
4157 const unsigned char *))
4158{
4159 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4160}
4161
4162/**
4163 * __dev_uc_unsync - Remove synchronized addresses from device
4164 * @dev: device to sync
4165 * @unsync: function to call if address should be removed
4166 *
4167 * Remove all addresses that were added to the device by dev_uc_sync().
4168 */
4169static inline void __dev_uc_unsync(struct net_device *dev,
4170 int (*unsync)(struct net_device *,
4171 const unsigned char *))
4172{
4173 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4174}
4175
4176/* Functions used for multicast addresses handling */
4177int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4178int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4179int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4180int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4181int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4182int dev_mc_sync(struct net_device *to, struct net_device *from);
4183int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4184void dev_mc_unsync(struct net_device *to, struct net_device *from);
4185void dev_mc_flush(struct net_device *dev);
4186void dev_mc_init(struct net_device *dev);
4187
4188/**
4189 * __dev_mc_sync - Synchonize device's multicast list
4190 * @dev: device to sync
4191 * @sync: function to call if address should be added
4192 * @unsync: function to call if address should be removed
4193 *
4194 * Add newly added addresses to the interface, and release
4195 * addresses that have been deleted.
4196 */
4197static inline int __dev_mc_sync(struct net_device *dev,
4198 int (*sync)(struct net_device *,
4199 const unsigned char *),
4200 int (*unsync)(struct net_device *,
4201 const unsigned char *))
4202{
4203 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4204}
4205
4206/**
4207 * __dev_mc_unsync - Remove synchronized addresses from device
4208 * @dev: device to sync
4209 * @unsync: function to call if address should be removed
4210 *
4211 * Remove all addresses that were added to the device by dev_mc_sync().
4212 */
4213static inline void __dev_mc_unsync(struct net_device *dev,
4214 int (*unsync)(struct net_device *,
4215 const unsigned char *))
4216{
4217 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4218}
4219
4220/* Functions used for secondary unicast and multicast support */
4221void dev_set_rx_mode(struct net_device *dev);
4222void __dev_set_rx_mode(struct net_device *dev);
4223int dev_set_promiscuity(struct net_device *dev, int inc);
4224int dev_set_allmulti(struct net_device *dev, int inc);
4225void netdev_state_change(struct net_device *dev);
4226void netdev_notify_peers(struct net_device *dev);
4227void netdev_features_change(struct net_device *dev);
4228/* Load a device via the kmod */
4229void dev_load(struct net *net, const char *name);
4230struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4231 struct rtnl_link_stats64 *storage);
4232void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4233 const struct net_device_stats *netdev_stats);
4234
4235extern int netdev_max_backlog;
4236extern int netdev_tstamp_prequeue;
4237extern int weight_p;
4238extern int dev_weight_rx_bias;
4239extern int dev_weight_tx_bias;
4240extern int dev_rx_weight;
4241extern int dev_tx_weight;
4242
4243bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4244struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4245 struct list_head **iter);
4246struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4247 struct list_head **iter);
4248
4249/* iterate through upper list, must be called under RCU read lock */
4250#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4251 for (iter = &(dev)->adj_list.upper, \
4252 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4253 updev; \
4254 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4255
4256int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4257 int (*fn)(struct net_device *upper_dev,
4258 void *data),
4259 void *data);
4260
4261bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4262 struct net_device *upper_dev);
4263
4264bool netdev_has_any_upper_dev(struct net_device *dev);
4265
4266void *netdev_lower_get_next_private(struct net_device *dev,
4267 struct list_head **iter);
4268void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4269 struct list_head **iter);
4270
4271#define netdev_for_each_lower_private(dev, priv, iter) \
4272 for (iter = (dev)->adj_list.lower.next, \
4273 priv = netdev_lower_get_next_private(dev, &(iter)); \
4274 priv; \
4275 priv = netdev_lower_get_next_private(dev, &(iter)))
4276
4277#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4278 for (iter = &(dev)->adj_list.lower, \
4279 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4280 priv; \
4281 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4282
4283void *netdev_lower_get_next(struct net_device *dev,
4284 struct list_head **iter);
4285
4286#define netdev_for_each_lower_dev(dev, ldev, iter) \
4287 for (iter = (dev)->adj_list.lower.next, \
4288 ldev = netdev_lower_get_next(dev, &(iter)); \
4289 ldev; \
4290 ldev = netdev_lower_get_next(dev, &(iter)))
4291
4292struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4293 struct list_head **iter);
4294struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4295 struct list_head **iter);
4296
4297int netdev_walk_all_lower_dev(struct net_device *dev,
4298 int (*fn)(struct net_device *lower_dev,
4299 void *data),
4300 void *data);
4301int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4302 int (*fn)(struct net_device *lower_dev,
4303 void *data),
4304 void *data);
4305
4306void *netdev_adjacent_get_private(struct list_head *adj_list);
4307void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4308struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4309struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4310int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4311 struct netlink_ext_ack *extack);
4312int netdev_master_upper_dev_link(struct net_device *dev,
4313 struct net_device *upper_dev,
4314 void *upper_priv, void *upper_info,
4315 struct netlink_ext_ack *extack);
4316void netdev_upper_dev_unlink(struct net_device *dev,
4317 struct net_device *upper_dev);
4318void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4319void *netdev_lower_dev_get_private(struct net_device *dev,
4320 struct net_device *lower_dev);
4321void netdev_lower_state_changed(struct net_device *lower_dev,
4322 void *lower_state_info);
4323
4324/* RSS keys are 40 or 52 bytes long */
4325#define NETDEV_RSS_KEY_LEN 52
4326extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4327void netdev_rss_key_fill(void *buffer, size_t len);
4328
4329int dev_get_nest_level(struct net_device *dev);
4330int skb_checksum_help(struct sk_buff *skb);
4331int skb_crc32c_csum_help(struct sk_buff *skb);
4332int skb_csum_hwoffload_help(struct sk_buff *skb,
4333 const netdev_features_t features);
4334
4335struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4336 netdev_features_t features, bool tx_path);
4337struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4338 netdev_features_t features);
4339
4340struct netdev_bonding_info {
4341 ifslave slave;
4342 ifbond master;
4343};
4344
4345struct netdev_notifier_bonding_info {
4346 struct netdev_notifier_info info; /* must be first */
4347 struct netdev_bonding_info bonding_info;
4348};
4349
4350void netdev_bonding_info_change(struct net_device *dev,
4351 struct netdev_bonding_info *bonding_info);
4352
4353static inline
4354struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4355{
4356 return __skb_gso_segment(skb, features, true);
4357}
4358__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4359
4360static inline bool can_checksum_protocol(netdev_features_t features,
4361 __be16 protocol)
4362{
4363 if (protocol == htons(ETH_P_FCOE))
4364 return !!(features & NETIF_F_FCOE_CRC);
4365
4366 /* Assume this is an IP checksum (not SCTP CRC) */
4367
4368 if (features & NETIF_F_HW_CSUM) {
4369 /* Can checksum everything */
4370 return true;
4371 }
4372
4373 switch (protocol) {
4374 case htons(ETH_P_IP):
4375 return !!(features & NETIF_F_IP_CSUM);
4376 case htons(ETH_P_IPV6):
4377 return !!(features & NETIF_F_IPV6_CSUM);
4378 default:
4379 return false;
4380 }
4381}
4382
4383#ifdef CONFIG_BUG
4384void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4385#else
4386static inline void netdev_rx_csum_fault(struct net_device *dev,
4387 struct sk_buff *skb)
4388{
4389}
4390#endif
4391/* rx skb timestamps */
4392void net_enable_timestamp(void);
4393void net_disable_timestamp(void);
4394
4395#ifdef CONFIG_PROC_FS
4396int __init dev_proc_init(void);
4397#else
4398#define dev_proc_init() 0
4399#endif
4400
4401static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4402 struct sk_buff *skb, struct net_device *dev,
4403 bool more)
4404{
4405 __this_cpu_write(softnet_data.xmit.more, more);
4406 return ops->ndo_start_xmit(skb, dev);
4407}
4408
4409static inline bool netdev_xmit_more(void)
4410{
4411 return __this_cpu_read(softnet_data.xmit.more);
4412}
4413
4414static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4415 struct netdev_queue *txq, bool more)
4416{
4417 const struct net_device_ops *ops = dev->netdev_ops;
4418 netdev_tx_t rc;
4419
4420 rc = __netdev_start_xmit(ops, skb, dev, more);
4421 if (rc == NETDEV_TX_OK)
4422 txq_trans_update(txq);
4423
4424 return rc;
4425}
4426
4427int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4428 const void *ns);
4429void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4430 const void *ns);
4431
4432static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4433{
4434 return netdev_class_create_file_ns(class_attr, NULL);
4435}
4436
4437static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4438{
4439 netdev_class_remove_file_ns(class_attr, NULL);
4440}
4441
4442extern const struct kobj_ns_type_operations net_ns_type_operations;
4443
4444const char *netdev_drivername(const struct net_device *dev);
4445
4446void linkwatch_run_queue(void);
4447
4448static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4449 netdev_features_t f2)
4450{
4451 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4452 if (f1 & NETIF_F_HW_CSUM)
4453 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4454 else
4455 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4456 }
4457
4458 return f1 & f2;
4459}
4460
4461static inline netdev_features_t netdev_get_wanted_features(
4462 struct net_device *dev)
4463{
4464 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4465}
4466netdev_features_t netdev_increment_features(netdev_features_t all,
4467 netdev_features_t one, netdev_features_t mask);
4468
4469/* Allow TSO being used on stacked device :
4470 * Performing the GSO segmentation before last device
4471 * is a performance improvement.
4472 */
4473static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4474 netdev_features_t mask)
4475{
4476 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4477}
4478
4479int __netdev_update_features(struct net_device *dev);
4480void netdev_update_features(struct net_device *dev);
4481void netdev_change_features(struct net_device *dev);
4482
4483void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4484 struct net_device *dev);
4485
4486netdev_features_t passthru_features_check(struct sk_buff *skb,
4487 struct net_device *dev,
4488 netdev_features_t features);
4489netdev_features_t netif_skb_features(struct sk_buff *skb);
4490
4491static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4492{
4493 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4494
4495 /* check flags correspondence */
4496 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4497 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4498 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4499 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4500 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4501 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4502 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4503 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4504 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4505 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4506 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4507 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4508 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4509 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4510 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4511 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4512 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4513 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4514
4515 return (features & feature) == feature;
4516}
4517
4518static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4519{
4520 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4521 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4522}
4523
4524static inline bool netif_needs_gso(struct sk_buff *skb,
4525 netdev_features_t features)
4526{
4527 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4528 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4529 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4530}
4531
4532static inline void netif_set_gso_max_size(struct net_device *dev,
4533 unsigned int size)
4534{
4535 dev->gso_max_size = size;
4536}
4537
4538static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4539 int pulled_hlen, u16 mac_offset,
4540 int mac_len)
4541{
4542 skb->protocol = protocol;
4543 skb->encapsulation = 1;
4544 skb_push(skb, pulled_hlen);
4545 skb_reset_transport_header(skb);
4546 skb->mac_header = mac_offset;
4547 skb->network_header = skb->mac_header + mac_len;
4548 skb->mac_len = mac_len;
4549}
4550
4551static inline bool netif_is_macsec(const struct net_device *dev)
4552{
4553 return dev->priv_flags & IFF_MACSEC;
4554}
4555
4556static inline bool netif_is_macvlan(const struct net_device *dev)
4557{
4558 return dev->priv_flags & IFF_MACVLAN;
4559}
4560
4561static inline bool netif_is_macvlan_port(const struct net_device *dev)
4562{
4563 return dev->priv_flags & IFF_MACVLAN_PORT;
4564}
4565
4566static inline bool netif_is_bond_master(const struct net_device *dev)
4567{
4568 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4569}
4570
4571static inline bool netif_is_bond_slave(const struct net_device *dev)
4572{
4573 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4574}
4575
4576static inline bool netif_supports_nofcs(struct net_device *dev)
4577{
4578 return dev->priv_flags & IFF_SUPP_NOFCS;
4579}
4580
4581static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4582{
4583 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4584}
4585
4586static inline bool netif_is_l3_master(const struct net_device *dev)
4587{
4588 return dev->priv_flags & IFF_L3MDEV_MASTER;
4589}
4590
4591static inline bool netif_is_l3_slave(const struct net_device *dev)
4592{
4593 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4594}
4595
4596static inline bool netif_is_bridge_master(const struct net_device *dev)
4597{
4598 return dev->priv_flags & IFF_EBRIDGE;
4599}
4600
4601static inline bool netif_is_bridge_port(const struct net_device *dev)
4602{
4603 return dev->priv_flags & IFF_BRIDGE_PORT;
4604}
4605
4606static inline bool netif_is_ovs_master(const struct net_device *dev)
4607{
4608 return dev->priv_flags & IFF_OPENVSWITCH;
4609}
4610
4611static inline bool netif_is_ovs_port(const struct net_device *dev)
4612{
4613 return dev->priv_flags & IFF_OVS_DATAPATH;
4614}
4615
4616static inline bool netif_is_team_master(const struct net_device *dev)
4617{
4618 return dev->priv_flags & IFF_TEAM;
4619}
4620
4621static inline bool netif_is_team_port(const struct net_device *dev)
4622{
4623 return dev->priv_flags & IFF_TEAM_PORT;
4624}
4625
4626static inline bool netif_is_lag_master(const struct net_device *dev)
4627{
4628 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4629}
4630
4631static inline bool netif_is_lag_port(const struct net_device *dev)
4632{
4633 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4634}
4635
4636static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4637{
4638 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4639}
4640
4641static inline bool netif_is_failover(const struct net_device *dev)
4642{
4643 return dev->priv_flags & IFF_FAILOVER;
4644}
4645
4646static inline bool netif_is_failover_slave(const struct net_device *dev)
4647{
4648 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4649}
4650
4651/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4652static inline void netif_keep_dst(struct net_device *dev)
4653{
4654 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4655}
4656
4657/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4658static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4659{
4660 /* TODO: reserve and use an additional IFF bit, if we get more users */
4661 return dev->priv_flags & IFF_MACSEC;
4662}
4663
4664extern struct pernet_operations __net_initdata loopback_net_ops;
4665
4666/* Logging, debugging and troubleshooting/diagnostic helpers. */
4667
4668/* netdev_printk helpers, similar to dev_printk */
4669
4670static inline const char *netdev_name(const struct net_device *dev)
4671{
4672 if (!dev->name[0] || strchr(dev->name, '%'))
4673 return "(unnamed net_device)";
4674 return dev->name;
4675}
4676
4677static inline bool netdev_unregistering(const struct net_device *dev)
4678{
4679 return dev->reg_state == NETREG_UNREGISTERING;
4680}
4681
4682static inline const char *netdev_reg_state(const struct net_device *dev)
4683{
4684 switch (dev->reg_state) {
4685 case NETREG_UNINITIALIZED: return " (uninitialized)";
4686 case NETREG_REGISTERED: return "";
4687 case NETREG_UNREGISTERING: return " (unregistering)";
4688 case NETREG_UNREGISTERED: return " (unregistered)";
4689 case NETREG_RELEASED: return " (released)";
4690 case NETREG_DUMMY: return " (dummy)";
4691 }
4692
4693 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4694 return " (unknown)";
4695}
4696
4697__printf(3, 4) __cold
4698void netdev_printk(const char *level, const struct net_device *dev,
4699 const char *format, ...);
4700__printf(2, 3) __cold
4701void netdev_emerg(const struct net_device *dev, const char *format, ...);
4702__printf(2, 3) __cold
4703void netdev_alert(const struct net_device *dev, const char *format, ...);
4704__printf(2, 3) __cold
4705void netdev_crit(const struct net_device *dev, const char *format, ...);
4706__printf(2, 3) __cold
4707void netdev_err(const struct net_device *dev, const char *format, ...);
4708__printf(2, 3) __cold
4709void netdev_warn(const struct net_device *dev, const char *format, ...);
4710__printf(2, 3) __cold
4711void netdev_notice(const struct net_device *dev, const char *format, ...);
4712__printf(2, 3) __cold
4713void netdev_info(const struct net_device *dev, const char *format, ...);
4714
4715#define netdev_level_once(level, dev, fmt, ...) \
4716do { \
4717 static bool __print_once __read_mostly; \
4718 \
4719 if (!__print_once) { \
4720 __print_once = true; \
4721 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4722 } \
4723} while (0)
4724
4725#define netdev_emerg_once(dev, fmt, ...) \
4726 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4727#define netdev_alert_once(dev, fmt, ...) \
4728 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4729#define netdev_crit_once(dev, fmt, ...) \
4730 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4731#define netdev_err_once(dev, fmt, ...) \
4732 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4733#define netdev_warn_once(dev, fmt, ...) \
4734 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4735#define netdev_notice_once(dev, fmt, ...) \
4736 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4737#define netdev_info_once(dev, fmt, ...) \
4738 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4739
4740#define MODULE_ALIAS_NETDEV(device) \
4741 MODULE_ALIAS("netdev-" device)
4742
4743#if defined(CONFIG_DYNAMIC_DEBUG)
4744#define netdev_dbg(__dev, format, args...) \
4745do { \
4746 dynamic_netdev_dbg(__dev, format, ##args); \
4747} while (0)
4748#elif defined(DEBUG)
4749#define netdev_dbg(__dev, format, args...) \
4750 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4751#else
4752#define netdev_dbg(__dev, format, args...) \
4753({ \
4754 if (0) \
4755 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4756})
4757#endif
4758
4759#if defined(VERBOSE_DEBUG)
4760#define netdev_vdbg netdev_dbg
4761#else
4762
4763#define netdev_vdbg(dev, format, args...) \
4764({ \
4765 if (0) \
4766 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4767 0; \
4768})
4769#endif
4770
4771/*
4772 * netdev_WARN() acts like dev_printk(), but with the key difference
4773 * of using a WARN/WARN_ON to get the message out, including the
4774 * file/line information and a backtrace.
4775 */
4776#define netdev_WARN(dev, format, args...) \
4777 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4778 netdev_reg_state(dev), ##args)
4779
4780#define netdev_WARN_ONCE(dev, format, args...) \
4781 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4782 netdev_reg_state(dev), ##args)
4783
4784/* netif printk helpers, similar to netdev_printk */
4785
4786#define netif_printk(priv, type, level, dev, fmt, args...) \
4787do { \
4788 if (netif_msg_##type(priv)) \
4789 netdev_printk(level, (dev), fmt, ##args); \
4790} while (0)
4791
4792#define netif_level(level, priv, type, dev, fmt, args...) \
4793do { \
4794 if (netif_msg_##type(priv)) \
4795 netdev_##level(dev, fmt, ##args); \
4796} while (0)
4797
4798#define netif_emerg(priv, type, dev, fmt, args...) \
4799 netif_level(emerg, priv, type, dev, fmt, ##args)
4800#define netif_alert(priv, type, dev, fmt, args...) \
4801 netif_level(alert, priv, type, dev, fmt, ##args)
4802#define netif_crit(priv, type, dev, fmt, args...) \
4803 netif_level(crit, priv, type, dev, fmt, ##args)
4804#define netif_err(priv, type, dev, fmt, args...) \
4805 netif_level(err, priv, type, dev, fmt, ##args)
4806#define netif_warn(priv, type, dev, fmt, args...) \
4807 netif_level(warn, priv, type, dev, fmt, ##args)
4808#define netif_notice(priv, type, dev, fmt, args...) \
4809 netif_level(notice, priv, type, dev, fmt, ##args)
4810#define netif_info(priv, type, dev, fmt, args...) \
4811 netif_level(info, priv, type, dev, fmt, ##args)
4812
4813#if defined(CONFIG_DYNAMIC_DEBUG)
4814#define netif_dbg(priv, type, netdev, format, args...) \
4815do { \
4816 if (netif_msg_##type(priv)) \
4817 dynamic_netdev_dbg(netdev, format, ##args); \
4818} while (0)
4819#elif defined(DEBUG)
4820#define netif_dbg(priv, type, dev, format, args...) \
4821 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4822#else
4823#define netif_dbg(priv, type, dev, format, args...) \
4824({ \
4825 if (0) \
4826 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4827 0; \
4828})
4829#endif
4830
4831/* if @cond then downgrade to debug, else print at @level */
4832#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4833 do { \
4834 if (cond) \
4835 netif_dbg(priv, type, netdev, fmt, ##args); \
4836 else \
4837 netif_ ## level(priv, type, netdev, fmt, ##args); \
4838 } while (0)
4839
4840#if defined(VERBOSE_DEBUG)
4841#define netif_vdbg netif_dbg
4842#else
4843#define netif_vdbg(priv, type, dev, format, args...) \
4844({ \
4845 if (0) \
4846 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4847 0; \
4848})
4849#endif
4850
4851/*
4852 * The list of packet types we will receive (as opposed to discard)
4853 * and the routines to invoke.
4854 *
4855 * Why 16. Because with 16 the only overlap we get on a hash of the
4856 * low nibble of the protocol value is RARP/SNAP/X.25.
4857 *
4858 * 0800 IP
4859 * 0001 802.3
4860 * 0002 AX.25
4861 * 0004 802.2
4862 * 8035 RARP
4863 * 0005 SNAP
4864 * 0805 X.25
4865 * 0806 ARP
4866 * 8137 IPX
4867 * 0009 Localtalk
4868 * 86DD IPv6
4869 */
4870#define PTYPE_HASH_SIZE (16)
4871#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4872
4873#endif /* _LINUX_NETDEVICE_H */