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 + READ_ONCE((dev)->needed_headroom)) \
301 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
302 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
303 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
304 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
307 int (*create) (struct sk_buff *skb, struct net_device *dev,
308 unsigned short type, const void *daddr,
309 const void *saddr, unsigned int len);
310 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
311 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
312 void (*cache_update)(struct hh_cache *hh,
313 const struct net_device *dev,
314 const unsigned char *haddr);
315 bool (*validate)(const char *ll_header, unsigned int len);
316 __be16 (*parse_protocol)(const struct sk_buff *skb);
319 /* These flag bits are private to the generic network queueing
320 * layer; they may not be explicitly referenced by any other
324 enum netdev_state_t {
326 __LINK_STATE_PRESENT,
327 __LINK_STATE_NOCARRIER,
328 __LINK_STATE_LINKWATCH_PENDING,
329 __LINK_STATE_DORMANT,
330 __LINK_STATE_TESTING,
334 struct list_head list;
339 * size of gro hash buckets, must less than bit number of
340 * napi_struct::gro_bitmask
342 #define GRO_HASH_BUCKETS 8
345 * Structure for NAPI scheduling similar to tasklet but with weighting
348 /* The poll_list must only be managed by the entity which
349 * changes the state of the NAPI_STATE_SCHED bit. This means
350 * whoever atomically sets that bit can add this napi_struct
351 * to the per-CPU poll_list, and whoever clears that bit
352 * can remove from the list right before clearing the bit.
354 struct list_head poll_list;
358 int defer_hard_irqs_count;
359 unsigned long gro_bitmask;
360 int (*poll)(struct napi_struct *, int);
361 #ifdef CONFIG_NETPOLL
364 struct net_device *dev;
365 struct gro_list gro_hash[GRO_HASH_BUCKETS];
367 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
368 int rx_count; /* length of rx_list */
369 struct hrtimer timer;
370 struct list_head dev_list;
371 struct hlist_node napi_hash_node;
372 unsigned int napi_id;
373 struct task_struct *thread;
377 NAPI_STATE_SCHED, /* Poll is scheduled */
378 NAPI_STATE_MISSED, /* reschedule a napi */
379 NAPI_STATE_DISABLE, /* Disable pending */
380 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
381 NAPI_STATE_LISTED, /* NAPI added to system lists */
382 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
383 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
384 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
385 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
386 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
390 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
391 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
392 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
393 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
394 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
395 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
396 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
397 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
398 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
399 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
409 typedef enum gro_result gro_result_t;
412 * enum rx_handler_result - Possible return values for rx_handlers.
413 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
415 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
416 * case skb->dev was changed by rx_handler.
417 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
418 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
420 * rx_handlers are functions called from inside __netif_receive_skb(), to do
421 * special processing of the skb, prior to delivery to protocol handlers.
423 * Currently, a net_device can only have a single rx_handler registered. Trying
424 * to register a second rx_handler will return -EBUSY.
426 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
427 * To unregister a rx_handler on a net_device, use
428 * netdev_rx_handler_unregister().
430 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
433 * If the rx_handler consumed the skb in some way, it should return
434 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
435 * the skb to be delivered in some other way.
437 * If the rx_handler changed skb->dev, to divert the skb to another
438 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
439 * new device will be called if it exists.
441 * If the rx_handler decides the skb should be ignored, it should return
442 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
443 * are registered on exact device (ptype->dev == skb->dev).
445 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
446 * delivered, it should return RX_HANDLER_PASS.
448 * A device without a registered rx_handler will behave as if rx_handler
449 * returned RX_HANDLER_PASS.
452 enum rx_handler_result {
458 typedef enum rx_handler_result rx_handler_result_t;
459 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
461 void __napi_schedule(struct napi_struct *n);
462 void __napi_schedule_irqoff(struct napi_struct *n);
464 static inline bool napi_disable_pending(struct napi_struct *n)
466 return test_bit(NAPI_STATE_DISABLE, &n->state);
469 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
471 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
474 bool napi_schedule_prep(struct napi_struct *n);
477 * napi_schedule - schedule NAPI poll
480 * Schedule NAPI poll routine to be called if it is not already
483 static inline void napi_schedule(struct napi_struct *n)
485 if (napi_schedule_prep(n))
490 * napi_schedule_irqoff - schedule NAPI poll
493 * Variant of napi_schedule(), assuming hard irqs are masked.
495 static inline void napi_schedule_irqoff(struct napi_struct *n)
497 if (napi_schedule_prep(n))
498 __napi_schedule_irqoff(n);
501 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
502 static inline bool napi_reschedule(struct napi_struct *napi)
504 if (napi_schedule_prep(napi)) {
505 __napi_schedule(napi);
511 bool napi_complete_done(struct napi_struct *n, int work_done);
513 * napi_complete - NAPI processing complete
516 * Mark NAPI processing as complete.
517 * Consider using napi_complete_done() instead.
518 * Return false if device should avoid rearming interrupts.
520 static inline bool napi_complete(struct napi_struct *n)
522 return napi_complete_done(n, 0);
525 int dev_set_threaded(struct net_device *dev, bool threaded);
528 * napi_disable - prevent NAPI from scheduling
531 * Stop NAPI from being scheduled on this context.
532 * Waits till any outstanding processing completes.
534 void napi_disable(struct napi_struct *n);
536 void napi_enable(struct napi_struct *n);
539 * napi_synchronize - wait until NAPI is not running
542 * Wait until NAPI is done being scheduled on this context.
543 * Waits till any outstanding processing completes but
544 * does not disable future activations.
546 static inline void napi_synchronize(const struct napi_struct *n)
548 if (IS_ENABLED(CONFIG_SMP))
549 while (test_bit(NAPI_STATE_SCHED, &n->state))
556 * napi_if_scheduled_mark_missed - if napi is running, set the
560 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
563 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
565 unsigned long val, new;
567 val = READ_ONCE(n->state);
569 if (val & NAPIF_STATE_DISABLE)
572 if (!(val & NAPIF_STATE_SCHED))
575 new = val | NAPIF_STATE_MISSED;
576 } while (!try_cmpxchg(&n->state, &val, new));
581 enum netdev_queue_state_t {
582 __QUEUE_STATE_DRV_XOFF,
583 __QUEUE_STATE_STACK_XOFF,
584 __QUEUE_STATE_FROZEN,
587 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
588 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
589 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
591 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
592 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
594 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
598 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
599 * netif_tx_* functions below are used to manipulate this flag. The
600 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
601 * queue independently. The netif_xmit_*stopped functions below are called
602 * to check if the queue has been stopped by the driver or stack (either
603 * of the XOFF bits are set in the state). Drivers should not need to call
604 * netif_xmit*stopped functions, they should only be using netif_tx_*.
607 struct netdev_queue {
611 struct net_device *dev;
612 netdevice_tracker dev_tracker;
614 struct Qdisc __rcu *qdisc;
615 struct Qdisc *qdisc_sleeping;
619 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
622 unsigned long tx_maxrate;
624 * Number of TX timeouts for this queue
625 * (/sys/class/net/DEV/Q/trans_timeout)
627 atomic_long_t trans_timeout;
629 /* Subordinate device that the queue has been assigned to */
630 struct net_device *sb_dev;
631 #ifdef CONFIG_XDP_SOCKETS
632 struct xsk_buff_pool *pool;
637 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
640 * Time (in jiffies) of last Tx
642 unsigned long trans_start;
649 } ____cacheline_aligned_in_smp;
651 extern int sysctl_fb_tunnels_only_for_init_net;
652 extern int sysctl_devconf_inherit_init_net;
655 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
656 * == 1 : For initns only
659 static inline bool net_has_fallback_tunnels(const struct net *net)
661 #if IS_ENABLED(CONFIG_SYSCTL)
662 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
664 return !fb_tunnels_only_for_init_net ||
665 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
671 static inline int net_inherit_devconf(void)
673 #if IS_ENABLED(CONFIG_SYSCTL)
674 return READ_ONCE(sysctl_devconf_inherit_init_net);
680 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
682 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
689 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
691 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
698 * This structure holds an RPS map which can be of variable length. The
699 * map is an array of CPUs.
706 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
709 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
710 * tail pointer for that CPU's input queue at the time of last enqueue, and
711 * a hardware filter index.
713 struct rps_dev_flow {
716 unsigned int last_qtail;
718 #define RPS_NO_FILTER 0xffff
721 * The rps_dev_flow_table structure contains a table of flow mappings.
723 struct rps_dev_flow_table {
726 struct rps_dev_flow flows[];
728 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
729 ((_num) * sizeof(struct rps_dev_flow)))
732 * The rps_sock_flow_table contains mappings of flows to the last CPU
733 * on which they were processed by the application (set in recvmsg).
734 * Each entry is a 32bit value. Upper part is the high-order bits
735 * of flow hash, lower part is CPU number.
736 * rps_cpu_mask is used to partition the space, depending on number of
737 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
738 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
739 * meaning we use 32-6=26 bits for the hash.
741 struct rps_sock_flow_table {
744 u32 ents[] ____cacheline_aligned_in_smp;
746 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
748 #define RPS_NO_CPU 0xffff
750 extern u32 rps_cpu_mask;
751 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
753 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
757 unsigned int index = hash & table->mask;
758 u32 val = hash & ~rps_cpu_mask;
760 /* We only give a hint, preemption can change CPU under us */
761 val |= raw_smp_processor_id();
763 if (table->ents[index] != val)
764 table->ents[index] = val;
768 #ifdef CONFIG_RFS_ACCEL
769 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
772 #endif /* CONFIG_RPS */
774 /* This structure contains an instance of an RX queue. */
775 struct netdev_rx_queue {
776 struct xdp_rxq_info xdp_rxq;
778 struct rps_map __rcu *rps_map;
779 struct rps_dev_flow_table __rcu *rps_flow_table;
782 struct net_device *dev;
783 netdevice_tracker dev_tracker;
785 #ifdef CONFIG_XDP_SOCKETS
786 struct xsk_buff_pool *pool;
788 } ____cacheline_aligned_in_smp;
791 * RX queue sysfs structures and functions.
793 struct rx_queue_attribute {
794 struct attribute attr;
795 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
796 ssize_t (*store)(struct netdev_rx_queue *queue,
797 const char *buf, size_t len);
800 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
809 * This structure holds an XPS map which can be of variable length. The
810 * map is an array of queues.
814 unsigned int alloc_len;
818 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
819 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
820 - sizeof(struct xps_map)) / sizeof(u16))
823 * This structure holds all XPS maps for device. Maps are indexed by CPU.
825 * We keep track of the number of cpus/rxqs used when the struct is allocated,
826 * in nr_ids. This will help not accessing out-of-bound memory.
828 * We keep track of the number of traffic classes used when the struct is
829 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
830 * not crossing its upper bound, as the original dev->num_tc can be updated in
833 struct xps_dev_maps {
837 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
840 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
841 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
843 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
844 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
846 #endif /* CONFIG_XPS */
848 #define TC_MAX_QUEUE 16
849 #define TC_BITMASK 15
850 /* HW offloaded queuing disciplines txq count and offset maps */
851 struct netdev_tc_txq {
856 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
858 * This structure is to hold information about the device
859 * configured to run FCoE protocol stack.
861 struct netdev_fcoe_hbainfo {
862 char manufacturer[64];
863 char serial_number[64];
864 char hardware_version[64];
865 char driver_version[64];
866 char optionrom_version[64];
867 char firmware_version[64];
869 char model_description[256];
873 #define MAX_PHYS_ITEM_ID_LEN 32
875 /* This structure holds a unique identifier to identify some
876 * physical item (port for example) used by a netdevice.
878 struct netdev_phys_item_id {
879 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
880 unsigned char id_len;
883 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
884 struct netdev_phys_item_id *b)
886 return a->id_len == b->id_len &&
887 memcmp(a->id, b->id, a->id_len) == 0;
890 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
892 struct net_device *sb_dev);
894 enum net_device_path_type {
895 DEV_PATH_ETHERNET = 0,
903 struct net_device_path {
904 enum net_device_path_type type;
905 const struct net_device *dev;
914 DEV_PATH_BR_VLAN_KEEP,
915 DEV_PATH_BR_VLAN_TAG,
916 DEV_PATH_BR_VLAN_UNTAG,
917 DEV_PATH_BR_VLAN_UNTAG_HW,
935 #define NET_DEVICE_PATH_STACK_MAX 5
936 #define NET_DEVICE_PATH_VLAN_MAX 2
938 struct net_device_path_stack {
940 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
943 struct net_device_path_ctx {
944 const struct net_device *dev;
951 } vlan[NET_DEVICE_PATH_VLAN_MAX];
956 TC_SETUP_QDISC_MQPRIO,
959 TC_SETUP_CLSMATCHALL,
969 TC_SETUP_QDISC_TAPRIO,
978 /* These structures hold the attributes of bpf state that are being passed
979 * to the netdevice through the bpf op.
981 enum bpf_netdev_command {
982 /* Set or clear a bpf program used in the earliest stages of packet
983 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
984 * is responsible for calling bpf_prog_put on any old progs that are
985 * stored. In case of error, the callee need not release the new prog
986 * reference, but on success it takes ownership and must bpf_prog_put
987 * when it is no longer used.
991 /* BPF program for offload callbacks, invoked at program load time. */
992 BPF_OFFLOAD_MAP_ALLOC,
993 BPF_OFFLOAD_MAP_FREE,
997 struct bpf_prog_offload_ops;
998 struct netlink_ext_ack;
1000 struct xdp_dev_bulk_queue;
1001 struct bpf_xdp_link;
1010 struct bpf_xdp_entity {
1011 struct bpf_prog *prog;
1012 struct bpf_xdp_link *link;
1016 enum bpf_netdev_command command;
1018 /* XDP_SETUP_PROG */
1021 struct bpf_prog *prog;
1022 struct netlink_ext_ack *extack;
1024 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1026 struct bpf_offloaded_map *offmap;
1028 /* XDP_SETUP_XSK_POOL */
1030 struct xsk_buff_pool *pool;
1036 /* Flags for ndo_xsk_wakeup. */
1037 #define XDP_WAKEUP_RX (1 << 0)
1038 #define XDP_WAKEUP_TX (1 << 1)
1040 #ifdef CONFIG_XFRM_OFFLOAD
1041 struct xfrmdev_ops {
1042 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1043 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1044 void (*xdo_dev_state_free) (struct xfrm_state *x);
1045 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1046 struct xfrm_state *x);
1047 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1048 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1049 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1050 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1051 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1055 struct dev_ifalias {
1056 struct rcu_head rcuhead;
1063 struct netdev_net_notifier {
1064 struct list_head list;
1065 struct notifier_block *nb;
1069 * This structure defines the management hooks for network devices.
1070 * The following hooks can be defined; unless noted otherwise, they are
1071 * optional and can be filled with a null pointer.
1073 * int (*ndo_init)(struct net_device *dev);
1074 * This function is called once when a network device is registered.
1075 * The network device can use this for any late stage initialization
1076 * or semantic validation. It can fail with an error code which will
1077 * be propagated back to register_netdev.
1079 * void (*ndo_uninit)(struct net_device *dev);
1080 * This function is called when device is unregistered or when registration
1081 * fails. It is not called if init fails.
1083 * int (*ndo_open)(struct net_device *dev);
1084 * This function is called when a network device transitions to the up
1087 * int (*ndo_stop)(struct net_device *dev);
1088 * This function is called when a network device transitions to the down
1091 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1092 * struct net_device *dev);
1093 * Called when a packet needs to be transmitted.
1094 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1095 * the queue before that can happen; it's for obsolete devices and weird
1096 * corner cases, but the stack really does a non-trivial amount
1097 * of useless work if you return NETDEV_TX_BUSY.
1098 * Required; cannot be NULL.
1100 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1101 * struct net_device *dev
1102 * netdev_features_t features);
1103 * Called by core transmit path to determine if device is capable of
1104 * performing offload operations on a given packet. This is to give
1105 * the device an opportunity to implement any restrictions that cannot
1106 * be otherwise expressed by feature flags. The check is called with
1107 * the set of features that the stack has calculated and it returns
1108 * those the driver believes to be appropriate.
1110 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1111 * struct net_device *sb_dev);
1112 * Called to decide which queue to use when device supports multiple
1115 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1116 * This function is called to allow device receiver to make
1117 * changes to configuration when multicast or promiscuous is enabled.
1119 * void (*ndo_set_rx_mode)(struct net_device *dev);
1120 * This function is called device changes address list filtering.
1121 * If driver handles unicast address filtering, it should set
1122 * IFF_UNICAST_FLT in its priv_flags.
1124 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1125 * This function is called when the Media Access Control address
1126 * needs to be changed. If this interface is not defined, the
1127 * MAC address can not be changed.
1129 * int (*ndo_validate_addr)(struct net_device *dev);
1130 * Test if Media Access Control address is valid for the device.
1132 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1133 * Old-style ioctl entry point. This is used internally by the
1134 * appletalk and ieee802154 subsystems but is no longer called by
1135 * the device ioctl handler.
1137 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1138 * Used by the bonding driver for its device specific ioctls:
1139 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1140 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1142 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1143 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1144 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1146 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1147 * Used to set network devices bus interface parameters. This interface
1148 * is retained for legacy reasons; new devices should use the bus
1149 * interface (PCI) for low level management.
1151 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1152 * Called when a user wants to change the Maximum Transfer Unit
1155 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1156 * Callback used when the transmitter has not made any progress
1157 * for dev->watchdog ticks.
1159 * void (*ndo_get_stats64)(struct net_device *dev,
1160 * struct rtnl_link_stats64 *storage);
1161 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1162 * Called when a user wants to get the network device usage
1163 * statistics. Drivers must do one of the following:
1164 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1165 * rtnl_link_stats64 structure passed by the caller.
1166 * 2. Define @ndo_get_stats to update a net_device_stats structure
1167 * (which should normally be dev->stats) and return a pointer to
1168 * it. The structure may be changed asynchronously only if each
1169 * field is written atomically.
1170 * 3. Update dev->stats asynchronously and atomically, and define
1171 * neither operation.
1173 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1174 * Return true if this device supports offload stats of this attr_id.
1176 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1178 * Get statistics for offload operations by attr_id. Write it into the
1179 * attr_data pointer.
1181 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1182 * If device supports VLAN filtering this function is called when a
1183 * VLAN id is registered.
1185 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1186 * If device supports VLAN filtering this function is called when a
1187 * VLAN id is unregistered.
1189 * void (*ndo_poll_controller)(struct net_device *dev);
1191 * SR-IOV management functions.
1192 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1193 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1194 * u8 qos, __be16 proto);
1195 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1197 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1198 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1199 * int (*ndo_get_vf_config)(struct net_device *dev,
1200 * int vf, struct ifla_vf_info *ivf);
1201 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1202 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1203 * struct nlattr *port[]);
1205 * Enable or disable the VF ability to query its RSS Redirection Table and
1206 * Hash Key. This is needed since on some devices VF share this information
1207 * with PF and querying it may introduce a theoretical security risk.
1208 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1209 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1210 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1212 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1213 * This is always called from the stack with the rtnl lock held and netif
1214 * tx queues stopped. This allows the netdevice to perform queue
1215 * management safely.
1217 * Fiber Channel over Ethernet (FCoE) offload functions.
1218 * int (*ndo_fcoe_enable)(struct net_device *dev);
1219 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1220 * so the underlying device can perform whatever needed configuration or
1221 * initialization to support acceleration of FCoE traffic.
1223 * int (*ndo_fcoe_disable)(struct net_device *dev);
1224 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1225 * so the underlying device can perform whatever needed clean-ups to
1226 * stop supporting acceleration of FCoE traffic.
1228 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1229 * struct scatterlist *sgl, unsigned int sgc);
1230 * Called when the FCoE Initiator wants to initialize an I/O that
1231 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1232 * perform necessary setup and returns 1 to indicate the device is set up
1233 * successfully to perform DDP on this I/O, otherwise this returns 0.
1235 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1236 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1237 * indicated by the FC exchange id 'xid', so the underlying device can
1238 * clean up and reuse resources for later DDP requests.
1240 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1241 * struct scatterlist *sgl, unsigned int sgc);
1242 * Called when the FCoE Target wants to initialize an I/O that
1243 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1244 * perform necessary setup and returns 1 to indicate the device is set up
1245 * successfully to perform DDP on this I/O, otherwise this returns 0.
1247 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1248 * struct netdev_fcoe_hbainfo *hbainfo);
1249 * Called when the FCoE Protocol stack wants information on the underlying
1250 * device. This information is utilized by the FCoE protocol stack to
1251 * register attributes with Fiber Channel management service as per the
1252 * FC-GS Fabric Device Management Information(FDMI) specification.
1254 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1255 * Called when the underlying device wants to override default World Wide
1256 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1257 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1258 * protocol stack to use.
1261 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1262 * u16 rxq_index, u32 flow_id);
1263 * Set hardware filter for RFS. rxq_index is the target queue index;
1264 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1265 * Return the filter ID on success, or a negative error code.
1267 * Slave management functions (for bridge, bonding, etc).
1268 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1269 * Called to make another netdev an underling.
1271 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1272 * Called to release previously enslaved netdev.
1274 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1275 * struct sk_buff *skb,
1277 * Get the xmit slave of master device. If all_slaves is true, function
1278 * assume all the slaves can transmit.
1280 * Feature/offload setting functions.
1281 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1282 * netdev_features_t features);
1283 * Adjusts the requested feature flags according to device-specific
1284 * constraints, and returns the resulting flags. Must not modify
1287 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1288 * Called to update device configuration to new features. Passed
1289 * feature set might be less than what was returned by ndo_fix_features()).
1290 * Must return >0 or -errno if it changed dev->features itself.
1292 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1293 * struct net_device *dev,
1294 * const unsigned char *addr, u16 vid, u16 flags,
1295 * struct netlink_ext_ack *extack);
1296 * Adds an FDB entry to dev for addr.
1297 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1298 * struct net_device *dev,
1299 * const unsigned char *addr, u16 vid)
1300 * Deletes the FDB entry from dev coresponding to addr.
1301 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1302 * struct net_device *dev,
1304 * struct netlink_ext_ack *extack);
1305 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1306 * struct net_device *dev, struct net_device *filter_dev,
1308 * Used to add FDB entries to dump requests. Implementers should add
1309 * entries to skb and update idx with the number of entries.
1311 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1312 * u16 flags, struct netlink_ext_ack *extack)
1313 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1314 * struct net_device *dev, u32 filter_mask,
1316 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1319 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1320 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1321 * which do not represent real hardware may define this to allow their
1322 * userspace components to manage their virtual carrier state. Devices
1323 * that determine carrier state from physical hardware properties (eg
1324 * network cables) or protocol-dependent mechanisms (eg
1325 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1327 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1328 * struct netdev_phys_item_id *ppid);
1329 * Called to get ID of physical port of this device. If driver does
1330 * not implement this, it is assumed that the hw is not able to have
1331 * multiple net devices on single physical port.
1333 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1334 * struct netdev_phys_item_id *ppid)
1335 * Called to get the parent ID of the physical port of this device.
1337 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1338 * struct net_device *dev)
1339 * Called by upper layer devices to accelerate switching or other
1340 * station functionality into hardware. 'pdev is the lowerdev
1341 * to use for the offload and 'dev' is the net device that will
1342 * back the offload. Returns a pointer to the private structure
1343 * the upper layer will maintain.
1344 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1345 * Called by upper layer device to delete the station created
1346 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1347 * the station and priv is the structure returned by the add
1349 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1350 * int queue_index, u32 maxrate);
1351 * Called when a user wants to set a max-rate limitation of specific
1353 * int (*ndo_get_iflink)(const struct net_device *dev);
1354 * Called to get the iflink value of this device.
1355 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1356 * This function is used to get egress tunnel information for given skb.
1357 * This is useful for retrieving outer tunnel header parameters while
1359 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1360 * This function is used to specify the headroom that the skb must
1361 * consider when allocation skb during packet reception. Setting
1362 * appropriate rx headroom value allows avoiding skb head copy on
1363 * forward. Setting a negative value resets the rx headroom to the
1365 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1366 * This function is used to set or query state related to XDP on the
1367 * netdevice and manage BPF offload. See definition of
1368 * enum bpf_netdev_command for details.
1369 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1371 * This function is used to submit @n XDP packets for transmit on a
1372 * netdevice. Returns number of frames successfully transmitted, frames
1373 * that got dropped are freed/returned via xdp_return_frame().
1374 * Returns negative number, means general error invoking ndo, meaning
1375 * no frames were xmit'ed and core-caller will free all frames.
1376 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1377 * struct xdp_buff *xdp);
1378 * Get the xmit slave of master device based on the xdp_buff.
1379 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1380 * This function is used to wake up the softirq, ksoftirqd or kthread
1381 * responsible for sending and/or receiving packets on a specific
1382 * queue id bound to an AF_XDP socket. The flags field specifies if
1383 * only RX, only Tx, or both should be woken up using the flags
1384 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1385 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1387 * Add, change, delete or get information on an IPv4 tunnel.
1388 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1389 * If a device is paired with a peer device, return the peer instance.
1390 * The caller must be under RCU read context.
1391 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1392 * Get the forwarding path to reach the real device from the HW destination address
1393 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1394 * const struct skb_shared_hwtstamps *hwtstamps,
1396 * Get hardware timestamp based on normal/adjustable time or free running
1397 * cycle counter. This function is required if physical clock supports a
1398 * free running cycle counter.
1400 struct net_device_ops {
1401 int (*ndo_init)(struct net_device *dev);
1402 void (*ndo_uninit)(struct net_device *dev);
1403 int (*ndo_open)(struct net_device *dev);
1404 int (*ndo_stop)(struct net_device *dev);
1405 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1406 struct net_device *dev);
1407 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1408 struct net_device *dev,
1409 netdev_features_t features);
1410 u16 (*ndo_select_queue)(struct net_device *dev,
1411 struct sk_buff *skb,
1412 struct net_device *sb_dev);
1413 void (*ndo_change_rx_flags)(struct net_device *dev,
1415 void (*ndo_set_rx_mode)(struct net_device *dev);
1416 int (*ndo_set_mac_address)(struct net_device *dev,
1418 int (*ndo_validate_addr)(struct net_device *dev);
1419 int (*ndo_do_ioctl)(struct net_device *dev,
1420 struct ifreq *ifr, int cmd);
1421 int (*ndo_eth_ioctl)(struct net_device *dev,
1422 struct ifreq *ifr, int cmd);
1423 int (*ndo_siocbond)(struct net_device *dev,
1424 struct ifreq *ifr, int cmd);
1425 int (*ndo_siocwandev)(struct net_device *dev,
1426 struct if_settings *ifs);
1427 int (*ndo_siocdevprivate)(struct net_device *dev,
1429 void __user *data, int cmd);
1430 int (*ndo_set_config)(struct net_device *dev,
1432 int (*ndo_change_mtu)(struct net_device *dev,
1434 int (*ndo_neigh_setup)(struct net_device *dev,
1435 struct neigh_parms *);
1436 void (*ndo_tx_timeout) (struct net_device *dev,
1437 unsigned int txqueue);
1439 void (*ndo_get_stats64)(struct net_device *dev,
1440 struct rtnl_link_stats64 *storage);
1441 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1442 int (*ndo_get_offload_stats)(int attr_id,
1443 const struct net_device *dev,
1445 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1447 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1448 __be16 proto, u16 vid);
1449 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1450 __be16 proto, u16 vid);
1451 #ifdef CONFIG_NET_POLL_CONTROLLER
1452 void (*ndo_poll_controller)(struct net_device *dev);
1453 int (*ndo_netpoll_setup)(struct net_device *dev,
1454 struct netpoll_info *info);
1455 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1457 int (*ndo_set_vf_mac)(struct net_device *dev,
1458 int queue, u8 *mac);
1459 int (*ndo_set_vf_vlan)(struct net_device *dev,
1460 int queue, u16 vlan,
1461 u8 qos, __be16 proto);
1462 int (*ndo_set_vf_rate)(struct net_device *dev,
1463 int vf, int min_tx_rate,
1465 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1466 int vf, bool setting);
1467 int (*ndo_set_vf_trust)(struct net_device *dev,
1468 int vf, bool setting);
1469 int (*ndo_get_vf_config)(struct net_device *dev,
1471 struct ifla_vf_info *ivf);
1472 int (*ndo_set_vf_link_state)(struct net_device *dev,
1473 int vf, int link_state);
1474 int (*ndo_get_vf_stats)(struct net_device *dev,
1476 struct ifla_vf_stats
1478 int (*ndo_set_vf_port)(struct net_device *dev,
1480 struct nlattr *port[]);
1481 int (*ndo_get_vf_port)(struct net_device *dev,
1482 int vf, struct sk_buff *skb);
1483 int (*ndo_get_vf_guid)(struct net_device *dev,
1485 struct ifla_vf_guid *node_guid,
1486 struct ifla_vf_guid *port_guid);
1487 int (*ndo_set_vf_guid)(struct net_device *dev,
1490 int (*ndo_set_vf_rss_query_en)(
1491 struct net_device *dev,
1492 int vf, bool setting);
1493 int (*ndo_setup_tc)(struct net_device *dev,
1494 enum tc_setup_type type,
1496 #if IS_ENABLED(CONFIG_FCOE)
1497 int (*ndo_fcoe_enable)(struct net_device *dev);
1498 int (*ndo_fcoe_disable)(struct net_device *dev);
1499 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1501 struct scatterlist *sgl,
1503 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1505 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1507 struct scatterlist *sgl,
1509 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1510 struct netdev_fcoe_hbainfo *hbainfo);
1513 #if IS_ENABLED(CONFIG_LIBFCOE)
1514 #define NETDEV_FCOE_WWNN 0
1515 #define NETDEV_FCOE_WWPN 1
1516 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1517 u64 *wwn, int type);
1520 #ifdef CONFIG_RFS_ACCEL
1521 int (*ndo_rx_flow_steer)(struct net_device *dev,
1522 const struct sk_buff *skb,
1526 int (*ndo_add_slave)(struct net_device *dev,
1527 struct net_device *slave_dev,
1528 struct netlink_ext_ack *extack);
1529 int (*ndo_del_slave)(struct net_device *dev,
1530 struct net_device *slave_dev);
1531 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1532 struct sk_buff *skb,
1534 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1536 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1537 netdev_features_t features);
1538 int (*ndo_set_features)(struct net_device *dev,
1539 netdev_features_t features);
1540 int (*ndo_neigh_construct)(struct net_device *dev,
1541 struct neighbour *n);
1542 void (*ndo_neigh_destroy)(struct net_device *dev,
1543 struct neighbour *n);
1545 int (*ndo_fdb_add)(struct ndmsg *ndm,
1546 struct nlattr *tb[],
1547 struct net_device *dev,
1548 const unsigned char *addr,
1551 struct netlink_ext_ack *extack);
1552 int (*ndo_fdb_del)(struct ndmsg *ndm,
1553 struct nlattr *tb[],
1554 struct net_device *dev,
1555 const unsigned char *addr,
1556 u16 vid, struct netlink_ext_ack *extack);
1557 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1558 struct nlattr *tb[],
1559 struct net_device *dev,
1561 struct netlink_ext_ack *extack);
1562 int (*ndo_fdb_dump)(struct sk_buff *skb,
1563 struct netlink_callback *cb,
1564 struct net_device *dev,
1565 struct net_device *filter_dev,
1567 int (*ndo_fdb_get)(struct sk_buff *skb,
1568 struct nlattr *tb[],
1569 struct net_device *dev,
1570 const unsigned char *addr,
1571 u16 vid, u32 portid, u32 seq,
1572 struct netlink_ext_ack *extack);
1573 int (*ndo_bridge_setlink)(struct net_device *dev,
1574 struct nlmsghdr *nlh,
1576 struct netlink_ext_ack *extack);
1577 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1579 struct net_device *dev,
1582 int (*ndo_bridge_dellink)(struct net_device *dev,
1583 struct nlmsghdr *nlh,
1585 int (*ndo_change_carrier)(struct net_device *dev,
1587 int (*ndo_get_phys_port_id)(struct net_device *dev,
1588 struct netdev_phys_item_id *ppid);
1589 int (*ndo_get_port_parent_id)(struct net_device *dev,
1590 struct netdev_phys_item_id *ppid);
1591 int (*ndo_get_phys_port_name)(struct net_device *dev,
1592 char *name, size_t len);
1593 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1594 struct net_device *dev);
1595 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1598 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1601 int (*ndo_get_iflink)(const struct net_device *dev);
1602 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1603 struct sk_buff *skb);
1604 void (*ndo_set_rx_headroom)(struct net_device *dev,
1605 int needed_headroom);
1606 int (*ndo_bpf)(struct net_device *dev,
1607 struct netdev_bpf *bpf);
1608 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1609 struct xdp_frame **xdp,
1611 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1612 struct xdp_buff *xdp);
1613 int (*ndo_xsk_wakeup)(struct net_device *dev,
1614 u32 queue_id, u32 flags);
1615 int (*ndo_tunnel_ctl)(struct net_device *dev,
1616 struct ip_tunnel_parm *p, int cmd);
1617 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1618 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1619 struct net_device_path *path);
1620 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1621 const struct skb_shared_hwtstamps *hwtstamps,
1625 struct xdp_metadata_ops {
1626 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp);
1627 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash,
1628 enum xdp_rss_hash_type *rss_type);
1632 * enum netdev_priv_flags - &struct net_device priv_flags
1634 * These are the &struct net_device, they are only set internally
1635 * by drivers and used in the kernel. These flags are invisible to
1636 * userspace; this means that the order of these flags can change
1637 * during any kernel release.
1639 * You should have a pretty good reason to be extending these flags.
1641 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1642 * @IFF_EBRIDGE: Ethernet bridging device
1643 * @IFF_BONDING: bonding master or slave
1644 * @IFF_ISATAP: ISATAP interface (RFC4214)
1645 * @IFF_WAN_HDLC: WAN HDLC device
1646 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1648 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1649 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1650 * @IFF_MACVLAN_PORT: device used as macvlan port
1651 * @IFF_BRIDGE_PORT: device used as bridge port
1652 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1653 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1654 * @IFF_UNICAST_FLT: Supports unicast filtering
1655 * @IFF_TEAM_PORT: device used as team port
1656 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1657 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1658 * change when it's running
1659 * @IFF_MACVLAN: Macvlan device
1660 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1661 * underlying stacked devices
1662 * @IFF_L3MDEV_MASTER: device is an L3 master device
1663 * @IFF_NO_QUEUE: device can run without qdisc attached
1664 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1665 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1666 * @IFF_TEAM: device is a team device
1667 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1668 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1669 * entity (i.e. the master device for bridged veth)
1670 * @IFF_MACSEC: device is a MACsec device
1671 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1672 * @IFF_FAILOVER: device is a failover master device
1673 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1674 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1675 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1676 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1677 * skb_headlen(skb) == 0 (data starts from frag0)
1678 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1680 enum netdev_priv_flags {
1681 IFF_802_1Q_VLAN = 1<<0,
1685 IFF_WAN_HDLC = 1<<4,
1686 IFF_XMIT_DST_RELEASE = 1<<5,
1687 IFF_DONT_BRIDGE = 1<<6,
1688 IFF_DISABLE_NETPOLL = 1<<7,
1689 IFF_MACVLAN_PORT = 1<<8,
1690 IFF_BRIDGE_PORT = 1<<9,
1691 IFF_OVS_DATAPATH = 1<<10,
1692 IFF_TX_SKB_SHARING = 1<<11,
1693 IFF_UNICAST_FLT = 1<<12,
1694 IFF_TEAM_PORT = 1<<13,
1695 IFF_SUPP_NOFCS = 1<<14,
1696 IFF_LIVE_ADDR_CHANGE = 1<<15,
1697 IFF_MACVLAN = 1<<16,
1698 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1699 IFF_L3MDEV_MASTER = 1<<18,
1700 IFF_NO_QUEUE = 1<<19,
1701 IFF_OPENVSWITCH = 1<<20,
1702 IFF_L3MDEV_SLAVE = 1<<21,
1704 IFF_RXFH_CONFIGURED = 1<<23,
1705 IFF_PHONY_HEADROOM = 1<<24,
1707 IFF_NO_RX_HANDLER = 1<<26,
1708 IFF_FAILOVER = 1<<27,
1709 IFF_FAILOVER_SLAVE = 1<<28,
1710 IFF_L3MDEV_RX_HANDLER = 1<<29,
1711 IFF_NO_ADDRCONF = BIT_ULL(30),
1712 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1713 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1716 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1717 #define IFF_EBRIDGE IFF_EBRIDGE
1718 #define IFF_BONDING IFF_BONDING
1719 #define IFF_ISATAP IFF_ISATAP
1720 #define IFF_WAN_HDLC IFF_WAN_HDLC
1721 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1722 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1723 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1724 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1725 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1726 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1727 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1728 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1729 #define IFF_TEAM_PORT IFF_TEAM_PORT
1730 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1731 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1732 #define IFF_MACVLAN IFF_MACVLAN
1733 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1734 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1735 #define IFF_NO_QUEUE IFF_NO_QUEUE
1736 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1737 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1738 #define IFF_TEAM IFF_TEAM
1739 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1740 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1741 #define IFF_MACSEC IFF_MACSEC
1742 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1743 #define IFF_FAILOVER IFF_FAILOVER
1744 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1745 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1746 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1748 /* Specifies the type of the struct net_device::ml_priv pointer */
1749 enum netdev_ml_priv_type {
1755 * struct net_device - The DEVICE structure.
1757 * Actually, this whole structure is a big mistake. It mixes I/O
1758 * data with strictly "high-level" data, and it has to know about
1759 * almost every data structure used in the INET module.
1761 * @name: This is the first field of the "visible" part of this structure
1762 * (i.e. as seen by users in the "Space.c" file). It is the name
1765 * @name_node: Name hashlist node
1766 * @ifalias: SNMP alias
1767 * @mem_end: Shared memory end
1768 * @mem_start: Shared memory start
1769 * @base_addr: Device I/O address
1770 * @irq: Device IRQ number
1772 * @state: Generic network queuing layer state, see netdev_state_t
1773 * @dev_list: The global list of network devices
1774 * @napi_list: List entry used for polling NAPI devices
1775 * @unreg_list: List entry when we are unregistering the
1776 * device; see the function unregister_netdev
1777 * @close_list: List entry used when we are closing the device
1778 * @ptype_all: Device-specific packet handlers for all protocols
1779 * @ptype_specific: Device-specific, protocol-specific packet handlers
1781 * @adj_list: Directly linked devices, like slaves for bonding
1782 * @features: Currently active device features
1783 * @hw_features: User-changeable features
1785 * @wanted_features: User-requested features
1786 * @vlan_features: Mask of features inheritable by VLAN devices
1788 * @hw_enc_features: Mask of features inherited by encapsulating devices
1789 * This field indicates what encapsulation
1790 * offloads the hardware is capable of doing,
1791 * and drivers will need to set them appropriately.
1793 * @mpls_features: Mask of features inheritable by MPLS
1794 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1796 * @ifindex: interface index
1797 * @group: The group the device belongs to
1799 * @stats: Statistics struct, which was left as a legacy, use
1800 * rtnl_link_stats64 instead
1802 * @core_stats: core networking counters,
1803 * do not use this in drivers
1804 * @carrier_up_count: Number of times the carrier has been up
1805 * @carrier_down_count: Number of times the carrier has been down
1807 * @wireless_handlers: List of functions to handle Wireless Extensions,
1809 * see <net/iw_handler.h> for details.
1810 * @wireless_data: Instance data managed by the core of wireless extensions
1812 * @netdev_ops: Includes several pointers to callbacks,
1813 * if one wants to override the ndo_*() functions
1814 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1815 * @ethtool_ops: Management operations
1816 * @l3mdev_ops: Layer 3 master device operations
1817 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1818 * discovery handling. Necessary for e.g. 6LoWPAN.
1819 * @xfrmdev_ops: Transformation offload operations
1820 * @tlsdev_ops: Transport Layer Security offload operations
1821 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1822 * of Layer 2 headers.
1824 * @flags: Interface flags (a la BSD)
1825 * @xdp_features: XDP capability supported by the device
1826 * @priv_flags: Like 'flags' but invisible to userspace,
1827 * see if.h for the definitions
1828 * @gflags: Global flags ( kept as legacy )
1829 * @padded: How much padding added by alloc_netdev()
1830 * @operstate: RFC2863 operstate
1831 * @link_mode: Mapping policy to operstate
1832 * @if_port: Selectable AUI, TP, ...
1834 * @mtu: Interface MTU value
1835 * @min_mtu: Interface Minimum MTU value
1836 * @max_mtu: Interface Maximum MTU value
1837 * @type: Interface hardware type
1838 * @hard_header_len: Maximum hardware header length.
1839 * @min_header_len: Minimum hardware header length
1841 * @needed_headroom: Extra headroom the hardware may need, but not in all
1842 * cases can this be guaranteed
1843 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1844 * cases can this be guaranteed. Some cases also use
1845 * LL_MAX_HEADER instead to allocate the skb
1847 * interface address info:
1849 * @perm_addr: Permanent hw address
1850 * @addr_assign_type: Hw address assignment type
1851 * @addr_len: Hardware address length
1852 * @upper_level: Maximum depth level of upper devices.
1853 * @lower_level: Maximum depth level of lower devices.
1854 * @neigh_priv_len: Used in neigh_alloc()
1855 * @dev_id: Used to differentiate devices that share
1856 * the same link layer address
1857 * @dev_port: Used to differentiate devices that share
1859 * @addr_list_lock: XXX: need comments on this one
1860 * @name_assign_type: network interface name assignment type
1861 * @uc_promisc: Counter that indicates promiscuous mode
1862 * has been enabled due to the need to listen to
1863 * additional unicast addresses in a device that
1864 * does not implement ndo_set_rx_mode()
1865 * @uc: unicast mac addresses
1866 * @mc: multicast mac addresses
1867 * @dev_addrs: list of device hw addresses
1868 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1869 * @promiscuity: Number of times the NIC is told to work in
1870 * promiscuous mode; if it becomes 0 the NIC will
1871 * exit promiscuous mode
1872 * @allmulti: Counter, enables or disables allmulticast mode
1874 * @vlan_info: VLAN info
1875 * @dsa_ptr: dsa specific data
1876 * @tipc_ptr: TIPC specific data
1877 * @atalk_ptr: AppleTalk link
1878 * @ip_ptr: IPv4 specific data
1879 * @ip6_ptr: IPv6 specific data
1880 * @ax25_ptr: AX.25 specific data
1881 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1882 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1884 * @mpls_ptr: mpls_dev struct pointer
1885 * @mctp_ptr: MCTP specific data
1887 * @dev_addr: Hw address (before bcast,
1888 * because most packets are unicast)
1890 * @_rx: Array of RX queues
1891 * @num_rx_queues: Number of RX queues
1892 * allocated at register_netdev() time
1893 * @real_num_rx_queues: Number of RX queues currently active in device
1894 * @xdp_prog: XDP sockets filter program pointer
1895 * @gro_flush_timeout: timeout for GRO layer in NAPI
1896 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1897 * allow to avoid NIC hard IRQ, on busy queues.
1899 * @rx_handler: handler for received packets
1900 * @rx_handler_data: XXX: need comments on this one
1901 * @miniq_ingress: ingress/clsact qdisc specific data for
1902 * ingress processing
1903 * @ingress_queue: XXX: need comments on this one
1904 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1905 * @broadcast: hw bcast address
1907 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1908 * indexed by RX queue number. Assigned by driver.
1909 * This must only be set if the ndo_rx_flow_steer
1910 * operation is defined
1911 * @index_hlist: Device index hash chain
1913 * @_tx: Array of TX queues
1914 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1915 * @real_num_tx_queues: Number of TX queues currently active in device
1916 * @qdisc: Root qdisc from userspace point of view
1917 * @tx_queue_len: Max frames per queue allowed
1918 * @tx_global_lock: XXX: need comments on this one
1919 * @xdp_bulkq: XDP device bulk queue
1920 * @xps_maps: all CPUs/RXQs maps for XPS device
1922 * @xps_maps: XXX: need comments on this one
1923 * @miniq_egress: clsact qdisc specific data for
1925 * @nf_hooks_egress: netfilter hooks executed for egress packets
1926 * @qdisc_hash: qdisc hash table
1927 * @watchdog_timeo: Represents the timeout that is used by
1928 * the watchdog (see dev_watchdog())
1929 * @watchdog_timer: List of timers
1931 * @proto_down_reason: reason a netdev interface is held down
1932 * @pcpu_refcnt: Number of references to this device
1933 * @dev_refcnt: Number of references to this device
1934 * @refcnt_tracker: Tracker directory for tracked references to this device
1935 * @todo_list: Delayed register/unregister
1936 * @link_watch_list: XXX: need comments on this one
1938 * @reg_state: Register/unregister state machine
1939 * @dismantle: Device is going to be freed
1940 * @rtnl_link_state: This enum represents the phases of creating
1943 * @needs_free_netdev: Should unregister perform free_netdev?
1944 * @priv_destructor: Called from unregister
1945 * @npinfo: XXX: need comments on this one
1946 * @nd_net: Network namespace this network device is inside
1948 * @ml_priv: Mid-layer private
1949 * @ml_priv_type: Mid-layer private type
1950 * @lstats: Loopback statistics
1951 * @tstats: Tunnel statistics
1952 * @dstats: Dummy statistics
1953 * @vstats: Virtual ethernet statistics
1958 * @dm_private: Drop monitor private
1960 * @dev: Class/net/name entry
1961 * @sysfs_groups: Space for optional device, statistics and wireless
1964 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1965 * @rtnl_link_ops: Rtnl_link_ops
1967 * @gso_max_size: Maximum size of generic segmentation offload
1968 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1969 * @gso_max_segs: Maximum number of segments that can be passed to the
1971 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1972 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1975 * @dcbnl_ops: Data Center Bridging netlink ops
1976 * @num_tc: Number of traffic classes in the net device
1977 * @tc_to_txq: XXX: need comments on this one
1978 * @prio_tc_map: XXX: need comments on this one
1980 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1982 * @priomap: XXX: need comments on this one
1983 * @phydev: Physical device may attach itself
1984 * for hardware timestamping
1985 * @sfp_bus: attached &struct sfp_bus structure.
1987 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1989 * @proto_down: protocol port state information can be sent to the
1990 * switch driver and used to set the phys state of the
1993 * @wol_enabled: Wake-on-LAN is enabled
1995 * @threaded: napi threaded mode is enabled
1997 * @net_notifier_list: List of per-net netdev notifier block
1998 * that follow this device when it is moved
1999 * to another network namespace.
2001 * @macsec_ops: MACsec offloading ops
2003 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2004 * offload capabilities of the device
2005 * @udp_tunnel_nic: UDP tunnel offload state
2006 * @xdp_state: stores info on attached XDP BPF programs
2008 * @nested_level: Used as a parameter of spin_lock_nested() of
2009 * dev->addr_list_lock.
2010 * @unlink_list: As netif_addr_lock() can be called recursively,
2011 * keep a list of interfaces to be deleted.
2012 * @gro_max_size: Maximum size of aggregated packet in generic
2013 * receive offload (GRO)
2014 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2015 * receive offload (GRO), for IPv4.
2017 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2018 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2019 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2020 * @dev_registered_tracker: tracker for reference held while
2022 * @offload_xstats_l3: L3 HW stats for this netdevice.
2024 * @devlink_port: Pointer to related devlink port structure.
2025 * Assigned by a driver before netdev registration using
2026 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2027 * during the time netdevice is registered.
2029 * FIXME: cleanup struct net_device such that network protocol info
2034 char name[IFNAMSIZ];
2035 struct netdev_name_node *name_node;
2036 struct dev_ifalias __rcu *ifalias;
2038 * I/O specific fields
2039 * FIXME: Merge these and struct ifmap into one
2041 unsigned long mem_end;
2042 unsigned long mem_start;
2043 unsigned long base_addr;
2046 * Some hardware also needs these fields (state,dev_list,
2047 * napi_list,unreg_list,close_list) but they are not
2048 * part of the usual set specified in Space.c.
2051 unsigned long state;
2053 struct list_head dev_list;
2054 struct list_head napi_list;
2055 struct list_head unreg_list;
2056 struct list_head close_list;
2057 struct list_head ptype_all;
2058 struct list_head ptype_specific;
2061 struct list_head upper;
2062 struct list_head lower;
2065 /* Read-mostly cache-line for fast-path access */
2067 xdp_features_t xdp_features;
2068 unsigned long long priv_flags;
2069 const struct net_device_ops *netdev_ops;
2070 const struct xdp_metadata_ops *xdp_metadata_ops;
2072 unsigned short gflags;
2073 unsigned short hard_header_len;
2075 /* Note : dev->mtu is often read without holding a lock.
2076 * Writers usually hold RTNL.
2077 * It is recommended to use READ_ONCE() to annotate the reads,
2078 * and to use WRITE_ONCE() to annotate the writes.
2081 unsigned short needed_headroom;
2082 unsigned short needed_tailroom;
2084 netdev_features_t features;
2085 netdev_features_t hw_features;
2086 netdev_features_t wanted_features;
2087 netdev_features_t vlan_features;
2088 netdev_features_t hw_enc_features;
2089 netdev_features_t mpls_features;
2090 netdev_features_t gso_partial_features;
2092 unsigned int min_mtu;
2093 unsigned int max_mtu;
2094 unsigned short type;
2095 unsigned char min_header_len;
2096 unsigned char name_assign_type;
2100 struct net_device_stats stats; /* not used by modern drivers */
2102 struct net_device_core_stats __percpu *core_stats;
2104 /* Stats to monitor link on/off, flapping */
2105 atomic_t carrier_up_count;
2106 atomic_t carrier_down_count;
2108 #ifdef CONFIG_WIRELESS_EXT
2109 const struct iw_handler_def *wireless_handlers;
2110 struct iw_public_data *wireless_data;
2112 const struct ethtool_ops *ethtool_ops;
2113 #ifdef CONFIG_NET_L3_MASTER_DEV
2114 const struct l3mdev_ops *l3mdev_ops;
2116 #if IS_ENABLED(CONFIG_IPV6)
2117 const struct ndisc_ops *ndisc_ops;
2120 #ifdef CONFIG_XFRM_OFFLOAD
2121 const struct xfrmdev_ops *xfrmdev_ops;
2124 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2125 const struct tlsdev_ops *tlsdev_ops;
2128 const struct header_ops *header_ops;
2130 unsigned char operstate;
2131 unsigned char link_mode;
2133 unsigned char if_port;
2136 /* Interface address info. */
2137 unsigned char perm_addr[MAX_ADDR_LEN];
2138 unsigned char addr_assign_type;
2139 unsigned char addr_len;
2140 unsigned char upper_level;
2141 unsigned char lower_level;
2143 unsigned short neigh_priv_len;
2144 unsigned short dev_id;
2145 unsigned short dev_port;
2146 unsigned short padded;
2148 spinlock_t addr_list_lock;
2151 struct netdev_hw_addr_list uc;
2152 struct netdev_hw_addr_list mc;
2153 struct netdev_hw_addr_list dev_addrs;
2156 struct kset *queues_kset;
2158 #ifdef CONFIG_LOCKDEP
2159 struct list_head unlink_list;
2161 unsigned int promiscuity;
2162 unsigned int allmulti;
2164 #ifdef CONFIG_LOCKDEP
2165 unsigned char nested_level;
2169 /* Protocol-specific pointers */
2171 struct in_device __rcu *ip_ptr;
2172 struct inet6_dev __rcu *ip6_ptr;
2173 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2174 struct vlan_info __rcu *vlan_info;
2176 #if IS_ENABLED(CONFIG_NET_DSA)
2177 struct dsa_port *dsa_ptr;
2179 #if IS_ENABLED(CONFIG_TIPC)
2180 struct tipc_bearer __rcu *tipc_ptr;
2182 #if IS_ENABLED(CONFIG_ATALK)
2185 #if IS_ENABLED(CONFIG_AX25)
2188 #if IS_ENABLED(CONFIG_CFG80211)
2189 struct wireless_dev *ieee80211_ptr;
2191 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2192 struct wpan_dev *ieee802154_ptr;
2194 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2195 struct mpls_dev __rcu *mpls_ptr;
2197 #if IS_ENABLED(CONFIG_MCTP)
2198 struct mctp_dev __rcu *mctp_ptr;
2202 * Cache lines mostly used on receive path (including eth_type_trans())
2204 /* Interface address info used in eth_type_trans() */
2205 const unsigned char *dev_addr;
2207 struct netdev_rx_queue *_rx;
2208 unsigned int num_rx_queues;
2209 unsigned int real_num_rx_queues;
2211 struct bpf_prog __rcu *xdp_prog;
2212 unsigned long gro_flush_timeout;
2213 int napi_defer_hard_irqs;
2214 #define GRO_LEGACY_MAX_SIZE 65536u
2215 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2216 * and shinfo->gso_segs is a 16bit field.
2218 #define GRO_MAX_SIZE (8 * 65535u)
2219 unsigned int gro_max_size;
2220 unsigned int gro_ipv4_max_size;
2221 rx_handler_func_t __rcu *rx_handler;
2222 void __rcu *rx_handler_data;
2224 #ifdef CONFIG_NET_CLS_ACT
2225 struct mini_Qdisc __rcu *miniq_ingress;
2227 struct netdev_queue __rcu *ingress_queue;
2228 #ifdef CONFIG_NETFILTER_INGRESS
2229 struct nf_hook_entries __rcu *nf_hooks_ingress;
2232 unsigned char broadcast[MAX_ADDR_LEN];
2233 #ifdef CONFIG_RFS_ACCEL
2234 struct cpu_rmap *rx_cpu_rmap;
2236 struct hlist_node index_hlist;
2239 * Cache lines mostly used on transmit path
2241 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2242 unsigned int num_tx_queues;
2243 unsigned int real_num_tx_queues;
2244 struct Qdisc __rcu *qdisc;
2245 unsigned int tx_queue_len;
2246 spinlock_t tx_global_lock;
2248 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2251 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2253 #ifdef CONFIG_NET_CLS_ACT
2254 struct mini_Qdisc __rcu *miniq_egress;
2256 #ifdef CONFIG_NETFILTER_EGRESS
2257 struct nf_hook_entries __rcu *nf_hooks_egress;
2260 #ifdef CONFIG_NET_SCHED
2261 DECLARE_HASHTABLE (qdisc_hash, 4);
2263 /* These may be needed for future network-power-down code. */
2264 struct timer_list watchdog_timer;
2267 u32 proto_down_reason;
2269 struct list_head todo_list;
2271 #ifdef CONFIG_PCPU_DEV_REFCNT
2272 int __percpu *pcpu_refcnt;
2274 refcount_t dev_refcnt;
2276 struct ref_tracker_dir refcnt_tracker;
2278 struct list_head link_watch_list;
2280 enum { NETREG_UNINITIALIZED=0,
2281 NETREG_REGISTERED, /* completed register_netdevice */
2282 NETREG_UNREGISTERING, /* called unregister_netdevice */
2283 NETREG_UNREGISTERED, /* completed unregister todo */
2284 NETREG_RELEASED, /* called free_netdev */
2285 NETREG_DUMMY, /* dummy device for NAPI poll */
2291 RTNL_LINK_INITIALIZED,
2292 RTNL_LINK_INITIALIZING,
2293 } rtnl_link_state:16;
2295 bool needs_free_netdev;
2296 void (*priv_destructor)(struct net_device *dev);
2298 #ifdef CONFIG_NETPOLL
2299 struct netpoll_info __rcu *npinfo;
2302 possible_net_t nd_net;
2304 /* mid-layer private */
2306 enum netdev_ml_priv_type ml_priv_type;
2309 struct pcpu_lstats __percpu *lstats;
2310 struct pcpu_sw_netstats __percpu *tstats;
2311 struct pcpu_dstats __percpu *dstats;
2314 #if IS_ENABLED(CONFIG_GARP)
2315 struct garp_port __rcu *garp_port;
2317 #if IS_ENABLED(CONFIG_MRP)
2318 struct mrp_port __rcu *mrp_port;
2320 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2321 struct dm_hw_stat_delta __rcu *dm_private;
2324 const struct attribute_group *sysfs_groups[4];
2325 const struct attribute_group *sysfs_rx_queue_group;
2327 const struct rtnl_link_ops *rtnl_link_ops;
2329 /* for setting kernel sock attribute on TCP connection setup */
2330 #define GSO_MAX_SEGS 65535u
2331 #define GSO_LEGACY_MAX_SIZE 65536u
2332 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2333 * and shinfo->gso_segs is a 16bit field.
2335 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2337 unsigned int gso_max_size;
2338 #define TSO_LEGACY_MAX_SIZE 65536
2339 #define TSO_MAX_SIZE UINT_MAX
2340 unsigned int tso_max_size;
2342 #define TSO_MAX_SEGS U16_MAX
2344 unsigned int gso_ipv4_max_size;
2347 const struct dcbnl_rtnl_ops *dcbnl_ops;
2350 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2351 u8 prio_tc_map[TC_BITMASK + 1];
2353 #if IS_ENABLED(CONFIG_FCOE)
2354 unsigned int fcoe_ddp_xid;
2356 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2357 struct netprio_map __rcu *priomap;
2359 struct phy_device *phydev;
2360 struct sfp_bus *sfp_bus;
2361 struct lock_class_key *qdisc_tx_busylock;
2363 unsigned wol_enabled:1;
2364 unsigned threaded:1;
2366 struct list_head net_notifier_list;
2368 #if IS_ENABLED(CONFIG_MACSEC)
2369 /* MACsec management functions */
2370 const struct macsec_ops *macsec_ops;
2372 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2373 struct udp_tunnel_nic *udp_tunnel_nic;
2375 /* protected by rtnl_lock */
2376 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2378 u8 dev_addr_shadow[MAX_ADDR_LEN];
2379 netdevice_tracker linkwatch_dev_tracker;
2380 netdevice_tracker watchdog_dev_tracker;
2381 netdevice_tracker dev_registered_tracker;
2382 struct rtnl_hw_stats64 *offload_xstats_l3;
2384 struct devlink_port *devlink_port;
2386 #define to_net_dev(d) container_of(d, struct net_device, dev)
2389 * Driver should use this to assign devlink port instance to a netdevice
2390 * before it registers the netdevice. Therefore devlink_port is static
2391 * during the netdev lifetime after it is registered.
2393 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2395 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2396 ((dev)->devlink_port = (port)); \
2399 static inline bool netif_elide_gro(const struct net_device *dev)
2401 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2406 #define NETDEV_ALIGN 32
2409 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2411 return dev->prio_tc_map[prio & TC_BITMASK];
2415 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2417 if (tc >= dev->num_tc)
2420 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2424 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2425 void netdev_reset_tc(struct net_device *dev);
2426 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2427 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2430 int netdev_get_num_tc(struct net_device *dev)
2435 static inline void net_prefetch(void *p)
2438 #if L1_CACHE_BYTES < 128
2439 prefetch((u8 *)p + L1_CACHE_BYTES);
2443 static inline void net_prefetchw(void *p)
2446 #if L1_CACHE_BYTES < 128
2447 prefetchw((u8 *)p + L1_CACHE_BYTES);
2451 void netdev_unbind_sb_channel(struct net_device *dev,
2452 struct net_device *sb_dev);
2453 int netdev_bind_sb_channel_queue(struct net_device *dev,
2454 struct net_device *sb_dev,
2455 u8 tc, u16 count, u16 offset);
2456 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2457 static inline int netdev_get_sb_channel(struct net_device *dev)
2459 return max_t(int, -dev->num_tc, 0);
2463 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2466 return &dev->_tx[index];
2469 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2470 const struct sk_buff *skb)
2472 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2475 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2476 void (*f)(struct net_device *,
2477 struct netdev_queue *,
2483 for (i = 0; i < dev->num_tx_queues; i++)
2484 f(dev, &dev->_tx[i], arg);
2487 #define netdev_lockdep_set_classes(dev) \
2489 static struct lock_class_key qdisc_tx_busylock_key; \
2490 static struct lock_class_key qdisc_xmit_lock_key; \
2491 static struct lock_class_key dev_addr_list_lock_key; \
2494 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2495 lockdep_set_class(&(dev)->addr_list_lock, \
2496 &dev_addr_list_lock_key); \
2497 for (i = 0; i < (dev)->num_tx_queues; i++) \
2498 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2499 &qdisc_xmit_lock_key); \
2502 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2503 struct net_device *sb_dev);
2504 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2505 struct sk_buff *skb,
2506 struct net_device *sb_dev);
2508 /* returns the headroom that the master device needs to take in account
2509 * when forwarding to this dev
2511 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2513 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2516 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2518 if (dev->netdev_ops->ndo_set_rx_headroom)
2519 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2522 /* set the device rx headroom to the dev's default */
2523 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2525 netdev_set_rx_headroom(dev, -1);
2528 static inline void *netdev_get_ml_priv(struct net_device *dev,
2529 enum netdev_ml_priv_type type)
2531 if (dev->ml_priv_type != type)
2534 return dev->ml_priv;
2537 static inline void netdev_set_ml_priv(struct net_device *dev,
2539 enum netdev_ml_priv_type type)
2541 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2542 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2543 dev->ml_priv_type, type);
2544 WARN(!dev->ml_priv_type && dev->ml_priv,
2545 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2547 dev->ml_priv = ml_priv;
2548 dev->ml_priv_type = type;
2552 * Net namespace inlines
2555 struct net *dev_net(const struct net_device *dev)
2557 return read_pnet(&dev->nd_net);
2561 void dev_net_set(struct net_device *dev, struct net *net)
2563 write_pnet(&dev->nd_net, net);
2567 * netdev_priv - access network device private data
2568 * @dev: network device
2570 * Get network device private data
2572 static inline void *netdev_priv(const struct net_device *dev)
2574 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2577 /* Set the sysfs physical device reference for the network logical device
2578 * if set prior to registration will cause a symlink during initialization.
2580 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2582 /* Set the sysfs device type for the network logical device to allow
2583 * fine-grained identification of different network device types. For
2584 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2586 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2588 /* Default NAPI poll() weight
2589 * Device drivers are strongly advised to not use bigger value
2591 #define NAPI_POLL_WEIGHT 64
2593 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2594 int (*poll)(struct napi_struct *, int), int weight);
2597 * netif_napi_add() - initialize a NAPI context
2598 * @dev: network device
2599 * @napi: NAPI context
2600 * @poll: polling function
2602 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2603 * *any* of the other NAPI-related functions.
2606 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2607 int (*poll)(struct napi_struct *, int))
2609 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2613 netif_napi_add_tx_weight(struct net_device *dev,
2614 struct napi_struct *napi,
2615 int (*poll)(struct napi_struct *, int),
2618 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2619 netif_napi_add_weight(dev, napi, poll, weight);
2623 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2624 * @dev: network device
2625 * @napi: NAPI context
2626 * @poll: polling function
2628 * This variant of netif_napi_add() should be used from drivers using NAPI
2629 * to exclusively poll a TX queue.
2630 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2632 static inline void netif_napi_add_tx(struct net_device *dev,
2633 struct napi_struct *napi,
2634 int (*poll)(struct napi_struct *, int))
2636 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2640 * __netif_napi_del - remove a NAPI context
2641 * @napi: NAPI context
2643 * Warning: caller must observe RCU grace period before freeing memory
2644 * containing @napi. Drivers might want to call this helper to combine
2645 * all the needed RCU grace periods into a single one.
2647 void __netif_napi_del(struct napi_struct *napi);
2650 * netif_napi_del - remove a NAPI context
2651 * @napi: NAPI context
2653 * netif_napi_del() removes a NAPI context from the network device NAPI list
2655 static inline void netif_napi_del(struct napi_struct *napi)
2657 __netif_napi_del(napi);
2661 struct packet_type {
2662 __be16 type; /* This is really htons(ether_type). */
2663 bool ignore_outgoing;
2664 struct net_device *dev; /* NULL is wildcarded here */
2665 netdevice_tracker dev_tracker;
2666 int (*func) (struct sk_buff *,
2667 struct net_device *,
2668 struct packet_type *,
2669 struct net_device *);
2670 void (*list_func) (struct list_head *,
2671 struct packet_type *,
2672 struct net_device *);
2673 bool (*id_match)(struct packet_type *ptype,
2675 struct net *af_packet_net;
2676 void *af_packet_priv;
2677 struct list_head list;
2680 struct offload_callbacks {
2681 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2682 netdev_features_t features);
2683 struct sk_buff *(*gro_receive)(struct list_head *head,
2684 struct sk_buff *skb);
2685 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2688 struct packet_offload {
2689 __be16 type; /* This is really htons(ether_type). */
2691 struct offload_callbacks callbacks;
2692 struct list_head list;
2695 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2696 struct pcpu_sw_netstats {
2697 u64_stats_t rx_packets;
2698 u64_stats_t rx_bytes;
2699 u64_stats_t tx_packets;
2700 u64_stats_t tx_bytes;
2701 struct u64_stats_sync syncp;
2702 } __aligned(4 * sizeof(u64));
2704 struct pcpu_lstats {
2705 u64_stats_t packets;
2707 struct u64_stats_sync syncp;
2708 } __aligned(2 * sizeof(u64));
2710 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2712 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2714 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2716 u64_stats_update_begin(&tstats->syncp);
2717 u64_stats_add(&tstats->rx_bytes, len);
2718 u64_stats_inc(&tstats->rx_packets);
2719 u64_stats_update_end(&tstats->syncp);
2722 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2723 unsigned int packets,
2726 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2728 u64_stats_update_begin(&tstats->syncp);
2729 u64_stats_add(&tstats->tx_bytes, len);
2730 u64_stats_add(&tstats->tx_packets, packets);
2731 u64_stats_update_end(&tstats->syncp);
2734 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2736 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2738 u64_stats_update_begin(&lstats->syncp);
2739 u64_stats_add(&lstats->bytes, len);
2740 u64_stats_inc(&lstats->packets);
2741 u64_stats_update_end(&lstats->syncp);
2744 #define __netdev_alloc_pcpu_stats(type, gfp) \
2746 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2749 for_each_possible_cpu(__cpu) { \
2750 typeof(type) *stat; \
2751 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2752 u64_stats_init(&stat->syncp); \
2758 #define netdev_alloc_pcpu_stats(type) \
2759 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2761 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2763 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2766 for_each_possible_cpu(__cpu) { \
2767 typeof(type) *stat; \
2768 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2769 u64_stats_init(&stat->syncp); \
2775 enum netdev_lag_tx_type {
2776 NETDEV_LAG_TX_TYPE_UNKNOWN,
2777 NETDEV_LAG_TX_TYPE_RANDOM,
2778 NETDEV_LAG_TX_TYPE_BROADCAST,
2779 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2780 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2781 NETDEV_LAG_TX_TYPE_HASH,
2784 enum netdev_lag_hash {
2785 NETDEV_LAG_HASH_NONE,
2787 NETDEV_LAG_HASH_L34,
2788 NETDEV_LAG_HASH_L23,
2789 NETDEV_LAG_HASH_E23,
2790 NETDEV_LAG_HASH_E34,
2791 NETDEV_LAG_HASH_VLAN_SRCMAC,
2792 NETDEV_LAG_HASH_UNKNOWN,
2795 struct netdev_lag_upper_info {
2796 enum netdev_lag_tx_type tx_type;
2797 enum netdev_lag_hash hash_type;
2800 struct netdev_lag_lower_state_info {
2805 #include <linux/notifier.h>
2807 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2808 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2812 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2814 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2815 detected a hardware crash and restarted
2816 - we can use this eg to kick tcp sessions
2818 NETDEV_CHANGE, /* Notify device state change */
2821 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2822 NETDEV_CHANGEADDR, /* notify after the address change */
2823 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2827 NETDEV_BONDING_FAILOVER,
2829 NETDEV_PRE_TYPE_CHANGE,
2830 NETDEV_POST_TYPE_CHANGE,
2834 NETDEV_NOTIFY_PEERS,
2838 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2839 NETDEV_CHANGEINFODATA,
2840 NETDEV_BONDING_INFO,
2841 NETDEV_PRECHANGEUPPER,
2842 NETDEV_CHANGELOWERSTATE,
2843 NETDEV_UDP_TUNNEL_PUSH_INFO,
2844 NETDEV_UDP_TUNNEL_DROP_INFO,
2845 NETDEV_CHANGE_TX_QUEUE_LEN,
2846 NETDEV_CVLAN_FILTER_PUSH_INFO,
2847 NETDEV_CVLAN_FILTER_DROP_INFO,
2848 NETDEV_SVLAN_FILTER_PUSH_INFO,
2849 NETDEV_SVLAN_FILTER_DROP_INFO,
2850 NETDEV_OFFLOAD_XSTATS_ENABLE,
2851 NETDEV_OFFLOAD_XSTATS_DISABLE,
2852 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2853 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2854 NETDEV_XDP_FEAT_CHANGE,
2856 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2858 int register_netdevice_notifier(struct notifier_block *nb);
2859 int unregister_netdevice_notifier(struct notifier_block *nb);
2860 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2861 int unregister_netdevice_notifier_net(struct net *net,
2862 struct notifier_block *nb);
2863 int register_netdevice_notifier_dev_net(struct net_device *dev,
2864 struct notifier_block *nb,
2865 struct netdev_net_notifier *nn);
2866 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2867 struct notifier_block *nb,
2868 struct netdev_net_notifier *nn);
2870 struct netdev_notifier_info {
2871 struct net_device *dev;
2872 struct netlink_ext_ack *extack;
2875 struct netdev_notifier_info_ext {
2876 struct netdev_notifier_info info; /* must be first */
2882 struct netdev_notifier_change_info {
2883 struct netdev_notifier_info info; /* must be first */
2884 unsigned int flags_changed;
2887 struct netdev_notifier_changeupper_info {
2888 struct netdev_notifier_info info; /* must be first */
2889 struct net_device *upper_dev; /* new upper dev */
2890 bool master; /* is upper dev master */
2891 bool linking; /* is the notification for link or unlink */
2892 void *upper_info; /* upper dev info */
2895 struct netdev_notifier_changelowerstate_info {
2896 struct netdev_notifier_info info; /* must be first */
2897 void *lower_state_info; /* is lower dev state */
2900 struct netdev_notifier_pre_changeaddr_info {
2901 struct netdev_notifier_info info; /* must be first */
2902 const unsigned char *dev_addr;
2905 enum netdev_offload_xstats_type {
2906 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2909 struct netdev_notifier_offload_xstats_info {
2910 struct netdev_notifier_info info; /* must be first */
2911 enum netdev_offload_xstats_type type;
2914 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2915 struct netdev_notifier_offload_xstats_rd *report_delta;
2916 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2917 struct netdev_notifier_offload_xstats_ru *report_used;
2921 int netdev_offload_xstats_enable(struct net_device *dev,
2922 enum netdev_offload_xstats_type type,
2923 struct netlink_ext_ack *extack);
2924 int netdev_offload_xstats_disable(struct net_device *dev,
2925 enum netdev_offload_xstats_type type);
2926 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2927 enum netdev_offload_xstats_type type);
2928 int netdev_offload_xstats_get(struct net_device *dev,
2929 enum netdev_offload_xstats_type type,
2930 struct rtnl_hw_stats64 *stats, bool *used,
2931 struct netlink_ext_ack *extack);
2933 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2934 const struct rtnl_hw_stats64 *stats);
2936 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2937 void netdev_offload_xstats_push_delta(struct net_device *dev,
2938 enum netdev_offload_xstats_type type,
2939 const struct rtnl_hw_stats64 *stats);
2941 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2942 struct net_device *dev)
2945 info->extack = NULL;
2948 static inline struct net_device *
2949 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2954 static inline struct netlink_ext_ack *
2955 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2957 return info->extack;
2960 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2963 extern rwlock_t dev_base_lock; /* Device list lock */
2965 #define for_each_netdev(net, d) \
2966 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2967 #define for_each_netdev_reverse(net, d) \
2968 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2969 #define for_each_netdev_rcu(net, d) \
2970 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2971 #define for_each_netdev_safe(net, d, n) \
2972 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2973 #define for_each_netdev_continue(net, d) \
2974 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2975 #define for_each_netdev_continue_reverse(net, d) \
2976 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2978 #define for_each_netdev_continue_rcu(net, d) \
2979 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2980 #define for_each_netdev_in_bond_rcu(bond, slave) \
2981 for_each_netdev_rcu(&init_net, slave) \
2982 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2983 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2985 static inline struct net_device *next_net_device(struct net_device *dev)
2987 struct list_head *lh;
2991 lh = dev->dev_list.next;
2992 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2995 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2997 struct list_head *lh;
3001 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3002 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3005 static inline struct net_device *first_net_device(struct net *net)
3007 return list_empty(&net->dev_base_head) ? NULL :
3008 net_device_entry(net->dev_base_head.next);
3011 static inline struct net_device *first_net_device_rcu(struct net *net)
3013 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3015 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3018 int netdev_boot_setup_check(struct net_device *dev);
3019 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3020 const char *hwaddr);
3021 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3022 void dev_add_pack(struct packet_type *pt);
3023 void dev_remove_pack(struct packet_type *pt);
3024 void __dev_remove_pack(struct packet_type *pt);
3025 void dev_add_offload(struct packet_offload *po);
3026 void dev_remove_offload(struct packet_offload *po);
3028 int dev_get_iflink(const struct net_device *dev);
3029 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3030 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3031 struct net_device_path_stack *stack);
3032 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3033 unsigned short mask);
3034 struct net_device *dev_get_by_name(struct net *net, const char *name);
3035 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3036 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3037 bool netdev_name_in_use(struct net *net, const char *name);
3038 int dev_alloc_name(struct net_device *dev, const char *name);
3039 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3040 void dev_close(struct net_device *dev);
3041 void dev_close_many(struct list_head *head, bool unlink);
3042 void dev_disable_lro(struct net_device *dev);
3043 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3044 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3045 struct net_device *sb_dev);
3046 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3047 struct net_device *sb_dev);
3049 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3050 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3052 static inline int dev_queue_xmit(struct sk_buff *skb)
3054 return __dev_queue_xmit(skb, NULL);
3057 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3058 struct net_device *sb_dev)
3060 return __dev_queue_xmit(skb, sb_dev);
3063 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3067 ret = __dev_direct_xmit(skb, queue_id);
3068 if (!dev_xmit_complete(ret))
3073 int register_netdevice(struct net_device *dev);
3074 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3075 void unregister_netdevice_many(struct list_head *head);
3076 static inline void unregister_netdevice(struct net_device *dev)
3078 unregister_netdevice_queue(dev, NULL);
3081 int netdev_refcnt_read(const struct net_device *dev);
3082 void free_netdev(struct net_device *dev);
3083 void netdev_freemem(struct net_device *dev);
3084 int init_dummy_netdev(struct net_device *dev);
3086 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3087 struct sk_buff *skb,
3089 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3091 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3092 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3093 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3094 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3095 int dev_restart(struct net_device *dev);
3098 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3099 unsigned short type,
3100 const void *daddr, const void *saddr,
3103 if (!dev->header_ops || !dev->header_ops->create)
3106 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3109 static inline int dev_parse_header(const struct sk_buff *skb,
3110 unsigned char *haddr)
3112 const struct net_device *dev = skb->dev;
3114 if (!dev->header_ops || !dev->header_ops->parse)
3116 return dev->header_ops->parse(skb, haddr);
3119 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3121 const struct net_device *dev = skb->dev;
3123 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3125 return dev->header_ops->parse_protocol(skb);
3128 /* ll_header must have at least hard_header_len allocated */
3129 static inline bool dev_validate_header(const struct net_device *dev,
3130 char *ll_header, int len)
3132 if (likely(len >= dev->hard_header_len))
3134 if (len < dev->min_header_len)
3137 if (capable(CAP_SYS_RAWIO)) {
3138 memset(ll_header + len, 0, dev->hard_header_len - len);
3142 if (dev->header_ops && dev->header_ops->validate)
3143 return dev->header_ops->validate(ll_header, len);
3148 static inline bool dev_has_header(const struct net_device *dev)
3150 return dev->header_ops && dev->header_ops->create;
3154 * Incoming packets are placed on per-CPU queues
3156 struct softnet_data {
3157 struct list_head poll_list;
3158 struct sk_buff_head process_queue;
3161 unsigned int processed;
3162 unsigned int time_squeeze;
3164 struct softnet_data *rps_ipi_list;
3166 #ifdef CONFIG_NET_FLOW_LIMIT
3167 struct sd_flow_limit __rcu *flow_limit;
3169 struct Qdisc *output_queue;
3170 struct Qdisc **output_queue_tailp;
3171 struct sk_buff *completion_queue;
3172 #ifdef CONFIG_XFRM_OFFLOAD
3173 struct sk_buff_head xfrm_backlog;
3175 /* written and read only by owning cpu: */
3179 #ifdef CONFIG_NET_EGRESS
3184 /* input_queue_head should be written by cpu owning this struct,
3185 * and only read by other cpus. Worth using a cache line.
3187 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3189 /* Elements below can be accessed between CPUs for RPS/RFS */
3190 call_single_data_t csd ____cacheline_aligned_in_smp;
3191 struct softnet_data *rps_ipi_next;
3193 unsigned int input_queue_tail;
3195 unsigned int received_rps;
3196 unsigned int dropped;
3197 struct sk_buff_head input_pkt_queue;
3198 struct napi_struct backlog;
3200 /* Another possibly contended cache line */
3201 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3203 int defer_ipi_scheduled;
3204 struct sk_buff *defer_list;
3205 call_single_data_t defer_csd;
3208 static inline void input_queue_head_incr(struct softnet_data *sd)
3211 sd->input_queue_head++;
3215 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3216 unsigned int *qtail)
3219 *qtail = ++sd->input_queue_tail;
3223 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3225 static inline int dev_recursion_level(void)
3227 return this_cpu_read(softnet_data.xmit.recursion);
3230 #define XMIT_RECURSION_LIMIT 8
3231 static inline bool dev_xmit_recursion(void)
3233 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3234 XMIT_RECURSION_LIMIT);
3237 static inline void dev_xmit_recursion_inc(void)
3239 __this_cpu_inc(softnet_data.xmit.recursion);
3242 static inline void dev_xmit_recursion_dec(void)
3244 __this_cpu_dec(softnet_data.xmit.recursion);
3247 void __netif_schedule(struct Qdisc *q);
3248 void netif_schedule_queue(struct netdev_queue *txq);
3250 static inline void netif_tx_schedule_all(struct net_device *dev)
3254 for (i = 0; i < dev->num_tx_queues; i++)
3255 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3258 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3260 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3264 * netif_start_queue - allow transmit
3265 * @dev: network device
3267 * Allow upper layers to call the device hard_start_xmit routine.
3269 static inline void netif_start_queue(struct net_device *dev)
3271 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3274 static inline void netif_tx_start_all_queues(struct net_device *dev)
3278 for (i = 0; i < dev->num_tx_queues; i++) {
3279 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3280 netif_tx_start_queue(txq);
3284 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3287 * netif_wake_queue - restart transmit
3288 * @dev: network device
3290 * Allow upper layers to call the device hard_start_xmit routine.
3291 * Used for flow control when transmit resources are available.
3293 static inline void netif_wake_queue(struct net_device *dev)
3295 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3298 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3302 for (i = 0; i < dev->num_tx_queues; i++) {
3303 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3304 netif_tx_wake_queue(txq);
3308 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3310 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3314 * netif_stop_queue - stop transmitted packets
3315 * @dev: network device
3317 * Stop upper layers calling the device hard_start_xmit routine.
3318 * Used for flow control when transmit resources are unavailable.
3320 static inline void netif_stop_queue(struct net_device *dev)
3322 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3325 void netif_tx_stop_all_queues(struct net_device *dev);
3327 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3329 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3333 * netif_queue_stopped - test if transmit queue is flowblocked
3334 * @dev: network device
3336 * Test if transmit queue on device is currently unable to send.
3338 static inline bool netif_queue_stopped(const struct net_device *dev)
3340 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3343 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3345 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3349 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3351 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3355 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3357 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3361 * netdev_queue_set_dql_min_limit - set dql minimum limit
3362 * @dev_queue: pointer to transmit queue
3363 * @min_limit: dql minimum limit
3365 * Forces xmit_more() to return true until the minimum threshold
3366 * defined by @min_limit is reached (or until the tx queue is
3367 * empty). Warning: to be use with care, misuse will impact the
3370 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3371 unsigned int min_limit)
3374 dev_queue->dql.min_limit = min_limit;
3379 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3380 * @dev_queue: pointer to transmit queue
3382 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3383 * to give appropriate hint to the CPU.
3385 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3388 prefetchw(&dev_queue->dql.num_queued);
3393 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3394 * @dev_queue: pointer to transmit queue
3396 * BQL enabled drivers might use this helper in their TX completion path,
3397 * to give appropriate hint to the CPU.
3399 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3402 prefetchw(&dev_queue->dql.limit);
3407 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3408 * @dev_queue: network device queue
3409 * @bytes: number of bytes queued to the device queue
3411 * Report the number of bytes queued for sending/completion to the network
3412 * device hardware queue. @bytes should be a good approximation and should
3413 * exactly match netdev_completed_queue() @bytes.
3414 * This is typically called once per packet, from ndo_start_xmit().
3416 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3420 dql_queued(&dev_queue->dql, bytes);
3422 if (likely(dql_avail(&dev_queue->dql) >= 0))
3425 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3428 * The XOFF flag must be set before checking the dql_avail below,
3429 * because in netdev_tx_completed_queue we update the dql_completed
3430 * before checking the XOFF flag.
3434 /* check again in case another CPU has just made room avail */
3435 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3436 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3440 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3441 * that they should not test BQL status themselves.
3442 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3444 * Returns true if the doorbell must be used to kick the NIC.
3446 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3452 dql_queued(&dev_queue->dql, bytes);
3454 return netif_tx_queue_stopped(dev_queue);
3456 netdev_tx_sent_queue(dev_queue, bytes);
3461 * netdev_sent_queue - report the number of bytes queued to hardware
3462 * @dev: network device
3463 * @bytes: number of bytes queued to the hardware device queue
3465 * Report the number of bytes queued for sending/completion to the network
3466 * device hardware queue#0. @bytes should be a good approximation and should
3467 * exactly match netdev_completed_queue() @bytes.
3468 * This is typically called once per packet, from ndo_start_xmit().
3470 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3472 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3475 static inline bool __netdev_sent_queue(struct net_device *dev,
3479 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3484 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3485 * @dev_queue: network device queue
3486 * @pkts: number of packets (currently ignored)
3487 * @bytes: number of bytes dequeued from the device queue
3489 * Must be called at most once per TX completion round (and not per
3490 * individual packet), so that BQL can adjust its limits appropriately.
3492 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3493 unsigned int pkts, unsigned int bytes)
3496 if (unlikely(!bytes))
3499 dql_completed(&dev_queue->dql, bytes);
3502 * Without the memory barrier there is a small possiblity that
3503 * netdev_tx_sent_queue will miss the update and cause the queue to
3504 * be stopped forever
3508 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3511 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3512 netif_schedule_queue(dev_queue);
3517 * netdev_completed_queue - report bytes and packets completed by device
3518 * @dev: network device
3519 * @pkts: actual number of packets sent over the medium
3520 * @bytes: actual number of bytes sent over the medium
3522 * Report the number of bytes and packets transmitted by the network device
3523 * hardware queue over the physical medium, @bytes must exactly match the
3524 * @bytes amount passed to netdev_sent_queue()
3526 static inline void netdev_completed_queue(struct net_device *dev,
3527 unsigned int pkts, unsigned int bytes)
3529 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3532 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3535 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3541 * netdev_reset_queue - reset the packets and bytes count of a network device
3542 * @dev_queue: network device
3544 * Reset the bytes and packet count of a network device and clear the
3545 * software flow control OFF bit for this network device
3547 static inline void netdev_reset_queue(struct net_device *dev_queue)
3549 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3553 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3554 * @dev: network device
3555 * @queue_index: given tx queue index
3557 * Returns 0 if given tx queue index >= number of device tx queues,
3558 * otherwise returns the originally passed tx queue index.
3560 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3562 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3563 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3564 dev->name, queue_index,
3565 dev->real_num_tx_queues);
3573 * netif_running - test if up
3574 * @dev: network device
3576 * Test if the device has been brought up.
3578 static inline bool netif_running(const struct net_device *dev)
3580 return test_bit(__LINK_STATE_START, &dev->state);
3584 * Routines to manage the subqueues on a device. We only need start,
3585 * stop, and a check if it's stopped. All other device management is
3586 * done at the overall netdevice level.
3587 * Also test the device if we're multiqueue.
3591 * netif_start_subqueue - allow sending packets on subqueue
3592 * @dev: network device
3593 * @queue_index: sub queue index
3595 * Start individual transmit queue of a device with multiple transmit queues.
3597 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3599 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3601 netif_tx_start_queue(txq);
3605 * netif_stop_subqueue - stop sending packets on subqueue
3606 * @dev: network device
3607 * @queue_index: sub queue index
3609 * Stop individual transmit queue of a device with multiple transmit queues.
3611 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3613 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3614 netif_tx_stop_queue(txq);
3618 * __netif_subqueue_stopped - test status of subqueue
3619 * @dev: network device
3620 * @queue_index: sub queue index
3622 * Check individual transmit queue of a device with multiple transmit queues.
3624 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3627 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3629 return netif_tx_queue_stopped(txq);
3633 * netif_subqueue_stopped - test status of subqueue
3634 * @dev: network device
3635 * @skb: sub queue buffer pointer
3637 * Check individual transmit queue of a device with multiple transmit queues.
3639 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3640 struct sk_buff *skb)
3642 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3646 * netif_wake_subqueue - allow sending packets on subqueue
3647 * @dev: network device
3648 * @queue_index: sub queue index
3650 * Resume individual transmit queue of a device with multiple transmit queues.
3652 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3654 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3656 netif_tx_wake_queue(txq);
3660 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3662 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3663 u16 index, enum xps_map_type type);
3666 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3667 * @j: CPU/Rx queue index
3668 * @mask: bitmask of all cpus/rx queues
3669 * @nr_bits: number of bits in the bitmask
3671 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3673 static inline bool netif_attr_test_mask(unsigned long j,
3674 const unsigned long *mask,
3675 unsigned int nr_bits)
3677 cpu_max_bits_warn(j, nr_bits);
3678 return test_bit(j, mask);
3682 * netif_attr_test_online - Test for online CPU/Rx queue
3683 * @j: CPU/Rx queue index
3684 * @online_mask: bitmask for CPUs/Rx queues that are online
3685 * @nr_bits: number of bits in the bitmask
3687 * Returns true if a CPU/Rx queue is online.
3689 static inline bool netif_attr_test_online(unsigned long j,
3690 const unsigned long *online_mask,
3691 unsigned int nr_bits)
3693 cpu_max_bits_warn(j, nr_bits);
3696 return test_bit(j, online_mask);
3698 return (j < nr_bits);
3702 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3703 * @n: CPU/Rx queue index
3704 * @srcp: the cpumask/Rx queue mask pointer
3705 * @nr_bits: number of bits in the bitmask
3707 * Returns >= nr_bits if no further CPUs/Rx queues set.
3709 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3710 unsigned int nr_bits)
3712 /* -1 is a legal arg here. */
3714 cpu_max_bits_warn(n, nr_bits);
3717 return find_next_bit(srcp, nr_bits, n + 1);
3723 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3724 * @n: CPU/Rx queue index
3725 * @src1p: the first CPUs/Rx queues mask pointer
3726 * @src2p: the second CPUs/Rx queues mask pointer
3727 * @nr_bits: number of bits in the bitmask
3729 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3731 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3732 const unsigned long *src2p,
3733 unsigned int nr_bits)
3735 /* -1 is a legal arg here. */
3737 cpu_max_bits_warn(n, nr_bits);
3740 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3742 return find_next_bit(src1p, nr_bits, n + 1);
3744 return find_next_bit(src2p, nr_bits, n + 1);
3749 static inline int netif_set_xps_queue(struct net_device *dev,
3750 const struct cpumask *mask,
3756 static inline int __netif_set_xps_queue(struct net_device *dev,
3757 const unsigned long *mask,
3758 u16 index, enum xps_map_type type)
3765 * netif_is_multiqueue - test if device has multiple transmit queues
3766 * @dev: network device
3768 * Check if device has multiple transmit queues
3770 static inline bool netif_is_multiqueue(const struct net_device *dev)
3772 return dev->num_tx_queues > 1;
3775 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3778 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3780 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3783 dev->real_num_rx_queues = rxqs;
3787 int netif_set_real_num_queues(struct net_device *dev,
3788 unsigned int txq, unsigned int rxq);
3790 static inline struct netdev_rx_queue *
3791 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3793 return dev->_rx + rxq;
3797 static inline unsigned int get_netdev_rx_queue_index(
3798 struct netdev_rx_queue *queue)
3800 struct net_device *dev = queue->dev;
3801 int index = queue - dev->_rx;
3803 BUG_ON(index >= dev->num_rx_queues);
3808 int netif_get_num_default_rss_queues(void);
3810 enum skb_free_reason {
3811 SKB_REASON_CONSUMED,
3815 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3816 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3819 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3820 * interrupt context or with hardware interrupts being disabled.
3821 * (in_hardirq() || irqs_disabled())
3823 * We provide four helpers that can be used in following contexts :
3825 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3826 * replacing kfree_skb(skb)
3828 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3829 * Typically used in place of consume_skb(skb) in TX completion path
3831 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3832 * replacing kfree_skb(skb)
3834 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3835 * and consumed a packet. Used in place of consume_skb(skb)
3837 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3839 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3842 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3844 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3847 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3849 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3852 static inline void dev_consume_skb_any(struct sk_buff *skb)
3854 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3857 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3858 struct bpf_prog *xdp_prog);
3859 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3860 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3861 int netif_rx(struct sk_buff *skb);
3862 int __netif_rx(struct sk_buff *skb);
3864 int netif_receive_skb(struct sk_buff *skb);
3865 int netif_receive_skb_core(struct sk_buff *skb);
3866 void netif_receive_skb_list_internal(struct list_head *head);
3867 void netif_receive_skb_list(struct list_head *head);
3868 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3869 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3870 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3871 void napi_get_frags_check(struct napi_struct *napi);
3872 gro_result_t napi_gro_frags(struct napi_struct *napi);
3873 struct packet_offload *gro_find_receive_by_type(__be16 type);
3874 struct packet_offload *gro_find_complete_by_type(__be16 type);
3876 static inline void napi_free_frags(struct napi_struct *napi)
3878 kfree_skb(napi->skb);
3882 bool netdev_is_rx_handler_busy(struct net_device *dev);
3883 int netdev_rx_handler_register(struct net_device *dev,
3884 rx_handler_func_t *rx_handler,
3885 void *rx_handler_data);
3886 void netdev_rx_handler_unregister(struct net_device *dev);
3888 bool dev_valid_name(const char *name);
3889 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3891 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3893 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3894 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3895 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3896 void __user *data, bool *need_copyout);
3897 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3898 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3899 unsigned int dev_get_flags(const struct net_device *);
3900 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3901 struct netlink_ext_ack *extack);
3902 int dev_change_flags(struct net_device *dev, unsigned int flags,
3903 struct netlink_ext_ack *extack);
3904 int dev_set_alias(struct net_device *, const char *, size_t);
3905 int dev_get_alias(const struct net_device *, char *, size_t);
3906 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3907 const char *pat, int new_ifindex);
3909 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3912 return __dev_change_net_namespace(dev, net, pat, 0);
3914 int __dev_set_mtu(struct net_device *, int);
3915 int dev_set_mtu(struct net_device *, int);
3916 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3917 struct netlink_ext_ack *extack);
3918 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3919 struct netlink_ext_ack *extack);
3920 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3921 struct netlink_ext_ack *extack);
3922 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3923 int dev_get_port_parent_id(struct net_device *dev,
3924 struct netdev_phys_item_id *ppid, bool recurse);
3925 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3926 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3927 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3928 struct netdev_queue *txq, int *ret);
3930 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3931 u8 dev_xdp_prog_count(struct net_device *dev);
3932 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3934 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3935 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3936 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3937 bool is_skb_forwardable(const struct net_device *dev,
3938 const struct sk_buff *skb);
3940 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3941 const struct sk_buff *skb,
3942 const bool check_mtu)
3944 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3947 if (!(dev->flags & IFF_UP))
3953 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3954 if (skb->len <= len)
3957 /* if TSO is enabled, we don't care about the length as the packet
3958 * could be forwarded without being segmented before
3960 if (skb_is_gso(skb))
3966 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3968 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3970 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3971 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3976 return netdev_core_stats_alloc(dev);
3979 #define DEV_CORE_STATS_INC(FIELD) \
3980 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3982 struct net_device_core_stats __percpu *p; \
3984 p = dev_core_stats(dev); \
3986 this_cpu_inc(p->FIELD); \
3988 DEV_CORE_STATS_INC(rx_dropped)
3989 DEV_CORE_STATS_INC(tx_dropped)
3990 DEV_CORE_STATS_INC(rx_nohandler)
3991 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3993 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3994 struct sk_buff *skb,
3995 const bool check_mtu)
3997 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3998 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3999 dev_core_stats_rx_dropped_inc(dev);
4004 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4009 bool dev_nit_active(struct net_device *dev);
4010 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4012 static inline void __dev_put(struct net_device *dev)
4015 #ifdef CONFIG_PCPU_DEV_REFCNT
4016 this_cpu_dec(*dev->pcpu_refcnt);
4018 refcount_dec(&dev->dev_refcnt);
4023 static inline void __dev_hold(struct net_device *dev)
4026 #ifdef CONFIG_PCPU_DEV_REFCNT
4027 this_cpu_inc(*dev->pcpu_refcnt);
4029 refcount_inc(&dev->dev_refcnt);
4034 static inline void __netdev_tracker_alloc(struct net_device *dev,
4035 netdevice_tracker *tracker,
4038 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4039 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4043 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4044 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4046 static inline void netdev_tracker_alloc(struct net_device *dev,
4047 netdevice_tracker *tracker, gfp_t gfp)
4049 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4050 refcount_dec(&dev->refcnt_tracker.no_tracker);
4051 __netdev_tracker_alloc(dev, tracker, gfp);
4055 static inline void netdev_tracker_free(struct net_device *dev,
4056 netdevice_tracker *tracker)
4058 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4059 ref_tracker_free(&dev->refcnt_tracker, tracker);
4063 static inline void netdev_hold(struct net_device *dev,
4064 netdevice_tracker *tracker, gfp_t gfp)
4068 __netdev_tracker_alloc(dev, tracker, gfp);
4072 static inline void netdev_put(struct net_device *dev,
4073 netdevice_tracker *tracker)
4076 netdev_tracker_free(dev, tracker);
4082 * dev_hold - get reference to device
4083 * @dev: network device
4085 * Hold reference to device to keep it from being freed.
4086 * Try using netdev_hold() instead.
4088 static inline void dev_hold(struct net_device *dev)
4090 netdev_hold(dev, NULL, GFP_ATOMIC);
4094 * dev_put - release reference to device
4095 * @dev: network device
4097 * Release reference to device to allow it to be freed.
4098 * Try using netdev_put() instead.
4100 static inline void dev_put(struct net_device *dev)
4102 netdev_put(dev, NULL);
4105 static inline void netdev_ref_replace(struct net_device *odev,
4106 struct net_device *ndev,
4107 netdevice_tracker *tracker,
4111 netdev_tracker_free(odev, tracker);
4117 __netdev_tracker_alloc(ndev, tracker, gfp);
4120 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4121 * and _off may be called from IRQ context, but it is caller
4122 * who is responsible for serialization of these calls.
4124 * The name carrier is inappropriate, these functions should really be
4125 * called netif_lowerlayer_*() because they represent the state of any
4126 * kind of lower layer not just hardware media.
4128 void linkwatch_fire_event(struct net_device *dev);
4131 * netif_carrier_ok - test if carrier present
4132 * @dev: network device
4134 * Check if carrier is present on device
4136 static inline bool netif_carrier_ok(const struct net_device *dev)
4138 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4141 unsigned long dev_trans_start(struct net_device *dev);
4143 void __netdev_watchdog_up(struct net_device *dev);
4145 void netif_carrier_on(struct net_device *dev);
4146 void netif_carrier_off(struct net_device *dev);
4147 void netif_carrier_event(struct net_device *dev);
4150 * netif_dormant_on - mark device as dormant.
4151 * @dev: network device
4153 * Mark device as dormant (as per RFC2863).
4155 * The dormant state indicates that the relevant interface is not
4156 * actually in a condition to pass packets (i.e., it is not 'up') but is
4157 * in a "pending" state, waiting for some external event. For "on-
4158 * demand" interfaces, this new state identifies the situation where the
4159 * interface is waiting for events to place it in the up state.
4161 static inline void netif_dormant_on(struct net_device *dev)
4163 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4164 linkwatch_fire_event(dev);
4168 * netif_dormant_off - set device as not dormant.
4169 * @dev: network device
4171 * Device is not in dormant state.
4173 static inline void netif_dormant_off(struct net_device *dev)
4175 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4176 linkwatch_fire_event(dev);
4180 * netif_dormant - test if device is dormant
4181 * @dev: network device
4183 * Check if device is dormant.
4185 static inline bool netif_dormant(const struct net_device *dev)
4187 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4192 * netif_testing_on - mark device as under test.
4193 * @dev: network device
4195 * Mark device as under test (as per RFC2863).
4197 * The testing state indicates that some test(s) must be performed on
4198 * the interface. After completion, of the test, the interface state
4199 * will change to up, dormant, or down, as appropriate.
4201 static inline void netif_testing_on(struct net_device *dev)
4203 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4204 linkwatch_fire_event(dev);
4208 * netif_testing_off - set device as not under test.
4209 * @dev: network device
4211 * Device is not in testing state.
4213 static inline void netif_testing_off(struct net_device *dev)
4215 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4216 linkwatch_fire_event(dev);
4220 * netif_testing - test if device is under test
4221 * @dev: network device
4223 * Check if device is under test
4225 static inline bool netif_testing(const struct net_device *dev)
4227 return test_bit(__LINK_STATE_TESTING, &dev->state);
4232 * netif_oper_up - test if device is operational
4233 * @dev: network device
4235 * Check if carrier is operational
4237 static inline bool netif_oper_up(const struct net_device *dev)
4239 return (dev->operstate == IF_OPER_UP ||
4240 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4244 * netif_device_present - is device available or removed
4245 * @dev: network device
4247 * Check if device has not been removed from system.
4249 static inline bool netif_device_present(const struct net_device *dev)
4251 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4254 void netif_device_detach(struct net_device *dev);
4256 void netif_device_attach(struct net_device *dev);
4259 * Network interface message level settings
4264 NETIF_MSG_PROBE_BIT,
4266 NETIF_MSG_TIMER_BIT,
4267 NETIF_MSG_IFDOWN_BIT,
4269 NETIF_MSG_RX_ERR_BIT,
4270 NETIF_MSG_TX_ERR_BIT,
4271 NETIF_MSG_TX_QUEUED_BIT,
4273 NETIF_MSG_TX_DONE_BIT,
4274 NETIF_MSG_RX_STATUS_BIT,
4275 NETIF_MSG_PKTDATA_BIT,
4279 /* When you add a new bit above, update netif_msg_class_names array
4280 * in net/ethtool/common.c
4282 NETIF_MSG_CLASS_COUNT,
4284 /* Both ethtool_ops interface and internal driver implementation use u32 */
4285 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4287 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4288 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4290 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4291 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4292 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4293 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4294 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4295 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4296 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4297 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4298 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4299 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4300 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4301 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4302 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4303 #define NETIF_MSG_HW __NETIF_MSG(HW)
4304 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4306 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4307 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4308 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4309 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4310 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4311 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4312 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4313 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4314 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4315 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4316 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4317 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4318 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4319 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4320 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4322 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4325 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4326 return default_msg_enable_bits;
4327 if (debug_value == 0) /* no output */
4329 /* set low N bits */
4330 return (1U << debug_value) - 1;
4333 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4335 spin_lock(&txq->_xmit_lock);
4336 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4337 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4340 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4342 __acquire(&txq->_xmit_lock);
4346 static inline void __netif_tx_release(struct netdev_queue *txq)
4348 __release(&txq->_xmit_lock);
4351 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4353 spin_lock_bh(&txq->_xmit_lock);
4354 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4355 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4358 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4360 bool ok = spin_trylock(&txq->_xmit_lock);
4363 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4364 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4369 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4371 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4372 WRITE_ONCE(txq->xmit_lock_owner, -1);
4373 spin_unlock(&txq->_xmit_lock);
4376 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4378 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4379 WRITE_ONCE(txq->xmit_lock_owner, -1);
4380 spin_unlock_bh(&txq->_xmit_lock);
4384 * txq->trans_start can be read locklessly from dev_watchdog()
4386 static inline void txq_trans_update(struct netdev_queue *txq)
4388 if (txq->xmit_lock_owner != -1)
4389 WRITE_ONCE(txq->trans_start, jiffies);
4392 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4394 unsigned long now = jiffies;
4396 if (READ_ONCE(txq->trans_start) != now)
4397 WRITE_ONCE(txq->trans_start, now);
4400 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4401 static inline void netif_trans_update(struct net_device *dev)
4403 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4405 txq_trans_cond_update(txq);
4409 * netif_tx_lock - grab network device transmit lock
4410 * @dev: network device
4412 * Get network device transmit lock
4414 void netif_tx_lock(struct net_device *dev);
4416 static inline void netif_tx_lock_bh(struct net_device *dev)
4422 void netif_tx_unlock(struct net_device *dev);
4424 static inline void netif_tx_unlock_bh(struct net_device *dev)
4426 netif_tx_unlock(dev);
4430 #define HARD_TX_LOCK(dev, txq, cpu) { \
4431 if ((dev->features & NETIF_F_LLTX) == 0) { \
4432 __netif_tx_lock(txq, cpu); \
4434 __netif_tx_acquire(txq); \
4438 #define HARD_TX_TRYLOCK(dev, txq) \
4439 (((dev->features & NETIF_F_LLTX) == 0) ? \
4440 __netif_tx_trylock(txq) : \
4441 __netif_tx_acquire(txq))
4443 #define HARD_TX_UNLOCK(dev, txq) { \
4444 if ((dev->features & NETIF_F_LLTX) == 0) { \
4445 __netif_tx_unlock(txq); \
4447 __netif_tx_release(txq); \
4451 static inline void netif_tx_disable(struct net_device *dev)
4457 cpu = smp_processor_id();
4458 spin_lock(&dev->tx_global_lock);
4459 for (i = 0; i < dev->num_tx_queues; i++) {
4460 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4462 __netif_tx_lock(txq, cpu);
4463 netif_tx_stop_queue(txq);
4464 __netif_tx_unlock(txq);
4466 spin_unlock(&dev->tx_global_lock);
4470 static inline void netif_addr_lock(struct net_device *dev)
4472 unsigned char nest_level = 0;
4474 #ifdef CONFIG_LOCKDEP
4475 nest_level = dev->nested_level;
4477 spin_lock_nested(&dev->addr_list_lock, nest_level);
4480 static inline void netif_addr_lock_bh(struct net_device *dev)
4482 unsigned char nest_level = 0;
4484 #ifdef CONFIG_LOCKDEP
4485 nest_level = dev->nested_level;
4488 spin_lock_nested(&dev->addr_list_lock, nest_level);
4491 static inline void netif_addr_unlock(struct net_device *dev)
4493 spin_unlock(&dev->addr_list_lock);
4496 static inline void netif_addr_unlock_bh(struct net_device *dev)
4498 spin_unlock_bh(&dev->addr_list_lock);
4502 * dev_addrs walker. Should be used only for read access. Call with
4503 * rcu_read_lock held.
4505 #define for_each_dev_addr(dev, ha) \
4506 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4508 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4510 void ether_setup(struct net_device *dev);
4512 /* Support for loadable net-drivers */
4513 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4514 unsigned char name_assign_type,
4515 void (*setup)(struct net_device *),
4516 unsigned int txqs, unsigned int rxqs);
4517 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4518 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4520 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4521 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4524 int register_netdev(struct net_device *dev);
4525 void unregister_netdev(struct net_device *dev);
4527 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4529 /* General hardware address lists handling functions */
4530 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4531 struct netdev_hw_addr_list *from_list, int addr_len);
4532 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4533 struct netdev_hw_addr_list *from_list, int addr_len);
4534 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4535 struct net_device *dev,
4536 int (*sync)(struct net_device *, const unsigned char *),
4537 int (*unsync)(struct net_device *,
4538 const unsigned char *));
4539 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4540 struct net_device *dev,
4541 int (*sync)(struct net_device *,
4542 const unsigned char *, int),
4543 int (*unsync)(struct net_device *,
4544 const unsigned char *, int));
4545 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4546 struct net_device *dev,
4547 int (*unsync)(struct net_device *,
4548 const unsigned char *, int));
4549 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4550 struct net_device *dev,
4551 int (*unsync)(struct net_device *,
4552 const unsigned char *));
4553 void __hw_addr_init(struct netdev_hw_addr_list *list);
4555 /* Functions used for device addresses handling */
4556 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4557 const void *addr, size_t len);
4560 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4562 dev_addr_mod(dev, 0, addr, len);
4565 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4567 __dev_addr_set(dev, addr, dev->addr_len);
4570 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4571 unsigned char addr_type);
4572 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4573 unsigned char addr_type);
4575 /* Functions used for unicast addresses handling */
4576 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4577 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4578 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4579 int dev_uc_sync(struct net_device *to, struct net_device *from);
4580 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4581 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4582 void dev_uc_flush(struct net_device *dev);
4583 void dev_uc_init(struct net_device *dev);
4586 * __dev_uc_sync - Synchonize device's unicast list
4587 * @dev: device to sync
4588 * @sync: function to call if address should be added
4589 * @unsync: function to call if address should be removed
4591 * Add newly added addresses to the interface, and release
4592 * addresses that have been deleted.
4594 static inline int __dev_uc_sync(struct net_device *dev,
4595 int (*sync)(struct net_device *,
4596 const unsigned char *),
4597 int (*unsync)(struct net_device *,
4598 const unsigned char *))
4600 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4604 * __dev_uc_unsync - Remove synchronized addresses from device
4605 * @dev: device to sync
4606 * @unsync: function to call if address should be removed
4608 * Remove all addresses that were added to the device by dev_uc_sync().
4610 static inline void __dev_uc_unsync(struct net_device *dev,
4611 int (*unsync)(struct net_device *,
4612 const unsigned char *))
4614 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4617 /* Functions used for multicast addresses handling */
4618 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4619 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4620 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4621 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4622 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4623 int dev_mc_sync(struct net_device *to, struct net_device *from);
4624 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4625 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4626 void dev_mc_flush(struct net_device *dev);
4627 void dev_mc_init(struct net_device *dev);
4630 * __dev_mc_sync - Synchonize device's multicast list
4631 * @dev: device to sync
4632 * @sync: function to call if address should be added
4633 * @unsync: function to call if address should be removed
4635 * Add newly added addresses to the interface, and release
4636 * addresses that have been deleted.
4638 static inline int __dev_mc_sync(struct net_device *dev,
4639 int (*sync)(struct net_device *,
4640 const unsigned char *),
4641 int (*unsync)(struct net_device *,
4642 const unsigned char *))
4644 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4648 * __dev_mc_unsync - Remove synchronized addresses from device
4649 * @dev: device to sync
4650 * @unsync: function to call if address should be removed
4652 * Remove all addresses that were added to the device by dev_mc_sync().
4654 static inline void __dev_mc_unsync(struct net_device *dev,
4655 int (*unsync)(struct net_device *,
4656 const unsigned char *))
4658 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4661 /* Functions used for secondary unicast and multicast support */
4662 void dev_set_rx_mode(struct net_device *dev);
4663 int dev_set_promiscuity(struct net_device *dev, int inc);
4664 int dev_set_allmulti(struct net_device *dev, int inc);
4665 void netdev_state_change(struct net_device *dev);
4666 void __netdev_notify_peers(struct net_device *dev);
4667 void netdev_notify_peers(struct net_device *dev);
4668 void netdev_features_change(struct net_device *dev);
4669 /* Load a device via the kmod */
4670 void dev_load(struct net *net, const char *name);
4671 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4672 struct rtnl_link_stats64 *storage);
4673 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4674 const struct net_device_stats *netdev_stats);
4675 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4676 const struct pcpu_sw_netstats __percpu *netstats);
4677 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4679 extern int netdev_max_backlog;
4680 extern int dev_rx_weight;
4681 extern int dev_tx_weight;
4682 extern int gro_normal_batch;
4685 NESTED_SYNC_IMM_BIT,
4686 NESTED_SYNC_TODO_BIT,
4689 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4690 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4692 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4693 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4695 struct netdev_nested_priv {
4696 unsigned char flags;
4700 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4701 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4702 struct list_head **iter);
4704 /* iterate through upper list, must be called under RCU read lock */
4705 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4706 for (iter = &(dev)->adj_list.upper, \
4707 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4709 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4711 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4712 int (*fn)(struct net_device *upper_dev,
4713 struct netdev_nested_priv *priv),
4714 struct netdev_nested_priv *priv);
4716 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4717 struct net_device *upper_dev);
4719 bool netdev_has_any_upper_dev(struct net_device *dev);
4721 void *netdev_lower_get_next_private(struct net_device *dev,
4722 struct list_head **iter);
4723 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4724 struct list_head **iter);
4726 #define netdev_for_each_lower_private(dev, priv, iter) \
4727 for (iter = (dev)->adj_list.lower.next, \
4728 priv = netdev_lower_get_next_private(dev, &(iter)); \
4730 priv = netdev_lower_get_next_private(dev, &(iter)))
4732 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4733 for (iter = &(dev)->adj_list.lower, \
4734 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4736 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4738 void *netdev_lower_get_next(struct net_device *dev,
4739 struct list_head **iter);
4741 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4742 for (iter = (dev)->adj_list.lower.next, \
4743 ldev = netdev_lower_get_next(dev, &(iter)); \
4745 ldev = netdev_lower_get_next(dev, &(iter)))
4747 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4748 struct list_head **iter);
4749 int netdev_walk_all_lower_dev(struct net_device *dev,
4750 int (*fn)(struct net_device *lower_dev,
4751 struct netdev_nested_priv *priv),
4752 struct netdev_nested_priv *priv);
4753 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4754 int (*fn)(struct net_device *lower_dev,
4755 struct netdev_nested_priv *priv),
4756 struct netdev_nested_priv *priv);
4758 void *netdev_adjacent_get_private(struct list_head *adj_list);
4759 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4760 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4761 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4762 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4763 struct netlink_ext_ack *extack);
4764 int netdev_master_upper_dev_link(struct net_device *dev,
4765 struct net_device *upper_dev,
4766 void *upper_priv, void *upper_info,
4767 struct netlink_ext_ack *extack);
4768 void netdev_upper_dev_unlink(struct net_device *dev,
4769 struct net_device *upper_dev);
4770 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4771 struct net_device *new_dev,
4772 struct net_device *dev,
4773 struct netlink_ext_ack *extack);
4774 void netdev_adjacent_change_commit(struct net_device *old_dev,
4775 struct net_device *new_dev,
4776 struct net_device *dev);
4777 void netdev_adjacent_change_abort(struct net_device *old_dev,
4778 struct net_device *new_dev,
4779 struct net_device *dev);
4780 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4781 void *netdev_lower_dev_get_private(struct net_device *dev,
4782 struct net_device *lower_dev);
4783 void netdev_lower_state_changed(struct net_device *lower_dev,
4784 void *lower_state_info);
4786 /* RSS keys are 40 or 52 bytes long */
4787 #define NETDEV_RSS_KEY_LEN 52
4788 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4789 void netdev_rss_key_fill(void *buffer, size_t len);
4791 int skb_checksum_help(struct sk_buff *skb);
4792 int skb_crc32c_csum_help(struct sk_buff *skb);
4793 int skb_csum_hwoffload_help(struct sk_buff *skb,
4794 const netdev_features_t features);
4796 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4797 netdev_features_t features, bool tx_path);
4798 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4799 netdev_features_t features, __be16 type);
4800 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4801 netdev_features_t features);
4803 struct netdev_bonding_info {
4808 struct netdev_notifier_bonding_info {
4809 struct netdev_notifier_info info; /* must be first */
4810 struct netdev_bonding_info bonding_info;
4813 void netdev_bonding_info_change(struct net_device *dev,
4814 struct netdev_bonding_info *bonding_info);
4816 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4817 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4819 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4826 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4828 return __skb_gso_segment(skb, features, true);
4830 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4832 static inline bool can_checksum_protocol(netdev_features_t features,
4835 if (protocol == htons(ETH_P_FCOE))
4836 return !!(features & NETIF_F_FCOE_CRC);
4838 /* Assume this is an IP checksum (not SCTP CRC) */
4840 if (features & NETIF_F_HW_CSUM) {
4841 /* Can checksum everything */
4846 case htons(ETH_P_IP):
4847 return !!(features & NETIF_F_IP_CSUM);
4848 case htons(ETH_P_IPV6):
4849 return !!(features & NETIF_F_IPV6_CSUM);
4856 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4858 static inline void netdev_rx_csum_fault(struct net_device *dev,
4859 struct sk_buff *skb)
4863 /* rx skb timestamps */
4864 void net_enable_timestamp(void);
4865 void net_disable_timestamp(void);
4867 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4868 const struct skb_shared_hwtstamps *hwtstamps,
4871 const struct net_device_ops *ops = dev->netdev_ops;
4873 if (ops->ndo_get_tstamp)
4874 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4876 return hwtstamps->hwtstamp;
4879 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4880 struct sk_buff *skb, struct net_device *dev,
4883 __this_cpu_write(softnet_data.xmit.more, more);
4884 return ops->ndo_start_xmit(skb, dev);
4887 static inline bool netdev_xmit_more(void)
4889 return __this_cpu_read(softnet_data.xmit.more);
4892 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4893 struct netdev_queue *txq, bool more)
4895 const struct net_device_ops *ops = dev->netdev_ops;
4898 rc = __netdev_start_xmit(ops, skb, dev, more);
4899 if (rc == NETDEV_TX_OK)
4900 txq_trans_update(txq);
4905 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4907 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4910 extern const struct kobj_ns_type_operations net_ns_type_operations;
4912 const char *netdev_drivername(const struct net_device *dev);
4914 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4915 netdev_features_t f2)
4917 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4918 if (f1 & NETIF_F_HW_CSUM)
4919 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4921 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4927 static inline netdev_features_t netdev_get_wanted_features(
4928 struct net_device *dev)
4930 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4932 netdev_features_t netdev_increment_features(netdev_features_t all,
4933 netdev_features_t one, netdev_features_t mask);
4935 /* Allow TSO being used on stacked device :
4936 * Performing the GSO segmentation before last device
4937 * is a performance improvement.
4939 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4940 netdev_features_t mask)
4942 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4945 int __netdev_update_features(struct net_device *dev);
4946 void netdev_update_features(struct net_device *dev);
4947 void netdev_change_features(struct net_device *dev);
4949 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4950 struct net_device *dev);
4952 netdev_features_t passthru_features_check(struct sk_buff *skb,
4953 struct net_device *dev,
4954 netdev_features_t features);
4955 netdev_features_t netif_skb_features(struct sk_buff *skb);
4957 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4959 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4961 /* check flags correspondence */
4962 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4963 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4964 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4965 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4966 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4967 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4968 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4969 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4970 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4971 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4972 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4973 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4974 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4975 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4976 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4977 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4978 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4979 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4980 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4982 return (features & feature) == feature;
4985 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4987 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4988 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4991 static inline bool netif_needs_gso(struct sk_buff *skb,
4992 netdev_features_t features)
4994 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4995 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4996 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4999 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5000 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5001 void netif_inherit_tso_max(struct net_device *to,
5002 const struct net_device *from);
5004 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5005 int pulled_hlen, u16 mac_offset,
5008 skb->protocol = protocol;
5009 skb->encapsulation = 1;
5010 skb_push(skb, pulled_hlen);
5011 skb_reset_transport_header(skb);
5012 skb->mac_header = mac_offset;
5013 skb->network_header = skb->mac_header + mac_len;
5014 skb->mac_len = mac_len;
5017 static inline bool netif_is_macsec(const struct net_device *dev)
5019 return dev->priv_flags & IFF_MACSEC;
5022 static inline bool netif_is_macvlan(const struct net_device *dev)
5024 return dev->priv_flags & IFF_MACVLAN;
5027 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5029 return dev->priv_flags & IFF_MACVLAN_PORT;
5032 static inline bool netif_is_bond_master(const struct net_device *dev)
5034 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5037 static inline bool netif_is_bond_slave(const struct net_device *dev)
5039 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5042 static inline bool netif_supports_nofcs(struct net_device *dev)
5044 return dev->priv_flags & IFF_SUPP_NOFCS;
5047 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5049 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5052 static inline bool netif_is_l3_master(const struct net_device *dev)
5054 return dev->priv_flags & IFF_L3MDEV_MASTER;
5057 static inline bool netif_is_l3_slave(const struct net_device *dev)
5059 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5062 static inline bool netif_is_bridge_master(const struct net_device *dev)
5064 return dev->priv_flags & IFF_EBRIDGE;
5067 static inline bool netif_is_bridge_port(const struct net_device *dev)
5069 return dev->priv_flags & IFF_BRIDGE_PORT;
5072 static inline bool netif_is_ovs_master(const struct net_device *dev)
5074 return dev->priv_flags & IFF_OPENVSWITCH;
5077 static inline bool netif_is_ovs_port(const struct net_device *dev)
5079 return dev->priv_flags & IFF_OVS_DATAPATH;
5082 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5084 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5087 static inline bool netif_is_team_master(const struct net_device *dev)
5089 return dev->priv_flags & IFF_TEAM;
5092 static inline bool netif_is_team_port(const struct net_device *dev)
5094 return dev->priv_flags & IFF_TEAM_PORT;
5097 static inline bool netif_is_lag_master(const struct net_device *dev)
5099 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5102 static inline bool netif_is_lag_port(const struct net_device *dev)
5104 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5107 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5109 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5112 static inline bool netif_is_failover(const struct net_device *dev)
5114 return dev->priv_flags & IFF_FAILOVER;
5117 static inline bool netif_is_failover_slave(const struct net_device *dev)
5119 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5122 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5123 static inline void netif_keep_dst(struct net_device *dev)
5125 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5128 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5129 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5131 /* TODO: reserve and use an additional IFF bit, if we get more users */
5132 return netif_is_macsec(dev);
5135 extern struct pernet_operations __net_initdata loopback_net_ops;
5137 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5139 /* netdev_printk helpers, similar to dev_printk */
5141 static inline const char *netdev_name(const struct net_device *dev)
5143 if (!dev->name[0] || strchr(dev->name, '%'))
5144 return "(unnamed net_device)";
5148 static inline const char *netdev_reg_state(const struct net_device *dev)
5150 switch (dev->reg_state) {
5151 case NETREG_UNINITIALIZED: return " (uninitialized)";
5152 case NETREG_REGISTERED: return "";
5153 case NETREG_UNREGISTERING: return " (unregistering)";
5154 case NETREG_UNREGISTERED: return " (unregistered)";
5155 case NETREG_RELEASED: return " (released)";
5156 case NETREG_DUMMY: return " (dummy)";
5159 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5160 return " (unknown)";
5163 #define MODULE_ALIAS_NETDEV(device) \
5164 MODULE_ALIAS("netdev-" device)
5167 * netdev_WARN() acts like dev_printk(), but with the key difference
5168 * of using a WARN/WARN_ON to get the message out, including the
5169 * file/line information and a backtrace.
5171 #define netdev_WARN(dev, format, args...) \
5172 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5173 netdev_reg_state(dev), ##args)
5175 #define netdev_WARN_ONCE(dev, format, args...) \
5176 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5177 netdev_reg_state(dev), ##args)
5180 * The list of packet types we will receive (as opposed to discard)
5181 * and the routines to invoke.
5183 * Why 16. Because with 16 the only overlap we get on a hash of the
5184 * low nibble of the protocol value is RARP/SNAP/X.25.
5198 #define PTYPE_HASH_SIZE (16)
5199 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5201 extern struct list_head ptype_all __read_mostly;
5202 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5204 extern struct net_device *blackhole_netdev;
5206 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5207 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5208 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5209 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5211 #endif /* _LINUX_NETDEVICE_H */