blob: ddc66ab8a1def2084d5d70b7bf52973da20c8451 [file] [log] [blame]
David Brazdil0f672f62019-12-10 10:32:29 +00001/* SPDX-License-Identifier: GPL-2.0-or-later */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002/*
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 *
Andrew Scullb4b6d4a2019-01-02 15:54:55 +000019 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31
32#include <linux/percpu.h>
33#include <linux/rculist.h>
34#include <linux/workqueue.h>
35#include <linux/dynamic_queue_limits.h>
36
37#include <linux/ethtool.h>
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <net/xdp.h>
44
45#include <linux/netdev_features.h>
46#include <linux/neighbour.h>
47#include <uapi/linux/netdevice.h>
48#include <uapi/linux/if_bonding.h>
49#include <uapi/linux/pkt_cls.h>
50#include <linux/hashtable.h>
51
52struct netpoll_info;
53struct device;
54struct phy_device;
55struct dsa_port;
56
57struct sfp_bus;
58/* 802.11 specific */
59struct wireless_dev;
60/* 802.15.4 specific */
61struct wpan_dev;
62struct mpls_dev;
63/* UDP Tunnel offloads */
64struct udp_tunnel_info;
65struct bpf_prog;
66struct xdp_buff;
67
68void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
70
71/* Backlog congestion levels */
72#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73#define NET_RX_DROP 1 /* packet dropped */
74
Olivier Deprez0e641232021-09-23 10:07:05 +020075#define MAX_NEST_DEV 8
76
Andrew Scullb4b6d4a2019-01-02 15:54:55 +000077/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
98#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99
100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
106/* Driver transmit return codes */
107#define NETDEV_TX_MASK 0xf0
108
109enum netdev_tx {
110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113};
114typedef enum netdev_tx netdev_tx_t;
115
116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
134/*
135 * Compute the worst-case header length according to the protocols
136 * used.
137 */
138
139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
147#else
148# define LL_MAX_HEADER 32
149#endif
150
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
161 */
162
163struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
193#ifdef CONFIG_RPS
194#include <linux/static_key.h>
David Brazdil0f672f62019-12-10 10:32:29 +0000195extern struct static_key_false rps_needed;
196extern struct static_key_false rfs_needed;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000197#endif
198
199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
211#define NETDEV_HW_ADDR_T_MULTICAST 5
212 bool global_use;
213 int sync_cnt;
214 int refcount;
215 int synced;
216 struct rcu_head rcu_head;
217};
218
219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
228
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231#define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233
234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236#define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238
239struct hh_cache {
240 unsigned int hh_len;
241 seqlock_t hh_lock;
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
David Brazdil0f672f62019-12-10 10:32:29 +0000275 __be16 (*parse_protocol)(const struct sk_buff *skb);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000276};
277
278/* These flag bits are private to the generic network queueing
279 * layer; they may not be explicitly referenced by any other
280 * code.
281 */
282
283enum netdev_state_t {
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
286 __LINK_STATE_NOCARRIER,
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
289};
290
291
292/*
293 * This structure holds boot-time configured netdevice settings. They
294 * are then used in the device probing.
295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
302int __init netdev_boot_setup(char *str);
303
304struct gro_list {
305 struct list_head list;
306 int count;
307};
308
309/*
310 * size of gro hash buckets, must less than bit number of
311 * napi_struct::gro_bitmask
312 */
313#define GRO_HASH_BUCKETS 8
314
315/*
316 * Structure for NAPI scheduling similar to tasklet but with weighting
317 */
318struct napi_struct {
319 /* The poll_list must only be managed by the entity which
320 * changes the state of the NAPI_STATE_SCHED bit. This means
321 * whoever atomically sets that bit can add this napi_struct
322 * to the per-CPU poll_list, and whoever clears that bit
323 * can remove from the list right before clearing the bit.
324 */
325 struct list_head poll_list;
326
327 unsigned long state;
328 int weight;
329 unsigned long gro_bitmask;
330 int (*poll)(struct napi_struct *, int);
331#ifdef CONFIG_NETPOLL
332 int poll_owner;
333#endif
334 struct net_device *dev;
335 struct gro_list gro_hash[GRO_HASH_BUCKETS];
336 struct sk_buff *skb;
David Brazdil0f672f62019-12-10 10:32:29 +0000337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
338 int rx_count; /* length of rx_list */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000339 struct hrtimer timer;
340 struct list_head dev_list;
341 struct hlist_node napi_hash_node;
342 unsigned int napi_id;
343};
344
345enum {
346 NAPI_STATE_SCHED, /* Poll is scheduled */
347 NAPI_STATE_MISSED, /* reschedule a napi */
348 NAPI_STATE_DISABLE, /* Disable pending */
349 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
350 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
351 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
352 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
353};
354
355enum {
356 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
357 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
358 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
359 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
360 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
361 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
362 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
363};
364
365enum gro_result {
366 GRO_MERGED,
367 GRO_MERGED_FREE,
368 GRO_HELD,
369 GRO_NORMAL,
370 GRO_DROP,
371 GRO_CONSUMED,
372};
373typedef enum gro_result gro_result_t;
374
375/*
376 * enum rx_handler_result - Possible return values for rx_handlers.
377 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
378 * further.
379 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
380 * case skb->dev was changed by rx_handler.
381 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
382 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
383 *
384 * rx_handlers are functions called from inside __netif_receive_skb(), to do
385 * special processing of the skb, prior to delivery to protocol handlers.
386 *
387 * Currently, a net_device can only have a single rx_handler registered. Trying
388 * to register a second rx_handler will return -EBUSY.
389 *
390 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
391 * To unregister a rx_handler on a net_device, use
392 * netdev_rx_handler_unregister().
393 *
394 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
395 * do with the skb.
396 *
397 * If the rx_handler consumed the skb in some way, it should return
398 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
399 * the skb to be delivered in some other way.
400 *
401 * If the rx_handler changed skb->dev, to divert the skb to another
402 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
403 * new device will be called if it exists.
404 *
405 * If the rx_handler decides the skb should be ignored, it should return
406 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
407 * are registered on exact device (ptype->dev == skb->dev).
408 *
409 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
410 * delivered, it should return RX_HANDLER_PASS.
411 *
412 * A device without a registered rx_handler will behave as if rx_handler
413 * returned RX_HANDLER_PASS.
414 */
415
416enum rx_handler_result {
417 RX_HANDLER_CONSUMED,
418 RX_HANDLER_ANOTHER,
419 RX_HANDLER_EXACT,
420 RX_HANDLER_PASS,
421};
422typedef enum rx_handler_result rx_handler_result_t;
423typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
424
425void __napi_schedule(struct napi_struct *n);
426void __napi_schedule_irqoff(struct napi_struct *n);
427
428static inline bool napi_disable_pending(struct napi_struct *n)
429{
430 return test_bit(NAPI_STATE_DISABLE, &n->state);
431}
432
433bool napi_schedule_prep(struct napi_struct *n);
434
435/**
436 * napi_schedule - schedule NAPI poll
437 * @n: NAPI context
438 *
439 * Schedule NAPI poll routine to be called if it is not already
440 * running.
441 */
442static inline void napi_schedule(struct napi_struct *n)
443{
444 if (napi_schedule_prep(n))
445 __napi_schedule(n);
446}
447
448/**
449 * napi_schedule_irqoff - schedule NAPI poll
450 * @n: NAPI context
451 *
452 * Variant of napi_schedule(), assuming hard irqs are masked.
453 */
454static inline void napi_schedule_irqoff(struct napi_struct *n)
455{
456 if (napi_schedule_prep(n))
457 __napi_schedule_irqoff(n);
458}
459
460/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
461static inline bool napi_reschedule(struct napi_struct *napi)
462{
463 if (napi_schedule_prep(napi)) {
464 __napi_schedule(napi);
465 return true;
466 }
467 return false;
468}
469
470bool napi_complete_done(struct napi_struct *n, int work_done);
471/**
472 * napi_complete - NAPI processing complete
473 * @n: NAPI context
474 *
475 * Mark NAPI processing as complete.
476 * Consider using napi_complete_done() instead.
477 * Return false if device should avoid rearming interrupts.
478 */
479static inline bool napi_complete(struct napi_struct *n)
480{
481 return napi_complete_done(n, 0);
482}
483
484/**
485 * napi_hash_del - remove a NAPI from global table
486 * @napi: NAPI context
487 *
488 * Warning: caller must observe RCU grace period
489 * before freeing memory containing @napi, if
490 * this function returns true.
491 * Note: core networking stack automatically calls it
492 * from netif_napi_del().
493 * Drivers might want to call this helper to combine all
494 * the needed RCU grace periods into a single one.
495 */
496bool napi_hash_del(struct napi_struct *napi);
497
498/**
499 * napi_disable - prevent NAPI from scheduling
500 * @n: NAPI context
501 *
502 * Stop NAPI from being scheduled on this context.
503 * Waits till any outstanding processing completes.
504 */
505void napi_disable(struct napi_struct *n);
506
507/**
508 * napi_enable - enable NAPI scheduling
509 * @n: NAPI context
510 *
511 * Resume NAPI from being scheduled on this context.
512 * Must be paired with napi_disable.
513 */
514static inline void napi_enable(struct napi_struct *n)
515{
516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
517 smp_mb__before_atomic();
518 clear_bit(NAPI_STATE_SCHED, &n->state);
519 clear_bit(NAPI_STATE_NPSVC, &n->state);
520}
521
522/**
523 * napi_synchronize - wait until NAPI is not running
524 * @n: NAPI context
525 *
526 * Wait until NAPI is done being scheduled on this context.
527 * Waits till any outstanding processing completes but
528 * does not disable future activations.
529 */
530static inline void napi_synchronize(const struct napi_struct *n)
531{
532 if (IS_ENABLED(CONFIG_SMP))
533 while (test_bit(NAPI_STATE_SCHED, &n->state))
534 msleep(1);
535 else
536 barrier();
537}
538
David Brazdil0f672f62019-12-10 10:32:29 +0000539/**
540 * napi_if_scheduled_mark_missed - if napi is running, set the
541 * NAPIF_STATE_MISSED
542 * @n: NAPI context
543 *
544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
545 * NAPI is scheduled.
546 **/
547static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
548{
549 unsigned long val, new;
550
551 do {
552 val = READ_ONCE(n->state);
553 if (val & NAPIF_STATE_DISABLE)
554 return true;
555
556 if (!(val & NAPIF_STATE_SCHED))
557 return false;
558
559 new = val | NAPIF_STATE_MISSED;
560 } while (cmpxchg(&n->state, val, new) != val);
561
562 return true;
563}
564
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000565enum netdev_queue_state_t {
566 __QUEUE_STATE_DRV_XOFF,
567 __QUEUE_STATE_STACK_XOFF,
568 __QUEUE_STATE_FROZEN,
569};
570
571#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
572#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
573#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
574
575#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
576#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
577 QUEUE_STATE_FROZEN)
578#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
579 QUEUE_STATE_FROZEN)
580
581/*
582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
583 * netif_tx_* functions below are used to manipulate this flag. The
584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
585 * queue independently. The netif_xmit_*stopped functions below are called
586 * to check if the queue has been stopped by the driver or stack (either
587 * of the XOFF bits are set in the state). Drivers should not need to call
588 * netif_xmit*stopped functions, they should only be using netif_tx_*.
589 */
590
591struct netdev_queue {
592/*
593 * read-mostly part
594 */
595 struct net_device *dev;
596 struct Qdisc __rcu *qdisc;
597 struct Qdisc *qdisc_sleeping;
598#ifdef CONFIG_SYSFS
599 struct kobject kobj;
600#endif
601#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
602 int numa_node;
603#endif
604 unsigned long tx_maxrate;
605 /*
606 * Number of TX timeouts for this queue
607 * (/sys/class/net/DEV/Q/trans_timeout)
608 */
609 unsigned long trans_timeout;
610
611 /* Subordinate device that the queue has been assigned to */
612 struct net_device *sb_dev;
David Brazdil0f672f62019-12-10 10:32:29 +0000613#ifdef CONFIG_XDP_SOCKETS
614 struct xdp_umem *umem;
615#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000616/*
617 * write-mostly part
618 */
619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
620 int xmit_lock_owner;
621 /*
622 * Time (in jiffies) of last Tx
623 */
624 unsigned long trans_start;
625
626 unsigned long state;
627
628#ifdef CONFIG_BQL
629 struct dql dql;
630#endif
631} ____cacheline_aligned_in_smp;
632
633extern int sysctl_fb_tunnels_only_for_init_net;
David Brazdil0f672f62019-12-10 10:32:29 +0000634extern int sysctl_devconf_inherit_init_net;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000635
636static inline bool net_has_fallback_tunnels(const struct net *net)
637{
638 return net == &init_net ||
639 !IS_ENABLED(CONFIG_SYSCTL) ||
640 !sysctl_fb_tunnels_only_for_init_net;
641}
642
643static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
644{
645#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
646 return q->numa_node;
647#else
648 return NUMA_NO_NODE;
649#endif
650}
651
652static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
653{
654#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
655 q->numa_node = node;
656#endif
657}
658
659#ifdef CONFIG_RPS
660/*
661 * This structure holds an RPS map which can be of variable length. The
662 * map is an array of CPUs.
663 */
664struct rps_map {
665 unsigned int len;
666 struct rcu_head rcu;
667 u16 cpus[0];
668};
669#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670
671/*
672 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
673 * tail pointer for that CPU's input queue at the time of last enqueue, and
674 * a hardware filter index.
675 */
676struct rps_dev_flow {
677 u16 cpu;
678 u16 filter;
679 unsigned int last_qtail;
680};
681#define RPS_NO_FILTER 0xffff
682
683/*
684 * The rps_dev_flow_table structure contains a table of flow mappings.
685 */
686struct rps_dev_flow_table {
687 unsigned int mask;
688 struct rcu_head rcu;
689 struct rps_dev_flow flows[0];
690};
691#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
692 ((_num) * sizeof(struct rps_dev_flow)))
693
694/*
695 * The rps_sock_flow_table contains mappings of flows to the last CPU
696 * on which they were processed by the application (set in recvmsg).
697 * Each entry is a 32bit value. Upper part is the high-order bits
698 * of flow hash, lower part is CPU number.
699 * rps_cpu_mask is used to partition the space, depending on number of
700 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
701 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
702 * meaning we use 32-6=26 bits for the hash.
703 */
704struct rps_sock_flow_table {
705 u32 mask;
706
707 u32 ents[0] ____cacheline_aligned_in_smp;
708};
709#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
710
711#define RPS_NO_CPU 0xffff
712
713extern u32 rps_cpu_mask;
714extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
715
716static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
717 u32 hash)
718{
719 if (table && hash) {
720 unsigned int index = hash & table->mask;
721 u32 val = hash & ~rps_cpu_mask;
722
723 /* We only give a hint, preemption can change CPU under us */
724 val |= raw_smp_processor_id();
725
726 if (table->ents[index] != val)
727 table->ents[index] = val;
728 }
729}
730
731#ifdef CONFIG_RFS_ACCEL
732bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 u16 filter_id);
734#endif
735#endif /* CONFIG_RPS */
736
737/* This structure contains an instance of an RX queue. */
738struct netdev_rx_queue {
739#ifdef CONFIG_RPS
740 struct rps_map __rcu *rps_map;
741 struct rps_dev_flow_table __rcu *rps_flow_table;
742#endif
743 struct kobject kobj;
744 struct net_device *dev;
745 struct xdp_rxq_info xdp_rxq;
David Brazdil0f672f62019-12-10 10:32:29 +0000746#ifdef CONFIG_XDP_SOCKETS
747 struct xdp_umem *umem;
748#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000749} ____cacheline_aligned_in_smp;
750
751/*
752 * RX queue sysfs structures and functions.
753 */
754struct rx_queue_attribute {
755 struct attribute attr;
756 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
757 ssize_t (*store)(struct netdev_rx_queue *queue,
758 const char *buf, size_t len);
759};
760
761#ifdef CONFIG_XPS
762/*
763 * This structure holds an XPS map which can be of variable length. The
764 * map is an array of queues.
765 */
766struct xps_map {
767 unsigned int len;
768 unsigned int alloc_len;
769 struct rcu_head rcu;
770 u16 queues[0];
771};
772#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
773#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
774 - sizeof(struct xps_map)) / sizeof(u16))
775
776/*
777 * This structure holds all XPS maps for device. Maps are indexed by CPU.
778 */
779struct xps_dev_maps {
780 struct rcu_head rcu;
781 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
782};
783
784#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
785 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
786
787#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
788 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
789
790#endif /* CONFIG_XPS */
791
792#define TC_MAX_QUEUE 16
793#define TC_BITMASK 15
794/* HW offloaded queuing disciplines txq count and offset maps */
795struct netdev_tc_txq {
796 u16 count;
797 u16 offset;
798};
799
800#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
801/*
802 * This structure is to hold information about the device
803 * configured to run FCoE protocol stack.
804 */
805struct netdev_fcoe_hbainfo {
806 char manufacturer[64];
807 char serial_number[64];
808 char hardware_version[64];
809 char driver_version[64];
810 char optionrom_version[64];
811 char firmware_version[64];
812 char model[256];
813 char model_description[256];
814};
815#endif
816
817#define MAX_PHYS_ITEM_ID_LEN 32
818
819/* This structure holds a unique identifier to identify some
820 * physical item (port for example) used by a netdevice.
821 */
822struct netdev_phys_item_id {
823 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
824 unsigned char id_len;
825};
826
827static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
828 struct netdev_phys_item_id *b)
829{
830 return a->id_len == b->id_len &&
831 memcmp(a->id, b->id, a->id_len) == 0;
832}
833
834typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
835 struct sk_buff *skb,
836 struct net_device *sb_dev);
837
838enum tc_setup_type {
839 TC_SETUP_QDISC_MQPRIO,
840 TC_SETUP_CLSU32,
841 TC_SETUP_CLSFLOWER,
842 TC_SETUP_CLSMATCHALL,
843 TC_SETUP_CLSBPF,
844 TC_SETUP_BLOCK,
845 TC_SETUP_QDISC_CBS,
846 TC_SETUP_QDISC_RED,
847 TC_SETUP_QDISC_PRIO,
848 TC_SETUP_QDISC_MQ,
849 TC_SETUP_QDISC_ETF,
David Brazdil0f672f62019-12-10 10:32:29 +0000850 TC_SETUP_ROOT_QDISC,
851 TC_SETUP_QDISC_GRED,
852 TC_SETUP_QDISC_TAPRIO,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000853};
854
855/* These structures hold the attributes of bpf state that are being passed
856 * to the netdevice through the bpf op.
857 */
858enum bpf_netdev_command {
859 /* Set or clear a bpf program used in the earliest stages of packet
860 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
861 * is responsible for calling bpf_prog_put on any old progs that are
862 * stored. In case of error, the callee need not release the new prog
863 * reference, but on success it takes ownership and must bpf_prog_put
864 * when it is no longer used.
865 */
866 XDP_SETUP_PROG,
867 XDP_SETUP_PROG_HW,
868 XDP_QUERY_PROG,
869 XDP_QUERY_PROG_HW,
870 /* BPF program for offload callbacks, invoked at program load time. */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000871 BPF_OFFLOAD_MAP_ALLOC,
872 BPF_OFFLOAD_MAP_FREE,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000873 XDP_SETUP_XSK_UMEM,
874};
875
876struct bpf_prog_offload_ops;
877struct netlink_ext_ack;
878struct xdp_umem;
879
880struct netdev_bpf {
881 enum bpf_netdev_command command;
882 union {
883 /* XDP_SETUP_PROG */
884 struct {
885 u32 flags;
886 struct bpf_prog *prog;
887 struct netlink_ext_ack *extack;
888 };
889 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
890 struct {
891 u32 prog_id;
892 /* flags with which program was installed */
893 u32 prog_flags;
894 };
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000895 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
896 struct {
897 struct bpf_offloaded_map *offmap;
898 };
David Brazdil0f672f62019-12-10 10:32:29 +0000899 /* XDP_SETUP_XSK_UMEM */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000900 struct {
David Brazdil0f672f62019-12-10 10:32:29 +0000901 struct xdp_umem *umem;
902 u16 queue_id;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000903 } xsk;
904 };
905};
906
David Brazdil0f672f62019-12-10 10:32:29 +0000907/* Flags for ndo_xsk_wakeup. */
908#define XDP_WAKEUP_RX (1 << 0)
909#define XDP_WAKEUP_TX (1 << 1)
910
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000911#ifdef CONFIG_XFRM_OFFLOAD
912struct xfrmdev_ops {
913 int (*xdo_dev_state_add) (struct xfrm_state *x);
914 void (*xdo_dev_state_delete) (struct xfrm_state *x);
915 void (*xdo_dev_state_free) (struct xfrm_state *x);
916 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
917 struct xfrm_state *x);
918 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
919};
920#endif
921
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000922struct dev_ifalias {
923 struct rcu_head rcuhead;
924 char ifalias[];
925};
926
David Brazdil0f672f62019-12-10 10:32:29 +0000927struct devlink;
928struct tlsdev_ops;
929
930
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000931/*
932 * This structure defines the management hooks for network devices.
933 * The following hooks can be defined; unless noted otherwise, they are
934 * optional and can be filled with a null pointer.
935 *
936 * int (*ndo_init)(struct net_device *dev);
937 * This function is called once when a network device is registered.
938 * The network device can use this for any late stage initialization
939 * or semantic validation. It can fail with an error code which will
940 * be propagated back to register_netdev.
941 *
942 * void (*ndo_uninit)(struct net_device *dev);
943 * This function is called when device is unregistered or when registration
944 * fails. It is not called if init fails.
945 *
946 * int (*ndo_open)(struct net_device *dev);
947 * This function is called when a network device transitions to the up
948 * state.
949 *
950 * int (*ndo_stop)(struct net_device *dev);
951 * This function is called when a network device transitions to the down
952 * state.
953 *
954 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
955 * struct net_device *dev);
956 * Called when a packet needs to be transmitted.
957 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
958 * the queue before that can happen; it's for obsolete devices and weird
959 * corner cases, but the stack really does a non-trivial amount
960 * of useless work if you return NETDEV_TX_BUSY.
961 * Required; cannot be NULL.
962 *
963 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
964 * struct net_device *dev
965 * netdev_features_t features);
966 * Called by core transmit path to determine if device is capable of
967 * performing offload operations on a given packet. This is to give
968 * the device an opportunity to implement any restrictions that cannot
969 * be otherwise expressed by feature flags. The check is called with
970 * the set of features that the stack has calculated and it returns
971 * those the driver believes to be appropriate.
972 *
973 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
David Brazdil0f672f62019-12-10 10:32:29 +0000974 * struct net_device *sb_dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000975 * Called to decide which queue to use when device supports multiple
976 * transmit queues.
977 *
978 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
979 * This function is called to allow device receiver to make
980 * changes to configuration when multicast or promiscuous is enabled.
981 *
982 * void (*ndo_set_rx_mode)(struct net_device *dev);
983 * This function is called device changes address list filtering.
984 * If driver handles unicast address filtering, it should set
985 * IFF_UNICAST_FLT in its priv_flags.
986 *
987 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
988 * This function is called when the Media Access Control address
989 * needs to be changed. If this interface is not defined, the
990 * MAC address can not be changed.
991 *
992 * int (*ndo_validate_addr)(struct net_device *dev);
993 * Test if Media Access Control address is valid for the device.
994 *
995 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
996 * Called when a user requests an ioctl which can't be handled by
997 * the generic interface code. If not defined ioctls return
998 * not supported error code.
999 *
1000 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1001 * Used to set network devices bus interface parameters. This interface
1002 * is retained for legacy reasons; new devices should use the bus
1003 * interface (PCI) for low level management.
1004 *
1005 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1006 * Called when a user wants to change the Maximum Transfer Unit
1007 * of a device.
1008 *
1009 * void (*ndo_tx_timeout)(struct net_device *dev);
1010 * Callback used when the transmitter has not made any progress
1011 * for dev->watchdog ticks.
1012 *
1013 * void (*ndo_get_stats64)(struct net_device *dev,
1014 * struct rtnl_link_stats64 *storage);
1015 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1016 * Called when a user wants to get the network device usage
1017 * statistics. Drivers must do one of the following:
1018 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1019 * rtnl_link_stats64 structure passed by the caller.
1020 * 2. Define @ndo_get_stats to update a net_device_stats structure
1021 * (which should normally be dev->stats) and return a pointer to
1022 * it. The structure may be changed asynchronously only if each
1023 * field is written atomically.
1024 * 3. Update dev->stats asynchronously and atomically, and define
1025 * neither operation.
1026 *
1027 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1028 * Return true if this device supports offload stats of this attr_id.
1029 *
1030 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1031 * void *attr_data)
1032 * Get statistics for offload operations by attr_id. Write it into the
1033 * attr_data pointer.
1034 *
1035 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1036 * If device supports VLAN filtering this function is called when a
1037 * VLAN id is registered.
1038 *
1039 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1040 * If device supports VLAN filtering this function is called when a
1041 * VLAN id is unregistered.
1042 *
1043 * void (*ndo_poll_controller)(struct net_device *dev);
1044 *
1045 * SR-IOV management functions.
1046 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1047 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1048 * u8 qos, __be16 proto);
1049 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1050 * int max_tx_rate);
1051 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1052 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1053 * int (*ndo_get_vf_config)(struct net_device *dev,
1054 * int vf, struct ifla_vf_info *ivf);
1055 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1056 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1057 * struct nlattr *port[]);
1058 *
1059 * Enable or disable the VF ability to query its RSS Redirection Table and
1060 * Hash Key. This is needed since on some devices VF share this information
1061 * with PF and querying it may introduce a theoretical security risk.
1062 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1063 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1064 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1065 * void *type_data);
1066 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1067 * This is always called from the stack with the rtnl lock held and netif
1068 * tx queues stopped. This allows the netdevice to perform queue
1069 * management safely.
1070 *
1071 * Fiber Channel over Ethernet (FCoE) offload functions.
1072 * int (*ndo_fcoe_enable)(struct net_device *dev);
1073 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1074 * so the underlying device can perform whatever needed configuration or
1075 * initialization to support acceleration of FCoE traffic.
1076 *
1077 * int (*ndo_fcoe_disable)(struct net_device *dev);
1078 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1079 * so the underlying device can perform whatever needed clean-ups to
1080 * stop supporting acceleration of FCoE traffic.
1081 *
1082 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1083 * struct scatterlist *sgl, unsigned int sgc);
1084 * Called when the FCoE Initiator wants to initialize an I/O that
1085 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1086 * perform necessary setup and returns 1 to indicate the device is set up
1087 * successfully to perform DDP on this I/O, otherwise this returns 0.
1088 *
1089 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1090 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1091 * indicated by the FC exchange id 'xid', so the underlying device can
1092 * clean up and reuse resources for later DDP requests.
1093 *
1094 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1095 * struct scatterlist *sgl, unsigned int sgc);
1096 * Called when the FCoE Target wants to initialize an I/O that
1097 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1098 * perform necessary setup and returns 1 to indicate the device is set up
1099 * successfully to perform DDP on this I/O, otherwise this returns 0.
1100 *
1101 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1102 * struct netdev_fcoe_hbainfo *hbainfo);
1103 * Called when the FCoE Protocol stack wants information on the underlying
1104 * device. This information is utilized by the FCoE protocol stack to
1105 * register attributes with Fiber Channel management service as per the
1106 * FC-GS Fabric Device Management Information(FDMI) specification.
1107 *
1108 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1109 * Called when the underlying device wants to override default World Wide
1110 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1111 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1112 * protocol stack to use.
1113 *
1114 * RFS acceleration.
1115 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1116 * u16 rxq_index, u32 flow_id);
1117 * Set hardware filter for RFS. rxq_index is the target queue index;
1118 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1119 * Return the filter ID on success, or a negative error code.
1120 *
1121 * Slave management functions (for bridge, bonding, etc).
1122 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1123 * Called to make another netdev an underling.
1124 *
1125 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1126 * Called to release previously enslaved netdev.
1127 *
1128 * Feature/offload setting functions.
1129 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1130 * netdev_features_t features);
1131 * Adjusts the requested feature flags according to device-specific
1132 * constraints, and returns the resulting flags. Must not modify
1133 * the device state.
1134 *
1135 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1136 * Called to update device configuration to new features. Passed
1137 * feature set might be less than what was returned by ndo_fix_features()).
1138 * Must return >0 or -errno if it changed dev->features itself.
1139 *
1140 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1141 * struct net_device *dev,
David Brazdil0f672f62019-12-10 10:32:29 +00001142 * const unsigned char *addr, u16 vid, u16 flags,
1143 * struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001144 * Adds an FDB entry to dev for addr.
1145 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1146 * struct net_device *dev,
1147 * const unsigned char *addr, u16 vid)
1148 * Deletes the FDB entry from dev coresponding to addr.
1149 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1150 * struct net_device *dev, struct net_device *filter_dev,
1151 * int *idx)
1152 * Used to add FDB entries to dump requests. Implementers should add
1153 * entries to skb and update idx with the number of entries.
1154 *
1155 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
David Brazdil0f672f62019-12-10 10:32:29 +00001156 * u16 flags, struct netlink_ext_ack *extack)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001157 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1158 * struct net_device *dev, u32 filter_mask,
1159 * int nlflags)
1160 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1161 * u16 flags);
1162 *
1163 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1164 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1165 * which do not represent real hardware may define this to allow their
1166 * userspace components to manage their virtual carrier state. Devices
1167 * that determine carrier state from physical hardware properties (eg
1168 * network cables) or protocol-dependent mechanisms (eg
1169 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1170 *
1171 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1172 * struct netdev_phys_item_id *ppid);
1173 * Called to get ID of physical port of this device. If driver does
1174 * not implement this, it is assumed that the hw is not able to have
1175 * multiple net devices on single physical port.
1176 *
David Brazdil0f672f62019-12-10 10:32:29 +00001177 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1178 * struct netdev_phys_item_id *ppid)
1179 * Called to get the parent ID of the physical port of this device.
1180 *
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001181 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1182 * struct udp_tunnel_info *ti);
1183 * Called by UDP tunnel to notify a driver about the UDP port and socket
1184 * address family that a UDP tunnel is listnening to. It is called only
1185 * when a new port starts listening. The operation is protected by the
1186 * RTNL.
1187 *
1188 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1189 * struct udp_tunnel_info *ti);
1190 * Called by UDP tunnel to notify the driver about a UDP port and socket
1191 * address family that the UDP tunnel is not listening to anymore. The
1192 * operation is protected by the RTNL.
1193 *
1194 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1195 * struct net_device *dev)
1196 * Called by upper layer devices to accelerate switching or other
1197 * station functionality into hardware. 'pdev is the lowerdev
1198 * to use for the offload and 'dev' is the net device that will
1199 * back the offload. Returns a pointer to the private structure
1200 * the upper layer will maintain.
1201 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1202 * Called by upper layer device to delete the station created
1203 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1204 * the station and priv is the structure returned by the add
1205 * operation.
1206 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1207 * int queue_index, u32 maxrate);
1208 * Called when a user wants to set a max-rate limitation of specific
1209 * TX queue.
1210 * int (*ndo_get_iflink)(const struct net_device *dev);
1211 * Called to get the iflink value of this device.
1212 * void (*ndo_change_proto_down)(struct net_device *dev,
1213 * bool proto_down);
1214 * This function is used to pass protocol port error state information
1215 * to the switch driver. The switch driver can react to the proto_down
1216 * by doing a phys down on the associated switch port.
1217 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1218 * This function is used to get egress tunnel information for given skb.
1219 * This is useful for retrieving outer tunnel header parameters while
1220 * sampling packet.
1221 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1222 * This function is used to specify the headroom that the skb must
1223 * consider when allocation skb during packet reception. Setting
1224 * appropriate rx headroom value allows avoiding skb head copy on
1225 * forward. Setting a negative value resets the rx headroom to the
1226 * default value.
1227 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1228 * This function is used to set or query state related to XDP on the
1229 * netdevice and manage BPF offload. See definition of
1230 * enum bpf_netdev_command for details.
1231 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1232 * u32 flags);
1233 * This function is used to submit @n XDP packets for transmit on a
1234 * netdevice. Returns number of frames successfully transmitted, frames
1235 * that got dropped are freed/returned via xdp_return_frame().
1236 * Returns negative number, means general error invoking ndo, meaning
1237 * no frames were xmit'ed and core-caller will free all frames.
David Brazdil0f672f62019-12-10 10:32:29 +00001238 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1239 * This function is used to wake up the softirq, ksoftirqd or kthread
1240 * responsible for sending and/or receiving packets on a specific
1241 * queue id bound to an AF_XDP socket. The flags field specifies if
1242 * only RX, only Tx, or both should be woken up using the flags
1243 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1244 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1245 * Get devlink port instance associated with a given netdev.
1246 * Called with a reference on the netdevice and devlink locks only,
1247 * rtnl_lock is not held.
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001248 */
1249struct net_device_ops {
1250 int (*ndo_init)(struct net_device *dev);
1251 void (*ndo_uninit)(struct net_device *dev);
1252 int (*ndo_open)(struct net_device *dev);
1253 int (*ndo_stop)(struct net_device *dev);
1254 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1255 struct net_device *dev);
1256 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1257 struct net_device *dev,
1258 netdev_features_t features);
1259 u16 (*ndo_select_queue)(struct net_device *dev,
1260 struct sk_buff *skb,
David Brazdil0f672f62019-12-10 10:32:29 +00001261 struct net_device *sb_dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001262 void (*ndo_change_rx_flags)(struct net_device *dev,
1263 int flags);
1264 void (*ndo_set_rx_mode)(struct net_device *dev);
1265 int (*ndo_set_mac_address)(struct net_device *dev,
1266 void *addr);
1267 int (*ndo_validate_addr)(struct net_device *dev);
1268 int (*ndo_do_ioctl)(struct net_device *dev,
1269 struct ifreq *ifr, int cmd);
1270 int (*ndo_set_config)(struct net_device *dev,
1271 struct ifmap *map);
1272 int (*ndo_change_mtu)(struct net_device *dev,
1273 int new_mtu);
1274 int (*ndo_neigh_setup)(struct net_device *dev,
1275 struct neigh_parms *);
1276 void (*ndo_tx_timeout) (struct net_device *dev);
1277
1278 void (*ndo_get_stats64)(struct net_device *dev,
1279 struct rtnl_link_stats64 *storage);
1280 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1281 int (*ndo_get_offload_stats)(int attr_id,
1282 const struct net_device *dev,
1283 void *attr_data);
1284 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1285
1286 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1287 __be16 proto, u16 vid);
1288 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1289 __be16 proto, u16 vid);
1290#ifdef CONFIG_NET_POLL_CONTROLLER
1291 void (*ndo_poll_controller)(struct net_device *dev);
1292 int (*ndo_netpoll_setup)(struct net_device *dev,
1293 struct netpoll_info *info);
1294 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1295#endif
1296 int (*ndo_set_vf_mac)(struct net_device *dev,
1297 int queue, u8 *mac);
1298 int (*ndo_set_vf_vlan)(struct net_device *dev,
1299 int queue, u16 vlan,
1300 u8 qos, __be16 proto);
1301 int (*ndo_set_vf_rate)(struct net_device *dev,
1302 int vf, int min_tx_rate,
1303 int max_tx_rate);
1304 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1305 int vf, bool setting);
1306 int (*ndo_set_vf_trust)(struct net_device *dev,
1307 int vf, bool setting);
1308 int (*ndo_get_vf_config)(struct net_device *dev,
1309 int vf,
1310 struct ifla_vf_info *ivf);
1311 int (*ndo_set_vf_link_state)(struct net_device *dev,
1312 int vf, int link_state);
1313 int (*ndo_get_vf_stats)(struct net_device *dev,
1314 int vf,
1315 struct ifla_vf_stats
1316 *vf_stats);
1317 int (*ndo_set_vf_port)(struct net_device *dev,
1318 int vf,
1319 struct nlattr *port[]);
1320 int (*ndo_get_vf_port)(struct net_device *dev,
1321 int vf, struct sk_buff *skb);
1322 int (*ndo_set_vf_guid)(struct net_device *dev,
1323 int vf, u64 guid,
1324 int guid_type);
1325 int (*ndo_set_vf_rss_query_en)(
1326 struct net_device *dev,
1327 int vf, bool setting);
1328 int (*ndo_setup_tc)(struct net_device *dev,
1329 enum tc_setup_type type,
1330 void *type_data);
1331#if IS_ENABLED(CONFIG_FCOE)
1332 int (*ndo_fcoe_enable)(struct net_device *dev);
1333 int (*ndo_fcoe_disable)(struct net_device *dev);
1334 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1335 u16 xid,
1336 struct scatterlist *sgl,
1337 unsigned int sgc);
1338 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1339 u16 xid);
1340 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1341 u16 xid,
1342 struct scatterlist *sgl,
1343 unsigned int sgc);
1344 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1345 struct netdev_fcoe_hbainfo *hbainfo);
1346#endif
1347
1348#if IS_ENABLED(CONFIG_LIBFCOE)
1349#define NETDEV_FCOE_WWNN 0
1350#define NETDEV_FCOE_WWPN 1
1351 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1352 u64 *wwn, int type);
1353#endif
1354
1355#ifdef CONFIG_RFS_ACCEL
1356 int (*ndo_rx_flow_steer)(struct net_device *dev,
1357 const struct sk_buff *skb,
1358 u16 rxq_index,
1359 u32 flow_id);
1360#endif
1361 int (*ndo_add_slave)(struct net_device *dev,
1362 struct net_device *slave_dev,
1363 struct netlink_ext_ack *extack);
1364 int (*ndo_del_slave)(struct net_device *dev,
1365 struct net_device *slave_dev);
1366 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1367 netdev_features_t features);
1368 int (*ndo_set_features)(struct net_device *dev,
1369 netdev_features_t features);
1370 int (*ndo_neigh_construct)(struct net_device *dev,
1371 struct neighbour *n);
1372 void (*ndo_neigh_destroy)(struct net_device *dev,
1373 struct neighbour *n);
1374
1375 int (*ndo_fdb_add)(struct ndmsg *ndm,
1376 struct nlattr *tb[],
1377 struct net_device *dev,
1378 const unsigned char *addr,
1379 u16 vid,
David Brazdil0f672f62019-12-10 10:32:29 +00001380 u16 flags,
1381 struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001382 int (*ndo_fdb_del)(struct ndmsg *ndm,
1383 struct nlattr *tb[],
1384 struct net_device *dev,
1385 const unsigned char *addr,
1386 u16 vid);
1387 int (*ndo_fdb_dump)(struct sk_buff *skb,
1388 struct netlink_callback *cb,
1389 struct net_device *dev,
1390 struct net_device *filter_dev,
1391 int *idx);
David Brazdil0f672f62019-12-10 10:32:29 +00001392 int (*ndo_fdb_get)(struct sk_buff *skb,
1393 struct nlattr *tb[],
1394 struct net_device *dev,
1395 const unsigned char *addr,
1396 u16 vid, u32 portid, u32 seq,
1397 struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001398 int (*ndo_bridge_setlink)(struct net_device *dev,
1399 struct nlmsghdr *nlh,
David Brazdil0f672f62019-12-10 10:32:29 +00001400 u16 flags,
1401 struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001402 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1403 u32 pid, u32 seq,
1404 struct net_device *dev,
1405 u32 filter_mask,
1406 int nlflags);
1407 int (*ndo_bridge_dellink)(struct net_device *dev,
1408 struct nlmsghdr *nlh,
1409 u16 flags);
1410 int (*ndo_change_carrier)(struct net_device *dev,
1411 bool new_carrier);
1412 int (*ndo_get_phys_port_id)(struct net_device *dev,
1413 struct netdev_phys_item_id *ppid);
David Brazdil0f672f62019-12-10 10:32:29 +00001414 int (*ndo_get_port_parent_id)(struct net_device *dev,
1415 struct netdev_phys_item_id *ppid);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001416 int (*ndo_get_phys_port_name)(struct net_device *dev,
1417 char *name, size_t len);
1418 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1419 struct udp_tunnel_info *ti);
1420 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1421 struct udp_tunnel_info *ti);
1422 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1423 struct net_device *dev);
1424 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1425 void *priv);
1426
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001427 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1428 int queue_index,
1429 u32 maxrate);
1430 int (*ndo_get_iflink)(const struct net_device *dev);
1431 int (*ndo_change_proto_down)(struct net_device *dev,
1432 bool proto_down);
1433 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1434 struct sk_buff *skb);
1435 void (*ndo_set_rx_headroom)(struct net_device *dev,
1436 int needed_headroom);
1437 int (*ndo_bpf)(struct net_device *dev,
1438 struct netdev_bpf *bpf);
1439 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1440 struct xdp_frame **xdp,
1441 u32 flags);
David Brazdil0f672f62019-12-10 10:32:29 +00001442 int (*ndo_xsk_wakeup)(struct net_device *dev,
1443 u32 queue_id, u32 flags);
1444 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001445};
1446
1447/**
1448 * enum net_device_priv_flags - &struct net_device priv_flags
1449 *
1450 * These are the &struct net_device, they are only set internally
1451 * by drivers and used in the kernel. These flags are invisible to
1452 * userspace; this means that the order of these flags can change
1453 * during any kernel release.
1454 *
1455 * You should have a pretty good reason to be extending these flags.
1456 *
1457 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1458 * @IFF_EBRIDGE: Ethernet bridging device
1459 * @IFF_BONDING: bonding master or slave
1460 * @IFF_ISATAP: ISATAP interface (RFC4214)
1461 * @IFF_WAN_HDLC: WAN HDLC device
1462 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1463 * release skb->dst
1464 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1465 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1466 * @IFF_MACVLAN_PORT: device used as macvlan port
1467 * @IFF_BRIDGE_PORT: device used as bridge port
1468 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1469 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1470 * @IFF_UNICAST_FLT: Supports unicast filtering
1471 * @IFF_TEAM_PORT: device used as team port
1472 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1473 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1474 * change when it's running
1475 * @IFF_MACVLAN: Macvlan device
1476 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1477 * underlying stacked devices
1478 * @IFF_L3MDEV_MASTER: device is an L3 master device
1479 * @IFF_NO_QUEUE: device can run without qdisc attached
1480 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1481 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1482 * @IFF_TEAM: device is a team device
1483 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1484 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1485 * entity (i.e. the master device for bridged veth)
1486 * @IFF_MACSEC: device is a MACsec device
1487 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1488 * @IFF_FAILOVER: device is a failover master device
1489 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
David Brazdil0f672f62019-12-10 10:32:29 +00001490 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1491 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001492 */
1493enum netdev_priv_flags {
1494 IFF_802_1Q_VLAN = 1<<0,
1495 IFF_EBRIDGE = 1<<1,
1496 IFF_BONDING = 1<<2,
1497 IFF_ISATAP = 1<<3,
1498 IFF_WAN_HDLC = 1<<4,
1499 IFF_XMIT_DST_RELEASE = 1<<5,
1500 IFF_DONT_BRIDGE = 1<<6,
1501 IFF_DISABLE_NETPOLL = 1<<7,
1502 IFF_MACVLAN_PORT = 1<<8,
1503 IFF_BRIDGE_PORT = 1<<9,
1504 IFF_OVS_DATAPATH = 1<<10,
1505 IFF_TX_SKB_SHARING = 1<<11,
1506 IFF_UNICAST_FLT = 1<<12,
1507 IFF_TEAM_PORT = 1<<13,
1508 IFF_SUPP_NOFCS = 1<<14,
1509 IFF_LIVE_ADDR_CHANGE = 1<<15,
1510 IFF_MACVLAN = 1<<16,
1511 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1512 IFF_L3MDEV_MASTER = 1<<18,
1513 IFF_NO_QUEUE = 1<<19,
1514 IFF_OPENVSWITCH = 1<<20,
1515 IFF_L3MDEV_SLAVE = 1<<21,
1516 IFF_TEAM = 1<<22,
1517 IFF_RXFH_CONFIGURED = 1<<23,
1518 IFF_PHONY_HEADROOM = 1<<24,
1519 IFF_MACSEC = 1<<25,
1520 IFF_NO_RX_HANDLER = 1<<26,
1521 IFF_FAILOVER = 1<<27,
1522 IFF_FAILOVER_SLAVE = 1<<28,
David Brazdil0f672f62019-12-10 10:32:29 +00001523 IFF_L3MDEV_RX_HANDLER = 1<<29,
1524 IFF_LIVE_RENAME_OK = 1<<30,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001525};
1526
1527#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1528#define IFF_EBRIDGE IFF_EBRIDGE
1529#define IFF_BONDING IFF_BONDING
1530#define IFF_ISATAP IFF_ISATAP
1531#define IFF_WAN_HDLC IFF_WAN_HDLC
1532#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1533#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1534#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1535#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1536#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1537#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1538#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1539#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1540#define IFF_TEAM_PORT IFF_TEAM_PORT
1541#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1542#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1543#define IFF_MACVLAN IFF_MACVLAN
1544#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1545#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1546#define IFF_NO_QUEUE IFF_NO_QUEUE
1547#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1548#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1549#define IFF_TEAM IFF_TEAM
1550#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1551#define IFF_MACSEC IFF_MACSEC
1552#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1553#define IFF_FAILOVER IFF_FAILOVER
1554#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
David Brazdil0f672f62019-12-10 10:32:29 +00001555#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1556#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001557
Olivier Deprez0e641232021-09-23 10:07:05 +02001558/* Specifies the type of the struct net_device::ml_priv pointer */
1559enum netdev_ml_priv_type {
1560 ML_PRIV_NONE,
1561 ML_PRIV_CAN,
1562};
1563
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001564/**
1565 * struct net_device - The DEVICE structure.
1566 *
1567 * Actually, this whole structure is a big mistake. It mixes I/O
1568 * data with strictly "high-level" data, and it has to know about
1569 * almost every data structure used in the INET module.
1570 *
1571 * @name: This is the first field of the "visible" part of this structure
1572 * (i.e. as seen by users in the "Space.c" file). It is the name
1573 * of the interface.
1574 *
1575 * @name_hlist: Device name hash chain, please keep it close to name[]
1576 * @ifalias: SNMP alias
1577 * @mem_end: Shared memory end
1578 * @mem_start: Shared memory start
1579 * @base_addr: Device I/O address
1580 * @irq: Device IRQ number
1581 *
1582 * @state: Generic network queuing layer state, see netdev_state_t
1583 * @dev_list: The global list of network devices
1584 * @napi_list: List entry used for polling NAPI devices
1585 * @unreg_list: List entry when we are unregistering the
1586 * device; see the function unregister_netdev
1587 * @close_list: List entry used when we are closing the device
1588 * @ptype_all: Device-specific packet handlers for all protocols
1589 * @ptype_specific: Device-specific, protocol-specific packet handlers
1590 *
1591 * @adj_list: Directly linked devices, like slaves for bonding
1592 * @features: Currently active device features
1593 * @hw_features: User-changeable features
1594 *
1595 * @wanted_features: User-requested features
1596 * @vlan_features: Mask of features inheritable by VLAN devices
1597 *
1598 * @hw_enc_features: Mask of features inherited by encapsulating devices
1599 * This field indicates what encapsulation
1600 * offloads the hardware is capable of doing,
1601 * and drivers will need to set them appropriately.
1602 *
1603 * @mpls_features: Mask of features inheritable by MPLS
1604 *
1605 * @ifindex: interface index
1606 * @group: The group the device belongs to
1607 *
1608 * @stats: Statistics struct, which was left as a legacy, use
1609 * rtnl_link_stats64 instead
1610 *
1611 * @rx_dropped: Dropped packets by core network,
1612 * do not use this in drivers
1613 * @tx_dropped: Dropped packets by core network,
1614 * do not use this in drivers
1615 * @rx_nohandler: nohandler dropped packets by core network on
1616 * inactive devices, do not use this in drivers
1617 * @carrier_up_count: Number of times the carrier has been up
1618 * @carrier_down_count: Number of times the carrier has been down
1619 *
1620 * @wireless_handlers: List of functions to handle Wireless Extensions,
1621 * instead of ioctl,
1622 * see <net/iw_handler.h> for details.
1623 * @wireless_data: Instance data managed by the core of wireless extensions
1624 *
1625 * @netdev_ops: Includes several pointers to callbacks,
1626 * if one wants to override the ndo_*() functions
1627 * @ethtool_ops: Management operations
1628 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1629 * discovery handling. Necessary for e.g. 6LoWPAN.
1630 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1631 * of Layer 2 headers.
1632 *
1633 * @flags: Interface flags (a la BSD)
1634 * @priv_flags: Like 'flags' but invisible to userspace,
1635 * see if.h for the definitions
1636 * @gflags: Global flags ( kept as legacy )
1637 * @padded: How much padding added by alloc_netdev()
1638 * @operstate: RFC2863 operstate
1639 * @link_mode: Mapping policy to operstate
1640 * @if_port: Selectable AUI, TP, ...
1641 * @dma: DMA channel
1642 * @mtu: Interface MTU value
1643 * @min_mtu: Interface Minimum MTU value
1644 * @max_mtu: Interface Maximum MTU value
1645 * @type: Interface hardware type
1646 * @hard_header_len: Maximum hardware header length.
1647 * @min_header_len: Minimum hardware header length
1648 *
1649 * @needed_headroom: Extra headroom the hardware may need, but not in all
1650 * cases can this be guaranteed
1651 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1652 * cases can this be guaranteed. Some cases also use
1653 * LL_MAX_HEADER instead to allocate the skb
1654 *
1655 * interface address info:
1656 *
1657 * @perm_addr: Permanent hw address
1658 * @addr_assign_type: Hw address assignment type
1659 * @addr_len: Hardware address length
David Brazdil0f672f62019-12-10 10:32:29 +00001660 * @upper_level: Maximum depth level of upper devices.
1661 * @lower_level: Maximum depth level of lower devices.
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001662 * @neigh_priv_len: Used in neigh_alloc()
1663 * @dev_id: Used to differentiate devices that share
1664 * the same link layer address
1665 * @dev_port: Used to differentiate devices that share
1666 * the same function
1667 * @addr_list_lock: XXX: need comments on this one
1668 * @uc_promisc: Counter that indicates promiscuous mode
1669 * has been enabled due to the need to listen to
1670 * additional unicast addresses in a device that
1671 * does not implement ndo_set_rx_mode()
1672 * @uc: unicast mac addresses
1673 * @mc: multicast mac addresses
1674 * @dev_addrs: list of device hw addresses
1675 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1676 * @promiscuity: Number of times the NIC is told to work in
1677 * promiscuous mode; if it becomes 0 the NIC will
1678 * exit promiscuous mode
1679 * @allmulti: Counter, enables or disables allmulticast mode
1680 *
1681 * @vlan_info: VLAN info
1682 * @dsa_ptr: dsa specific data
1683 * @tipc_ptr: TIPC specific data
1684 * @atalk_ptr: AppleTalk link
1685 * @ip_ptr: IPv4 specific data
1686 * @dn_ptr: DECnet specific data
1687 * @ip6_ptr: IPv6 specific data
1688 * @ax25_ptr: AX.25 specific data
1689 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1690 *
1691 * @dev_addr: Hw address (before bcast,
1692 * because most packets are unicast)
1693 *
1694 * @_rx: Array of RX queues
1695 * @num_rx_queues: Number of RX queues
1696 * allocated at register_netdev() time
1697 * @real_num_rx_queues: Number of RX queues currently active in device
1698 *
1699 * @rx_handler: handler for received packets
1700 * @rx_handler_data: XXX: need comments on this one
1701 * @miniq_ingress: ingress/clsact qdisc specific data for
1702 * ingress processing
1703 * @ingress_queue: XXX: need comments on this one
1704 * @broadcast: hw bcast address
1705 *
1706 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1707 * indexed by RX queue number. Assigned by driver.
1708 * This must only be set if the ndo_rx_flow_steer
1709 * operation is defined
1710 * @index_hlist: Device index hash chain
1711 *
1712 * @_tx: Array of TX queues
1713 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1714 * @real_num_tx_queues: Number of TX queues currently active in device
1715 * @qdisc: Root qdisc from userspace point of view
1716 * @tx_queue_len: Max frames per queue allowed
1717 * @tx_global_lock: XXX: need comments on this one
1718 *
1719 * @xps_maps: XXX: need comments on this one
1720 * @miniq_egress: clsact qdisc specific data for
1721 * egress processing
1722 * @watchdog_timeo: Represents the timeout that is used by
1723 * the watchdog (see dev_watchdog())
1724 * @watchdog_timer: List of timers
1725 *
1726 * @pcpu_refcnt: Number of references to this device
1727 * @todo_list: Delayed register/unregister
1728 * @link_watch_list: XXX: need comments on this one
1729 *
1730 * @reg_state: Register/unregister state machine
1731 * @dismantle: Device is going to be freed
1732 * @rtnl_link_state: This enum represents the phases of creating
1733 * a new link
1734 *
1735 * @needs_free_netdev: Should unregister perform free_netdev?
1736 * @priv_destructor: Called from unregister
1737 * @npinfo: XXX: need comments on this one
1738 * @nd_net: Network namespace this network device is inside
1739 *
1740 * @ml_priv: Mid-layer private
Olivier Deprez0e641232021-09-23 10:07:05 +02001741 * @ml_priv_type: Mid-layer private type
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001742 * @lstats: Loopback statistics
1743 * @tstats: Tunnel statistics
1744 * @dstats: Dummy statistics
1745 * @vstats: Virtual ethernet statistics
1746 *
1747 * @garp_port: GARP
1748 * @mrp_port: MRP
1749 *
1750 * @dev: Class/net/name entry
1751 * @sysfs_groups: Space for optional device, statistics and wireless
1752 * sysfs groups
1753 *
1754 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1755 * @rtnl_link_ops: Rtnl_link_ops
1756 *
1757 * @gso_max_size: Maximum size of generic segmentation offload
1758 * @gso_max_segs: Maximum number of segments that can be passed to the
1759 * NIC for GSO
1760 *
1761 * @dcbnl_ops: Data Center Bridging netlink ops
1762 * @num_tc: Number of traffic classes in the net device
1763 * @tc_to_txq: XXX: need comments on this one
1764 * @prio_tc_map: XXX: need comments on this one
1765 *
1766 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1767 *
1768 * @priomap: XXX: need comments on this one
1769 * @phydev: Physical device may attach itself
1770 * for hardware timestamping
1771 * @sfp_bus: attached &struct sfp_bus structure.
David Brazdil0f672f62019-12-10 10:32:29 +00001772 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
Olivier Deprez0e641232021-09-23 10:07:05 +02001773 * spinlock
David Brazdil0f672f62019-12-10 10:32:29 +00001774 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1775 * @qdisc_xmit_lock_key: lockdep class annotating
1776 * netdev_queue->_xmit_lock spinlock
1777 * @addr_list_lock_key: lockdep class annotating
1778 * net_device->addr_list_lock spinlock
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001779 *
1780 * @proto_down: protocol port state information can be sent to the
1781 * switch driver and used to set the phys state of the
1782 * switch port.
1783 *
1784 * @wol_enabled: Wake-on-LAN is enabled
1785 *
1786 * FIXME: cleanup struct net_device such that network protocol info
1787 * moves out.
1788 */
1789
1790struct net_device {
1791 char name[IFNAMSIZ];
1792 struct hlist_node name_hlist;
1793 struct dev_ifalias __rcu *ifalias;
1794 /*
1795 * I/O specific fields
1796 * FIXME: Merge these and struct ifmap into one
1797 */
1798 unsigned long mem_end;
1799 unsigned long mem_start;
1800 unsigned long base_addr;
1801 int irq;
1802
1803 /*
1804 * Some hardware also needs these fields (state,dev_list,
1805 * napi_list,unreg_list,close_list) but they are not
1806 * part of the usual set specified in Space.c.
1807 */
1808
1809 unsigned long state;
1810
1811 struct list_head dev_list;
1812 struct list_head napi_list;
1813 struct list_head unreg_list;
1814 struct list_head close_list;
1815 struct list_head ptype_all;
1816 struct list_head ptype_specific;
1817
1818 struct {
1819 struct list_head upper;
1820 struct list_head lower;
1821 } adj_list;
1822
1823 netdev_features_t features;
1824 netdev_features_t hw_features;
1825 netdev_features_t wanted_features;
1826 netdev_features_t vlan_features;
1827 netdev_features_t hw_enc_features;
1828 netdev_features_t mpls_features;
1829 netdev_features_t gso_partial_features;
1830
1831 int ifindex;
1832 int group;
1833
1834 struct net_device_stats stats;
1835
1836 atomic_long_t rx_dropped;
1837 atomic_long_t tx_dropped;
1838 atomic_long_t rx_nohandler;
1839
1840 /* Stats to monitor link on/off, flapping */
1841 atomic_t carrier_up_count;
1842 atomic_t carrier_down_count;
1843
1844#ifdef CONFIG_WIRELESS_EXT
1845 const struct iw_handler_def *wireless_handlers;
1846 struct iw_public_data *wireless_data;
1847#endif
1848 const struct net_device_ops *netdev_ops;
1849 const struct ethtool_ops *ethtool_ops;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001850#ifdef CONFIG_NET_L3_MASTER_DEV
1851 const struct l3mdev_ops *l3mdev_ops;
1852#endif
1853#if IS_ENABLED(CONFIG_IPV6)
1854 const struct ndisc_ops *ndisc_ops;
1855#endif
1856
1857#ifdef CONFIG_XFRM_OFFLOAD
1858 const struct xfrmdev_ops *xfrmdev_ops;
1859#endif
1860
1861#if IS_ENABLED(CONFIG_TLS_DEVICE)
1862 const struct tlsdev_ops *tlsdev_ops;
1863#endif
1864
1865 const struct header_ops *header_ops;
1866
1867 unsigned int flags;
1868 unsigned int priv_flags;
1869
1870 unsigned short gflags;
1871 unsigned short padded;
1872
1873 unsigned char operstate;
1874 unsigned char link_mode;
1875
1876 unsigned char if_port;
1877 unsigned char dma;
1878
Olivier Deprez0e641232021-09-23 10:07:05 +02001879 /* Note : dev->mtu is often read without holding a lock.
1880 * Writers usually hold RTNL.
1881 * It is recommended to use READ_ONCE() to annotate the reads,
1882 * and to use WRITE_ONCE() to annotate the writes.
1883 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001884 unsigned int mtu;
1885 unsigned int min_mtu;
1886 unsigned int max_mtu;
1887 unsigned short type;
1888 unsigned short hard_header_len;
1889 unsigned char min_header_len;
1890
1891 unsigned short needed_headroom;
1892 unsigned short needed_tailroom;
1893
1894 /* Interface address info. */
1895 unsigned char perm_addr[MAX_ADDR_LEN];
1896 unsigned char addr_assign_type;
1897 unsigned char addr_len;
David Brazdil0f672f62019-12-10 10:32:29 +00001898 unsigned char upper_level;
1899 unsigned char lower_level;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001900 unsigned short neigh_priv_len;
1901 unsigned short dev_id;
1902 unsigned short dev_port;
1903 spinlock_t addr_list_lock;
1904 unsigned char name_assign_type;
1905 bool uc_promisc;
1906 struct netdev_hw_addr_list uc;
1907 struct netdev_hw_addr_list mc;
1908 struct netdev_hw_addr_list dev_addrs;
1909
1910#ifdef CONFIG_SYSFS
1911 struct kset *queues_kset;
1912#endif
1913 unsigned int promiscuity;
1914 unsigned int allmulti;
1915
1916
1917 /* Protocol-specific pointers */
1918
1919#if IS_ENABLED(CONFIG_VLAN_8021Q)
1920 struct vlan_info __rcu *vlan_info;
1921#endif
1922#if IS_ENABLED(CONFIG_NET_DSA)
1923 struct dsa_port *dsa_ptr;
1924#endif
1925#if IS_ENABLED(CONFIG_TIPC)
1926 struct tipc_bearer __rcu *tipc_ptr;
1927#endif
1928#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1929 void *atalk_ptr;
1930#endif
1931 struct in_device __rcu *ip_ptr;
1932#if IS_ENABLED(CONFIG_DECNET)
1933 struct dn_dev __rcu *dn_ptr;
1934#endif
1935 struct inet6_dev __rcu *ip6_ptr;
1936#if IS_ENABLED(CONFIG_AX25)
1937 void *ax25_ptr;
1938#endif
1939 struct wireless_dev *ieee80211_ptr;
1940 struct wpan_dev *ieee802154_ptr;
1941#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1942 struct mpls_dev __rcu *mpls_ptr;
1943#endif
1944
1945/*
1946 * Cache lines mostly used on receive path (including eth_type_trans())
1947 */
1948 /* Interface address info used in eth_type_trans() */
1949 unsigned char *dev_addr;
1950
1951 struct netdev_rx_queue *_rx;
1952 unsigned int num_rx_queues;
1953 unsigned int real_num_rx_queues;
1954
1955 struct bpf_prog __rcu *xdp_prog;
1956 unsigned long gro_flush_timeout;
1957 rx_handler_func_t __rcu *rx_handler;
1958 void __rcu *rx_handler_data;
1959
1960#ifdef CONFIG_NET_CLS_ACT
1961 struct mini_Qdisc __rcu *miniq_ingress;
1962#endif
1963 struct netdev_queue __rcu *ingress_queue;
1964#ifdef CONFIG_NETFILTER_INGRESS
1965 struct nf_hook_entries __rcu *nf_hooks_ingress;
1966#endif
1967
1968 unsigned char broadcast[MAX_ADDR_LEN];
1969#ifdef CONFIG_RFS_ACCEL
1970 struct cpu_rmap *rx_cpu_rmap;
1971#endif
1972 struct hlist_node index_hlist;
1973
1974/*
1975 * Cache lines mostly used on transmit path
1976 */
1977 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1978 unsigned int num_tx_queues;
1979 unsigned int real_num_tx_queues;
1980 struct Qdisc *qdisc;
1981#ifdef CONFIG_NET_SCHED
1982 DECLARE_HASHTABLE (qdisc_hash, 4);
1983#endif
1984 unsigned int tx_queue_len;
1985 spinlock_t tx_global_lock;
1986 int watchdog_timeo;
1987
1988#ifdef CONFIG_XPS
1989 struct xps_dev_maps __rcu *xps_cpus_map;
1990 struct xps_dev_maps __rcu *xps_rxqs_map;
1991#endif
1992#ifdef CONFIG_NET_CLS_ACT
1993 struct mini_Qdisc __rcu *miniq_egress;
1994#endif
1995
1996 /* These may be needed for future network-power-down code. */
1997 struct timer_list watchdog_timer;
1998
1999 int __percpu *pcpu_refcnt;
2000 struct list_head todo_list;
2001
2002 struct list_head link_watch_list;
2003
2004 enum { NETREG_UNINITIALIZED=0,
2005 NETREG_REGISTERED, /* completed register_netdevice */
2006 NETREG_UNREGISTERING, /* called unregister_netdevice */
2007 NETREG_UNREGISTERED, /* completed unregister todo */
2008 NETREG_RELEASED, /* called free_netdev */
2009 NETREG_DUMMY, /* dummy device for NAPI poll */
2010 } reg_state:8;
2011
2012 bool dismantle;
2013
2014 enum {
2015 RTNL_LINK_INITIALIZED,
2016 RTNL_LINK_INITIALIZING,
2017 } rtnl_link_state:16;
2018
2019 bool needs_free_netdev;
2020 void (*priv_destructor)(struct net_device *dev);
2021
2022#ifdef CONFIG_NETPOLL
2023 struct netpoll_info __rcu *npinfo;
2024#endif
2025
2026 possible_net_t nd_net;
2027
2028 /* mid-layer private */
Olivier Deprez0e641232021-09-23 10:07:05 +02002029 void *ml_priv;
2030 enum netdev_ml_priv_type ml_priv_type;
2031
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002032 union {
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002033 struct pcpu_lstats __percpu *lstats;
2034 struct pcpu_sw_netstats __percpu *tstats;
2035 struct pcpu_dstats __percpu *dstats;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002036 };
2037
2038#if IS_ENABLED(CONFIG_GARP)
2039 struct garp_port __rcu *garp_port;
2040#endif
2041#if IS_ENABLED(CONFIG_MRP)
2042 struct mrp_port __rcu *mrp_port;
2043#endif
2044
2045 struct device dev;
2046 const struct attribute_group *sysfs_groups[4];
2047 const struct attribute_group *sysfs_rx_queue_group;
2048
2049 const struct rtnl_link_ops *rtnl_link_ops;
2050
2051 /* for setting kernel sock attribute on TCP connection setup */
2052#define GSO_MAX_SIZE 65536
2053 unsigned int gso_max_size;
2054#define GSO_MAX_SEGS 65535
2055 u16 gso_max_segs;
2056
2057#ifdef CONFIG_DCB
2058 const struct dcbnl_rtnl_ops *dcbnl_ops;
2059#endif
2060 s16 num_tc;
2061 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2062 u8 prio_tc_map[TC_BITMASK + 1];
2063
2064#if IS_ENABLED(CONFIG_FCOE)
2065 unsigned int fcoe_ddp_xid;
2066#endif
2067#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2068 struct netprio_map __rcu *priomap;
2069#endif
2070 struct phy_device *phydev;
2071 struct sfp_bus *sfp_bus;
David Brazdil0f672f62019-12-10 10:32:29 +00002072 struct lock_class_key qdisc_tx_busylock_key;
2073 struct lock_class_key qdisc_running_key;
2074 struct lock_class_key qdisc_xmit_lock_key;
2075 struct lock_class_key addr_list_lock_key;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002076 bool proto_down;
2077 unsigned wol_enabled:1;
2078};
2079#define to_net_dev(d) container_of(d, struct net_device, dev)
2080
2081static inline bool netif_elide_gro(const struct net_device *dev)
2082{
2083 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2084 return true;
2085 return false;
2086}
2087
2088#define NETDEV_ALIGN 32
2089
2090static inline
2091int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2092{
2093 return dev->prio_tc_map[prio & TC_BITMASK];
2094}
2095
2096static inline
2097int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2098{
2099 if (tc >= dev->num_tc)
2100 return -EINVAL;
2101
2102 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2103 return 0;
2104}
2105
2106int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2107void netdev_reset_tc(struct net_device *dev);
2108int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2109int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2110
2111static inline
2112int netdev_get_num_tc(struct net_device *dev)
2113{
2114 return dev->num_tc;
2115}
2116
2117void netdev_unbind_sb_channel(struct net_device *dev,
2118 struct net_device *sb_dev);
2119int netdev_bind_sb_channel_queue(struct net_device *dev,
2120 struct net_device *sb_dev,
2121 u8 tc, u16 count, u16 offset);
2122int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2123static inline int netdev_get_sb_channel(struct net_device *dev)
2124{
2125 return max_t(int, -dev->num_tc, 0);
2126}
2127
2128static inline
2129struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2130 unsigned int index)
2131{
2132 return &dev->_tx[index];
2133}
2134
2135static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2136 const struct sk_buff *skb)
2137{
2138 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2139}
2140
2141static inline void netdev_for_each_tx_queue(struct net_device *dev,
2142 void (*f)(struct net_device *,
2143 struct netdev_queue *,
2144 void *),
2145 void *arg)
2146{
2147 unsigned int i;
2148
2149 for (i = 0; i < dev->num_tx_queues; i++)
2150 f(dev, &dev->_tx[i], arg);
2151}
2152
David Brazdil0f672f62019-12-10 10:32:29 +00002153u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2154 struct net_device *sb_dev);
2155struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2156 struct sk_buff *skb,
2157 struct net_device *sb_dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002158
2159/* returns the headroom that the master device needs to take in account
2160 * when forwarding to this dev
2161 */
2162static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2163{
2164 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2165}
2166
2167static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2168{
2169 if (dev->netdev_ops->ndo_set_rx_headroom)
2170 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2171}
2172
2173/* set the device rx headroom to the dev's default */
2174static inline void netdev_reset_rx_headroom(struct net_device *dev)
2175{
2176 netdev_set_rx_headroom(dev, -1);
2177}
2178
Olivier Deprez0e641232021-09-23 10:07:05 +02002179static inline void *netdev_get_ml_priv(struct net_device *dev,
2180 enum netdev_ml_priv_type type)
2181{
2182 if (dev->ml_priv_type != type)
2183 return NULL;
2184
2185 return dev->ml_priv;
2186}
2187
2188static inline void netdev_set_ml_priv(struct net_device *dev,
2189 void *ml_priv,
2190 enum netdev_ml_priv_type type)
2191{
2192 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2193 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2194 dev->ml_priv_type, type);
2195 WARN(!dev->ml_priv_type && dev->ml_priv,
2196 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2197
2198 dev->ml_priv = ml_priv;
2199 dev->ml_priv_type = type;
2200}
2201
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002202/*
2203 * Net namespace inlines
2204 */
2205static inline
2206struct net *dev_net(const struct net_device *dev)
2207{
2208 return read_pnet(&dev->nd_net);
2209}
2210
2211static inline
2212void dev_net_set(struct net_device *dev, struct net *net)
2213{
2214 write_pnet(&dev->nd_net, net);
2215}
2216
2217/**
2218 * netdev_priv - access network device private data
2219 * @dev: network device
2220 *
2221 * Get network device private data
2222 */
2223static inline void *netdev_priv(const struct net_device *dev)
2224{
2225 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2226}
2227
2228/* Set the sysfs physical device reference for the network logical device
2229 * if set prior to registration will cause a symlink during initialization.
2230 */
2231#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2232
2233/* Set the sysfs device type for the network logical device to allow
2234 * fine-grained identification of different network device types. For
2235 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2236 */
2237#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2238
2239/* Default NAPI poll() weight
2240 * Device drivers are strongly advised to not use bigger value
2241 */
2242#define NAPI_POLL_WEIGHT 64
2243
2244/**
2245 * netif_napi_add - initialize a NAPI context
2246 * @dev: network device
2247 * @napi: NAPI context
2248 * @poll: polling function
2249 * @weight: default weight
2250 *
2251 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2252 * *any* of the other NAPI-related functions.
2253 */
2254void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2255 int (*poll)(struct napi_struct *, int), int weight);
2256
2257/**
2258 * netif_tx_napi_add - initialize a NAPI context
2259 * @dev: network device
2260 * @napi: NAPI context
2261 * @poll: polling function
2262 * @weight: default weight
2263 *
2264 * This variant of netif_napi_add() should be used from drivers using NAPI
2265 * to exclusively poll a TX queue.
2266 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2267 */
2268static inline void netif_tx_napi_add(struct net_device *dev,
2269 struct napi_struct *napi,
2270 int (*poll)(struct napi_struct *, int),
2271 int weight)
2272{
2273 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2274 netif_napi_add(dev, napi, poll, weight);
2275}
2276
2277/**
2278 * netif_napi_del - remove a NAPI context
2279 * @napi: NAPI context
2280 *
2281 * netif_napi_del() removes a NAPI context from the network device NAPI list
2282 */
2283void netif_napi_del(struct napi_struct *napi);
2284
2285struct napi_gro_cb {
2286 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2287 void *frag0;
2288
2289 /* Length of frag0. */
2290 unsigned int frag0_len;
2291
2292 /* This indicates where we are processing relative to skb->data. */
2293 int data_offset;
2294
2295 /* This is non-zero if the packet cannot be merged with the new skb. */
2296 u16 flush;
2297
2298 /* Save the IP ID here and check when we get to the transport layer */
2299 u16 flush_id;
2300
2301 /* Number of segments aggregated. */
2302 u16 count;
2303
2304 /* Start offset for remote checksum offload */
2305 u16 gro_remcsum_start;
2306
2307 /* jiffies when first packet was created/queued */
2308 unsigned long age;
2309
2310 /* Used in ipv6_gro_receive() and foo-over-udp */
2311 u16 proto;
2312
2313 /* This is non-zero if the packet may be of the same flow. */
2314 u8 same_flow:1;
2315
2316 /* Used in tunnel GRO receive */
2317 u8 encap_mark:1;
2318
2319 /* GRO checksum is valid */
2320 u8 csum_valid:1;
2321
2322 /* Number of checksums via CHECKSUM_UNNECESSARY */
2323 u8 csum_cnt:3;
2324
2325 /* Free the skb? */
2326 u8 free:2;
2327#define NAPI_GRO_FREE 1
2328#define NAPI_GRO_FREE_STOLEN_HEAD 2
2329
2330 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2331 u8 is_ipv6:1;
2332
2333 /* Used in GRE, set in fou/gue_gro_receive */
2334 u8 is_fou:1;
2335
2336 /* Used to determine if flush_id can be ignored */
2337 u8 is_atomic:1;
2338
2339 /* Number of gro_receive callbacks this packet already went through */
2340 u8 recursion_counter:4;
2341
2342 /* 1 bit hole */
2343
2344 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2345 __wsum csum;
2346
2347 /* used in skb_gro_receive() slow path */
2348 struct sk_buff *last;
2349};
2350
2351#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2352
2353#define GRO_RECURSION_LIMIT 15
2354static inline int gro_recursion_inc_test(struct sk_buff *skb)
2355{
2356 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2357}
2358
2359typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2360static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2361 struct list_head *head,
2362 struct sk_buff *skb)
2363{
2364 if (unlikely(gro_recursion_inc_test(skb))) {
2365 NAPI_GRO_CB(skb)->flush |= 1;
2366 return NULL;
2367 }
2368
2369 return cb(head, skb);
2370}
2371
2372typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2373 struct sk_buff *);
2374static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2375 struct sock *sk,
2376 struct list_head *head,
2377 struct sk_buff *skb)
2378{
2379 if (unlikely(gro_recursion_inc_test(skb))) {
2380 NAPI_GRO_CB(skb)->flush |= 1;
2381 return NULL;
2382 }
2383
2384 return cb(sk, head, skb);
2385}
2386
2387struct packet_type {
2388 __be16 type; /* This is really htons(ether_type). */
David Brazdil0f672f62019-12-10 10:32:29 +00002389 bool ignore_outgoing;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002390 struct net_device *dev; /* NULL is wildcarded here */
2391 int (*func) (struct sk_buff *,
2392 struct net_device *,
2393 struct packet_type *,
2394 struct net_device *);
2395 void (*list_func) (struct list_head *,
2396 struct packet_type *,
2397 struct net_device *);
2398 bool (*id_match)(struct packet_type *ptype,
2399 struct sock *sk);
2400 void *af_packet_priv;
2401 struct list_head list;
2402};
2403
2404struct offload_callbacks {
2405 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2406 netdev_features_t features);
2407 struct sk_buff *(*gro_receive)(struct list_head *head,
2408 struct sk_buff *skb);
2409 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2410};
2411
2412struct packet_offload {
2413 __be16 type; /* This is really htons(ether_type). */
2414 u16 priority;
2415 struct offload_callbacks callbacks;
2416 struct list_head list;
2417};
2418
2419/* often modified stats are per-CPU, other are shared (netdev->stats) */
2420struct pcpu_sw_netstats {
2421 u64 rx_packets;
2422 u64 rx_bytes;
2423 u64 tx_packets;
2424 u64 tx_bytes;
2425 struct u64_stats_sync syncp;
David Brazdil0f672f62019-12-10 10:32:29 +00002426} __aligned(4 * sizeof(u64));
2427
2428struct pcpu_lstats {
2429 u64 packets;
2430 u64 bytes;
2431 struct u64_stats_sync syncp;
2432} __aligned(2 * sizeof(u64));
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002433
2434#define __netdev_alloc_pcpu_stats(type, gfp) \
2435({ \
2436 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2437 if (pcpu_stats) { \
2438 int __cpu; \
2439 for_each_possible_cpu(__cpu) { \
2440 typeof(type) *stat; \
2441 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2442 u64_stats_init(&stat->syncp); \
2443 } \
2444 } \
2445 pcpu_stats; \
2446})
2447
2448#define netdev_alloc_pcpu_stats(type) \
2449 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2450
2451enum netdev_lag_tx_type {
2452 NETDEV_LAG_TX_TYPE_UNKNOWN,
2453 NETDEV_LAG_TX_TYPE_RANDOM,
2454 NETDEV_LAG_TX_TYPE_BROADCAST,
2455 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2456 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2457 NETDEV_LAG_TX_TYPE_HASH,
2458};
2459
2460enum netdev_lag_hash {
2461 NETDEV_LAG_HASH_NONE,
2462 NETDEV_LAG_HASH_L2,
2463 NETDEV_LAG_HASH_L34,
2464 NETDEV_LAG_HASH_L23,
2465 NETDEV_LAG_HASH_E23,
2466 NETDEV_LAG_HASH_E34,
2467 NETDEV_LAG_HASH_UNKNOWN,
2468};
2469
2470struct netdev_lag_upper_info {
2471 enum netdev_lag_tx_type tx_type;
2472 enum netdev_lag_hash hash_type;
2473};
2474
2475struct netdev_lag_lower_state_info {
2476 u8 link_up : 1,
2477 tx_enabled : 1;
2478};
2479
2480#include <linux/notifier.h>
2481
2482/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2483 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2484 * adding new types.
2485 */
2486enum netdev_cmd {
2487 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2488 NETDEV_DOWN,
2489 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2490 detected a hardware crash and restarted
2491 - we can use this eg to kick tcp sessions
2492 once done */
2493 NETDEV_CHANGE, /* Notify device state change */
2494 NETDEV_REGISTER,
2495 NETDEV_UNREGISTER,
2496 NETDEV_CHANGEMTU, /* notify after mtu change happened */
David Brazdil0f672f62019-12-10 10:32:29 +00002497 NETDEV_CHANGEADDR, /* notify after the address change */
2498 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002499 NETDEV_GOING_DOWN,
2500 NETDEV_CHANGENAME,
2501 NETDEV_FEAT_CHANGE,
2502 NETDEV_BONDING_FAILOVER,
2503 NETDEV_PRE_UP,
2504 NETDEV_PRE_TYPE_CHANGE,
2505 NETDEV_POST_TYPE_CHANGE,
2506 NETDEV_POST_INIT,
2507 NETDEV_RELEASE,
2508 NETDEV_NOTIFY_PEERS,
2509 NETDEV_JOIN,
2510 NETDEV_CHANGEUPPER,
2511 NETDEV_RESEND_IGMP,
2512 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2513 NETDEV_CHANGEINFODATA,
2514 NETDEV_BONDING_INFO,
2515 NETDEV_PRECHANGEUPPER,
2516 NETDEV_CHANGELOWERSTATE,
2517 NETDEV_UDP_TUNNEL_PUSH_INFO,
2518 NETDEV_UDP_TUNNEL_DROP_INFO,
2519 NETDEV_CHANGE_TX_QUEUE_LEN,
2520 NETDEV_CVLAN_FILTER_PUSH_INFO,
2521 NETDEV_CVLAN_FILTER_DROP_INFO,
2522 NETDEV_SVLAN_FILTER_PUSH_INFO,
2523 NETDEV_SVLAN_FILTER_DROP_INFO,
2524};
2525const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2526
2527int register_netdevice_notifier(struct notifier_block *nb);
2528int unregister_netdevice_notifier(struct notifier_block *nb);
2529
2530struct netdev_notifier_info {
2531 struct net_device *dev;
2532 struct netlink_ext_ack *extack;
2533};
2534
2535struct netdev_notifier_info_ext {
2536 struct netdev_notifier_info info; /* must be first */
2537 union {
2538 u32 mtu;
2539 } ext;
2540};
2541
2542struct netdev_notifier_change_info {
2543 struct netdev_notifier_info info; /* must be first */
2544 unsigned int flags_changed;
2545};
2546
2547struct netdev_notifier_changeupper_info {
2548 struct netdev_notifier_info info; /* must be first */
2549 struct net_device *upper_dev; /* new upper dev */
2550 bool master; /* is upper dev master */
2551 bool linking; /* is the notification for link or unlink */
2552 void *upper_info; /* upper dev info */
2553};
2554
2555struct netdev_notifier_changelowerstate_info {
2556 struct netdev_notifier_info info; /* must be first */
2557 void *lower_state_info; /* is lower dev state */
2558};
2559
David Brazdil0f672f62019-12-10 10:32:29 +00002560struct netdev_notifier_pre_changeaddr_info {
2561 struct netdev_notifier_info info; /* must be first */
2562 const unsigned char *dev_addr;
2563};
2564
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002565static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2566 struct net_device *dev)
2567{
2568 info->dev = dev;
2569 info->extack = NULL;
2570}
2571
2572static inline struct net_device *
2573netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2574{
2575 return info->dev;
2576}
2577
2578static inline struct netlink_ext_ack *
2579netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2580{
2581 return info->extack;
2582}
2583
2584int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2585
2586
2587extern rwlock_t dev_base_lock; /* Device list lock */
2588
2589#define for_each_netdev(net, d) \
2590 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2591#define for_each_netdev_reverse(net, d) \
2592 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2593#define for_each_netdev_rcu(net, d) \
2594 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2595#define for_each_netdev_safe(net, d, n) \
2596 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2597#define for_each_netdev_continue(net, d) \
2598 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2599#define for_each_netdev_continue_rcu(net, d) \
2600 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2601#define for_each_netdev_in_bond_rcu(bond, slave) \
2602 for_each_netdev_rcu(&init_net, slave) \
2603 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2604#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2605
2606static inline struct net_device *next_net_device(struct net_device *dev)
2607{
2608 struct list_head *lh;
2609 struct net *net;
2610
2611 net = dev_net(dev);
2612 lh = dev->dev_list.next;
2613 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2614}
2615
2616static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2617{
2618 struct list_head *lh;
2619 struct net *net;
2620
2621 net = dev_net(dev);
2622 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2623 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2624}
2625
2626static inline struct net_device *first_net_device(struct net *net)
2627{
2628 return list_empty(&net->dev_base_head) ? NULL :
2629 net_device_entry(net->dev_base_head.next);
2630}
2631
2632static inline struct net_device *first_net_device_rcu(struct net *net)
2633{
2634 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2635
2636 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2637}
2638
2639int netdev_boot_setup_check(struct net_device *dev);
2640unsigned long netdev_boot_base(const char *prefix, int unit);
2641struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2642 const char *hwaddr);
2643struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2644struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2645void dev_add_pack(struct packet_type *pt);
2646void dev_remove_pack(struct packet_type *pt);
2647void __dev_remove_pack(struct packet_type *pt);
2648void dev_add_offload(struct packet_offload *po);
2649void dev_remove_offload(struct packet_offload *po);
2650
2651int dev_get_iflink(const struct net_device *dev);
2652int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2653struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2654 unsigned short mask);
2655struct net_device *dev_get_by_name(struct net *net, const char *name);
2656struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2657struct net_device *__dev_get_by_name(struct net *net, const char *name);
2658int dev_alloc_name(struct net_device *dev, const char *name);
David Brazdil0f672f62019-12-10 10:32:29 +00002659int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002660void dev_close(struct net_device *dev);
2661void dev_close_many(struct list_head *head, bool unlink);
2662void dev_disable_lro(struct net_device *dev);
2663int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2664u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
David Brazdil0f672f62019-12-10 10:32:29 +00002665 struct net_device *sb_dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002666u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
David Brazdil0f672f62019-12-10 10:32:29 +00002667 struct net_device *sb_dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002668int dev_queue_xmit(struct sk_buff *skb);
2669int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2670int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2671int register_netdevice(struct net_device *dev);
2672void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2673void unregister_netdevice_many(struct list_head *head);
2674static inline void unregister_netdevice(struct net_device *dev)
2675{
2676 unregister_netdevice_queue(dev, NULL);
2677}
2678
2679int netdev_refcnt_read(const struct net_device *dev);
2680void free_netdev(struct net_device *dev);
2681void netdev_freemem(struct net_device *dev);
2682void synchronize_net(void);
2683int init_dummy_netdev(struct net_device *dev);
2684
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002685struct net_device *dev_get_by_index(struct net *net, int ifindex);
2686struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2687struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2688struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2689int netdev_get_name(struct net *net, char *name, int ifindex);
2690int dev_restart(struct net_device *dev);
2691int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2692
2693static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2694{
2695 return NAPI_GRO_CB(skb)->data_offset;
2696}
2697
2698static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2699{
2700 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2701}
2702
2703static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2704{
2705 NAPI_GRO_CB(skb)->data_offset += len;
2706}
2707
2708static inline void *skb_gro_header_fast(struct sk_buff *skb,
2709 unsigned int offset)
2710{
2711 return NAPI_GRO_CB(skb)->frag0 + offset;
2712}
2713
2714static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2715{
2716 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2717}
2718
2719static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2720{
2721 NAPI_GRO_CB(skb)->frag0 = NULL;
2722 NAPI_GRO_CB(skb)->frag0_len = 0;
2723}
2724
2725static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2726 unsigned int offset)
2727{
2728 if (!pskb_may_pull(skb, hlen))
2729 return NULL;
2730
2731 skb_gro_frag0_invalidate(skb);
2732 return skb->data + offset;
2733}
2734
2735static inline void *skb_gro_network_header(struct sk_buff *skb)
2736{
2737 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2738 skb_network_offset(skb);
2739}
2740
2741static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2742 const void *start, unsigned int len)
2743{
2744 if (NAPI_GRO_CB(skb)->csum_valid)
2745 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2746 csum_partial(start, len, 0));
2747}
2748
2749/* GRO checksum functions. These are logical equivalents of the normal
2750 * checksum functions (in skbuff.h) except that they operate on the GRO
2751 * offsets and fields in sk_buff.
2752 */
2753
2754__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2755
2756static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2757{
2758 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2759}
2760
2761static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2762 bool zero_okay,
2763 __sum16 check)
2764{
2765 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2766 skb_checksum_start_offset(skb) <
2767 skb_gro_offset(skb)) &&
2768 !skb_at_gro_remcsum_start(skb) &&
2769 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2770 (!zero_okay || check));
2771}
2772
2773static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2774 __wsum psum)
2775{
2776 if (NAPI_GRO_CB(skb)->csum_valid &&
2777 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2778 return 0;
2779
2780 NAPI_GRO_CB(skb)->csum = psum;
2781
2782 return __skb_gro_checksum_complete(skb);
2783}
2784
2785static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2786{
2787 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2788 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2789 NAPI_GRO_CB(skb)->csum_cnt--;
2790 } else {
2791 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2792 * verified a new top level checksum or an encapsulated one
2793 * during GRO. This saves work if we fallback to normal path.
2794 */
2795 __skb_incr_checksum_unnecessary(skb);
2796 }
2797}
2798
2799#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2800 compute_pseudo) \
2801({ \
2802 __sum16 __ret = 0; \
2803 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2804 __ret = __skb_gro_checksum_validate_complete(skb, \
2805 compute_pseudo(skb, proto)); \
2806 if (!__ret) \
2807 skb_gro_incr_csum_unnecessary(skb); \
2808 __ret; \
2809})
2810
2811#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2812 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2813
2814#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2815 compute_pseudo) \
2816 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2817
2818#define skb_gro_checksum_simple_validate(skb) \
2819 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2820
2821static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2822{
2823 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2824 !NAPI_GRO_CB(skb)->csum_valid);
2825}
2826
2827static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2828 __sum16 check, __wsum pseudo)
2829{
2830 NAPI_GRO_CB(skb)->csum = ~pseudo;
2831 NAPI_GRO_CB(skb)->csum_valid = 1;
2832}
2833
2834#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2835do { \
2836 if (__skb_gro_checksum_convert_check(skb)) \
2837 __skb_gro_checksum_convert(skb, check, \
2838 compute_pseudo(skb, proto)); \
2839} while (0)
2840
2841struct gro_remcsum {
2842 int offset;
2843 __wsum delta;
2844};
2845
2846static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2847{
2848 grc->offset = 0;
2849 grc->delta = 0;
2850}
2851
2852static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2853 unsigned int off, size_t hdrlen,
2854 int start, int offset,
2855 struct gro_remcsum *grc,
2856 bool nopartial)
2857{
2858 __wsum delta;
2859 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2860
2861 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2862
2863 if (!nopartial) {
2864 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2865 return ptr;
2866 }
2867
2868 ptr = skb_gro_header_fast(skb, off);
2869 if (skb_gro_header_hard(skb, off + plen)) {
2870 ptr = skb_gro_header_slow(skb, off + plen, off);
2871 if (!ptr)
2872 return NULL;
2873 }
2874
2875 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2876 start, offset);
2877
2878 /* Adjust skb->csum since we changed the packet */
2879 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2880
2881 grc->offset = off + hdrlen + offset;
2882 grc->delta = delta;
2883
2884 return ptr;
2885}
2886
2887static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2888 struct gro_remcsum *grc)
2889{
2890 void *ptr;
2891 size_t plen = grc->offset + sizeof(u16);
2892
2893 if (!grc->delta)
2894 return;
2895
2896 ptr = skb_gro_header_fast(skb, grc->offset);
2897 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2898 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2899 if (!ptr)
2900 return;
2901 }
2902
2903 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2904}
2905
2906#ifdef CONFIG_XFRM_OFFLOAD
2907static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2908{
2909 if (PTR_ERR(pp) != -EINPROGRESS)
2910 NAPI_GRO_CB(skb)->flush |= flush;
2911}
2912static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2913 struct sk_buff *pp,
2914 int flush,
2915 struct gro_remcsum *grc)
2916{
2917 if (PTR_ERR(pp) != -EINPROGRESS) {
2918 NAPI_GRO_CB(skb)->flush |= flush;
2919 skb_gro_remcsum_cleanup(skb, grc);
2920 skb->remcsum_offload = 0;
2921 }
2922}
2923#else
2924static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2925{
2926 NAPI_GRO_CB(skb)->flush |= flush;
2927}
2928static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2929 struct sk_buff *pp,
2930 int flush,
2931 struct gro_remcsum *grc)
2932{
2933 NAPI_GRO_CB(skb)->flush |= flush;
2934 skb_gro_remcsum_cleanup(skb, grc);
2935 skb->remcsum_offload = 0;
2936}
2937#endif
2938
2939static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2940 unsigned short type,
2941 const void *daddr, const void *saddr,
2942 unsigned int len)
2943{
2944 if (!dev->header_ops || !dev->header_ops->create)
2945 return 0;
2946
2947 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2948}
2949
2950static inline int dev_parse_header(const struct sk_buff *skb,
2951 unsigned char *haddr)
2952{
2953 const struct net_device *dev = skb->dev;
2954
2955 if (!dev->header_ops || !dev->header_ops->parse)
2956 return 0;
2957 return dev->header_ops->parse(skb, haddr);
2958}
2959
David Brazdil0f672f62019-12-10 10:32:29 +00002960static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2961{
2962 const struct net_device *dev = skb->dev;
2963
2964 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2965 return 0;
2966 return dev->header_ops->parse_protocol(skb);
2967}
2968
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002969/* ll_header must have at least hard_header_len allocated */
2970static inline bool dev_validate_header(const struct net_device *dev,
2971 char *ll_header, int len)
2972{
2973 if (likely(len >= dev->hard_header_len))
2974 return true;
2975 if (len < dev->min_header_len)
2976 return false;
2977
2978 if (capable(CAP_SYS_RAWIO)) {
2979 memset(ll_header + len, 0, dev->hard_header_len - len);
2980 return true;
2981 }
2982
2983 if (dev->header_ops && dev->header_ops->validate)
2984 return dev->header_ops->validate(ll_header, len);
2985
2986 return false;
2987}
2988
2989typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2990 int len, int size);
2991int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2992static inline int unregister_gifconf(unsigned int family)
2993{
2994 return register_gifconf(family, NULL);
2995}
2996
2997#ifdef CONFIG_NET_FLOW_LIMIT
2998#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2999struct sd_flow_limit {
3000 u64 count;
3001 unsigned int num_buckets;
3002 unsigned int history_head;
3003 u16 history[FLOW_LIMIT_HISTORY];
3004 u8 buckets[];
3005};
3006
3007extern int netdev_flow_limit_table_len;
3008#endif /* CONFIG_NET_FLOW_LIMIT */
3009
3010/*
3011 * Incoming packets are placed on per-CPU queues
3012 */
3013struct softnet_data {
3014 struct list_head poll_list;
3015 struct sk_buff_head process_queue;
3016
3017 /* stats */
3018 unsigned int processed;
3019 unsigned int time_squeeze;
3020 unsigned int received_rps;
3021#ifdef CONFIG_RPS
3022 struct softnet_data *rps_ipi_list;
3023#endif
3024#ifdef CONFIG_NET_FLOW_LIMIT
3025 struct sd_flow_limit __rcu *flow_limit;
3026#endif
3027 struct Qdisc *output_queue;
3028 struct Qdisc **output_queue_tailp;
3029 struct sk_buff *completion_queue;
3030#ifdef CONFIG_XFRM_OFFLOAD
3031 struct sk_buff_head xfrm_backlog;
3032#endif
David Brazdil0f672f62019-12-10 10:32:29 +00003033 /* written and read only by owning cpu: */
3034 struct {
3035 u16 recursion;
3036 u8 more;
3037 } xmit;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003038#ifdef CONFIG_RPS
3039 /* input_queue_head should be written by cpu owning this struct,
3040 * and only read by other cpus. Worth using a cache line.
3041 */
3042 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3043
3044 /* Elements below can be accessed between CPUs for RPS/RFS */
3045 call_single_data_t csd ____cacheline_aligned_in_smp;
3046 struct softnet_data *rps_ipi_next;
3047 unsigned int cpu;
3048 unsigned int input_queue_tail;
3049#endif
3050 unsigned int dropped;
3051 struct sk_buff_head input_pkt_queue;
3052 struct napi_struct backlog;
3053
3054};
3055
3056static inline void input_queue_head_incr(struct softnet_data *sd)
3057{
3058#ifdef CONFIG_RPS
3059 sd->input_queue_head++;
3060#endif
3061}
3062
3063static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3064 unsigned int *qtail)
3065{
3066#ifdef CONFIG_RPS
3067 *qtail = ++sd->input_queue_tail;
3068#endif
3069}
3070
3071DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3072
David Brazdil0f672f62019-12-10 10:32:29 +00003073static inline int dev_recursion_level(void)
3074{
3075 return this_cpu_read(softnet_data.xmit.recursion);
3076}
3077
Olivier Deprez0e641232021-09-23 10:07:05 +02003078#define XMIT_RECURSION_LIMIT 8
David Brazdil0f672f62019-12-10 10:32:29 +00003079static inline bool dev_xmit_recursion(void)
3080{
3081 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3082 XMIT_RECURSION_LIMIT);
3083}
3084
3085static inline void dev_xmit_recursion_inc(void)
3086{
3087 __this_cpu_inc(softnet_data.xmit.recursion);
3088}
3089
3090static inline void dev_xmit_recursion_dec(void)
3091{
3092 __this_cpu_dec(softnet_data.xmit.recursion);
3093}
3094
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003095void __netif_schedule(struct Qdisc *q);
3096void netif_schedule_queue(struct netdev_queue *txq);
3097
3098static inline void netif_tx_schedule_all(struct net_device *dev)
3099{
3100 unsigned int i;
3101
3102 for (i = 0; i < dev->num_tx_queues; i++)
3103 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3104}
3105
3106static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3107{
3108 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3109}
3110
3111/**
3112 * netif_start_queue - allow transmit
3113 * @dev: network device
3114 *
3115 * Allow upper layers to call the device hard_start_xmit routine.
3116 */
3117static inline void netif_start_queue(struct net_device *dev)
3118{
3119 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3120}
3121
3122static inline void netif_tx_start_all_queues(struct net_device *dev)
3123{
3124 unsigned int i;
3125
3126 for (i = 0; i < dev->num_tx_queues; i++) {
3127 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3128 netif_tx_start_queue(txq);
3129 }
3130}
3131
3132void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3133
3134/**
3135 * netif_wake_queue - restart transmit
3136 * @dev: network device
3137 *
3138 * Allow upper layers to call the device hard_start_xmit routine.
3139 * Used for flow control when transmit resources are available.
3140 */
3141static inline void netif_wake_queue(struct net_device *dev)
3142{
3143 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3144}
3145
3146static inline void netif_tx_wake_all_queues(struct net_device *dev)
3147{
3148 unsigned int i;
3149
3150 for (i = 0; i < dev->num_tx_queues; i++) {
3151 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3152 netif_tx_wake_queue(txq);
3153 }
3154}
3155
3156static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3157{
3158 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3159}
3160
3161/**
3162 * netif_stop_queue - stop transmitted packets
3163 * @dev: network device
3164 *
3165 * Stop upper layers calling the device hard_start_xmit routine.
3166 * Used for flow control when transmit resources are unavailable.
3167 */
3168static inline void netif_stop_queue(struct net_device *dev)
3169{
3170 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3171}
3172
3173void netif_tx_stop_all_queues(struct net_device *dev);
David Brazdil0f672f62019-12-10 10:32:29 +00003174void netdev_update_lockdep_key(struct net_device *dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003175
3176static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3177{
3178 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3179}
3180
3181/**
3182 * netif_queue_stopped - test if transmit queue is flowblocked
3183 * @dev: network device
3184 *
3185 * Test if transmit queue on device is currently unable to send.
3186 */
3187static inline bool netif_queue_stopped(const struct net_device *dev)
3188{
3189 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3190}
3191
3192static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3193{
3194 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3195}
3196
3197static inline bool
3198netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3199{
3200 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3201}
3202
3203static inline bool
3204netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3205{
3206 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3207}
3208
3209/**
3210 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3211 * @dev_queue: pointer to transmit queue
3212 *
3213 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3214 * to give appropriate hint to the CPU.
3215 */
3216static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3217{
3218#ifdef CONFIG_BQL
3219 prefetchw(&dev_queue->dql.num_queued);
3220#endif
3221}
3222
3223/**
3224 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3225 * @dev_queue: pointer to transmit queue
3226 *
3227 * BQL enabled drivers might use this helper in their TX completion path,
3228 * to give appropriate hint to the CPU.
3229 */
3230static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3231{
3232#ifdef CONFIG_BQL
3233 prefetchw(&dev_queue->dql.limit);
3234#endif
3235}
3236
3237static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3238 unsigned int bytes)
3239{
3240#ifdef CONFIG_BQL
3241 dql_queued(&dev_queue->dql, bytes);
3242
3243 if (likely(dql_avail(&dev_queue->dql) >= 0))
3244 return;
3245
3246 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3247
3248 /*
3249 * The XOFF flag must be set before checking the dql_avail below,
3250 * because in netdev_tx_completed_queue we update the dql_completed
3251 * before checking the XOFF flag.
3252 */
3253 smp_mb();
3254
3255 /* check again in case another CPU has just made room avail */
3256 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3257 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3258#endif
3259}
3260
David Brazdil0f672f62019-12-10 10:32:29 +00003261/* Variant of netdev_tx_sent_queue() for drivers that are aware
3262 * that they should not test BQL status themselves.
3263 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3264 * skb of a batch.
3265 * Returns true if the doorbell must be used to kick the NIC.
3266 */
3267static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3268 unsigned int bytes,
3269 bool xmit_more)
3270{
3271 if (xmit_more) {
3272#ifdef CONFIG_BQL
3273 dql_queued(&dev_queue->dql, bytes);
3274#endif
3275 return netif_tx_queue_stopped(dev_queue);
3276 }
3277 netdev_tx_sent_queue(dev_queue, bytes);
3278 return true;
3279}
3280
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003281/**
3282 * netdev_sent_queue - report the number of bytes queued to hardware
3283 * @dev: network device
3284 * @bytes: number of bytes queued to the hardware device queue
3285 *
3286 * Report the number of bytes queued for sending/completion to the network
3287 * device hardware queue. @bytes should be a good approximation and should
3288 * exactly match netdev_completed_queue() @bytes
3289 */
3290static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3291{
3292 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3293}
3294
David Brazdil0f672f62019-12-10 10:32:29 +00003295static inline bool __netdev_sent_queue(struct net_device *dev,
3296 unsigned int bytes,
3297 bool xmit_more)
3298{
3299 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3300 xmit_more);
3301}
3302
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003303static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3304 unsigned int pkts, unsigned int bytes)
3305{
3306#ifdef CONFIG_BQL
3307 if (unlikely(!bytes))
3308 return;
3309
3310 dql_completed(&dev_queue->dql, bytes);
3311
3312 /*
3313 * Without the memory barrier there is a small possiblity that
3314 * netdev_tx_sent_queue will miss the update and cause the queue to
3315 * be stopped forever
3316 */
3317 smp_mb();
3318
David Brazdil0f672f62019-12-10 10:32:29 +00003319 if (unlikely(dql_avail(&dev_queue->dql) < 0))
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003320 return;
3321
3322 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3323 netif_schedule_queue(dev_queue);
3324#endif
3325}
3326
3327/**
3328 * netdev_completed_queue - report bytes and packets completed by device
3329 * @dev: network device
3330 * @pkts: actual number of packets sent over the medium
3331 * @bytes: actual number of bytes sent over the medium
3332 *
3333 * Report the number of bytes and packets transmitted by the network device
3334 * hardware queue over the physical medium, @bytes must exactly match the
3335 * @bytes amount passed to netdev_sent_queue()
3336 */
3337static inline void netdev_completed_queue(struct net_device *dev,
3338 unsigned int pkts, unsigned int bytes)
3339{
3340 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3341}
3342
3343static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3344{
3345#ifdef CONFIG_BQL
3346 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3347 dql_reset(&q->dql);
3348#endif
3349}
3350
3351/**
3352 * netdev_reset_queue - reset the packets and bytes count of a network device
3353 * @dev_queue: network device
3354 *
3355 * Reset the bytes and packet count of a network device and clear the
3356 * software flow control OFF bit for this network device
3357 */
3358static inline void netdev_reset_queue(struct net_device *dev_queue)
3359{
3360 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3361}
3362
3363/**
3364 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3365 * @dev: network device
3366 * @queue_index: given tx queue index
3367 *
3368 * Returns 0 if given tx queue index >= number of device tx queues,
3369 * otherwise returns the originally passed tx queue index.
3370 */
3371static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3372{
3373 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3374 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3375 dev->name, queue_index,
3376 dev->real_num_tx_queues);
3377 return 0;
3378 }
3379
3380 return queue_index;
3381}
3382
3383/**
3384 * netif_running - test if up
3385 * @dev: network device
3386 *
3387 * Test if the device has been brought up.
3388 */
3389static inline bool netif_running(const struct net_device *dev)
3390{
3391 return test_bit(__LINK_STATE_START, &dev->state);
3392}
3393
3394/*
3395 * Routines to manage the subqueues on a device. We only need start,
3396 * stop, and a check if it's stopped. All other device management is
3397 * done at the overall netdevice level.
3398 * Also test the device if we're multiqueue.
3399 */
3400
3401/**
3402 * netif_start_subqueue - allow sending packets on subqueue
3403 * @dev: network device
3404 * @queue_index: sub queue index
3405 *
3406 * Start individual transmit queue of a device with multiple transmit queues.
3407 */
3408static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3409{
3410 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3411
3412 netif_tx_start_queue(txq);
3413}
3414
3415/**
3416 * netif_stop_subqueue - stop sending packets on subqueue
3417 * @dev: network device
3418 * @queue_index: sub queue index
3419 *
3420 * Stop individual transmit queue of a device with multiple transmit queues.
3421 */
3422static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3423{
3424 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3425 netif_tx_stop_queue(txq);
3426}
3427
3428/**
3429 * netif_subqueue_stopped - test status of subqueue
3430 * @dev: network device
3431 * @queue_index: sub queue index
3432 *
3433 * Check individual transmit queue of a device with multiple transmit queues.
3434 */
3435static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3436 u16 queue_index)
3437{
3438 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3439
3440 return netif_tx_queue_stopped(txq);
3441}
3442
3443static inline bool netif_subqueue_stopped(const struct net_device *dev,
3444 struct sk_buff *skb)
3445{
3446 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3447}
3448
3449/**
3450 * netif_wake_subqueue - allow sending packets on subqueue
3451 * @dev: network device
3452 * @queue_index: sub queue index
3453 *
3454 * Resume individual transmit queue of a device with multiple transmit queues.
3455 */
3456static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3457{
3458 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3459
3460 netif_tx_wake_queue(txq);
3461}
3462
3463#ifdef CONFIG_XPS
3464int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3465 u16 index);
3466int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3467 u16 index, bool is_rxqs_map);
3468
3469/**
3470 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3471 * @j: CPU/Rx queue index
3472 * @mask: bitmask of all cpus/rx queues
3473 * @nr_bits: number of bits in the bitmask
3474 *
3475 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3476 */
3477static inline bool netif_attr_test_mask(unsigned long j,
3478 const unsigned long *mask,
3479 unsigned int nr_bits)
3480{
3481 cpu_max_bits_warn(j, nr_bits);
3482 return test_bit(j, mask);
3483}
3484
3485/**
3486 * netif_attr_test_online - Test for online CPU/Rx queue
3487 * @j: CPU/Rx queue index
3488 * @online_mask: bitmask for CPUs/Rx queues that are online
3489 * @nr_bits: number of bits in the bitmask
3490 *
3491 * Returns true if a CPU/Rx queue is online.
3492 */
3493static inline bool netif_attr_test_online(unsigned long j,
3494 const unsigned long *online_mask,
3495 unsigned int nr_bits)
3496{
3497 cpu_max_bits_warn(j, nr_bits);
3498
3499 if (online_mask)
3500 return test_bit(j, online_mask);
3501
3502 return (j < nr_bits);
3503}
3504
3505/**
3506 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3507 * @n: CPU/Rx queue index
3508 * @srcp: the cpumask/Rx queue mask pointer
3509 * @nr_bits: number of bits in the bitmask
3510 *
3511 * Returns >= nr_bits if no further CPUs/Rx queues set.
3512 */
3513static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3514 unsigned int nr_bits)
3515{
3516 /* -1 is a legal arg here. */
3517 if (n != -1)
3518 cpu_max_bits_warn(n, nr_bits);
3519
3520 if (srcp)
3521 return find_next_bit(srcp, nr_bits, n + 1);
3522
3523 return n + 1;
3524}
3525
3526/**
3527 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3528 * @n: CPU/Rx queue index
3529 * @src1p: the first CPUs/Rx queues mask pointer
3530 * @src2p: the second CPUs/Rx queues mask pointer
3531 * @nr_bits: number of bits in the bitmask
3532 *
3533 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3534 */
3535static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3536 const unsigned long *src2p,
3537 unsigned int nr_bits)
3538{
3539 /* -1 is a legal arg here. */
3540 if (n != -1)
3541 cpu_max_bits_warn(n, nr_bits);
3542
3543 if (src1p && src2p)
3544 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3545 else if (src1p)
3546 return find_next_bit(src1p, nr_bits, n + 1);
3547 else if (src2p)
3548 return find_next_bit(src2p, nr_bits, n + 1);
3549
3550 return n + 1;
3551}
3552#else
3553static inline int netif_set_xps_queue(struct net_device *dev,
3554 const struct cpumask *mask,
3555 u16 index)
3556{
3557 return 0;
3558}
3559
3560static inline int __netif_set_xps_queue(struct net_device *dev,
3561 const unsigned long *mask,
3562 u16 index, bool is_rxqs_map)
3563{
3564 return 0;
3565}
3566#endif
3567
3568/**
3569 * netif_is_multiqueue - test if device has multiple transmit queues
3570 * @dev: network device
3571 *
3572 * Check if device has multiple transmit queues
3573 */
3574static inline bool netif_is_multiqueue(const struct net_device *dev)
3575{
3576 return dev->num_tx_queues > 1;
3577}
3578
3579int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3580
3581#ifdef CONFIG_SYSFS
3582int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3583#else
3584static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3585 unsigned int rxqs)
3586{
3587 dev->real_num_rx_queues = rxqs;
3588 return 0;
3589}
3590#endif
3591
3592static inline struct netdev_rx_queue *
3593__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3594{
3595 return dev->_rx + rxq;
3596}
3597
3598#ifdef CONFIG_SYSFS
3599static inline unsigned int get_netdev_rx_queue_index(
3600 struct netdev_rx_queue *queue)
3601{
3602 struct net_device *dev = queue->dev;
3603 int index = queue - dev->_rx;
3604
3605 BUG_ON(index >= dev->num_rx_queues);
3606 return index;
3607}
3608#endif
3609
3610#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3611int netif_get_num_default_rss_queues(void);
3612
3613enum skb_free_reason {
3614 SKB_REASON_CONSUMED,
3615 SKB_REASON_DROPPED,
3616};
3617
3618void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3619void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3620
3621/*
3622 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3623 * interrupt context or with hardware interrupts being disabled.
3624 * (in_irq() || irqs_disabled())
3625 *
3626 * We provide four helpers that can be used in following contexts :
3627 *
3628 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3629 * replacing kfree_skb(skb)
3630 *
3631 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3632 * Typically used in place of consume_skb(skb) in TX completion path
3633 *
3634 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3635 * replacing kfree_skb(skb)
3636 *
3637 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3638 * and consumed a packet. Used in place of consume_skb(skb)
3639 */
3640static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3641{
3642 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3643}
3644
3645static inline void dev_consume_skb_irq(struct sk_buff *skb)
3646{
3647 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3648}
3649
3650static inline void dev_kfree_skb_any(struct sk_buff *skb)
3651{
3652 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3653}
3654
3655static inline void dev_consume_skb_any(struct sk_buff *skb)
3656{
3657 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3658}
3659
3660void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3661int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3662int netif_rx(struct sk_buff *skb);
3663int netif_rx_ni(struct sk_buff *skb);
3664int netif_receive_skb(struct sk_buff *skb);
3665int netif_receive_skb_core(struct sk_buff *skb);
3666void netif_receive_skb_list(struct list_head *head);
3667gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3668void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3669struct sk_buff *napi_get_frags(struct napi_struct *napi);
3670gro_result_t napi_gro_frags(struct napi_struct *napi);
3671struct packet_offload *gro_find_receive_by_type(__be16 type);
3672struct packet_offload *gro_find_complete_by_type(__be16 type);
3673
3674static inline void napi_free_frags(struct napi_struct *napi)
3675{
3676 kfree_skb(napi->skb);
3677 napi->skb = NULL;
3678}
3679
3680bool netdev_is_rx_handler_busy(struct net_device *dev);
3681int netdev_rx_handler_register(struct net_device *dev,
3682 rx_handler_func_t *rx_handler,
3683 void *rx_handler_data);
3684void netdev_rx_handler_unregister(struct net_device *dev);
3685
3686bool dev_valid_name(const char *name);
Olivier Deprez0e641232021-09-23 10:07:05 +02003687static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3688{
3689 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3690}
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003691int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3692 bool *need_copyout);
3693int dev_ifconf(struct net *net, struct ifconf *, int);
3694int dev_ethtool(struct net *net, struct ifreq *);
3695unsigned int dev_get_flags(const struct net_device *);
David Brazdil0f672f62019-12-10 10:32:29 +00003696int __dev_change_flags(struct net_device *dev, unsigned int flags,
3697 struct netlink_ext_ack *extack);
3698int dev_change_flags(struct net_device *dev, unsigned int flags,
3699 struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003700void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3701 unsigned int gchanges);
3702int dev_change_name(struct net_device *, const char *);
3703int dev_set_alias(struct net_device *, const char *, size_t);
3704int dev_get_alias(const struct net_device *, char *, size_t);
3705int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3706int __dev_set_mtu(struct net_device *, int);
Olivier Deprez0e641232021-09-23 10:07:05 +02003707int dev_validate_mtu(struct net_device *dev, int mtu,
3708 struct netlink_ext_ack *extack);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003709int dev_set_mtu_ext(struct net_device *dev, int mtu,
3710 struct netlink_ext_ack *extack);
3711int dev_set_mtu(struct net_device *, int);
3712int dev_change_tx_queue_len(struct net_device *, unsigned long);
3713void dev_set_group(struct net_device *, int);
David Brazdil0f672f62019-12-10 10:32:29 +00003714int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3715 struct netlink_ext_ack *extack);
3716int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3717 struct netlink_ext_ack *extack);
Olivier Deprez0e641232021-09-23 10:07:05 +02003718int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3719 struct netlink_ext_ack *extack);
3720int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003721int dev_change_carrier(struct net_device *, bool new_carrier);
3722int dev_get_phys_port_id(struct net_device *dev,
3723 struct netdev_phys_item_id *ppid);
3724int dev_get_phys_port_name(struct net_device *dev,
3725 char *name, size_t len);
David Brazdil0f672f62019-12-10 10:32:29 +00003726int dev_get_port_parent_id(struct net_device *dev,
3727 struct netdev_phys_item_id *ppid, bool recurse);
3728bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003729int dev_change_proto_down(struct net_device *dev, bool proto_down);
David Brazdil0f672f62019-12-10 10:32:29 +00003730int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003731struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3732struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3733 struct netdev_queue *txq, int *ret);
3734
3735typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3736int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3737 int fd, u32 flags);
3738u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3739 enum bpf_netdev_command cmd);
3740int xdp_umem_query(struct net_device *dev, u16 queue_id);
3741
3742int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3743int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3744bool is_skb_forwardable(const struct net_device *dev,
3745 const struct sk_buff *skb);
3746
3747static __always_inline int ____dev_forward_skb(struct net_device *dev,
3748 struct sk_buff *skb)
3749{
3750 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3751 unlikely(!is_skb_forwardable(dev, skb))) {
3752 atomic_long_inc(&dev->rx_dropped);
3753 kfree_skb(skb);
3754 return NET_RX_DROP;
3755 }
3756
3757 skb_scrub_packet(skb, true);
3758 skb->priority = 0;
3759 return 0;
3760}
3761
David Brazdil0f672f62019-12-10 10:32:29 +00003762bool dev_nit_active(struct net_device *dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003763void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3764
3765extern int netdev_budget;
3766extern unsigned int netdev_budget_usecs;
3767
3768/* Called by rtnetlink.c:rtnl_unlock() */
3769void netdev_run_todo(void);
3770
3771/**
3772 * dev_put - release reference to device
3773 * @dev: network device
3774 *
3775 * Release reference to device to allow it to be freed.
3776 */
3777static inline void dev_put(struct net_device *dev)
3778{
3779 this_cpu_dec(*dev->pcpu_refcnt);
3780}
3781
3782/**
3783 * dev_hold - get reference to device
3784 * @dev: network device
3785 *
3786 * Hold reference to device to keep it from being freed.
3787 */
3788static inline void dev_hold(struct net_device *dev)
3789{
3790 this_cpu_inc(*dev->pcpu_refcnt);
3791}
3792
3793/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3794 * and _off may be called from IRQ context, but it is caller
3795 * who is responsible for serialization of these calls.
3796 *
3797 * The name carrier is inappropriate, these functions should really be
3798 * called netif_lowerlayer_*() because they represent the state of any
3799 * kind of lower layer not just hardware media.
3800 */
3801
3802void linkwatch_init_dev(struct net_device *dev);
3803void linkwatch_fire_event(struct net_device *dev);
3804void linkwatch_forget_dev(struct net_device *dev);
3805
3806/**
3807 * netif_carrier_ok - test if carrier present
3808 * @dev: network device
3809 *
3810 * Check if carrier is present on device
3811 */
3812static inline bool netif_carrier_ok(const struct net_device *dev)
3813{
3814 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3815}
3816
3817unsigned long dev_trans_start(struct net_device *dev);
3818
3819void __netdev_watchdog_up(struct net_device *dev);
3820
3821void netif_carrier_on(struct net_device *dev);
3822
3823void netif_carrier_off(struct net_device *dev);
3824
3825/**
3826 * netif_dormant_on - mark device as dormant.
3827 * @dev: network device
3828 *
3829 * Mark device as dormant (as per RFC2863).
3830 *
3831 * The dormant state indicates that the relevant interface is not
3832 * actually in a condition to pass packets (i.e., it is not 'up') but is
3833 * in a "pending" state, waiting for some external event. For "on-
3834 * demand" interfaces, this new state identifies the situation where the
3835 * interface is waiting for events to place it in the up state.
3836 */
3837static inline void netif_dormant_on(struct net_device *dev)
3838{
3839 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3840 linkwatch_fire_event(dev);
3841}
3842
3843/**
3844 * netif_dormant_off - set device as not dormant.
3845 * @dev: network device
3846 *
3847 * Device is not in dormant state.
3848 */
3849static inline void netif_dormant_off(struct net_device *dev)
3850{
3851 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3852 linkwatch_fire_event(dev);
3853}
3854
3855/**
3856 * netif_dormant - test if device is dormant
3857 * @dev: network device
3858 *
3859 * Check if device is dormant.
3860 */
3861static inline bool netif_dormant(const struct net_device *dev)
3862{
3863 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3864}
3865
3866
3867/**
3868 * netif_oper_up - test if device is operational
3869 * @dev: network device
3870 *
3871 * Check if carrier is operational
3872 */
3873static inline bool netif_oper_up(const struct net_device *dev)
3874{
3875 return (dev->operstate == IF_OPER_UP ||
3876 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3877}
3878
3879/**
3880 * netif_device_present - is device available or removed
3881 * @dev: network device
3882 *
3883 * Check if device has not been removed from system.
3884 */
3885static inline bool netif_device_present(struct net_device *dev)
3886{
3887 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3888}
3889
3890void netif_device_detach(struct net_device *dev);
3891
3892void netif_device_attach(struct net_device *dev);
3893
3894/*
3895 * Network interface message level settings
3896 */
3897
3898enum {
3899 NETIF_MSG_DRV = 0x0001,
3900 NETIF_MSG_PROBE = 0x0002,
3901 NETIF_MSG_LINK = 0x0004,
3902 NETIF_MSG_TIMER = 0x0008,
3903 NETIF_MSG_IFDOWN = 0x0010,
3904 NETIF_MSG_IFUP = 0x0020,
3905 NETIF_MSG_RX_ERR = 0x0040,
3906 NETIF_MSG_TX_ERR = 0x0080,
3907 NETIF_MSG_TX_QUEUED = 0x0100,
3908 NETIF_MSG_INTR = 0x0200,
3909 NETIF_MSG_TX_DONE = 0x0400,
3910 NETIF_MSG_RX_STATUS = 0x0800,
3911 NETIF_MSG_PKTDATA = 0x1000,
3912 NETIF_MSG_HW = 0x2000,
3913 NETIF_MSG_WOL = 0x4000,
3914};
3915
3916#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3917#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3918#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3919#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3920#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3921#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3922#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3923#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3924#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3925#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3926#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3927#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3928#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3929#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3930#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3931
3932static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3933{
3934 /* use default */
3935 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3936 return default_msg_enable_bits;
3937 if (debug_value == 0) /* no output */
3938 return 0;
3939 /* set low N bits */
David Brazdil0f672f62019-12-10 10:32:29 +00003940 return (1U << debug_value) - 1;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003941}
3942
3943static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3944{
3945 spin_lock(&txq->_xmit_lock);
3946 txq->xmit_lock_owner = cpu;
3947}
3948
3949static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3950{
3951 __acquire(&txq->_xmit_lock);
3952 return true;
3953}
3954
3955static inline void __netif_tx_release(struct netdev_queue *txq)
3956{
3957 __release(&txq->_xmit_lock);
3958}
3959
3960static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3961{
3962 spin_lock_bh(&txq->_xmit_lock);
3963 txq->xmit_lock_owner = smp_processor_id();
3964}
3965
3966static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3967{
3968 bool ok = spin_trylock(&txq->_xmit_lock);
3969 if (likely(ok))
3970 txq->xmit_lock_owner = smp_processor_id();
3971 return ok;
3972}
3973
3974static inline void __netif_tx_unlock(struct netdev_queue *txq)
3975{
3976 txq->xmit_lock_owner = -1;
3977 spin_unlock(&txq->_xmit_lock);
3978}
3979
3980static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3981{
3982 txq->xmit_lock_owner = -1;
3983 spin_unlock_bh(&txq->_xmit_lock);
3984}
3985
3986static inline void txq_trans_update(struct netdev_queue *txq)
3987{
3988 if (txq->xmit_lock_owner != -1)
3989 txq->trans_start = jiffies;
3990}
3991
3992/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3993static inline void netif_trans_update(struct net_device *dev)
3994{
3995 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3996
3997 if (txq->trans_start != jiffies)
3998 txq->trans_start = jiffies;
3999}
4000
4001/**
4002 * netif_tx_lock - grab network device transmit lock
4003 * @dev: network device
4004 *
4005 * Get network device transmit lock
4006 */
4007static inline void netif_tx_lock(struct net_device *dev)
4008{
4009 unsigned int i;
4010 int cpu;
4011
4012 spin_lock(&dev->tx_global_lock);
4013 cpu = smp_processor_id();
4014 for (i = 0; i < dev->num_tx_queues; i++) {
4015 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4016
4017 /* We are the only thread of execution doing a
4018 * freeze, but we have to grab the _xmit_lock in
4019 * order to synchronize with threads which are in
4020 * the ->hard_start_xmit() handler and already
4021 * checked the frozen bit.
4022 */
4023 __netif_tx_lock(txq, cpu);
4024 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4025 __netif_tx_unlock(txq);
4026 }
4027}
4028
4029static inline void netif_tx_lock_bh(struct net_device *dev)
4030{
4031 local_bh_disable();
4032 netif_tx_lock(dev);
4033}
4034
4035static inline void netif_tx_unlock(struct net_device *dev)
4036{
4037 unsigned int i;
4038
4039 for (i = 0; i < dev->num_tx_queues; i++) {
4040 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4041
4042 /* No need to grab the _xmit_lock here. If the
4043 * queue is not stopped for another reason, we
4044 * force a schedule.
4045 */
4046 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4047 netif_schedule_queue(txq);
4048 }
4049 spin_unlock(&dev->tx_global_lock);
4050}
4051
4052static inline void netif_tx_unlock_bh(struct net_device *dev)
4053{
4054 netif_tx_unlock(dev);
4055 local_bh_enable();
4056}
4057
4058#define HARD_TX_LOCK(dev, txq, cpu) { \
4059 if ((dev->features & NETIF_F_LLTX) == 0) { \
4060 __netif_tx_lock(txq, cpu); \
4061 } else { \
4062 __netif_tx_acquire(txq); \
4063 } \
4064}
4065
4066#define HARD_TX_TRYLOCK(dev, txq) \
4067 (((dev->features & NETIF_F_LLTX) == 0) ? \
4068 __netif_tx_trylock(txq) : \
4069 __netif_tx_acquire(txq))
4070
4071#define HARD_TX_UNLOCK(dev, txq) { \
4072 if ((dev->features & NETIF_F_LLTX) == 0) { \
4073 __netif_tx_unlock(txq); \
4074 } else { \
4075 __netif_tx_release(txq); \
4076 } \
4077}
4078
4079static inline void netif_tx_disable(struct net_device *dev)
4080{
4081 unsigned int i;
4082 int cpu;
4083
4084 local_bh_disable();
4085 cpu = smp_processor_id();
Olivier Deprez0e641232021-09-23 10:07:05 +02004086 spin_lock(&dev->tx_global_lock);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004087 for (i = 0; i < dev->num_tx_queues; i++) {
4088 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4089
4090 __netif_tx_lock(txq, cpu);
4091 netif_tx_stop_queue(txq);
4092 __netif_tx_unlock(txq);
4093 }
Olivier Deprez0e641232021-09-23 10:07:05 +02004094 spin_unlock(&dev->tx_global_lock);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004095 local_bh_enable();
4096}
4097
4098static inline void netif_addr_lock(struct net_device *dev)
4099{
4100 spin_lock(&dev->addr_list_lock);
4101}
4102
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004103static inline void netif_addr_lock_bh(struct net_device *dev)
4104{
4105 spin_lock_bh(&dev->addr_list_lock);
4106}
4107
4108static inline void netif_addr_unlock(struct net_device *dev)
4109{
4110 spin_unlock(&dev->addr_list_lock);
4111}
4112
4113static inline void netif_addr_unlock_bh(struct net_device *dev)
4114{
4115 spin_unlock_bh(&dev->addr_list_lock);
4116}
4117
4118/*
4119 * dev_addrs walker. Should be used only for read access. Call with
4120 * rcu_read_lock held.
4121 */
4122#define for_each_dev_addr(dev, ha) \
4123 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4124
4125/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4126
4127void ether_setup(struct net_device *dev);
4128
4129/* Support for loadable net-drivers */
4130struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4131 unsigned char name_assign_type,
4132 void (*setup)(struct net_device *),
4133 unsigned int txqs, unsigned int rxqs);
4134int dev_get_valid_name(struct net *net, struct net_device *dev,
4135 const char *name);
4136
4137#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4138 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4139
4140#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4141 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4142 count)
4143
4144int register_netdev(struct net_device *dev);
4145void unregister_netdev(struct net_device *dev);
4146
4147/* General hardware address lists handling functions */
4148int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4149 struct netdev_hw_addr_list *from_list, int addr_len);
4150void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4151 struct netdev_hw_addr_list *from_list, int addr_len);
4152int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4153 struct net_device *dev,
4154 int (*sync)(struct net_device *, const unsigned char *),
4155 int (*unsync)(struct net_device *,
4156 const unsigned char *));
David Brazdil0f672f62019-12-10 10:32:29 +00004157int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4158 struct net_device *dev,
4159 int (*sync)(struct net_device *,
4160 const unsigned char *, int),
4161 int (*unsync)(struct net_device *,
4162 const unsigned char *, int));
4163void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4164 struct net_device *dev,
4165 int (*unsync)(struct net_device *,
4166 const unsigned char *, int));
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004167void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4168 struct net_device *dev,
4169 int (*unsync)(struct net_device *,
4170 const unsigned char *));
4171void __hw_addr_init(struct netdev_hw_addr_list *list);
4172
4173/* Functions used for device addresses handling */
4174int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4175 unsigned char addr_type);
4176int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4177 unsigned char addr_type);
4178void dev_addr_flush(struct net_device *dev);
4179int dev_addr_init(struct net_device *dev);
4180
4181/* Functions used for unicast addresses handling */
4182int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4183int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4184int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4185int dev_uc_sync(struct net_device *to, struct net_device *from);
4186int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4187void dev_uc_unsync(struct net_device *to, struct net_device *from);
4188void dev_uc_flush(struct net_device *dev);
4189void dev_uc_init(struct net_device *dev);
4190
4191/**
4192 * __dev_uc_sync - Synchonize device's unicast list
4193 * @dev: device to sync
4194 * @sync: function to call if address should be added
4195 * @unsync: function to call if address should be removed
4196 *
4197 * Add newly added addresses to the interface, and release
4198 * addresses that have been deleted.
4199 */
4200static inline int __dev_uc_sync(struct net_device *dev,
4201 int (*sync)(struct net_device *,
4202 const unsigned char *),
4203 int (*unsync)(struct net_device *,
4204 const unsigned char *))
4205{
4206 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4207}
4208
4209/**
4210 * __dev_uc_unsync - Remove synchronized addresses from device
4211 * @dev: device to sync
4212 * @unsync: function to call if address should be removed
4213 *
4214 * Remove all addresses that were added to the device by dev_uc_sync().
4215 */
4216static inline void __dev_uc_unsync(struct net_device *dev,
4217 int (*unsync)(struct net_device *,
4218 const unsigned char *))
4219{
4220 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4221}
4222
4223/* Functions used for multicast addresses handling */
4224int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4225int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4226int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4227int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4228int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4229int dev_mc_sync(struct net_device *to, struct net_device *from);
4230int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4231void dev_mc_unsync(struct net_device *to, struct net_device *from);
4232void dev_mc_flush(struct net_device *dev);
4233void dev_mc_init(struct net_device *dev);
4234
4235/**
4236 * __dev_mc_sync - Synchonize device's multicast list
4237 * @dev: device to sync
4238 * @sync: function to call if address should be added
4239 * @unsync: function to call if address should be removed
4240 *
4241 * Add newly added addresses to the interface, and release
4242 * addresses that have been deleted.
4243 */
4244static inline int __dev_mc_sync(struct net_device *dev,
4245 int (*sync)(struct net_device *,
4246 const unsigned char *),
4247 int (*unsync)(struct net_device *,
4248 const unsigned char *))
4249{
4250 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4251}
4252
4253/**
4254 * __dev_mc_unsync - Remove synchronized addresses from device
4255 * @dev: device to sync
4256 * @unsync: function to call if address should be removed
4257 *
4258 * Remove all addresses that were added to the device by dev_mc_sync().
4259 */
4260static inline void __dev_mc_unsync(struct net_device *dev,
4261 int (*unsync)(struct net_device *,
4262 const unsigned char *))
4263{
4264 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4265}
4266
4267/* Functions used for secondary unicast and multicast support */
4268void dev_set_rx_mode(struct net_device *dev);
4269void __dev_set_rx_mode(struct net_device *dev);
4270int dev_set_promiscuity(struct net_device *dev, int inc);
4271int dev_set_allmulti(struct net_device *dev, int inc);
4272void netdev_state_change(struct net_device *dev);
4273void netdev_notify_peers(struct net_device *dev);
4274void netdev_features_change(struct net_device *dev);
4275/* Load a device via the kmod */
4276void dev_load(struct net *net, const char *name);
4277struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4278 struct rtnl_link_stats64 *storage);
4279void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4280 const struct net_device_stats *netdev_stats);
4281
4282extern int netdev_max_backlog;
4283extern int netdev_tstamp_prequeue;
4284extern int weight_p;
4285extern int dev_weight_rx_bias;
4286extern int dev_weight_tx_bias;
4287extern int dev_rx_weight;
4288extern int dev_tx_weight;
David Brazdil0f672f62019-12-10 10:32:29 +00004289extern int gro_normal_batch;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004290
4291bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4292struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4293 struct list_head **iter);
4294struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4295 struct list_head **iter);
4296
4297/* iterate through upper list, must be called under RCU read lock */
4298#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4299 for (iter = &(dev)->adj_list.upper, \
4300 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4301 updev; \
4302 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4303
4304int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4305 int (*fn)(struct net_device *upper_dev,
4306 void *data),
4307 void *data);
4308
4309bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4310 struct net_device *upper_dev);
4311
4312bool netdev_has_any_upper_dev(struct net_device *dev);
4313
4314void *netdev_lower_get_next_private(struct net_device *dev,
4315 struct list_head **iter);
4316void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4317 struct list_head **iter);
4318
4319#define netdev_for_each_lower_private(dev, priv, iter) \
4320 for (iter = (dev)->adj_list.lower.next, \
4321 priv = netdev_lower_get_next_private(dev, &(iter)); \
4322 priv; \
4323 priv = netdev_lower_get_next_private(dev, &(iter)))
4324
4325#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4326 for (iter = &(dev)->adj_list.lower, \
4327 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4328 priv; \
4329 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4330
4331void *netdev_lower_get_next(struct net_device *dev,
4332 struct list_head **iter);
4333
4334#define netdev_for_each_lower_dev(dev, ldev, iter) \
4335 for (iter = (dev)->adj_list.lower.next, \
4336 ldev = netdev_lower_get_next(dev, &(iter)); \
4337 ldev; \
4338 ldev = netdev_lower_get_next(dev, &(iter)))
4339
Olivier Deprez0e641232021-09-23 10:07:05 +02004340struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004341 struct list_head **iter);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004342int netdev_walk_all_lower_dev(struct net_device *dev,
4343 int (*fn)(struct net_device *lower_dev,
4344 void *data),
4345 void *data);
4346int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4347 int (*fn)(struct net_device *lower_dev,
4348 void *data),
4349 void *data);
4350
4351void *netdev_adjacent_get_private(struct list_head *adj_list);
4352void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4353struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4354struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4355int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4356 struct netlink_ext_ack *extack);
4357int netdev_master_upper_dev_link(struct net_device *dev,
4358 struct net_device *upper_dev,
4359 void *upper_priv, void *upper_info,
4360 struct netlink_ext_ack *extack);
4361void netdev_upper_dev_unlink(struct net_device *dev,
4362 struct net_device *upper_dev);
David Brazdil0f672f62019-12-10 10:32:29 +00004363int netdev_adjacent_change_prepare(struct net_device *old_dev,
4364 struct net_device *new_dev,
4365 struct net_device *dev,
4366 struct netlink_ext_ack *extack);
4367void netdev_adjacent_change_commit(struct net_device *old_dev,
4368 struct net_device *new_dev,
4369 struct net_device *dev);
4370void netdev_adjacent_change_abort(struct net_device *old_dev,
4371 struct net_device *new_dev,
4372 struct net_device *dev);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004373void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4374void *netdev_lower_dev_get_private(struct net_device *dev,
4375 struct net_device *lower_dev);
4376void netdev_lower_state_changed(struct net_device *lower_dev,
4377 void *lower_state_info);
4378
4379/* RSS keys are 40 or 52 bytes long */
4380#define NETDEV_RSS_KEY_LEN 52
4381extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4382void netdev_rss_key_fill(void *buffer, size_t len);
4383
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004384int skb_checksum_help(struct sk_buff *skb);
4385int skb_crc32c_csum_help(struct sk_buff *skb);
4386int skb_csum_hwoffload_help(struct sk_buff *skb,
4387 const netdev_features_t features);
4388
4389struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4390 netdev_features_t features, bool tx_path);
4391struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4392 netdev_features_t features);
4393
4394struct netdev_bonding_info {
4395 ifslave slave;
4396 ifbond master;
4397};
4398
4399struct netdev_notifier_bonding_info {
4400 struct netdev_notifier_info info; /* must be first */
4401 struct netdev_bonding_info bonding_info;
4402};
4403
4404void netdev_bonding_info_change(struct net_device *dev,
4405 struct netdev_bonding_info *bonding_info);
4406
4407static inline
4408struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4409{
4410 return __skb_gso_segment(skb, features, true);
4411}
4412__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4413
4414static inline bool can_checksum_protocol(netdev_features_t features,
4415 __be16 protocol)
4416{
4417 if (protocol == htons(ETH_P_FCOE))
4418 return !!(features & NETIF_F_FCOE_CRC);
4419
4420 /* Assume this is an IP checksum (not SCTP CRC) */
4421
4422 if (features & NETIF_F_HW_CSUM) {
4423 /* Can checksum everything */
4424 return true;
4425 }
4426
4427 switch (protocol) {
4428 case htons(ETH_P_IP):
4429 return !!(features & NETIF_F_IP_CSUM);
4430 case htons(ETH_P_IPV6):
4431 return !!(features & NETIF_F_IPV6_CSUM);
4432 default:
4433 return false;
4434 }
4435}
4436
4437#ifdef CONFIG_BUG
David Brazdil0f672f62019-12-10 10:32:29 +00004438void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004439#else
David Brazdil0f672f62019-12-10 10:32:29 +00004440static inline void netdev_rx_csum_fault(struct net_device *dev,
4441 struct sk_buff *skb)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004442{
4443}
4444#endif
4445/* rx skb timestamps */
4446void net_enable_timestamp(void);
4447void net_disable_timestamp(void);
4448
4449#ifdef CONFIG_PROC_FS
4450int __init dev_proc_init(void);
4451#else
4452#define dev_proc_init() 0
4453#endif
4454
4455static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4456 struct sk_buff *skb, struct net_device *dev,
4457 bool more)
4458{
David Brazdil0f672f62019-12-10 10:32:29 +00004459 __this_cpu_write(softnet_data.xmit.more, more);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004460 return ops->ndo_start_xmit(skb, dev);
4461}
4462
David Brazdil0f672f62019-12-10 10:32:29 +00004463static inline bool netdev_xmit_more(void)
4464{
4465 return __this_cpu_read(softnet_data.xmit.more);
4466}
4467
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004468static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4469 struct netdev_queue *txq, bool more)
4470{
4471 const struct net_device_ops *ops = dev->netdev_ops;
David Brazdil0f672f62019-12-10 10:32:29 +00004472 netdev_tx_t rc;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004473
4474 rc = __netdev_start_xmit(ops, skb, dev, more);
4475 if (rc == NETDEV_TX_OK)
4476 txq_trans_update(txq);
4477
4478 return rc;
4479}
4480
4481int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4482 const void *ns);
4483void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4484 const void *ns);
4485
4486static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4487{
4488 return netdev_class_create_file_ns(class_attr, NULL);
4489}
4490
4491static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4492{
4493 netdev_class_remove_file_ns(class_attr, NULL);
4494}
4495
4496extern const struct kobj_ns_type_operations net_ns_type_operations;
4497
4498const char *netdev_drivername(const struct net_device *dev);
4499
4500void linkwatch_run_queue(void);
4501
4502static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4503 netdev_features_t f2)
4504{
4505 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4506 if (f1 & NETIF_F_HW_CSUM)
4507 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4508 else
4509 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4510 }
4511
4512 return f1 & f2;
4513}
4514
4515static inline netdev_features_t netdev_get_wanted_features(
4516 struct net_device *dev)
4517{
4518 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4519}
4520netdev_features_t netdev_increment_features(netdev_features_t all,
4521 netdev_features_t one, netdev_features_t mask);
4522
4523/* Allow TSO being used on stacked device :
4524 * Performing the GSO segmentation before last device
4525 * is a performance improvement.
4526 */
4527static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4528 netdev_features_t mask)
4529{
4530 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4531}
4532
4533int __netdev_update_features(struct net_device *dev);
4534void netdev_update_features(struct net_device *dev);
4535void netdev_change_features(struct net_device *dev);
4536
4537void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4538 struct net_device *dev);
4539
4540netdev_features_t passthru_features_check(struct sk_buff *skb,
4541 struct net_device *dev,
4542 netdev_features_t features);
4543netdev_features_t netif_skb_features(struct sk_buff *skb);
4544
4545static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4546{
4547 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4548
4549 /* check flags correspondence */
4550 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4551 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4552 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4553 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4554 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4555 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4556 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4557 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4558 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4559 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4560 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4561 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4562 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4563 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4564 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4565 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4566 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4567 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4568
4569 return (features & feature) == feature;
4570}
4571
4572static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4573{
4574 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4575 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4576}
4577
4578static inline bool netif_needs_gso(struct sk_buff *skb,
4579 netdev_features_t features)
4580{
4581 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4582 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4583 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4584}
4585
4586static inline void netif_set_gso_max_size(struct net_device *dev,
4587 unsigned int size)
4588{
4589 dev->gso_max_size = size;
4590}
4591
4592static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4593 int pulled_hlen, u16 mac_offset,
4594 int mac_len)
4595{
4596 skb->protocol = protocol;
4597 skb->encapsulation = 1;
4598 skb_push(skb, pulled_hlen);
4599 skb_reset_transport_header(skb);
4600 skb->mac_header = mac_offset;
4601 skb->network_header = skb->mac_header + mac_len;
4602 skb->mac_len = mac_len;
4603}
4604
4605static inline bool netif_is_macsec(const struct net_device *dev)
4606{
4607 return dev->priv_flags & IFF_MACSEC;
4608}
4609
4610static inline bool netif_is_macvlan(const struct net_device *dev)
4611{
4612 return dev->priv_flags & IFF_MACVLAN;
4613}
4614
4615static inline bool netif_is_macvlan_port(const struct net_device *dev)
4616{
4617 return dev->priv_flags & IFF_MACVLAN_PORT;
4618}
4619
4620static inline bool netif_is_bond_master(const struct net_device *dev)
4621{
4622 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4623}
4624
4625static inline bool netif_is_bond_slave(const struct net_device *dev)
4626{
4627 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4628}
4629
4630static inline bool netif_supports_nofcs(struct net_device *dev)
4631{
4632 return dev->priv_flags & IFF_SUPP_NOFCS;
4633}
4634
David Brazdil0f672f62019-12-10 10:32:29 +00004635static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4636{
4637 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4638}
4639
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004640static inline bool netif_is_l3_master(const struct net_device *dev)
4641{
4642 return dev->priv_flags & IFF_L3MDEV_MASTER;
4643}
4644
4645static inline bool netif_is_l3_slave(const struct net_device *dev)
4646{
4647 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4648}
4649
4650static inline bool netif_is_bridge_master(const struct net_device *dev)
4651{
4652 return dev->priv_flags & IFF_EBRIDGE;
4653}
4654
4655static inline bool netif_is_bridge_port(const struct net_device *dev)
4656{
4657 return dev->priv_flags & IFF_BRIDGE_PORT;
4658}
4659
4660static inline bool netif_is_ovs_master(const struct net_device *dev)
4661{
4662 return dev->priv_flags & IFF_OPENVSWITCH;
4663}
4664
4665static inline bool netif_is_ovs_port(const struct net_device *dev)
4666{
4667 return dev->priv_flags & IFF_OVS_DATAPATH;
4668}
4669
4670static inline bool netif_is_team_master(const struct net_device *dev)
4671{
4672 return dev->priv_flags & IFF_TEAM;
4673}
4674
4675static inline bool netif_is_team_port(const struct net_device *dev)
4676{
4677 return dev->priv_flags & IFF_TEAM_PORT;
4678}
4679
4680static inline bool netif_is_lag_master(const struct net_device *dev)
4681{
4682 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4683}
4684
4685static inline bool netif_is_lag_port(const struct net_device *dev)
4686{
4687 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4688}
4689
4690static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4691{
4692 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4693}
4694
4695static inline bool netif_is_failover(const struct net_device *dev)
4696{
4697 return dev->priv_flags & IFF_FAILOVER;
4698}
4699
4700static inline bool netif_is_failover_slave(const struct net_device *dev)
4701{
4702 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4703}
4704
4705/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4706static inline void netif_keep_dst(struct net_device *dev)
4707{
4708 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4709}
4710
4711/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4712static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4713{
4714 /* TODO: reserve and use an additional IFF bit, if we get more users */
4715 return dev->priv_flags & IFF_MACSEC;
4716}
4717
4718extern struct pernet_operations __net_initdata loopback_net_ops;
4719
4720/* Logging, debugging and troubleshooting/diagnostic helpers. */
4721
4722/* netdev_printk helpers, similar to dev_printk */
4723
4724static inline const char *netdev_name(const struct net_device *dev)
4725{
4726 if (!dev->name[0] || strchr(dev->name, '%'))
4727 return "(unnamed net_device)";
4728 return dev->name;
4729}
4730
4731static inline bool netdev_unregistering(const struct net_device *dev)
4732{
4733 return dev->reg_state == NETREG_UNREGISTERING;
4734}
4735
4736static inline const char *netdev_reg_state(const struct net_device *dev)
4737{
4738 switch (dev->reg_state) {
4739 case NETREG_UNINITIALIZED: return " (uninitialized)";
4740 case NETREG_REGISTERED: return "";
4741 case NETREG_UNREGISTERING: return " (unregistering)";
4742 case NETREG_UNREGISTERED: return " (unregistered)";
4743 case NETREG_RELEASED: return " (released)";
4744 case NETREG_DUMMY: return " (dummy)";
4745 }
4746
4747 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4748 return " (unknown)";
4749}
4750
David Brazdil0f672f62019-12-10 10:32:29 +00004751__printf(3, 4) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004752void netdev_printk(const char *level, const struct net_device *dev,
4753 const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004754__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004755void netdev_emerg(const struct net_device *dev, const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004756__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004757void netdev_alert(const struct net_device *dev, const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004758__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004759void netdev_crit(const struct net_device *dev, const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004760__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004761void netdev_err(const struct net_device *dev, const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004762__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004763void netdev_warn(const struct net_device *dev, const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004764__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004765void netdev_notice(const struct net_device *dev, const char *format, ...);
David Brazdil0f672f62019-12-10 10:32:29 +00004766__printf(2, 3) __cold
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004767void netdev_info(const struct net_device *dev, const char *format, ...);
4768
4769#define netdev_level_once(level, dev, fmt, ...) \
4770do { \
4771 static bool __print_once __read_mostly; \
4772 \
4773 if (!__print_once) { \
4774 __print_once = true; \
4775 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4776 } \
4777} while (0)
4778
4779#define netdev_emerg_once(dev, fmt, ...) \
4780 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4781#define netdev_alert_once(dev, fmt, ...) \
4782 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4783#define netdev_crit_once(dev, fmt, ...) \
4784 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4785#define netdev_err_once(dev, fmt, ...) \
4786 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4787#define netdev_warn_once(dev, fmt, ...) \
4788 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4789#define netdev_notice_once(dev, fmt, ...) \
4790 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4791#define netdev_info_once(dev, fmt, ...) \
4792 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4793
4794#define MODULE_ALIAS_NETDEV(device) \
4795 MODULE_ALIAS("netdev-" device)
4796
4797#if defined(CONFIG_DYNAMIC_DEBUG)
4798#define netdev_dbg(__dev, format, args...) \
4799do { \
4800 dynamic_netdev_dbg(__dev, format, ##args); \
4801} while (0)
4802#elif defined(DEBUG)
4803#define netdev_dbg(__dev, format, args...) \
4804 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4805#else
4806#define netdev_dbg(__dev, format, args...) \
4807({ \
4808 if (0) \
4809 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4810})
4811#endif
4812
4813#if defined(VERBOSE_DEBUG)
4814#define netdev_vdbg netdev_dbg
4815#else
4816
4817#define netdev_vdbg(dev, format, args...) \
4818({ \
4819 if (0) \
4820 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4821 0; \
4822})
4823#endif
4824
4825/*
4826 * netdev_WARN() acts like dev_printk(), but with the key difference
4827 * of using a WARN/WARN_ON to get the message out, including the
4828 * file/line information and a backtrace.
4829 */
4830#define netdev_WARN(dev, format, args...) \
4831 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4832 netdev_reg_state(dev), ##args)
4833
4834#define netdev_WARN_ONCE(dev, format, args...) \
4835 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4836 netdev_reg_state(dev), ##args)
4837
4838/* netif printk helpers, similar to netdev_printk */
4839
4840#define netif_printk(priv, type, level, dev, fmt, args...) \
4841do { \
4842 if (netif_msg_##type(priv)) \
4843 netdev_printk(level, (dev), fmt, ##args); \
4844} while (0)
4845
4846#define netif_level(level, priv, type, dev, fmt, args...) \
4847do { \
4848 if (netif_msg_##type(priv)) \
4849 netdev_##level(dev, fmt, ##args); \
4850} while (0)
4851
4852#define netif_emerg(priv, type, dev, fmt, args...) \
4853 netif_level(emerg, priv, type, dev, fmt, ##args)
4854#define netif_alert(priv, type, dev, fmt, args...) \
4855 netif_level(alert, priv, type, dev, fmt, ##args)
4856#define netif_crit(priv, type, dev, fmt, args...) \
4857 netif_level(crit, priv, type, dev, fmt, ##args)
4858#define netif_err(priv, type, dev, fmt, args...) \
4859 netif_level(err, priv, type, dev, fmt, ##args)
4860#define netif_warn(priv, type, dev, fmt, args...) \
4861 netif_level(warn, priv, type, dev, fmt, ##args)
4862#define netif_notice(priv, type, dev, fmt, args...) \
4863 netif_level(notice, priv, type, dev, fmt, ##args)
4864#define netif_info(priv, type, dev, fmt, args...) \
4865 netif_level(info, priv, type, dev, fmt, ##args)
4866
4867#if defined(CONFIG_DYNAMIC_DEBUG)
4868#define netif_dbg(priv, type, netdev, format, args...) \
4869do { \
4870 if (netif_msg_##type(priv)) \
4871 dynamic_netdev_dbg(netdev, format, ##args); \
4872} while (0)
4873#elif defined(DEBUG)
4874#define netif_dbg(priv, type, dev, format, args...) \
4875 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4876#else
4877#define netif_dbg(priv, type, dev, format, args...) \
4878({ \
4879 if (0) \
4880 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4881 0; \
4882})
4883#endif
4884
4885/* if @cond then downgrade to debug, else print at @level */
4886#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4887 do { \
4888 if (cond) \
4889 netif_dbg(priv, type, netdev, fmt, ##args); \
4890 else \
4891 netif_ ## level(priv, type, netdev, fmt, ##args); \
4892 } while (0)
4893
4894#if defined(VERBOSE_DEBUG)
4895#define netif_vdbg netif_dbg
4896#else
4897#define netif_vdbg(priv, type, dev, format, args...) \
4898({ \
4899 if (0) \
4900 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4901 0; \
4902})
4903#endif
4904
4905/*
4906 * The list of packet types we will receive (as opposed to discard)
4907 * and the routines to invoke.
4908 *
4909 * Why 16. Because with 16 the only overlap we get on a hash of the
4910 * low nibble of the protocol value is RARP/SNAP/X.25.
4911 *
4912 * 0800 IP
4913 * 0001 802.3
4914 * 0002 AX.25
4915 * 0004 802.2
4916 * 8035 RARP
4917 * 0005 SNAP
4918 * 0805 X.25
4919 * 0806 ARP
4920 * 8137 IPX
4921 * 0009 Localtalk
4922 * 86DD IPv6
4923 */
4924#define PTYPE_HASH_SIZE (16)
4925#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4926
David Brazdil0f672f62019-12-10 10:32:29 +00004927extern struct net_device *blackhole_netdev;
4928
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004929#endif /* _LINUX_NETDEVICE_H */