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 list_head rx_list; /* Pending GRO_NORMAL skbs */
336 int rx_count; /* length of rx_list */
337 struct hrtimer timer;
338 struct list_head dev_list;
339 struct hlist_node napi_hash_node;
340 unsigned int napi_id;
341};
342
343enum {
344 NAPI_STATE_SCHED, /* Poll is scheduled */
345 NAPI_STATE_MISSED, /* reschedule a napi */
346 NAPI_STATE_DISABLE, /* Disable pending */
347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
351};
352
353enum {
354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
361};
362
363enum gro_result {
364 GRO_MERGED,
365 GRO_MERGED_FREE,
366 GRO_HELD,
367 GRO_NORMAL,
368 GRO_DROP,
369 GRO_CONSUMED,
370};
371typedef enum gro_result gro_result_t;
372
373/*
374 * enum rx_handler_result - Possible return values for rx_handlers.
375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
376 * further.
377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
378 * case skb->dev was changed by rx_handler.
379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
381 *
382 * rx_handlers are functions called from inside __netif_receive_skb(), to do
383 * special processing of the skb, prior to delivery to protocol handlers.
384 *
385 * Currently, a net_device can only have a single rx_handler registered. Trying
386 * to register a second rx_handler will return -EBUSY.
387 *
388 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
389 * To unregister a rx_handler on a net_device, use
390 * netdev_rx_handler_unregister().
391 *
392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
393 * do with the skb.
394 *
395 * If the rx_handler consumed the skb in some way, it should return
396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
397 * the skb to be delivered in some other way.
398 *
399 * If the rx_handler changed skb->dev, to divert the skb to another
400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
401 * new device will be called if it exists.
402 *
403 * If the rx_handler decides the skb should be ignored, it should return
404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
405 * are registered on exact device (ptype->dev == skb->dev).
406 *
407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
408 * delivered, it should return RX_HANDLER_PASS.
409 *
410 * A device without a registered rx_handler will behave as if rx_handler
411 * returned RX_HANDLER_PASS.
412 */
413
414enum rx_handler_result {
415 RX_HANDLER_CONSUMED,
416 RX_HANDLER_ANOTHER,
417 RX_HANDLER_EXACT,
418 RX_HANDLER_PASS,
419};
420typedef enum rx_handler_result rx_handler_result_t;
421typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
422
423void __napi_schedule(struct napi_struct *n);
424void __napi_schedule_irqoff(struct napi_struct *n);
425
426static inline bool napi_disable_pending(struct napi_struct *n)
427{
428 return test_bit(NAPI_STATE_DISABLE, &n->state);
429}
430
431bool napi_schedule_prep(struct napi_struct *n);
432
433/**
434 * napi_schedule - schedule NAPI poll
435 * @n: NAPI context
436 *
437 * Schedule NAPI poll routine to be called if it is not already
438 * running.
439 */
440static inline void napi_schedule(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule(n);
444}
445
446/**
447 * napi_schedule_irqoff - schedule NAPI poll
448 * @n: NAPI context
449 *
450 * Variant of napi_schedule(), assuming hard irqs are masked.
451 */
452static inline void napi_schedule_irqoff(struct napi_struct *n)
453{
454 if (napi_schedule_prep(n))
455 __napi_schedule_irqoff(n);
456}
457
458/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
459static inline bool napi_reschedule(struct napi_struct *napi)
460{
461 if (napi_schedule_prep(napi)) {
462 __napi_schedule(napi);
463 return true;
464 }
465 return false;
466}
467
468bool napi_complete_done(struct napi_struct *n, int work_done);
469/**
470 * napi_complete - NAPI processing complete
471 * @n: NAPI context
472 *
473 * Mark NAPI processing as complete.
474 * Consider using napi_complete_done() instead.
475 * Return false if device should avoid rearming interrupts.
476 */
477static inline bool napi_complete(struct napi_struct *n)
478{
479 return napi_complete_done(n, 0);
480}
481
482/**
483 * napi_hash_del - remove a NAPI from global table
484 * @napi: NAPI context
485 *
486 * Warning: caller must observe RCU grace period
487 * before freeing memory containing @napi, if
488 * this function returns true.
489 * Note: core networking stack automatically calls it
490 * from netif_napi_del().
491 * Drivers might want to call this helper to combine all
492 * the needed RCU grace periods into a single one.
493 */
494bool napi_hash_del(struct napi_struct *napi);
495
496/**
497 * napi_disable - prevent NAPI from scheduling
498 * @n: NAPI context
499 *
500 * Stop NAPI from being scheduled on this context.
501 * Waits till any outstanding processing completes.
502 */
503void napi_disable(struct napi_struct *n);
504
505/**
506 * napi_enable - enable NAPI scheduling
507 * @n: NAPI context
508 *
509 * Resume NAPI from being scheduled on this context.
510 * Must be paired with napi_disable.
511 */
512static inline void napi_enable(struct napi_struct *n)
513{
514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
515 smp_mb__before_atomic();
516 clear_bit(NAPI_STATE_SCHED, &n->state);
517 clear_bit(NAPI_STATE_NPSVC, &n->state);
518}
519
520/**
521 * napi_synchronize - wait until NAPI is not running
522 * @n: NAPI context
523 *
524 * Wait until NAPI is done being scheduled on this context.
525 * Waits till any outstanding processing completes but
526 * does not disable future activations.
527 */
528static inline void napi_synchronize(const struct napi_struct *n)
529{
530 if (IS_ENABLED(CONFIG_SMP))
531 while (test_bit(NAPI_STATE_SCHED, &n->state))
532 msleep(1);
533 else
534 barrier();
535}
536
537/**
538 * napi_if_scheduled_mark_missed - if napi is running, set the
539 * NAPIF_STATE_MISSED
540 * @n: NAPI context
541 *
542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
543 * NAPI is scheduled.
544 **/
545static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
546{
547 unsigned long val, new;
548
549 do {
550 val = READ_ONCE(n->state);
551 if (val & NAPIF_STATE_DISABLE)
552 return true;
553
554 if (!(val & NAPIF_STATE_SCHED))
555 return false;
556
557 new = val | NAPIF_STATE_MISSED;
558 } while (cmpxchg(&n->state, val, new) != val);
559
560 return true;
561}
562
563enum netdev_queue_state_t {
564 __QUEUE_STATE_DRV_XOFF,
565 __QUEUE_STATE_STACK_XOFF,
566 __QUEUE_STATE_FROZEN,
567};
568
569#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
570#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
571#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
572
573#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
574#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
575 QUEUE_STATE_FROZEN)
576#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
577 QUEUE_STATE_FROZEN)
578
579/*
580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
581 * netif_tx_* functions below are used to manipulate this flag. The
582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
583 * queue independently. The netif_xmit_*stopped functions below are called
584 * to check if the queue has been stopped by the driver or stack (either
585 * of the XOFF bits are set in the state). Drivers should not need to call
586 * netif_xmit*stopped functions, they should only be using netif_tx_*.
587 */
588
589struct netdev_queue {
590/*
591 * read-mostly part
592 */
593 struct net_device *dev;
594 struct Qdisc __rcu *qdisc;
595 struct Qdisc *qdisc_sleeping;
596#ifdef CONFIG_SYSFS
597 struct kobject kobj;
598#endif
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 int numa_node;
601#endif
602 unsigned long tx_maxrate;
603 /*
604 * Number of TX timeouts for this queue
605 * (/sys/class/net/DEV/Q/trans_timeout)
606 */
607 unsigned long trans_timeout;
608
609 /* Subordinate device that the queue has been assigned to */
610 struct net_device *sb_dev;
611#ifdef CONFIG_XDP_SOCKETS
612 struct xdp_umem *umem;
613#endif
614/*
615 * write-mostly part
616 */
617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
618 int xmit_lock_owner;
619 /*
620 * Time (in jiffies) of last Tx
621 */
622 unsigned long trans_start;
623
624 unsigned long state;
625
626#ifdef CONFIG_BQL
627 struct dql dql;
628#endif
629} ____cacheline_aligned_in_smp;
630
631extern int sysctl_fb_tunnels_only_for_init_net;
632extern int sysctl_devconf_inherit_init_net;
633
634static inline bool net_has_fallback_tunnels(const struct net *net)
635{
636 return net == &init_net ||
637 !IS_ENABLED(CONFIG_SYSCTL) ||
638 !sysctl_fb_tunnels_only_for_init_net;
639}
640
641static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
642{
643#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
644 return q->numa_node;
645#else
646 return NUMA_NO_NODE;
647#endif
648}
649
650static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
651{
652#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
653 q->numa_node = node;
654#endif
655}
656
657#ifdef CONFIG_RPS
658/*
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
661 */
662struct rps_map {
663 unsigned int len;
664 struct rcu_head rcu;
665 u16 cpus[0];
666};
667#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668
669/*
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
673 */
674struct rps_dev_flow {
675 u16 cpu;
676 u16 filter;
677 unsigned int last_qtail;
678};
679#define RPS_NO_FILTER 0xffff
680
681/*
682 * The rps_dev_flow_table structure contains a table of flow mappings.
683 */
684struct rps_dev_flow_table {
685 unsigned int mask;
686 struct rcu_head rcu;
687 struct rps_dev_flow flows[0];
688};
689#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
691
692/*
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
701 */
702struct rps_sock_flow_table {
703 u32 mask;
704
705 u32 ents[0] ____cacheline_aligned_in_smp;
706};
707#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
708
709#define RPS_NO_CPU 0xffff
710
711extern u32 rps_cpu_mask;
712extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
713
714static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
715 u32 hash)
716{
717 if (table && hash) {
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
720
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
723
724 if (table->ents[index] != val)
725 table->ents[index] = val;
726 }
727}
728
729#ifdef CONFIG_RFS_ACCEL
730bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 u16 filter_id);
732#endif
733#endif /* CONFIG_RPS */
734
735/* This structure contains an instance of an RX queue. */
736struct netdev_rx_queue {
737#ifdef CONFIG_RPS
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
740#endif
741 struct kobject kobj;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744#ifdef CONFIG_XDP_SOCKETS
745 struct xdp_umem *umem;
746#endif
747} ____cacheline_aligned_in_smp;
748
749/*
750 * RX queue sysfs structures and functions.
751 */
752struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
757};
758
759#ifdef CONFIG_XPS
760/*
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
763 */
764struct xps_map {
765 unsigned int len;
766 unsigned int alloc_len;
767 struct rcu_head rcu;
768 u16 queues[0];
769};
770#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
773
774/*
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
776 */
777struct xps_dev_maps {
778 struct rcu_head rcu;
779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
780};
781
782#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
784
785#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
787
788#endif /* CONFIG_XPS */
789
790#define TC_MAX_QUEUE 16
791#define TC_BITMASK 15
792/* HW offloaded queuing disciplines txq count and offset maps */
793struct netdev_tc_txq {
794 u16 count;
795 u16 offset;
796};
797
798#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
799/*
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
802 */
803struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
810 char model[256];
811 char model_description[256];
812};
813#endif
814
815#define MAX_PHYS_ITEM_ID_LEN 32
816
817/* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
819 */
820struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 unsigned char id_len;
823};
824
825static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
827{
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
830}
831
832typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
833 struct sk_buff *skb,
834 struct net_device *sb_dev);
835
836enum tc_setup_type {
837 TC_SETUP_QDISC_MQPRIO,
838 TC_SETUP_CLSU32,
839 TC_SETUP_CLSFLOWER,
840 TC_SETUP_CLSMATCHALL,
841 TC_SETUP_CLSBPF,
842 TC_SETUP_BLOCK,
843 TC_SETUP_QDISC_CBS,
844 TC_SETUP_QDISC_RED,
845 TC_SETUP_QDISC_PRIO,
846 TC_SETUP_QDISC_MQ,
847 TC_SETUP_QDISC_ETF,
848 TC_SETUP_ROOT_QDISC,
849 TC_SETUP_QDISC_GRED,
850 TC_SETUP_QDISC_TAPRIO,
851 TC_SETUP_FT,
852 TC_SETUP_QDISC_ETS,
853 TC_SETUP_QDISC_TBF,
854};
855
856/* These structures hold the attributes of bpf state that are being passed
857 * to the netdevice through the bpf op.
858 */
859enum bpf_netdev_command {
860 /* Set or clear a bpf program used in the earliest stages of packet
861 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
862 * is responsible for calling bpf_prog_put on any old progs that are
863 * stored. In case of error, the callee need not release the new prog
864 * reference, but on success it takes ownership and must bpf_prog_put
865 * when it is no longer used.
866 */
867 XDP_SETUP_PROG,
868 XDP_SETUP_PROG_HW,
869 XDP_QUERY_PROG,
870 XDP_QUERY_PROG_HW,
871 /* BPF program for offload callbacks, invoked at program load time. */
872 BPF_OFFLOAD_MAP_ALLOC,
873 BPF_OFFLOAD_MAP_FREE,
874 XDP_SETUP_XSK_UMEM,
875};
876
877struct bpf_prog_offload_ops;
878struct netlink_ext_ack;
879struct xdp_umem;
880struct xdp_dev_bulk_queue;
881
882struct netdev_bpf {
883 enum bpf_netdev_command command;
884 union {
885 /* XDP_SETUP_PROG */
886 struct {
887 u32 flags;
888 struct bpf_prog *prog;
889 struct netlink_ext_ack *extack;
890 };
891 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
892 struct {
893 u32 prog_id;
894 /* flags with which program was installed */
895 u32 prog_flags;
896 };
897 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
898 struct {
899 struct bpf_offloaded_map *offmap;
900 };
901 /* XDP_SETUP_XSK_UMEM */
902 struct {
903 struct xdp_umem *umem;
904 u16 queue_id;
905 } xsk;
906 };
907};
908
909/* Flags for ndo_xsk_wakeup. */
910#define XDP_WAKEUP_RX (1 << 0)
911#define XDP_WAKEUP_TX (1 << 1)
912
913#ifdef CONFIG_XFRM_OFFLOAD
914struct xfrmdev_ops {
915 int (*xdo_dev_state_add) (struct xfrm_state *x);
916 void (*xdo_dev_state_delete) (struct xfrm_state *x);
917 void (*xdo_dev_state_free) (struct xfrm_state *x);
918 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
919 struct xfrm_state *x);
920 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
921};
922#endif
923
924struct dev_ifalias {
925 struct rcu_head rcuhead;
926 char ifalias[];
927};
928
929struct devlink;
930struct tlsdev_ops;
931
932struct netdev_name_node {
933 struct hlist_node hlist;
934 struct list_head list;
935 struct net_device *dev;
936 const char *name;
937};
938
939int netdev_name_node_alt_create(struct net_device *dev, const char *name);
940int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
941
942struct netdev_net_notifier {
943 struct list_head list;
944 struct notifier_block *nb;
945};
946
947/*
948 * This structure defines the management hooks for network devices.
949 * The following hooks can be defined; unless noted otherwise, they are
950 * optional and can be filled with a null pointer.
951 *
952 * int (*ndo_init)(struct net_device *dev);
953 * This function is called once when a network device is registered.
954 * The network device can use this for any late stage initialization
955 * or semantic validation. It can fail with an error code which will
956 * be propagated back to register_netdev.
957 *
958 * void (*ndo_uninit)(struct net_device *dev);
959 * This function is called when device is unregistered or when registration
960 * fails. It is not called if init fails.
961 *
962 * int (*ndo_open)(struct net_device *dev);
963 * This function is called when a network device transitions to the up
964 * state.
965 *
966 * int (*ndo_stop)(struct net_device *dev);
967 * This function is called when a network device transitions to the down
968 * state.
969 *
970 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
971 * struct net_device *dev);
972 * Called when a packet needs to be transmitted.
973 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
974 * the queue before that can happen; it's for obsolete devices and weird
975 * corner cases, but the stack really does a non-trivial amount
976 * of useless work if you return NETDEV_TX_BUSY.
977 * Required; cannot be NULL.
978 *
979 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
980 * struct net_device *dev
981 * netdev_features_t features);
982 * Called by core transmit path to determine if device is capable of
983 * performing offload operations on a given packet. This is to give
984 * the device an opportunity to implement any restrictions that cannot
985 * be otherwise expressed by feature flags. The check is called with
986 * the set of features that the stack has calculated and it returns
987 * those the driver believes to be appropriate.
988 *
989 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
990 * struct net_device *sb_dev);
991 * Called to decide which queue to use when device supports multiple
992 * transmit queues.
993 *
994 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
995 * This function is called to allow device receiver to make
996 * changes to configuration when multicast or promiscuous is enabled.
997 *
998 * void (*ndo_set_rx_mode)(struct net_device *dev);
999 * This function is called device changes address list filtering.
1000 * If driver handles unicast address filtering, it should set
1001 * IFF_UNICAST_FLT in its priv_flags.
1002 *
1003 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1004 * This function is called when the Media Access Control address
1005 * needs to be changed. If this interface is not defined, the
1006 * MAC address can not be changed.
1007 *
1008 * int (*ndo_validate_addr)(struct net_device *dev);
1009 * Test if Media Access Control address is valid for the device.
1010 *
1011 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1012 * Called when a user requests an ioctl which can't be handled by
1013 * the generic interface code. If not defined ioctls return
1014 * not supported error code.
1015 *
1016 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1017 * Used to set network devices bus interface parameters. This interface
1018 * is retained for legacy reasons; new devices should use the bus
1019 * interface (PCI) for low level management.
1020 *
1021 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1022 * Called when a user wants to change the Maximum Transfer Unit
1023 * of a device.
1024 *
1025 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1026 * Callback used when the transmitter has not made any progress
1027 * for dev->watchdog ticks.
1028 *
1029 * void (*ndo_get_stats64)(struct net_device *dev,
1030 * struct rtnl_link_stats64 *storage);
1031 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1032 * Called when a user wants to get the network device usage
1033 * statistics. Drivers must do one of the following:
1034 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1035 * rtnl_link_stats64 structure passed by the caller.
1036 * 2. Define @ndo_get_stats to update a net_device_stats structure
1037 * (which should normally be dev->stats) and return a pointer to
1038 * it. The structure may be changed asynchronously only if each
1039 * field is written atomically.
1040 * 3. Update dev->stats asynchronously and atomically, and define
1041 * neither operation.
1042 *
1043 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1044 * Return true if this device supports offload stats of this attr_id.
1045 *
1046 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1047 * void *attr_data)
1048 * Get statistics for offload operations by attr_id. Write it into the
1049 * attr_data pointer.
1050 *
1051 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1052 * If device supports VLAN filtering this function is called when a
1053 * VLAN id is registered.
1054 *
1055 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1056 * If device supports VLAN filtering this function is called when a
1057 * VLAN id is unregistered.
1058 *
1059 * void (*ndo_poll_controller)(struct net_device *dev);
1060 *
1061 * SR-IOV management functions.
1062 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1063 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1064 * u8 qos, __be16 proto);
1065 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1066 * int max_tx_rate);
1067 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1068 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1069 * int (*ndo_get_vf_config)(struct net_device *dev,
1070 * int vf, struct ifla_vf_info *ivf);
1071 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1072 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1073 * struct nlattr *port[]);
1074 *
1075 * Enable or disable the VF ability to query its RSS Redirection Table and
1076 * Hash Key. This is needed since on some devices VF share this information
1077 * with PF and querying it may introduce a theoretical security risk.
1078 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1079 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1080 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1081 * void *type_data);
1082 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1083 * This is always called from the stack with the rtnl lock held and netif
1084 * tx queues stopped. This allows the netdevice to perform queue
1085 * management safely.
1086 *
1087 * Fiber Channel over Ethernet (FCoE) offload functions.
1088 * int (*ndo_fcoe_enable)(struct net_device *dev);
1089 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1090 * so the underlying device can perform whatever needed configuration or
1091 * initialization to support acceleration of FCoE traffic.
1092 *
1093 * int (*ndo_fcoe_disable)(struct net_device *dev);
1094 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1095 * so the underlying device can perform whatever needed clean-ups to
1096 * stop supporting acceleration of FCoE traffic.
1097 *
1098 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1099 * struct scatterlist *sgl, unsigned int sgc);
1100 * Called when the FCoE Initiator wants to initialize an I/O that
1101 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1102 * perform necessary setup and returns 1 to indicate the device is set up
1103 * successfully to perform DDP on this I/O, otherwise this returns 0.
1104 *
1105 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1106 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1107 * indicated by the FC exchange id 'xid', so the underlying device can
1108 * clean up and reuse resources for later DDP requests.
1109 *
1110 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1111 * struct scatterlist *sgl, unsigned int sgc);
1112 * Called when the FCoE Target wants to initialize an I/O that
1113 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1114 * perform necessary setup and returns 1 to indicate the device is set up
1115 * successfully to perform DDP on this I/O, otherwise this returns 0.
1116 *
1117 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1118 * struct netdev_fcoe_hbainfo *hbainfo);
1119 * Called when the FCoE Protocol stack wants information on the underlying
1120 * device. This information is utilized by the FCoE protocol stack to
1121 * register attributes with Fiber Channel management service as per the
1122 * FC-GS Fabric Device Management Information(FDMI) specification.
1123 *
1124 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1125 * Called when the underlying device wants to override default World Wide
1126 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1127 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1128 * protocol stack to use.
1129 *
1130 * RFS acceleration.
1131 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1132 * u16 rxq_index, u32 flow_id);
1133 * Set hardware filter for RFS. rxq_index is the target queue index;
1134 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1135 * Return the filter ID on success, or a negative error code.
1136 *
1137 * Slave management functions (for bridge, bonding, etc).
1138 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1139 * Called to make another netdev an underling.
1140 *
1141 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1142 * Called to release previously enslaved netdev.
1143 *
1144 * Feature/offload setting functions.
1145 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1146 * netdev_features_t features);
1147 * Adjusts the requested feature flags according to device-specific
1148 * constraints, and returns the resulting flags. Must not modify
1149 * the device state.
1150 *
1151 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1152 * Called to update device configuration to new features. Passed
1153 * feature set might be less than what was returned by ndo_fix_features()).
1154 * Must return >0 or -errno if it changed dev->features itself.
1155 *
1156 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1157 * struct net_device *dev,
1158 * const unsigned char *addr, u16 vid, u16 flags,
1159 * struct netlink_ext_ack *extack);
1160 * Adds an FDB entry to dev for addr.
1161 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1162 * struct net_device *dev,
1163 * const unsigned char *addr, u16 vid)
1164 * Deletes the FDB entry from dev coresponding to addr.
1165 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1166 * struct net_device *dev, struct net_device *filter_dev,
1167 * int *idx)
1168 * Used to add FDB entries to dump requests. Implementers should add
1169 * entries to skb and update idx with the number of entries.
1170 *
1171 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1172 * u16 flags, struct netlink_ext_ack *extack)
1173 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1174 * struct net_device *dev, u32 filter_mask,
1175 * int nlflags)
1176 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1177 * u16 flags);
1178 *
1179 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1180 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1181 * which do not represent real hardware may define this to allow their
1182 * userspace components to manage their virtual carrier state. Devices
1183 * that determine carrier state from physical hardware properties (eg
1184 * network cables) or protocol-dependent mechanisms (eg
1185 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1186 *
1187 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1188 * struct netdev_phys_item_id *ppid);
1189 * Called to get ID of physical port of this device. If driver does
1190 * not implement this, it is assumed that the hw is not able to have
1191 * multiple net devices on single physical port.
1192 *
1193 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1194 * struct netdev_phys_item_id *ppid)
1195 * Called to get the parent ID of the physical port of this device.
1196 *
1197 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1198 * struct udp_tunnel_info *ti);
1199 * Called by UDP tunnel to notify a driver about the UDP port and socket
1200 * address family that a UDP tunnel is listnening to. It is called only
1201 * when a new port starts listening. The operation is protected by the
1202 * RTNL.
1203 *
1204 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1205 * struct udp_tunnel_info *ti);
1206 * Called by UDP tunnel to notify the driver about a UDP port and socket
1207 * address family that the UDP tunnel is not listening to anymore. The
1208 * operation is protected by the RTNL.
1209 *
1210 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1211 * struct net_device *dev)
1212 * Called by upper layer devices to accelerate switching or other
1213 * station functionality into hardware. 'pdev is the lowerdev
1214 * to use for the offload and 'dev' is the net device that will
1215 * back the offload. Returns a pointer to the private structure
1216 * the upper layer will maintain.
1217 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1218 * Called by upper layer device to delete the station created
1219 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1220 * the station and priv is the structure returned by the add
1221 * operation.
1222 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1223 * int queue_index, u32 maxrate);
1224 * Called when a user wants to set a max-rate limitation of specific
1225 * TX queue.
1226 * int (*ndo_get_iflink)(const struct net_device *dev);
1227 * Called to get the iflink value of this device.
1228 * void (*ndo_change_proto_down)(struct net_device *dev,
1229 * bool proto_down);
1230 * This function is used to pass protocol port error state information
1231 * to the switch driver. The switch driver can react to the proto_down
1232 * by doing a phys down on the associated switch port.
1233 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1234 * This function is used to get egress tunnel information for given skb.
1235 * This is useful for retrieving outer tunnel header parameters while
1236 * sampling packet.
1237 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1238 * This function is used to specify the headroom that the skb must
1239 * consider when allocation skb during packet reception. Setting
1240 * appropriate rx headroom value allows avoiding skb head copy on
1241 * forward. Setting a negative value resets the rx headroom to the
1242 * default value.
1243 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1244 * This function is used to set or query state related to XDP on the
1245 * netdevice and manage BPF offload. See definition of
1246 * enum bpf_netdev_command for details.
1247 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1248 * u32 flags);
1249 * This function is used to submit @n XDP packets for transmit on a
1250 * netdevice. Returns number of frames successfully transmitted, frames
1251 * that got dropped are freed/returned via xdp_return_frame().
1252 * Returns negative number, means general error invoking ndo, meaning
1253 * no frames were xmit'ed and core-caller will free all frames.
1254 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1255 * This function is used to wake up the softirq, ksoftirqd or kthread
1256 * responsible for sending and/or receiving packets on a specific
1257 * queue id bound to an AF_XDP socket. The flags field specifies if
1258 * only RX, only Tx, or both should be woken up using the flags
1259 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1260 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1261 * Get devlink port instance associated with a given netdev.
1262 * Called with a reference on the netdevice and devlink locks only,
1263 * rtnl_lock is not held.
1264 */
1265struct net_device_ops {
1266 int (*ndo_init)(struct net_device *dev);
1267 void (*ndo_uninit)(struct net_device *dev);
1268 int (*ndo_open)(struct net_device *dev);
1269 int (*ndo_stop)(struct net_device *dev);
1270 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1271 struct net_device *dev);
1272 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1273 struct net_device *dev,
1274 netdev_features_t features);
1275 u16 (*ndo_select_queue)(struct net_device *dev,
1276 struct sk_buff *skb,
1277 struct net_device *sb_dev);
1278 void (*ndo_change_rx_flags)(struct net_device *dev,
1279 int flags);
1280 void (*ndo_set_rx_mode)(struct net_device *dev);
1281 int (*ndo_set_mac_address)(struct net_device *dev,
1282 void *addr);
1283 int (*ndo_validate_addr)(struct net_device *dev);
1284 int (*ndo_do_ioctl)(struct net_device *dev,
1285 struct ifreq *ifr, int cmd);
1286 int (*ndo_set_config)(struct net_device *dev,
1287 struct ifmap *map);
1288 int (*ndo_change_mtu)(struct net_device *dev,
1289 int new_mtu);
1290 int (*ndo_neigh_setup)(struct net_device *dev,
1291 struct neigh_parms *);
1292 void (*ndo_tx_timeout) (struct net_device *dev,
1293 unsigned int txqueue);
1294
1295 void (*ndo_get_stats64)(struct net_device *dev,
1296 struct rtnl_link_stats64 *storage);
1297 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1298 int (*ndo_get_offload_stats)(int attr_id,
1299 const struct net_device *dev,
1300 void *attr_data);
1301 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1302
1303 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1304 __be16 proto, u16 vid);
1305 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1306 __be16 proto, u16 vid);
1307#ifdef CONFIG_NET_POLL_CONTROLLER
1308 void (*ndo_poll_controller)(struct net_device *dev);
1309 int (*ndo_netpoll_setup)(struct net_device *dev,
1310 struct netpoll_info *info);
1311 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1312#endif
1313 int (*ndo_set_vf_mac)(struct net_device *dev,
1314 int queue, u8 *mac);
1315 int (*ndo_set_vf_vlan)(struct net_device *dev,
1316 int queue, u16 vlan,
1317 u8 qos, __be16 proto);
1318 int (*ndo_set_vf_rate)(struct net_device *dev,
1319 int vf, int min_tx_rate,
1320 int max_tx_rate);
1321 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1322 int vf, bool setting);
1323 int (*ndo_set_vf_trust)(struct net_device *dev,
1324 int vf, bool setting);
1325 int (*ndo_get_vf_config)(struct net_device *dev,
1326 int vf,
1327 struct ifla_vf_info *ivf);
1328 int (*ndo_set_vf_link_state)(struct net_device *dev,
1329 int vf, int link_state);
1330 int (*ndo_get_vf_stats)(struct net_device *dev,
1331 int vf,
1332 struct ifla_vf_stats
1333 *vf_stats);
1334 int (*ndo_set_vf_port)(struct net_device *dev,
1335 int vf,
1336 struct nlattr *port[]);
1337 int (*ndo_get_vf_port)(struct net_device *dev,
1338 int vf, struct sk_buff *skb);
1339 int (*ndo_get_vf_guid)(struct net_device *dev,
1340 int vf,
1341 struct ifla_vf_guid *node_guid,
1342 struct ifla_vf_guid *port_guid);
1343 int (*ndo_set_vf_guid)(struct net_device *dev,
1344 int vf, u64 guid,
1345 int guid_type);
1346 int (*ndo_set_vf_rss_query_en)(
1347 struct net_device *dev,
1348 int vf, bool setting);
1349 int (*ndo_setup_tc)(struct net_device *dev,
1350 enum tc_setup_type type,
1351 void *type_data);
1352#if IS_ENABLED(CONFIG_FCOE)
1353 int (*ndo_fcoe_enable)(struct net_device *dev);
1354 int (*ndo_fcoe_disable)(struct net_device *dev);
1355 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1356 u16 xid,
1357 struct scatterlist *sgl,
1358 unsigned int sgc);
1359 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1360 u16 xid);
1361 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1362 u16 xid,
1363 struct scatterlist *sgl,
1364 unsigned int sgc);
1365 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1366 struct netdev_fcoe_hbainfo *hbainfo);
1367#endif
1368
1369#if IS_ENABLED(CONFIG_LIBFCOE)
1370#define NETDEV_FCOE_WWNN 0
1371#define NETDEV_FCOE_WWPN 1
1372 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1373 u64 *wwn, int type);
1374#endif
1375
1376#ifdef CONFIG_RFS_ACCEL
1377 int (*ndo_rx_flow_steer)(struct net_device *dev,
1378 const struct sk_buff *skb,
1379 u16 rxq_index,
1380 u32 flow_id);
1381#endif
1382 int (*ndo_add_slave)(struct net_device *dev,
1383 struct net_device *slave_dev,
1384 struct netlink_ext_ack *extack);
1385 int (*ndo_del_slave)(struct net_device *dev,
1386 struct net_device *slave_dev);
1387 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1388 netdev_features_t features);
1389 int (*ndo_set_features)(struct net_device *dev,
1390 netdev_features_t features);
1391 int (*ndo_neigh_construct)(struct net_device *dev,
1392 struct neighbour *n);
1393 void (*ndo_neigh_destroy)(struct net_device *dev,
1394 struct neighbour *n);
1395
1396 int (*ndo_fdb_add)(struct ndmsg *ndm,
1397 struct nlattr *tb[],
1398 struct net_device *dev,
1399 const unsigned char *addr,
1400 u16 vid,
1401 u16 flags,
1402 struct netlink_ext_ack *extack);
1403 int (*ndo_fdb_del)(struct ndmsg *ndm,
1404 struct nlattr *tb[],
1405 struct net_device *dev,
1406 const unsigned char *addr,
1407 u16 vid);
1408 int (*ndo_fdb_dump)(struct sk_buff *skb,
1409 struct netlink_callback *cb,
1410 struct net_device *dev,
1411 struct net_device *filter_dev,
1412 int *idx);
1413 int (*ndo_fdb_get)(struct sk_buff *skb,
1414 struct nlattr *tb[],
1415 struct net_device *dev,
1416 const unsigned char *addr,
1417 u16 vid, u32 portid, u32 seq,
1418 struct netlink_ext_ack *extack);
1419 int (*ndo_bridge_setlink)(struct net_device *dev,
1420 struct nlmsghdr *nlh,
1421 u16 flags,
1422 struct netlink_ext_ack *extack);
1423 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1424 u32 pid, u32 seq,
1425 struct net_device *dev,
1426 u32 filter_mask,
1427 int nlflags);
1428 int (*ndo_bridge_dellink)(struct net_device *dev,
1429 struct nlmsghdr *nlh,
1430 u16 flags);
1431 int (*ndo_change_carrier)(struct net_device *dev,
1432 bool new_carrier);
1433 int (*ndo_get_phys_port_id)(struct net_device *dev,
1434 struct netdev_phys_item_id *ppid);
1435 int (*ndo_get_port_parent_id)(struct net_device *dev,
1436 struct netdev_phys_item_id *ppid);
1437 int (*ndo_get_phys_port_name)(struct net_device *dev,
1438 char *name, size_t len);
1439 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1440 struct udp_tunnel_info *ti);
1441 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1442 struct udp_tunnel_info *ti);
1443 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1444 struct net_device *dev);
1445 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1446 void *priv);
1447
1448 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1449 int queue_index,
1450 u32 maxrate);
1451 int (*ndo_get_iflink)(const struct net_device *dev);
1452 int (*ndo_change_proto_down)(struct net_device *dev,
1453 bool proto_down);
1454 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1455 struct sk_buff *skb);
1456 void (*ndo_set_rx_headroom)(struct net_device *dev,
1457 int needed_headroom);
1458 int (*ndo_bpf)(struct net_device *dev,
1459 struct netdev_bpf *bpf);
1460 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1461 struct xdp_frame **xdp,
1462 u32 flags);
1463 int (*ndo_xsk_wakeup)(struct net_device *dev,
1464 u32 queue_id, u32 flags);
1465 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1466};
1467
1468/**
1469 * enum net_device_priv_flags - &struct net_device priv_flags
1470 *
1471 * These are the &struct net_device, they are only set internally
1472 * by drivers and used in the kernel. These flags are invisible to
1473 * userspace; this means that the order of these flags can change
1474 * during any kernel release.
1475 *
1476 * You should have a pretty good reason to be extending these flags.
1477 *
1478 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1479 * @IFF_EBRIDGE: Ethernet bridging device
1480 * @IFF_BONDING: bonding master or slave
1481 * @IFF_ISATAP: ISATAP interface (RFC4214)
1482 * @IFF_WAN_HDLC: WAN HDLC device
1483 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1484 * release skb->dst
1485 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1486 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1487 * @IFF_MACVLAN_PORT: device used as macvlan port
1488 * @IFF_BRIDGE_PORT: device used as bridge port
1489 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1490 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1491 * @IFF_UNICAST_FLT: Supports unicast filtering
1492 * @IFF_TEAM_PORT: device used as team port
1493 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1494 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1495 * change when it's running
1496 * @IFF_MACVLAN: Macvlan device
1497 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1498 * underlying stacked devices
1499 * @IFF_L3MDEV_MASTER: device is an L3 master device
1500 * @IFF_NO_QUEUE: device can run without qdisc attached
1501 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1502 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1503 * @IFF_TEAM: device is a team device
1504 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1505 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1506 * entity (i.e. the master device for bridged veth)
1507 * @IFF_MACSEC: device is a MACsec device
1508 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1509 * @IFF_FAILOVER: device is a failover master device
1510 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1511 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1512 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1513 */
1514enum netdev_priv_flags {
1515 IFF_802_1Q_VLAN = 1<<0,
1516 IFF_EBRIDGE = 1<<1,
1517 IFF_BONDING = 1<<2,
1518 IFF_ISATAP = 1<<3,
1519 IFF_WAN_HDLC = 1<<4,
1520 IFF_XMIT_DST_RELEASE = 1<<5,
1521 IFF_DONT_BRIDGE = 1<<6,
1522 IFF_DISABLE_NETPOLL = 1<<7,
1523 IFF_MACVLAN_PORT = 1<<8,
1524 IFF_BRIDGE_PORT = 1<<9,
1525 IFF_OVS_DATAPATH = 1<<10,
1526 IFF_TX_SKB_SHARING = 1<<11,
1527 IFF_UNICAST_FLT = 1<<12,
1528 IFF_TEAM_PORT = 1<<13,
1529 IFF_SUPP_NOFCS = 1<<14,
1530 IFF_LIVE_ADDR_CHANGE = 1<<15,
1531 IFF_MACVLAN = 1<<16,
1532 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1533 IFF_L3MDEV_MASTER = 1<<18,
1534 IFF_NO_QUEUE = 1<<19,
1535 IFF_OPENVSWITCH = 1<<20,
1536 IFF_L3MDEV_SLAVE = 1<<21,
1537 IFF_TEAM = 1<<22,
1538 IFF_RXFH_CONFIGURED = 1<<23,
1539 IFF_PHONY_HEADROOM = 1<<24,
1540 IFF_MACSEC = 1<<25,
1541 IFF_NO_RX_HANDLER = 1<<26,
1542 IFF_FAILOVER = 1<<27,
1543 IFF_FAILOVER_SLAVE = 1<<28,
1544 IFF_L3MDEV_RX_HANDLER = 1<<29,
1545 IFF_LIVE_RENAME_OK = 1<<30,
1546};
1547
1548#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1549#define IFF_EBRIDGE IFF_EBRIDGE
1550#define IFF_BONDING IFF_BONDING
1551#define IFF_ISATAP IFF_ISATAP
1552#define IFF_WAN_HDLC IFF_WAN_HDLC
1553#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1554#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1555#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1556#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1557#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1558#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1559#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1560#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1561#define IFF_TEAM_PORT IFF_TEAM_PORT
1562#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1563#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1564#define IFF_MACVLAN IFF_MACVLAN
1565#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1566#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1567#define IFF_NO_QUEUE IFF_NO_QUEUE
1568#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1569#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1570#define IFF_TEAM IFF_TEAM
1571#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1572#define IFF_MACSEC IFF_MACSEC
1573#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1574#define IFF_FAILOVER IFF_FAILOVER
1575#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1576#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1577#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1578
1579/**
1580 * struct net_device - The DEVICE structure.
1581 *
1582 * Actually, this whole structure is a big mistake. It mixes I/O
1583 * data with strictly "high-level" data, and it has to know about
1584 * almost every data structure used in the INET module.
1585 *
1586 * @name: This is the first field of the "visible" part of this structure
1587 * (i.e. as seen by users in the "Space.c" file). It is the name
1588 * of the interface.
1589 *
1590 * @name_node: Name hashlist node
1591 * @ifalias: SNMP alias
1592 * @mem_end: Shared memory end
1593 * @mem_start: Shared memory start
1594 * @base_addr: Device I/O address
1595 * @irq: Device IRQ number
1596 *
1597 * @state: Generic network queuing layer state, see netdev_state_t
1598 * @dev_list: The global list of network devices
1599 * @napi_list: List entry used for polling NAPI devices
1600 * @unreg_list: List entry when we are unregistering the
1601 * device; see the function unregister_netdev
1602 * @close_list: List entry used when we are closing the device
1603 * @ptype_all: Device-specific packet handlers for all protocols
1604 * @ptype_specific: Device-specific, protocol-specific packet handlers
1605 *
1606 * @adj_list: Directly linked devices, like slaves for bonding
1607 * @features: Currently active device features
1608 * @hw_features: User-changeable features
1609 *
1610 * @wanted_features: User-requested features
1611 * @vlan_features: Mask of features inheritable by VLAN devices
1612 *
1613 * @hw_enc_features: Mask of features inherited by encapsulating devices
1614 * This field indicates what encapsulation
1615 * offloads the hardware is capable of doing,
1616 * and drivers will need to set them appropriately.
1617 *
1618 * @mpls_features: Mask of features inheritable by MPLS
1619 *
1620 * @ifindex: interface index
1621 * @group: The group the device belongs to
1622 *
1623 * @stats: Statistics struct, which was left as a legacy, use
1624 * rtnl_link_stats64 instead
1625 *
1626 * @rx_dropped: Dropped packets by core network,
1627 * do not use this in drivers
1628 * @tx_dropped: Dropped packets by core network,
1629 * do not use this in drivers
1630 * @rx_nohandler: nohandler dropped packets by core network on
1631 * inactive devices, do not use this in drivers
1632 * @carrier_up_count: Number of times the carrier has been up
1633 * @carrier_down_count: Number of times the carrier has been down
1634 *
1635 * @wireless_handlers: List of functions to handle Wireless Extensions,
1636 * instead of ioctl,
1637 * see <net/iw_handler.h> for details.
1638 * @wireless_data: Instance data managed by the core of wireless extensions
1639 *
1640 * @netdev_ops: Includes several pointers to callbacks,
1641 * if one wants to override the ndo_*() functions
1642 * @ethtool_ops: Management operations
1643 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1644 * discovery handling. Necessary for e.g. 6LoWPAN.
1645 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1646 * of Layer 2 headers.
1647 *
1648 * @flags: Interface flags (a la BSD)
1649 * @priv_flags: Like 'flags' but invisible to userspace,
1650 * see if.h for the definitions
1651 * @gflags: Global flags ( kept as legacy )
1652 * @padded: How much padding added by alloc_netdev()
1653 * @operstate: RFC2863 operstate
1654 * @link_mode: Mapping policy to operstate
1655 * @if_port: Selectable AUI, TP, ...
1656 * @dma: DMA channel
1657 * @mtu: Interface MTU value
1658 * @min_mtu: Interface Minimum MTU value
1659 * @max_mtu: Interface Maximum MTU value
1660 * @type: Interface hardware type
1661 * @hard_header_len: Maximum hardware header length.
1662 * @min_header_len: Minimum hardware header length
1663 *
1664 * @needed_headroom: Extra headroom the hardware may need, but not in all
1665 * cases can this be guaranteed
1666 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1667 * cases can this be guaranteed. Some cases also use
1668 * LL_MAX_HEADER instead to allocate the skb
1669 *
1670 * interface address info:
1671 *
1672 * @perm_addr: Permanent hw address
1673 * @addr_assign_type: Hw address assignment type
1674 * @addr_len: Hardware address length
1675 * @upper_level: Maximum depth level of upper devices.
1676 * @lower_level: Maximum depth level of lower devices.
1677 * @neigh_priv_len: Used in neigh_alloc()
1678 * @dev_id: Used to differentiate devices that share
1679 * the same link layer address
1680 * @dev_port: Used to differentiate devices that share
1681 * the same function
1682 * @addr_list_lock: XXX: need comments on this one
1683 * @uc_promisc: Counter that indicates promiscuous mode
1684 * has been enabled due to the need to listen to
1685 * additional unicast addresses in a device that
1686 * does not implement ndo_set_rx_mode()
1687 * @uc: unicast mac addresses
1688 * @mc: multicast mac addresses
1689 * @dev_addrs: list of device hw addresses
1690 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1691 * @promiscuity: Number of times the NIC is told to work in
1692 * promiscuous mode; if it becomes 0 the NIC will
1693 * exit promiscuous mode
1694 * @allmulti: Counter, enables or disables allmulticast mode
1695 *
1696 * @vlan_info: VLAN info
1697 * @dsa_ptr: dsa specific data
1698 * @tipc_ptr: TIPC specific data
1699 * @atalk_ptr: AppleTalk link
1700 * @ip_ptr: IPv4 specific data
1701 * @dn_ptr: DECnet specific data
1702 * @ip6_ptr: IPv6 specific data
1703 * @ax25_ptr: AX.25 specific data
1704 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1705 *
1706 * @dev_addr: Hw address (before bcast,
1707 * because most packets are unicast)
1708 *
1709 * @_rx: Array of RX queues
1710 * @num_rx_queues: Number of RX queues
1711 * allocated at register_netdev() time
1712 * @real_num_rx_queues: Number of RX queues currently active in device
1713 *
1714 * @rx_handler: handler for received packets
1715 * @rx_handler_data: XXX: need comments on this one
1716 * @miniq_ingress: ingress/clsact qdisc specific data for
1717 * ingress processing
1718 * @ingress_queue: XXX: need comments on this one
1719 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1720 * @broadcast: hw bcast address
1721 *
1722 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1723 * indexed by RX queue number. Assigned by driver.
1724 * This must only be set if the ndo_rx_flow_steer
1725 * operation is defined
1726 * @index_hlist: Device index hash chain
1727 *
1728 * @_tx: Array of TX queues
1729 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1730 * @real_num_tx_queues: Number of TX queues currently active in device
1731 * @qdisc: Root qdisc from userspace point of view
1732 * @tx_queue_len: Max frames per queue allowed
1733 * @tx_global_lock: XXX: need comments on this one
1734 *
1735 * @xps_maps: XXX: need comments on this one
1736 * @miniq_egress: clsact qdisc specific data for
1737 * egress processing
1738 * @watchdog_timeo: Represents the timeout that is used by
1739 * the watchdog (see dev_watchdog())
1740 * @watchdog_timer: List of timers
1741 *
1742 * @pcpu_refcnt: Number of references to this device
1743 * @todo_list: Delayed register/unregister
1744 * @link_watch_list: XXX: need comments on this one
1745 *
1746 * @reg_state: Register/unregister state machine
1747 * @dismantle: Device is going to be freed
1748 * @rtnl_link_state: This enum represents the phases of creating
1749 * a new link
1750 *
1751 * @needs_free_netdev: Should unregister perform free_netdev?
1752 * @priv_destructor: Called from unregister
1753 * @npinfo: XXX: need comments on this one
1754 * @nd_net: Network namespace this network device is inside
1755 *
1756 * @ml_priv: Mid-layer private
1757 * @lstats: Loopback statistics
1758 * @tstats: Tunnel statistics
1759 * @dstats: Dummy statistics
1760 * @vstats: Virtual ethernet statistics
1761 *
1762 * @garp_port: GARP
1763 * @mrp_port: MRP
1764 *
1765 * @dev: Class/net/name entry
1766 * @sysfs_groups: Space for optional device, statistics and wireless
1767 * sysfs groups
1768 *
1769 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1770 * @rtnl_link_ops: Rtnl_link_ops
1771 *
1772 * @gso_max_size: Maximum size of generic segmentation offload
1773 * @gso_max_segs: Maximum number of segments that can be passed to the
1774 * NIC for GSO
1775 *
1776 * @dcbnl_ops: Data Center Bridging netlink ops
1777 * @num_tc: Number of traffic classes in the net device
1778 * @tc_to_txq: XXX: need comments on this one
1779 * @prio_tc_map: XXX: need comments on this one
1780 *
1781 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1782 *
1783 * @priomap: XXX: need comments on this one
1784 * @phydev: Physical device may attach itself
1785 * for hardware timestamping
1786 * @sfp_bus: attached &struct sfp_bus structure.
1787 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1788 * spinlock
1789 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1790 * @qdisc_xmit_lock_key: lockdep class annotating
1791 * netdev_queue->_xmit_lock spinlock
1792 * @addr_list_lock_key: lockdep class annotating
1793 * net_device->addr_list_lock spinlock
1794 *
1795 * @proto_down: protocol port state information can be sent to the
1796 * switch driver and used to set the phys state of the
1797 * switch port.
1798 *
1799 * @wol_enabled: Wake-on-LAN is enabled
1800 *
1801 * @net_notifier_list: List of per-net netdev notifier block
1802 * that follow this device when it is moved
1803 * to another network namespace.
1804 *
1805 * FIXME: cleanup struct net_device such that network protocol info
1806 * moves out.
1807 */
1808
1809struct net_device {
1810 char name[IFNAMSIZ];
1811 struct netdev_name_node *name_node;
1812 struct dev_ifalias __rcu *ifalias;
1813 /*
1814 * I/O specific fields
1815 * FIXME: Merge these and struct ifmap into one
1816 */
1817 unsigned long mem_end;
1818 unsigned long mem_start;
1819 unsigned long base_addr;
1820 int irq;
1821
1822 /*
1823 * Some hardware also needs these fields (state,dev_list,
1824 * napi_list,unreg_list,close_list) but they are not
1825 * part of the usual set specified in Space.c.
1826 */
1827
1828 unsigned long state;
1829
1830 struct list_head dev_list;
1831 struct list_head napi_list;
1832 struct list_head unreg_list;
1833 struct list_head close_list;
1834 struct list_head ptype_all;
1835 struct list_head ptype_specific;
1836
1837 struct {
1838 struct list_head upper;
1839 struct list_head lower;
1840 } adj_list;
1841
1842 netdev_features_t features;
1843 netdev_features_t hw_features;
1844 netdev_features_t wanted_features;
1845 netdev_features_t vlan_features;
1846 netdev_features_t hw_enc_features;
1847 netdev_features_t mpls_features;
1848 netdev_features_t gso_partial_features;
1849
1850 int ifindex;
1851 int group;
1852
1853 struct net_device_stats stats;
1854
1855 atomic_long_t rx_dropped;
1856 atomic_long_t tx_dropped;
1857 atomic_long_t rx_nohandler;
1858
1859 /* Stats to monitor link on/off, flapping */
1860 atomic_t carrier_up_count;
1861 atomic_t carrier_down_count;
1862
1863#ifdef CONFIG_WIRELESS_EXT
1864 const struct iw_handler_def *wireless_handlers;
1865 struct iw_public_data *wireless_data;
1866#endif
1867 const struct net_device_ops *netdev_ops;
1868 const struct ethtool_ops *ethtool_ops;
1869#ifdef CONFIG_NET_L3_MASTER_DEV
1870 const struct l3mdev_ops *l3mdev_ops;
1871#endif
1872#if IS_ENABLED(CONFIG_IPV6)
1873 const struct ndisc_ops *ndisc_ops;
1874#endif
1875
1876#ifdef CONFIG_XFRM_OFFLOAD
1877 const struct xfrmdev_ops *xfrmdev_ops;
1878#endif
1879
1880#if IS_ENABLED(CONFIG_TLS_DEVICE)
1881 const struct tlsdev_ops *tlsdev_ops;
1882#endif
1883
1884 const struct header_ops *header_ops;
1885
1886 unsigned int flags;
1887 unsigned int priv_flags;
1888
1889 unsigned short gflags;
1890 unsigned short padded;
1891
1892 unsigned char operstate;
1893 unsigned char link_mode;
1894
1895 unsigned char if_port;
1896 unsigned char dma;
1897
1898 /* Note : dev->mtu is often read without holding a lock.
1899 * Writers usually hold RTNL.
1900 * It is recommended to use READ_ONCE() to annotate the reads,
1901 * and to use WRITE_ONCE() to annotate the writes.
1902 */
1903 unsigned int mtu;
1904 unsigned int min_mtu;
1905 unsigned int max_mtu;
1906 unsigned short type;
1907 unsigned short hard_header_len;
1908 unsigned char min_header_len;
1909
1910 unsigned short needed_headroom;
1911 unsigned short needed_tailroom;
1912
1913 /* Interface address info. */
1914 unsigned char perm_addr[MAX_ADDR_LEN];
1915 unsigned char addr_assign_type;
1916 unsigned char addr_len;
1917 unsigned char upper_level;
1918 unsigned char lower_level;
1919 unsigned short neigh_priv_len;
1920 unsigned short dev_id;
1921 unsigned short dev_port;
1922 spinlock_t addr_list_lock;
1923 unsigned char name_assign_type;
1924 bool uc_promisc;
1925 struct netdev_hw_addr_list uc;
1926 struct netdev_hw_addr_list mc;
1927 struct netdev_hw_addr_list dev_addrs;
1928
1929#ifdef CONFIG_SYSFS
1930 struct kset *queues_kset;
1931#endif
1932 unsigned int promiscuity;
1933 unsigned int allmulti;
1934
1935
1936 /* Protocol-specific pointers */
1937
1938#if IS_ENABLED(CONFIG_VLAN_8021Q)
1939 struct vlan_info __rcu *vlan_info;
1940#endif
1941#if IS_ENABLED(CONFIG_NET_DSA)
1942 struct dsa_port *dsa_ptr;
1943#endif
1944#if IS_ENABLED(CONFIG_TIPC)
1945 struct tipc_bearer __rcu *tipc_ptr;
1946#endif
1947#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1948 void *atalk_ptr;
1949#endif
1950 struct in_device __rcu *ip_ptr;
1951#if IS_ENABLED(CONFIG_DECNET)
1952 struct dn_dev __rcu *dn_ptr;
1953#endif
1954 struct inet6_dev __rcu *ip6_ptr;
1955#if IS_ENABLED(CONFIG_AX25)
1956 void *ax25_ptr;
1957#endif
1958 struct wireless_dev *ieee80211_ptr;
1959 struct wpan_dev *ieee802154_ptr;
1960#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1961 struct mpls_dev __rcu *mpls_ptr;
1962#endif
1963
1964/*
1965 * Cache lines mostly used on receive path (including eth_type_trans())
1966 */
1967 /* Interface address info used in eth_type_trans() */
1968 unsigned char *dev_addr;
1969
1970 struct netdev_rx_queue *_rx;
1971 unsigned int num_rx_queues;
1972 unsigned int real_num_rx_queues;
1973
1974 struct bpf_prog __rcu *xdp_prog;
1975 unsigned long gro_flush_timeout;
1976 rx_handler_func_t __rcu *rx_handler;
1977 void __rcu *rx_handler_data;
1978
1979#ifdef CONFIG_NET_CLS_ACT
1980 struct mini_Qdisc __rcu *miniq_ingress;
1981#endif
1982 struct netdev_queue __rcu *ingress_queue;
1983#ifdef CONFIG_NETFILTER_INGRESS
1984 struct nf_hook_entries __rcu *nf_hooks_ingress;
1985#endif
1986
1987 unsigned char broadcast[MAX_ADDR_LEN];
1988#ifdef CONFIG_RFS_ACCEL
1989 struct cpu_rmap *rx_cpu_rmap;
1990#endif
1991 struct hlist_node index_hlist;
1992
1993/*
1994 * Cache lines mostly used on transmit path
1995 */
1996 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1997 unsigned int num_tx_queues;
1998 unsigned int real_num_tx_queues;
1999 struct Qdisc *qdisc;
2000 unsigned int tx_queue_len;
2001 spinlock_t tx_global_lock;
2002
2003 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2004
2005#ifdef CONFIG_XPS
2006 struct xps_dev_maps __rcu *xps_cpus_map;
2007 struct xps_dev_maps __rcu *xps_rxqs_map;
2008#endif
2009#ifdef CONFIG_NET_CLS_ACT
2010 struct mini_Qdisc __rcu *miniq_egress;
2011#endif
2012
2013#ifdef CONFIG_NET_SCHED
2014 DECLARE_HASHTABLE (qdisc_hash, 4);
2015#endif
2016 /* These may be needed for future network-power-down code. */
2017 struct timer_list watchdog_timer;
2018 int watchdog_timeo;
2019
2020 struct list_head todo_list;
2021 int __percpu *pcpu_refcnt;
2022
2023 struct list_head link_watch_list;
2024
2025 enum { NETREG_UNINITIALIZED=0,
2026 NETREG_REGISTERED, /* completed register_netdevice */
2027 NETREG_UNREGISTERING, /* called unregister_netdevice */
2028 NETREG_UNREGISTERED, /* completed unregister todo */
2029 NETREG_RELEASED, /* called free_netdev */
2030 NETREG_DUMMY, /* dummy device for NAPI poll */
2031 } reg_state:8;
2032
2033 bool dismantle;
2034
2035 enum {
2036 RTNL_LINK_INITIALIZED,
2037 RTNL_LINK_INITIALIZING,
2038 } rtnl_link_state:16;
2039
2040 bool needs_free_netdev;
2041 void (*priv_destructor)(struct net_device *dev);
2042
2043#ifdef CONFIG_NETPOLL
2044 struct netpoll_info __rcu *npinfo;
2045#endif
2046
2047 possible_net_t nd_net;
2048
2049 /* mid-layer private */
2050 union {
2051 void *ml_priv;
2052 struct pcpu_lstats __percpu *lstats;
2053 struct pcpu_sw_netstats __percpu *tstats;
2054 struct pcpu_dstats __percpu *dstats;
2055 };
2056
2057#if IS_ENABLED(CONFIG_GARP)
2058 struct garp_port __rcu *garp_port;
2059#endif
2060#if IS_ENABLED(CONFIG_MRP)
2061 struct mrp_port __rcu *mrp_port;
2062#endif
2063
2064 struct device dev;
2065 const struct attribute_group *sysfs_groups[4];
2066 const struct attribute_group *sysfs_rx_queue_group;
2067
2068 const struct rtnl_link_ops *rtnl_link_ops;
2069
2070 /* for setting kernel sock attribute on TCP connection setup */
2071#define GSO_MAX_SIZE 65536
2072 unsigned int gso_max_size;
2073#define GSO_MAX_SEGS 65535
2074 u16 gso_max_segs;
2075
2076#ifdef CONFIG_DCB
2077 const struct dcbnl_rtnl_ops *dcbnl_ops;
2078#endif
2079 s16 num_tc;
2080 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2081 u8 prio_tc_map[TC_BITMASK + 1];
2082
2083#if IS_ENABLED(CONFIG_FCOE)
2084 unsigned int fcoe_ddp_xid;
2085#endif
2086#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2087 struct netprio_map __rcu *priomap;
2088#endif
2089 struct phy_device *phydev;
2090 struct sfp_bus *sfp_bus;
2091 struct lock_class_key qdisc_tx_busylock_key;
2092 struct lock_class_key qdisc_running_key;
2093 struct lock_class_key qdisc_xmit_lock_key;
2094 struct lock_class_key addr_list_lock_key;
2095 bool proto_down;
2096 unsigned wol_enabled:1;
2097
2098 struct list_head net_notifier_list;
2099};
2100#define to_net_dev(d) container_of(d, struct net_device, dev)
2101
2102static inline bool netif_elide_gro(const struct net_device *dev)
2103{
2104 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2105 return true;
2106 return false;
2107}
2108
2109#define NETDEV_ALIGN 32
2110
2111static inline
2112int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2113{
2114 return dev->prio_tc_map[prio & TC_BITMASK];
2115}
2116
2117static inline
2118int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2119{
2120 if (tc >= dev->num_tc)
2121 return -EINVAL;
2122
2123 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2124 return 0;
2125}
2126
2127int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2128void netdev_reset_tc(struct net_device *dev);
2129int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2130int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2131
2132static inline
2133int netdev_get_num_tc(struct net_device *dev)
2134{
2135 return dev->num_tc;
2136}
2137
2138void netdev_unbind_sb_channel(struct net_device *dev,
2139 struct net_device *sb_dev);
2140int netdev_bind_sb_channel_queue(struct net_device *dev,
2141 struct net_device *sb_dev,
2142 u8 tc, u16 count, u16 offset);
2143int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2144static inline int netdev_get_sb_channel(struct net_device *dev)
2145{
2146 return max_t(int, -dev->num_tc, 0);
2147}
2148
2149static inline
2150struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2151 unsigned int index)
2152{
2153 return &dev->_tx[index];
2154}
2155
2156static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2157 const struct sk_buff *skb)
2158{
2159 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2160}
2161
2162static inline void netdev_for_each_tx_queue(struct net_device *dev,
2163 void (*f)(struct net_device *,
2164 struct netdev_queue *,
2165 void *),
2166 void *arg)
2167{
2168 unsigned int i;
2169
2170 for (i = 0; i < dev->num_tx_queues; i++)
2171 f(dev, &dev->_tx[i], arg);
2172}
2173
2174u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2175 struct net_device *sb_dev);
2176struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2177 struct sk_buff *skb,
2178 struct net_device *sb_dev);
2179
2180/* returns the headroom that the master device needs to take in account
2181 * when forwarding to this dev
2182 */
2183static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2184{
2185 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2186}
2187
2188static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2189{
2190 if (dev->netdev_ops->ndo_set_rx_headroom)
2191 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2192}
2193
2194/* set the device rx headroom to the dev's default */
2195static inline void netdev_reset_rx_headroom(struct net_device *dev)
2196{
2197 netdev_set_rx_headroom(dev, -1);
2198}
2199
2200/*
2201 * Net namespace inlines
2202 */
2203static inline
2204struct net *dev_net(const struct net_device *dev)
2205{
2206 return read_pnet(&dev->nd_net);
2207}
2208
2209static inline
2210void dev_net_set(struct net_device *dev, struct net *net)
2211{
2212 write_pnet(&dev->nd_net, net);
2213}
2214
2215/**
2216 * netdev_priv - access network device private data
2217 * @dev: network device
2218 *
2219 * Get network device private data
2220 */
2221static inline void *netdev_priv(const struct net_device *dev)
2222{
2223 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2224}
2225
2226/* Set the sysfs physical device reference for the network logical device
2227 * if set prior to registration will cause a symlink during initialization.
2228 */
2229#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2230
2231/* Set the sysfs device type for the network logical device to allow
2232 * fine-grained identification of different network device types. For
2233 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2234 */
2235#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2236
2237/* Default NAPI poll() weight
2238 * Device drivers are strongly advised to not use bigger value
2239 */
2240#define NAPI_POLL_WEIGHT 64
2241
2242/**
2243 * netif_napi_add - initialize a NAPI context
2244 * @dev: network device
2245 * @napi: NAPI context
2246 * @poll: polling function
2247 * @weight: default weight
2248 *
2249 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2250 * *any* of the other NAPI-related functions.
2251 */
2252void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2253 int (*poll)(struct napi_struct *, int), int weight);
2254
2255/**
2256 * netif_tx_napi_add - initialize a NAPI context
2257 * @dev: network device
2258 * @napi: NAPI context
2259 * @poll: polling function
2260 * @weight: default weight
2261 *
2262 * This variant of netif_napi_add() should be used from drivers using NAPI
2263 * to exclusively poll a TX queue.
2264 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2265 */
2266static inline void netif_tx_napi_add(struct net_device *dev,
2267 struct napi_struct *napi,
2268 int (*poll)(struct napi_struct *, int),
2269 int weight)
2270{
2271 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2272 netif_napi_add(dev, napi, poll, weight);
2273}
2274
2275/**
2276 * netif_napi_del - remove a NAPI context
2277 * @napi: NAPI context
2278 *
2279 * netif_napi_del() removes a NAPI context from the network device NAPI list
2280 */
2281void netif_napi_del(struct napi_struct *napi);
2282
2283struct napi_gro_cb {
2284 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2285 void *frag0;
2286
2287 /* Length of frag0. */
2288 unsigned int frag0_len;
2289
2290 /* This indicates where we are processing relative to skb->data. */
2291 int data_offset;
2292
2293 /* This is non-zero if the packet cannot be merged with the new skb. */
2294 u16 flush;
2295
2296 /* Save the IP ID here and check when we get to the transport layer */
2297 u16 flush_id;
2298
2299 /* Number of segments aggregated. */
2300 u16 count;
2301
2302 /* Start offset for remote checksum offload */
2303 u16 gro_remcsum_start;
2304
2305 /* jiffies when first packet was created/queued */
2306 unsigned long age;
2307
2308 /* Used in ipv6_gro_receive() and foo-over-udp */
2309 u16 proto;
2310
2311 /* This is non-zero if the packet may be of the same flow. */
2312 u8 same_flow:1;
2313
2314 /* Used in tunnel GRO receive */
2315 u8 encap_mark:1;
2316
2317 /* GRO checksum is valid */
2318 u8 csum_valid:1;
2319
2320 /* Number of checksums via CHECKSUM_UNNECESSARY */
2321 u8 csum_cnt:3;
2322
2323 /* Free the skb? */
2324 u8 free:2;
2325#define NAPI_GRO_FREE 1
2326#define NAPI_GRO_FREE_STOLEN_HEAD 2
2327
2328 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2329 u8 is_ipv6:1;
2330
2331 /* Used in GRE, set in fou/gue_gro_receive */
2332 u8 is_fou:1;
2333
2334 /* Used to determine if flush_id can be ignored */
2335 u8 is_atomic:1;
2336
2337 /* Number of gro_receive callbacks this packet already went through */
2338 u8 recursion_counter:4;
2339
2340 /* GRO is done by frag_list pointer chaining. */
2341 u8 is_flist:1;
2342
2343 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2344 __wsum csum;
2345
2346 /* used in skb_gro_receive() slow path */
2347 struct sk_buff *last;
2348};
2349
2350#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2351
2352#define GRO_RECURSION_LIMIT 15
2353static inline int gro_recursion_inc_test(struct sk_buff *skb)
2354{
2355 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2356}
2357
2358typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2359static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2360 struct list_head *head,
2361 struct sk_buff *skb)
2362{
2363 if (unlikely(gro_recursion_inc_test(skb))) {
2364 NAPI_GRO_CB(skb)->flush |= 1;
2365 return NULL;
2366 }
2367
2368 return cb(head, skb);
2369}
2370
2371typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2372 struct sk_buff *);
2373static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2374 struct sock *sk,
2375 struct list_head *head,
2376 struct sk_buff *skb)
2377{
2378 if (unlikely(gro_recursion_inc_test(skb))) {
2379 NAPI_GRO_CB(skb)->flush |= 1;
2380 return NULL;
2381 }
2382
2383 return cb(sk, head, skb);
2384}
2385
2386struct packet_type {
2387 __be16 type; /* This is really htons(ether_type). */
2388 bool ignore_outgoing;
2389 struct net_device *dev; /* NULL is wildcarded here */
2390 int (*func) (struct sk_buff *,
2391 struct net_device *,
2392 struct packet_type *,
2393 struct net_device *);
2394 void (*list_func) (struct list_head *,
2395 struct packet_type *,
2396 struct net_device *);
2397 bool (*id_match)(struct packet_type *ptype,
2398 struct sock *sk);
2399 void *af_packet_priv;
2400 struct list_head list;
2401};
2402
2403struct offload_callbacks {
2404 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2405 netdev_features_t features);
2406 struct sk_buff *(*gro_receive)(struct list_head *head,
2407 struct sk_buff *skb);
2408 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2409};
2410
2411struct packet_offload {
2412 __be16 type; /* This is really htons(ether_type). */
2413 u16 priority;
2414 struct offload_callbacks callbacks;
2415 struct list_head list;
2416};
2417
2418/* often modified stats are per-CPU, other are shared (netdev->stats) */
2419struct pcpu_sw_netstats {
2420 u64 rx_packets;
2421 u64 rx_bytes;
2422 u64 tx_packets;
2423 u64 tx_bytes;
2424 struct u64_stats_sync syncp;
2425} __aligned(4 * sizeof(u64));
2426
2427struct pcpu_lstats {
2428 u64_stats_t packets;
2429 u64_stats_t bytes;
2430 struct u64_stats_sync syncp;
2431} __aligned(2 * sizeof(u64));
2432
2433void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2434
2435static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2436{
2437 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2438
2439 u64_stats_update_begin(&lstats->syncp);
2440 u64_stats_add(&lstats->bytes, len);
2441 u64_stats_inc(&lstats->packets);
2442 u64_stats_update_end(&lstats->syncp);
2443}
2444
2445#define __netdev_alloc_pcpu_stats(type, gfp) \
2446({ \
2447 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2448 if (pcpu_stats) { \
2449 int __cpu; \
2450 for_each_possible_cpu(__cpu) { \
2451 typeof(type) *stat; \
2452 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2453 u64_stats_init(&stat->syncp); \
2454 } \
2455 } \
2456 pcpu_stats; \
2457})
2458
2459#define netdev_alloc_pcpu_stats(type) \
2460 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2461
2462enum netdev_lag_tx_type {
2463 NETDEV_LAG_TX_TYPE_UNKNOWN,
2464 NETDEV_LAG_TX_TYPE_RANDOM,
2465 NETDEV_LAG_TX_TYPE_BROADCAST,
2466 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2467 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2468 NETDEV_LAG_TX_TYPE_HASH,
2469};
2470
2471enum netdev_lag_hash {
2472 NETDEV_LAG_HASH_NONE,
2473 NETDEV_LAG_HASH_L2,
2474 NETDEV_LAG_HASH_L34,
2475 NETDEV_LAG_HASH_L23,
2476 NETDEV_LAG_HASH_E23,
2477 NETDEV_LAG_HASH_E34,
2478 NETDEV_LAG_HASH_UNKNOWN,
2479};
2480
2481struct netdev_lag_upper_info {
2482 enum netdev_lag_tx_type tx_type;
2483 enum netdev_lag_hash hash_type;
2484};
2485
2486struct netdev_lag_lower_state_info {
2487 u8 link_up : 1,
2488 tx_enabled : 1;
2489};
2490
2491#include <linux/notifier.h>
2492
2493/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2494 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2495 * adding new types.
2496 */
2497enum netdev_cmd {
2498 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2499 NETDEV_DOWN,
2500 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2501 detected a hardware crash and restarted
2502 - we can use this eg to kick tcp sessions
2503 once done */
2504 NETDEV_CHANGE, /* Notify device state change */
2505 NETDEV_REGISTER,
2506 NETDEV_UNREGISTER,
2507 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2508 NETDEV_CHANGEADDR, /* notify after the address change */
2509 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2510 NETDEV_GOING_DOWN,
2511 NETDEV_CHANGENAME,
2512 NETDEV_FEAT_CHANGE,
2513 NETDEV_BONDING_FAILOVER,
2514 NETDEV_PRE_UP,
2515 NETDEV_PRE_TYPE_CHANGE,
2516 NETDEV_POST_TYPE_CHANGE,
2517 NETDEV_POST_INIT,
2518 NETDEV_RELEASE,
2519 NETDEV_NOTIFY_PEERS,
2520 NETDEV_JOIN,
2521 NETDEV_CHANGEUPPER,
2522 NETDEV_RESEND_IGMP,
2523 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2524 NETDEV_CHANGEINFODATA,
2525 NETDEV_BONDING_INFO,
2526 NETDEV_PRECHANGEUPPER,
2527 NETDEV_CHANGELOWERSTATE,
2528 NETDEV_UDP_TUNNEL_PUSH_INFO,
2529 NETDEV_UDP_TUNNEL_DROP_INFO,
2530 NETDEV_CHANGE_TX_QUEUE_LEN,
2531 NETDEV_CVLAN_FILTER_PUSH_INFO,
2532 NETDEV_CVLAN_FILTER_DROP_INFO,
2533 NETDEV_SVLAN_FILTER_PUSH_INFO,
2534 NETDEV_SVLAN_FILTER_DROP_INFO,
2535};
2536const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2537
2538int register_netdevice_notifier(struct notifier_block *nb);
2539int unregister_netdevice_notifier(struct notifier_block *nb);
2540int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2541int unregister_netdevice_notifier_net(struct net *net,
2542 struct notifier_block *nb);
2543int register_netdevice_notifier_dev_net(struct net_device *dev,
2544 struct notifier_block *nb,
2545 struct netdev_net_notifier *nn);
2546int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2547 struct notifier_block *nb,
2548 struct netdev_net_notifier *nn);
2549
2550struct netdev_notifier_info {
2551 struct net_device *dev;
2552 struct netlink_ext_ack *extack;
2553};
2554
2555struct netdev_notifier_info_ext {
2556 struct netdev_notifier_info info; /* must be first */
2557 union {
2558 u32 mtu;
2559 } ext;
2560};
2561
2562struct netdev_notifier_change_info {
2563 struct netdev_notifier_info info; /* must be first */
2564 unsigned int flags_changed;
2565};
2566
2567struct netdev_notifier_changeupper_info {
2568 struct netdev_notifier_info info; /* must be first */
2569 struct net_device *upper_dev; /* new upper dev */
2570 bool master; /* is upper dev master */
2571 bool linking; /* is the notification for link or unlink */
2572 void *upper_info; /* upper dev info */
2573};
2574
2575struct netdev_notifier_changelowerstate_info {
2576 struct netdev_notifier_info info; /* must be first */
2577 void *lower_state_info; /* is lower dev state */
2578};
2579
2580struct netdev_notifier_pre_changeaddr_info {
2581 struct netdev_notifier_info info; /* must be first */
2582 const unsigned char *dev_addr;
2583};
2584
2585static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2586 struct net_device *dev)
2587{
2588 info->dev = dev;
2589 info->extack = NULL;
2590}
2591
2592static inline struct net_device *
2593netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2594{
2595 return info->dev;
2596}
2597
2598static inline struct netlink_ext_ack *
2599netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2600{
2601 return info->extack;
2602}
2603
2604int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2605
2606
2607extern rwlock_t dev_base_lock; /* Device list lock */
2608
2609#define for_each_netdev(net, d) \
2610 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2611#define for_each_netdev_reverse(net, d) \
2612 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2613#define for_each_netdev_rcu(net, d) \
2614 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2615#define for_each_netdev_safe(net, d, n) \
2616 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2617#define for_each_netdev_continue(net, d) \
2618 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2619#define for_each_netdev_continue_reverse(net, d) \
2620 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2621 dev_list)
2622#define for_each_netdev_continue_rcu(net, d) \
2623 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2624#define for_each_netdev_in_bond_rcu(bond, slave) \
2625 for_each_netdev_rcu(&init_net, slave) \
2626 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2627#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2628
2629static inline struct net_device *next_net_device(struct net_device *dev)
2630{
2631 struct list_head *lh;
2632 struct net *net;
2633
2634 net = dev_net(dev);
2635 lh = dev->dev_list.next;
2636 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2637}
2638
2639static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2640{
2641 struct list_head *lh;
2642 struct net *net;
2643
2644 net = dev_net(dev);
2645 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2646 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2647}
2648
2649static inline struct net_device *first_net_device(struct net *net)
2650{
2651 return list_empty(&net->dev_base_head) ? NULL :
2652 net_device_entry(net->dev_base_head.next);
2653}
2654
2655static inline struct net_device *first_net_device_rcu(struct net *net)
2656{
2657 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2658
2659 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2660}
2661
2662int netdev_boot_setup_check(struct net_device *dev);
2663unsigned long netdev_boot_base(const char *prefix, int unit);
2664struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2665 const char *hwaddr);
2666struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2667struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2668void dev_add_pack(struct packet_type *pt);
2669void dev_remove_pack(struct packet_type *pt);
2670void __dev_remove_pack(struct packet_type *pt);
2671void dev_add_offload(struct packet_offload *po);
2672void dev_remove_offload(struct packet_offload *po);
2673
2674int dev_get_iflink(const struct net_device *dev);
2675int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2676struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2677 unsigned short mask);
2678struct net_device *dev_get_by_name(struct net *net, const char *name);
2679struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2680struct net_device *__dev_get_by_name(struct net *net, const char *name);
2681int dev_alloc_name(struct net_device *dev, const char *name);
2682int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2683void dev_close(struct net_device *dev);
2684void dev_close_many(struct list_head *head, bool unlink);
2685void dev_disable_lro(struct net_device *dev);
2686int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2687u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2688 struct net_device *sb_dev);
2689u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2690 struct net_device *sb_dev);
2691int dev_queue_xmit(struct sk_buff *skb);
2692int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2693int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2694int register_netdevice(struct net_device *dev);
2695void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2696void unregister_netdevice_many(struct list_head *head);
2697static inline void unregister_netdevice(struct net_device *dev)
2698{
2699 unregister_netdevice_queue(dev, NULL);
2700}
2701
2702int netdev_refcnt_read(const struct net_device *dev);
2703void free_netdev(struct net_device *dev);
2704void netdev_freemem(struct net_device *dev);
2705void synchronize_net(void);
2706int init_dummy_netdev(struct net_device *dev);
2707
2708struct net_device *dev_get_by_index(struct net *net, int ifindex);
2709struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2710struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2711struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2712int netdev_get_name(struct net *net, char *name, int ifindex);
2713int dev_restart(struct net_device *dev);
2714int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2715int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2716
2717static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2718{
2719 return NAPI_GRO_CB(skb)->data_offset;
2720}
2721
2722static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2723{
2724 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2725}
2726
2727static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2728{
2729 NAPI_GRO_CB(skb)->data_offset += len;
2730}
2731
2732static inline void *skb_gro_header_fast(struct sk_buff *skb,
2733 unsigned int offset)
2734{
2735 return NAPI_GRO_CB(skb)->frag0 + offset;
2736}
2737
2738static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2739{
2740 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2741}
2742
2743static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2744{
2745 NAPI_GRO_CB(skb)->frag0 = NULL;
2746 NAPI_GRO_CB(skb)->frag0_len = 0;
2747}
2748
2749static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2750 unsigned int offset)
2751{
2752 if (!pskb_may_pull(skb, hlen))
2753 return NULL;
2754
2755 skb_gro_frag0_invalidate(skb);
2756 return skb->data + offset;
2757}
2758
2759static inline void *skb_gro_network_header(struct sk_buff *skb)
2760{
2761 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2762 skb_network_offset(skb);
2763}
2764
2765static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2766 const void *start, unsigned int len)
2767{
2768 if (NAPI_GRO_CB(skb)->csum_valid)
2769 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2770 csum_partial(start, len, 0));
2771}
2772
2773/* GRO checksum functions. These are logical equivalents of the normal
2774 * checksum functions (in skbuff.h) except that they operate on the GRO
2775 * offsets and fields in sk_buff.
2776 */
2777
2778__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2779
2780static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2781{
2782 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2783}
2784
2785static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2786 bool zero_okay,
2787 __sum16 check)
2788{
2789 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2790 skb_checksum_start_offset(skb) <
2791 skb_gro_offset(skb)) &&
2792 !skb_at_gro_remcsum_start(skb) &&
2793 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2794 (!zero_okay || check));
2795}
2796
2797static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2798 __wsum psum)
2799{
2800 if (NAPI_GRO_CB(skb)->csum_valid &&
2801 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2802 return 0;
2803
2804 NAPI_GRO_CB(skb)->csum = psum;
2805
2806 return __skb_gro_checksum_complete(skb);
2807}
2808
2809static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2810{
2811 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2812 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2813 NAPI_GRO_CB(skb)->csum_cnt--;
2814 } else {
2815 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2816 * verified a new top level checksum or an encapsulated one
2817 * during GRO. This saves work if we fallback to normal path.
2818 */
2819 __skb_incr_checksum_unnecessary(skb);
2820 }
2821}
2822
2823#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2824 compute_pseudo) \
2825({ \
2826 __sum16 __ret = 0; \
2827 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2828 __ret = __skb_gro_checksum_validate_complete(skb, \
2829 compute_pseudo(skb, proto)); \
2830 if (!__ret) \
2831 skb_gro_incr_csum_unnecessary(skb); \
2832 __ret; \
2833})
2834
2835#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2836 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2837
2838#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2839 compute_pseudo) \
2840 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2841
2842#define skb_gro_checksum_simple_validate(skb) \
2843 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2844
2845static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2846{
2847 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2848 !NAPI_GRO_CB(skb)->csum_valid);
2849}
2850
2851static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2852 __wsum pseudo)
2853{
2854 NAPI_GRO_CB(skb)->csum = ~pseudo;
2855 NAPI_GRO_CB(skb)->csum_valid = 1;
2856}
2857
2858#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
2859do { \
2860 if (__skb_gro_checksum_convert_check(skb)) \
2861 __skb_gro_checksum_convert(skb, \
2862 compute_pseudo(skb, proto)); \
2863} while (0)
2864
2865struct gro_remcsum {
2866 int offset;
2867 __wsum delta;
2868};
2869
2870static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2871{
2872 grc->offset = 0;
2873 grc->delta = 0;
2874}
2875
2876static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2877 unsigned int off, size_t hdrlen,
2878 int start, int offset,
2879 struct gro_remcsum *grc,
2880 bool nopartial)
2881{
2882 __wsum delta;
2883 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2884
2885 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2886
2887 if (!nopartial) {
2888 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2889 return ptr;
2890 }
2891
2892 ptr = skb_gro_header_fast(skb, off);
2893 if (skb_gro_header_hard(skb, off + plen)) {
2894 ptr = skb_gro_header_slow(skb, off + plen, off);
2895 if (!ptr)
2896 return NULL;
2897 }
2898
2899 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2900 start, offset);
2901
2902 /* Adjust skb->csum since we changed the packet */
2903 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2904
2905 grc->offset = off + hdrlen + offset;
2906 grc->delta = delta;
2907
2908 return ptr;
2909}
2910
2911static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2912 struct gro_remcsum *grc)
2913{
2914 void *ptr;
2915 size_t plen = grc->offset + sizeof(u16);
2916
2917 if (!grc->delta)
2918 return;
2919
2920 ptr = skb_gro_header_fast(skb, grc->offset);
2921 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2922 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2923 if (!ptr)
2924 return;
2925 }
2926
2927 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2928}
2929
2930#ifdef CONFIG_XFRM_OFFLOAD
2931static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2932{
2933 if (PTR_ERR(pp) != -EINPROGRESS)
2934 NAPI_GRO_CB(skb)->flush |= flush;
2935}
2936static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2937 struct sk_buff *pp,
2938 int flush,
2939 struct gro_remcsum *grc)
2940{
2941 if (PTR_ERR(pp) != -EINPROGRESS) {
2942 NAPI_GRO_CB(skb)->flush |= flush;
2943 skb_gro_remcsum_cleanup(skb, grc);
2944 skb->remcsum_offload = 0;
2945 }
2946}
2947#else
2948static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2949{
2950 NAPI_GRO_CB(skb)->flush |= flush;
2951}
2952static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2953 struct sk_buff *pp,
2954 int flush,
2955 struct gro_remcsum *grc)
2956{
2957 NAPI_GRO_CB(skb)->flush |= flush;
2958 skb_gro_remcsum_cleanup(skb, grc);
2959 skb->remcsum_offload = 0;
2960}
2961#endif
2962
2963static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2964 unsigned short type,
2965 const void *daddr, const void *saddr,
2966 unsigned int len)
2967{
2968 if (!dev->header_ops || !dev->header_ops->create)
2969 return 0;
2970
2971 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2972}
2973
2974static inline int dev_parse_header(const struct sk_buff *skb,
2975 unsigned char *haddr)
2976{
2977 const struct net_device *dev = skb->dev;
2978
2979 if (!dev->header_ops || !dev->header_ops->parse)
2980 return 0;
2981 return dev->header_ops->parse(skb, haddr);
2982}
2983
2984static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2985{
2986 const struct net_device *dev = skb->dev;
2987
2988 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2989 return 0;
2990 return dev->header_ops->parse_protocol(skb);
2991}
2992
2993/* ll_header must have at least hard_header_len allocated */
2994static inline bool dev_validate_header(const struct net_device *dev,
2995 char *ll_header, int len)
2996{
2997 if (likely(len >= dev->hard_header_len))
2998 return true;
2999 if (len < dev->min_header_len)
3000 return false;
3001
3002 if (capable(CAP_SYS_RAWIO)) {
3003 memset(ll_header + len, 0, dev->hard_header_len - len);
3004 return true;
3005 }
3006
3007 if (dev->header_ops && dev->header_ops->validate)
3008 return dev->header_ops->validate(ll_header, len);
3009
3010 return false;
3011}
3012
3013typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3014 int len, int size);
3015int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3016static inline int unregister_gifconf(unsigned int family)
3017{
3018 return register_gifconf(family, NULL);
3019}
3020
3021#ifdef CONFIG_NET_FLOW_LIMIT
3022#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3023struct sd_flow_limit {
3024 u64 count;
3025 unsigned int num_buckets;
3026 unsigned int history_head;
3027 u16 history[FLOW_LIMIT_HISTORY];
3028 u8 buckets[];
3029};
3030
3031extern int netdev_flow_limit_table_len;
3032#endif /* CONFIG_NET_FLOW_LIMIT */
3033
3034/*
3035 * Incoming packets are placed on per-CPU queues
3036 */
3037struct softnet_data {
3038 struct list_head poll_list;
3039 struct sk_buff_head process_queue;
3040
3041 /* stats */
3042 unsigned int processed;
3043 unsigned int time_squeeze;
3044 unsigned int received_rps;
3045#ifdef CONFIG_RPS
3046 struct softnet_data *rps_ipi_list;
3047#endif
3048#ifdef CONFIG_NET_FLOW_LIMIT
3049 struct sd_flow_limit __rcu *flow_limit;
3050#endif
3051 struct Qdisc *output_queue;
3052 struct Qdisc **output_queue_tailp;
3053 struct sk_buff *completion_queue;
3054#ifdef CONFIG_XFRM_OFFLOAD
3055 struct sk_buff_head xfrm_backlog;
3056#endif
3057 /* written and read only by owning cpu: */
3058 struct {
3059 u16 recursion;
3060 u8 more;
3061 } xmit;
3062#ifdef CONFIG_RPS
3063 /* input_queue_head should be written by cpu owning this struct,
3064 * and only read by other cpus. Worth using a cache line.
3065 */
3066 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3067
3068 /* Elements below can be accessed between CPUs for RPS/RFS */
3069 call_single_data_t csd ____cacheline_aligned_in_smp;
3070 struct softnet_data *rps_ipi_next;
3071 unsigned int cpu;
3072 unsigned int input_queue_tail;
3073#endif
3074 unsigned int dropped;
3075 struct sk_buff_head input_pkt_queue;
3076 struct napi_struct backlog;
3077
3078};
3079
3080static inline void input_queue_head_incr(struct softnet_data *sd)
3081{
3082#ifdef CONFIG_RPS
3083 sd->input_queue_head++;
3084#endif
3085}
3086
3087static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3088 unsigned int *qtail)
3089{
3090#ifdef CONFIG_RPS
3091 *qtail = ++sd->input_queue_tail;
3092#endif
3093}
3094
3095DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3096
3097static inline int dev_recursion_level(void)
3098{
3099 return this_cpu_read(softnet_data.xmit.recursion);
3100}
3101
3102#define XMIT_RECURSION_LIMIT 10
3103static inline bool dev_xmit_recursion(void)
3104{
3105 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3106 XMIT_RECURSION_LIMIT);
3107}
3108
3109static inline void dev_xmit_recursion_inc(void)
3110{
3111 __this_cpu_inc(softnet_data.xmit.recursion);
3112}
3113
3114static inline void dev_xmit_recursion_dec(void)
3115{
3116 __this_cpu_dec(softnet_data.xmit.recursion);
3117}
3118
3119void __netif_schedule(struct Qdisc *q);
3120void netif_schedule_queue(struct netdev_queue *txq);
3121
3122static inline void netif_tx_schedule_all(struct net_device *dev)
3123{
3124 unsigned int i;
3125
3126 for (i = 0; i < dev->num_tx_queues; i++)
3127 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3128}
3129
3130static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3131{
3132 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3133}
3134
3135/**
3136 * netif_start_queue - allow transmit
3137 * @dev: network device
3138 *
3139 * Allow upper layers to call the device hard_start_xmit routine.
3140 */
3141static inline void netif_start_queue(struct net_device *dev)
3142{
3143 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3144}
3145
3146static inline void netif_tx_start_all_queues(struct net_device *dev)
3147{
3148 unsigned int i;
3149
3150 for (i = 0; i < dev->num_tx_queues; i++) {
3151 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3152 netif_tx_start_queue(txq);
3153 }
3154}
3155
3156void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3157
3158/**
3159 * netif_wake_queue - restart transmit
3160 * @dev: network device
3161 *
3162 * Allow upper layers to call the device hard_start_xmit routine.
3163 * Used for flow control when transmit resources are available.
3164 */
3165static inline void netif_wake_queue(struct net_device *dev)
3166{
3167 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3168}
3169
3170static inline void netif_tx_wake_all_queues(struct net_device *dev)
3171{
3172 unsigned int i;
3173
3174 for (i = 0; i < dev->num_tx_queues; i++) {
3175 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3176 netif_tx_wake_queue(txq);
3177 }
3178}
3179
3180static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3181{
3182 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3183}
3184
3185/**
3186 * netif_stop_queue - stop transmitted packets
3187 * @dev: network device
3188 *
3189 * Stop upper layers calling the device hard_start_xmit routine.
3190 * Used for flow control when transmit resources are unavailable.
3191 */
3192static inline void netif_stop_queue(struct net_device *dev)
3193{
3194 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3195}
3196
3197void netif_tx_stop_all_queues(struct net_device *dev);
3198void netdev_update_lockdep_key(struct net_device *dev);
3199
3200static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3201{
3202 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3203}
3204
3205/**
3206 * netif_queue_stopped - test if transmit queue is flowblocked
3207 * @dev: network device
3208 *
3209 * Test if transmit queue on device is currently unable to send.
3210 */
3211static inline bool netif_queue_stopped(const struct net_device *dev)
3212{
3213 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3214}
3215
3216static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3217{
3218 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3219}
3220
3221static inline bool
3222netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3223{
3224 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3225}
3226
3227static inline bool
3228netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3229{
3230 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3231}
3232
3233/**
3234 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3235 * @dev_queue: pointer to transmit queue
3236 *
3237 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3238 * to give appropriate hint to the CPU.
3239 */
3240static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3241{
3242#ifdef CONFIG_BQL
3243 prefetchw(&dev_queue->dql.num_queued);
3244#endif
3245}
3246
3247/**
3248 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3249 * @dev_queue: pointer to transmit queue
3250 *
3251 * BQL enabled drivers might use this helper in their TX completion path,
3252 * to give appropriate hint to the CPU.
3253 */
3254static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3255{
3256#ifdef CONFIG_BQL
3257 prefetchw(&dev_queue->dql.limit);
3258#endif
3259}
3260
3261static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3262 unsigned int bytes)
3263{
3264#ifdef CONFIG_BQL
3265 dql_queued(&dev_queue->dql, bytes);
3266
3267 if (likely(dql_avail(&dev_queue->dql) >= 0))
3268 return;
3269
3270 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3271
3272 /*
3273 * The XOFF flag must be set before checking the dql_avail below,
3274 * because in netdev_tx_completed_queue we update the dql_completed
3275 * before checking the XOFF flag.
3276 */
3277 smp_mb();
3278
3279 /* check again in case another CPU has just made room avail */
3280 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3281 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3282#endif
3283}
3284
3285/* Variant of netdev_tx_sent_queue() for drivers that are aware
3286 * that they should not test BQL status themselves.
3287 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3288 * skb of a batch.
3289 * Returns true if the doorbell must be used to kick the NIC.
3290 */
3291static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3292 unsigned int bytes,
3293 bool xmit_more)
3294{
3295 if (xmit_more) {
3296#ifdef CONFIG_BQL
3297 dql_queued(&dev_queue->dql, bytes);
3298#endif
3299 return netif_tx_queue_stopped(dev_queue);
3300 }
3301 netdev_tx_sent_queue(dev_queue, bytes);
3302 return true;
3303}
3304
3305/**
3306 * netdev_sent_queue - report the number of bytes queued to hardware
3307 * @dev: network device
3308 * @bytes: number of bytes queued to the hardware device queue
3309 *
3310 * Report the number of bytes queued for sending/completion to the network
3311 * device hardware queue. @bytes should be a good approximation and should
3312 * exactly match netdev_completed_queue() @bytes
3313 */
3314static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3315{
3316 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3317}
3318
3319static inline bool __netdev_sent_queue(struct net_device *dev,
3320 unsigned int bytes,
3321 bool xmit_more)
3322{
3323 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3324 xmit_more);
3325}
3326
3327static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3328 unsigned int pkts, unsigned int bytes)
3329{
3330#ifdef CONFIG_BQL
3331 if (unlikely(!bytes))
3332 return;
3333
3334 dql_completed(&dev_queue->dql, bytes);
3335
3336 /*
3337 * Without the memory barrier there is a small possiblity that
3338 * netdev_tx_sent_queue will miss the update and cause the queue to
3339 * be stopped forever
3340 */
3341 smp_mb();
3342
3343 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3344 return;
3345
3346 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3347 netif_schedule_queue(dev_queue);
3348#endif
3349}
3350
3351/**
3352 * netdev_completed_queue - report bytes and packets completed by device
3353 * @dev: network device
3354 * @pkts: actual number of packets sent over the medium
3355 * @bytes: actual number of bytes sent over the medium
3356 *
3357 * Report the number of bytes and packets transmitted by the network device
3358 * hardware queue over the physical medium, @bytes must exactly match the
3359 * @bytes amount passed to netdev_sent_queue()
3360 */
3361static inline void netdev_completed_queue(struct net_device *dev,
3362 unsigned int pkts, unsigned int bytes)
3363{
3364 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3365}
3366
3367static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3368{
3369#ifdef CONFIG_BQL
3370 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3371 dql_reset(&q->dql);
3372#endif
3373}
3374
3375/**
3376 * netdev_reset_queue - reset the packets and bytes count of a network device
3377 * @dev_queue: network device
3378 *
3379 * Reset the bytes and packet count of a network device and clear the
3380 * software flow control OFF bit for this network device
3381 */
3382static inline void netdev_reset_queue(struct net_device *dev_queue)
3383{
3384 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3385}
3386
3387/**
3388 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3389 * @dev: network device
3390 * @queue_index: given tx queue index
3391 *
3392 * Returns 0 if given tx queue index >= number of device tx queues,
3393 * otherwise returns the originally passed tx queue index.
3394 */
3395static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3396{
3397 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3398 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3399 dev->name, queue_index,
3400 dev->real_num_tx_queues);
3401 return 0;
3402 }
3403
3404 return queue_index;
3405}
3406
3407/**
3408 * netif_running - test if up
3409 * @dev: network device
3410 *
3411 * Test if the device has been brought up.
3412 */
3413static inline bool netif_running(const struct net_device *dev)
3414{
3415 return test_bit(__LINK_STATE_START, &dev->state);
3416}
3417
3418/*
3419 * Routines to manage the subqueues on a device. We only need start,
3420 * stop, and a check if it's stopped. All other device management is
3421 * done at the overall netdevice level.
3422 * Also test the device if we're multiqueue.
3423 */
3424
3425/**
3426 * netif_start_subqueue - allow sending packets on subqueue
3427 * @dev: network device
3428 * @queue_index: sub queue index
3429 *
3430 * Start individual transmit queue of a device with multiple transmit queues.
3431 */
3432static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3433{
3434 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3435
3436 netif_tx_start_queue(txq);
3437}
3438
3439/**
3440 * netif_stop_subqueue - stop sending packets on subqueue
3441 * @dev: network device
3442 * @queue_index: sub queue index
3443 *
3444 * Stop individual transmit queue of a device with multiple transmit queues.
3445 */
3446static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3447{
3448 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3449 netif_tx_stop_queue(txq);
3450}
3451
3452/**
3453 * netif_subqueue_stopped - test status of subqueue
3454 * @dev: network device
3455 * @queue_index: sub queue index
3456 *
3457 * Check individual transmit queue of a device with multiple transmit queues.
3458 */
3459static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3460 u16 queue_index)
3461{
3462 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3463
3464 return netif_tx_queue_stopped(txq);
3465}
3466
3467static inline bool netif_subqueue_stopped(const struct net_device *dev,
3468 struct sk_buff *skb)
3469{
3470 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3471}
3472
3473/**
3474 * netif_wake_subqueue - allow sending packets on subqueue
3475 * @dev: network device
3476 * @queue_index: sub queue index
3477 *
3478 * Resume individual transmit queue of a device with multiple transmit queues.
3479 */
3480static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3481{
3482 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3483
3484 netif_tx_wake_queue(txq);
3485}
3486
3487#ifdef CONFIG_XPS
3488int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3489 u16 index);
3490int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3491 u16 index, bool is_rxqs_map);
3492
3493/**
3494 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3495 * @j: CPU/Rx queue index
3496 * @mask: bitmask of all cpus/rx queues
3497 * @nr_bits: number of bits in the bitmask
3498 *
3499 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3500 */
3501static inline bool netif_attr_test_mask(unsigned long j,
3502 const unsigned long *mask,
3503 unsigned int nr_bits)
3504{
3505 cpu_max_bits_warn(j, nr_bits);
3506 return test_bit(j, mask);
3507}
3508
3509/**
3510 * netif_attr_test_online - Test for online CPU/Rx queue
3511 * @j: CPU/Rx queue index
3512 * @online_mask: bitmask for CPUs/Rx queues that are online
3513 * @nr_bits: number of bits in the bitmask
3514 *
3515 * Returns true if a CPU/Rx queue is online.
3516 */
3517static inline bool netif_attr_test_online(unsigned long j,
3518 const unsigned long *online_mask,
3519 unsigned int nr_bits)
3520{
3521 cpu_max_bits_warn(j, nr_bits);
3522
3523 if (online_mask)
3524 return test_bit(j, online_mask);
3525
3526 return (j < nr_bits);
3527}
3528
3529/**
3530 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3531 * @n: CPU/Rx queue index
3532 * @srcp: the cpumask/Rx queue mask pointer
3533 * @nr_bits: number of bits in the bitmask
3534 *
3535 * Returns >= nr_bits if no further CPUs/Rx queues set.
3536 */
3537static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3538 unsigned int nr_bits)
3539{
3540 /* -1 is a legal arg here. */
3541 if (n != -1)
3542 cpu_max_bits_warn(n, nr_bits);
3543
3544 if (srcp)
3545 return find_next_bit(srcp, nr_bits, n + 1);
3546
3547 return n + 1;
3548}
3549
3550/**
3551 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3552 * @n: CPU/Rx queue index
3553 * @src1p: the first CPUs/Rx queues mask pointer
3554 * @src2p: the second CPUs/Rx queues mask pointer
3555 * @nr_bits: number of bits in the bitmask
3556 *
3557 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3558 */
3559static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3560 const unsigned long *src2p,
3561 unsigned int nr_bits)
3562{
3563 /* -1 is a legal arg here. */
3564 if (n != -1)
3565 cpu_max_bits_warn(n, nr_bits);
3566
3567 if (src1p && src2p)
3568 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3569 else if (src1p)
3570 return find_next_bit(src1p, nr_bits, n + 1);
3571 else if (src2p)
3572 return find_next_bit(src2p, nr_bits, n + 1);
3573
3574 return n + 1;
3575}
3576#else
3577static inline int netif_set_xps_queue(struct net_device *dev,
3578 const struct cpumask *mask,
3579 u16 index)
3580{
3581 return 0;
3582}
3583
3584static inline int __netif_set_xps_queue(struct net_device *dev,
3585 const unsigned long *mask,
3586 u16 index, bool is_rxqs_map)
3587{
3588 return 0;
3589}
3590#endif
3591
3592/**
3593 * netif_is_multiqueue - test if device has multiple transmit queues
3594 * @dev: network device
3595 *
3596 * Check if device has multiple transmit queues
3597 */
3598static inline bool netif_is_multiqueue(const struct net_device *dev)
3599{
3600 return dev->num_tx_queues > 1;
3601}
3602
3603int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3604
3605#ifdef CONFIG_SYSFS
3606int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3607#else
3608static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3609 unsigned int rxqs)
3610{
3611 dev->real_num_rx_queues = rxqs;
3612 return 0;
3613}
3614#endif
3615
3616static inline struct netdev_rx_queue *
3617__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3618{
3619 return dev->_rx + rxq;
3620}
3621
3622#ifdef CONFIG_SYSFS
3623static inline unsigned int get_netdev_rx_queue_index(
3624 struct netdev_rx_queue *queue)
3625{
3626 struct net_device *dev = queue->dev;
3627 int index = queue - dev->_rx;
3628
3629 BUG_ON(index >= dev->num_rx_queues);
3630 return index;
3631}
3632#endif
3633
3634#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3635int netif_get_num_default_rss_queues(void);
3636
3637enum skb_free_reason {
3638 SKB_REASON_CONSUMED,
3639 SKB_REASON_DROPPED,
3640};
3641
3642void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3643void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3644
3645/*
3646 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3647 * interrupt context or with hardware interrupts being disabled.
3648 * (in_irq() || irqs_disabled())
3649 *
3650 * We provide four helpers that can be used in following contexts :
3651 *
3652 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3653 * replacing kfree_skb(skb)
3654 *
3655 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3656 * Typically used in place of consume_skb(skb) in TX completion path
3657 *
3658 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3659 * replacing kfree_skb(skb)
3660 *
3661 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3662 * and consumed a packet. Used in place of consume_skb(skb)
3663 */
3664static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3665{
3666 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3667}
3668
3669static inline void dev_consume_skb_irq(struct sk_buff *skb)
3670{
3671 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3672}
3673
3674static inline void dev_kfree_skb_any(struct sk_buff *skb)
3675{
3676 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3677}
3678
3679static inline void dev_consume_skb_any(struct sk_buff *skb)
3680{
3681 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3682}
3683
3684void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3685int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3686int netif_rx(struct sk_buff *skb);
3687int netif_rx_ni(struct sk_buff *skb);
3688int netif_receive_skb(struct sk_buff *skb);
3689int netif_receive_skb_core(struct sk_buff *skb);
3690void netif_receive_skb_list(struct list_head *head);
3691gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3692void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3693struct sk_buff *napi_get_frags(struct napi_struct *napi);
3694gro_result_t napi_gro_frags(struct napi_struct *napi);
3695struct packet_offload *gro_find_receive_by_type(__be16 type);
3696struct packet_offload *gro_find_complete_by_type(__be16 type);
3697
3698static inline void napi_free_frags(struct napi_struct *napi)
3699{
3700 kfree_skb(napi->skb);
3701 napi->skb = NULL;
3702}
3703
3704bool netdev_is_rx_handler_busy(struct net_device *dev);
3705int netdev_rx_handler_register(struct net_device *dev,
3706 rx_handler_func_t *rx_handler,
3707 void *rx_handler_data);
3708void netdev_rx_handler_unregister(struct net_device *dev);
3709
3710bool dev_valid_name(const char *name);
3711int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3712 bool *need_copyout);
3713int dev_ifconf(struct net *net, struct ifconf *, int);
3714int dev_ethtool(struct net *net, struct ifreq *);
3715unsigned int dev_get_flags(const struct net_device *);
3716int __dev_change_flags(struct net_device *dev, unsigned int flags,
3717 struct netlink_ext_ack *extack);
3718int dev_change_flags(struct net_device *dev, unsigned int flags,
3719 struct netlink_ext_ack *extack);
3720void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3721 unsigned int gchanges);
3722int dev_change_name(struct net_device *, const char *);
3723int dev_set_alias(struct net_device *, const char *, size_t);
3724int dev_get_alias(const struct net_device *, char *, size_t);
3725int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3726int __dev_set_mtu(struct net_device *, int);
3727int dev_validate_mtu(struct net_device *dev, int mtu,
3728 struct netlink_ext_ack *extack);
3729int dev_set_mtu_ext(struct net_device *dev, int mtu,
3730 struct netlink_ext_ack *extack);
3731int dev_set_mtu(struct net_device *, int);
3732int dev_change_tx_queue_len(struct net_device *, unsigned long);
3733void dev_set_group(struct net_device *, int);
3734int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3735 struct netlink_ext_ack *extack);
3736int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3737 struct netlink_ext_ack *extack);
3738int dev_change_carrier(struct net_device *, bool new_carrier);
3739int dev_get_phys_port_id(struct net_device *dev,
3740 struct netdev_phys_item_id *ppid);
3741int dev_get_phys_port_name(struct net_device *dev,
3742 char *name, size_t len);
3743int dev_get_port_parent_id(struct net_device *dev,
3744 struct netdev_phys_item_id *ppid, bool recurse);
3745bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3746int dev_change_proto_down(struct net_device *dev, bool proto_down);
3747int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3748struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3749struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3750 struct netdev_queue *txq, int *ret);
3751
3752typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3753int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3754 int fd, u32 flags);
3755u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3756 enum bpf_netdev_command cmd);
3757int xdp_umem_query(struct net_device *dev, u16 queue_id);
3758
3759int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3760int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3761bool is_skb_forwardable(const struct net_device *dev,
3762 const struct sk_buff *skb);
3763
3764static __always_inline int ____dev_forward_skb(struct net_device *dev,
3765 struct sk_buff *skb)
3766{
3767 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3768 unlikely(!is_skb_forwardable(dev, skb))) {
3769 atomic_long_inc(&dev->rx_dropped);
3770 kfree_skb(skb);
3771 return NET_RX_DROP;
3772 }
3773
3774 skb_scrub_packet(skb, true);
3775 skb->priority = 0;
3776 return 0;
3777}
3778
3779bool dev_nit_active(struct net_device *dev);
3780void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3781
3782extern int netdev_budget;
3783extern unsigned int netdev_budget_usecs;
3784
3785/* Called by rtnetlink.c:rtnl_unlock() */
3786void netdev_run_todo(void);
3787
3788/**
3789 * dev_put - release reference to device
3790 * @dev: network device
3791 *
3792 * Release reference to device to allow it to be freed.
3793 */
3794static inline void dev_put(struct net_device *dev)
3795{
3796 this_cpu_dec(*dev->pcpu_refcnt);
3797}
3798
3799/**
3800 * dev_hold - get reference to device
3801 * @dev: network device
3802 *
3803 * Hold reference to device to keep it from being freed.
3804 */
3805static inline void dev_hold(struct net_device *dev)
3806{
3807 this_cpu_inc(*dev->pcpu_refcnt);
3808}
3809
3810/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3811 * and _off may be called from IRQ context, but it is caller
3812 * who is responsible for serialization of these calls.
3813 *
3814 * The name carrier is inappropriate, these functions should really be
3815 * called netif_lowerlayer_*() because they represent the state of any
3816 * kind of lower layer not just hardware media.
3817 */
3818
3819void linkwatch_init_dev(struct net_device *dev);
3820void linkwatch_fire_event(struct net_device *dev);
3821void linkwatch_forget_dev(struct net_device *dev);
3822
3823/**
3824 * netif_carrier_ok - test if carrier present
3825 * @dev: network device
3826 *
3827 * Check if carrier is present on device
3828 */
3829static inline bool netif_carrier_ok(const struct net_device *dev)
3830{
3831 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3832}
3833
3834unsigned long dev_trans_start(struct net_device *dev);
3835
3836void __netdev_watchdog_up(struct net_device *dev);
3837
3838void netif_carrier_on(struct net_device *dev);
3839
3840void netif_carrier_off(struct net_device *dev);
3841
3842/**
3843 * netif_dormant_on - mark device as dormant.
3844 * @dev: network device
3845 *
3846 * Mark device as dormant (as per RFC2863).
3847 *
3848 * The dormant state indicates that the relevant interface is not
3849 * actually in a condition to pass packets (i.e., it is not 'up') but is
3850 * in a "pending" state, waiting for some external event. For "on-
3851 * demand" interfaces, this new state identifies the situation where the
3852 * interface is waiting for events to place it in the up state.
3853 */
3854static inline void netif_dormant_on(struct net_device *dev)
3855{
3856 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3857 linkwatch_fire_event(dev);
3858}
3859
3860/**
3861 * netif_dormant_off - set device as not dormant.
3862 * @dev: network device
3863 *
3864 * Device is not in dormant state.
3865 */
3866static inline void netif_dormant_off(struct net_device *dev)
3867{
3868 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3869 linkwatch_fire_event(dev);
3870}
3871
3872/**
3873 * netif_dormant - test if device is dormant
3874 * @dev: network device
3875 *
3876 * Check if device is dormant.
3877 */
3878static inline bool netif_dormant(const struct net_device *dev)
3879{
3880 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3881}
3882
3883
3884/**
3885 * netif_oper_up - test if device is operational
3886 * @dev: network device
3887 *
3888 * Check if carrier is operational
3889 */
3890static inline bool netif_oper_up(const struct net_device *dev)
3891{
3892 return (dev->operstate == IF_OPER_UP ||
3893 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3894}
3895
3896/**
3897 * netif_device_present - is device available or removed
3898 * @dev: network device
3899 *
3900 * Check if device has not been removed from system.
3901 */
3902static inline bool netif_device_present(struct net_device *dev)
3903{
3904 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3905}
3906
3907void netif_device_detach(struct net_device *dev);
3908
3909void netif_device_attach(struct net_device *dev);
3910
3911/*
3912 * Network interface message level settings
3913 */
3914
3915enum {
3916 NETIF_MSG_DRV_BIT,
3917 NETIF_MSG_PROBE_BIT,
3918 NETIF_MSG_LINK_BIT,
3919 NETIF_MSG_TIMER_BIT,
3920 NETIF_MSG_IFDOWN_BIT,
3921 NETIF_MSG_IFUP_BIT,
3922 NETIF_MSG_RX_ERR_BIT,
3923 NETIF_MSG_TX_ERR_BIT,
3924 NETIF_MSG_TX_QUEUED_BIT,
3925 NETIF_MSG_INTR_BIT,
3926 NETIF_MSG_TX_DONE_BIT,
3927 NETIF_MSG_RX_STATUS_BIT,
3928 NETIF_MSG_PKTDATA_BIT,
3929 NETIF_MSG_HW_BIT,
3930 NETIF_MSG_WOL_BIT,
3931
3932 /* When you add a new bit above, update netif_msg_class_names array
3933 * in net/ethtool/common.c
3934 */
3935 NETIF_MSG_CLASS_COUNT,
3936};
3937/* Both ethtool_ops interface and internal driver implementation use u32 */
3938static_assert(NETIF_MSG_CLASS_COUNT <= 32);
3939
3940#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
3941#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
3942
3943#define NETIF_MSG_DRV __NETIF_MSG(DRV)
3944#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
3945#define NETIF_MSG_LINK __NETIF_MSG(LINK)
3946#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
3947#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
3948#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
3949#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
3950#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
3951#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
3952#define NETIF_MSG_INTR __NETIF_MSG(INTR)
3953#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
3954#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
3955#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
3956#define NETIF_MSG_HW __NETIF_MSG(HW)
3957#define NETIF_MSG_WOL __NETIF_MSG(WOL)
3958
3959#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3960#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3961#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3962#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3963#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3964#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3965#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3966#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3967#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3968#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3969#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3970#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3971#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3972#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3973#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3974
3975static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3976{
3977 /* use default */
3978 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3979 return default_msg_enable_bits;
3980 if (debug_value == 0) /* no output */
3981 return 0;
3982 /* set low N bits */
3983 return (1U << debug_value) - 1;
3984}
3985
3986static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3987{
3988 spin_lock(&txq->_xmit_lock);
3989 txq->xmit_lock_owner = cpu;
3990}
3991
3992static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3993{
3994 __acquire(&txq->_xmit_lock);
3995 return true;
3996}
3997
3998static inline void __netif_tx_release(struct netdev_queue *txq)
3999{
4000 __release(&txq->_xmit_lock);
4001}
4002
4003static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4004{
4005 spin_lock_bh(&txq->_xmit_lock);
4006 txq->xmit_lock_owner = smp_processor_id();
4007}
4008
4009static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4010{
4011 bool ok = spin_trylock(&txq->_xmit_lock);
4012 if (likely(ok))
4013 txq->xmit_lock_owner = smp_processor_id();
4014 return ok;
4015}
4016
4017static inline void __netif_tx_unlock(struct netdev_queue *txq)
4018{
4019 txq->xmit_lock_owner = -1;
4020 spin_unlock(&txq->_xmit_lock);
4021}
4022
4023static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4024{
4025 txq->xmit_lock_owner = -1;
4026 spin_unlock_bh(&txq->_xmit_lock);
4027}
4028
4029static inline void txq_trans_update(struct netdev_queue *txq)
4030{
4031 if (txq->xmit_lock_owner != -1)
4032 txq->trans_start = jiffies;
4033}
4034
4035/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4036static inline void netif_trans_update(struct net_device *dev)
4037{
4038 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4039
4040 if (txq->trans_start != jiffies)
4041 txq->trans_start = jiffies;
4042}
4043
4044/**
4045 * netif_tx_lock - grab network device transmit lock
4046 * @dev: network device
4047 *
4048 * Get network device transmit lock
4049 */
4050static inline void netif_tx_lock(struct net_device *dev)
4051{
4052 unsigned int i;
4053 int cpu;
4054
4055 spin_lock(&dev->tx_global_lock);
4056 cpu = smp_processor_id();
4057 for (i = 0; i < dev->num_tx_queues; i++) {
4058 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4059
4060 /* We are the only thread of execution doing a
4061 * freeze, but we have to grab the _xmit_lock in
4062 * order to synchronize with threads which are in
4063 * the ->hard_start_xmit() handler and already
4064 * checked the frozen bit.
4065 */
4066 __netif_tx_lock(txq, cpu);
4067 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4068 __netif_tx_unlock(txq);
4069 }
4070}
4071
4072static inline void netif_tx_lock_bh(struct net_device *dev)
4073{
4074 local_bh_disable();
4075 netif_tx_lock(dev);
4076}
4077
4078static inline void netif_tx_unlock(struct net_device *dev)
4079{
4080 unsigned int i;
4081
4082 for (i = 0; i < dev->num_tx_queues; i++) {
4083 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4084
4085 /* No need to grab the _xmit_lock here. If the
4086 * queue is not stopped for another reason, we
4087 * force a schedule.
4088 */
4089 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4090 netif_schedule_queue(txq);
4091 }
4092 spin_unlock(&dev->tx_global_lock);
4093}
4094
4095static inline void netif_tx_unlock_bh(struct net_device *dev)
4096{
4097 netif_tx_unlock(dev);
4098 local_bh_enable();
4099}
4100
4101#define HARD_TX_LOCK(dev, txq, cpu) { \
4102 if ((dev->features & NETIF_F_LLTX) == 0) { \
4103 __netif_tx_lock(txq, cpu); \
4104 } else { \
4105 __netif_tx_acquire(txq); \
4106 } \
4107}
4108
4109#define HARD_TX_TRYLOCK(dev, txq) \
4110 (((dev->features & NETIF_F_LLTX) == 0) ? \
4111 __netif_tx_trylock(txq) : \
4112 __netif_tx_acquire(txq))
4113
4114#define HARD_TX_UNLOCK(dev, txq) { \
4115 if ((dev->features & NETIF_F_LLTX) == 0) { \
4116 __netif_tx_unlock(txq); \
4117 } else { \
4118 __netif_tx_release(txq); \
4119 } \
4120}
4121
4122static inline void netif_tx_disable(struct net_device *dev)
4123{
4124 unsigned int i;
4125 int cpu;
4126
4127 local_bh_disable();
4128 cpu = smp_processor_id();
4129 for (i = 0; i < dev->num_tx_queues; i++) {
4130 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4131
4132 __netif_tx_lock(txq, cpu);
4133 netif_tx_stop_queue(txq);
4134 __netif_tx_unlock(txq);
4135 }
4136 local_bh_enable();
4137}
4138
4139static inline void netif_addr_lock(struct net_device *dev)
4140{
4141 spin_lock(&dev->addr_list_lock);
4142}
4143
4144static inline void netif_addr_lock_bh(struct net_device *dev)
4145{
4146 spin_lock_bh(&dev->addr_list_lock);
4147}
4148
4149static inline void netif_addr_unlock(struct net_device *dev)
4150{
4151 spin_unlock(&dev->addr_list_lock);
4152}
4153
4154static inline void netif_addr_unlock_bh(struct net_device *dev)
4155{
4156 spin_unlock_bh(&dev->addr_list_lock);
4157}
4158
4159/*
4160 * dev_addrs walker. Should be used only for read access. Call with
4161 * rcu_read_lock held.
4162 */
4163#define for_each_dev_addr(dev, ha) \
4164 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4165
4166/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4167
4168void ether_setup(struct net_device *dev);
4169
4170/* Support for loadable net-drivers */
4171struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4172 unsigned char name_assign_type,
4173 void (*setup)(struct net_device *),
4174 unsigned int txqs, unsigned int rxqs);
4175#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4176 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4177
4178#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4179 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4180 count)
4181
4182int register_netdev(struct net_device *dev);
4183void unregister_netdev(struct net_device *dev);
4184
4185/* General hardware address lists handling functions */
4186int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4187 struct netdev_hw_addr_list *from_list, int addr_len);
4188void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4189 struct netdev_hw_addr_list *from_list, int addr_len);
4190int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4191 struct net_device *dev,
4192 int (*sync)(struct net_device *, const unsigned char *),
4193 int (*unsync)(struct net_device *,
4194 const unsigned char *));
4195int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4196 struct net_device *dev,
4197 int (*sync)(struct net_device *,
4198 const unsigned char *, int),
4199 int (*unsync)(struct net_device *,
4200 const unsigned char *, int));
4201void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4202 struct net_device *dev,
4203 int (*unsync)(struct net_device *,
4204 const unsigned char *, int));
4205void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4206 struct net_device *dev,
4207 int (*unsync)(struct net_device *,
4208 const unsigned char *));
4209void __hw_addr_init(struct netdev_hw_addr_list *list);
4210
4211/* Functions used for device addresses handling */
4212int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4213 unsigned char addr_type);
4214int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4215 unsigned char addr_type);
4216void dev_addr_flush(struct net_device *dev);
4217int dev_addr_init(struct net_device *dev);
4218
4219/* Functions used for unicast addresses handling */
4220int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4221int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4222int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4223int dev_uc_sync(struct net_device *to, struct net_device *from);
4224int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4225void dev_uc_unsync(struct net_device *to, struct net_device *from);
4226void dev_uc_flush(struct net_device *dev);
4227void dev_uc_init(struct net_device *dev);
4228
4229/**
4230 * __dev_uc_sync - Synchonize device's unicast list
4231 * @dev: device to sync
4232 * @sync: function to call if address should be added
4233 * @unsync: function to call if address should be removed
4234 *
4235 * Add newly added addresses to the interface, and release
4236 * addresses that have been deleted.
4237 */
4238static inline int __dev_uc_sync(struct net_device *dev,
4239 int (*sync)(struct net_device *,
4240 const unsigned char *),
4241 int (*unsync)(struct net_device *,
4242 const unsigned char *))
4243{
4244 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4245}
4246
4247/**
4248 * __dev_uc_unsync - Remove synchronized addresses from device
4249 * @dev: device to sync
4250 * @unsync: function to call if address should be removed
4251 *
4252 * Remove all addresses that were added to the device by dev_uc_sync().
4253 */
4254static inline void __dev_uc_unsync(struct net_device *dev,
4255 int (*unsync)(struct net_device *,
4256 const unsigned char *))
4257{
4258 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4259}
4260
4261/* Functions used for multicast addresses handling */
4262int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4263int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4264int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4265int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4266int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4267int dev_mc_sync(struct net_device *to, struct net_device *from);
4268int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4269void dev_mc_unsync(struct net_device *to, struct net_device *from);
4270void dev_mc_flush(struct net_device *dev);
4271void dev_mc_init(struct net_device *dev);
4272
4273/**
4274 * __dev_mc_sync - Synchonize device's multicast list
4275 * @dev: device to sync
4276 * @sync: function to call if address should be added
4277 * @unsync: function to call if address should be removed
4278 *
4279 * Add newly added addresses to the interface, and release
4280 * addresses that have been deleted.
4281 */
4282static inline int __dev_mc_sync(struct net_device *dev,
4283 int (*sync)(struct net_device *,
4284 const unsigned char *),
4285 int (*unsync)(struct net_device *,
4286 const unsigned char *))
4287{
4288 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4289}
4290
4291/**
4292 * __dev_mc_unsync - Remove synchronized addresses from device
4293 * @dev: device to sync
4294 * @unsync: function to call if address should be removed
4295 *
4296 * Remove all addresses that were added to the device by dev_mc_sync().
4297 */
4298static inline void __dev_mc_unsync(struct net_device *dev,
4299 int (*unsync)(struct net_device *,
4300 const unsigned char *))
4301{
4302 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4303}
4304
4305/* Functions used for secondary unicast and multicast support */
4306void dev_set_rx_mode(struct net_device *dev);
4307void __dev_set_rx_mode(struct net_device *dev);
4308int dev_set_promiscuity(struct net_device *dev, int inc);
4309int dev_set_allmulti(struct net_device *dev, int inc);
4310void netdev_state_change(struct net_device *dev);
4311void netdev_notify_peers(struct net_device *dev);
4312void netdev_features_change(struct net_device *dev);
4313/* Load a device via the kmod */
4314void dev_load(struct net *net, const char *name);
4315struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4316 struct rtnl_link_stats64 *storage);
4317void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4318 const struct net_device_stats *netdev_stats);
4319
4320extern int netdev_max_backlog;
4321extern int netdev_tstamp_prequeue;
4322extern int weight_p;
4323extern int dev_weight_rx_bias;
4324extern int dev_weight_tx_bias;
4325extern int dev_rx_weight;
4326extern int dev_tx_weight;
4327extern int gro_normal_batch;
4328
4329bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4330struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4331 struct list_head **iter);
4332struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4333 struct list_head **iter);
4334
4335/* iterate through upper list, must be called under RCU read lock */
4336#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4337 for (iter = &(dev)->adj_list.upper, \
4338 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4339 updev; \
4340 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4341
4342int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4343 int (*fn)(struct net_device *upper_dev,
4344 void *data),
4345 void *data);
4346
4347bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4348 struct net_device *upper_dev);
4349
4350bool netdev_has_any_upper_dev(struct net_device *dev);
4351
4352void *netdev_lower_get_next_private(struct net_device *dev,
4353 struct list_head **iter);
4354void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4355 struct list_head **iter);
4356
4357#define netdev_for_each_lower_private(dev, priv, iter) \
4358 for (iter = (dev)->adj_list.lower.next, \
4359 priv = netdev_lower_get_next_private(dev, &(iter)); \
4360 priv; \
4361 priv = netdev_lower_get_next_private(dev, &(iter)))
4362
4363#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4364 for (iter = &(dev)->adj_list.lower, \
4365 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4366 priv; \
4367 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4368
4369void *netdev_lower_get_next(struct net_device *dev,
4370 struct list_head **iter);
4371
4372#define netdev_for_each_lower_dev(dev, ldev, iter) \
4373 for (iter = (dev)->adj_list.lower.next, \
4374 ldev = netdev_lower_get_next(dev, &(iter)); \
4375 ldev; \
4376 ldev = netdev_lower_get_next(dev, &(iter)))
4377
4378struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4379 struct list_head **iter);
4380struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4381 struct list_head **iter);
4382
4383int netdev_walk_all_lower_dev(struct net_device *dev,
4384 int (*fn)(struct net_device *lower_dev,
4385 void *data),
4386 void *data);
4387int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4388 int (*fn)(struct net_device *lower_dev,
4389 void *data),
4390 void *data);
4391
4392void *netdev_adjacent_get_private(struct list_head *adj_list);
4393void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4394struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4395struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4396int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4397 struct netlink_ext_ack *extack);
4398int netdev_master_upper_dev_link(struct net_device *dev,
4399 struct net_device *upper_dev,
4400 void *upper_priv, void *upper_info,
4401 struct netlink_ext_ack *extack);
4402void netdev_upper_dev_unlink(struct net_device *dev,
4403 struct net_device *upper_dev);
4404int netdev_adjacent_change_prepare(struct net_device *old_dev,
4405 struct net_device *new_dev,
4406 struct net_device *dev,
4407 struct netlink_ext_ack *extack);
4408void netdev_adjacent_change_commit(struct net_device *old_dev,
4409 struct net_device *new_dev,
4410 struct net_device *dev);
4411void netdev_adjacent_change_abort(struct net_device *old_dev,
4412 struct net_device *new_dev,
4413 struct net_device *dev);
4414void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4415void *netdev_lower_dev_get_private(struct net_device *dev,
4416 struct net_device *lower_dev);
4417void netdev_lower_state_changed(struct net_device *lower_dev,
4418 void *lower_state_info);
4419
4420/* RSS keys are 40 or 52 bytes long */
4421#define NETDEV_RSS_KEY_LEN 52
4422extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4423void netdev_rss_key_fill(void *buffer, size_t len);
4424
4425int skb_checksum_help(struct sk_buff *skb);
4426int skb_crc32c_csum_help(struct sk_buff *skb);
4427int skb_csum_hwoffload_help(struct sk_buff *skb,
4428 const netdev_features_t features);
4429
4430struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4431 netdev_features_t features, bool tx_path);
4432struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4433 netdev_features_t features);
4434
4435struct netdev_bonding_info {
4436 ifslave slave;
4437 ifbond master;
4438};
4439
4440struct netdev_notifier_bonding_info {
4441 struct netdev_notifier_info info; /* must be first */
4442 struct netdev_bonding_info bonding_info;
4443};
4444
4445void netdev_bonding_info_change(struct net_device *dev,
4446 struct netdev_bonding_info *bonding_info);
4447
4448#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4449void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4450#else
4451static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4452 const void *data)
4453{
4454}
4455#endif
4456
4457static inline
4458struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4459{
4460 return __skb_gso_segment(skb, features, true);
4461}
4462__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4463
4464static inline bool can_checksum_protocol(netdev_features_t features,
4465 __be16 protocol)
4466{
4467 if (protocol == htons(ETH_P_FCOE))
4468 return !!(features & NETIF_F_FCOE_CRC);
4469
4470 /* Assume this is an IP checksum (not SCTP CRC) */
4471
4472 if (features & NETIF_F_HW_CSUM) {
4473 /* Can checksum everything */
4474 return true;
4475 }
4476
4477 switch (protocol) {
4478 case htons(ETH_P_IP):
4479 return !!(features & NETIF_F_IP_CSUM);
4480 case htons(ETH_P_IPV6):
4481 return !!(features & NETIF_F_IPV6_CSUM);
4482 default:
4483 return false;
4484 }
4485}
4486
4487#ifdef CONFIG_BUG
4488void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4489#else
4490static inline void netdev_rx_csum_fault(struct net_device *dev,
4491 struct sk_buff *skb)
4492{
4493}
4494#endif
4495/* rx skb timestamps */
4496void net_enable_timestamp(void);
4497void net_disable_timestamp(void);
4498
4499#ifdef CONFIG_PROC_FS
4500int __init dev_proc_init(void);
4501#else
4502#define dev_proc_init() 0
4503#endif
4504
4505static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4506 struct sk_buff *skb, struct net_device *dev,
4507 bool more)
4508{
4509 __this_cpu_write(softnet_data.xmit.more, more);
4510 return ops->ndo_start_xmit(skb, dev);
4511}
4512
4513static inline bool netdev_xmit_more(void)
4514{
4515 return __this_cpu_read(softnet_data.xmit.more);
4516}
4517
4518static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4519 struct netdev_queue *txq, bool more)
4520{
4521 const struct net_device_ops *ops = dev->netdev_ops;
4522 netdev_tx_t rc;
4523
4524 rc = __netdev_start_xmit(ops, skb, dev, more);
4525 if (rc == NETDEV_TX_OK)
4526 txq_trans_update(txq);
4527
4528 return rc;
4529}
4530
4531int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4532 const void *ns);
4533void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4534 const void *ns);
4535
4536static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4537{
4538 return netdev_class_create_file_ns(class_attr, NULL);
4539}
4540
4541static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4542{
4543 netdev_class_remove_file_ns(class_attr, NULL);
4544}
4545
4546extern const struct kobj_ns_type_operations net_ns_type_operations;
4547
4548const char *netdev_drivername(const struct net_device *dev);
4549
4550void linkwatch_run_queue(void);
4551
4552static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4553 netdev_features_t f2)
4554{
4555 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4556 if (f1 & NETIF_F_HW_CSUM)
4557 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4558 else
4559 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4560 }
4561
4562 return f1 & f2;
4563}
4564
4565static inline netdev_features_t netdev_get_wanted_features(
4566 struct net_device *dev)
4567{
4568 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4569}
4570netdev_features_t netdev_increment_features(netdev_features_t all,
4571 netdev_features_t one, netdev_features_t mask);
4572
4573/* Allow TSO being used on stacked device :
4574 * Performing the GSO segmentation before last device
4575 * is a performance improvement.
4576 */
4577static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4578 netdev_features_t mask)
4579{
4580 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4581}
4582
4583int __netdev_update_features(struct net_device *dev);
4584void netdev_update_features(struct net_device *dev);
4585void netdev_change_features(struct net_device *dev);
4586
4587void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4588 struct net_device *dev);
4589
4590netdev_features_t passthru_features_check(struct sk_buff *skb,
4591 struct net_device *dev,
4592 netdev_features_t features);
4593netdev_features_t netif_skb_features(struct sk_buff *skb);
4594
4595static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4596{
4597 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4598
4599 /* check flags correspondence */
4600 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4601 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4602 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4603 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4604 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4605 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4606 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4607 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4608 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4609 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4610 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4611 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4612 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4613 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4614 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4615 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4616 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4617 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4618 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4619
4620 return (features & feature) == feature;
4621}
4622
4623static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4624{
4625 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4626 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4627}
4628
4629static inline bool netif_needs_gso(struct sk_buff *skb,
4630 netdev_features_t features)
4631{
4632 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4633 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4634 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4635}
4636
4637static inline void netif_set_gso_max_size(struct net_device *dev,
4638 unsigned int size)
4639{
4640 dev->gso_max_size = size;
4641}
4642
4643static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4644 int pulled_hlen, u16 mac_offset,
4645 int mac_len)
4646{
4647 skb->protocol = protocol;
4648 skb->encapsulation = 1;
4649 skb_push(skb, pulled_hlen);
4650 skb_reset_transport_header(skb);
4651 skb->mac_header = mac_offset;
4652 skb->network_header = skb->mac_header + mac_len;
4653 skb->mac_len = mac_len;
4654}
4655
4656static inline bool netif_is_macsec(const struct net_device *dev)
4657{
4658 return dev->priv_flags & IFF_MACSEC;
4659}
4660
4661static inline bool netif_is_macvlan(const struct net_device *dev)
4662{
4663 return dev->priv_flags & IFF_MACVLAN;
4664}
4665
4666static inline bool netif_is_macvlan_port(const struct net_device *dev)
4667{
4668 return dev->priv_flags & IFF_MACVLAN_PORT;
4669}
4670
4671static inline bool netif_is_bond_master(const struct net_device *dev)
4672{
4673 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4674}
4675
4676static inline bool netif_is_bond_slave(const struct net_device *dev)
4677{
4678 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4679}
4680
4681static inline bool netif_supports_nofcs(struct net_device *dev)
4682{
4683 return dev->priv_flags & IFF_SUPP_NOFCS;
4684}
4685
4686static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4687{
4688 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4689}
4690
4691static inline bool netif_is_l3_master(const struct net_device *dev)
4692{
4693 return dev->priv_flags & IFF_L3MDEV_MASTER;
4694}
4695
4696static inline bool netif_is_l3_slave(const struct net_device *dev)
4697{
4698 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4699}
4700
4701static inline bool netif_is_bridge_master(const struct net_device *dev)
4702{
4703 return dev->priv_flags & IFF_EBRIDGE;
4704}
4705
4706static inline bool netif_is_bridge_port(const struct net_device *dev)
4707{
4708 return dev->priv_flags & IFF_BRIDGE_PORT;
4709}
4710
4711static inline bool netif_is_ovs_master(const struct net_device *dev)
4712{
4713 return dev->priv_flags & IFF_OPENVSWITCH;
4714}
4715
4716static inline bool netif_is_ovs_port(const struct net_device *dev)
4717{
4718 return dev->priv_flags & IFF_OVS_DATAPATH;
4719}
4720
4721static inline bool netif_is_team_master(const struct net_device *dev)
4722{
4723 return dev->priv_flags & IFF_TEAM;
4724}
4725
4726static inline bool netif_is_team_port(const struct net_device *dev)
4727{
4728 return dev->priv_flags & IFF_TEAM_PORT;
4729}
4730
4731static inline bool netif_is_lag_master(const struct net_device *dev)
4732{
4733 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4734}
4735
4736static inline bool netif_is_lag_port(const struct net_device *dev)
4737{
4738 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4739}
4740
4741static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4742{
4743 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4744}
4745
4746static inline bool netif_is_failover(const struct net_device *dev)
4747{
4748 return dev->priv_flags & IFF_FAILOVER;
4749}
4750
4751static inline bool netif_is_failover_slave(const struct net_device *dev)
4752{
4753 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4754}
4755
4756/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4757static inline void netif_keep_dst(struct net_device *dev)
4758{
4759 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4760}
4761
4762/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4763static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4764{
4765 /* TODO: reserve and use an additional IFF bit, if we get more users */
4766 return dev->priv_flags & IFF_MACSEC;
4767}
4768
4769extern struct pernet_operations __net_initdata loopback_net_ops;
4770
4771/* Logging, debugging and troubleshooting/diagnostic helpers. */
4772
4773/* netdev_printk helpers, similar to dev_printk */
4774
4775static inline const char *netdev_name(const struct net_device *dev)
4776{
4777 if (!dev->name[0] || strchr(dev->name, '%'))
4778 return "(unnamed net_device)";
4779 return dev->name;
4780}
4781
4782static inline bool netdev_unregistering(const struct net_device *dev)
4783{
4784 return dev->reg_state == NETREG_UNREGISTERING;
4785}
4786
4787static inline const char *netdev_reg_state(const struct net_device *dev)
4788{
4789 switch (dev->reg_state) {
4790 case NETREG_UNINITIALIZED: return " (uninitialized)";
4791 case NETREG_REGISTERED: return "";
4792 case NETREG_UNREGISTERING: return " (unregistering)";
4793 case NETREG_UNREGISTERED: return " (unregistered)";
4794 case NETREG_RELEASED: return " (released)";
4795 case NETREG_DUMMY: return " (dummy)";
4796 }
4797
4798 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4799 return " (unknown)";
4800}
4801
4802__printf(3, 4) __cold
4803void netdev_printk(const char *level, const struct net_device *dev,
4804 const char *format, ...);
4805__printf(2, 3) __cold
4806void netdev_emerg(const struct net_device *dev, const char *format, ...);
4807__printf(2, 3) __cold
4808void netdev_alert(const struct net_device *dev, const char *format, ...);
4809__printf(2, 3) __cold
4810void netdev_crit(const struct net_device *dev, const char *format, ...);
4811__printf(2, 3) __cold
4812void netdev_err(const struct net_device *dev, const char *format, ...);
4813__printf(2, 3) __cold
4814void netdev_warn(const struct net_device *dev, const char *format, ...);
4815__printf(2, 3) __cold
4816void netdev_notice(const struct net_device *dev, const char *format, ...);
4817__printf(2, 3) __cold
4818void netdev_info(const struct net_device *dev, const char *format, ...);
4819
4820#define netdev_level_once(level, dev, fmt, ...) \
4821do { \
4822 static bool __print_once __read_mostly; \
4823 \
4824 if (!__print_once) { \
4825 __print_once = true; \
4826 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4827 } \
4828} while (0)
4829
4830#define netdev_emerg_once(dev, fmt, ...) \
4831 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4832#define netdev_alert_once(dev, fmt, ...) \
4833 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4834#define netdev_crit_once(dev, fmt, ...) \
4835 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4836#define netdev_err_once(dev, fmt, ...) \
4837 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4838#define netdev_warn_once(dev, fmt, ...) \
4839 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4840#define netdev_notice_once(dev, fmt, ...) \
4841 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4842#define netdev_info_once(dev, fmt, ...) \
4843 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4844
4845#define MODULE_ALIAS_NETDEV(device) \
4846 MODULE_ALIAS("netdev-" device)
4847
4848#if defined(CONFIG_DYNAMIC_DEBUG)
4849#define netdev_dbg(__dev, format, args...) \
4850do { \
4851 dynamic_netdev_dbg(__dev, format, ##args); \
4852} while (0)
4853#elif defined(DEBUG)
4854#define netdev_dbg(__dev, format, args...) \
4855 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4856#else
4857#define netdev_dbg(__dev, format, args...) \
4858({ \
4859 if (0) \
4860 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4861})
4862#endif
4863
4864#if defined(VERBOSE_DEBUG)
4865#define netdev_vdbg netdev_dbg
4866#else
4867
4868#define netdev_vdbg(dev, format, args...) \
4869({ \
4870 if (0) \
4871 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4872 0; \
4873})
4874#endif
4875
4876/*
4877 * netdev_WARN() acts like dev_printk(), but with the key difference
4878 * of using a WARN/WARN_ON to get the message out, including the
4879 * file/line information and a backtrace.
4880 */
4881#define netdev_WARN(dev, format, args...) \
4882 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4883 netdev_reg_state(dev), ##args)
4884
4885#define netdev_WARN_ONCE(dev, format, args...) \
4886 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4887 netdev_reg_state(dev), ##args)
4888
4889/* netif printk helpers, similar to netdev_printk */
4890
4891#define netif_printk(priv, type, level, dev, fmt, args...) \
4892do { \
4893 if (netif_msg_##type(priv)) \
4894 netdev_printk(level, (dev), fmt, ##args); \
4895} while (0)
4896
4897#define netif_level(level, priv, type, dev, fmt, args...) \
4898do { \
4899 if (netif_msg_##type(priv)) \
4900 netdev_##level(dev, fmt, ##args); \
4901} while (0)
4902
4903#define netif_emerg(priv, type, dev, fmt, args...) \
4904 netif_level(emerg, priv, type, dev, fmt, ##args)
4905#define netif_alert(priv, type, dev, fmt, args...) \
4906 netif_level(alert, priv, type, dev, fmt, ##args)
4907#define netif_crit(priv, type, dev, fmt, args...) \
4908 netif_level(crit, priv, type, dev, fmt, ##args)
4909#define netif_err(priv, type, dev, fmt, args...) \
4910 netif_level(err, priv, type, dev, fmt, ##args)
4911#define netif_warn(priv, type, dev, fmt, args...) \
4912 netif_level(warn, priv, type, dev, fmt, ##args)
4913#define netif_notice(priv, type, dev, fmt, args...) \
4914 netif_level(notice, priv, type, dev, fmt, ##args)
4915#define netif_info(priv, type, dev, fmt, args...) \
4916 netif_level(info, priv, type, dev, fmt, ##args)
4917
4918#if defined(CONFIG_DYNAMIC_DEBUG)
4919#define netif_dbg(priv, type, netdev, format, args...) \
4920do { \
4921 if (netif_msg_##type(priv)) \
4922 dynamic_netdev_dbg(netdev, format, ##args); \
4923} while (0)
4924#elif defined(DEBUG)
4925#define netif_dbg(priv, type, dev, format, args...) \
4926 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4927#else
4928#define netif_dbg(priv, type, dev, format, args...) \
4929({ \
4930 if (0) \
4931 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4932 0; \
4933})
4934#endif
4935
4936/* if @cond then downgrade to debug, else print at @level */
4937#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4938 do { \
4939 if (cond) \
4940 netif_dbg(priv, type, netdev, fmt, ##args); \
4941 else \
4942 netif_ ## level(priv, type, netdev, fmt, ##args); \
4943 } while (0)
4944
4945#if defined(VERBOSE_DEBUG)
4946#define netif_vdbg netif_dbg
4947#else
4948#define netif_vdbg(priv, type, dev, format, args...) \
4949({ \
4950 if (0) \
4951 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4952 0; \
4953})
4954#endif
4955
4956/*
4957 * The list of packet types we will receive (as opposed to discard)
4958 * and the routines to invoke.
4959 *
4960 * Why 16. Because with 16 the only overlap we get on a hash of the
4961 * low nibble of the protocol value is RARP/SNAP/X.25.
4962 *
4963 * 0800 IP
4964 * 0001 802.3
4965 * 0002 AX.25
4966 * 0004 802.2
4967 * 8035 RARP
4968 * 0005 SNAP
4969 * 0805 X.25
4970 * 0806 ARP
4971 * 8137 IPX
4972 * 0009 Localtalk
4973 * 86DD IPv6
4974 */
4975#define PTYPE_HASH_SIZE (16)
4976#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4977
4978extern struct net_device *blackhole_netdev;
4979
4980#endif /* _LINUX_NETDEVICE_H */