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 <net/net_namespace.h>
38#ifdef CONFIG_DCB
39#include <net/dcbnl.h>
40#endif
41#include <net/netprio_cgroup.h>
42#include <net/xdp.h>
43
44#include <linux/netdev_features.h>
45#include <linux/neighbour.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <linux/hashtable.h>
50
51struct netpoll_info;
52struct device;
53struct ethtool_ops;
54struct phy_device;
55struct dsa_port;
56struct ip_tunnel_parm;
57struct macsec_context;
58struct macsec_ops;
59
60struct sfp_bus;
61/* 802.11 specific */
62struct wireless_dev;
63/* 802.15.4 specific */
64struct wpan_dev;
65struct mpls_dev;
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
68struct udp_tunnel_nic_info;
69struct udp_tunnel_nic;
70struct bpf_prog;
71struct xdp_buff;
72
73void synchronize_net(void);
74void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
76
77/* Backlog congestion levels */
78#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79#define NET_RX_DROP 1 /* packet dropped */
80
81#define MAX_NEST_DEV 8
82
83/*
84 * Transmit return codes: transmit return codes originate from three different
85 * namespaces:
86 *
87 * - qdisc return codes
88 * - driver transmit return codes
89 * - errno values
90 *
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
98 */
99
100/* qdisc ->enqueue() return codes. */
101#define NET_XMIT_SUCCESS 0x00
102#define NET_XMIT_DROP 0x01 /* skb dropped */
103#define NET_XMIT_CN 0x02 /* congestion notification */
104#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
105
106/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
111
112/* Driver transmit return codes */
113#define NETDEV_TX_MASK 0xf0
114
115enum netdev_tx {
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
119};
120typedef enum netdev_tx netdev_tx_t;
121
122/*
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
125 */
126static inline bool dev_xmit_complete(int rc)
127{
128 /*
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
133 */
134 if (likely(rc < NET_XMIT_MASK))
135 return true;
136
137 return false;
138}
139
140/*
141 * Compute the worst-case header length according to the protocols
142 * used.
143 */
144
145#if defined(CONFIG_HYPERV_NET)
146# define LL_MAX_HEADER 128
147#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148# if defined(CONFIG_MAC80211_MESH)
149# define LL_MAX_HEADER 128
150# else
151# define LL_MAX_HEADER 96
152# endif
153#else
154# define LL_MAX_HEADER 32
155#endif
156
157#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159#define MAX_HEADER LL_MAX_HEADER
160#else
161#define MAX_HEADER (LL_MAX_HEADER + 48)
162#endif
163
164/*
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
167 */
168
169struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
193};
194
195
196#include <linux/cache.h>
197#include <linux/skbuff.h>
198
199#ifdef CONFIG_RPS
200#include <linux/static_key.h>
201extern struct static_key_false rps_needed;
202extern struct static_key_false rfs_needed;
203#endif
204
205struct neighbour;
206struct neigh_parms;
207struct sk_buff;
208
209struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
212 unsigned char type;
213#define NETDEV_HW_ADDR_T_LAN 1
214#define NETDEV_HW_ADDR_T_SAN 2
215#define NETDEV_HW_ADDR_T_UNICAST 3
216#define NETDEV_HW_ADDR_T_MULTICAST 4
217 bool global_use;
218 int sync_cnt;
219 int refcount;
220 int synced;
221 struct rcu_head rcu_head;
222};
223
224struct netdev_hw_addr_list {
225 struct list_head list;
226 int count;
227};
228
229#define netdev_hw_addr_list_count(l) ((l)->count)
230#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231#define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
233
234#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236#define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
238
239#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241#define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243
244struct hh_cache {
245 unsigned int hh_len;
246 seqlock_t hh_lock;
247
248 /* cached hardware header; allow for machine alignment needs. */
249#define HH_DATA_MOD 16
250#define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252#define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
255};
256
257/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
258 * Alternative is:
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
261 *
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
264 */
265#define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
269
270struct header_ops {
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
281};
282
283/* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
285 * code.
286 */
287
288enum netdev_state_t {
289 __LINK_STATE_START,
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
295};
296
297
298/*
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
301 */
302struct netdev_boot_setup {
303 char name[IFNAMSIZ];
304 struct ifmap map;
305};
306#define NETDEV_BOOT_SETUP_MAX 8
307
308int __init netdev_boot_setup(char *str);
309
310struct gro_list {
311 struct list_head list;
312 int count;
313};
314
315/*
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
318 */
319#define GRO_HASH_BUCKETS 8
320
321/*
322 * Structure for NAPI scheduling similar to tasklet but with weighting
323 */
324struct napi_struct {
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
330 */
331 struct list_head poll_list;
332
333 unsigned long state;
334 int weight;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338#ifdef CONFIG_NETPOLL
339 int poll_owner;
340#endif
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
343 struct sk_buff *skb;
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
350 struct task_struct *thread;
351};
352
353enum {
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_MISSED, /* reschedule a napi */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
358 NAPI_STATE_LISTED, /* NAPI added to system lists */
359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
363 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
364};
365
366enum {
367 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
368 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
369 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
370 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
371 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
372 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
373 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
374 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
375 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
376 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
377};
378
379enum gro_result {
380 GRO_MERGED,
381 GRO_MERGED_FREE,
382 GRO_HELD,
383 GRO_NORMAL,
384 GRO_CONSUMED,
385};
386typedef enum gro_result gro_result_t;
387
388/*
389 * enum rx_handler_result - Possible return values for rx_handlers.
390 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
391 * further.
392 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
393 * case skb->dev was changed by rx_handler.
394 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
395 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
396 *
397 * rx_handlers are functions called from inside __netif_receive_skb(), to do
398 * special processing of the skb, prior to delivery to protocol handlers.
399 *
400 * Currently, a net_device can only have a single rx_handler registered. Trying
401 * to register a second rx_handler will return -EBUSY.
402 *
403 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
404 * To unregister a rx_handler on a net_device, use
405 * netdev_rx_handler_unregister().
406 *
407 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
408 * do with the skb.
409 *
410 * If the rx_handler consumed the skb in some way, it should return
411 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
412 * the skb to be delivered in some other way.
413 *
414 * If the rx_handler changed skb->dev, to divert the skb to another
415 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
416 * new device will be called if it exists.
417 *
418 * If the rx_handler decides the skb should be ignored, it should return
419 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
420 * are registered on exact device (ptype->dev == skb->dev).
421 *
422 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
423 * delivered, it should return RX_HANDLER_PASS.
424 *
425 * A device without a registered rx_handler will behave as if rx_handler
426 * returned RX_HANDLER_PASS.
427 */
428
429enum rx_handler_result {
430 RX_HANDLER_CONSUMED,
431 RX_HANDLER_ANOTHER,
432 RX_HANDLER_EXACT,
433 RX_HANDLER_PASS,
434};
435typedef enum rx_handler_result rx_handler_result_t;
436typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
437
438void __napi_schedule(struct napi_struct *n);
439void __napi_schedule_irqoff(struct napi_struct *n);
440
441static inline bool napi_disable_pending(struct napi_struct *n)
442{
443 return test_bit(NAPI_STATE_DISABLE, &n->state);
444}
445
446static inline bool napi_prefer_busy_poll(struct napi_struct *n)
447{
448 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
449}
450
451bool napi_schedule_prep(struct napi_struct *n);
452
453/**
454 * napi_schedule - schedule NAPI poll
455 * @n: NAPI context
456 *
457 * Schedule NAPI poll routine to be called if it is not already
458 * running.
459 */
460static inline void napi_schedule(struct napi_struct *n)
461{
462 if (napi_schedule_prep(n))
463 __napi_schedule(n);
464}
465
466/**
467 * napi_schedule_irqoff - schedule NAPI poll
468 * @n: NAPI context
469 *
470 * Variant of napi_schedule(), assuming hard irqs are masked.
471 */
472static inline void napi_schedule_irqoff(struct napi_struct *n)
473{
474 if (napi_schedule_prep(n))
475 __napi_schedule_irqoff(n);
476}
477
478/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
479static inline bool napi_reschedule(struct napi_struct *napi)
480{
481 if (napi_schedule_prep(napi)) {
482 __napi_schedule(napi);
483 return true;
484 }
485 return false;
486}
487
488bool napi_complete_done(struct napi_struct *n, int work_done);
489/**
490 * napi_complete - NAPI processing complete
491 * @n: NAPI context
492 *
493 * Mark NAPI processing as complete.
494 * Consider using napi_complete_done() instead.
495 * Return false if device should avoid rearming interrupts.
496 */
497static inline bool napi_complete(struct napi_struct *n)
498{
499 return napi_complete_done(n, 0);
500}
501
502int dev_set_threaded(struct net_device *dev, bool threaded);
503
504/**
505 * napi_disable - prevent NAPI from scheduling
506 * @n: NAPI context
507 *
508 * Stop NAPI from being scheduled on this context.
509 * Waits till any outstanding processing completes.
510 */
511void napi_disable(struct napi_struct *n);
512
513void napi_enable(struct napi_struct *n);
514
515/**
516 * napi_synchronize - wait until NAPI is not running
517 * @n: NAPI context
518 *
519 * Wait until NAPI is done being scheduled on this context.
520 * Waits till any outstanding processing completes but
521 * does not disable future activations.
522 */
523static inline void napi_synchronize(const struct napi_struct *n)
524{
525 if (IS_ENABLED(CONFIG_SMP))
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 msleep(1);
528 else
529 barrier();
530}
531
532/**
533 * napi_if_scheduled_mark_missed - if napi is running, set the
534 * NAPIF_STATE_MISSED
535 * @n: NAPI context
536 *
537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
538 * NAPI is scheduled.
539 **/
540static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
541{
542 unsigned long val, new;
543
544 do {
545 val = READ_ONCE(n->state);
546 if (val & NAPIF_STATE_DISABLE)
547 return true;
548
549 if (!(val & NAPIF_STATE_SCHED))
550 return false;
551
552 new = val | NAPIF_STATE_MISSED;
553 } while (cmpxchg(&n->state, val, new) != val);
554
555 return true;
556}
557
558enum netdev_queue_state_t {
559 __QUEUE_STATE_DRV_XOFF,
560 __QUEUE_STATE_STACK_XOFF,
561 __QUEUE_STATE_FROZEN,
562};
563
564#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
565#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
566#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
567
568#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
569#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
570 QUEUE_STATE_FROZEN)
571#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
572 QUEUE_STATE_FROZEN)
573
574/*
575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
576 * netif_tx_* functions below are used to manipulate this flag. The
577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
578 * queue independently. The netif_xmit_*stopped functions below are called
579 * to check if the queue has been stopped by the driver or stack (either
580 * of the XOFF bits are set in the state). Drivers should not need to call
581 * netif_xmit*stopped functions, they should only be using netif_tx_*.
582 */
583
584struct netdev_queue {
585/*
586 * read-mostly part
587 */
588 struct net_device *dev;
589 struct Qdisc __rcu *qdisc;
590 struct Qdisc *qdisc_sleeping;
591#ifdef CONFIG_SYSFS
592 struct kobject kobj;
593#endif
594#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
595 int numa_node;
596#endif
597 unsigned long tx_maxrate;
598 /*
599 * Number of TX timeouts for this queue
600 * (/sys/class/net/DEV/Q/trans_timeout)
601 */
602 unsigned long trans_timeout;
603
604 /* Subordinate device that the queue has been assigned to */
605 struct net_device *sb_dev;
606#ifdef CONFIG_XDP_SOCKETS
607 struct xsk_buff_pool *pool;
608#endif
609/*
610 * write-mostly part
611 */
612 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
613 int xmit_lock_owner;
614 /*
615 * Time (in jiffies) of last Tx
616 */
617 unsigned long trans_start;
618
619 unsigned long state;
620
621#ifdef CONFIG_BQL
622 struct dql dql;
623#endif
624} ____cacheline_aligned_in_smp;
625
626extern int sysctl_fb_tunnels_only_for_init_net;
627extern int sysctl_devconf_inherit_init_net;
628
629/*
630 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
631 * == 1 : For initns only
632 * == 2 : For none.
633 */
634static inline bool net_has_fallback_tunnels(const struct net *net)
635{
636 return !IS_ENABLED(CONFIG_SYSCTL) ||
637 !sysctl_fb_tunnels_only_for_init_net ||
638 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
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[];
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[];
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[] ____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 xsk_buff_pool *pool;
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[];
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[]; /* 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 TC_SETUP_QDISC_FIFO,
855 TC_SETUP_QDISC_HTB,
856};
857
858/* These structures hold the attributes of bpf state that are being passed
859 * to the netdevice through the bpf op.
860 */
861enum bpf_netdev_command {
862 /* Set or clear a bpf program used in the earliest stages of packet
863 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
864 * is responsible for calling bpf_prog_put on any old progs that are
865 * stored. In case of error, the callee need not release the new prog
866 * reference, but on success it takes ownership and must bpf_prog_put
867 * when it is no longer used.
868 */
869 XDP_SETUP_PROG,
870 XDP_SETUP_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_POOL,
875};
876
877struct bpf_prog_offload_ops;
878struct netlink_ext_ack;
879struct xdp_umem;
880struct xdp_dev_bulk_queue;
881struct bpf_xdp_link;
882
883enum bpf_xdp_mode {
884 XDP_MODE_SKB = 0,
885 XDP_MODE_DRV = 1,
886 XDP_MODE_HW = 2,
887 __MAX_XDP_MODE
888};
889
890struct bpf_xdp_entity {
891 struct bpf_prog *prog;
892 struct bpf_xdp_link *link;
893};
894
895struct netdev_bpf {
896 enum bpf_netdev_command command;
897 union {
898 /* XDP_SETUP_PROG */
899 struct {
900 u32 flags;
901 struct bpf_prog *prog;
902 struct netlink_ext_ack *extack;
903 };
904 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
905 struct {
906 struct bpf_offloaded_map *offmap;
907 };
908 /* XDP_SETUP_XSK_POOL */
909 struct {
910 struct xsk_buff_pool *pool;
911 u16 queue_id;
912 } xsk;
913 };
914};
915
916/* Flags for ndo_xsk_wakeup. */
917#define XDP_WAKEUP_RX (1 << 0)
918#define XDP_WAKEUP_TX (1 << 1)
919
920#ifdef CONFIG_XFRM_OFFLOAD
921struct xfrmdev_ops {
922 int (*xdo_dev_state_add) (struct xfrm_state *x);
923 void (*xdo_dev_state_delete) (struct xfrm_state *x);
924 void (*xdo_dev_state_free) (struct xfrm_state *x);
925 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
926 struct xfrm_state *x);
927 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
928};
929#endif
930
931struct dev_ifalias {
932 struct rcu_head rcuhead;
933 char ifalias[];
934};
935
936struct devlink;
937struct tlsdev_ops;
938
939struct netdev_name_node {
940 struct hlist_node hlist;
941 struct list_head list;
942 struct net_device *dev;
943 const char *name;
944};
945
946int netdev_name_node_alt_create(struct net_device *dev, const char *name);
947int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
948
949struct netdev_net_notifier {
950 struct list_head list;
951 struct notifier_block *nb;
952};
953
954/*
955 * This structure defines the management hooks for network devices.
956 * The following hooks can be defined; unless noted otherwise, they are
957 * optional and can be filled with a null pointer.
958 *
959 * int (*ndo_init)(struct net_device *dev);
960 * This function is called once when a network device is registered.
961 * The network device can use this for any late stage initialization
962 * or semantic validation. It can fail with an error code which will
963 * be propagated back to register_netdev.
964 *
965 * void (*ndo_uninit)(struct net_device *dev);
966 * This function is called when device is unregistered or when registration
967 * fails. It is not called if init fails.
968 *
969 * int (*ndo_open)(struct net_device *dev);
970 * This function is called when a network device transitions to the up
971 * state.
972 *
973 * int (*ndo_stop)(struct net_device *dev);
974 * This function is called when a network device transitions to the down
975 * state.
976 *
977 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
978 * struct net_device *dev);
979 * Called when a packet needs to be transmitted.
980 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
981 * the queue before that can happen; it's for obsolete devices and weird
982 * corner cases, but the stack really does a non-trivial amount
983 * of useless work if you return NETDEV_TX_BUSY.
984 * Required; cannot be NULL.
985 *
986 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
987 * struct net_device *dev
988 * netdev_features_t features);
989 * Called by core transmit path to determine if device is capable of
990 * performing offload operations on a given packet. This is to give
991 * the device an opportunity to implement any restrictions that cannot
992 * be otherwise expressed by feature flags. The check is called with
993 * the set of features that the stack has calculated and it returns
994 * those the driver believes to be appropriate.
995 *
996 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
997 * struct net_device *sb_dev);
998 * Called to decide which queue to use when device supports multiple
999 * transmit queues.
1000 *
1001 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1002 * This function is called to allow device receiver to make
1003 * changes to configuration when multicast or promiscuous is enabled.
1004 *
1005 * void (*ndo_set_rx_mode)(struct net_device *dev);
1006 * This function is called device changes address list filtering.
1007 * If driver handles unicast address filtering, it should set
1008 * IFF_UNICAST_FLT in its priv_flags.
1009 *
1010 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1011 * This function is called when the Media Access Control address
1012 * needs to be changed. If this interface is not defined, the
1013 * MAC address can not be changed.
1014 *
1015 * int (*ndo_validate_addr)(struct net_device *dev);
1016 * Test if Media Access Control address is valid for the device.
1017 *
1018 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1019 * Called when a user requests an ioctl which can't be handled by
1020 * the generic interface code. If not defined ioctls return
1021 * not supported error code.
1022 *
1023 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1024 * Used to set network devices bus interface parameters. This interface
1025 * is retained for legacy reasons; new devices should use the bus
1026 * interface (PCI) for low level management.
1027 *
1028 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1029 * Called when a user wants to change the Maximum Transfer Unit
1030 * of a device.
1031 *
1032 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1033 * Callback used when the transmitter has not made any progress
1034 * for dev->watchdog ticks.
1035 *
1036 * void (*ndo_get_stats64)(struct net_device *dev,
1037 * struct rtnl_link_stats64 *storage);
1038 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1039 * Called when a user wants to get the network device usage
1040 * statistics. Drivers must do one of the following:
1041 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1042 * rtnl_link_stats64 structure passed by the caller.
1043 * 2. Define @ndo_get_stats to update a net_device_stats structure
1044 * (which should normally be dev->stats) and return a pointer to
1045 * it. The structure may be changed asynchronously only if each
1046 * field is written atomically.
1047 * 3. Update dev->stats asynchronously and atomically, and define
1048 * neither operation.
1049 *
1050 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1051 * Return true if this device supports offload stats of this attr_id.
1052 *
1053 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1054 * void *attr_data)
1055 * Get statistics for offload operations by attr_id. Write it into the
1056 * attr_data pointer.
1057 *
1058 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1059 * If device supports VLAN filtering this function is called when a
1060 * VLAN id is registered.
1061 *
1062 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1063 * If device supports VLAN filtering this function is called when a
1064 * VLAN id is unregistered.
1065 *
1066 * void (*ndo_poll_controller)(struct net_device *dev);
1067 *
1068 * SR-IOV management functions.
1069 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1070 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1071 * u8 qos, __be16 proto);
1072 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1073 * int max_tx_rate);
1074 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1075 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1076 * int (*ndo_get_vf_config)(struct net_device *dev,
1077 * int vf, struct ifla_vf_info *ivf);
1078 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1079 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1080 * struct nlattr *port[]);
1081 *
1082 * Enable or disable the VF ability to query its RSS Redirection Table and
1083 * Hash Key. This is needed since on some devices VF share this information
1084 * with PF and querying it may introduce a theoretical security risk.
1085 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1086 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1087 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1088 * void *type_data);
1089 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1090 * This is always called from the stack with the rtnl lock held and netif
1091 * tx queues stopped. This allows the netdevice to perform queue
1092 * management safely.
1093 *
1094 * Fiber Channel over Ethernet (FCoE) offload functions.
1095 * int (*ndo_fcoe_enable)(struct net_device *dev);
1096 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1097 * so the underlying device can perform whatever needed configuration or
1098 * initialization to support acceleration of FCoE traffic.
1099 *
1100 * int (*ndo_fcoe_disable)(struct net_device *dev);
1101 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1102 * so the underlying device can perform whatever needed clean-ups to
1103 * stop supporting acceleration of FCoE traffic.
1104 *
1105 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1106 * struct scatterlist *sgl, unsigned int sgc);
1107 * Called when the FCoE Initiator wants to initialize an I/O that
1108 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1109 * perform necessary setup and returns 1 to indicate the device is set up
1110 * successfully to perform DDP on this I/O, otherwise this returns 0.
1111 *
1112 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1113 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1114 * indicated by the FC exchange id 'xid', so the underlying device can
1115 * clean up and reuse resources for later DDP requests.
1116 *
1117 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1118 * struct scatterlist *sgl, unsigned int sgc);
1119 * Called when the FCoE Target wants to initialize an I/O that
1120 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1121 * perform necessary setup and returns 1 to indicate the device is set up
1122 * successfully to perform DDP on this I/O, otherwise this returns 0.
1123 *
1124 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1125 * struct netdev_fcoe_hbainfo *hbainfo);
1126 * Called when the FCoE Protocol stack wants information on the underlying
1127 * device. This information is utilized by the FCoE protocol stack to
1128 * register attributes with Fiber Channel management service as per the
1129 * FC-GS Fabric Device Management Information(FDMI) specification.
1130 *
1131 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1132 * Called when the underlying device wants to override default World Wide
1133 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1134 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1135 * protocol stack to use.
1136 *
1137 * RFS acceleration.
1138 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1139 * u16 rxq_index, u32 flow_id);
1140 * Set hardware filter for RFS. rxq_index is the target queue index;
1141 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1142 * Return the filter ID on success, or a negative error code.
1143 *
1144 * Slave management functions (for bridge, bonding, etc).
1145 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1146 * Called to make another netdev an underling.
1147 *
1148 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1149 * Called to release previously enslaved netdev.
1150 *
1151 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1152 * struct sk_buff *skb,
1153 * bool all_slaves);
1154 * Get the xmit slave of master device. If all_slaves is true, function
1155 * assume all the slaves can transmit.
1156 *
1157 * Feature/offload setting functions.
1158 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1159 * netdev_features_t features);
1160 * Adjusts the requested feature flags according to device-specific
1161 * constraints, and returns the resulting flags. Must not modify
1162 * the device state.
1163 *
1164 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1165 * Called to update device configuration to new features. Passed
1166 * feature set might be less than what was returned by ndo_fix_features()).
1167 * Must return >0 or -errno if it changed dev->features itself.
1168 *
1169 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1170 * struct net_device *dev,
1171 * const unsigned char *addr, u16 vid, u16 flags,
1172 * struct netlink_ext_ack *extack);
1173 * Adds an FDB entry to dev for addr.
1174 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1175 * struct net_device *dev,
1176 * const unsigned char *addr, u16 vid)
1177 * Deletes the FDB entry from dev coresponding to addr.
1178 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1179 * struct net_device *dev, struct net_device *filter_dev,
1180 * int *idx)
1181 * Used to add FDB entries to dump requests. Implementers should add
1182 * entries to skb and update idx with the number of entries.
1183 *
1184 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1185 * u16 flags, struct netlink_ext_ack *extack)
1186 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1187 * struct net_device *dev, u32 filter_mask,
1188 * int nlflags)
1189 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1190 * u16 flags);
1191 *
1192 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1193 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1194 * which do not represent real hardware may define this to allow their
1195 * userspace components to manage their virtual carrier state. Devices
1196 * that determine carrier state from physical hardware properties (eg
1197 * network cables) or protocol-dependent mechanisms (eg
1198 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1199 *
1200 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1201 * struct netdev_phys_item_id *ppid);
1202 * Called to get ID of physical port of this device. If driver does
1203 * not implement this, it is assumed that the hw is not able to have
1204 * multiple net devices on single physical port.
1205 *
1206 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1207 * struct netdev_phys_item_id *ppid)
1208 * Called to get the parent ID of the physical port of this device.
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 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1265 * int cmd);
1266 * Add, change, delete or get information on an IPv4 tunnel.
1267 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1268 * If a device is paired with a peer device, return the peer instance.
1269 * The caller must be under RCU read context.
1270 */
1271struct net_device_ops {
1272 int (*ndo_init)(struct net_device *dev);
1273 void (*ndo_uninit)(struct net_device *dev);
1274 int (*ndo_open)(struct net_device *dev);
1275 int (*ndo_stop)(struct net_device *dev);
1276 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1277 struct net_device *dev);
1278 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1279 struct net_device *dev,
1280 netdev_features_t features);
1281 u16 (*ndo_select_queue)(struct net_device *dev,
1282 struct sk_buff *skb,
1283 struct net_device *sb_dev);
1284 void (*ndo_change_rx_flags)(struct net_device *dev,
1285 int flags);
1286 void (*ndo_set_rx_mode)(struct net_device *dev);
1287 int (*ndo_set_mac_address)(struct net_device *dev,
1288 void *addr);
1289 int (*ndo_validate_addr)(struct net_device *dev);
1290 int (*ndo_do_ioctl)(struct net_device *dev,
1291 struct ifreq *ifr, int cmd);
1292 int (*ndo_set_config)(struct net_device *dev,
1293 struct ifmap *map);
1294 int (*ndo_change_mtu)(struct net_device *dev,
1295 int new_mtu);
1296 int (*ndo_neigh_setup)(struct net_device *dev,
1297 struct neigh_parms *);
1298 void (*ndo_tx_timeout) (struct net_device *dev,
1299 unsigned int txqueue);
1300
1301 void (*ndo_get_stats64)(struct net_device *dev,
1302 struct rtnl_link_stats64 *storage);
1303 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1304 int (*ndo_get_offload_stats)(int attr_id,
1305 const struct net_device *dev,
1306 void *attr_data);
1307 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1308
1309 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1310 __be16 proto, u16 vid);
1311 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1312 __be16 proto, u16 vid);
1313#ifdef CONFIG_NET_POLL_CONTROLLER
1314 void (*ndo_poll_controller)(struct net_device *dev);
1315 int (*ndo_netpoll_setup)(struct net_device *dev,
1316 struct netpoll_info *info);
1317 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1318#endif
1319 int (*ndo_set_vf_mac)(struct net_device *dev,
1320 int queue, u8 *mac);
1321 int (*ndo_set_vf_vlan)(struct net_device *dev,
1322 int queue, u16 vlan,
1323 u8 qos, __be16 proto);
1324 int (*ndo_set_vf_rate)(struct net_device *dev,
1325 int vf, int min_tx_rate,
1326 int max_tx_rate);
1327 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1328 int vf, bool setting);
1329 int (*ndo_set_vf_trust)(struct net_device *dev,
1330 int vf, bool setting);
1331 int (*ndo_get_vf_config)(struct net_device *dev,
1332 int vf,
1333 struct ifla_vf_info *ivf);
1334 int (*ndo_set_vf_link_state)(struct net_device *dev,
1335 int vf, int link_state);
1336 int (*ndo_get_vf_stats)(struct net_device *dev,
1337 int vf,
1338 struct ifla_vf_stats
1339 *vf_stats);
1340 int (*ndo_set_vf_port)(struct net_device *dev,
1341 int vf,
1342 struct nlattr *port[]);
1343 int (*ndo_get_vf_port)(struct net_device *dev,
1344 int vf, struct sk_buff *skb);
1345 int (*ndo_get_vf_guid)(struct net_device *dev,
1346 int vf,
1347 struct ifla_vf_guid *node_guid,
1348 struct ifla_vf_guid *port_guid);
1349 int (*ndo_set_vf_guid)(struct net_device *dev,
1350 int vf, u64 guid,
1351 int guid_type);
1352 int (*ndo_set_vf_rss_query_en)(
1353 struct net_device *dev,
1354 int vf, bool setting);
1355 int (*ndo_setup_tc)(struct net_device *dev,
1356 enum tc_setup_type type,
1357 void *type_data);
1358#if IS_ENABLED(CONFIG_FCOE)
1359 int (*ndo_fcoe_enable)(struct net_device *dev);
1360 int (*ndo_fcoe_disable)(struct net_device *dev);
1361 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1362 u16 xid,
1363 struct scatterlist *sgl,
1364 unsigned int sgc);
1365 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1366 u16 xid);
1367 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1368 u16 xid,
1369 struct scatterlist *sgl,
1370 unsigned int sgc);
1371 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1372 struct netdev_fcoe_hbainfo *hbainfo);
1373#endif
1374
1375#if IS_ENABLED(CONFIG_LIBFCOE)
1376#define NETDEV_FCOE_WWNN 0
1377#define NETDEV_FCOE_WWPN 1
1378 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1379 u64 *wwn, int type);
1380#endif
1381
1382#ifdef CONFIG_RFS_ACCEL
1383 int (*ndo_rx_flow_steer)(struct net_device *dev,
1384 const struct sk_buff *skb,
1385 u16 rxq_index,
1386 u32 flow_id);
1387#endif
1388 int (*ndo_add_slave)(struct net_device *dev,
1389 struct net_device *slave_dev,
1390 struct netlink_ext_ack *extack);
1391 int (*ndo_del_slave)(struct net_device *dev,
1392 struct net_device *slave_dev);
1393 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1394 struct sk_buff *skb,
1395 bool all_slaves);
1396 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1397 struct sock *sk);
1398 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1399 netdev_features_t features);
1400 int (*ndo_set_features)(struct net_device *dev,
1401 netdev_features_t features);
1402 int (*ndo_neigh_construct)(struct net_device *dev,
1403 struct neighbour *n);
1404 void (*ndo_neigh_destroy)(struct net_device *dev,
1405 struct neighbour *n);
1406
1407 int (*ndo_fdb_add)(struct ndmsg *ndm,
1408 struct nlattr *tb[],
1409 struct net_device *dev,
1410 const unsigned char *addr,
1411 u16 vid,
1412 u16 flags,
1413 struct netlink_ext_ack *extack);
1414 int (*ndo_fdb_del)(struct ndmsg *ndm,
1415 struct nlattr *tb[],
1416 struct net_device *dev,
1417 const unsigned char *addr,
1418 u16 vid);
1419 int (*ndo_fdb_dump)(struct sk_buff *skb,
1420 struct netlink_callback *cb,
1421 struct net_device *dev,
1422 struct net_device *filter_dev,
1423 int *idx);
1424 int (*ndo_fdb_get)(struct sk_buff *skb,
1425 struct nlattr *tb[],
1426 struct net_device *dev,
1427 const unsigned char *addr,
1428 u16 vid, u32 portid, u32 seq,
1429 struct netlink_ext_ack *extack);
1430 int (*ndo_bridge_setlink)(struct net_device *dev,
1431 struct nlmsghdr *nlh,
1432 u16 flags,
1433 struct netlink_ext_ack *extack);
1434 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1435 u32 pid, u32 seq,
1436 struct net_device *dev,
1437 u32 filter_mask,
1438 int nlflags);
1439 int (*ndo_bridge_dellink)(struct net_device *dev,
1440 struct nlmsghdr *nlh,
1441 u16 flags);
1442 int (*ndo_change_carrier)(struct net_device *dev,
1443 bool new_carrier);
1444 int (*ndo_get_phys_port_id)(struct net_device *dev,
1445 struct netdev_phys_item_id *ppid);
1446 int (*ndo_get_port_parent_id)(struct net_device *dev,
1447 struct netdev_phys_item_id *ppid);
1448 int (*ndo_get_phys_port_name)(struct net_device *dev,
1449 char *name, size_t len);
1450 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1451 struct net_device *dev);
1452 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1453 void *priv);
1454
1455 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1456 int queue_index,
1457 u32 maxrate);
1458 int (*ndo_get_iflink)(const struct net_device *dev);
1459 int (*ndo_change_proto_down)(struct net_device *dev,
1460 bool proto_down);
1461 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1462 struct sk_buff *skb);
1463 void (*ndo_set_rx_headroom)(struct net_device *dev,
1464 int needed_headroom);
1465 int (*ndo_bpf)(struct net_device *dev,
1466 struct netdev_bpf *bpf);
1467 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1468 struct xdp_frame **xdp,
1469 u32 flags);
1470 int (*ndo_xsk_wakeup)(struct net_device *dev,
1471 u32 queue_id, u32 flags);
1472 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1473 int (*ndo_tunnel_ctl)(struct net_device *dev,
1474 struct ip_tunnel_parm *p, int cmd);
1475 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1476};
1477
1478/**
1479 * enum netdev_priv_flags - &struct net_device priv_flags
1480 *
1481 * These are the &struct net_device, they are only set internally
1482 * by drivers and used in the kernel. These flags are invisible to
1483 * userspace; this means that the order of these flags can change
1484 * during any kernel release.
1485 *
1486 * You should have a pretty good reason to be extending these flags.
1487 *
1488 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1489 * @IFF_EBRIDGE: Ethernet bridging device
1490 * @IFF_BONDING: bonding master or slave
1491 * @IFF_ISATAP: ISATAP interface (RFC4214)
1492 * @IFF_WAN_HDLC: WAN HDLC device
1493 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1494 * release skb->dst
1495 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1496 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1497 * @IFF_MACVLAN_PORT: device used as macvlan port
1498 * @IFF_BRIDGE_PORT: device used as bridge port
1499 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1500 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1501 * @IFF_UNICAST_FLT: Supports unicast filtering
1502 * @IFF_TEAM_PORT: device used as team port
1503 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1504 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1505 * change when it's running
1506 * @IFF_MACVLAN: Macvlan device
1507 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1508 * underlying stacked devices
1509 * @IFF_L3MDEV_MASTER: device is an L3 master device
1510 * @IFF_NO_QUEUE: device can run without qdisc attached
1511 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1512 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1513 * @IFF_TEAM: device is a team device
1514 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1515 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1516 * entity (i.e. the master device for bridged veth)
1517 * @IFF_MACSEC: device is a MACsec device
1518 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1519 * @IFF_FAILOVER: device is a failover master device
1520 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1521 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1522 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1523 */
1524enum netdev_priv_flags {
1525 IFF_802_1Q_VLAN = 1<<0,
1526 IFF_EBRIDGE = 1<<1,
1527 IFF_BONDING = 1<<2,
1528 IFF_ISATAP = 1<<3,
1529 IFF_WAN_HDLC = 1<<4,
1530 IFF_XMIT_DST_RELEASE = 1<<5,
1531 IFF_DONT_BRIDGE = 1<<6,
1532 IFF_DISABLE_NETPOLL = 1<<7,
1533 IFF_MACVLAN_PORT = 1<<8,
1534 IFF_BRIDGE_PORT = 1<<9,
1535 IFF_OVS_DATAPATH = 1<<10,
1536 IFF_TX_SKB_SHARING = 1<<11,
1537 IFF_UNICAST_FLT = 1<<12,
1538 IFF_TEAM_PORT = 1<<13,
1539 IFF_SUPP_NOFCS = 1<<14,
1540 IFF_LIVE_ADDR_CHANGE = 1<<15,
1541 IFF_MACVLAN = 1<<16,
1542 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1543 IFF_L3MDEV_MASTER = 1<<18,
1544 IFF_NO_QUEUE = 1<<19,
1545 IFF_OPENVSWITCH = 1<<20,
1546 IFF_L3MDEV_SLAVE = 1<<21,
1547 IFF_TEAM = 1<<22,
1548 IFF_RXFH_CONFIGURED = 1<<23,
1549 IFF_PHONY_HEADROOM = 1<<24,
1550 IFF_MACSEC = 1<<25,
1551 IFF_NO_RX_HANDLER = 1<<26,
1552 IFF_FAILOVER = 1<<27,
1553 IFF_FAILOVER_SLAVE = 1<<28,
1554 IFF_L3MDEV_RX_HANDLER = 1<<29,
1555 IFF_LIVE_RENAME_OK = 1<<30,
1556};
1557
1558#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1559#define IFF_EBRIDGE IFF_EBRIDGE
1560#define IFF_BONDING IFF_BONDING
1561#define IFF_ISATAP IFF_ISATAP
1562#define IFF_WAN_HDLC IFF_WAN_HDLC
1563#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1564#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1565#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1566#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1567#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1568#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1569#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1570#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1571#define IFF_TEAM_PORT IFF_TEAM_PORT
1572#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1573#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1574#define IFF_MACVLAN IFF_MACVLAN
1575#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1576#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1577#define IFF_NO_QUEUE IFF_NO_QUEUE
1578#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1579#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1580#define IFF_TEAM IFF_TEAM
1581#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1582#define IFF_MACSEC IFF_MACSEC
1583#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1584#define IFF_FAILOVER IFF_FAILOVER
1585#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1586#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1587#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1588
1589/* Specifies the type of the struct net_device::ml_priv pointer */
1590enum netdev_ml_priv_type {
1591 ML_PRIV_NONE,
1592 ML_PRIV_CAN,
1593};
1594
1595/**
1596 * struct net_device - The DEVICE structure.
1597 *
1598 * Actually, this whole structure is a big mistake. It mixes I/O
1599 * data with strictly "high-level" data, and it has to know about
1600 * almost every data structure used in the INET module.
1601 *
1602 * @name: This is the first field of the "visible" part of this structure
1603 * (i.e. as seen by users in the "Space.c" file). It is the name
1604 * of the interface.
1605 *
1606 * @name_node: Name hashlist node
1607 * @ifalias: SNMP alias
1608 * @mem_end: Shared memory end
1609 * @mem_start: Shared memory start
1610 * @base_addr: Device I/O address
1611 * @irq: Device IRQ number
1612 *
1613 * @state: Generic network queuing layer state, see netdev_state_t
1614 * @dev_list: The global list of network devices
1615 * @napi_list: List entry used for polling NAPI devices
1616 * @unreg_list: List entry when we are unregistering the
1617 * device; see the function unregister_netdev
1618 * @close_list: List entry used when we are closing the device
1619 * @ptype_all: Device-specific packet handlers for all protocols
1620 * @ptype_specific: Device-specific, protocol-specific packet handlers
1621 *
1622 * @adj_list: Directly linked devices, like slaves for bonding
1623 * @features: Currently active device features
1624 * @hw_features: User-changeable features
1625 *
1626 * @wanted_features: User-requested features
1627 * @vlan_features: Mask of features inheritable by VLAN devices
1628 *
1629 * @hw_enc_features: Mask of features inherited by encapsulating devices
1630 * This field indicates what encapsulation
1631 * offloads the hardware is capable of doing,
1632 * and drivers will need to set them appropriately.
1633 *
1634 * @mpls_features: Mask of features inheritable by MPLS
1635 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1636 *
1637 * @ifindex: interface index
1638 * @group: The group the device belongs to
1639 *
1640 * @stats: Statistics struct, which was left as a legacy, use
1641 * rtnl_link_stats64 instead
1642 *
1643 * @rx_dropped: Dropped packets by core network,
1644 * do not use this in drivers
1645 * @tx_dropped: Dropped packets by core network,
1646 * do not use this in drivers
1647 * @rx_nohandler: nohandler dropped packets by core network on
1648 * inactive devices, do not use this in drivers
1649 * @carrier_up_count: Number of times the carrier has been up
1650 * @carrier_down_count: Number of times the carrier has been down
1651 *
1652 * @wireless_handlers: List of functions to handle Wireless Extensions,
1653 * instead of ioctl,
1654 * see <net/iw_handler.h> for details.
1655 * @wireless_data: Instance data managed by the core of wireless extensions
1656 *
1657 * @netdev_ops: Includes several pointers to callbacks,
1658 * if one wants to override the ndo_*() functions
1659 * @ethtool_ops: Management operations
1660 * @l3mdev_ops: Layer 3 master device operations
1661 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1662 * discovery handling. Necessary for e.g. 6LoWPAN.
1663 * @xfrmdev_ops: Transformation offload operations
1664 * @tlsdev_ops: Transport Layer Security offload operations
1665 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1666 * of Layer 2 headers.
1667 *
1668 * @flags: Interface flags (a la BSD)
1669 * @priv_flags: Like 'flags' but invisible to userspace,
1670 * see if.h for the definitions
1671 * @gflags: Global flags ( kept as legacy )
1672 * @padded: How much padding added by alloc_netdev()
1673 * @operstate: RFC2863 operstate
1674 * @link_mode: Mapping policy to operstate
1675 * @if_port: Selectable AUI, TP, ...
1676 * @dma: DMA channel
1677 * @mtu: Interface MTU value
1678 * @min_mtu: Interface Minimum MTU value
1679 * @max_mtu: Interface Maximum MTU value
1680 * @type: Interface hardware type
1681 * @hard_header_len: Maximum hardware header length.
1682 * @min_header_len: Minimum hardware header length
1683 *
1684 * @needed_headroom: Extra headroom the hardware may need, but not in all
1685 * cases can this be guaranteed
1686 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1687 * cases can this be guaranteed. Some cases also use
1688 * LL_MAX_HEADER instead to allocate the skb
1689 *
1690 * interface address info:
1691 *
1692 * @perm_addr: Permanent hw address
1693 * @addr_assign_type: Hw address assignment type
1694 * @addr_len: Hardware address length
1695 * @upper_level: Maximum depth level of upper devices.
1696 * @lower_level: Maximum depth level of lower devices.
1697 * @neigh_priv_len: Used in neigh_alloc()
1698 * @dev_id: Used to differentiate devices that share
1699 * the same link layer address
1700 * @dev_port: Used to differentiate devices that share
1701 * the same function
1702 * @addr_list_lock: XXX: need comments on this one
1703 * @name_assign_type: network interface name assignment type
1704 * @uc_promisc: Counter that indicates promiscuous mode
1705 * has been enabled due to the need to listen to
1706 * additional unicast addresses in a device that
1707 * does not implement ndo_set_rx_mode()
1708 * @uc: unicast mac addresses
1709 * @mc: multicast mac addresses
1710 * @dev_addrs: list of device hw addresses
1711 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1712 * @promiscuity: Number of times the NIC is told to work in
1713 * promiscuous mode; if it becomes 0 the NIC will
1714 * exit promiscuous mode
1715 * @allmulti: Counter, enables or disables allmulticast mode
1716 *
1717 * @vlan_info: VLAN info
1718 * @dsa_ptr: dsa specific data
1719 * @tipc_ptr: TIPC specific data
1720 * @atalk_ptr: AppleTalk link
1721 * @ip_ptr: IPv4 specific data
1722 * @dn_ptr: DECnet specific data
1723 * @ip6_ptr: IPv6 specific data
1724 * @ax25_ptr: AX.25 specific data
1725 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1726 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1727 * device struct
1728 * @mpls_ptr: mpls_dev struct pointer
1729 *
1730 * @dev_addr: Hw address (before bcast,
1731 * because most packets are unicast)
1732 *
1733 * @_rx: Array of RX queues
1734 * @num_rx_queues: Number of RX queues
1735 * allocated at register_netdev() time
1736 * @real_num_rx_queues: Number of RX queues currently active in device
1737 * @xdp_prog: XDP sockets filter program pointer
1738 * @gro_flush_timeout: timeout for GRO layer in NAPI
1739 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1740 * allow to avoid NIC hard IRQ, on busy queues.
1741 *
1742 * @rx_handler: handler for received packets
1743 * @rx_handler_data: XXX: need comments on this one
1744 * @miniq_ingress: ingress/clsact qdisc specific data for
1745 * ingress processing
1746 * @ingress_queue: XXX: need comments on this one
1747 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1748 * @broadcast: hw bcast address
1749 *
1750 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1751 * indexed by RX queue number. Assigned by driver.
1752 * This must only be set if the ndo_rx_flow_steer
1753 * operation is defined
1754 * @index_hlist: Device index hash chain
1755 *
1756 * @_tx: Array of TX queues
1757 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1758 * @real_num_tx_queues: Number of TX queues currently active in device
1759 * @qdisc: Root qdisc from userspace point of view
1760 * @tx_queue_len: Max frames per queue allowed
1761 * @tx_global_lock: XXX: need comments on this one
1762 * @xdp_bulkq: XDP device bulk queue
1763 * @xps_cpus_map: all CPUs map for XPS device
1764 * @xps_rxqs_map: all RXQs map for XPS device
1765 *
1766 * @xps_maps: XXX: need comments on this one
1767 * @miniq_egress: clsact qdisc specific data for
1768 * egress processing
1769 * @qdisc_hash: qdisc hash table
1770 * @watchdog_timeo: Represents the timeout that is used by
1771 * the watchdog (see dev_watchdog())
1772 * @watchdog_timer: List of timers
1773 *
1774 * @proto_down_reason: reason a netdev interface is held down
1775 * @pcpu_refcnt: Number of references to this device
1776 * @todo_list: Delayed register/unregister
1777 * @link_watch_list: XXX: need comments on this one
1778 *
1779 * @reg_state: Register/unregister state machine
1780 * @dismantle: Device is going to be freed
1781 * @rtnl_link_state: This enum represents the phases of creating
1782 * a new link
1783 *
1784 * @needs_free_netdev: Should unregister perform free_netdev?
1785 * @priv_destructor: Called from unregister
1786 * @npinfo: XXX: need comments on this one
1787 * @nd_net: Network namespace this network device is inside
1788 *
1789 * @ml_priv: Mid-layer private
1790 * @ml_priv_type: Mid-layer private type
1791 * @lstats: Loopback statistics
1792 * @tstats: Tunnel statistics
1793 * @dstats: Dummy statistics
1794 * @vstats: Virtual ethernet statistics
1795 *
1796 * @garp_port: GARP
1797 * @mrp_port: MRP
1798 *
1799 * @dev: Class/net/name entry
1800 * @sysfs_groups: Space for optional device, statistics and wireless
1801 * sysfs groups
1802 *
1803 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1804 * @rtnl_link_ops: Rtnl_link_ops
1805 *
1806 * @gso_max_size: Maximum size of generic segmentation offload
1807 * @gso_max_segs: Maximum number of segments that can be passed to the
1808 * NIC for GSO
1809 *
1810 * @dcbnl_ops: Data Center Bridging netlink ops
1811 * @num_tc: Number of traffic classes in the net device
1812 * @tc_to_txq: XXX: need comments on this one
1813 * @prio_tc_map: XXX: need comments on this one
1814 *
1815 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1816 *
1817 * @priomap: XXX: need comments on this one
1818 * @phydev: Physical device may attach itself
1819 * for hardware timestamping
1820 * @sfp_bus: attached &struct sfp_bus structure.
1821 *
1822 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1823 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1824 *
1825 * @proto_down: protocol port state information can be sent to the
1826 * switch driver and used to set the phys state of the
1827 * switch port.
1828 *
1829 * @wol_enabled: Wake-on-LAN is enabled
1830 *
1831 * @threaded: napi threaded mode is enabled
1832 *
1833 * @net_notifier_list: List of per-net netdev notifier block
1834 * that follow this device when it is moved
1835 * to another network namespace.
1836 *
1837 * @macsec_ops: MACsec offloading ops
1838 *
1839 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1840 * offload capabilities of the device
1841 * @udp_tunnel_nic: UDP tunnel offload state
1842 * @xdp_state: stores info on attached XDP BPF programs
1843 *
1844 * @nested_level: Used as as a parameter of spin_lock_nested() of
1845 * dev->addr_list_lock.
1846 * @unlink_list: As netif_addr_lock() can be called recursively,
1847 * keep a list of interfaces to be deleted.
1848 *
1849 * FIXME: cleanup struct net_device such that network protocol info
1850 * moves out.
1851 */
1852
1853struct net_device {
1854 char name[IFNAMSIZ];
1855 struct netdev_name_node *name_node;
1856 struct dev_ifalias __rcu *ifalias;
1857 /*
1858 * I/O specific fields
1859 * FIXME: Merge these and struct ifmap into one
1860 */
1861 unsigned long mem_end;
1862 unsigned long mem_start;
1863 unsigned long base_addr;
1864
1865 /*
1866 * Some hardware also needs these fields (state,dev_list,
1867 * napi_list,unreg_list,close_list) but they are not
1868 * part of the usual set specified in Space.c.
1869 */
1870
1871 unsigned long state;
1872
1873 struct list_head dev_list;
1874 struct list_head napi_list;
1875 struct list_head unreg_list;
1876 struct list_head close_list;
1877 struct list_head ptype_all;
1878 struct list_head ptype_specific;
1879
1880 struct {
1881 struct list_head upper;
1882 struct list_head lower;
1883 } adj_list;
1884
1885 /* Read-mostly cache-line for fast-path access */
1886 unsigned int flags;
1887 unsigned int priv_flags;
1888 const struct net_device_ops *netdev_ops;
1889 int ifindex;
1890 unsigned short gflags;
1891 unsigned short hard_header_len;
1892
1893 /* Note : dev->mtu is often read without holding a lock.
1894 * Writers usually hold RTNL.
1895 * It is recommended to use READ_ONCE() to annotate the reads,
1896 * and to use WRITE_ONCE() to annotate the writes.
1897 */
1898 unsigned int mtu;
1899 unsigned short needed_headroom;
1900 unsigned short needed_tailroom;
1901
1902 netdev_features_t features;
1903 netdev_features_t hw_features;
1904 netdev_features_t wanted_features;
1905 netdev_features_t vlan_features;
1906 netdev_features_t hw_enc_features;
1907 netdev_features_t mpls_features;
1908 netdev_features_t gso_partial_features;
1909
1910 unsigned int min_mtu;
1911 unsigned int max_mtu;
1912 unsigned short type;
1913 unsigned char min_header_len;
1914 unsigned char name_assign_type;
1915
1916 int group;
1917
1918 struct net_device_stats stats; /* not used by modern drivers */
1919
1920 atomic_long_t rx_dropped;
1921 atomic_long_t tx_dropped;
1922 atomic_long_t rx_nohandler;
1923
1924 /* Stats to monitor link on/off, flapping */
1925 atomic_t carrier_up_count;
1926 atomic_t carrier_down_count;
1927
1928#ifdef CONFIG_WIRELESS_EXT
1929 const struct iw_handler_def *wireless_handlers;
1930 struct iw_public_data *wireless_data;
1931#endif
1932 const struct ethtool_ops *ethtool_ops;
1933#ifdef CONFIG_NET_L3_MASTER_DEV
1934 const struct l3mdev_ops *l3mdev_ops;
1935#endif
1936#if IS_ENABLED(CONFIG_IPV6)
1937 const struct ndisc_ops *ndisc_ops;
1938#endif
1939
1940#ifdef CONFIG_XFRM_OFFLOAD
1941 const struct xfrmdev_ops *xfrmdev_ops;
1942#endif
1943
1944#if IS_ENABLED(CONFIG_TLS_DEVICE)
1945 const struct tlsdev_ops *tlsdev_ops;
1946#endif
1947
1948 const struct header_ops *header_ops;
1949
1950 unsigned char operstate;
1951 unsigned char link_mode;
1952
1953 unsigned char if_port;
1954 unsigned char dma;
1955
1956 /* Interface address info. */
1957 unsigned char perm_addr[MAX_ADDR_LEN];
1958 unsigned char addr_assign_type;
1959 unsigned char addr_len;
1960 unsigned char upper_level;
1961 unsigned char lower_level;
1962
1963 unsigned short neigh_priv_len;
1964 unsigned short dev_id;
1965 unsigned short dev_port;
1966 unsigned short padded;
1967
1968 spinlock_t addr_list_lock;
1969 int irq;
1970
1971 struct netdev_hw_addr_list uc;
1972 struct netdev_hw_addr_list mc;
1973 struct netdev_hw_addr_list dev_addrs;
1974
1975#ifdef CONFIG_SYSFS
1976 struct kset *queues_kset;
1977#endif
1978#ifdef CONFIG_LOCKDEP
1979 struct list_head unlink_list;
1980#endif
1981 unsigned int promiscuity;
1982 unsigned int allmulti;
1983 bool uc_promisc;
1984#ifdef CONFIG_LOCKDEP
1985 unsigned char nested_level;
1986#endif
1987
1988
1989 /* Protocol-specific pointers */
1990
1991#if IS_ENABLED(CONFIG_VLAN_8021Q)
1992 struct vlan_info __rcu *vlan_info;
1993#endif
1994#if IS_ENABLED(CONFIG_NET_DSA)
1995 struct dsa_port *dsa_ptr;
1996#endif
1997#if IS_ENABLED(CONFIG_TIPC)
1998 struct tipc_bearer __rcu *tipc_ptr;
1999#endif
2000#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2001 void *atalk_ptr;
2002#endif
2003 struct in_device __rcu *ip_ptr;
2004#if IS_ENABLED(CONFIG_DECNET)
2005 struct dn_dev __rcu *dn_ptr;
2006#endif
2007 struct inet6_dev __rcu *ip6_ptr;
2008#if IS_ENABLED(CONFIG_AX25)
2009 void *ax25_ptr;
2010#endif
2011 struct wireless_dev *ieee80211_ptr;
2012 struct wpan_dev *ieee802154_ptr;
2013#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2014 struct mpls_dev __rcu *mpls_ptr;
2015#endif
2016
2017/*
2018 * Cache lines mostly used on receive path (including eth_type_trans())
2019 */
2020 /* Interface address info used in eth_type_trans() */
2021 unsigned char *dev_addr;
2022
2023 struct netdev_rx_queue *_rx;
2024 unsigned int num_rx_queues;
2025 unsigned int real_num_rx_queues;
2026
2027 struct bpf_prog __rcu *xdp_prog;
2028 unsigned long gro_flush_timeout;
2029 int napi_defer_hard_irqs;
2030 rx_handler_func_t __rcu *rx_handler;
2031 void __rcu *rx_handler_data;
2032
2033#ifdef CONFIG_NET_CLS_ACT
2034 struct mini_Qdisc __rcu *miniq_ingress;
2035#endif
2036 struct netdev_queue __rcu *ingress_queue;
2037#ifdef CONFIG_NETFILTER_INGRESS
2038 struct nf_hook_entries __rcu *nf_hooks_ingress;
2039#endif
2040
2041 unsigned char broadcast[MAX_ADDR_LEN];
2042#ifdef CONFIG_RFS_ACCEL
2043 struct cpu_rmap *rx_cpu_rmap;
2044#endif
2045 struct hlist_node index_hlist;
2046
2047/*
2048 * Cache lines mostly used on transmit path
2049 */
2050 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2051 unsigned int num_tx_queues;
2052 unsigned int real_num_tx_queues;
2053 struct Qdisc *qdisc;
2054 unsigned int tx_queue_len;
2055 spinlock_t tx_global_lock;
2056
2057 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2058
2059#ifdef CONFIG_XPS
2060 struct xps_dev_maps __rcu *xps_cpus_map;
2061 struct xps_dev_maps __rcu *xps_rxqs_map;
2062#endif
2063#ifdef CONFIG_NET_CLS_ACT
2064 struct mini_Qdisc __rcu *miniq_egress;
2065#endif
2066
2067#ifdef CONFIG_NET_SCHED
2068 DECLARE_HASHTABLE (qdisc_hash, 4);
2069#endif
2070 /* These may be needed for future network-power-down code. */
2071 struct timer_list watchdog_timer;
2072 int watchdog_timeo;
2073
2074 u32 proto_down_reason;
2075
2076 struct list_head todo_list;
2077 int __percpu *pcpu_refcnt;
2078
2079 struct list_head link_watch_list;
2080
2081 enum { NETREG_UNINITIALIZED=0,
2082 NETREG_REGISTERED, /* completed register_netdevice */
2083 NETREG_UNREGISTERING, /* called unregister_netdevice */
2084 NETREG_UNREGISTERED, /* completed unregister todo */
2085 NETREG_RELEASED, /* called free_netdev */
2086 NETREG_DUMMY, /* dummy device for NAPI poll */
2087 } reg_state:8;
2088
2089 bool dismantle;
2090
2091 enum {
2092 RTNL_LINK_INITIALIZED,
2093 RTNL_LINK_INITIALIZING,
2094 } rtnl_link_state:16;
2095
2096 bool needs_free_netdev;
2097 void (*priv_destructor)(struct net_device *dev);
2098
2099#ifdef CONFIG_NETPOLL
2100 struct netpoll_info __rcu *npinfo;
2101#endif
2102
2103 possible_net_t nd_net;
2104
2105 /* mid-layer private */
2106 void *ml_priv;
2107 enum netdev_ml_priv_type ml_priv_type;
2108
2109 union {
2110 struct pcpu_lstats __percpu *lstats;
2111 struct pcpu_sw_netstats __percpu *tstats;
2112 struct pcpu_dstats __percpu *dstats;
2113 };
2114
2115#if IS_ENABLED(CONFIG_GARP)
2116 struct garp_port __rcu *garp_port;
2117#endif
2118#if IS_ENABLED(CONFIG_MRP)
2119 struct mrp_port __rcu *mrp_port;
2120#endif
2121
2122 struct device dev;
2123 const struct attribute_group *sysfs_groups[4];
2124 const struct attribute_group *sysfs_rx_queue_group;
2125
2126 const struct rtnl_link_ops *rtnl_link_ops;
2127
2128 /* for setting kernel sock attribute on TCP connection setup */
2129#define GSO_MAX_SIZE 65536
2130 unsigned int gso_max_size;
2131#define GSO_MAX_SEGS 65535
2132 u16 gso_max_segs;
2133
2134#ifdef CONFIG_DCB
2135 const struct dcbnl_rtnl_ops *dcbnl_ops;
2136#endif
2137 s16 num_tc;
2138 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2139 u8 prio_tc_map[TC_BITMASK + 1];
2140
2141#if IS_ENABLED(CONFIG_FCOE)
2142 unsigned int fcoe_ddp_xid;
2143#endif
2144#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2145 struct netprio_map __rcu *priomap;
2146#endif
2147 struct phy_device *phydev;
2148 struct sfp_bus *sfp_bus;
2149 struct lock_class_key *qdisc_tx_busylock;
2150 struct lock_class_key *qdisc_running_key;
2151 bool proto_down;
2152 unsigned wol_enabled:1;
2153 unsigned threaded:1;
2154
2155 struct list_head net_notifier_list;
2156
2157#if IS_ENABLED(CONFIG_MACSEC)
2158 /* MACsec management functions */
2159 const struct macsec_ops *macsec_ops;
2160#endif
2161 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2162 struct udp_tunnel_nic *udp_tunnel_nic;
2163
2164 /* protected by rtnl_lock */
2165 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2166};
2167#define to_net_dev(d) container_of(d, struct net_device, dev)
2168
2169static inline bool netif_elide_gro(const struct net_device *dev)
2170{
2171 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2172 return true;
2173 return false;
2174}
2175
2176#define NETDEV_ALIGN 32
2177
2178static inline
2179int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2180{
2181 return dev->prio_tc_map[prio & TC_BITMASK];
2182}
2183
2184static inline
2185int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2186{
2187 if (tc >= dev->num_tc)
2188 return -EINVAL;
2189
2190 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2191 return 0;
2192}
2193
2194int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2195void netdev_reset_tc(struct net_device *dev);
2196int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2197int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2198
2199static inline
2200int netdev_get_num_tc(struct net_device *dev)
2201{
2202 return dev->num_tc;
2203}
2204
2205static inline void net_prefetch(void *p)
2206{
2207 prefetch(p);
2208#if L1_CACHE_BYTES < 128
2209 prefetch((u8 *)p + L1_CACHE_BYTES);
2210#endif
2211}
2212
2213static inline void net_prefetchw(void *p)
2214{
2215 prefetchw(p);
2216#if L1_CACHE_BYTES < 128
2217 prefetchw((u8 *)p + L1_CACHE_BYTES);
2218#endif
2219}
2220
2221void netdev_unbind_sb_channel(struct net_device *dev,
2222 struct net_device *sb_dev);
2223int netdev_bind_sb_channel_queue(struct net_device *dev,
2224 struct net_device *sb_dev,
2225 u8 tc, u16 count, u16 offset);
2226int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2227static inline int netdev_get_sb_channel(struct net_device *dev)
2228{
2229 return max_t(int, -dev->num_tc, 0);
2230}
2231
2232static inline
2233struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2234 unsigned int index)
2235{
2236 return &dev->_tx[index];
2237}
2238
2239static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2240 const struct sk_buff *skb)
2241{
2242 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2243}
2244
2245static inline void netdev_for_each_tx_queue(struct net_device *dev,
2246 void (*f)(struct net_device *,
2247 struct netdev_queue *,
2248 void *),
2249 void *arg)
2250{
2251 unsigned int i;
2252
2253 for (i = 0; i < dev->num_tx_queues; i++)
2254 f(dev, &dev->_tx[i], arg);
2255}
2256
2257#define netdev_lockdep_set_classes(dev) \
2258{ \
2259 static struct lock_class_key qdisc_tx_busylock_key; \
2260 static struct lock_class_key qdisc_running_key; \
2261 static struct lock_class_key qdisc_xmit_lock_key; \
2262 static struct lock_class_key dev_addr_list_lock_key; \
2263 unsigned int i; \
2264 \
2265 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2266 (dev)->qdisc_running_key = &qdisc_running_key; \
2267 lockdep_set_class(&(dev)->addr_list_lock, \
2268 &dev_addr_list_lock_key); \
2269 for (i = 0; i < (dev)->num_tx_queues; i++) \
2270 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2271 &qdisc_xmit_lock_key); \
2272}
2273
2274u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2275 struct net_device *sb_dev);
2276struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2277 struct sk_buff *skb,
2278 struct net_device *sb_dev);
2279
2280/* returns the headroom that the master device needs to take in account
2281 * when forwarding to this dev
2282 */
2283static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2284{
2285 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2286}
2287
2288static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2289{
2290 if (dev->netdev_ops->ndo_set_rx_headroom)
2291 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2292}
2293
2294/* set the device rx headroom to the dev's default */
2295static inline void netdev_reset_rx_headroom(struct net_device *dev)
2296{
2297 netdev_set_rx_headroom(dev, -1);
2298}
2299
2300static inline void *netdev_get_ml_priv(struct net_device *dev,
2301 enum netdev_ml_priv_type type)
2302{
2303 if (dev->ml_priv_type != type)
2304 return NULL;
2305
2306 return dev->ml_priv;
2307}
2308
2309static inline void netdev_set_ml_priv(struct net_device *dev,
2310 void *ml_priv,
2311 enum netdev_ml_priv_type type)
2312{
2313 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2314 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2315 dev->ml_priv_type, type);
2316 WARN(!dev->ml_priv_type && dev->ml_priv,
2317 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2318
2319 dev->ml_priv = ml_priv;
2320 dev->ml_priv_type = type;
2321}
2322
2323/*
2324 * Net namespace inlines
2325 */
2326static inline
2327struct net *dev_net(const struct net_device *dev)
2328{
2329 return read_pnet(&dev->nd_net);
2330}
2331
2332static inline
2333void dev_net_set(struct net_device *dev, struct net *net)
2334{
2335 write_pnet(&dev->nd_net, net);
2336}
2337
2338/**
2339 * netdev_priv - access network device private data
2340 * @dev: network device
2341 *
2342 * Get network device private data
2343 */
2344static inline void *netdev_priv(const struct net_device *dev)
2345{
2346 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2347}
2348
2349/* Set the sysfs physical device reference for the network logical device
2350 * if set prior to registration will cause a symlink during initialization.
2351 */
2352#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2353
2354/* Set the sysfs device type for the network logical device to allow
2355 * fine-grained identification of different network device types. For
2356 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2357 */
2358#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2359
2360/* Default NAPI poll() weight
2361 * Device drivers are strongly advised to not use bigger value
2362 */
2363#define NAPI_POLL_WEIGHT 64
2364
2365/**
2366 * netif_napi_add - initialize a NAPI context
2367 * @dev: network device
2368 * @napi: NAPI context
2369 * @poll: polling function
2370 * @weight: default weight
2371 *
2372 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2373 * *any* of the other NAPI-related functions.
2374 */
2375void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2376 int (*poll)(struct napi_struct *, int), int weight);
2377
2378/**
2379 * netif_tx_napi_add - initialize a NAPI context
2380 * @dev: network device
2381 * @napi: NAPI context
2382 * @poll: polling function
2383 * @weight: default weight
2384 *
2385 * This variant of netif_napi_add() should be used from drivers using NAPI
2386 * to exclusively poll a TX queue.
2387 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2388 */
2389static inline void netif_tx_napi_add(struct net_device *dev,
2390 struct napi_struct *napi,
2391 int (*poll)(struct napi_struct *, int),
2392 int weight)
2393{
2394 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2395 netif_napi_add(dev, napi, poll, weight);
2396}
2397
2398/**
2399 * __netif_napi_del - remove a NAPI context
2400 * @napi: NAPI context
2401 *
2402 * Warning: caller must observe RCU grace period before freeing memory
2403 * containing @napi. Drivers might want to call this helper to combine
2404 * all the needed RCU grace periods into a single one.
2405 */
2406void __netif_napi_del(struct napi_struct *napi);
2407
2408/**
2409 * netif_napi_del - remove a NAPI context
2410 * @napi: NAPI context
2411 *
2412 * netif_napi_del() removes a NAPI context from the network device NAPI list
2413 */
2414static inline void netif_napi_del(struct napi_struct *napi)
2415{
2416 __netif_napi_del(napi);
2417 synchronize_net();
2418}
2419
2420struct napi_gro_cb {
2421 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2422 void *frag0;
2423
2424 /* Length of frag0. */
2425 unsigned int frag0_len;
2426
2427 /* This indicates where we are processing relative to skb->data. */
2428 int data_offset;
2429
2430 /* This is non-zero if the packet cannot be merged with the new skb. */
2431 u16 flush;
2432
2433 /* Save the IP ID here and check when we get to the transport layer */
2434 u16 flush_id;
2435
2436 /* Number of segments aggregated. */
2437 u16 count;
2438
2439 /* Start offset for remote checksum offload */
2440 u16 gro_remcsum_start;
2441
2442 /* jiffies when first packet was created/queued */
2443 unsigned long age;
2444
2445 /* Used in ipv6_gro_receive() and foo-over-udp */
2446 u16 proto;
2447
2448 /* This is non-zero if the packet may be of the same flow. */
2449 u8 same_flow:1;
2450
2451 /* Used in tunnel GRO receive */
2452 u8 encap_mark:1;
2453
2454 /* GRO checksum is valid */
2455 u8 csum_valid:1;
2456
2457 /* Number of checksums via CHECKSUM_UNNECESSARY */
2458 u8 csum_cnt:3;
2459
2460 /* Free the skb? */
2461 u8 free:2;
2462#define NAPI_GRO_FREE 1
2463#define NAPI_GRO_FREE_STOLEN_HEAD 2
2464
2465 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2466 u8 is_ipv6:1;
2467
2468 /* Used in GRE, set in fou/gue_gro_receive */
2469 u8 is_fou:1;
2470
2471 /* Used to determine if flush_id can be ignored */
2472 u8 is_atomic:1;
2473
2474 /* Number of gro_receive callbacks this packet already went through */
2475 u8 recursion_counter:4;
2476
2477 /* GRO is done by frag_list pointer chaining. */
2478 u8 is_flist:1;
2479
2480 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2481 __wsum csum;
2482
2483 /* used in skb_gro_receive() slow path */
2484 struct sk_buff *last;
2485};
2486
2487#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2488
2489#define GRO_RECURSION_LIMIT 15
2490static inline int gro_recursion_inc_test(struct sk_buff *skb)
2491{
2492 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2493}
2494
2495typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2496static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2497 struct list_head *head,
2498 struct sk_buff *skb)
2499{
2500 if (unlikely(gro_recursion_inc_test(skb))) {
2501 NAPI_GRO_CB(skb)->flush |= 1;
2502 return NULL;
2503 }
2504
2505 return cb(head, skb);
2506}
2507
2508typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2509 struct sk_buff *);
2510static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2511 struct sock *sk,
2512 struct list_head *head,
2513 struct sk_buff *skb)
2514{
2515 if (unlikely(gro_recursion_inc_test(skb))) {
2516 NAPI_GRO_CB(skb)->flush |= 1;
2517 return NULL;
2518 }
2519
2520 return cb(sk, head, skb);
2521}
2522
2523struct packet_type {
2524 __be16 type; /* This is really htons(ether_type). */
2525 bool ignore_outgoing;
2526 struct net_device *dev; /* NULL is wildcarded here */
2527 int (*func) (struct sk_buff *,
2528 struct net_device *,
2529 struct packet_type *,
2530 struct net_device *);
2531 void (*list_func) (struct list_head *,
2532 struct packet_type *,
2533 struct net_device *);
2534 bool (*id_match)(struct packet_type *ptype,
2535 struct sock *sk);
2536 void *af_packet_priv;
2537 struct list_head list;
2538};
2539
2540struct offload_callbacks {
2541 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2542 netdev_features_t features);
2543 struct sk_buff *(*gro_receive)(struct list_head *head,
2544 struct sk_buff *skb);
2545 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2546};
2547
2548struct packet_offload {
2549 __be16 type; /* This is really htons(ether_type). */
2550 u16 priority;
2551 struct offload_callbacks callbacks;
2552 struct list_head list;
2553};
2554
2555/* often modified stats are per-CPU, other are shared (netdev->stats) */
2556struct pcpu_sw_netstats {
2557 u64 rx_packets;
2558 u64 rx_bytes;
2559 u64 tx_packets;
2560 u64 tx_bytes;
2561 struct u64_stats_sync syncp;
2562} __aligned(4 * sizeof(u64));
2563
2564struct pcpu_lstats {
2565 u64_stats_t packets;
2566 u64_stats_t bytes;
2567 struct u64_stats_sync syncp;
2568} __aligned(2 * sizeof(u64));
2569
2570void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2571
2572static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2573{
2574 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2575
2576 u64_stats_update_begin(&tstats->syncp);
2577 tstats->rx_bytes += len;
2578 tstats->rx_packets++;
2579 u64_stats_update_end(&tstats->syncp);
2580}
2581
2582static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2583 unsigned int packets,
2584 unsigned int len)
2585{
2586 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2587
2588 u64_stats_update_begin(&tstats->syncp);
2589 tstats->tx_bytes += len;
2590 tstats->tx_packets += packets;
2591 u64_stats_update_end(&tstats->syncp);
2592}
2593
2594static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2595{
2596 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2597
2598 u64_stats_update_begin(&lstats->syncp);
2599 u64_stats_add(&lstats->bytes, len);
2600 u64_stats_inc(&lstats->packets);
2601 u64_stats_update_end(&lstats->syncp);
2602}
2603
2604#define __netdev_alloc_pcpu_stats(type, gfp) \
2605({ \
2606 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2607 if (pcpu_stats) { \
2608 int __cpu; \
2609 for_each_possible_cpu(__cpu) { \
2610 typeof(type) *stat; \
2611 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2612 u64_stats_init(&stat->syncp); \
2613 } \
2614 } \
2615 pcpu_stats; \
2616})
2617
2618#define netdev_alloc_pcpu_stats(type) \
2619 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2620
2621#define devm_netdev_alloc_pcpu_stats(dev, type) \
2622({ \
2623 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2624 if (pcpu_stats) { \
2625 int __cpu; \
2626 for_each_possible_cpu(__cpu) { \
2627 typeof(type) *stat; \
2628 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2629 u64_stats_init(&stat->syncp); \
2630 } \
2631 } \
2632 pcpu_stats; \
2633})
2634
2635enum netdev_lag_tx_type {
2636 NETDEV_LAG_TX_TYPE_UNKNOWN,
2637 NETDEV_LAG_TX_TYPE_RANDOM,
2638 NETDEV_LAG_TX_TYPE_BROADCAST,
2639 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2640 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2641 NETDEV_LAG_TX_TYPE_HASH,
2642};
2643
2644enum netdev_lag_hash {
2645 NETDEV_LAG_HASH_NONE,
2646 NETDEV_LAG_HASH_L2,
2647 NETDEV_LAG_HASH_L34,
2648 NETDEV_LAG_HASH_L23,
2649 NETDEV_LAG_HASH_E23,
2650 NETDEV_LAG_HASH_E34,
2651 NETDEV_LAG_HASH_VLAN_SRCMAC,
2652 NETDEV_LAG_HASH_UNKNOWN,
2653};
2654
2655struct netdev_lag_upper_info {
2656 enum netdev_lag_tx_type tx_type;
2657 enum netdev_lag_hash hash_type;
2658};
2659
2660struct netdev_lag_lower_state_info {
2661 u8 link_up : 1,
2662 tx_enabled : 1;
2663};
2664
2665#include <linux/notifier.h>
2666
2667/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2668 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2669 * adding new types.
2670 */
2671enum netdev_cmd {
2672 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2673 NETDEV_DOWN,
2674 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2675 detected a hardware crash and restarted
2676 - we can use this eg to kick tcp sessions
2677 once done */
2678 NETDEV_CHANGE, /* Notify device state change */
2679 NETDEV_REGISTER,
2680 NETDEV_UNREGISTER,
2681 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2682 NETDEV_CHANGEADDR, /* notify after the address change */
2683 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2684 NETDEV_GOING_DOWN,
2685 NETDEV_CHANGENAME,
2686 NETDEV_FEAT_CHANGE,
2687 NETDEV_BONDING_FAILOVER,
2688 NETDEV_PRE_UP,
2689 NETDEV_PRE_TYPE_CHANGE,
2690 NETDEV_POST_TYPE_CHANGE,
2691 NETDEV_POST_INIT,
2692 NETDEV_RELEASE,
2693 NETDEV_NOTIFY_PEERS,
2694 NETDEV_JOIN,
2695 NETDEV_CHANGEUPPER,
2696 NETDEV_RESEND_IGMP,
2697 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2698 NETDEV_CHANGEINFODATA,
2699 NETDEV_BONDING_INFO,
2700 NETDEV_PRECHANGEUPPER,
2701 NETDEV_CHANGELOWERSTATE,
2702 NETDEV_UDP_TUNNEL_PUSH_INFO,
2703 NETDEV_UDP_TUNNEL_DROP_INFO,
2704 NETDEV_CHANGE_TX_QUEUE_LEN,
2705 NETDEV_CVLAN_FILTER_PUSH_INFO,
2706 NETDEV_CVLAN_FILTER_DROP_INFO,
2707 NETDEV_SVLAN_FILTER_PUSH_INFO,
2708 NETDEV_SVLAN_FILTER_DROP_INFO,
2709};
2710const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2711
2712int register_netdevice_notifier(struct notifier_block *nb);
2713int unregister_netdevice_notifier(struct notifier_block *nb);
2714int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2715int unregister_netdevice_notifier_net(struct net *net,
2716 struct notifier_block *nb);
2717int register_netdevice_notifier_dev_net(struct net_device *dev,
2718 struct notifier_block *nb,
2719 struct netdev_net_notifier *nn);
2720int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2721 struct notifier_block *nb,
2722 struct netdev_net_notifier *nn);
2723
2724struct netdev_notifier_info {
2725 struct net_device *dev;
2726 struct netlink_ext_ack *extack;
2727};
2728
2729struct netdev_notifier_info_ext {
2730 struct netdev_notifier_info info; /* must be first */
2731 union {
2732 u32 mtu;
2733 } ext;
2734};
2735
2736struct netdev_notifier_change_info {
2737 struct netdev_notifier_info info; /* must be first */
2738 unsigned int flags_changed;
2739};
2740
2741struct netdev_notifier_changeupper_info {
2742 struct netdev_notifier_info info; /* must be first */
2743 struct net_device *upper_dev; /* new upper dev */
2744 bool master; /* is upper dev master */
2745 bool linking; /* is the notification for link or unlink */
2746 void *upper_info; /* upper dev info */
2747};
2748
2749struct netdev_notifier_changelowerstate_info {
2750 struct netdev_notifier_info info; /* must be first */
2751 void *lower_state_info; /* is lower dev state */
2752};
2753
2754struct netdev_notifier_pre_changeaddr_info {
2755 struct netdev_notifier_info info; /* must be first */
2756 const unsigned char *dev_addr;
2757};
2758
2759static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2760 struct net_device *dev)
2761{
2762 info->dev = dev;
2763 info->extack = NULL;
2764}
2765
2766static inline struct net_device *
2767netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2768{
2769 return info->dev;
2770}
2771
2772static inline struct netlink_ext_ack *
2773netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2774{
2775 return info->extack;
2776}
2777
2778int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2779
2780
2781extern rwlock_t dev_base_lock; /* Device list lock */
2782
2783#define for_each_netdev(net, d) \
2784 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2785#define for_each_netdev_reverse(net, d) \
2786 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2787#define for_each_netdev_rcu(net, d) \
2788 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2789#define for_each_netdev_safe(net, d, n) \
2790 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2791#define for_each_netdev_continue(net, d) \
2792 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2793#define for_each_netdev_continue_reverse(net, d) \
2794 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2795 dev_list)
2796#define for_each_netdev_continue_rcu(net, d) \
2797 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2798#define for_each_netdev_in_bond_rcu(bond, slave) \
2799 for_each_netdev_rcu(&init_net, slave) \
2800 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2801#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2802
2803static inline struct net_device *next_net_device(struct net_device *dev)
2804{
2805 struct list_head *lh;
2806 struct net *net;
2807
2808 net = dev_net(dev);
2809 lh = dev->dev_list.next;
2810 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2811}
2812
2813static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2814{
2815 struct list_head *lh;
2816 struct net *net;
2817
2818 net = dev_net(dev);
2819 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2820 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2821}
2822
2823static inline struct net_device *first_net_device(struct net *net)
2824{
2825 return list_empty(&net->dev_base_head) ? NULL :
2826 net_device_entry(net->dev_base_head.next);
2827}
2828
2829static inline struct net_device *first_net_device_rcu(struct net *net)
2830{
2831 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2832
2833 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2834}
2835
2836int netdev_boot_setup_check(struct net_device *dev);
2837unsigned long netdev_boot_base(const char *prefix, int unit);
2838struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2839 const char *hwaddr);
2840struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2841void dev_add_pack(struct packet_type *pt);
2842void dev_remove_pack(struct packet_type *pt);
2843void __dev_remove_pack(struct packet_type *pt);
2844void dev_add_offload(struct packet_offload *po);
2845void dev_remove_offload(struct packet_offload *po);
2846
2847int dev_get_iflink(const struct net_device *dev);
2848int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2849struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2850 unsigned short mask);
2851struct net_device *dev_get_by_name(struct net *net, const char *name);
2852struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2853struct net_device *__dev_get_by_name(struct net *net, const char *name);
2854int dev_alloc_name(struct net_device *dev, const char *name);
2855int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2856void dev_close(struct net_device *dev);
2857void dev_close_many(struct list_head *head, bool unlink);
2858void dev_disable_lro(struct net_device *dev);
2859int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2860u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2861 struct net_device *sb_dev);
2862u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2863 struct net_device *sb_dev);
2864
2865int dev_queue_xmit(struct sk_buff *skb);
2866int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2867int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2868
2869static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2870{
2871 int ret;
2872
2873 ret = __dev_direct_xmit(skb, queue_id);
2874 if (!dev_xmit_complete(ret))
2875 kfree_skb(skb);
2876 return ret;
2877}
2878
2879int register_netdevice(struct net_device *dev);
2880void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2881void unregister_netdevice_many(struct list_head *head);
2882static inline void unregister_netdevice(struct net_device *dev)
2883{
2884 unregister_netdevice_queue(dev, NULL);
2885}
2886
2887int netdev_refcnt_read(const struct net_device *dev);
2888void free_netdev(struct net_device *dev);
2889void netdev_freemem(struct net_device *dev);
2890int init_dummy_netdev(struct net_device *dev);
2891
2892struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2893 struct sk_buff *skb,
2894 bool all_slaves);
2895struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2896 struct sock *sk);
2897struct net_device *dev_get_by_index(struct net *net, int ifindex);
2898struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2899struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2900struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2901int netdev_get_name(struct net *net, char *name, int ifindex);
2902int dev_restart(struct net_device *dev);
2903int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2904int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2905
2906static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2907{
2908 return NAPI_GRO_CB(skb)->data_offset;
2909}
2910
2911static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2912{
2913 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2914}
2915
2916static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2917{
2918 NAPI_GRO_CB(skb)->data_offset += len;
2919}
2920
2921static inline void *skb_gro_header_fast(struct sk_buff *skb,
2922 unsigned int offset)
2923{
2924 return NAPI_GRO_CB(skb)->frag0 + offset;
2925}
2926
2927static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2928{
2929 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2930}
2931
2932static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2933{
2934 NAPI_GRO_CB(skb)->frag0 = NULL;
2935 NAPI_GRO_CB(skb)->frag0_len = 0;
2936}
2937
2938static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2939 unsigned int offset)
2940{
2941 if (!pskb_may_pull(skb, hlen))
2942 return NULL;
2943
2944 skb_gro_frag0_invalidate(skb);
2945 return skb->data + offset;
2946}
2947
2948static inline void *skb_gro_network_header(struct sk_buff *skb)
2949{
2950 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2951 skb_network_offset(skb);
2952}
2953
2954static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2955 const void *start, unsigned int len)
2956{
2957 if (NAPI_GRO_CB(skb)->csum_valid)
2958 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2959 csum_partial(start, len, 0));
2960}
2961
2962/* GRO checksum functions. These are logical equivalents of the normal
2963 * checksum functions (in skbuff.h) except that they operate on the GRO
2964 * offsets and fields in sk_buff.
2965 */
2966
2967__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2968
2969static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2970{
2971 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2972}
2973
2974static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2975 bool zero_okay,
2976 __sum16 check)
2977{
2978 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2979 skb_checksum_start_offset(skb) <
2980 skb_gro_offset(skb)) &&
2981 !skb_at_gro_remcsum_start(skb) &&
2982 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2983 (!zero_okay || check));
2984}
2985
2986static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2987 __wsum psum)
2988{
2989 if (NAPI_GRO_CB(skb)->csum_valid &&
2990 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2991 return 0;
2992
2993 NAPI_GRO_CB(skb)->csum = psum;
2994
2995 return __skb_gro_checksum_complete(skb);
2996}
2997
2998static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2999{
3000 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3001 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3002 NAPI_GRO_CB(skb)->csum_cnt--;
3003 } else {
3004 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3005 * verified a new top level checksum or an encapsulated one
3006 * during GRO. This saves work if we fallback to normal path.
3007 */
3008 __skb_incr_checksum_unnecessary(skb);
3009 }
3010}
3011
3012#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3013 compute_pseudo) \
3014({ \
3015 __sum16 __ret = 0; \
3016 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3017 __ret = __skb_gro_checksum_validate_complete(skb, \
3018 compute_pseudo(skb, proto)); \
3019 if (!__ret) \
3020 skb_gro_incr_csum_unnecessary(skb); \
3021 __ret; \
3022})
3023
3024#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3025 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3026
3027#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3028 compute_pseudo) \
3029 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3030
3031#define skb_gro_checksum_simple_validate(skb) \
3032 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3033
3034static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3035{
3036 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3037 !NAPI_GRO_CB(skb)->csum_valid);
3038}
3039
3040static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3041 __wsum pseudo)
3042{
3043 NAPI_GRO_CB(skb)->csum = ~pseudo;
3044 NAPI_GRO_CB(skb)->csum_valid = 1;
3045}
3046
3047#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3048do { \
3049 if (__skb_gro_checksum_convert_check(skb)) \
3050 __skb_gro_checksum_convert(skb, \
3051 compute_pseudo(skb, proto)); \
3052} while (0)
3053
3054struct gro_remcsum {
3055 int offset;
3056 __wsum delta;
3057};
3058
3059static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3060{
3061 grc->offset = 0;
3062 grc->delta = 0;
3063}
3064
3065static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3066 unsigned int off, size_t hdrlen,
3067 int start, int offset,
3068 struct gro_remcsum *grc,
3069 bool nopartial)
3070{
3071 __wsum delta;
3072 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3073
3074 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3075
3076 if (!nopartial) {
3077 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3078 return ptr;
3079 }
3080
3081 ptr = skb_gro_header_fast(skb, off);
3082 if (skb_gro_header_hard(skb, off + plen)) {
3083 ptr = skb_gro_header_slow(skb, off + plen, off);
3084 if (!ptr)
3085 return NULL;
3086 }
3087
3088 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3089 start, offset);
3090
3091 /* Adjust skb->csum since we changed the packet */
3092 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3093
3094 grc->offset = off + hdrlen + offset;
3095 grc->delta = delta;
3096
3097 return ptr;
3098}
3099
3100static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3101 struct gro_remcsum *grc)
3102{
3103 void *ptr;
3104 size_t plen = grc->offset + sizeof(u16);
3105
3106 if (!grc->delta)
3107 return;
3108
3109 ptr = skb_gro_header_fast(skb, grc->offset);
3110 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3111 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3112 if (!ptr)
3113 return;
3114 }
3115
3116 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3117}
3118
3119#ifdef CONFIG_XFRM_OFFLOAD
3120static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3121{
3122 if (PTR_ERR(pp) != -EINPROGRESS)
3123 NAPI_GRO_CB(skb)->flush |= flush;
3124}
3125static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3126 struct sk_buff *pp,
3127 int flush,
3128 struct gro_remcsum *grc)
3129{
3130 if (PTR_ERR(pp) != -EINPROGRESS) {
3131 NAPI_GRO_CB(skb)->flush |= flush;
3132 skb_gro_remcsum_cleanup(skb, grc);
3133 skb->remcsum_offload = 0;
3134 }
3135}
3136#else
3137static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3138{
3139 NAPI_GRO_CB(skb)->flush |= flush;
3140}
3141static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3142 struct sk_buff *pp,
3143 int flush,
3144 struct gro_remcsum *grc)
3145{
3146 NAPI_GRO_CB(skb)->flush |= flush;
3147 skb_gro_remcsum_cleanup(skb, grc);
3148 skb->remcsum_offload = 0;
3149}
3150#endif
3151
3152static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3153 unsigned short type,
3154 const void *daddr, const void *saddr,
3155 unsigned int len)
3156{
3157 if (!dev->header_ops || !dev->header_ops->create)
3158 return 0;
3159
3160 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3161}
3162
3163static inline int dev_parse_header(const struct sk_buff *skb,
3164 unsigned char *haddr)
3165{
3166 const struct net_device *dev = skb->dev;
3167
3168 if (!dev->header_ops || !dev->header_ops->parse)
3169 return 0;
3170 return dev->header_ops->parse(skb, haddr);
3171}
3172
3173static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3174{
3175 const struct net_device *dev = skb->dev;
3176
3177 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3178 return 0;
3179 return dev->header_ops->parse_protocol(skb);
3180}
3181
3182/* ll_header must have at least hard_header_len allocated */
3183static inline bool dev_validate_header(const struct net_device *dev,
3184 char *ll_header, int len)
3185{
3186 if (likely(len >= dev->hard_header_len))
3187 return true;
3188 if (len < dev->min_header_len)
3189 return false;
3190
3191 if (capable(CAP_SYS_RAWIO)) {
3192 memset(ll_header + len, 0, dev->hard_header_len - len);
3193 return true;
3194 }
3195
3196 if (dev->header_ops && dev->header_ops->validate)
3197 return dev->header_ops->validate(ll_header, len);
3198
3199 return false;
3200}
3201
3202static inline bool dev_has_header(const struct net_device *dev)
3203{
3204 return dev->header_ops && dev->header_ops->create;
3205}
3206
3207typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3208 int len, int size);
3209int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3210static inline int unregister_gifconf(unsigned int family)
3211{
3212 return register_gifconf(family, NULL);
3213}
3214
3215#ifdef CONFIG_NET_FLOW_LIMIT
3216#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3217struct sd_flow_limit {
3218 u64 count;
3219 unsigned int num_buckets;
3220 unsigned int history_head;
3221 u16 history[FLOW_LIMIT_HISTORY];
3222 u8 buckets[];
3223};
3224
3225extern int netdev_flow_limit_table_len;
3226#endif /* CONFIG_NET_FLOW_LIMIT */
3227
3228/*
3229 * Incoming packets are placed on per-CPU queues
3230 */
3231struct softnet_data {
3232 struct list_head poll_list;
3233 struct sk_buff_head process_queue;
3234
3235 /* stats */
3236 unsigned int processed;
3237 unsigned int time_squeeze;
3238 unsigned int received_rps;
3239#ifdef CONFIG_RPS
3240 struct softnet_data *rps_ipi_list;
3241#endif
3242#ifdef CONFIG_NET_FLOW_LIMIT
3243 struct sd_flow_limit __rcu *flow_limit;
3244#endif
3245 struct Qdisc *output_queue;
3246 struct Qdisc **output_queue_tailp;
3247 struct sk_buff *completion_queue;
3248#ifdef CONFIG_XFRM_OFFLOAD
3249 struct sk_buff_head xfrm_backlog;
3250#endif
3251 /* written and read only by owning cpu: */
3252 struct {
3253 u16 recursion;
3254 u8 more;
3255 } xmit;
3256#ifdef CONFIG_RPS
3257 /* input_queue_head should be written by cpu owning this struct,
3258 * and only read by other cpus. Worth using a cache line.
3259 */
3260 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3261
3262 /* Elements below can be accessed between CPUs for RPS/RFS */
3263 call_single_data_t csd ____cacheline_aligned_in_smp;
3264 struct softnet_data *rps_ipi_next;
3265 unsigned int cpu;
3266 unsigned int input_queue_tail;
3267#endif
3268 unsigned int dropped;
3269 struct sk_buff_head input_pkt_queue;
3270 struct napi_struct backlog;
3271
3272};
3273
3274static inline void input_queue_head_incr(struct softnet_data *sd)
3275{
3276#ifdef CONFIG_RPS
3277 sd->input_queue_head++;
3278#endif
3279}
3280
3281static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3282 unsigned int *qtail)
3283{
3284#ifdef CONFIG_RPS
3285 *qtail = ++sd->input_queue_tail;
3286#endif
3287}
3288
3289DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3290
3291static inline int dev_recursion_level(void)
3292{
3293 return this_cpu_read(softnet_data.xmit.recursion);
3294}
3295
3296#define XMIT_RECURSION_LIMIT 8
3297static inline bool dev_xmit_recursion(void)
3298{
3299 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3300 XMIT_RECURSION_LIMIT);
3301}
3302
3303static inline void dev_xmit_recursion_inc(void)
3304{
3305 __this_cpu_inc(softnet_data.xmit.recursion);
3306}
3307
3308static inline void dev_xmit_recursion_dec(void)
3309{
3310 __this_cpu_dec(softnet_data.xmit.recursion);
3311}
3312
3313void __netif_schedule(struct Qdisc *q);
3314void netif_schedule_queue(struct netdev_queue *txq);
3315
3316static inline void netif_tx_schedule_all(struct net_device *dev)
3317{
3318 unsigned int i;
3319
3320 for (i = 0; i < dev->num_tx_queues; i++)
3321 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3322}
3323
3324static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3325{
3326 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3327}
3328
3329/**
3330 * netif_start_queue - allow transmit
3331 * @dev: network device
3332 *
3333 * Allow upper layers to call the device hard_start_xmit routine.
3334 */
3335static inline void netif_start_queue(struct net_device *dev)
3336{
3337 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3338}
3339
3340static inline void netif_tx_start_all_queues(struct net_device *dev)
3341{
3342 unsigned int i;
3343
3344 for (i = 0; i < dev->num_tx_queues; i++) {
3345 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3346 netif_tx_start_queue(txq);
3347 }
3348}
3349
3350void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3351
3352/**
3353 * netif_wake_queue - restart transmit
3354 * @dev: network device
3355 *
3356 * Allow upper layers to call the device hard_start_xmit routine.
3357 * Used for flow control when transmit resources are available.
3358 */
3359static inline void netif_wake_queue(struct net_device *dev)
3360{
3361 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3362}
3363
3364static inline void netif_tx_wake_all_queues(struct net_device *dev)
3365{
3366 unsigned int i;
3367
3368 for (i = 0; i < dev->num_tx_queues; i++) {
3369 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3370 netif_tx_wake_queue(txq);
3371 }
3372}
3373
3374static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3375{
3376 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3377}
3378
3379/**
3380 * netif_stop_queue - stop transmitted packets
3381 * @dev: network device
3382 *
3383 * Stop upper layers calling the device hard_start_xmit routine.
3384 * Used for flow control when transmit resources are unavailable.
3385 */
3386static inline void netif_stop_queue(struct net_device *dev)
3387{
3388 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3389}
3390
3391void netif_tx_stop_all_queues(struct net_device *dev);
3392
3393static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3394{
3395 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3396}
3397
3398/**
3399 * netif_queue_stopped - test if transmit queue is flowblocked
3400 * @dev: network device
3401 *
3402 * Test if transmit queue on device is currently unable to send.
3403 */
3404static inline bool netif_queue_stopped(const struct net_device *dev)
3405{
3406 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3407}
3408
3409static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3410{
3411 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3412}
3413
3414static inline bool
3415netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3416{
3417 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3418}
3419
3420static inline bool
3421netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3422{
3423 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3424}
3425
3426/**
3427 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3428 * @dev_queue: pointer to transmit queue
3429 *
3430 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3431 * to give appropriate hint to the CPU.
3432 */
3433static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3434{
3435#ifdef CONFIG_BQL
3436 prefetchw(&dev_queue->dql.num_queued);
3437#endif
3438}
3439
3440/**
3441 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3442 * @dev_queue: pointer to transmit queue
3443 *
3444 * BQL enabled drivers might use this helper in their TX completion path,
3445 * to give appropriate hint to the CPU.
3446 */
3447static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3448{
3449#ifdef CONFIG_BQL
3450 prefetchw(&dev_queue->dql.limit);
3451#endif
3452}
3453
3454static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3455 unsigned int bytes)
3456{
3457#ifdef CONFIG_BQL
3458 dql_queued(&dev_queue->dql, bytes);
3459
3460 if (likely(dql_avail(&dev_queue->dql) >= 0))
3461 return;
3462
3463 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3464
3465 /*
3466 * The XOFF flag must be set before checking the dql_avail below,
3467 * because in netdev_tx_completed_queue we update the dql_completed
3468 * before checking the XOFF flag.
3469 */
3470 smp_mb();
3471
3472 /* check again in case another CPU has just made room avail */
3473 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3474 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3475#endif
3476}
3477
3478/* Variant of netdev_tx_sent_queue() for drivers that are aware
3479 * that they should not test BQL status themselves.
3480 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3481 * skb of a batch.
3482 * Returns true if the doorbell must be used to kick the NIC.
3483 */
3484static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3485 unsigned int bytes,
3486 bool xmit_more)
3487{
3488 if (xmit_more) {
3489#ifdef CONFIG_BQL
3490 dql_queued(&dev_queue->dql, bytes);
3491#endif
3492 return netif_tx_queue_stopped(dev_queue);
3493 }
3494 netdev_tx_sent_queue(dev_queue, bytes);
3495 return true;
3496}
3497
3498/**
3499 * netdev_sent_queue - report the number of bytes queued to hardware
3500 * @dev: network device
3501 * @bytes: number of bytes queued to the hardware device queue
3502 *
3503 * Report the number of bytes queued for sending/completion to the network
3504 * device hardware queue. @bytes should be a good approximation and should
3505 * exactly match netdev_completed_queue() @bytes
3506 */
3507static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3508{
3509 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3510}
3511
3512static inline bool __netdev_sent_queue(struct net_device *dev,
3513 unsigned int bytes,
3514 bool xmit_more)
3515{
3516 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3517 xmit_more);
3518}
3519
3520static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3521 unsigned int pkts, unsigned int bytes)
3522{
3523#ifdef CONFIG_BQL
3524 if (unlikely(!bytes))
3525 return;
3526
3527 dql_completed(&dev_queue->dql, bytes);
3528
3529 /*
3530 * Without the memory barrier there is a small possiblity that
3531 * netdev_tx_sent_queue will miss the update and cause the queue to
3532 * be stopped forever
3533 */
3534 smp_mb();
3535
3536 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3537 return;
3538
3539 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3540 netif_schedule_queue(dev_queue);
3541#endif
3542}
3543
3544/**
3545 * netdev_completed_queue - report bytes and packets completed by device
3546 * @dev: network device
3547 * @pkts: actual number of packets sent over the medium
3548 * @bytes: actual number of bytes sent over the medium
3549 *
3550 * Report the number of bytes and packets transmitted by the network device
3551 * hardware queue over the physical medium, @bytes must exactly match the
3552 * @bytes amount passed to netdev_sent_queue()
3553 */
3554static inline void netdev_completed_queue(struct net_device *dev,
3555 unsigned int pkts, unsigned int bytes)
3556{
3557 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3558}
3559
3560static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3561{
3562#ifdef CONFIG_BQL
3563 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3564 dql_reset(&q->dql);
3565#endif
3566}
3567
3568/**
3569 * netdev_reset_queue - reset the packets and bytes count of a network device
3570 * @dev_queue: network device
3571 *
3572 * Reset the bytes and packet count of a network device and clear the
3573 * software flow control OFF bit for this network device
3574 */
3575static inline void netdev_reset_queue(struct net_device *dev_queue)
3576{
3577 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3578}
3579
3580/**
3581 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3582 * @dev: network device
3583 * @queue_index: given tx queue index
3584 *
3585 * Returns 0 if given tx queue index >= number of device tx queues,
3586 * otherwise returns the originally passed tx queue index.
3587 */
3588static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3589{
3590 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3591 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3592 dev->name, queue_index,
3593 dev->real_num_tx_queues);
3594 return 0;
3595 }
3596
3597 return queue_index;
3598}
3599
3600/**
3601 * netif_running - test if up
3602 * @dev: network device
3603 *
3604 * Test if the device has been brought up.
3605 */
3606static inline bool netif_running(const struct net_device *dev)
3607{
3608 return test_bit(__LINK_STATE_START, &dev->state);
3609}
3610
3611/*
3612 * Routines to manage the subqueues on a device. We only need start,
3613 * stop, and a check if it's stopped. All other device management is
3614 * done at the overall netdevice level.
3615 * Also test the device if we're multiqueue.
3616 */
3617
3618/**
3619 * netif_start_subqueue - allow sending packets on subqueue
3620 * @dev: network device
3621 * @queue_index: sub queue index
3622 *
3623 * Start individual transmit queue of a device with multiple transmit queues.
3624 */
3625static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3626{
3627 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3628
3629 netif_tx_start_queue(txq);
3630}
3631
3632/**
3633 * netif_stop_subqueue - stop sending packets on subqueue
3634 * @dev: network device
3635 * @queue_index: sub queue index
3636 *
3637 * Stop individual transmit queue of a device with multiple transmit queues.
3638 */
3639static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3640{
3641 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3642 netif_tx_stop_queue(txq);
3643}
3644
3645/**
3646 * __netif_subqueue_stopped - test status of subqueue
3647 * @dev: network device
3648 * @queue_index: sub queue index
3649 *
3650 * Check individual transmit queue of a device with multiple transmit queues.
3651 */
3652static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3653 u16 queue_index)
3654{
3655 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3656
3657 return netif_tx_queue_stopped(txq);
3658}
3659
3660/**
3661 * netif_subqueue_stopped - test status of subqueue
3662 * @dev: network device
3663 * @skb: sub queue buffer pointer
3664 *
3665 * Check individual transmit queue of a device with multiple transmit queues.
3666 */
3667static inline bool netif_subqueue_stopped(const struct net_device *dev,
3668 struct sk_buff *skb)
3669{
3670 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3671}
3672
3673/**
3674 * netif_wake_subqueue - allow sending packets on subqueue
3675 * @dev: network device
3676 * @queue_index: sub queue index
3677 *
3678 * Resume individual transmit queue of a device with multiple transmit queues.
3679 */
3680static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3681{
3682 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3683
3684 netif_tx_wake_queue(txq);
3685}
3686
3687#ifdef CONFIG_XPS
3688int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3689 u16 index);
3690int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3691 u16 index, bool is_rxqs_map);
3692
3693/**
3694 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3695 * @j: CPU/Rx queue index
3696 * @mask: bitmask of all cpus/rx queues
3697 * @nr_bits: number of bits in the bitmask
3698 *
3699 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3700 */
3701static inline bool netif_attr_test_mask(unsigned long j,
3702 const unsigned long *mask,
3703 unsigned int nr_bits)
3704{
3705 cpu_max_bits_warn(j, nr_bits);
3706 return test_bit(j, mask);
3707}
3708
3709/**
3710 * netif_attr_test_online - Test for online CPU/Rx queue
3711 * @j: CPU/Rx queue index
3712 * @online_mask: bitmask for CPUs/Rx queues that are online
3713 * @nr_bits: number of bits in the bitmask
3714 *
3715 * Returns true if a CPU/Rx queue is online.
3716 */
3717static inline bool netif_attr_test_online(unsigned long j,
3718 const unsigned long *online_mask,
3719 unsigned int nr_bits)
3720{
3721 cpu_max_bits_warn(j, nr_bits);
3722
3723 if (online_mask)
3724 return test_bit(j, online_mask);
3725
3726 return (j < nr_bits);
3727}
3728
3729/**
3730 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3731 * @n: CPU/Rx queue index
3732 * @srcp: the cpumask/Rx queue mask pointer
3733 * @nr_bits: number of bits in the bitmask
3734 *
3735 * Returns >= nr_bits if no further CPUs/Rx queues set.
3736 */
3737static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3738 unsigned int nr_bits)
3739{
3740 /* -1 is a legal arg here. */
3741 if (n != -1)
3742 cpu_max_bits_warn(n, nr_bits);
3743
3744 if (srcp)
3745 return find_next_bit(srcp, nr_bits, n + 1);
3746
3747 return n + 1;
3748}
3749
3750/**
3751 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3752 * @n: CPU/Rx queue index
3753 * @src1p: the first CPUs/Rx queues mask pointer
3754 * @src2p: the second CPUs/Rx queues mask pointer
3755 * @nr_bits: number of bits in the bitmask
3756 *
3757 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3758 */
3759static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3760 const unsigned long *src2p,
3761 unsigned int nr_bits)
3762{
3763 /* -1 is a legal arg here. */
3764 if (n != -1)
3765 cpu_max_bits_warn(n, nr_bits);
3766
3767 if (src1p && src2p)
3768 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3769 else if (src1p)
3770 return find_next_bit(src1p, nr_bits, n + 1);
3771 else if (src2p)
3772 return find_next_bit(src2p, nr_bits, n + 1);
3773
3774 return n + 1;
3775}
3776#else
3777static inline int netif_set_xps_queue(struct net_device *dev,
3778 const struct cpumask *mask,
3779 u16 index)
3780{
3781 return 0;
3782}
3783
3784static inline int __netif_set_xps_queue(struct net_device *dev,
3785 const unsigned long *mask,
3786 u16 index, bool is_rxqs_map)
3787{
3788 return 0;
3789}
3790#endif
3791
3792/**
3793 * netif_is_multiqueue - test if device has multiple transmit queues
3794 * @dev: network device
3795 *
3796 * Check if device has multiple transmit queues
3797 */
3798static inline bool netif_is_multiqueue(const struct net_device *dev)
3799{
3800 return dev->num_tx_queues > 1;
3801}
3802
3803int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3804
3805#ifdef CONFIG_SYSFS
3806int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3807#else
3808static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3809 unsigned int rxqs)
3810{
3811 dev->real_num_rx_queues = rxqs;
3812 return 0;
3813}
3814#endif
3815
3816static inline struct netdev_rx_queue *
3817__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3818{
3819 return dev->_rx + rxq;
3820}
3821
3822#ifdef CONFIG_SYSFS
3823static inline unsigned int get_netdev_rx_queue_index(
3824 struct netdev_rx_queue *queue)
3825{
3826 struct net_device *dev = queue->dev;
3827 int index = queue - dev->_rx;
3828
3829 BUG_ON(index >= dev->num_rx_queues);
3830 return index;
3831}
3832#endif
3833
3834#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3835int netif_get_num_default_rss_queues(void);
3836
3837enum skb_free_reason {
3838 SKB_REASON_CONSUMED,
3839 SKB_REASON_DROPPED,
3840};
3841
3842void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3843void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3844
3845/*
3846 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3847 * interrupt context or with hardware interrupts being disabled.
3848 * (in_irq() || irqs_disabled())
3849 *
3850 * We provide four helpers that can be used in following contexts :
3851 *
3852 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3853 * replacing kfree_skb(skb)
3854 *
3855 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3856 * Typically used in place of consume_skb(skb) in TX completion path
3857 *
3858 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3859 * replacing kfree_skb(skb)
3860 *
3861 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3862 * and consumed a packet. Used in place of consume_skb(skb)
3863 */
3864static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3865{
3866 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3867}
3868
3869static inline void dev_consume_skb_irq(struct sk_buff *skb)
3870{
3871 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3872}
3873
3874static inline void dev_kfree_skb_any(struct sk_buff *skb)
3875{
3876 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3877}
3878
3879static inline void dev_consume_skb_any(struct sk_buff *skb)
3880{
3881 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3882}
3883
3884void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3885int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3886int netif_rx(struct sk_buff *skb);
3887int netif_rx_ni(struct sk_buff *skb);
3888int netif_rx_any_context(struct sk_buff *skb);
3889int netif_receive_skb(struct sk_buff *skb);
3890int netif_receive_skb_core(struct sk_buff *skb);
3891void netif_receive_skb_list(struct list_head *head);
3892gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3893void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3894struct sk_buff *napi_get_frags(struct napi_struct *napi);
3895gro_result_t napi_gro_frags(struct napi_struct *napi);
3896struct packet_offload *gro_find_receive_by_type(__be16 type);
3897struct packet_offload *gro_find_complete_by_type(__be16 type);
3898
3899static inline void napi_free_frags(struct napi_struct *napi)
3900{
3901 kfree_skb(napi->skb);
3902 napi->skb = NULL;
3903}
3904
3905bool netdev_is_rx_handler_busy(struct net_device *dev);
3906int netdev_rx_handler_register(struct net_device *dev,
3907 rx_handler_func_t *rx_handler,
3908 void *rx_handler_data);
3909void netdev_rx_handler_unregister(struct net_device *dev);
3910
3911bool dev_valid_name(const char *name);
3912int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3913 bool *need_copyout);
3914int dev_ifconf(struct net *net, struct ifconf *, int);
3915int dev_ethtool(struct net *net, struct ifreq *);
3916unsigned int dev_get_flags(const struct net_device *);
3917int __dev_change_flags(struct net_device *dev, unsigned int flags,
3918 struct netlink_ext_ack *extack);
3919int dev_change_flags(struct net_device *dev, unsigned int flags,
3920 struct netlink_ext_ack *extack);
3921void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3922 unsigned int gchanges);
3923int dev_change_name(struct net_device *, const char *);
3924int dev_set_alias(struct net_device *, const char *, size_t);
3925int dev_get_alias(const struct net_device *, char *, size_t);
3926int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3927int __dev_set_mtu(struct net_device *, int);
3928int dev_validate_mtu(struct net_device *dev, int mtu,
3929 struct netlink_ext_ack *extack);
3930int dev_set_mtu_ext(struct net_device *dev, int mtu,
3931 struct netlink_ext_ack *extack);
3932int dev_set_mtu(struct net_device *, int);
3933int dev_change_tx_queue_len(struct net_device *, unsigned long);
3934void dev_set_group(struct net_device *, int);
3935int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3936 struct netlink_ext_ack *extack);
3937int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3938 struct netlink_ext_ack *extack);
3939int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3940 struct netlink_ext_ack *extack);
3941int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3942int dev_change_carrier(struct net_device *, bool new_carrier);
3943int dev_get_phys_port_id(struct net_device *dev,
3944 struct netdev_phys_item_id *ppid);
3945int dev_get_phys_port_name(struct net_device *dev,
3946 char *name, size_t len);
3947int dev_get_port_parent_id(struct net_device *dev,
3948 struct netdev_phys_item_id *ppid, bool recurse);
3949bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3950int dev_change_proto_down(struct net_device *dev, bool proto_down);
3951int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3952void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3953 u32 value);
3954struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3955struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3956 struct netdev_queue *txq, int *ret);
3957
3958typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3959int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3960 int fd, int expected_fd, u32 flags);
3961int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3962u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3963
3964int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3965int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3966int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3967bool is_skb_forwardable(const struct net_device *dev,
3968 const struct sk_buff *skb);
3969
3970static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3971 const struct sk_buff *skb,
3972 const bool check_mtu)
3973{
3974 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3975 unsigned int len;
3976
3977 if (!(dev->flags & IFF_UP))
3978 return false;
3979
3980 if (!check_mtu)
3981 return true;
3982
3983 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3984 if (skb->len <= len)
3985 return true;
3986
3987 /* if TSO is enabled, we don't care about the length as the packet
3988 * could be forwarded without being segmented before
3989 */
3990 if (skb_is_gso(skb))
3991 return true;
3992
3993 return false;
3994}
3995
3996static __always_inline int ____dev_forward_skb(struct net_device *dev,
3997 struct sk_buff *skb,
3998 const bool check_mtu)
3999{
4000 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4001 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4002 atomic_long_inc(&dev->rx_dropped);
4003 kfree_skb(skb);
4004 return NET_RX_DROP;
4005 }
4006
4007 skb_scrub_packet(skb, true);
4008 skb->priority = 0;
4009 return 0;
4010}
4011
4012bool dev_nit_active(struct net_device *dev);
4013void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4014
4015extern int netdev_budget;
4016extern unsigned int netdev_budget_usecs;
4017
4018/* Called by rtnetlink.c:rtnl_unlock() */
4019void netdev_run_todo(void);
4020
4021/**
4022 * dev_put - release reference to device
4023 * @dev: network device
4024 *
4025 * Release reference to device to allow it to be freed.
4026 */
4027static inline void dev_put(struct net_device *dev)
4028{
4029 this_cpu_dec(*dev->pcpu_refcnt);
4030}
4031
4032/**
4033 * dev_hold - get reference to device
4034 * @dev: network device
4035 *
4036 * Hold reference to device to keep it from being freed.
4037 */
4038static inline void dev_hold(struct net_device *dev)
4039{
4040 this_cpu_inc(*dev->pcpu_refcnt);
4041}
4042
4043/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4044 * and _off may be called from IRQ context, but it is caller
4045 * who is responsible for serialization of these calls.
4046 *
4047 * The name carrier is inappropriate, these functions should really be
4048 * called netif_lowerlayer_*() because they represent the state of any
4049 * kind of lower layer not just hardware media.
4050 */
4051
4052void linkwatch_init_dev(struct net_device *dev);
4053void linkwatch_fire_event(struct net_device *dev);
4054void linkwatch_forget_dev(struct net_device *dev);
4055
4056/**
4057 * netif_carrier_ok - test if carrier present
4058 * @dev: network device
4059 *
4060 * Check if carrier is present on device
4061 */
4062static inline bool netif_carrier_ok(const struct net_device *dev)
4063{
4064 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4065}
4066
4067unsigned long dev_trans_start(struct net_device *dev);
4068
4069void __netdev_watchdog_up(struct net_device *dev);
4070
4071void netif_carrier_on(struct net_device *dev);
4072
4073void netif_carrier_off(struct net_device *dev);
4074
4075/**
4076 * netif_dormant_on - mark device as dormant.
4077 * @dev: network device
4078 *
4079 * Mark device as dormant (as per RFC2863).
4080 *
4081 * The dormant state indicates that the relevant interface is not
4082 * actually in a condition to pass packets (i.e., it is not 'up') but is
4083 * in a "pending" state, waiting for some external event. For "on-
4084 * demand" interfaces, this new state identifies the situation where the
4085 * interface is waiting for events to place it in the up state.
4086 */
4087static inline void netif_dormant_on(struct net_device *dev)
4088{
4089 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4090 linkwatch_fire_event(dev);
4091}
4092
4093/**
4094 * netif_dormant_off - set device as not dormant.
4095 * @dev: network device
4096 *
4097 * Device is not in dormant state.
4098 */
4099static inline void netif_dormant_off(struct net_device *dev)
4100{
4101 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4102 linkwatch_fire_event(dev);
4103}
4104
4105/**
4106 * netif_dormant - test if device is dormant
4107 * @dev: network device
4108 *
4109 * Check if device is dormant.
4110 */
4111static inline bool netif_dormant(const struct net_device *dev)
4112{
4113 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4114}
4115
4116
4117/**
4118 * netif_testing_on - mark device as under test.
4119 * @dev: network device
4120 *
4121 * Mark device as under test (as per RFC2863).
4122 *
4123 * The testing state indicates that some test(s) must be performed on
4124 * the interface. After completion, of the test, the interface state
4125 * will change to up, dormant, or down, as appropriate.
4126 */
4127static inline void netif_testing_on(struct net_device *dev)
4128{
4129 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4130 linkwatch_fire_event(dev);
4131}
4132
4133/**
4134 * netif_testing_off - set device as not under test.
4135 * @dev: network device
4136 *
4137 * Device is not in testing state.
4138 */
4139static inline void netif_testing_off(struct net_device *dev)
4140{
4141 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4142 linkwatch_fire_event(dev);
4143}
4144
4145/**
4146 * netif_testing - test if device is under test
4147 * @dev: network device
4148 *
4149 * Check if device is under test
4150 */
4151static inline bool netif_testing(const struct net_device *dev)
4152{
4153 return test_bit(__LINK_STATE_TESTING, &dev->state);
4154}
4155
4156
4157/**
4158 * netif_oper_up - test if device is operational
4159 * @dev: network device
4160 *
4161 * Check if carrier is operational
4162 */
4163static inline bool netif_oper_up(const struct net_device *dev)
4164{
4165 return (dev->operstate == IF_OPER_UP ||
4166 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4167}
4168
4169/**
4170 * netif_device_present - is device available or removed
4171 * @dev: network device
4172 *
4173 * Check if device has not been removed from system.
4174 */
4175static inline bool netif_device_present(struct net_device *dev)
4176{
4177 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4178}
4179
4180void netif_device_detach(struct net_device *dev);
4181
4182void netif_device_attach(struct net_device *dev);
4183
4184/*
4185 * Network interface message level settings
4186 */
4187
4188enum {
4189 NETIF_MSG_DRV_BIT,
4190 NETIF_MSG_PROBE_BIT,
4191 NETIF_MSG_LINK_BIT,
4192 NETIF_MSG_TIMER_BIT,
4193 NETIF_MSG_IFDOWN_BIT,
4194 NETIF_MSG_IFUP_BIT,
4195 NETIF_MSG_RX_ERR_BIT,
4196 NETIF_MSG_TX_ERR_BIT,
4197 NETIF_MSG_TX_QUEUED_BIT,
4198 NETIF_MSG_INTR_BIT,
4199 NETIF_MSG_TX_DONE_BIT,
4200 NETIF_MSG_RX_STATUS_BIT,
4201 NETIF_MSG_PKTDATA_BIT,
4202 NETIF_MSG_HW_BIT,
4203 NETIF_MSG_WOL_BIT,
4204
4205 /* When you add a new bit above, update netif_msg_class_names array
4206 * in net/ethtool/common.c
4207 */
4208 NETIF_MSG_CLASS_COUNT,
4209};
4210/* Both ethtool_ops interface and internal driver implementation use u32 */
4211static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4212
4213#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4214#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4215
4216#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4217#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4218#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4219#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4220#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4221#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4222#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4223#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4224#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4225#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4226#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4227#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4228#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4229#define NETIF_MSG_HW __NETIF_MSG(HW)
4230#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4231
4232#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4233#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4234#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4235#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4236#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4237#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4238#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4239#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4240#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4241#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4242#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4243#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4244#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4245#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4246#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4247
4248static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4249{
4250 /* use default */
4251 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4252 return default_msg_enable_bits;
4253 if (debug_value == 0) /* no output */
4254 return 0;
4255 /* set low N bits */
4256 return (1U << debug_value) - 1;
4257}
4258
4259static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4260{
4261 spin_lock(&txq->_xmit_lock);
4262 txq->xmit_lock_owner = cpu;
4263}
4264
4265static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4266{
4267 __acquire(&txq->_xmit_lock);
4268 return true;
4269}
4270
4271static inline void __netif_tx_release(struct netdev_queue *txq)
4272{
4273 __release(&txq->_xmit_lock);
4274}
4275
4276static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4277{
4278 spin_lock_bh(&txq->_xmit_lock);
4279 txq->xmit_lock_owner = smp_processor_id();
4280}
4281
4282static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4283{
4284 bool ok = spin_trylock(&txq->_xmit_lock);
4285 if (likely(ok))
4286 txq->xmit_lock_owner = smp_processor_id();
4287 return ok;
4288}
4289
4290static inline void __netif_tx_unlock(struct netdev_queue *txq)
4291{
4292 txq->xmit_lock_owner = -1;
4293 spin_unlock(&txq->_xmit_lock);
4294}
4295
4296static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4297{
4298 txq->xmit_lock_owner = -1;
4299 spin_unlock_bh(&txq->_xmit_lock);
4300}
4301
4302static inline void txq_trans_update(struct netdev_queue *txq)
4303{
4304 if (txq->xmit_lock_owner != -1)
4305 txq->trans_start = jiffies;
4306}
4307
4308/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4309static inline void netif_trans_update(struct net_device *dev)
4310{
4311 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4312
4313 if (txq->trans_start != jiffies)
4314 txq->trans_start = jiffies;
4315}
4316
4317/**
4318 * netif_tx_lock - grab network device transmit lock
4319 * @dev: network device
4320 *
4321 * Get network device transmit lock
4322 */
4323static inline void netif_tx_lock(struct net_device *dev)
4324{
4325 unsigned int i;
4326 int cpu;
4327
4328 spin_lock(&dev->tx_global_lock);
4329 cpu = smp_processor_id();
4330 for (i = 0; i < dev->num_tx_queues; i++) {
4331 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4332
4333 /* We are the only thread of execution doing a
4334 * freeze, but we have to grab the _xmit_lock in
4335 * order to synchronize with threads which are in
4336 * the ->hard_start_xmit() handler and already
4337 * checked the frozen bit.
4338 */
4339 __netif_tx_lock(txq, cpu);
4340 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4341 __netif_tx_unlock(txq);
4342 }
4343}
4344
4345static inline void netif_tx_lock_bh(struct net_device *dev)
4346{
4347 local_bh_disable();
4348 netif_tx_lock(dev);
4349}
4350
4351static inline void netif_tx_unlock(struct net_device *dev)
4352{
4353 unsigned int i;
4354
4355 for (i = 0; i < dev->num_tx_queues; i++) {
4356 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4357
4358 /* No need to grab the _xmit_lock here. If the
4359 * queue is not stopped for another reason, we
4360 * force a schedule.
4361 */
4362 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4363 netif_schedule_queue(txq);
4364 }
4365 spin_unlock(&dev->tx_global_lock);
4366}
4367
4368static inline void netif_tx_unlock_bh(struct net_device *dev)
4369{
4370 netif_tx_unlock(dev);
4371 local_bh_enable();
4372}
4373
4374#define HARD_TX_LOCK(dev, txq, cpu) { \
4375 if ((dev->features & NETIF_F_LLTX) == 0) { \
4376 __netif_tx_lock(txq, cpu); \
4377 } else { \
4378 __netif_tx_acquire(txq); \
4379 } \
4380}
4381
4382#define HARD_TX_TRYLOCK(dev, txq) \
4383 (((dev->features & NETIF_F_LLTX) == 0) ? \
4384 __netif_tx_trylock(txq) : \
4385 __netif_tx_acquire(txq))
4386
4387#define HARD_TX_UNLOCK(dev, txq) { \
4388 if ((dev->features & NETIF_F_LLTX) == 0) { \
4389 __netif_tx_unlock(txq); \
4390 } else { \
4391 __netif_tx_release(txq); \
4392 } \
4393}
4394
4395static inline void netif_tx_disable(struct net_device *dev)
4396{
4397 unsigned int i;
4398 int cpu;
4399
4400 local_bh_disable();
4401 cpu = smp_processor_id();
4402 spin_lock(&dev->tx_global_lock);
4403 for (i = 0; i < dev->num_tx_queues; i++) {
4404 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4405
4406 __netif_tx_lock(txq, cpu);
4407 netif_tx_stop_queue(txq);
4408 __netif_tx_unlock(txq);
4409 }
4410 spin_unlock(&dev->tx_global_lock);
4411 local_bh_enable();
4412}
4413
4414static inline void netif_addr_lock(struct net_device *dev)
4415{
4416 unsigned char nest_level = 0;
4417
4418#ifdef CONFIG_LOCKDEP
4419 nest_level = dev->nested_level;
4420#endif
4421 spin_lock_nested(&dev->addr_list_lock, nest_level);
4422}
4423
4424static inline void netif_addr_lock_bh(struct net_device *dev)
4425{
4426 unsigned char nest_level = 0;
4427
4428#ifdef CONFIG_LOCKDEP
4429 nest_level = dev->nested_level;
4430#endif
4431 local_bh_disable();
4432 spin_lock_nested(&dev->addr_list_lock, nest_level);
4433}
4434
4435static inline void netif_addr_unlock(struct net_device *dev)
4436{
4437 spin_unlock(&dev->addr_list_lock);
4438}
4439
4440static inline void netif_addr_unlock_bh(struct net_device *dev)
4441{
4442 spin_unlock_bh(&dev->addr_list_lock);
4443}
4444
4445/*
4446 * dev_addrs walker. Should be used only for read access. Call with
4447 * rcu_read_lock held.
4448 */
4449#define for_each_dev_addr(dev, ha) \
4450 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4451
4452/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4453
4454void ether_setup(struct net_device *dev);
4455
4456/* Support for loadable net-drivers */
4457struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4458 unsigned char name_assign_type,
4459 void (*setup)(struct net_device *),
4460 unsigned int txqs, unsigned int rxqs);
4461#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4462 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4463
4464#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4465 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4466 count)
4467
4468int register_netdev(struct net_device *dev);
4469void unregister_netdev(struct net_device *dev);
4470
4471int devm_register_netdev(struct device *dev, struct net_device *ndev);
4472
4473/* General hardware address lists handling functions */
4474int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4475 struct netdev_hw_addr_list *from_list, int addr_len);
4476void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4477 struct netdev_hw_addr_list *from_list, int addr_len);
4478int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4479 struct net_device *dev,
4480 int (*sync)(struct net_device *, const unsigned char *),
4481 int (*unsync)(struct net_device *,
4482 const unsigned char *));
4483int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4484 struct net_device *dev,
4485 int (*sync)(struct net_device *,
4486 const unsigned char *, int),
4487 int (*unsync)(struct net_device *,
4488 const unsigned char *, int));
4489void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4490 struct net_device *dev,
4491 int (*unsync)(struct net_device *,
4492 const unsigned char *, int));
4493void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4494 struct net_device *dev,
4495 int (*unsync)(struct net_device *,
4496 const unsigned char *));
4497void __hw_addr_init(struct netdev_hw_addr_list *list);
4498
4499/* Functions used for device addresses handling */
4500int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4501 unsigned char addr_type);
4502int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4503 unsigned char addr_type);
4504void dev_addr_flush(struct net_device *dev);
4505int dev_addr_init(struct net_device *dev);
4506
4507/* Functions used for unicast addresses handling */
4508int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4509int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4510int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4511int dev_uc_sync(struct net_device *to, struct net_device *from);
4512int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4513void dev_uc_unsync(struct net_device *to, struct net_device *from);
4514void dev_uc_flush(struct net_device *dev);
4515void dev_uc_init(struct net_device *dev);
4516
4517/**
4518 * __dev_uc_sync - Synchonize device's unicast list
4519 * @dev: device to sync
4520 * @sync: function to call if address should be added
4521 * @unsync: function to call if address should be removed
4522 *
4523 * Add newly added addresses to the interface, and release
4524 * addresses that have been deleted.
4525 */
4526static inline int __dev_uc_sync(struct net_device *dev,
4527 int (*sync)(struct net_device *,
4528 const unsigned char *),
4529 int (*unsync)(struct net_device *,
4530 const unsigned char *))
4531{
4532 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4533}
4534
4535/**
4536 * __dev_uc_unsync - Remove synchronized addresses from device
4537 * @dev: device to sync
4538 * @unsync: function to call if address should be removed
4539 *
4540 * Remove all addresses that were added to the device by dev_uc_sync().
4541 */
4542static inline void __dev_uc_unsync(struct net_device *dev,
4543 int (*unsync)(struct net_device *,
4544 const unsigned char *))
4545{
4546 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4547}
4548
4549/* Functions used for multicast addresses handling */
4550int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4551int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4552int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4553int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4554int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4555int dev_mc_sync(struct net_device *to, struct net_device *from);
4556int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4557void dev_mc_unsync(struct net_device *to, struct net_device *from);
4558void dev_mc_flush(struct net_device *dev);
4559void dev_mc_init(struct net_device *dev);
4560
4561/**
4562 * __dev_mc_sync - Synchonize device's multicast list
4563 * @dev: device to sync
4564 * @sync: function to call if address should be added
4565 * @unsync: function to call if address should be removed
4566 *
4567 * Add newly added addresses to the interface, and release
4568 * addresses that have been deleted.
4569 */
4570static inline int __dev_mc_sync(struct net_device *dev,
4571 int (*sync)(struct net_device *,
4572 const unsigned char *),
4573 int (*unsync)(struct net_device *,
4574 const unsigned char *))
4575{
4576 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4577}
4578
4579/**
4580 * __dev_mc_unsync - Remove synchronized addresses from device
4581 * @dev: device to sync
4582 * @unsync: function to call if address should be removed
4583 *
4584 * Remove all addresses that were added to the device by dev_mc_sync().
4585 */
4586static inline void __dev_mc_unsync(struct net_device *dev,
4587 int (*unsync)(struct net_device *,
4588 const unsigned char *))
4589{
4590 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4591}
4592
4593/* Functions used for secondary unicast and multicast support */
4594void dev_set_rx_mode(struct net_device *dev);
4595void __dev_set_rx_mode(struct net_device *dev);
4596int dev_set_promiscuity(struct net_device *dev, int inc);
4597int dev_set_allmulti(struct net_device *dev, int inc);
4598void netdev_state_change(struct net_device *dev);
4599void __netdev_notify_peers(struct net_device *dev);
4600void netdev_notify_peers(struct net_device *dev);
4601void netdev_features_change(struct net_device *dev);
4602/* Load a device via the kmod */
4603void dev_load(struct net *net, const char *name);
4604struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4605 struct rtnl_link_stats64 *storage);
4606void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4607 const struct net_device_stats *netdev_stats);
4608void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4609 const struct pcpu_sw_netstats __percpu *netstats);
4610void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4611
4612extern int netdev_max_backlog;
4613extern int netdev_tstamp_prequeue;
4614extern int weight_p;
4615extern int dev_weight_rx_bias;
4616extern int dev_weight_tx_bias;
4617extern int dev_rx_weight;
4618extern int dev_tx_weight;
4619extern int gro_normal_batch;
4620
4621enum {
4622 NESTED_SYNC_IMM_BIT,
4623 NESTED_SYNC_TODO_BIT,
4624};
4625
4626#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4627#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4628
4629#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4630#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4631
4632struct netdev_nested_priv {
4633 unsigned char flags;
4634 void *data;
4635};
4636
4637bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4638struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4639 struct list_head **iter);
4640struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4641 struct list_head **iter);
4642
4643#ifdef CONFIG_LOCKDEP
4644static LIST_HEAD(net_unlink_list);
4645
4646static inline void net_unlink_todo(struct net_device *dev)
4647{
4648 if (list_empty(&dev->unlink_list))
4649 list_add_tail(&dev->unlink_list, &net_unlink_list);
4650}
4651#endif
4652
4653/* iterate through upper list, must be called under RCU read lock */
4654#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4655 for (iter = &(dev)->adj_list.upper, \
4656 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4657 updev; \
4658 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4659
4660int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4661 int (*fn)(struct net_device *upper_dev,
4662 struct netdev_nested_priv *priv),
4663 struct netdev_nested_priv *priv);
4664
4665bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4666 struct net_device *upper_dev);
4667
4668bool netdev_has_any_upper_dev(struct net_device *dev);
4669
4670void *netdev_lower_get_next_private(struct net_device *dev,
4671 struct list_head **iter);
4672void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4673 struct list_head **iter);
4674
4675#define netdev_for_each_lower_private(dev, priv, iter) \
4676 for (iter = (dev)->adj_list.lower.next, \
4677 priv = netdev_lower_get_next_private(dev, &(iter)); \
4678 priv; \
4679 priv = netdev_lower_get_next_private(dev, &(iter)))
4680
4681#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4682 for (iter = &(dev)->adj_list.lower, \
4683 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4684 priv; \
4685 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4686
4687void *netdev_lower_get_next(struct net_device *dev,
4688 struct list_head **iter);
4689
4690#define netdev_for_each_lower_dev(dev, ldev, iter) \
4691 for (iter = (dev)->adj_list.lower.next, \
4692 ldev = netdev_lower_get_next(dev, &(iter)); \
4693 ldev; \
4694 ldev = netdev_lower_get_next(dev, &(iter)))
4695
4696struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4697 struct list_head **iter);
4698int netdev_walk_all_lower_dev(struct net_device *dev,
4699 int (*fn)(struct net_device *lower_dev,
4700 struct netdev_nested_priv *priv),
4701 struct netdev_nested_priv *priv);
4702int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4703 int (*fn)(struct net_device *lower_dev,
4704 struct netdev_nested_priv *priv),
4705 struct netdev_nested_priv *priv);
4706
4707void *netdev_adjacent_get_private(struct list_head *adj_list);
4708void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4709struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4710struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4711int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4712 struct netlink_ext_ack *extack);
4713int netdev_master_upper_dev_link(struct net_device *dev,
4714 struct net_device *upper_dev,
4715 void *upper_priv, void *upper_info,
4716 struct netlink_ext_ack *extack);
4717void netdev_upper_dev_unlink(struct net_device *dev,
4718 struct net_device *upper_dev);
4719int netdev_adjacent_change_prepare(struct net_device *old_dev,
4720 struct net_device *new_dev,
4721 struct net_device *dev,
4722 struct netlink_ext_ack *extack);
4723void netdev_adjacent_change_commit(struct net_device *old_dev,
4724 struct net_device *new_dev,
4725 struct net_device *dev);
4726void netdev_adjacent_change_abort(struct net_device *old_dev,
4727 struct net_device *new_dev,
4728 struct net_device *dev);
4729void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4730void *netdev_lower_dev_get_private(struct net_device *dev,
4731 struct net_device *lower_dev);
4732void netdev_lower_state_changed(struct net_device *lower_dev,
4733 void *lower_state_info);
4734
4735/* RSS keys are 40 or 52 bytes long */
4736#define NETDEV_RSS_KEY_LEN 52
4737extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4738void netdev_rss_key_fill(void *buffer, size_t len);
4739
4740int skb_checksum_help(struct sk_buff *skb);
4741int skb_crc32c_csum_help(struct sk_buff *skb);
4742int skb_csum_hwoffload_help(struct sk_buff *skb,
4743 const netdev_features_t features);
4744
4745struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4746 netdev_features_t features, bool tx_path);
4747struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4748 netdev_features_t features);
4749
4750struct netdev_bonding_info {
4751 ifslave slave;
4752 ifbond master;
4753};
4754
4755struct netdev_notifier_bonding_info {
4756 struct netdev_notifier_info info; /* must be first */
4757 struct netdev_bonding_info bonding_info;
4758};
4759
4760void netdev_bonding_info_change(struct net_device *dev,
4761 struct netdev_bonding_info *bonding_info);
4762
4763#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4764void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4765#else
4766static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4767 const void *data)
4768{
4769}
4770#endif
4771
4772static inline
4773struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4774{
4775 return __skb_gso_segment(skb, features, true);
4776}
4777__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4778
4779static inline bool can_checksum_protocol(netdev_features_t features,
4780 __be16 protocol)
4781{
4782 if (protocol == htons(ETH_P_FCOE))
4783 return !!(features & NETIF_F_FCOE_CRC);
4784
4785 /* Assume this is an IP checksum (not SCTP CRC) */
4786
4787 if (features & NETIF_F_HW_CSUM) {
4788 /* Can checksum everything */
4789 return true;
4790 }
4791
4792 switch (protocol) {
4793 case htons(ETH_P_IP):
4794 return !!(features & NETIF_F_IP_CSUM);
4795 case htons(ETH_P_IPV6):
4796 return !!(features & NETIF_F_IPV6_CSUM);
4797 default:
4798 return false;
4799 }
4800}
4801
4802#ifdef CONFIG_BUG
4803void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4804#else
4805static inline void netdev_rx_csum_fault(struct net_device *dev,
4806 struct sk_buff *skb)
4807{
4808}
4809#endif
4810/* rx skb timestamps */
4811void net_enable_timestamp(void);
4812void net_disable_timestamp(void);
4813
4814#ifdef CONFIG_PROC_FS
4815int __init dev_proc_init(void);
4816#else
4817#define dev_proc_init() 0
4818#endif
4819
4820static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4821 struct sk_buff *skb, struct net_device *dev,
4822 bool more)
4823{
4824 __this_cpu_write(softnet_data.xmit.more, more);
4825 return ops->ndo_start_xmit(skb, dev);
4826}
4827
4828static inline bool netdev_xmit_more(void)
4829{
4830 return __this_cpu_read(softnet_data.xmit.more);
4831}
4832
4833static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4834 struct netdev_queue *txq, bool more)
4835{
4836 const struct net_device_ops *ops = dev->netdev_ops;
4837 netdev_tx_t rc;
4838
4839 rc = __netdev_start_xmit(ops, skb, dev, more);
4840 if (rc == NETDEV_TX_OK)
4841 txq_trans_update(txq);
4842
4843 return rc;
4844}
4845
4846int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4847 const void *ns);
4848void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4849 const void *ns);
4850
4851extern const struct kobj_ns_type_operations net_ns_type_operations;
4852
4853const char *netdev_drivername(const struct net_device *dev);
4854
4855void linkwatch_run_queue(void);
4856
4857static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4858 netdev_features_t f2)
4859{
4860 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4861 if (f1 & NETIF_F_HW_CSUM)
4862 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4863 else
4864 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4865 }
4866
4867 return f1 & f2;
4868}
4869
4870static inline netdev_features_t netdev_get_wanted_features(
4871 struct net_device *dev)
4872{
4873 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4874}
4875netdev_features_t netdev_increment_features(netdev_features_t all,
4876 netdev_features_t one, netdev_features_t mask);
4877
4878/* Allow TSO being used on stacked device :
4879 * Performing the GSO segmentation before last device
4880 * is a performance improvement.
4881 */
4882static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4883 netdev_features_t mask)
4884{
4885 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4886}
4887
4888int __netdev_update_features(struct net_device *dev);
4889void netdev_update_features(struct net_device *dev);
4890void netdev_change_features(struct net_device *dev);
4891
4892void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4893 struct net_device *dev);
4894
4895netdev_features_t passthru_features_check(struct sk_buff *skb,
4896 struct net_device *dev,
4897 netdev_features_t features);
4898netdev_features_t netif_skb_features(struct sk_buff *skb);
4899
4900static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4901{
4902 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4903
4904 /* check flags correspondence */
4905 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4906 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4907 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4908 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4909 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4910 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4911 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4912 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4913 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4914 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4915 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4916 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4917 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4918 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4919 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4920 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4921 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4922 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4923 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4924
4925 return (features & feature) == feature;
4926}
4927
4928static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4929{
4930 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4931 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4932}
4933
4934static inline bool netif_needs_gso(struct sk_buff *skb,
4935 netdev_features_t features)
4936{
4937 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4938 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4939 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4940}
4941
4942static inline void netif_set_gso_max_size(struct net_device *dev,
4943 unsigned int size)
4944{
4945 dev->gso_max_size = size;
4946}
4947
4948static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4949 int pulled_hlen, u16 mac_offset,
4950 int mac_len)
4951{
4952 skb->protocol = protocol;
4953 skb->encapsulation = 1;
4954 skb_push(skb, pulled_hlen);
4955 skb_reset_transport_header(skb);
4956 skb->mac_header = mac_offset;
4957 skb->network_header = skb->mac_header + mac_len;
4958 skb->mac_len = mac_len;
4959}
4960
4961static inline bool netif_is_macsec(const struct net_device *dev)
4962{
4963 return dev->priv_flags & IFF_MACSEC;
4964}
4965
4966static inline bool netif_is_macvlan(const struct net_device *dev)
4967{
4968 return dev->priv_flags & IFF_MACVLAN;
4969}
4970
4971static inline bool netif_is_macvlan_port(const struct net_device *dev)
4972{
4973 return dev->priv_flags & IFF_MACVLAN_PORT;
4974}
4975
4976static inline bool netif_is_bond_master(const struct net_device *dev)
4977{
4978 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4979}
4980
4981static inline bool netif_is_bond_slave(const struct net_device *dev)
4982{
4983 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4984}
4985
4986static inline bool netif_supports_nofcs(struct net_device *dev)
4987{
4988 return dev->priv_flags & IFF_SUPP_NOFCS;
4989}
4990
4991static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4992{
4993 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4994}
4995
4996static inline bool netif_is_l3_master(const struct net_device *dev)
4997{
4998 return dev->priv_flags & IFF_L3MDEV_MASTER;
4999}
5000
5001static inline bool netif_is_l3_slave(const struct net_device *dev)
5002{
5003 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5004}
5005
5006static inline bool netif_is_bridge_master(const struct net_device *dev)
5007{
5008 return dev->priv_flags & IFF_EBRIDGE;
5009}
5010
5011static inline bool netif_is_bridge_port(const struct net_device *dev)
5012{
5013 return dev->priv_flags & IFF_BRIDGE_PORT;
5014}
5015
5016static inline bool netif_is_ovs_master(const struct net_device *dev)
5017{
5018 return dev->priv_flags & IFF_OPENVSWITCH;
5019}
5020
5021static inline bool netif_is_ovs_port(const struct net_device *dev)
5022{
5023 return dev->priv_flags & IFF_OVS_DATAPATH;
5024}
5025
5026static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5027{
5028 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5029}
5030
5031static inline bool netif_is_team_master(const struct net_device *dev)
5032{
5033 return dev->priv_flags & IFF_TEAM;
5034}
5035
5036static inline bool netif_is_team_port(const struct net_device *dev)
5037{
5038 return dev->priv_flags & IFF_TEAM_PORT;
5039}
5040
5041static inline bool netif_is_lag_master(const struct net_device *dev)
5042{
5043 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5044}
5045
5046static inline bool netif_is_lag_port(const struct net_device *dev)
5047{
5048 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5049}
5050
5051static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5052{
5053 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5054}
5055
5056static inline bool netif_is_failover(const struct net_device *dev)
5057{
5058 return dev->priv_flags & IFF_FAILOVER;
5059}
5060
5061static inline bool netif_is_failover_slave(const struct net_device *dev)
5062{
5063 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5064}
5065
5066/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5067static inline void netif_keep_dst(struct net_device *dev)
5068{
5069 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5070}
5071
5072/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5073static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5074{
5075 /* TODO: reserve and use an additional IFF bit, if we get more users */
5076 return dev->priv_flags & IFF_MACSEC;
5077}
5078
5079extern struct pernet_operations __net_initdata loopback_net_ops;
5080
5081/* Logging, debugging and troubleshooting/diagnostic helpers. */
5082
5083/* netdev_printk helpers, similar to dev_printk */
5084
5085static inline const char *netdev_name(const struct net_device *dev)
5086{
5087 if (!dev->name[0] || strchr(dev->name, '%'))
5088 return "(unnamed net_device)";
5089 return dev->name;
5090}
5091
5092static inline bool netdev_unregistering(const struct net_device *dev)
5093{
5094 return dev->reg_state == NETREG_UNREGISTERING;
5095}
5096
5097static inline const char *netdev_reg_state(const struct net_device *dev)
5098{
5099 switch (dev->reg_state) {
5100 case NETREG_UNINITIALIZED: return " (uninitialized)";
5101 case NETREG_REGISTERED: return "";
5102 case NETREG_UNREGISTERING: return " (unregistering)";
5103 case NETREG_UNREGISTERED: return " (unregistered)";
5104 case NETREG_RELEASED: return " (released)";
5105 case NETREG_DUMMY: return " (dummy)";
5106 }
5107
5108 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5109 return " (unknown)";
5110}
5111
5112__printf(3, 4) __cold
5113void netdev_printk(const char *level, const struct net_device *dev,
5114 const char *format, ...);
5115__printf(2, 3) __cold
5116void netdev_emerg(const struct net_device *dev, const char *format, ...);
5117__printf(2, 3) __cold
5118void netdev_alert(const struct net_device *dev, const char *format, ...);
5119__printf(2, 3) __cold
5120void netdev_crit(const struct net_device *dev, const char *format, ...);
5121__printf(2, 3) __cold
5122void netdev_err(const struct net_device *dev, const char *format, ...);
5123__printf(2, 3) __cold
5124void netdev_warn(const struct net_device *dev, const char *format, ...);
5125__printf(2, 3) __cold
5126void netdev_notice(const struct net_device *dev, const char *format, ...);
5127__printf(2, 3) __cold
5128void netdev_info(const struct net_device *dev, const char *format, ...);
5129
5130#define netdev_level_once(level, dev, fmt, ...) \
5131do { \
5132 static bool __print_once __read_mostly; \
5133 \
5134 if (!__print_once) { \
5135 __print_once = true; \
5136 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5137 } \
5138} while (0)
5139
5140#define netdev_emerg_once(dev, fmt, ...) \
5141 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5142#define netdev_alert_once(dev, fmt, ...) \
5143 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5144#define netdev_crit_once(dev, fmt, ...) \
5145 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5146#define netdev_err_once(dev, fmt, ...) \
5147 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5148#define netdev_warn_once(dev, fmt, ...) \
5149 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5150#define netdev_notice_once(dev, fmt, ...) \
5151 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5152#define netdev_info_once(dev, fmt, ...) \
5153 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5154
5155#define MODULE_ALIAS_NETDEV(device) \
5156 MODULE_ALIAS("netdev-" device)
5157
5158#if defined(CONFIG_DYNAMIC_DEBUG) || \
5159 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5160#define netdev_dbg(__dev, format, args...) \
5161do { \
5162 dynamic_netdev_dbg(__dev, format, ##args); \
5163} while (0)
5164#elif defined(DEBUG)
5165#define netdev_dbg(__dev, format, args...) \
5166 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5167#else
5168#define netdev_dbg(__dev, format, args...) \
5169({ \
5170 if (0) \
5171 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5172})
5173#endif
5174
5175#if defined(VERBOSE_DEBUG)
5176#define netdev_vdbg netdev_dbg
5177#else
5178
5179#define netdev_vdbg(dev, format, args...) \
5180({ \
5181 if (0) \
5182 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5183 0; \
5184})
5185#endif
5186
5187/*
5188 * netdev_WARN() acts like dev_printk(), but with the key difference
5189 * of using a WARN/WARN_ON to get the message out, including the
5190 * file/line information and a backtrace.
5191 */
5192#define netdev_WARN(dev, format, args...) \
5193 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5194 netdev_reg_state(dev), ##args)
5195
5196#define netdev_WARN_ONCE(dev, format, args...) \
5197 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5198 netdev_reg_state(dev), ##args)
5199
5200/* netif printk helpers, similar to netdev_printk */
5201
5202#define netif_printk(priv, type, level, dev, fmt, args...) \
5203do { \
5204 if (netif_msg_##type(priv)) \
5205 netdev_printk(level, (dev), fmt, ##args); \
5206} while (0)
5207
5208#define netif_level(level, priv, type, dev, fmt, args...) \
5209do { \
5210 if (netif_msg_##type(priv)) \
5211 netdev_##level(dev, fmt, ##args); \
5212} while (0)
5213
5214#define netif_emerg(priv, type, dev, fmt, args...) \
5215 netif_level(emerg, priv, type, dev, fmt, ##args)
5216#define netif_alert(priv, type, dev, fmt, args...) \
5217 netif_level(alert, priv, type, dev, fmt, ##args)
5218#define netif_crit(priv, type, dev, fmt, args...) \
5219 netif_level(crit, priv, type, dev, fmt, ##args)
5220#define netif_err(priv, type, dev, fmt, args...) \
5221 netif_level(err, priv, type, dev, fmt, ##args)
5222#define netif_warn(priv, type, dev, fmt, args...) \
5223 netif_level(warn, priv, type, dev, fmt, ##args)
5224#define netif_notice(priv, type, dev, fmt, args...) \
5225 netif_level(notice, priv, type, dev, fmt, ##args)
5226#define netif_info(priv, type, dev, fmt, args...) \
5227 netif_level(info, priv, type, dev, fmt, ##args)
5228
5229#if defined(CONFIG_DYNAMIC_DEBUG) || \
5230 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5231#define netif_dbg(priv, type, netdev, format, args...) \
5232do { \
5233 if (netif_msg_##type(priv)) \
5234 dynamic_netdev_dbg(netdev, format, ##args); \
5235} while (0)
5236#elif defined(DEBUG)
5237#define netif_dbg(priv, type, dev, format, args...) \
5238 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5239#else
5240#define netif_dbg(priv, type, dev, format, args...) \
5241({ \
5242 if (0) \
5243 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5244 0; \
5245})
5246#endif
5247
5248/* if @cond then downgrade to debug, else print at @level */
5249#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5250 do { \
5251 if (cond) \
5252 netif_dbg(priv, type, netdev, fmt, ##args); \
5253 else \
5254 netif_ ## level(priv, type, netdev, fmt, ##args); \
5255 } while (0)
5256
5257#if defined(VERBOSE_DEBUG)
5258#define netif_vdbg netif_dbg
5259#else
5260#define netif_vdbg(priv, type, dev, format, args...) \
5261({ \
5262 if (0) \
5263 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5264 0; \
5265})
5266#endif
5267
5268/*
5269 * The list of packet types we will receive (as opposed to discard)
5270 * and the routines to invoke.
5271 *
5272 * Why 16. Because with 16 the only overlap we get on a hash of the
5273 * low nibble of the protocol value is RARP/SNAP/X.25.
5274 *
5275 * 0800 IP
5276 * 0001 802.3
5277 * 0002 AX.25
5278 * 0004 802.2
5279 * 8035 RARP
5280 * 0005 SNAP
5281 * 0805 X.25
5282 * 0806 ARP
5283 * 8137 IPX
5284 * 0009 Localtalk
5285 * 86DD IPv6
5286 */
5287#define PTYPE_HASH_SIZE (16)
5288#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5289
5290extern struct net_device *blackhole_netdev;
5291
5292#endif /* _LINUX_NETDEVICE_H */