1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
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>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 #include <asm/local.h>
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
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 <uapi/linux/netdev.h>
51 #include <linux/hashtable.h>
52 #include <linux/rbtree.h>
53 #include <net/net_trackers.h>
54 #include <net/net_debug.h>
61 struct ip_tunnel_parm;
62 struct macsec_context;
64 struct netdev_name_node;
69 /* 802.15.4 specific */
72 /* UDP Tunnel offloads */
73 struct udp_tunnel_info;
74 struct udp_tunnel_nic_info;
75 struct udp_tunnel_nic;
80 void synchronize_net(void);
81 void netdev_set_default_ethtool_ops(struct net_device *dev,
82 const struct ethtool_ops *ops);
83 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
85 /* Backlog congestion levels */
86 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
87 #define NET_RX_DROP 1 /* packet dropped */
89 #define MAX_NEST_DEV 8
92 * Transmit return codes: transmit return codes originate from three different
95 * - qdisc return codes
96 * - driver transmit return codes
99 * Drivers are allowed to return any one of those in their hard_start_xmit()
100 * function. Real network devices commonly used with qdiscs should only return
101 * the driver transmit return codes though - when qdiscs are used, the actual
102 * transmission happens asynchronously, so the value is not propagated to
103 * higher layers. Virtual network devices transmit synchronously; in this case
104 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
105 * others are propagated to higher layers.
108 /* qdisc ->enqueue() return codes. */
109 #define NET_XMIT_SUCCESS 0x00
110 #define NET_XMIT_DROP 0x01 /* skb dropped */
111 #define NET_XMIT_CN 0x02 /* congestion notification */
112 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
114 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
115 * indicates that the device will soon be dropping packets, or already drops
116 * some packets of the same priority; prompting us to send less aggressively. */
117 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
118 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
120 /* Driver transmit return codes */
121 #define NETDEV_TX_MASK 0xf0
124 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
125 NETDEV_TX_OK = 0x00, /* driver took care of packet */
126 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
128 typedef enum netdev_tx netdev_tx_t;
131 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
132 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
134 static inline bool dev_xmit_complete(int rc)
137 * Positive cases with an skb consumed by a driver:
138 * - successful transmission (rc == NETDEV_TX_OK)
139 * - error while transmitting (rc < 0)
140 * - error while queueing to a different device (rc & NET_XMIT_MASK)
142 if (likely(rc < NET_XMIT_MASK))
149 * Compute the worst-case header length according to the protocols
153 #if defined(CONFIG_HYPERV_NET)
154 # define LL_MAX_HEADER 128
155 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
156 # if defined(CONFIG_MAC80211_MESH)
157 # define LL_MAX_HEADER 128
159 # define LL_MAX_HEADER 96
162 # define LL_MAX_HEADER 32
165 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
166 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
167 #define MAX_HEADER LL_MAX_HEADER
169 #define MAX_HEADER (LL_MAX_HEADER + 48)
173 * Old network device statistics. Fields are native words
174 * (unsigned long) so they can be read and written atomically.
177 #define NET_DEV_STAT(FIELD) \
179 unsigned long FIELD; \
180 atomic_long_t __##FIELD; \
183 struct net_device_stats {
184 NET_DEV_STAT(rx_packets);
185 NET_DEV_STAT(tx_packets);
186 NET_DEV_STAT(rx_bytes);
187 NET_DEV_STAT(tx_bytes);
188 NET_DEV_STAT(rx_errors);
189 NET_DEV_STAT(tx_errors);
190 NET_DEV_STAT(rx_dropped);
191 NET_DEV_STAT(tx_dropped);
192 NET_DEV_STAT(multicast);
193 NET_DEV_STAT(collisions);
194 NET_DEV_STAT(rx_length_errors);
195 NET_DEV_STAT(rx_over_errors);
196 NET_DEV_STAT(rx_crc_errors);
197 NET_DEV_STAT(rx_frame_errors);
198 NET_DEV_STAT(rx_fifo_errors);
199 NET_DEV_STAT(rx_missed_errors);
200 NET_DEV_STAT(tx_aborted_errors);
201 NET_DEV_STAT(tx_carrier_errors);
202 NET_DEV_STAT(tx_fifo_errors);
203 NET_DEV_STAT(tx_heartbeat_errors);
204 NET_DEV_STAT(tx_window_errors);
205 NET_DEV_STAT(rx_compressed);
206 NET_DEV_STAT(tx_compressed);
210 /* per-cpu stats, allocated on demand.
211 * Try to fit them in a single cache line, for dev_get_stats() sake.
213 struct net_device_core_stats {
214 unsigned long rx_dropped;
215 unsigned long tx_dropped;
216 unsigned long rx_nohandler;
217 unsigned long rx_otherhost_dropped;
218 } __aligned(4 * sizeof(unsigned long));
220 #include <linux/cache.h>
221 #include <linux/skbuff.h>
224 #include <linux/static_key.h>
225 extern struct static_key_false rps_needed;
226 extern struct static_key_false rfs_needed;
233 struct netdev_hw_addr {
234 struct list_head list;
236 unsigned char addr[MAX_ADDR_LEN];
238 #define NETDEV_HW_ADDR_T_LAN 1
239 #define NETDEV_HW_ADDR_T_SAN 2
240 #define NETDEV_HW_ADDR_T_UNICAST 3
241 #define NETDEV_HW_ADDR_T_MULTICAST 4
246 struct rcu_head rcu_head;
249 struct netdev_hw_addr_list {
250 struct list_head list;
253 /* Auxiliary tree for faster lookup on addition and deletion */
257 #define netdev_hw_addr_list_count(l) ((l)->count)
258 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
259 #define netdev_hw_addr_list_for_each(ha, l) \
260 list_for_each_entry(ha, &(l)->list, list)
262 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
263 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
264 #define netdev_for_each_uc_addr(ha, dev) \
265 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
266 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
267 netdev_for_each_uc_addr((_ha), (_dev)) \
270 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
271 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
272 #define netdev_for_each_mc_addr(ha, dev) \
273 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
274 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
275 netdev_for_each_mc_addr((_ha), (_dev)) \
282 /* cached hardware header; allow for machine alignment needs. */
283 #define HH_DATA_MOD 16
284 #define HH_DATA_OFF(__len) \
285 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
286 #define HH_DATA_ALIGN(__len) \
287 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
288 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
291 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
293 * dev->hard_header_len ? (dev->hard_header_len +
294 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
296 * We could use other alignment values, but we must maintain the
297 * relationship HH alignment <= LL alignment.
299 #define LL_RESERVED_SPACE(dev) \
300 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
301 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
302 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
305 int (*create) (struct sk_buff *skb, struct net_device *dev,
306 unsigned short type, const void *daddr,
307 const void *saddr, unsigned int len);
308 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
309 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
310 void (*cache_update)(struct hh_cache *hh,
311 const struct net_device *dev,
312 const unsigned char *haddr);
313 bool (*validate)(const char *ll_header, unsigned int len);
314 __be16 (*parse_protocol)(const struct sk_buff *skb);
317 /* These flag bits are private to the generic network queueing
318 * layer; they may not be explicitly referenced by any other
322 enum netdev_state_t {
324 __LINK_STATE_PRESENT,
325 __LINK_STATE_NOCARRIER,
326 __LINK_STATE_LINKWATCH_PENDING,
327 __LINK_STATE_DORMANT,
328 __LINK_STATE_TESTING,
332 struct list_head list;
337 * size of gro hash buckets, must less than bit number of
338 * napi_struct::gro_bitmask
340 #define GRO_HASH_BUCKETS 8
343 * Structure for NAPI scheduling similar to tasklet but with weighting
346 /* The poll_list must only be managed by the entity which
347 * changes the state of the NAPI_STATE_SCHED bit. This means
348 * whoever atomically sets that bit can add this napi_struct
349 * to the per-CPU poll_list, and whoever clears that bit
350 * can remove from the list right before clearing the bit.
352 struct list_head poll_list;
356 int defer_hard_irqs_count;
357 unsigned long gro_bitmask;
358 int (*poll)(struct napi_struct *, int);
359 #ifdef CONFIG_NETPOLL
362 struct net_device *dev;
363 struct gro_list gro_hash[GRO_HASH_BUCKETS];
365 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
366 int rx_count; /* length of rx_list */
367 struct hrtimer timer;
368 struct list_head dev_list;
369 struct hlist_node napi_hash_node;
370 unsigned int napi_id;
371 struct task_struct *thread;
375 NAPI_STATE_SCHED, /* Poll is scheduled */
376 NAPI_STATE_MISSED, /* reschedule a napi */
377 NAPI_STATE_DISABLE, /* Disable pending */
378 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
379 NAPI_STATE_LISTED, /* NAPI added to system lists */
380 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
381 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
382 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
383 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
384 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
388 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
389 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
390 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
391 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
392 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
393 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
394 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
395 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
396 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
397 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
407 typedef enum gro_result gro_result_t;
410 * enum rx_handler_result - Possible return values for rx_handlers.
411 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
413 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
414 * case skb->dev was changed by rx_handler.
415 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
416 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
418 * rx_handlers are functions called from inside __netif_receive_skb(), to do
419 * special processing of the skb, prior to delivery to protocol handlers.
421 * Currently, a net_device can only have a single rx_handler registered. Trying
422 * to register a second rx_handler will return -EBUSY.
424 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
425 * To unregister a rx_handler on a net_device, use
426 * netdev_rx_handler_unregister().
428 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
431 * If the rx_handler consumed the skb in some way, it should return
432 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
433 * the skb to be delivered in some other way.
435 * If the rx_handler changed skb->dev, to divert the skb to another
436 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
437 * new device will be called if it exists.
439 * If the rx_handler decides the skb should be ignored, it should return
440 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
441 * are registered on exact device (ptype->dev == skb->dev).
443 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
444 * delivered, it should return RX_HANDLER_PASS.
446 * A device without a registered rx_handler will behave as if rx_handler
447 * returned RX_HANDLER_PASS.
450 enum rx_handler_result {
456 typedef enum rx_handler_result rx_handler_result_t;
457 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
459 void __napi_schedule(struct napi_struct *n);
460 void __napi_schedule_irqoff(struct napi_struct *n);
462 static inline bool napi_disable_pending(struct napi_struct *n)
464 return test_bit(NAPI_STATE_DISABLE, &n->state);
467 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
469 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
472 bool napi_schedule_prep(struct napi_struct *n);
475 * napi_schedule - schedule NAPI poll
478 * Schedule NAPI poll routine to be called if it is not already
481 static inline void napi_schedule(struct napi_struct *n)
483 if (napi_schedule_prep(n))
488 * napi_schedule_irqoff - schedule NAPI poll
491 * Variant of napi_schedule(), assuming hard irqs are masked.
493 static inline void napi_schedule_irqoff(struct napi_struct *n)
495 if (napi_schedule_prep(n))
496 __napi_schedule_irqoff(n);
499 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
500 static inline bool napi_reschedule(struct napi_struct *napi)
502 if (napi_schedule_prep(napi)) {
503 __napi_schedule(napi);
509 bool napi_complete_done(struct napi_struct *n, int work_done);
511 * napi_complete - NAPI processing complete
514 * Mark NAPI processing as complete.
515 * Consider using napi_complete_done() instead.
516 * Return false if device should avoid rearming interrupts.
518 static inline bool napi_complete(struct napi_struct *n)
520 return napi_complete_done(n, 0);
523 int dev_set_threaded(struct net_device *dev, bool threaded);
526 * napi_disable - prevent NAPI from scheduling
529 * Stop NAPI from being scheduled on this context.
530 * Waits till any outstanding processing completes.
532 void napi_disable(struct napi_struct *n);
534 void napi_enable(struct napi_struct *n);
537 * napi_synchronize - wait until NAPI is not running
540 * Wait until NAPI is done being scheduled on this context.
541 * Waits till any outstanding processing completes but
542 * does not disable future activations.
544 static inline void napi_synchronize(const struct napi_struct *n)
546 if (IS_ENABLED(CONFIG_SMP))
547 while (test_bit(NAPI_STATE_SCHED, &n->state))
554 * napi_if_scheduled_mark_missed - if napi is running, set the
558 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
561 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
563 unsigned long val, new;
565 val = READ_ONCE(n->state);
567 if (val & NAPIF_STATE_DISABLE)
570 if (!(val & NAPIF_STATE_SCHED))
573 new = val | NAPIF_STATE_MISSED;
574 } while (!try_cmpxchg(&n->state, &val, new));
579 enum netdev_queue_state_t {
580 __QUEUE_STATE_DRV_XOFF,
581 __QUEUE_STATE_STACK_XOFF,
582 __QUEUE_STATE_FROZEN,
585 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
586 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
587 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
589 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
590 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
592 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
596 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
597 * netif_tx_* functions below are used to manipulate this flag. The
598 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
599 * queue independently. The netif_xmit_*stopped functions below are called
600 * to check if the queue has been stopped by the driver or stack (either
601 * of the XOFF bits are set in the state). Drivers should not need to call
602 * netif_xmit*stopped functions, they should only be using netif_tx_*.
605 struct netdev_queue {
609 struct net_device *dev;
610 netdevice_tracker dev_tracker;
612 struct Qdisc __rcu *qdisc;
613 struct Qdisc *qdisc_sleeping;
617 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
620 unsigned long tx_maxrate;
622 * Number of TX timeouts for this queue
623 * (/sys/class/net/DEV/Q/trans_timeout)
625 atomic_long_t trans_timeout;
627 /* Subordinate device that the queue has been assigned to */
628 struct net_device *sb_dev;
629 #ifdef CONFIG_XDP_SOCKETS
630 struct xsk_buff_pool *pool;
635 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
638 * Time (in jiffies) of last Tx
640 unsigned long trans_start;
647 } ____cacheline_aligned_in_smp;
649 extern int sysctl_fb_tunnels_only_for_init_net;
650 extern int sysctl_devconf_inherit_init_net;
653 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
654 * == 1 : For initns only
657 static inline bool net_has_fallback_tunnels(const struct net *net)
659 #if IS_ENABLED(CONFIG_SYSCTL)
660 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
662 return !fb_tunnels_only_for_init_net ||
663 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
669 static inline int net_inherit_devconf(void)
671 #if IS_ENABLED(CONFIG_SYSCTL)
672 return READ_ONCE(sysctl_devconf_inherit_init_net);
678 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
680 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
687 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
689 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
696 * This structure holds an RPS map which can be of variable length. The
697 * map is an array of CPUs.
704 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
707 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
708 * tail pointer for that CPU's input queue at the time of last enqueue, and
709 * a hardware filter index.
711 struct rps_dev_flow {
714 unsigned int last_qtail;
716 #define RPS_NO_FILTER 0xffff
719 * The rps_dev_flow_table structure contains a table of flow mappings.
721 struct rps_dev_flow_table {
724 struct rps_dev_flow flows[];
726 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
727 ((_num) * sizeof(struct rps_dev_flow)))
730 * The rps_sock_flow_table contains mappings of flows to the last CPU
731 * on which they were processed by the application (set in recvmsg).
732 * Each entry is a 32bit value. Upper part is the high-order bits
733 * of flow hash, lower part is CPU number.
734 * rps_cpu_mask is used to partition the space, depending on number of
735 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
736 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
737 * meaning we use 32-6=26 bits for the hash.
739 struct rps_sock_flow_table {
742 u32 ents[] ____cacheline_aligned_in_smp;
744 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
746 #define RPS_NO_CPU 0xffff
748 extern u32 rps_cpu_mask;
749 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
751 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
755 unsigned int index = hash & table->mask;
756 u32 val = hash & ~rps_cpu_mask;
758 /* We only give a hint, preemption can change CPU under us */
759 val |= raw_smp_processor_id();
761 if (table->ents[index] != val)
762 table->ents[index] = val;
766 #ifdef CONFIG_RFS_ACCEL
767 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
770 #endif /* CONFIG_RPS */
772 /* This structure contains an instance of an RX queue. */
773 struct netdev_rx_queue {
774 struct xdp_rxq_info xdp_rxq;
776 struct rps_map __rcu *rps_map;
777 struct rps_dev_flow_table __rcu *rps_flow_table;
780 struct net_device *dev;
781 netdevice_tracker dev_tracker;
783 #ifdef CONFIG_XDP_SOCKETS
784 struct xsk_buff_pool *pool;
786 } ____cacheline_aligned_in_smp;
789 * RX queue sysfs structures and functions.
791 struct rx_queue_attribute {
792 struct attribute attr;
793 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
794 ssize_t (*store)(struct netdev_rx_queue *queue,
795 const char *buf, size_t len);
798 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
807 * This structure holds an XPS map which can be of variable length. The
808 * map is an array of queues.
812 unsigned int alloc_len;
816 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
817 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
818 - sizeof(struct xps_map)) / sizeof(u16))
821 * This structure holds all XPS maps for device. Maps are indexed by CPU.
823 * We keep track of the number of cpus/rxqs used when the struct is allocated,
824 * in nr_ids. This will help not accessing out-of-bound memory.
826 * We keep track of the number of traffic classes used when the struct is
827 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
828 * not crossing its upper bound, as the original dev->num_tc can be updated in
831 struct xps_dev_maps {
835 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
838 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
839 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
841 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
842 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
844 #endif /* CONFIG_XPS */
846 #define TC_MAX_QUEUE 16
847 #define TC_BITMASK 15
848 /* HW offloaded queuing disciplines txq count and offset maps */
849 struct netdev_tc_txq {
854 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
856 * This structure is to hold information about the device
857 * configured to run FCoE protocol stack.
859 struct netdev_fcoe_hbainfo {
860 char manufacturer[64];
861 char serial_number[64];
862 char hardware_version[64];
863 char driver_version[64];
864 char optionrom_version[64];
865 char firmware_version[64];
867 char model_description[256];
871 #define MAX_PHYS_ITEM_ID_LEN 32
873 /* This structure holds a unique identifier to identify some
874 * physical item (port for example) used by a netdevice.
876 struct netdev_phys_item_id {
877 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
878 unsigned char id_len;
881 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
882 struct netdev_phys_item_id *b)
884 return a->id_len == b->id_len &&
885 memcmp(a->id, b->id, a->id_len) == 0;
888 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
890 struct net_device *sb_dev);
892 enum net_device_path_type {
893 DEV_PATH_ETHERNET = 0,
901 struct net_device_path {
902 enum net_device_path_type type;
903 const struct net_device *dev;
912 DEV_PATH_BR_VLAN_KEEP,
913 DEV_PATH_BR_VLAN_TAG,
914 DEV_PATH_BR_VLAN_UNTAG,
915 DEV_PATH_BR_VLAN_UNTAG_HW,
933 #define NET_DEVICE_PATH_STACK_MAX 5
934 #define NET_DEVICE_PATH_VLAN_MAX 2
936 struct net_device_path_stack {
938 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
941 struct net_device_path_ctx {
942 const struct net_device *dev;
949 } vlan[NET_DEVICE_PATH_VLAN_MAX];
954 TC_SETUP_QDISC_MQPRIO,
957 TC_SETUP_CLSMATCHALL,
967 TC_SETUP_QDISC_TAPRIO,
976 /* These structures hold the attributes of bpf state that are being passed
977 * to the netdevice through the bpf op.
979 enum bpf_netdev_command {
980 /* Set or clear a bpf program used in the earliest stages of packet
981 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
982 * is responsible for calling bpf_prog_put on any old progs that are
983 * stored. In case of error, the callee need not release the new prog
984 * reference, but on success it takes ownership and must bpf_prog_put
985 * when it is no longer used.
989 /* BPF program for offload callbacks, invoked at program load time. */
990 BPF_OFFLOAD_MAP_ALLOC,
991 BPF_OFFLOAD_MAP_FREE,
995 struct bpf_prog_offload_ops;
996 struct netlink_ext_ack;
998 struct xdp_dev_bulk_queue;
1008 struct bpf_xdp_entity {
1009 struct bpf_prog *prog;
1010 struct bpf_xdp_link *link;
1014 enum bpf_netdev_command command;
1016 /* XDP_SETUP_PROG */
1019 struct bpf_prog *prog;
1020 struct netlink_ext_ack *extack;
1022 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1024 struct bpf_offloaded_map *offmap;
1026 /* XDP_SETUP_XSK_POOL */
1028 struct xsk_buff_pool *pool;
1034 /* Flags for ndo_xsk_wakeup. */
1035 #define XDP_WAKEUP_RX (1 << 0)
1036 #define XDP_WAKEUP_TX (1 << 1)
1038 #ifdef CONFIG_XFRM_OFFLOAD
1039 struct xfrmdev_ops {
1040 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1041 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1042 void (*xdo_dev_state_free) (struct xfrm_state *x);
1043 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1044 struct xfrm_state *x);
1045 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1046 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1047 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1048 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1049 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1053 struct dev_ifalias {
1054 struct rcu_head rcuhead;
1061 struct netdev_net_notifier {
1062 struct list_head list;
1063 struct notifier_block *nb;
1067 * This structure defines the management hooks for network devices.
1068 * The following hooks can be defined; unless noted otherwise, they are
1069 * optional and can be filled with a null pointer.
1071 * int (*ndo_init)(struct net_device *dev);
1072 * This function is called once when a network device is registered.
1073 * The network device can use this for any late stage initialization
1074 * or semantic validation. It can fail with an error code which will
1075 * be propagated back to register_netdev.
1077 * void (*ndo_uninit)(struct net_device *dev);
1078 * This function is called when device is unregistered or when registration
1079 * fails. It is not called if init fails.
1081 * int (*ndo_open)(struct net_device *dev);
1082 * This function is called when a network device transitions to the up
1085 * int (*ndo_stop)(struct net_device *dev);
1086 * This function is called when a network device transitions to the down
1089 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1090 * struct net_device *dev);
1091 * Called when a packet needs to be transmitted.
1092 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1093 * the queue before that can happen; it's for obsolete devices and weird
1094 * corner cases, but the stack really does a non-trivial amount
1095 * of useless work if you return NETDEV_TX_BUSY.
1096 * Required; cannot be NULL.
1098 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1099 * struct net_device *dev
1100 * netdev_features_t features);
1101 * Called by core transmit path to determine if device is capable of
1102 * performing offload operations on a given packet. This is to give
1103 * the device an opportunity to implement any restrictions that cannot
1104 * be otherwise expressed by feature flags. The check is called with
1105 * the set of features that the stack has calculated and it returns
1106 * those the driver believes to be appropriate.
1108 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1109 * struct net_device *sb_dev);
1110 * Called to decide which queue to use when device supports multiple
1113 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1114 * This function is called to allow device receiver to make
1115 * changes to configuration when multicast or promiscuous is enabled.
1117 * void (*ndo_set_rx_mode)(struct net_device *dev);
1118 * This function is called device changes address list filtering.
1119 * If driver handles unicast address filtering, it should set
1120 * IFF_UNICAST_FLT in its priv_flags.
1122 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1123 * This function is called when the Media Access Control address
1124 * needs to be changed. If this interface is not defined, the
1125 * MAC address can not be changed.
1127 * int (*ndo_validate_addr)(struct net_device *dev);
1128 * Test if Media Access Control address is valid for the device.
1130 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1131 * Old-style ioctl entry point. This is used internally by the
1132 * appletalk and ieee802154 subsystems but is no longer called by
1133 * the device ioctl handler.
1135 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1136 * Used by the bonding driver for its device specific ioctls:
1137 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1138 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1140 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1141 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1142 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1144 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1145 * Used to set network devices bus interface parameters. This interface
1146 * is retained for legacy reasons; new devices should use the bus
1147 * interface (PCI) for low level management.
1149 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1150 * Called when a user wants to change the Maximum Transfer Unit
1153 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1154 * Callback used when the transmitter has not made any progress
1155 * for dev->watchdog ticks.
1157 * void (*ndo_get_stats64)(struct net_device *dev,
1158 * struct rtnl_link_stats64 *storage);
1159 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1160 * Called when a user wants to get the network device usage
1161 * statistics. Drivers must do one of the following:
1162 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1163 * rtnl_link_stats64 structure passed by the caller.
1164 * 2. Define @ndo_get_stats to update a net_device_stats structure
1165 * (which should normally be dev->stats) and return a pointer to
1166 * it. The structure may be changed asynchronously only if each
1167 * field is written atomically.
1168 * 3. Update dev->stats asynchronously and atomically, and define
1169 * neither operation.
1171 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1172 * Return true if this device supports offload stats of this attr_id.
1174 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1176 * Get statistics for offload operations by attr_id. Write it into the
1177 * attr_data pointer.
1179 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1180 * If device supports VLAN filtering this function is called when a
1181 * VLAN id is registered.
1183 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1184 * If device supports VLAN filtering this function is called when a
1185 * VLAN id is unregistered.
1187 * void (*ndo_poll_controller)(struct net_device *dev);
1189 * SR-IOV management functions.
1190 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1191 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1192 * u8 qos, __be16 proto);
1193 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1195 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1196 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1197 * int (*ndo_get_vf_config)(struct net_device *dev,
1198 * int vf, struct ifla_vf_info *ivf);
1199 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1200 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1201 * struct nlattr *port[]);
1203 * Enable or disable the VF ability to query its RSS Redirection Table and
1204 * Hash Key. This is needed since on some devices VF share this information
1205 * with PF and querying it may introduce a theoretical security risk.
1206 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1207 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1208 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1210 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1211 * This is always called from the stack with the rtnl lock held and netif
1212 * tx queues stopped. This allows the netdevice to perform queue
1213 * management safely.
1215 * Fiber Channel over Ethernet (FCoE) offload functions.
1216 * int (*ndo_fcoe_enable)(struct net_device *dev);
1217 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1218 * so the underlying device can perform whatever needed configuration or
1219 * initialization to support acceleration of FCoE traffic.
1221 * int (*ndo_fcoe_disable)(struct net_device *dev);
1222 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1223 * so the underlying device can perform whatever needed clean-ups to
1224 * stop supporting acceleration of FCoE traffic.
1226 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1227 * struct scatterlist *sgl, unsigned int sgc);
1228 * Called when the FCoE Initiator wants to initialize an I/O that
1229 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1230 * perform necessary setup and returns 1 to indicate the device is set up
1231 * successfully to perform DDP on this I/O, otherwise this returns 0.
1233 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1234 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1235 * indicated by the FC exchange id 'xid', so the underlying device can
1236 * clean up and reuse resources for later DDP requests.
1238 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1239 * struct scatterlist *sgl, unsigned int sgc);
1240 * Called when the FCoE Target wants to initialize an I/O that
1241 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1242 * perform necessary setup and returns 1 to indicate the device is set up
1243 * successfully to perform DDP on this I/O, otherwise this returns 0.
1245 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1246 * struct netdev_fcoe_hbainfo *hbainfo);
1247 * Called when the FCoE Protocol stack wants information on the underlying
1248 * device. This information is utilized by the FCoE protocol stack to
1249 * register attributes with Fiber Channel management service as per the
1250 * FC-GS Fabric Device Management Information(FDMI) specification.
1252 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1253 * Called when the underlying device wants to override default World Wide
1254 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1255 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1256 * protocol stack to use.
1259 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1260 * u16 rxq_index, u32 flow_id);
1261 * Set hardware filter for RFS. rxq_index is the target queue index;
1262 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1263 * Return the filter ID on success, or a negative error code.
1265 * Slave management functions (for bridge, bonding, etc).
1266 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1267 * Called to make another netdev an underling.
1269 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1270 * Called to release previously enslaved netdev.
1272 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1273 * struct sk_buff *skb,
1275 * Get the xmit slave of master device. If all_slaves is true, function
1276 * assume all the slaves can transmit.
1278 * Feature/offload setting functions.
1279 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1280 * netdev_features_t features);
1281 * Adjusts the requested feature flags according to device-specific
1282 * constraints, and returns the resulting flags. Must not modify
1285 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1286 * Called to update device configuration to new features. Passed
1287 * feature set might be less than what was returned by ndo_fix_features()).
1288 * Must return >0 or -errno if it changed dev->features itself.
1290 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1291 * struct net_device *dev,
1292 * const unsigned char *addr, u16 vid, u16 flags,
1293 * struct netlink_ext_ack *extack);
1294 * Adds an FDB entry to dev for addr.
1295 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1296 * struct net_device *dev,
1297 * const unsigned char *addr, u16 vid)
1298 * Deletes the FDB entry from dev coresponding to addr.
1299 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1300 * struct net_device *dev,
1302 * struct netlink_ext_ack *extack);
1303 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1304 * struct net_device *dev, struct net_device *filter_dev,
1306 * Used to add FDB entries to dump requests. Implementers should add
1307 * entries to skb and update idx with the number of entries.
1309 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1310 * u16 flags, struct netlink_ext_ack *extack)
1311 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1312 * struct net_device *dev, u32 filter_mask,
1314 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1317 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1318 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1319 * which do not represent real hardware may define this to allow their
1320 * userspace components to manage their virtual carrier state. Devices
1321 * that determine carrier state from physical hardware properties (eg
1322 * network cables) or protocol-dependent mechanisms (eg
1323 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1325 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1326 * struct netdev_phys_item_id *ppid);
1327 * Called to get ID of physical port of this device. If driver does
1328 * not implement this, it is assumed that the hw is not able to have
1329 * multiple net devices on single physical port.
1331 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1332 * struct netdev_phys_item_id *ppid)
1333 * Called to get the parent ID of the physical port of this device.
1335 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1336 * struct net_device *dev)
1337 * Called by upper layer devices to accelerate switching or other
1338 * station functionality into hardware. 'pdev is the lowerdev
1339 * to use for the offload and 'dev' is the net device that will
1340 * back the offload. Returns a pointer to the private structure
1341 * the upper layer will maintain.
1342 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1343 * Called by upper layer device to delete the station created
1344 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1345 * the station and priv is the structure returned by the add
1347 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1348 * int queue_index, u32 maxrate);
1349 * Called when a user wants to set a max-rate limitation of specific
1351 * int (*ndo_get_iflink)(const struct net_device *dev);
1352 * Called to get the iflink value of this device.
1353 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1354 * This function is used to get egress tunnel information for given skb.
1355 * This is useful for retrieving outer tunnel header parameters while
1357 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1358 * This function is used to specify the headroom that the skb must
1359 * consider when allocation skb during packet reception. Setting
1360 * appropriate rx headroom value allows avoiding skb head copy on
1361 * forward. Setting a negative value resets the rx headroom to the
1363 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1364 * This function is used to set or query state related to XDP on the
1365 * netdevice and manage BPF offload. See definition of
1366 * enum bpf_netdev_command for details.
1367 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1369 * This function is used to submit @n XDP packets for transmit on a
1370 * netdevice. Returns number of frames successfully transmitted, frames
1371 * that got dropped are freed/returned via xdp_return_frame().
1372 * Returns negative number, means general error invoking ndo, meaning
1373 * no frames were xmit'ed and core-caller will free all frames.
1374 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1375 * struct xdp_buff *xdp);
1376 * Get the xmit slave of master device based on the xdp_buff.
1377 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1378 * This function is used to wake up the softirq, ksoftirqd or kthread
1379 * responsible for sending and/or receiving packets on a specific
1380 * queue id bound to an AF_XDP socket. The flags field specifies if
1381 * only RX, only Tx, or both should be woken up using the flags
1382 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1383 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1385 * Add, change, delete or get information on an IPv4 tunnel.
1386 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1387 * If a device is paired with a peer device, return the peer instance.
1388 * The caller must be under RCU read context.
1389 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1390 * Get the forwarding path to reach the real device from the HW destination address
1391 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1392 * const struct skb_shared_hwtstamps *hwtstamps,
1394 * Get hardware timestamp based on normal/adjustable time or free running
1395 * cycle counter. This function is required if physical clock supports a
1396 * free running cycle counter.
1398 struct net_device_ops {
1399 int (*ndo_init)(struct net_device *dev);
1400 void (*ndo_uninit)(struct net_device *dev);
1401 int (*ndo_open)(struct net_device *dev);
1402 int (*ndo_stop)(struct net_device *dev);
1403 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1404 struct net_device *dev);
1405 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1406 struct net_device *dev,
1407 netdev_features_t features);
1408 u16 (*ndo_select_queue)(struct net_device *dev,
1409 struct sk_buff *skb,
1410 struct net_device *sb_dev);
1411 void (*ndo_change_rx_flags)(struct net_device *dev,
1413 void (*ndo_set_rx_mode)(struct net_device *dev);
1414 int (*ndo_set_mac_address)(struct net_device *dev,
1416 int (*ndo_validate_addr)(struct net_device *dev);
1417 int (*ndo_do_ioctl)(struct net_device *dev,
1418 struct ifreq *ifr, int cmd);
1419 int (*ndo_eth_ioctl)(struct net_device *dev,
1420 struct ifreq *ifr, int cmd);
1421 int (*ndo_siocbond)(struct net_device *dev,
1422 struct ifreq *ifr, int cmd);
1423 int (*ndo_siocwandev)(struct net_device *dev,
1424 struct if_settings *ifs);
1425 int (*ndo_siocdevprivate)(struct net_device *dev,
1427 void __user *data, int cmd);
1428 int (*ndo_set_config)(struct net_device *dev,
1430 int (*ndo_change_mtu)(struct net_device *dev,
1432 int (*ndo_neigh_setup)(struct net_device *dev,
1433 struct neigh_parms *);
1434 void (*ndo_tx_timeout) (struct net_device *dev,
1435 unsigned int txqueue);
1437 void (*ndo_get_stats64)(struct net_device *dev,
1438 struct rtnl_link_stats64 *storage);
1439 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1440 int (*ndo_get_offload_stats)(int attr_id,
1441 const struct net_device *dev,
1443 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1445 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1446 __be16 proto, u16 vid);
1447 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1448 __be16 proto, u16 vid);
1449 #ifdef CONFIG_NET_POLL_CONTROLLER
1450 void (*ndo_poll_controller)(struct net_device *dev);
1451 int (*ndo_netpoll_setup)(struct net_device *dev,
1452 struct netpoll_info *info);
1453 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1455 int (*ndo_set_vf_mac)(struct net_device *dev,
1456 int queue, u8 *mac);
1457 int (*ndo_set_vf_vlan)(struct net_device *dev,
1458 int queue, u16 vlan,
1459 u8 qos, __be16 proto);
1460 int (*ndo_set_vf_rate)(struct net_device *dev,
1461 int vf, int min_tx_rate,
1463 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1464 int vf, bool setting);
1465 int (*ndo_set_vf_trust)(struct net_device *dev,
1466 int vf, bool setting);
1467 int (*ndo_get_vf_config)(struct net_device *dev,
1469 struct ifla_vf_info *ivf);
1470 int (*ndo_set_vf_link_state)(struct net_device *dev,
1471 int vf, int link_state);
1472 int (*ndo_get_vf_stats)(struct net_device *dev,
1474 struct ifla_vf_stats
1476 int (*ndo_set_vf_port)(struct net_device *dev,
1478 struct nlattr *port[]);
1479 int (*ndo_get_vf_port)(struct net_device *dev,
1480 int vf, struct sk_buff *skb);
1481 int (*ndo_get_vf_guid)(struct net_device *dev,
1483 struct ifla_vf_guid *node_guid,
1484 struct ifla_vf_guid *port_guid);
1485 int (*ndo_set_vf_guid)(struct net_device *dev,
1488 int (*ndo_set_vf_rss_query_en)(
1489 struct net_device *dev,
1490 int vf, bool setting);
1491 int (*ndo_setup_tc)(struct net_device *dev,
1492 enum tc_setup_type type,
1494 #if IS_ENABLED(CONFIG_FCOE)
1495 int (*ndo_fcoe_enable)(struct net_device *dev);
1496 int (*ndo_fcoe_disable)(struct net_device *dev);
1497 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1499 struct scatterlist *sgl,
1501 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1503 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1505 struct scatterlist *sgl,
1507 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1508 struct netdev_fcoe_hbainfo *hbainfo);
1511 #if IS_ENABLED(CONFIG_LIBFCOE)
1512 #define NETDEV_FCOE_WWNN 0
1513 #define NETDEV_FCOE_WWPN 1
1514 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1515 u64 *wwn, int type);
1518 #ifdef CONFIG_RFS_ACCEL
1519 int (*ndo_rx_flow_steer)(struct net_device *dev,
1520 const struct sk_buff *skb,
1524 int (*ndo_add_slave)(struct net_device *dev,
1525 struct net_device *slave_dev,
1526 struct netlink_ext_ack *extack);
1527 int (*ndo_del_slave)(struct net_device *dev,
1528 struct net_device *slave_dev);
1529 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1530 struct sk_buff *skb,
1532 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1534 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1535 netdev_features_t features);
1536 int (*ndo_set_features)(struct net_device *dev,
1537 netdev_features_t features);
1538 int (*ndo_neigh_construct)(struct net_device *dev,
1539 struct neighbour *n);
1540 void (*ndo_neigh_destroy)(struct net_device *dev,
1541 struct neighbour *n);
1543 int (*ndo_fdb_add)(struct ndmsg *ndm,
1544 struct nlattr *tb[],
1545 struct net_device *dev,
1546 const unsigned char *addr,
1549 struct netlink_ext_ack *extack);
1550 int (*ndo_fdb_del)(struct ndmsg *ndm,
1551 struct nlattr *tb[],
1552 struct net_device *dev,
1553 const unsigned char *addr,
1554 u16 vid, struct netlink_ext_ack *extack);
1555 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1556 struct nlattr *tb[],
1557 struct net_device *dev,
1559 struct netlink_ext_ack *extack);
1560 int (*ndo_fdb_dump)(struct sk_buff *skb,
1561 struct netlink_callback *cb,
1562 struct net_device *dev,
1563 struct net_device *filter_dev,
1565 int (*ndo_fdb_get)(struct sk_buff *skb,
1566 struct nlattr *tb[],
1567 struct net_device *dev,
1568 const unsigned char *addr,
1569 u16 vid, u32 portid, u32 seq,
1570 struct netlink_ext_ack *extack);
1571 int (*ndo_bridge_setlink)(struct net_device *dev,
1572 struct nlmsghdr *nlh,
1574 struct netlink_ext_ack *extack);
1575 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1577 struct net_device *dev,
1580 int (*ndo_bridge_dellink)(struct net_device *dev,
1581 struct nlmsghdr *nlh,
1583 int (*ndo_change_carrier)(struct net_device *dev,
1585 int (*ndo_get_phys_port_id)(struct net_device *dev,
1586 struct netdev_phys_item_id *ppid);
1587 int (*ndo_get_port_parent_id)(struct net_device *dev,
1588 struct netdev_phys_item_id *ppid);
1589 int (*ndo_get_phys_port_name)(struct net_device *dev,
1590 char *name, size_t len);
1591 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1592 struct net_device *dev);
1593 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1596 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1599 int (*ndo_get_iflink)(const struct net_device *dev);
1600 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1601 struct sk_buff *skb);
1602 void (*ndo_set_rx_headroom)(struct net_device *dev,
1603 int needed_headroom);
1604 int (*ndo_bpf)(struct net_device *dev,
1605 struct netdev_bpf *bpf);
1606 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1607 struct xdp_frame **xdp,
1609 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1610 struct xdp_buff *xdp);
1611 int (*ndo_xsk_wakeup)(struct net_device *dev,
1612 u32 queue_id, u32 flags);
1613 int (*ndo_tunnel_ctl)(struct net_device *dev,
1614 struct ip_tunnel_parm *p, int cmd);
1615 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1616 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1617 struct net_device_path *path);
1618 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1619 const struct skb_shared_hwtstamps *hwtstamps,
1623 struct xdp_metadata_ops {
1624 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp);
1625 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash);
1629 * enum netdev_priv_flags - &struct net_device priv_flags
1631 * These are the &struct net_device, they are only set internally
1632 * by drivers and used in the kernel. These flags are invisible to
1633 * userspace; this means that the order of these flags can change
1634 * during any kernel release.
1636 * You should have a pretty good reason to be extending these flags.
1638 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1639 * @IFF_EBRIDGE: Ethernet bridging device
1640 * @IFF_BONDING: bonding master or slave
1641 * @IFF_ISATAP: ISATAP interface (RFC4214)
1642 * @IFF_WAN_HDLC: WAN HDLC device
1643 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1645 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1646 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1647 * @IFF_MACVLAN_PORT: device used as macvlan port
1648 * @IFF_BRIDGE_PORT: device used as bridge port
1649 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1650 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1651 * @IFF_UNICAST_FLT: Supports unicast filtering
1652 * @IFF_TEAM_PORT: device used as team port
1653 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1654 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1655 * change when it's running
1656 * @IFF_MACVLAN: Macvlan device
1657 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1658 * underlying stacked devices
1659 * @IFF_L3MDEV_MASTER: device is an L3 master device
1660 * @IFF_NO_QUEUE: device can run without qdisc attached
1661 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1662 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1663 * @IFF_TEAM: device is a team device
1664 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1665 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1666 * entity (i.e. the master device for bridged veth)
1667 * @IFF_MACSEC: device is a MACsec device
1668 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1669 * @IFF_FAILOVER: device is a failover master device
1670 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1671 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1672 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1673 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1674 * skb_headlen(skb) == 0 (data starts from frag0)
1675 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1677 enum netdev_priv_flags {
1678 IFF_802_1Q_VLAN = 1<<0,
1682 IFF_WAN_HDLC = 1<<4,
1683 IFF_XMIT_DST_RELEASE = 1<<5,
1684 IFF_DONT_BRIDGE = 1<<6,
1685 IFF_DISABLE_NETPOLL = 1<<7,
1686 IFF_MACVLAN_PORT = 1<<8,
1687 IFF_BRIDGE_PORT = 1<<9,
1688 IFF_OVS_DATAPATH = 1<<10,
1689 IFF_TX_SKB_SHARING = 1<<11,
1690 IFF_UNICAST_FLT = 1<<12,
1691 IFF_TEAM_PORT = 1<<13,
1692 IFF_SUPP_NOFCS = 1<<14,
1693 IFF_LIVE_ADDR_CHANGE = 1<<15,
1694 IFF_MACVLAN = 1<<16,
1695 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1696 IFF_L3MDEV_MASTER = 1<<18,
1697 IFF_NO_QUEUE = 1<<19,
1698 IFF_OPENVSWITCH = 1<<20,
1699 IFF_L3MDEV_SLAVE = 1<<21,
1701 IFF_RXFH_CONFIGURED = 1<<23,
1702 IFF_PHONY_HEADROOM = 1<<24,
1704 IFF_NO_RX_HANDLER = 1<<26,
1705 IFF_FAILOVER = 1<<27,
1706 IFF_FAILOVER_SLAVE = 1<<28,
1707 IFF_L3MDEV_RX_HANDLER = 1<<29,
1708 IFF_NO_ADDRCONF = BIT_ULL(30),
1709 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1710 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1713 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1714 #define IFF_EBRIDGE IFF_EBRIDGE
1715 #define IFF_BONDING IFF_BONDING
1716 #define IFF_ISATAP IFF_ISATAP
1717 #define IFF_WAN_HDLC IFF_WAN_HDLC
1718 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1719 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1720 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1721 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1722 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1723 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1724 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1725 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1726 #define IFF_TEAM_PORT IFF_TEAM_PORT
1727 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1728 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1729 #define IFF_MACVLAN IFF_MACVLAN
1730 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1731 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1732 #define IFF_NO_QUEUE IFF_NO_QUEUE
1733 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1734 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1735 #define IFF_TEAM IFF_TEAM
1736 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1737 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1738 #define IFF_MACSEC IFF_MACSEC
1739 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1740 #define IFF_FAILOVER IFF_FAILOVER
1741 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1742 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1743 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1745 /* Specifies the type of the struct net_device::ml_priv pointer */
1746 enum netdev_ml_priv_type {
1752 * struct net_device - The DEVICE structure.
1754 * Actually, this whole structure is a big mistake. It mixes I/O
1755 * data with strictly "high-level" data, and it has to know about
1756 * almost every data structure used in the INET module.
1758 * @name: This is the first field of the "visible" part of this structure
1759 * (i.e. as seen by users in the "Space.c" file). It is the name
1762 * @name_node: Name hashlist node
1763 * @ifalias: SNMP alias
1764 * @mem_end: Shared memory end
1765 * @mem_start: Shared memory start
1766 * @base_addr: Device I/O address
1767 * @irq: Device IRQ number
1769 * @state: Generic network queuing layer state, see netdev_state_t
1770 * @dev_list: The global list of network devices
1771 * @napi_list: List entry used for polling NAPI devices
1772 * @unreg_list: List entry when we are unregistering the
1773 * device; see the function unregister_netdev
1774 * @close_list: List entry used when we are closing the device
1775 * @ptype_all: Device-specific packet handlers for all protocols
1776 * @ptype_specific: Device-specific, protocol-specific packet handlers
1778 * @adj_list: Directly linked devices, like slaves for bonding
1779 * @features: Currently active device features
1780 * @hw_features: User-changeable features
1782 * @wanted_features: User-requested features
1783 * @vlan_features: Mask of features inheritable by VLAN devices
1785 * @hw_enc_features: Mask of features inherited by encapsulating devices
1786 * This field indicates what encapsulation
1787 * offloads the hardware is capable of doing,
1788 * and drivers will need to set them appropriately.
1790 * @mpls_features: Mask of features inheritable by MPLS
1791 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1793 * @ifindex: interface index
1794 * @group: The group the device belongs to
1796 * @stats: Statistics struct, which was left as a legacy, use
1797 * rtnl_link_stats64 instead
1799 * @core_stats: core networking counters,
1800 * do not use this in drivers
1801 * @carrier_up_count: Number of times the carrier has been up
1802 * @carrier_down_count: Number of times the carrier has been down
1804 * @wireless_handlers: List of functions to handle Wireless Extensions,
1806 * see <net/iw_handler.h> for details.
1807 * @wireless_data: Instance data managed by the core of wireless extensions
1809 * @netdev_ops: Includes several pointers to callbacks,
1810 * if one wants to override the ndo_*() functions
1811 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1812 * @ethtool_ops: Management operations
1813 * @l3mdev_ops: Layer 3 master device operations
1814 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1815 * discovery handling. Necessary for e.g. 6LoWPAN.
1816 * @xfrmdev_ops: Transformation offload operations
1817 * @tlsdev_ops: Transport Layer Security offload operations
1818 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1819 * of Layer 2 headers.
1821 * @flags: Interface flags (a la BSD)
1822 * @xdp_features: XDP capability supported by the device
1823 * @priv_flags: Like 'flags' but invisible to userspace,
1824 * see if.h for the definitions
1825 * @gflags: Global flags ( kept as legacy )
1826 * @padded: How much padding added by alloc_netdev()
1827 * @operstate: RFC2863 operstate
1828 * @link_mode: Mapping policy to operstate
1829 * @if_port: Selectable AUI, TP, ...
1831 * @mtu: Interface MTU value
1832 * @min_mtu: Interface Minimum MTU value
1833 * @max_mtu: Interface Maximum MTU value
1834 * @type: Interface hardware type
1835 * @hard_header_len: Maximum hardware header length.
1836 * @min_header_len: Minimum hardware header length
1838 * @needed_headroom: Extra headroom the hardware may need, but not in all
1839 * cases can this be guaranteed
1840 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1841 * cases can this be guaranteed. Some cases also use
1842 * LL_MAX_HEADER instead to allocate the skb
1844 * interface address info:
1846 * @perm_addr: Permanent hw address
1847 * @addr_assign_type: Hw address assignment type
1848 * @addr_len: Hardware address length
1849 * @upper_level: Maximum depth level of upper devices.
1850 * @lower_level: Maximum depth level of lower devices.
1851 * @neigh_priv_len: Used in neigh_alloc()
1852 * @dev_id: Used to differentiate devices that share
1853 * the same link layer address
1854 * @dev_port: Used to differentiate devices that share
1856 * @addr_list_lock: XXX: need comments on this one
1857 * @name_assign_type: network interface name assignment type
1858 * @uc_promisc: Counter that indicates promiscuous mode
1859 * has been enabled due to the need to listen to
1860 * additional unicast addresses in a device that
1861 * does not implement ndo_set_rx_mode()
1862 * @uc: unicast mac addresses
1863 * @mc: multicast mac addresses
1864 * @dev_addrs: list of device hw addresses
1865 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1866 * @promiscuity: Number of times the NIC is told to work in
1867 * promiscuous mode; if it becomes 0 the NIC will
1868 * exit promiscuous mode
1869 * @allmulti: Counter, enables or disables allmulticast mode
1871 * @vlan_info: VLAN info
1872 * @dsa_ptr: dsa specific data
1873 * @tipc_ptr: TIPC specific data
1874 * @atalk_ptr: AppleTalk link
1875 * @ip_ptr: IPv4 specific data
1876 * @ip6_ptr: IPv6 specific data
1877 * @ax25_ptr: AX.25 specific data
1878 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1879 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1881 * @mpls_ptr: mpls_dev struct pointer
1882 * @mctp_ptr: MCTP specific data
1884 * @dev_addr: Hw address (before bcast,
1885 * because most packets are unicast)
1887 * @_rx: Array of RX queues
1888 * @num_rx_queues: Number of RX queues
1889 * allocated at register_netdev() time
1890 * @real_num_rx_queues: Number of RX queues currently active in device
1891 * @xdp_prog: XDP sockets filter program pointer
1892 * @gro_flush_timeout: timeout for GRO layer in NAPI
1893 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1894 * allow to avoid NIC hard IRQ, on busy queues.
1896 * @rx_handler: handler for received packets
1897 * @rx_handler_data: XXX: need comments on this one
1898 * @miniq_ingress: ingress/clsact qdisc specific data for
1899 * ingress processing
1900 * @ingress_queue: XXX: need comments on this one
1901 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1902 * @broadcast: hw bcast address
1904 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1905 * indexed by RX queue number. Assigned by driver.
1906 * This must only be set if the ndo_rx_flow_steer
1907 * operation is defined
1908 * @index_hlist: Device index hash chain
1910 * @_tx: Array of TX queues
1911 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1912 * @real_num_tx_queues: Number of TX queues currently active in device
1913 * @qdisc: Root qdisc from userspace point of view
1914 * @tx_queue_len: Max frames per queue allowed
1915 * @tx_global_lock: XXX: need comments on this one
1916 * @xdp_bulkq: XDP device bulk queue
1917 * @xps_maps: all CPUs/RXQs maps for XPS device
1919 * @xps_maps: XXX: need comments on this one
1920 * @miniq_egress: clsact qdisc specific data for
1922 * @nf_hooks_egress: netfilter hooks executed for egress packets
1923 * @qdisc_hash: qdisc hash table
1924 * @watchdog_timeo: Represents the timeout that is used by
1925 * the watchdog (see dev_watchdog())
1926 * @watchdog_timer: List of timers
1928 * @proto_down_reason: reason a netdev interface is held down
1929 * @pcpu_refcnt: Number of references to this device
1930 * @dev_refcnt: Number of references to this device
1931 * @refcnt_tracker: Tracker directory for tracked references to this device
1932 * @todo_list: Delayed register/unregister
1933 * @link_watch_list: XXX: need comments on this one
1935 * @reg_state: Register/unregister state machine
1936 * @dismantle: Device is going to be freed
1937 * @rtnl_link_state: This enum represents the phases of creating
1940 * @needs_free_netdev: Should unregister perform free_netdev?
1941 * @priv_destructor: Called from unregister
1942 * @npinfo: XXX: need comments on this one
1943 * @nd_net: Network namespace this network device is inside
1945 * @ml_priv: Mid-layer private
1946 * @ml_priv_type: Mid-layer private type
1947 * @lstats: Loopback statistics
1948 * @tstats: Tunnel statistics
1949 * @dstats: Dummy statistics
1950 * @vstats: Virtual ethernet statistics
1955 * @dm_private: Drop monitor private
1957 * @dev: Class/net/name entry
1958 * @sysfs_groups: Space for optional device, statistics and wireless
1961 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1962 * @rtnl_link_ops: Rtnl_link_ops
1964 * @gso_max_size: Maximum size of generic segmentation offload
1965 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1966 * @gso_max_segs: Maximum number of segments that can be passed to the
1968 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1969 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1972 * @dcbnl_ops: Data Center Bridging netlink ops
1973 * @num_tc: Number of traffic classes in the net device
1974 * @tc_to_txq: XXX: need comments on this one
1975 * @prio_tc_map: XXX: need comments on this one
1977 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1979 * @priomap: XXX: need comments on this one
1980 * @phydev: Physical device may attach itself
1981 * for hardware timestamping
1982 * @sfp_bus: attached &struct sfp_bus structure.
1984 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1986 * @proto_down: protocol port state information can be sent to the
1987 * switch driver and used to set the phys state of the
1990 * @wol_enabled: Wake-on-LAN is enabled
1992 * @threaded: napi threaded mode is enabled
1994 * @net_notifier_list: List of per-net netdev notifier block
1995 * that follow this device when it is moved
1996 * to another network namespace.
1998 * @macsec_ops: MACsec offloading ops
2000 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2001 * offload capabilities of the device
2002 * @udp_tunnel_nic: UDP tunnel offload state
2003 * @xdp_state: stores info on attached XDP BPF programs
2005 * @nested_level: Used as a parameter of spin_lock_nested() of
2006 * dev->addr_list_lock.
2007 * @unlink_list: As netif_addr_lock() can be called recursively,
2008 * keep a list of interfaces to be deleted.
2009 * @gro_max_size: Maximum size of aggregated packet in generic
2010 * receive offload (GRO)
2011 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2012 * receive offload (GRO), for IPv4.
2014 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2015 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2016 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2017 * @dev_registered_tracker: tracker for reference held while
2019 * @offload_xstats_l3: L3 HW stats for this netdevice.
2021 * @devlink_port: Pointer to related devlink port structure.
2022 * Assigned by a driver before netdev registration using
2023 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2024 * during the time netdevice is registered.
2026 * FIXME: cleanup struct net_device such that network protocol info
2031 char name[IFNAMSIZ];
2032 struct netdev_name_node *name_node;
2033 struct dev_ifalias __rcu *ifalias;
2035 * I/O specific fields
2036 * FIXME: Merge these and struct ifmap into one
2038 unsigned long mem_end;
2039 unsigned long mem_start;
2040 unsigned long base_addr;
2043 * Some hardware also needs these fields (state,dev_list,
2044 * napi_list,unreg_list,close_list) but they are not
2045 * part of the usual set specified in Space.c.
2048 unsigned long state;
2050 struct list_head dev_list;
2051 struct list_head napi_list;
2052 struct list_head unreg_list;
2053 struct list_head close_list;
2054 struct list_head ptype_all;
2055 struct list_head ptype_specific;
2058 struct list_head upper;
2059 struct list_head lower;
2062 /* Read-mostly cache-line for fast-path access */
2064 xdp_features_t xdp_features;
2065 unsigned long long priv_flags;
2066 const struct net_device_ops *netdev_ops;
2067 const struct xdp_metadata_ops *xdp_metadata_ops;
2069 unsigned short gflags;
2070 unsigned short hard_header_len;
2072 /* Note : dev->mtu is often read without holding a lock.
2073 * Writers usually hold RTNL.
2074 * It is recommended to use READ_ONCE() to annotate the reads,
2075 * and to use WRITE_ONCE() to annotate the writes.
2078 unsigned short needed_headroom;
2079 unsigned short needed_tailroom;
2081 netdev_features_t features;
2082 netdev_features_t hw_features;
2083 netdev_features_t wanted_features;
2084 netdev_features_t vlan_features;
2085 netdev_features_t hw_enc_features;
2086 netdev_features_t mpls_features;
2087 netdev_features_t gso_partial_features;
2089 unsigned int min_mtu;
2090 unsigned int max_mtu;
2091 unsigned short type;
2092 unsigned char min_header_len;
2093 unsigned char name_assign_type;
2097 struct net_device_stats stats; /* not used by modern drivers */
2099 struct net_device_core_stats __percpu *core_stats;
2101 /* Stats to monitor link on/off, flapping */
2102 atomic_t carrier_up_count;
2103 atomic_t carrier_down_count;
2105 #ifdef CONFIG_WIRELESS_EXT
2106 const struct iw_handler_def *wireless_handlers;
2107 struct iw_public_data *wireless_data;
2109 const struct ethtool_ops *ethtool_ops;
2110 #ifdef CONFIG_NET_L3_MASTER_DEV
2111 const struct l3mdev_ops *l3mdev_ops;
2113 #if IS_ENABLED(CONFIG_IPV6)
2114 const struct ndisc_ops *ndisc_ops;
2117 #ifdef CONFIG_XFRM_OFFLOAD
2118 const struct xfrmdev_ops *xfrmdev_ops;
2121 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2122 const struct tlsdev_ops *tlsdev_ops;
2125 const struct header_ops *header_ops;
2127 unsigned char operstate;
2128 unsigned char link_mode;
2130 unsigned char if_port;
2133 /* Interface address info. */
2134 unsigned char perm_addr[MAX_ADDR_LEN];
2135 unsigned char addr_assign_type;
2136 unsigned char addr_len;
2137 unsigned char upper_level;
2138 unsigned char lower_level;
2140 unsigned short neigh_priv_len;
2141 unsigned short dev_id;
2142 unsigned short dev_port;
2143 unsigned short padded;
2145 spinlock_t addr_list_lock;
2148 struct netdev_hw_addr_list uc;
2149 struct netdev_hw_addr_list mc;
2150 struct netdev_hw_addr_list dev_addrs;
2153 struct kset *queues_kset;
2155 #ifdef CONFIG_LOCKDEP
2156 struct list_head unlink_list;
2158 unsigned int promiscuity;
2159 unsigned int allmulti;
2161 #ifdef CONFIG_LOCKDEP
2162 unsigned char nested_level;
2166 /* Protocol-specific pointers */
2168 struct in_device __rcu *ip_ptr;
2169 struct inet6_dev __rcu *ip6_ptr;
2170 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2171 struct vlan_info __rcu *vlan_info;
2173 #if IS_ENABLED(CONFIG_NET_DSA)
2174 struct dsa_port *dsa_ptr;
2176 #if IS_ENABLED(CONFIG_TIPC)
2177 struct tipc_bearer __rcu *tipc_ptr;
2179 #if IS_ENABLED(CONFIG_ATALK)
2182 #if IS_ENABLED(CONFIG_AX25)
2185 #if IS_ENABLED(CONFIG_CFG80211)
2186 struct wireless_dev *ieee80211_ptr;
2188 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2189 struct wpan_dev *ieee802154_ptr;
2191 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2192 struct mpls_dev __rcu *mpls_ptr;
2194 #if IS_ENABLED(CONFIG_MCTP)
2195 struct mctp_dev __rcu *mctp_ptr;
2199 * Cache lines mostly used on receive path (including eth_type_trans())
2201 /* Interface address info used in eth_type_trans() */
2202 const unsigned char *dev_addr;
2204 struct netdev_rx_queue *_rx;
2205 unsigned int num_rx_queues;
2206 unsigned int real_num_rx_queues;
2208 struct bpf_prog __rcu *xdp_prog;
2209 unsigned long gro_flush_timeout;
2210 int napi_defer_hard_irqs;
2211 #define GRO_LEGACY_MAX_SIZE 65536u
2212 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2213 * and shinfo->gso_segs is a 16bit field.
2215 #define GRO_MAX_SIZE (8 * 65535u)
2216 unsigned int gro_max_size;
2217 unsigned int gro_ipv4_max_size;
2218 rx_handler_func_t __rcu *rx_handler;
2219 void __rcu *rx_handler_data;
2221 #ifdef CONFIG_NET_CLS_ACT
2222 struct mini_Qdisc __rcu *miniq_ingress;
2224 struct netdev_queue __rcu *ingress_queue;
2225 #ifdef CONFIG_NETFILTER_INGRESS
2226 struct nf_hook_entries __rcu *nf_hooks_ingress;
2229 unsigned char broadcast[MAX_ADDR_LEN];
2230 #ifdef CONFIG_RFS_ACCEL
2231 struct cpu_rmap *rx_cpu_rmap;
2233 struct hlist_node index_hlist;
2236 * Cache lines mostly used on transmit path
2238 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2239 unsigned int num_tx_queues;
2240 unsigned int real_num_tx_queues;
2241 struct Qdisc __rcu *qdisc;
2242 unsigned int tx_queue_len;
2243 spinlock_t tx_global_lock;
2245 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2248 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2250 #ifdef CONFIG_NET_CLS_ACT
2251 struct mini_Qdisc __rcu *miniq_egress;
2253 #ifdef CONFIG_NETFILTER_EGRESS
2254 struct nf_hook_entries __rcu *nf_hooks_egress;
2257 #ifdef CONFIG_NET_SCHED
2258 DECLARE_HASHTABLE (qdisc_hash, 4);
2260 /* These may be needed for future network-power-down code. */
2261 struct timer_list watchdog_timer;
2264 u32 proto_down_reason;
2266 struct list_head todo_list;
2268 #ifdef CONFIG_PCPU_DEV_REFCNT
2269 int __percpu *pcpu_refcnt;
2271 refcount_t dev_refcnt;
2273 struct ref_tracker_dir refcnt_tracker;
2275 struct list_head link_watch_list;
2277 enum { NETREG_UNINITIALIZED=0,
2278 NETREG_REGISTERED, /* completed register_netdevice */
2279 NETREG_UNREGISTERING, /* called unregister_netdevice */
2280 NETREG_UNREGISTERED, /* completed unregister todo */
2281 NETREG_RELEASED, /* called free_netdev */
2282 NETREG_DUMMY, /* dummy device for NAPI poll */
2288 RTNL_LINK_INITIALIZED,
2289 RTNL_LINK_INITIALIZING,
2290 } rtnl_link_state:16;
2292 bool needs_free_netdev;
2293 void (*priv_destructor)(struct net_device *dev);
2295 #ifdef CONFIG_NETPOLL
2296 struct netpoll_info __rcu *npinfo;
2299 possible_net_t nd_net;
2301 /* mid-layer private */
2303 enum netdev_ml_priv_type ml_priv_type;
2306 struct pcpu_lstats __percpu *lstats;
2307 struct pcpu_sw_netstats __percpu *tstats;
2308 struct pcpu_dstats __percpu *dstats;
2311 #if IS_ENABLED(CONFIG_GARP)
2312 struct garp_port __rcu *garp_port;
2314 #if IS_ENABLED(CONFIG_MRP)
2315 struct mrp_port __rcu *mrp_port;
2317 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2318 struct dm_hw_stat_delta __rcu *dm_private;
2321 const struct attribute_group *sysfs_groups[4];
2322 const struct attribute_group *sysfs_rx_queue_group;
2324 const struct rtnl_link_ops *rtnl_link_ops;
2326 /* for setting kernel sock attribute on TCP connection setup */
2327 #define GSO_MAX_SEGS 65535u
2328 #define GSO_LEGACY_MAX_SIZE 65536u
2329 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2330 * and shinfo->gso_segs is a 16bit field.
2332 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2334 unsigned int gso_max_size;
2335 #define TSO_LEGACY_MAX_SIZE 65536
2336 #define TSO_MAX_SIZE UINT_MAX
2337 unsigned int tso_max_size;
2339 #define TSO_MAX_SEGS U16_MAX
2341 unsigned int gso_ipv4_max_size;
2344 const struct dcbnl_rtnl_ops *dcbnl_ops;
2347 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2348 u8 prio_tc_map[TC_BITMASK + 1];
2350 #if IS_ENABLED(CONFIG_FCOE)
2351 unsigned int fcoe_ddp_xid;
2353 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2354 struct netprio_map __rcu *priomap;
2356 struct phy_device *phydev;
2357 struct sfp_bus *sfp_bus;
2358 struct lock_class_key *qdisc_tx_busylock;
2360 unsigned wol_enabled:1;
2361 unsigned threaded:1;
2363 struct list_head net_notifier_list;
2365 #if IS_ENABLED(CONFIG_MACSEC)
2366 /* MACsec management functions */
2367 const struct macsec_ops *macsec_ops;
2369 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2370 struct udp_tunnel_nic *udp_tunnel_nic;
2372 /* protected by rtnl_lock */
2373 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2375 u8 dev_addr_shadow[MAX_ADDR_LEN];
2376 netdevice_tracker linkwatch_dev_tracker;
2377 netdevice_tracker watchdog_dev_tracker;
2378 netdevice_tracker dev_registered_tracker;
2379 struct rtnl_hw_stats64 *offload_xstats_l3;
2381 struct devlink_port *devlink_port;
2383 #define to_net_dev(d) container_of(d, struct net_device, dev)
2386 * Driver should use this to assign devlink port instance to a netdevice
2387 * before it registers the netdevice. Therefore devlink_port is static
2388 * during the netdev lifetime after it is registered.
2390 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2392 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2393 ((dev)->devlink_port = (port)); \
2396 static inline bool netif_elide_gro(const struct net_device *dev)
2398 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2403 #define NETDEV_ALIGN 32
2406 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2408 return dev->prio_tc_map[prio & TC_BITMASK];
2412 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2414 if (tc >= dev->num_tc)
2417 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2421 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2422 void netdev_reset_tc(struct net_device *dev);
2423 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2424 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2427 int netdev_get_num_tc(struct net_device *dev)
2432 static inline void net_prefetch(void *p)
2435 #if L1_CACHE_BYTES < 128
2436 prefetch((u8 *)p + L1_CACHE_BYTES);
2440 static inline void net_prefetchw(void *p)
2443 #if L1_CACHE_BYTES < 128
2444 prefetchw((u8 *)p + L1_CACHE_BYTES);
2448 void netdev_unbind_sb_channel(struct net_device *dev,
2449 struct net_device *sb_dev);
2450 int netdev_bind_sb_channel_queue(struct net_device *dev,
2451 struct net_device *sb_dev,
2452 u8 tc, u16 count, u16 offset);
2453 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2454 static inline int netdev_get_sb_channel(struct net_device *dev)
2456 return max_t(int, -dev->num_tc, 0);
2460 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2463 return &dev->_tx[index];
2466 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2467 const struct sk_buff *skb)
2469 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2472 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2473 void (*f)(struct net_device *,
2474 struct netdev_queue *,
2480 for (i = 0; i < dev->num_tx_queues; i++)
2481 f(dev, &dev->_tx[i], arg);
2484 #define netdev_lockdep_set_classes(dev) \
2486 static struct lock_class_key qdisc_tx_busylock_key; \
2487 static struct lock_class_key qdisc_xmit_lock_key; \
2488 static struct lock_class_key dev_addr_list_lock_key; \
2491 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2492 lockdep_set_class(&(dev)->addr_list_lock, \
2493 &dev_addr_list_lock_key); \
2494 for (i = 0; i < (dev)->num_tx_queues; i++) \
2495 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2496 &qdisc_xmit_lock_key); \
2499 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2500 struct net_device *sb_dev);
2501 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2502 struct sk_buff *skb,
2503 struct net_device *sb_dev);
2505 /* returns the headroom that the master device needs to take in account
2506 * when forwarding to this dev
2508 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2510 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2513 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2515 if (dev->netdev_ops->ndo_set_rx_headroom)
2516 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2519 /* set the device rx headroom to the dev's default */
2520 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2522 netdev_set_rx_headroom(dev, -1);
2525 static inline void *netdev_get_ml_priv(struct net_device *dev,
2526 enum netdev_ml_priv_type type)
2528 if (dev->ml_priv_type != type)
2531 return dev->ml_priv;
2534 static inline void netdev_set_ml_priv(struct net_device *dev,
2536 enum netdev_ml_priv_type type)
2538 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2539 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2540 dev->ml_priv_type, type);
2541 WARN(!dev->ml_priv_type && dev->ml_priv,
2542 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2544 dev->ml_priv = ml_priv;
2545 dev->ml_priv_type = type;
2549 * Net namespace inlines
2552 struct net *dev_net(const struct net_device *dev)
2554 return read_pnet(&dev->nd_net);
2558 void dev_net_set(struct net_device *dev, struct net *net)
2560 write_pnet(&dev->nd_net, net);
2564 * netdev_priv - access network device private data
2565 * @dev: network device
2567 * Get network device private data
2569 static inline void *netdev_priv(const struct net_device *dev)
2571 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2574 /* Set the sysfs physical device reference for the network logical device
2575 * if set prior to registration will cause a symlink during initialization.
2577 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2579 /* Set the sysfs device type for the network logical device to allow
2580 * fine-grained identification of different network device types. For
2581 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2583 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2585 /* Default NAPI poll() weight
2586 * Device drivers are strongly advised to not use bigger value
2588 #define NAPI_POLL_WEIGHT 64
2590 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2591 int (*poll)(struct napi_struct *, int), int weight);
2594 * netif_napi_add() - initialize a NAPI context
2595 * @dev: network device
2596 * @napi: NAPI context
2597 * @poll: polling function
2599 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2600 * *any* of the other NAPI-related functions.
2603 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2604 int (*poll)(struct napi_struct *, int))
2606 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2610 netif_napi_add_tx_weight(struct net_device *dev,
2611 struct napi_struct *napi,
2612 int (*poll)(struct napi_struct *, int),
2615 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2616 netif_napi_add_weight(dev, napi, poll, weight);
2620 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2621 * @dev: network device
2622 * @napi: NAPI context
2623 * @poll: polling function
2625 * This variant of netif_napi_add() should be used from drivers using NAPI
2626 * to exclusively poll a TX queue.
2627 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2629 static inline void netif_napi_add_tx(struct net_device *dev,
2630 struct napi_struct *napi,
2631 int (*poll)(struct napi_struct *, int))
2633 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2637 * __netif_napi_del - remove a NAPI context
2638 * @napi: NAPI context
2640 * Warning: caller must observe RCU grace period before freeing memory
2641 * containing @napi. Drivers might want to call this helper to combine
2642 * all the needed RCU grace periods into a single one.
2644 void __netif_napi_del(struct napi_struct *napi);
2647 * netif_napi_del - remove a NAPI context
2648 * @napi: NAPI context
2650 * netif_napi_del() removes a NAPI context from the network device NAPI list
2652 static inline void netif_napi_del(struct napi_struct *napi)
2654 __netif_napi_del(napi);
2658 struct packet_type {
2659 __be16 type; /* This is really htons(ether_type). */
2660 bool ignore_outgoing;
2661 struct net_device *dev; /* NULL is wildcarded here */
2662 netdevice_tracker dev_tracker;
2663 int (*func) (struct sk_buff *,
2664 struct net_device *,
2665 struct packet_type *,
2666 struct net_device *);
2667 void (*list_func) (struct list_head *,
2668 struct packet_type *,
2669 struct net_device *);
2670 bool (*id_match)(struct packet_type *ptype,
2672 struct net *af_packet_net;
2673 void *af_packet_priv;
2674 struct list_head list;
2677 struct offload_callbacks {
2678 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2679 netdev_features_t features);
2680 struct sk_buff *(*gro_receive)(struct list_head *head,
2681 struct sk_buff *skb);
2682 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2685 struct packet_offload {
2686 __be16 type; /* This is really htons(ether_type). */
2688 struct offload_callbacks callbacks;
2689 struct list_head list;
2692 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2693 struct pcpu_sw_netstats {
2694 u64_stats_t rx_packets;
2695 u64_stats_t rx_bytes;
2696 u64_stats_t tx_packets;
2697 u64_stats_t tx_bytes;
2698 struct u64_stats_sync syncp;
2699 } __aligned(4 * sizeof(u64));
2701 struct pcpu_lstats {
2702 u64_stats_t packets;
2704 struct u64_stats_sync syncp;
2705 } __aligned(2 * sizeof(u64));
2707 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2709 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2711 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2713 u64_stats_update_begin(&tstats->syncp);
2714 u64_stats_add(&tstats->rx_bytes, len);
2715 u64_stats_inc(&tstats->rx_packets);
2716 u64_stats_update_end(&tstats->syncp);
2719 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2720 unsigned int packets,
2723 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2725 u64_stats_update_begin(&tstats->syncp);
2726 u64_stats_add(&tstats->tx_bytes, len);
2727 u64_stats_add(&tstats->tx_packets, packets);
2728 u64_stats_update_end(&tstats->syncp);
2731 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2733 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2735 u64_stats_update_begin(&lstats->syncp);
2736 u64_stats_add(&lstats->bytes, len);
2737 u64_stats_inc(&lstats->packets);
2738 u64_stats_update_end(&lstats->syncp);
2741 #define __netdev_alloc_pcpu_stats(type, gfp) \
2743 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2746 for_each_possible_cpu(__cpu) { \
2747 typeof(type) *stat; \
2748 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2749 u64_stats_init(&stat->syncp); \
2755 #define netdev_alloc_pcpu_stats(type) \
2756 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2758 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2760 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2763 for_each_possible_cpu(__cpu) { \
2764 typeof(type) *stat; \
2765 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2766 u64_stats_init(&stat->syncp); \
2772 enum netdev_lag_tx_type {
2773 NETDEV_LAG_TX_TYPE_UNKNOWN,
2774 NETDEV_LAG_TX_TYPE_RANDOM,
2775 NETDEV_LAG_TX_TYPE_BROADCAST,
2776 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2777 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2778 NETDEV_LAG_TX_TYPE_HASH,
2781 enum netdev_lag_hash {
2782 NETDEV_LAG_HASH_NONE,
2784 NETDEV_LAG_HASH_L34,
2785 NETDEV_LAG_HASH_L23,
2786 NETDEV_LAG_HASH_E23,
2787 NETDEV_LAG_HASH_E34,
2788 NETDEV_LAG_HASH_VLAN_SRCMAC,
2789 NETDEV_LAG_HASH_UNKNOWN,
2792 struct netdev_lag_upper_info {
2793 enum netdev_lag_tx_type tx_type;
2794 enum netdev_lag_hash hash_type;
2797 struct netdev_lag_lower_state_info {
2802 #include <linux/notifier.h>
2804 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2805 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2809 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2811 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2812 detected a hardware crash and restarted
2813 - we can use this eg to kick tcp sessions
2815 NETDEV_CHANGE, /* Notify device state change */
2818 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2819 NETDEV_CHANGEADDR, /* notify after the address change */
2820 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2824 NETDEV_BONDING_FAILOVER,
2826 NETDEV_PRE_TYPE_CHANGE,
2827 NETDEV_POST_TYPE_CHANGE,
2831 NETDEV_NOTIFY_PEERS,
2835 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2836 NETDEV_CHANGEINFODATA,
2837 NETDEV_BONDING_INFO,
2838 NETDEV_PRECHANGEUPPER,
2839 NETDEV_CHANGELOWERSTATE,
2840 NETDEV_UDP_TUNNEL_PUSH_INFO,
2841 NETDEV_UDP_TUNNEL_DROP_INFO,
2842 NETDEV_CHANGE_TX_QUEUE_LEN,
2843 NETDEV_CVLAN_FILTER_PUSH_INFO,
2844 NETDEV_CVLAN_FILTER_DROP_INFO,
2845 NETDEV_SVLAN_FILTER_PUSH_INFO,
2846 NETDEV_SVLAN_FILTER_DROP_INFO,
2847 NETDEV_OFFLOAD_XSTATS_ENABLE,
2848 NETDEV_OFFLOAD_XSTATS_DISABLE,
2849 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2850 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2851 NETDEV_XDP_FEAT_CHANGE,
2853 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2855 int register_netdevice_notifier(struct notifier_block *nb);
2856 int unregister_netdevice_notifier(struct notifier_block *nb);
2857 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2858 int unregister_netdevice_notifier_net(struct net *net,
2859 struct notifier_block *nb);
2860 int register_netdevice_notifier_dev_net(struct net_device *dev,
2861 struct notifier_block *nb,
2862 struct netdev_net_notifier *nn);
2863 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2864 struct notifier_block *nb,
2865 struct netdev_net_notifier *nn);
2867 struct netdev_notifier_info {
2868 struct net_device *dev;
2869 struct netlink_ext_ack *extack;
2872 struct netdev_notifier_info_ext {
2873 struct netdev_notifier_info info; /* must be first */
2879 struct netdev_notifier_change_info {
2880 struct netdev_notifier_info info; /* must be first */
2881 unsigned int flags_changed;
2884 struct netdev_notifier_changeupper_info {
2885 struct netdev_notifier_info info; /* must be first */
2886 struct net_device *upper_dev; /* new upper dev */
2887 bool master; /* is upper dev master */
2888 bool linking; /* is the notification for link or unlink */
2889 void *upper_info; /* upper dev info */
2892 struct netdev_notifier_changelowerstate_info {
2893 struct netdev_notifier_info info; /* must be first */
2894 void *lower_state_info; /* is lower dev state */
2897 struct netdev_notifier_pre_changeaddr_info {
2898 struct netdev_notifier_info info; /* must be first */
2899 const unsigned char *dev_addr;
2902 enum netdev_offload_xstats_type {
2903 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2906 struct netdev_notifier_offload_xstats_info {
2907 struct netdev_notifier_info info; /* must be first */
2908 enum netdev_offload_xstats_type type;
2911 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2912 struct netdev_notifier_offload_xstats_rd *report_delta;
2913 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2914 struct netdev_notifier_offload_xstats_ru *report_used;
2918 int netdev_offload_xstats_enable(struct net_device *dev,
2919 enum netdev_offload_xstats_type type,
2920 struct netlink_ext_ack *extack);
2921 int netdev_offload_xstats_disable(struct net_device *dev,
2922 enum netdev_offload_xstats_type type);
2923 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2924 enum netdev_offload_xstats_type type);
2925 int netdev_offload_xstats_get(struct net_device *dev,
2926 enum netdev_offload_xstats_type type,
2927 struct rtnl_hw_stats64 *stats, bool *used,
2928 struct netlink_ext_ack *extack);
2930 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2931 const struct rtnl_hw_stats64 *stats);
2933 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2934 void netdev_offload_xstats_push_delta(struct net_device *dev,
2935 enum netdev_offload_xstats_type type,
2936 const struct rtnl_hw_stats64 *stats);
2938 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2939 struct net_device *dev)
2942 info->extack = NULL;
2945 static inline struct net_device *
2946 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2951 static inline struct netlink_ext_ack *
2952 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2954 return info->extack;
2957 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2960 extern rwlock_t dev_base_lock; /* Device list lock */
2962 #define for_each_netdev(net, d) \
2963 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2964 #define for_each_netdev_reverse(net, d) \
2965 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2966 #define for_each_netdev_rcu(net, d) \
2967 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2968 #define for_each_netdev_safe(net, d, n) \
2969 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2970 #define for_each_netdev_continue(net, d) \
2971 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2972 #define for_each_netdev_continue_reverse(net, d) \
2973 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2975 #define for_each_netdev_continue_rcu(net, d) \
2976 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2977 #define for_each_netdev_in_bond_rcu(bond, slave) \
2978 for_each_netdev_rcu(&init_net, slave) \
2979 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2980 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2982 static inline struct net_device *next_net_device(struct net_device *dev)
2984 struct list_head *lh;
2988 lh = dev->dev_list.next;
2989 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2992 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2994 struct list_head *lh;
2998 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2999 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3002 static inline struct net_device *first_net_device(struct net *net)
3004 return list_empty(&net->dev_base_head) ? NULL :
3005 net_device_entry(net->dev_base_head.next);
3008 static inline struct net_device *first_net_device_rcu(struct net *net)
3010 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3012 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3015 int netdev_boot_setup_check(struct net_device *dev);
3016 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3017 const char *hwaddr);
3018 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3019 void dev_add_pack(struct packet_type *pt);
3020 void dev_remove_pack(struct packet_type *pt);
3021 void __dev_remove_pack(struct packet_type *pt);
3022 void dev_add_offload(struct packet_offload *po);
3023 void dev_remove_offload(struct packet_offload *po);
3025 int dev_get_iflink(const struct net_device *dev);
3026 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3027 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3028 struct net_device_path_stack *stack);
3029 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3030 unsigned short mask);
3031 struct net_device *dev_get_by_name(struct net *net, const char *name);
3032 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3033 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3034 bool netdev_name_in_use(struct net *net, const char *name);
3035 int dev_alloc_name(struct net_device *dev, const char *name);
3036 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3037 void dev_close(struct net_device *dev);
3038 void dev_close_many(struct list_head *head, bool unlink);
3039 void dev_disable_lro(struct net_device *dev);
3040 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3041 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3042 struct net_device *sb_dev);
3043 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3044 struct net_device *sb_dev);
3046 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3047 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3049 static inline int dev_queue_xmit(struct sk_buff *skb)
3051 return __dev_queue_xmit(skb, NULL);
3054 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3055 struct net_device *sb_dev)
3057 return __dev_queue_xmit(skb, sb_dev);
3060 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3064 ret = __dev_direct_xmit(skb, queue_id);
3065 if (!dev_xmit_complete(ret))
3070 int register_netdevice(struct net_device *dev);
3071 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3072 void unregister_netdevice_many(struct list_head *head);
3073 static inline void unregister_netdevice(struct net_device *dev)
3075 unregister_netdevice_queue(dev, NULL);
3078 int netdev_refcnt_read(const struct net_device *dev);
3079 void free_netdev(struct net_device *dev);
3080 void netdev_freemem(struct net_device *dev);
3081 int init_dummy_netdev(struct net_device *dev);
3083 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3084 struct sk_buff *skb,
3086 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3088 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3089 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3090 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3091 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3092 int dev_restart(struct net_device *dev);
3095 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3096 unsigned short type,
3097 const void *daddr, const void *saddr,
3100 if (!dev->header_ops || !dev->header_ops->create)
3103 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3106 static inline int dev_parse_header(const struct sk_buff *skb,
3107 unsigned char *haddr)
3109 const struct net_device *dev = skb->dev;
3111 if (!dev->header_ops || !dev->header_ops->parse)
3113 return dev->header_ops->parse(skb, haddr);
3116 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3118 const struct net_device *dev = skb->dev;
3120 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3122 return dev->header_ops->parse_protocol(skb);
3125 /* ll_header must have at least hard_header_len allocated */
3126 static inline bool dev_validate_header(const struct net_device *dev,
3127 char *ll_header, int len)
3129 if (likely(len >= dev->hard_header_len))
3131 if (len < dev->min_header_len)
3134 if (capable(CAP_SYS_RAWIO)) {
3135 memset(ll_header + len, 0, dev->hard_header_len - len);
3139 if (dev->header_ops && dev->header_ops->validate)
3140 return dev->header_ops->validate(ll_header, len);
3145 static inline bool dev_has_header(const struct net_device *dev)
3147 return dev->header_ops && dev->header_ops->create;
3151 * Incoming packets are placed on per-CPU queues
3153 struct softnet_data {
3154 struct list_head poll_list;
3155 struct sk_buff_head process_queue;
3158 unsigned int processed;
3159 unsigned int time_squeeze;
3161 struct softnet_data *rps_ipi_list;
3163 #ifdef CONFIG_NET_FLOW_LIMIT
3164 struct sd_flow_limit __rcu *flow_limit;
3166 struct Qdisc *output_queue;
3167 struct Qdisc **output_queue_tailp;
3168 struct sk_buff *completion_queue;
3169 #ifdef CONFIG_XFRM_OFFLOAD
3170 struct sk_buff_head xfrm_backlog;
3172 /* written and read only by owning cpu: */
3176 #ifdef CONFIG_NET_EGRESS
3181 /* input_queue_head should be written by cpu owning this struct,
3182 * and only read by other cpus. Worth using a cache line.
3184 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3186 /* Elements below can be accessed between CPUs for RPS/RFS */
3187 call_single_data_t csd ____cacheline_aligned_in_smp;
3188 struct softnet_data *rps_ipi_next;
3190 unsigned int input_queue_tail;
3192 unsigned int received_rps;
3193 unsigned int dropped;
3194 struct sk_buff_head input_pkt_queue;
3195 struct napi_struct backlog;
3197 /* Another possibly contended cache line */
3198 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3200 int defer_ipi_scheduled;
3201 struct sk_buff *defer_list;
3202 call_single_data_t defer_csd;
3205 static inline void input_queue_head_incr(struct softnet_data *sd)
3208 sd->input_queue_head++;
3212 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3213 unsigned int *qtail)
3216 *qtail = ++sd->input_queue_tail;
3220 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3222 static inline int dev_recursion_level(void)
3224 return this_cpu_read(softnet_data.xmit.recursion);
3227 #define XMIT_RECURSION_LIMIT 8
3228 static inline bool dev_xmit_recursion(void)
3230 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3231 XMIT_RECURSION_LIMIT);
3234 static inline void dev_xmit_recursion_inc(void)
3236 __this_cpu_inc(softnet_data.xmit.recursion);
3239 static inline void dev_xmit_recursion_dec(void)
3241 __this_cpu_dec(softnet_data.xmit.recursion);
3244 void __netif_schedule(struct Qdisc *q);
3245 void netif_schedule_queue(struct netdev_queue *txq);
3247 static inline void netif_tx_schedule_all(struct net_device *dev)
3251 for (i = 0; i < dev->num_tx_queues; i++)
3252 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3255 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3257 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3261 * netif_start_queue - allow transmit
3262 * @dev: network device
3264 * Allow upper layers to call the device hard_start_xmit routine.
3266 static inline void netif_start_queue(struct net_device *dev)
3268 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3271 static inline void netif_tx_start_all_queues(struct net_device *dev)
3275 for (i = 0; i < dev->num_tx_queues; i++) {
3276 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3277 netif_tx_start_queue(txq);
3281 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3284 * netif_wake_queue - restart transmit
3285 * @dev: network device
3287 * Allow upper layers to call the device hard_start_xmit routine.
3288 * Used for flow control when transmit resources are available.
3290 static inline void netif_wake_queue(struct net_device *dev)
3292 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3295 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3299 for (i = 0; i < dev->num_tx_queues; i++) {
3300 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3301 netif_tx_wake_queue(txq);
3305 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3307 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3311 * netif_stop_queue - stop transmitted packets
3312 * @dev: network device
3314 * Stop upper layers calling the device hard_start_xmit routine.
3315 * Used for flow control when transmit resources are unavailable.
3317 static inline void netif_stop_queue(struct net_device *dev)
3319 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3322 void netif_tx_stop_all_queues(struct net_device *dev);
3324 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3326 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3330 * netif_queue_stopped - test if transmit queue is flowblocked
3331 * @dev: network device
3333 * Test if transmit queue on device is currently unable to send.
3335 static inline bool netif_queue_stopped(const struct net_device *dev)
3337 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3340 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3342 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3346 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3348 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3352 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3354 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3358 * netdev_queue_set_dql_min_limit - set dql minimum limit
3359 * @dev_queue: pointer to transmit queue
3360 * @min_limit: dql minimum limit
3362 * Forces xmit_more() to return true until the minimum threshold
3363 * defined by @min_limit is reached (or until the tx queue is
3364 * empty). Warning: to be use with care, misuse will impact the
3367 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3368 unsigned int min_limit)
3371 dev_queue->dql.min_limit = min_limit;
3376 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3377 * @dev_queue: pointer to transmit queue
3379 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3380 * to give appropriate hint to the CPU.
3382 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3385 prefetchw(&dev_queue->dql.num_queued);
3390 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3391 * @dev_queue: pointer to transmit queue
3393 * BQL enabled drivers might use this helper in their TX completion path,
3394 * to give appropriate hint to the CPU.
3396 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3399 prefetchw(&dev_queue->dql.limit);
3404 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3405 * @dev_queue: network device queue
3406 * @bytes: number of bytes queued to the device queue
3408 * Report the number of bytes queued for sending/completion to the network
3409 * device hardware queue. @bytes should be a good approximation and should
3410 * exactly match netdev_completed_queue() @bytes.
3411 * This is typically called once per packet, from ndo_start_xmit().
3413 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3417 dql_queued(&dev_queue->dql, bytes);
3419 if (likely(dql_avail(&dev_queue->dql) >= 0))
3422 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3425 * The XOFF flag must be set before checking the dql_avail below,
3426 * because in netdev_tx_completed_queue we update the dql_completed
3427 * before checking the XOFF flag.
3431 /* check again in case another CPU has just made room avail */
3432 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3433 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3437 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3438 * that they should not test BQL status themselves.
3439 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3441 * Returns true if the doorbell must be used to kick the NIC.
3443 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3449 dql_queued(&dev_queue->dql, bytes);
3451 return netif_tx_queue_stopped(dev_queue);
3453 netdev_tx_sent_queue(dev_queue, bytes);
3458 * netdev_sent_queue - report the number of bytes queued to hardware
3459 * @dev: network device
3460 * @bytes: number of bytes queued to the hardware device queue
3462 * Report the number of bytes queued for sending/completion to the network
3463 * device hardware queue#0. @bytes should be a good approximation and should
3464 * exactly match netdev_completed_queue() @bytes.
3465 * This is typically called once per packet, from ndo_start_xmit().
3467 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3469 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3472 static inline bool __netdev_sent_queue(struct net_device *dev,
3476 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3481 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3482 * @dev_queue: network device queue
3483 * @pkts: number of packets (currently ignored)
3484 * @bytes: number of bytes dequeued from the device queue
3486 * Must be called at most once per TX completion round (and not per
3487 * individual packet), so that BQL can adjust its limits appropriately.
3489 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3490 unsigned int pkts, unsigned int bytes)
3493 if (unlikely(!bytes))
3496 dql_completed(&dev_queue->dql, bytes);
3499 * Without the memory barrier there is a small possiblity that
3500 * netdev_tx_sent_queue will miss the update and cause the queue to
3501 * be stopped forever
3505 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3508 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3509 netif_schedule_queue(dev_queue);
3514 * netdev_completed_queue - report bytes and packets completed by device
3515 * @dev: network device
3516 * @pkts: actual number of packets sent over the medium
3517 * @bytes: actual number of bytes sent over the medium
3519 * Report the number of bytes and packets transmitted by the network device
3520 * hardware queue over the physical medium, @bytes must exactly match the
3521 * @bytes amount passed to netdev_sent_queue()
3523 static inline void netdev_completed_queue(struct net_device *dev,
3524 unsigned int pkts, unsigned int bytes)
3526 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3529 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3532 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3538 * netdev_reset_queue - reset the packets and bytes count of a network device
3539 * @dev_queue: network device
3541 * Reset the bytes and packet count of a network device and clear the
3542 * software flow control OFF bit for this network device
3544 static inline void netdev_reset_queue(struct net_device *dev_queue)
3546 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3550 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3551 * @dev: network device
3552 * @queue_index: given tx queue index
3554 * Returns 0 if given tx queue index >= number of device tx queues,
3555 * otherwise returns the originally passed tx queue index.
3557 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3559 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3560 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3561 dev->name, queue_index,
3562 dev->real_num_tx_queues);
3570 * netif_running - test if up
3571 * @dev: network device
3573 * Test if the device has been brought up.
3575 static inline bool netif_running(const struct net_device *dev)
3577 return test_bit(__LINK_STATE_START, &dev->state);
3581 * Routines to manage the subqueues on a device. We only need start,
3582 * stop, and a check if it's stopped. All other device management is
3583 * done at the overall netdevice level.
3584 * Also test the device if we're multiqueue.
3588 * netif_start_subqueue - allow sending packets on subqueue
3589 * @dev: network device
3590 * @queue_index: sub queue index
3592 * Start individual transmit queue of a device with multiple transmit queues.
3594 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3596 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3598 netif_tx_start_queue(txq);
3602 * netif_stop_subqueue - stop sending packets on subqueue
3603 * @dev: network device
3604 * @queue_index: sub queue index
3606 * Stop individual transmit queue of a device with multiple transmit queues.
3608 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3611 netif_tx_stop_queue(txq);
3615 * __netif_subqueue_stopped - test status of subqueue
3616 * @dev: network device
3617 * @queue_index: sub queue index
3619 * Check individual transmit queue of a device with multiple transmit queues.
3621 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3624 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3626 return netif_tx_queue_stopped(txq);
3630 * netif_subqueue_stopped - test status of subqueue
3631 * @dev: network device
3632 * @skb: sub queue buffer pointer
3634 * Check individual transmit queue of a device with multiple transmit queues.
3636 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3637 struct sk_buff *skb)
3639 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3643 * netif_wake_subqueue - allow sending packets on subqueue
3644 * @dev: network device
3645 * @queue_index: sub queue index
3647 * Resume individual transmit queue of a device with multiple transmit queues.
3649 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3651 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3653 netif_tx_wake_queue(txq);
3657 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3659 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3660 u16 index, enum xps_map_type type);
3663 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3664 * @j: CPU/Rx queue index
3665 * @mask: bitmask of all cpus/rx queues
3666 * @nr_bits: number of bits in the bitmask
3668 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3670 static inline bool netif_attr_test_mask(unsigned long j,
3671 const unsigned long *mask,
3672 unsigned int nr_bits)
3674 cpu_max_bits_warn(j, nr_bits);
3675 return test_bit(j, mask);
3679 * netif_attr_test_online - Test for online CPU/Rx queue
3680 * @j: CPU/Rx queue index
3681 * @online_mask: bitmask for CPUs/Rx queues that are online
3682 * @nr_bits: number of bits in the bitmask
3684 * Returns true if a CPU/Rx queue is online.
3686 static inline bool netif_attr_test_online(unsigned long j,
3687 const unsigned long *online_mask,
3688 unsigned int nr_bits)
3690 cpu_max_bits_warn(j, nr_bits);
3693 return test_bit(j, online_mask);
3695 return (j < nr_bits);
3699 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3700 * @n: CPU/Rx queue index
3701 * @srcp: the cpumask/Rx queue mask pointer
3702 * @nr_bits: number of bits in the bitmask
3704 * Returns >= nr_bits if no further CPUs/Rx queues set.
3706 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3707 unsigned int nr_bits)
3709 /* -1 is a legal arg here. */
3711 cpu_max_bits_warn(n, nr_bits);
3714 return find_next_bit(srcp, nr_bits, n + 1);
3720 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3721 * @n: CPU/Rx queue index
3722 * @src1p: the first CPUs/Rx queues mask pointer
3723 * @src2p: the second CPUs/Rx queues mask pointer
3724 * @nr_bits: number of bits in the bitmask
3726 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3728 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3729 const unsigned long *src2p,
3730 unsigned int nr_bits)
3732 /* -1 is a legal arg here. */
3734 cpu_max_bits_warn(n, nr_bits);
3737 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3739 return find_next_bit(src1p, nr_bits, n + 1);
3741 return find_next_bit(src2p, nr_bits, n + 1);
3746 static inline int netif_set_xps_queue(struct net_device *dev,
3747 const struct cpumask *mask,
3753 static inline int __netif_set_xps_queue(struct net_device *dev,
3754 const unsigned long *mask,
3755 u16 index, enum xps_map_type type)
3762 * netif_is_multiqueue - test if device has multiple transmit queues
3763 * @dev: network device
3765 * Check if device has multiple transmit queues
3767 static inline bool netif_is_multiqueue(const struct net_device *dev)
3769 return dev->num_tx_queues > 1;
3772 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3775 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3777 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3780 dev->real_num_rx_queues = rxqs;
3784 int netif_set_real_num_queues(struct net_device *dev,
3785 unsigned int txq, unsigned int rxq);
3787 static inline struct netdev_rx_queue *
3788 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3790 return dev->_rx + rxq;
3794 static inline unsigned int get_netdev_rx_queue_index(
3795 struct netdev_rx_queue *queue)
3797 struct net_device *dev = queue->dev;
3798 int index = queue - dev->_rx;
3800 BUG_ON(index >= dev->num_rx_queues);
3805 int netif_get_num_default_rss_queues(void);
3807 enum skb_free_reason {
3808 SKB_REASON_CONSUMED,
3812 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3813 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3816 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3817 * interrupt context or with hardware interrupts being disabled.
3818 * (in_hardirq() || irqs_disabled())
3820 * We provide four helpers that can be used in following contexts :
3822 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3823 * replacing kfree_skb(skb)
3825 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3826 * Typically used in place of consume_skb(skb) in TX completion path
3828 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3829 * replacing kfree_skb(skb)
3831 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3832 * and consumed a packet. Used in place of consume_skb(skb)
3834 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3836 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3839 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3841 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3844 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3846 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3849 static inline void dev_consume_skb_any(struct sk_buff *skb)
3851 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3854 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3855 struct bpf_prog *xdp_prog);
3856 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3857 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3858 int netif_rx(struct sk_buff *skb);
3859 int __netif_rx(struct sk_buff *skb);
3861 int netif_receive_skb(struct sk_buff *skb);
3862 int netif_receive_skb_core(struct sk_buff *skb);
3863 void netif_receive_skb_list_internal(struct list_head *head);
3864 void netif_receive_skb_list(struct list_head *head);
3865 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3866 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3867 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3868 void napi_get_frags_check(struct napi_struct *napi);
3869 gro_result_t napi_gro_frags(struct napi_struct *napi);
3870 struct packet_offload *gro_find_receive_by_type(__be16 type);
3871 struct packet_offload *gro_find_complete_by_type(__be16 type);
3873 static inline void napi_free_frags(struct napi_struct *napi)
3875 kfree_skb(napi->skb);
3879 bool netdev_is_rx_handler_busy(struct net_device *dev);
3880 int netdev_rx_handler_register(struct net_device *dev,
3881 rx_handler_func_t *rx_handler,
3882 void *rx_handler_data);
3883 void netdev_rx_handler_unregister(struct net_device *dev);
3885 bool dev_valid_name(const char *name);
3886 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3888 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3890 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3891 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3892 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3893 void __user *data, bool *need_copyout);
3894 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3895 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3896 unsigned int dev_get_flags(const struct net_device *);
3897 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3898 struct netlink_ext_ack *extack);
3899 int dev_change_flags(struct net_device *dev, unsigned int flags,
3900 struct netlink_ext_ack *extack);
3901 int dev_set_alias(struct net_device *, const char *, size_t);
3902 int dev_get_alias(const struct net_device *, char *, size_t);
3903 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3904 const char *pat, int new_ifindex);
3906 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3909 return __dev_change_net_namespace(dev, net, pat, 0);
3911 int __dev_set_mtu(struct net_device *, int);
3912 int dev_set_mtu(struct net_device *, int);
3913 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3914 struct netlink_ext_ack *extack);
3915 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3916 struct netlink_ext_ack *extack);
3917 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3918 struct netlink_ext_ack *extack);
3919 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3920 int dev_get_port_parent_id(struct net_device *dev,
3921 struct netdev_phys_item_id *ppid, bool recurse);
3922 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3923 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3924 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3925 struct netdev_queue *txq, int *ret);
3927 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3928 u8 dev_xdp_prog_count(struct net_device *dev);
3929 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3931 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3932 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3933 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3934 bool is_skb_forwardable(const struct net_device *dev,
3935 const struct sk_buff *skb);
3937 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3938 const struct sk_buff *skb,
3939 const bool check_mtu)
3941 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3944 if (!(dev->flags & IFF_UP))
3950 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3951 if (skb->len <= len)
3954 /* if TSO is enabled, we don't care about the length as the packet
3955 * could be forwarded without being segmented before
3957 if (skb_is_gso(skb))
3963 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3965 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3967 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3968 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3973 return netdev_core_stats_alloc(dev);
3976 #define DEV_CORE_STATS_INC(FIELD) \
3977 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3979 struct net_device_core_stats __percpu *p; \
3981 p = dev_core_stats(dev); \
3983 this_cpu_inc(p->FIELD); \
3985 DEV_CORE_STATS_INC(rx_dropped)
3986 DEV_CORE_STATS_INC(tx_dropped)
3987 DEV_CORE_STATS_INC(rx_nohandler)
3988 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3990 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3991 struct sk_buff *skb,
3992 const bool check_mtu)
3994 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3995 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3996 dev_core_stats_rx_dropped_inc(dev);
4001 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4006 bool dev_nit_active(struct net_device *dev);
4007 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4009 static inline void __dev_put(struct net_device *dev)
4012 #ifdef CONFIG_PCPU_DEV_REFCNT
4013 this_cpu_dec(*dev->pcpu_refcnt);
4015 refcount_dec(&dev->dev_refcnt);
4020 static inline void __dev_hold(struct net_device *dev)
4023 #ifdef CONFIG_PCPU_DEV_REFCNT
4024 this_cpu_inc(*dev->pcpu_refcnt);
4026 refcount_inc(&dev->dev_refcnt);
4031 static inline void __netdev_tracker_alloc(struct net_device *dev,
4032 netdevice_tracker *tracker,
4035 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4036 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4040 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4041 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4043 static inline void netdev_tracker_alloc(struct net_device *dev,
4044 netdevice_tracker *tracker, gfp_t gfp)
4046 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4047 refcount_dec(&dev->refcnt_tracker.no_tracker);
4048 __netdev_tracker_alloc(dev, tracker, gfp);
4052 static inline void netdev_tracker_free(struct net_device *dev,
4053 netdevice_tracker *tracker)
4055 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4056 ref_tracker_free(&dev->refcnt_tracker, tracker);
4060 static inline void netdev_hold(struct net_device *dev,
4061 netdevice_tracker *tracker, gfp_t gfp)
4065 __netdev_tracker_alloc(dev, tracker, gfp);
4069 static inline void netdev_put(struct net_device *dev,
4070 netdevice_tracker *tracker)
4073 netdev_tracker_free(dev, tracker);
4079 * dev_hold - get reference to device
4080 * @dev: network device
4082 * Hold reference to device to keep it from being freed.
4083 * Try using netdev_hold() instead.
4085 static inline void dev_hold(struct net_device *dev)
4087 netdev_hold(dev, NULL, GFP_ATOMIC);
4091 * dev_put - release reference to device
4092 * @dev: network device
4094 * Release reference to device to allow it to be freed.
4095 * Try using netdev_put() instead.
4097 static inline void dev_put(struct net_device *dev)
4099 netdev_put(dev, NULL);
4102 static inline void netdev_ref_replace(struct net_device *odev,
4103 struct net_device *ndev,
4104 netdevice_tracker *tracker,
4108 netdev_tracker_free(odev, tracker);
4114 __netdev_tracker_alloc(ndev, tracker, gfp);
4117 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4118 * and _off may be called from IRQ context, but it is caller
4119 * who is responsible for serialization of these calls.
4121 * The name carrier is inappropriate, these functions should really be
4122 * called netif_lowerlayer_*() because they represent the state of any
4123 * kind of lower layer not just hardware media.
4125 void linkwatch_fire_event(struct net_device *dev);
4128 * netif_carrier_ok - test if carrier present
4129 * @dev: network device
4131 * Check if carrier is present on device
4133 static inline bool netif_carrier_ok(const struct net_device *dev)
4135 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4138 unsigned long dev_trans_start(struct net_device *dev);
4140 void __netdev_watchdog_up(struct net_device *dev);
4142 void netif_carrier_on(struct net_device *dev);
4143 void netif_carrier_off(struct net_device *dev);
4144 void netif_carrier_event(struct net_device *dev);
4147 * netif_dormant_on - mark device as dormant.
4148 * @dev: network device
4150 * Mark device as dormant (as per RFC2863).
4152 * The dormant state indicates that the relevant interface is not
4153 * actually in a condition to pass packets (i.e., it is not 'up') but is
4154 * in a "pending" state, waiting for some external event. For "on-
4155 * demand" interfaces, this new state identifies the situation where the
4156 * interface is waiting for events to place it in the up state.
4158 static inline void netif_dormant_on(struct net_device *dev)
4160 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4161 linkwatch_fire_event(dev);
4165 * netif_dormant_off - set device as not dormant.
4166 * @dev: network device
4168 * Device is not in dormant state.
4170 static inline void netif_dormant_off(struct net_device *dev)
4172 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4173 linkwatch_fire_event(dev);
4177 * netif_dormant - test if device is dormant
4178 * @dev: network device
4180 * Check if device is dormant.
4182 static inline bool netif_dormant(const struct net_device *dev)
4184 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4189 * netif_testing_on - mark device as under test.
4190 * @dev: network device
4192 * Mark device as under test (as per RFC2863).
4194 * The testing state indicates that some test(s) must be performed on
4195 * the interface. After completion, of the test, the interface state
4196 * will change to up, dormant, or down, as appropriate.
4198 static inline void netif_testing_on(struct net_device *dev)
4200 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4201 linkwatch_fire_event(dev);
4205 * netif_testing_off - set device as not under test.
4206 * @dev: network device
4208 * Device is not in testing state.
4210 static inline void netif_testing_off(struct net_device *dev)
4212 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4213 linkwatch_fire_event(dev);
4217 * netif_testing - test if device is under test
4218 * @dev: network device
4220 * Check if device is under test
4222 static inline bool netif_testing(const struct net_device *dev)
4224 return test_bit(__LINK_STATE_TESTING, &dev->state);
4229 * netif_oper_up - test if device is operational
4230 * @dev: network device
4232 * Check if carrier is operational
4234 static inline bool netif_oper_up(const struct net_device *dev)
4236 return (dev->operstate == IF_OPER_UP ||
4237 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4241 * netif_device_present - is device available or removed
4242 * @dev: network device
4244 * Check if device has not been removed from system.
4246 static inline bool netif_device_present(const struct net_device *dev)
4248 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4251 void netif_device_detach(struct net_device *dev);
4253 void netif_device_attach(struct net_device *dev);
4256 * Network interface message level settings
4261 NETIF_MSG_PROBE_BIT,
4263 NETIF_MSG_TIMER_BIT,
4264 NETIF_MSG_IFDOWN_BIT,
4266 NETIF_MSG_RX_ERR_BIT,
4267 NETIF_MSG_TX_ERR_BIT,
4268 NETIF_MSG_TX_QUEUED_BIT,
4270 NETIF_MSG_TX_DONE_BIT,
4271 NETIF_MSG_RX_STATUS_BIT,
4272 NETIF_MSG_PKTDATA_BIT,
4276 /* When you add a new bit above, update netif_msg_class_names array
4277 * in net/ethtool/common.c
4279 NETIF_MSG_CLASS_COUNT,
4281 /* Both ethtool_ops interface and internal driver implementation use u32 */
4282 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4284 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4285 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4287 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4288 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4289 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4290 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4291 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4292 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4293 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4294 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4295 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4296 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4297 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4298 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4299 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4300 #define NETIF_MSG_HW __NETIF_MSG(HW)
4301 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4303 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4304 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4305 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4306 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4307 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4308 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4309 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4310 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4311 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4312 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4313 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4314 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4315 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4316 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4317 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4319 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4322 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4323 return default_msg_enable_bits;
4324 if (debug_value == 0) /* no output */
4326 /* set low N bits */
4327 return (1U << debug_value) - 1;
4330 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4332 spin_lock(&txq->_xmit_lock);
4333 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4334 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4337 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4339 __acquire(&txq->_xmit_lock);
4343 static inline void __netif_tx_release(struct netdev_queue *txq)
4345 __release(&txq->_xmit_lock);
4348 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4350 spin_lock_bh(&txq->_xmit_lock);
4351 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4352 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4355 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4357 bool ok = spin_trylock(&txq->_xmit_lock);
4360 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4361 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4366 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4368 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4369 WRITE_ONCE(txq->xmit_lock_owner, -1);
4370 spin_unlock(&txq->_xmit_lock);
4373 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4375 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4376 WRITE_ONCE(txq->xmit_lock_owner, -1);
4377 spin_unlock_bh(&txq->_xmit_lock);
4381 * txq->trans_start can be read locklessly from dev_watchdog()
4383 static inline void txq_trans_update(struct netdev_queue *txq)
4385 if (txq->xmit_lock_owner != -1)
4386 WRITE_ONCE(txq->trans_start, jiffies);
4389 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4391 unsigned long now = jiffies;
4393 if (READ_ONCE(txq->trans_start) != now)
4394 WRITE_ONCE(txq->trans_start, now);
4397 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4398 static inline void netif_trans_update(struct net_device *dev)
4400 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4402 txq_trans_cond_update(txq);
4406 * netif_tx_lock - grab network device transmit lock
4407 * @dev: network device
4409 * Get network device transmit lock
4411 void netif_tx_lock(struct net_device *dev);
4413 static inline void netif_tx_lock_bh(struct net_device *dev)
4419 void netif_tx_unlock(struct net_device *dev);
4421 static inline void netif_tx_unlock_bh(struct net_device *dev)
4423 netif_tx_unlock(dev);
4427 #define HARD_TX_LOCK(dev, txq, cpu) { \
4428 if ((dev->features & NETIF_F_LLTX) == 0) { \
4429 __netif_tx_lock(txq, cpu); \
4431 __netif_tx_acquire(txq); \
4435 #define HARD_TX_TRYLOCK(dev, txq) \
4436 (((dev->features & NETIF_F_LLTX) == 0) ? \
4437 __netif_tx_trylock(txq) : \
4438 __netif_tx_acquire(txq))
4440 #define HARD_TX_UNLOCK(dev, txq) { \
4441 if ((dev->features & NETIF_F_LLTX) == 0) { \
4442 __netif_tx_unlock(txq); \
4444 __netif_tx_release(txq); \
4448 static inline void netif_tx_disable(struct net_device *dev)
4454 cpu = smp_processor_id();
4455 spin_lock(&dev->tx_global_lock);
4456 for (i = 0; i < dev->num_tx_queues; i++) {
4457 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4459 __netif_tx_lock(txq, cpu);
4460 netif_tx_stop_queue(txq);
4461 __netif_tx_unlock(txq);
4463 spin_unlock(&dev->tx_global_lock);
4467 static inline void netif_addr_lock(struct net_device *dev)
4469 unsigned char nest_level = 0;
4471 #ifdef CONFIG_LOCKDEP
4472 nest_level = dev->nested_level;
4474 spin_lock_nested(&dev->addr_list_lock, nest_level);
4477 static inline void netif_addr_lock_bh(struct net_device *dev)
4479 unsigned char nest_level = 0;
4481 #ifdef CONFIG_LOCKDEP
4482 nest_level = dev->nested_level;
4485 spin_lock_nested(&dev->addr_list_lock, nest_level);
4488 static inline void netif_addr_unlock(struct net_device *dev)
4490 spin_unlock(&dev->addr_list_lock);
4493 static inline void netif_addr_unlock_bh(struct net_device *dev)
4495 spin_unlock_bh(&dev->addr_list_lock);
4499 * dev_addrs walker. Should be used only for read access. Call with
4500 * rcu_read_lock held.
4502 #define for_each_dev_addr(dev, ha) \
4503 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4505 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4507 void ether_setup(struct net_device *dev);
4509 /* Support for loadable net-drivers */
4510 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4511 unsigned char name_assign_type,
4512 void (*setup)(struct net_device *),
4513 unsigned int txqs, unsigned int rxqs);
4514 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4515 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4517 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4518 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4521 int register_netdev(struct net_device *dev);
4522 void unregister_netdev(struct net_device *dev);
4524 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4526 /* General hardware address lists handling functions */
4527 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4528 struct netdev_hw_addr_list *from_list, int addr_len);
4529 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4530 struct netdev_hw_addr_list *from_list, int addr_len);
4531 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4532 struct net_device *dev,
4533 int (*sync)(struct net_device *, const unsigned char *),
4534 int (*unsync)(struct net_device *,
4535 const unsigned char *));
4536 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4537 struct net_device *dev,
4538 int (*sync)(struct net_device *,
4539 const unsigned char *, int),
4540 int (*unsync)(struct net_device *,
4541 const unsigned char *, int));
4542 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4543 struct net_device *dev,
4544 int (*unsync)(struct net_device *,
4545 const unsigned char *, int));
4546 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4547 struct net_device *dev,
4548 int (*unsync)(struct net_device *,
4549 const unsigned char *));
4550 void __hw_addr_init(struct netdev_hw_addr_list *list);
4552 /* Functions used for device addresses handling */
4553 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4554 const void *addr, size_t len);
4557 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4559 dev_addr_mod(dev, 0, addr, len);
4562 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4564 __dev_addr_set(dev, addr, dev->addr_len);
4567 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4568 unsigned char addr_type);
4569 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4570 unsigned char addr_type);
4572 /* Functions used for unicast addresses handling */
4573 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4574 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4575 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4576 int dev_uc_sync(struct net_device *to, struct net_device *from);
4577 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4578 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4579 void dev_uc_flush(struct net_device *dev);
4580 void dev_uc_init(struct net_device *dev);
4583 * __dev_uc_sync - Synchonize device's unicast list
4584 * @dev: device to sync
4585 * @sync: function to call if address should be added
4586 * @unsync: function to call if address should be removed
4588 * Add newly added addresses to the interface, and release
4589 * addresses that have been deleted.
4591 static inline int __dev_uc_sync(struct net_device *dev,
4592 int (*sync)(struct net_device *,
4593 const unsigned char *),
4594 int (*unsync)(struct net_device *,
4595 const unsigned char *))
4597 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4601 * __dev_uc_unsync - Remove synchronized addresses from device
4602 * @dev: device to sync
4603 * @unsync: function to call if address should be removed
4605 * Remove all addresses that were added to the device by dev_uc_sync().
4607 static inline void __dev_uc_unsync(struct net_device *dev,
4608 int (*unsync)(struct net_device *,
4609 const unsigned char *))
4611 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4614 /* Functions used for multicast addresses handling */
4615 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4616 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4617 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4618 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4619 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4620 int dev_mc_sync(struct net_device *to, struct net_device *from);
4621 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4622 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4623 void dev_mc_flush(struct net_device *dev);
4624 void dev_mc_init(struct net_device *dev);
4627 * __dev_mc_sync - Synchonize device's multicast list
4628 * @dev: device to sync
4629 * @sync: function to call if address should be added
4630 * @unsync: function to call if address should be removed
4632 * Add newly added addresses to the interface, and release
4633 * addresses that have been deleted.
4635 static inline int __dev_mc_sync(struct net_device *dev,
4636 int (*sync)(struct net_device *,
4637 const unsigned char *),
4638 int (*unsync)(struct net_device *,
4639 const unsigned char *))
4641 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4645 * __dev_mc_unsync - Remove synchronized addresses from device
4646 * @dev: device to sync
4647 * @unsync: function to call if address should be removed
4649 * Remove all addresses that were added to the device by dev_mc_sync().
4651 static inline void __dev_mc_unsync(struct net_device *dev,
4652 int (*unsync)(struct net_device *,
4653 const unsigned char *))
4655 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4658 /* Functions used for secondary unicast and multicast support */
4659 void dev_set_rx_mode(struct net_device *dev);
4660 int dev_set_promiscuity(struct net_device *dev, int inc);
4661 int dev_set_allmulti(struct net_device *dev, int inc);
4662 void netdev_state_change(struct net_device *dev);
4663 void __netdev_notify_peers(struct net_device *dev);
4664 void netdev_notify_peers(struct net_device *dev);
4665 void netdev_features_change(struct net_device *dev);
4666 /* Load a device via the kmod */
4667 void dev_load(struct net *net, const char *name);
4668 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4669 struct rtnl_link_stats64 *storage);
4670 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4671 const struct net_device_stats *netdev_stats);
4672 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4673 const struct pcpu_sw_netstats __percpu *netstats);
4674 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4676 extern int netdev_max_backlog;
4677 extern int dev_rx_weight;
4678 extern int dev_tx_weight;
4679 extern int gro_normal_batch;
4682 NESTED_SYNC_IMM_BIT,
4683 NESTED_SYNC_TODO_BIT,
4686 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4687 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4689 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4690 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4692 struct netdev_nested_priv {
4693 unsigned char flags;
4697 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4698 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4699 struct list_head **iter);
4701 /* iterate through upper list, must be called under RCU read lock */
4702 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4703 for (iter = &(dev)->adj_list.upper, \
4704 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4706 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4708 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4709 int (*fn)(struct net_device *upper_dev,
4710 struct netdev_nested_priv *priv),
4711 struct netdev_nested_priv *priv);
4713 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4714 struct net_device *upper_dev);
4716 bool netdev_has_any_upper_dev(struct net_device *dev);
4718 void *netdev_lower_get_next_private(struct net_device *dev,
4719 struct list_head **iter);
4720 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4721 struct list_head **iter);
4723 #define netdev_for_each_lower_private(dev, priv, iter) \
4724 for (iter = (dev)->adj_list.lower.next, \
4725 priv = netdev_lower_get_next_private(dev, &(iter)); \
4727 priv = netdev_lower_get_next_private(dev, &(iter)))
4729 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4730 for (iter = &(dev)->adj_list.lower, \
4731 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4733 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4735 void *netdev_lower_get_next(struct net_device *dev,
4736 struct list_head **iter);
4738 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4739 for (iter = (dev)->adj_list.lower.next, \
4740 ldev = netdev_lower_get_next(dev, &(iter)); \
4742 ldev = netdev_lower_get_next(dev, &(iter)))
4744 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4745 struct list_head **iter);
4746 int netdev_walk_all_lower_dev(struct net_device *dev,
4747 int (*fn)(struct net_device *lower_dev,
4748 struct netdev_nested_priv *priv),
4749 struct netdev_nested_priv *priv);
4750 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4751 int (*fn)(struct net_device *lower_dev,
4752 struct netdev_nested_priv *priv),
4753 struct netdev_nested_priv *priv);
4755 void *netdev_adjacent_get_private(struct list_head *adj_list);
4756 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4757 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4758 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4759 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4760 struct netlink_ext_ack *extack);
4761 int netdev_master_upper_dev_link(struct net_device *dev,
4762 struct net_device *upper_dev,
4763 void *upper_priv, void *upper_info,
4764 struct netlink_ext_ack *extack);
4765 void netdev_upper_dev_unlink(struct net_device *dev,
4766 struct net_device *upper_dev);
4767 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4768 struct net_device *new_dev,
4769 struct net_device *dev,
4770 struct netlink_ext_ack *extack);
4771 void netdev_adjacent_change_commit(struct net_device *old_dev,
4772 struct net_device *new_dev,
4773 struct net_device *dev);
4774 void netdev_adjacent_change_abort(struct net_device *old_dev,
4775 struct net_device *new_dev,
4776 struct net_device *dev);
4777 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4778 void *netdev_lower_dev_get_private(struct net_device *dev,
4779 struct net_device *lower_dev);
4780 void netdev_lower_state_changed(struct net_device *lower_dev,
4781 void *lower_state_info);
4783 /* RSS keys are 40 or 52 bytes long */
4784 #define NETDEV_RSS_KEY_LEN 52
4785 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4786 void netdev_rss_key_fill(void *buffer, size_t len);
4788 int skb_checksum_help(struct sk_buff *skb);
4789 int skb_crc32c_csum_help(struct sk_buff *skb);
4790 int skb_csum_hwoffload_help(struct sk_buff *skb,
4791 const netdev_features_t features);
4793 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4794 netdev_features_t features, bool tx_path);
4795 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4796 netdev_features_t features, __be16 type);
4797 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4798 netdev_features_t features);
4800 struct netdev_bonding_info {
4805 struct netdev_notifier_bonding_info {
4806 struct netdev_notifier_info info; /* must be first */
4807 struct netdev_bonding_info bonding_info;
4810 void netdev_bonding_info_change(struct net_device *dev,
4811 struct netdev_bonding_info *bonding_info);
4813 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4814 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4816 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4823 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4825 return __skb_gso_segment(skb, features, true);
4827 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4829 static inline bool can_checksum_protocol(netdev_features_t features,
4832 if (protocol == htons(ETH_P_FCOE))
4833 return !!(features & NETIF_F_FCOE_CRC);
4835 /* Assume this is an IP checksum (not SCTP CRC) */
4837 if (features & NETIF_F_HW_CSUM) {
4838 /* Can checksum everything */
4843 case htons(ETH_P_IP):
4844 return !!(features & NETIF_F_IP_CSUM);
4845 case htons(ETH_P_IPV6):
4846 return !!(features & NETIF_F_IPV6_CSUM);
4853 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4855 static inline void netdev_rx_csum_fault(struct net_device *dev,
4856 struct sk_buff *skb)
4860 /* rx skb timestamps */
4861 void net_enable_timestamp(void);
4862 void net_disable_timestamp(void);
4864 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4865 const struct skb_shared_hwtstamps *hwtstamps,
4868 const struct net_device_ops *ops = dev->netdev_ops;
4870 if (ops->ndo_get_tstamp)
4871 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4873 return hwtstamps->hwtstamp;
4876 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4877 struct sk_buff *skb, struct net_device *dev,
4880 __this_cpu_write(softnet_data.xmit.more, more);
4881 return ops->ndo_start_xmit(skb, dev);
4884 static inline bool netdev_xmit_more(void)
4886 return __this_cpu_read(softnet_data.xmit.more);
4889 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4890 struct netdev_queue *txq, bool more)
4892 const struct net_device_ops *ops = dev->netdev_ops;
4895 rc = __netdev_start_xmit(ops, skb, dev, more);
4896 if (rc == NETDEV_TX_OK)
4897 txq_trans_update(txq);
4902 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4904 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4907 extern const struct kobj_ns_type_operations net_ns_type_operations;
4909 const char *netdev_drivername(const struct net_device *dev);
4911 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4912 netdev_features_t f2)
4914 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4915 if (f1 & NETIF_F_HW_CSUM)
4916 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4918 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4924 static inline netdev_features_t netdev_get_wanted_features(
4925 struct net_device *dev)
4927 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4929 netdev_features_t netdev_increment_features(netdev_features_t all,
4930 netdev_features_t one, netdev_features_t mask);
4932 /* Allow TSO being used on stacked device :
4933 * Performing the GSO segmentation before last device
4934 * is a performance improvement.
4936 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4937 netdev_features_t mask)
4939 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4942 int __netdev_update_features(struct net_device *dev);
4943 void netdev_update_features(struct net_device *dev);
4944 void netdev_change_features(struct net_device *dev);
4946 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4947 struct net_device *dev);
4949 netdev_features_t passthru_features_check(struct sk_buff *skb,
4950 struct net_device *dev,
4951 netdev_features_t features);
4952 netdev_features_t netif_skb_features(struct sk_buff *skb);
4954 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4956 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4958 /* check flags correspondence */
4959 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4960 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4961 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4962 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4963 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4964 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4965 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4966 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4967 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4968 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4969 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4970 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4971 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4972 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4973 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4974 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4975 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4976 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4977 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4979 return (features & feature) == feature;
4982 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4984 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4985 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4988 static inline bool netif_needs_gso(struct sk_buff *skb,
4989 netdev_features_t features)
4991 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4992 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4993 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4996 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4997 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
4998 void netif_inherit_tso_max(struct net_device *to,
4999 const struct net_device *from);
5001 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5002 int pulled_hlen, u16 mac_offset,
5005 skb->protocol = protocol;
5006 skb->encapsulation = 1;
5007 skb_push(skb, pulled_hlen);
5008 skb_reset_transport_header(skb);
5009 skb->mac_header = mac_offset;
5010 skb->network_header = skb->mac_header + mac_len;
5011 skb->mac_len = mac_len;
5014 static inline bool netif_is_macsec(const struct net_device *dev)
5016 return dev->priv_flags & IFF_MACSEC;
5019 static inline bool netif_is_macvlan(const struct net_device *dev)
5021 return dev->priv_flags & IFF_MACVLAN;
5024 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5026 return dev->priv_flags & IFF_MACVLAN_PORT;
5029 static inline bool netif_is_bond_master(const struct net_device *dev)
5031 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5034 static inline bool netif_is_bond_slave(const struct net_device *dev)
5036 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5039 static inline bool netif_supports_nofcs(struct net_device *dev)
5041 return dev->priv_flags & IFF_SUPP_NOFCS;
5044 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5046 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5049 static inline bool netif_is_l3_master(const struct net_device *dev)
5051 return dev->priv_flags & IFF_L3MDEV_MASTER;
5054 static inline bool netif_is_l3_slave(const struct net_device *dev)
5056 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5059 static inline bool netif_is_bridge_master(const struct net_device *dev)
5061 return dev->priv_flags & IFF_EBRIDGE;
5064 static inline bool netif_is_bridge_port(const struct net_device *dev)
5066 return dev->priv_flags & IFF_BRIDGE_PORT;
5069 static inline bool netif_is_ovs_master(const struct net_device *dev)
5071 return dev->priv_flags & IFF_OPENVSWITCH;
5074 static inline bool netif_is_ovs_port(const struct net_device *dev)
5076 return dev->priv_flags & IFF_OVS_DATAPATH;
5079 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5081 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5084 static inline bool netif_is_team_master(const struct net_device *dev)
5086 return dev->priv_flags & IFF_TEAM;
5089 static inline bool netif_is_team_port(const struct net_device *dev)
5091 return dev->priv_flags & IFF_TEAM_PORT;
5094 static inline bool netif_is_lag_master(const struct net_device *dev)
5096 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5099 static inline bool netif_is_lag_port(const struct net_device *dev)
5101 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5104 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5106 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5109 static inline bool netif_is_failover(const struct net_device *dev)
5111 return dev->priv_flags & IFF_FAILOVER;
5114 static inline bool netif_is_failover_slave(const struct net_device *dev)
5116 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5119 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5120 static inline void netif_keep_dst(struct net_device *dev)
5122 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5125 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5126 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5128 /* TODO: reserve and use an additional IFF bit, if we get more users */
5129 return netif_is_macsec(dev);
5132 extern struct pernet_operations __net_initdata loopback_net_ops;
5134 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5136 /* netdev_printk helpers, similar to dev_printk */
5138 static inline const char *netdev_name(const struct net_device *dev)
5140 if (!dev->name[0] || strchr(dev->name, '%'))
5141 return "(unnamed net_device)";
5145 static inline const char *netdev_reg_state(const struct net_device *dev)
5147 switch (dev->reg_state) {
5148 case NETREG_UNINITIALIZED: return " (uninitialized)";
5149 case NETREG_REGISTERED: return "";
5150 case NETREG_UNREGISTERING: return " (unregistering)";
5151 case NETREG_UNREGISTERED: return " (unregistered)";
5152 case NETREG_RELEASED: return " (released)";
5153 case NETREG_DUMMY: return " (dummy)";
5156 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5157 return " (unknown)";
5160 #define MODULE_ALIAS_NETDEV(device) \
5161 MODULE_ALIAS("netdev-" device)
5164 * netdev_WARN() acts like dev_printk(), but with the key difference
5165 * of using a WARN/WARN_ON to get the message out, including the
5166 * file/line information and a backtrace.
5168 #define netdev_WARN(dev, format, args...) \
5169 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5170 netdev_reg_state(dev), ##args)
5172 #define netdev_WARN_ONCE(dev, format, args...) \
5173 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5174 netdev_reg_state(dev), ##args)
5177 * The list of packet types we will receive (as opposed to discard)
5178 * and the routines to invoke.
5180 * Why 16. Because with 16 the only overlap we get on a hash of the
5181 * low nibble of the protocol value is RARP/SNAP/X.25.
5195 #define PTYPE_HASH_SIZE (16)
5196 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5198 extern struct list_head ptype_all __read_mostly;
5199 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5201 extern struct net_device *blackhole_netdev;
5203 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5204 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5205 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5206 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5208 #endif /* _LINUX_NETDEVICE_H */