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 <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
60 struct ip_tunnel_parm;
61 struct macsec_context;
63 struct netdev_name_node;
68 /* 802.15.4 specific */
71 /* UDP Tunnel offloads */
72 struct udp_tunnel_info;
73 struct udp_tunnel_nic_info;
74 struct udp_tunnel_nic;
78 void synchronize_net(void);
79 void netdev_set_default_ethtool_ops(struct net_device *dev,
80 const struct ethtool_ops *ops);
82 /* Backlog congestion levels */
83 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
84 #define NET_RX_DROP 1 /* packet dropped */
86 #define MAX_NEST_DEV 8
89 * Transmit return codes: transmit return codes originate from three different
92 * - qdisc return codes
93 * - driver transmit return codes
96 * Drivers are allowed to return any one of those in their hard_start_xmit()
97 * function. Real network devices commonly used with qdiscs should only return
98 * the driver transmit return codes though - when qdiscs are used, the actual
99 * transmission happens asynchronously, so the value is not propagated to
100 * higher layers. Virtual network devices transmit synchronously; in this case
101 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
102 * others are propagated to higher layers.
105 /* qdisc ->enqueue() return codes. */
106 #define NET_XMIT_SUCCESS 0x00
107 #define NET_XMIT_DROP 0x01 /* skb dropped */
108 #define NET_XMIT_CN 0x02 /* congestion notification */
109 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
111 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
112 * indicates that the device will soon be dropping packets, or already drops
113 * some packets of the same priority; prompting us to send less aggressively. */
114 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
115 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
117 /* Driver transmit return codes */
118 #define NETDEV_TX_MASK 0xf0
121 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
122 NETDEV_TX_OK = 0x00, /* driver took care of packet */
123 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
125 typedef enum netdev_tx netdev_tx_t;
128 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
129 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
131 static inline bool dev_xmit_complete(int rc)
134 * Positive cases with an skb consumed by a driver:
135 * - successful transmission (rc == NETDEV_TX_OK)
136 * - error while transmitting (rc < 0)
137 * - error while queueing to a different device (rc & NET_XMIT_MASK)
139 if (likely(rc < NET_XMIT_MASK))
146 * Compute the worst-case header length according to the protocols
150 #if defined(CONFIG_HYPERV_NET)
151 # define LL_MAX_HEADER 128
152 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
153 # if defined(CONFIG_MAC80211_MESH)
154 # define LL_MAX_HEADER 128
156 # define LL_MAX_HEADER 96
159 # define LL_MAX_HEADER 32
162 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
163 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
164 #define MAX_HEADER LL_MAX_HEADER
166 #define MAX_HEADER (LL_MAX_HEADER + 48)
170 * Old network device statistics. Fields are native words
171 * (unsigned long) so they can be read and written atomically.
174 #define NET_DEV_STAT(FIELD) \
176 unsigned long FIELD; \
177 atomic_long_t __##FIELD; \
180 struct net_device_stats {
181 NET_DEV_STAT(rx_packets);
182 NET_DEV_STAT(tx_packets);
183 NET_DEV_STAT(rx_bytes);
184 NET_DEV_STAT(tx_bytes);
185 NET_DEV_STAT(rx_errors);
186 NET_DEV_STAT(tx_errors);
187 NET_DEV_STAT(rx_dropped);
188 NET_DEV_STAT(tx_dropped);
189 NET_DEV_STAT(multicast);
190 NET_DEV_STAT(collisions);
191 NET_DEV_STAT(rx_length_errors);
192 NET_DEV_STAT(rx_over_errors);
193 NET_DEV_STAT(rx_crc_errors);
194 NET_DEV_STAT(rx_frame_errors);
195 NET_DEV_STAT(rx_fifo_errors);
196 NET_DEV_STAT(rx_missed_errors);
197 NET_DEV_STAT(tx_aborted_errors);
198 NET_DEV_STAT(tx_carrier_errors);
199 NET_DEV_STAT(tx_fifo_errors);
200 NET_DEV_STAT(tx_heartbeat_errors);
201 NET_DEV_STAT(tx_window_errors);
202 NET_DEV_STAT(rx_compressed);
203 NET_DEV_STAT(tx_compressed);
207 /* per-cpu stats, allocated on demand.
208 * Try to fit them in a single cache line, for dev_get_stats() sake.
210 struct net_device_core_stats {
211 unsigned long rx_dropped;
212 unsigned long tx_dropped;
213 unsigned long rx_nohandler;
214 unsigned long rx_otherhost_dropped;
215 } __aligned(4 * sizeof(unsigned long));
217 #include <linux/cache.h>
218 #include <linux/skbuff.h>
221 #include <linux/static_key.h>
222 extern struct static_key_false rps_needed;
223 extern struct static_key_false rfs_needed;
230 struct netdev_hw_addr {
231 struct list_head list;
233 unsigned char addr[MAX_ADDR_LEN];
235 #define NETDEV_HW_ADDR_T_LAN 1
236 #define NETDEV_HW_ADDR_T_SAN 2
237 #define NETDEV_HW_ADDR_T_UNICAST 3
238 #define NETDEV_HW_ADDR_T_MULTICAST 4
243 struct rcu_head rcu_head;
246 struct netdev_hw_addr_list {
247 struct list_head list;
250 /* Auxiliary tree for faster lookup on addition and deletion */
254 #define netdev_hw_addr_list_count(l) ((l)->count)
255 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
256 #define netdev_hw_addr_list_for_each(ha, l) \
257 list_for_each_entry(ha, &(l)->list, list)
259 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
260 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
261 #define netdev_for_each_uc_addr(ha, dev) \
262 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
263 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
264 netdev_for_each_uc_addr((_ha), (_dev)) \
267 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
268 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
269 #define netdev_for_each_mc_addr(ha, dev) \
270 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
271 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
272 netdev_for_each_mc_addr((_ha), (_dev)) \
279 /* cached hardware header; allow for machine alignment needs. */
280 #define HH_DATA_MOD 16
281 #define HH_DATA_OFF(__len) \
282 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
283 #define HH_DATA_ALIGN(__len) \
284 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
285 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
288 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
290 * dev->hard_header_len ? (dev->hard_header_len +
291 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
293 * We could use other alignment values, but we must maintain the
294 * relationship HH alignment <= LL alignment.
296 #define LL_RESERVED_SPACE(dev) \
297 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
298 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
299 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
300 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
301 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
304 int (*create) (struct sk_buff *skb, struct net_device *dev,
305 unsigned short type, const void *daddr,
306 const void *saddr, unsigned int len);
307 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
308 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
309 void (*cache_update)(struct hh_cache *hh,
310 const struct net_device *dev,
311 const unsigned char *haddr);
312 bool (*validate)(const char *ll_header, unsigned int len);
313 __be16 (*parse_protocol)(const struct sk_buff *skb);
316 /* These flag bits are private to the generic network queueing
317 * layer; they may not be explicitly referenced by any other
321 enum netdev_state_t {
323 __LINK_STATE_PRESENT,
324 __LINK_STATE_NOCARRIER,
325 __LINK_STATE_LINKWATCH_PENDING,
326 __LINK_STATE_DORMANT,
327 __LINK_STATE_TESTING,
331 struct list_head list;
336 * size of gro hash buckets, must less than bit number of
337 * napi_struct::gro_bitmask
339 #define GRO_HASH_BUCKETS 8
342 * Structure for NAPI scheduling similar to tasklet but with weighting
345 /* The poll_list must only be managed by the entity which
346 * changes the state of the NAPI_STATE_SCHED bit. This means
347 * whoever atomically sets that bit can add this napi_struct
348 * to the per-CPU poll_list, and whoever clears that bit
349 * can remove from the list right before clearing the bit.
351 struct list_head poll_list;
355 int defer_hard_irqs_count;
356 unsigned long gro_bitmask;
357 int (*poll)(struct napi_struct *, int);
358 #ifdef CONFIG_NETPOLL
361 struct net_device *dev;
362 struct gro_list gro_hash[GRO_HASH_BUCKETS];
364 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
365 int rx_count; /* length of rx_list */
366 struct hrtimer timer;
367 struct list_head dev_list;
368 struct hlist_node napi_hash_node;
369 unsigned int napi_id;
370 struct task_struct *thread;
374 NAPI_STATE_SCHED, /* Poll is scheduled */
375 NAPI_STATE_MISSED, /* reschedule a napi */
376 NAPI_STATE_DISABLE, /* Disable pending */
377 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
378 NAPI_STATE_LISTED, /* NAPI added to system lists */
379 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
380 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
381 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
382 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
383 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
387 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
388 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
389 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
390 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
391 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
392 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
393 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
394 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
395 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
396 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
406 typedef enum gro_result gro_result_t;
409 * enum rx_handler_result - Possible return values for rx_handlers.
410 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
412 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
413 * case skb->dev was changed by rx_handler.
414 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
415 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
417 * rx_handlers are functions called from inside __netif_receive_skb(), to do
418 * special processing of the skb, prior to delivery to protocol handlers.
420 * Currently, a net_device can only have a single rx_handler registered. Trying
421 * to register a second rx_handler will return -EBUSY.
423 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
424 * To unregister a rx_handler on a net_device, use
425 * netdev_rx_handler_unregister().
427 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
430 * If the rx_handler consumed the skb in some way, it should return
431 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
432 * the skb to be delivered in some other way.
434 * If the rx_handler changed skb->dev, to divert the skb to another
435 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
436 * new device will be called if it exists.
438 * If the rx_handler decides the skb should be ignored, it should return
439 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
440 * are registered on exact device (ptype->dev == skb->dev).
442 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
443 * delivered, it should return RX_HANDLER_PASS.
445 * A device without a registered rx_handler will behave as if rx_handler
446 * returned RX_HANDLER_PASS.
449 enum rx_handler_result {
455 typedef enum rx_handler_result rx_handler_result_t;
456 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
458 void __napi_schedule(struct napi_struct *n);
459 void __napi_schedule_irqoff(struct napi_struct *n);
461 static inline bool napi_disable_pending(struct napi_struct *n)
463 return test_bit(NAPI_STATE_DISABLE, &n->state);
466 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
468 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
471 bool napi_schedule_prep(struct napi_struct *n);
474 * napi_schedule - schedule NAPI poll
477 * Schedule NAPI poll routine to be called if it is not already
480 static inline void napi_schedule(struct napi_struct *n)
482 if (napi_schedule_prep(n))
487 * napi_schedule_irqoff - schedule NAPI poll
490 * Variant of napi_schedule(), assuming hard irqs are masked.
492 static inline void napi_schedule_irqoff(struct napi_struct *n)
494 if (napi_schedule_prep(n))
495 __napi_schedule_irqoff(n);
498 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
499 static inline bool napi_reschedule(struct napi_struct *napi)
501 if (napi_schedule_prep(napi)) {
502 __napi_schedule(napi);
508 bool napi_complete_done(struct napi_struct *n, int work_done);
510 * napi_complete - NAPI processing complete
513 * Mark NAPI processing as complete.
514 * Consider using napi_complete_done() instead.
515 * Return false if device should avoid rearming interrupts.
517 static inline bool napi_complete(struct napi_struct *n)
519 return napi_complete_done(n, 0);
522 int dev_set_threaded(struct net_device *dev, bool threaded);
525 * napi_disable - prevent NAPI from scheduling
528 * Stop NAPI from being scheduled on this context.
529 * Waits till any outstanding processing completes.
531 void napi_disable(struct napi_struct *n);
533 void napi_enable(struct napi_struct *n);
536 * napi_synchronize - wait until NAPI is not running
539 * Wait until NAPI is done being scheduled on this context.
540 * Waits till any outstanding processing completes but
541 * does not disable future activations.
543 static inline void napi_synchronize(const struct napi_struct *n)
545 if (IS_ENABLED(CONFIG_SMP))
546 while (test_bit(NAPI_STATE_SCHED, &n->state))
553 * napi_if_scheduled_mark_missed - if napi is running, set the
557 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
560 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
562 unsigned long val, new;
564 val = READ_ONCE(n->state);
566 if (val & NAPIF_STATE_DISABLE)
569 if (!(val & NAPIF_STATE_SCHED))
572 new = val | NAPIF_STATE_MISSED;
573 } while (!try_cmpxchg(&n->state, &val, new));
578 enum netdev_queue_state_t {
579 __QUEUE_STATE_DRV_XOFF,
580 __QUEUE_STATE_STACK_XOFF,
581 __QUEUE_STATE_FROZEN,
584 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
585 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
586 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
588 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
589 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
591 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
595 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
596 * netif_tx_* functions below are used to manipulate this flag. The
597 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
598 * queue independently. The netif_xmit_*stopped functions below are called
599 * to check if the queue has been stopped by the driver or stack (either
600 * of the XOFF bits are set in the state). Drivers should not need to call
601 * netif_xmit*stopped functions, they should only be using netif_tx_*.
604 struct netdev_queue {
608 struct net_device *dev;
609 netdevice_tracker dev_tracker;
611 struct Qdisc __rcu *qdisc;
612 struct Qdisc __rcu *qdisc_sleeping;
616 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
619 unsigned long tx_maxrate;
621 * Number of TX timeouts for this queue
622 * (/sys/class/net/DEV/Q/trans_timeout)
624 atomic_long_t trans_timeout;
626 /* Subordinate device that the queue has been assigned to */
627 struct net_device *sb_dev;
628 #ifdef CONFIG_XDP_SOCKETS
629 struct xsk_buff_pool *pool;
634 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
637 * Time (in jiffies) of last Tx
639 unsigned long trans_start;
646 } ____cacheline_aligned_in_smp;
648 extern int sysctl_fb_tunnels_only_for_init_net;
649 extern int sysctl_devconf_inherit_init_net;
652 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
653 * == 1 : For initns only
656 static inline bool net_has_fallback_tunnels(const struct net *net)
658 #if IS_ENABLED(CONFIG_SYSCTL)
659 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
661 return !fb_tunnels_only_for_init_net ||
662 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
668 static inline int net_inherit_devconf(void)
670 #if IS_ENABLED(CONFIG_SYSCTL)
671 return READ_ONCE(sysctl_devconf_inherit_init_net);
677 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
679 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
686 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
688 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
695 * This structure holds an RPS map which can be of variable length. The
696 * map is an array of CPUs.
703 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
706 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
707 * tail pointer for that CPU's input queue at the time of last enqueue, and
708 * a hardware filter index.
710 struct rps_dev_flow {
713 unsigned int last_qtail;
715 #define RPS_NO_FILTER 0xffff
718 * The rps_dev_flow_table structure contains a table of flow mappings.
720 struct rps_dev_flow_table {
723 struct rps_dev_flow flows[];
725 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
726 ((_num) * sizeof(struct rps_dev_flow)))
729 * The rps_sock_flow_table contains mappings of flows to the last CPU
730 * on which they were processed by the application (set in recvmsg).
731 * Each entry is a 32bit value. Upper part is the high-order bits
732 * of flow hash, lower part is CPU number.
733 * rps_cpu_mask is used to partition the space, depending on number of
734 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
735 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
736 * meaning we use 32-6=26 bits for the hash.
738 struct rps_sock_flow_table {
741 u32 ents[] ____cacheline_aligned_in_smp;
743 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
745 #define RPS_NO_CPU 0xffff
747 extern u32 rps_cpu_mask;
748 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
750 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
754 unsigned int index = hash & table->mask;
755 u32 val = hash & ~rps_cpu_mask;
757 /* We only give a hint, preemption can change CPU under us */
758 val |= raw_smp_processor_id();
760 /* The following WRITE_ONCE() is paired with the READ_ONCE()
761 * here, and another one in get_rps_cpu().
763 if (READ_ONCE(table->ents[index]) != val)
764 WRITE_ONCE(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);
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);
1051 struct dev_ifalias {
1052 struct rcu_head rcuhead;
1059 struct netdev_net_notifier {
1060 struct list_head list;
1061 struct notifier_block *nb;
1065 * This structure defines the management hooks for network devices.
1066 * The following hooks can be defined; unless noted otherwise, they are
1067 * optional and can be filled with a null pointer.
1069 * int (*ndo_init)(struct net_device *dev);
1070 * This function is called once when a network device is registered.
1071 * The network device can use this for any late stage initialization
1072 * or semantic validation. It can fail with an error code which will
1073 * be propagated back to register_netdev.
1075 * void (*ndo_uninit)(struct net_device *dev);
1076 * This function is called when device is unregistered or when registration
1077 * fails. It is not called if init fails.
1079 * int (*ndo_open)(struct net_device *dev);
1080 * This function is called when a network device transitions to the up
1083 * int (*ndo_stop)(struct net_device *dev);
1084 * This function is called when a network device transitions to the down
1087 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1088 * struct net_device *dev);
1089 * Called when a packet needs to be transmitted.
1090 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1091 * the queue before that can happen; it's for obsolete devices and weird
1092 * corner cases, but the stack really does a non-trivial amount
1093 * of useless work if you return NETDEV_TX_BUSY.
1094 * Required; cannot be NULL.
1096 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1097 * struct net_device *dev
1098 * netdev_features_t features);
1099 * Called by core transmit path to determine if device is capable of
1100 * performing offload operations on a given packet. This is to give
1101 * the device an opportunity to implement any restrictions that cannot
1102 * be otherwise expressed by feature flags. The check is called with
1103 * the set of features that the stack has calculated and it returns
1104 * those the driver believes to be appropriate.
1106 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1107 * struct net_device *sb_dev);
1108 * Called to decide which queue to use when device supports multiple
1111 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1112 * This function is called to allow device receiver to make
1113 * changes to configuration when multicast or promiscuous is enabled.
1115 * void (*ndo_set_rx_mode)(struct net_device *dev);
1116 * This function is called device changes address list filtering.
1117 * If driver handles unicast address filtering, it should set
1118 * IFF_UNICAST_FLT in its priv_flags.
1120 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1121 * This function is called when the Media Access Control address
1122 * needs to be changed. If this interface is not defined, the
1123 * MAC address can not be changed.
1125 * int (*ndo_validate_addr)(struct net_device *dev);
1126 * Test if Media Access Control address is valid for the device.
1128 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1129 * Old-style ioctl entry point. This is used internally by the
1130 * appletalk and ieee802154 subsystems but is no longer called by
1131 * the device ioctl handler.
1133 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1134 * Used by the bonding driver for its device specific ioctls:
1135 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1136 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1138 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1139 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1140 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1142 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1143 * Used to set network devices bus interface parameters. This interface
1144 * is retained for legacy reasons; new devices should use the bus
1145 * interface (PCI) for low level management.
1147 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1148 * Called when a user wants to change the Maximum Transfer Unit
1151 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1152 * Callback used when the transmitter has not made any progress
1153 * for dev->watchdog ticks.
1155 * void (*ndo_get_stats64)(struct net_device *dev,
1156 * struct rtnl_link_stats64 *storage);
1157 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1158 * Called when a user wants to get the network device usage
1159 * statistics. Drivers must do one of the following:
1160 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1161 * rtnl_link_stats64 structure passed by the caller.
1162 * 2. Define @ndo_get_stats to update a net_device_stats structure
1163 * (which should normally be dev->stats) and return a pointer to
1164 * it. The structure may be changed asynchronously only if each
1165 * field is written atomically.
1166 * 3. Update dev->stats asynchronously and atomically, and define
1167 * neither operation.
1169 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1170 * Return true if this device supports offload stats of this attr_id.
1172 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1174 * Get statistics for offload operations by attr_id. Write it into the
1175 * attr_data pointer.
1177 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1178 * If device supports VLAN filtering this function is called when a
1179 * VLAN id is registered.
1181 * int (*ndo_vlan_rx_kill_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 unregistered.
1185 * void (*ndo_poll_controller)(struct net_device *dev);
1187 * SR-IOV management functions.
1188 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1189 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1190 * u8 qos, __be16 proto);
1191 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1193 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1194 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1195 * int (*ndo_get_vf_config)(struct net_device *dev,
1196 * int vf, struct ifla_vf_info *ivf);
1197 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1198 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1199 * struct nlattr *port[]);
1201 * Enable or disable the VF ability to query its RSS Redirection Table and
1202 * Hash Key. This is needed since on some devices VF share this information
1203 * with PF and querying it may introduce a theoretical security risk.
1204 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1205 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1206 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1208 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1209 * This is always called from the stack with the rtnl lock held and netif
1210 * tx queues stopped. This allows the netdevice to perform queue
1211 * management safely.
1213 * Fiber Channel over Ethernet (FCoE) offload functions.
1214 * int (*ndo_fcoe_enable)(struct net_device *dev);
1215 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1216 * so the underlying device can perform whatever needed configuration or
1217 * initialization to support acceleration of FCoE traffic.
1219 * int (*ndo_fcoe_disable)(struct net_device *dev);
1220 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1221 * so the underlying device can perform whatever needed clean-ups to
1222 * stop supporting acceleration of FCoE traffic.
1224 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1225 * struct scatterlist *sgl, unsigned int sgc);
1226 * Called when the FCoE Initiator wants to initialize an I/O that
1227 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1228 * perform necessary setup and returns 1 to indicate the device is set up
1229 * successfully to perform DDP on this I/O, otherwise this returns 0.
1231 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1232 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1233 * indicated by the FC exchange id 'xid', so the underlying device can
1234 * clean up and reuse resources for later DDP requests.
1236 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1237 * struct scatterlist *sgl, unsigned int sgc);
1238 * Called when the FCoE Target wants to initialize an I/O that
1239 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1240 * perform necessary setup and returns 1 to indicate the device is set up
1241 * successfully to perform DDP on this I/O, otherwise this returns 0.
1243 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1244 * struct netdev_fcoe_hbainfo *hbainfo);
1245 * Called when the FCoE Protocol stack wants information on the underlying
1246 * device. This information is utilized by the FCoE protocol stack to
1247 * register attributes with Fiber Channel management service as per the
1248 * FC-GS Fabric Device Management Information(FDMI) specification.
1250 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1251 * Called when the underlying device wants to override default World Wide
1252 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1253 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1254 * protocol stack to use.
1257 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1258 * u16 rxq_index, u32 flow_id);
1259 * Set hardware filter for RFS. rxq_index is the target queue index;
1260 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1261 * Return the filter ID on success, or a negative error code.
1263 * Slave management functions (for bridge, bonding, etc).
1264 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1265 * Called to make another netdev an underling.
1267 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1268 * Called to release previously enslaved netdev.
1270 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1271 * struct sk_buff *skb,
1273 * Get the xmit slave of master device. If all_slaves is true, function
1274 * assume all the slaves can transmit.
1276 * Feature/offload setting functions.
1277 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1278 * netdev_features_t features);
1279 * Adjusts the requested feature flags according to device-specific
1280 * constraints, and returns the resulting flags. Must not modify
1283 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1284 * Called to update device configuration to new features. Passed
1285 * feature set might be less than what was returned by ndo_fix_features()).
1286 * Must return >0 or -errno if it changed dev->features itself.
1288 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1289 * struct net_device *dev,
1290 * const unsigned char *addr, u16 vid, u16 flags,
1291 * struct netlink_ext_ack *extack);
1292 * Adds an FDB entry to dev for addr.
1293 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1294 * struct net_device *dev,
1295 * const unsigned char *addr, u16 vid)
1296 * Deletes the FDB entry from dev coresponding to addr.
1297 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1298 * struct net_device *dev,
1300 * struct netlink_ext_ack *extack);
1301 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1302 * struct net_device *dev, struct net_device *filter_dev,
1304 * Used to add FDB entries to dump requests. Implementers should add
1305 * entries to skb and update idx with the number of entries.
1307 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1308 * u16 flags, struct netlink_ext_ack *extack)
1309 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1310 * struct net_device *dev, u32 filter_mask,
1312 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1315 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1316 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1317 * which do not represent real hardware may define this to allow their
1318 * userspace components to manage their virtual carrier state. Devices
1319 * that determine carrier state from physical hardware properties (eg
1320 * network cables) or protocol-dependent mechanisms (eg
1321 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1323 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1324 * struct netdev_phys_item_id *ppid);
1325 * Called to get ID of physical port of this device. If driver does
1326 * not implement this, it is assumed that the hw is not able to have
1327 * multiple net devices on single physical port.
1329 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1330 * struct netdev_phys_item_id *ppid)
1331 * Called to get the parent ID of the physical port of this device.
1333 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1334 * struct net_device *dev)
1335 * Called by upper layer devices to accelerate switching or other
1336 * station functionality into hardware. 'pdev is the lowerdev
1337 * to use for the offload and 'dev' is the net device that will
1338 * back the offload. Returns a pointer to the private structure
1339 * the upper layer will maintain.
1340 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1341 * Called by upper layer device to delete the station created
1342 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1343 * the station and priv is the structure returned by the add
1345 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1346 * int queue_index, u32 maxrate);
1347 * Called when a user wants to set a max-rate limitation of specific
1349 * int (*ndo_get_iflink)(const struct net_device *dev);
1350 * Called to get the iflink value of this device.
1351 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1352 * This function is used to get egress tunnel information for given skb.
1353 * This is useful for retrieving outer tunnel header parameters while
1355 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1356 * This function is used to specify the headroom that the skb must
1357 * consider when allocation skb during packet reception. Setting
1358 * appropriate rx headroom value allows avoiding skb head copy on
1359 * forward. Setting a negative value resets the rx headroom to the
1361 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1362 * This function is used to set or query state related to XDP on the
1363 * netdevice and manage BPF offload. See definition of
1364 * enum bpf_netdev_command for details.
1365 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1367 * This function is used to submit @n XDP packets for transmit on a
1368 * netdevice. Returns number of frames successfully transmitted, frames
1369 * that got dropped are freed/returned via xdp_return_frame().
1370 * Returns negative number, means general error invoking ndo, meaning
1371 * no frames were xmit'ed and core-caller will free all frames.
1372 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1373 * struct xdp_buff *xdp);
1374 * Get the xmit slave of master device based on the xdp_buff.
1375 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1376 * This function is used to wake up the softirq, ksoftirqd or kthread
1377 * responsible for sending and/or receiving packets on a specific
1378 * queue id bound to an AF_XDP socket. The flags field specifies if
1379 * only RX, only Tx, or both should be woken up using the flags
1380 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1381 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1382 * Get devlink port instance associated with a given netdev.
1383 * Called with a reference on the netdevice and devlink locks only,
1384 * rtnl_lock is not held.
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 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1616 int (*ndo_tunnel_ctl)(struct net_device *dev,
1617 struct ip_tunnel_parm *p, int cmd);
1618 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1619 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1620 struct net_device_path *path);
1621 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1622 const struct skb_shared_hwtstamps *hwtstamps,
1627 * enum netdev_priv_flags - &struct net_device priv_flags
1629 * These are the &struct net_device, they are only set internally
1630 * by drivers and used in the kernel. These flags are invisible to
1631 * userspace; this means that the order of these flags can change
1632 * during any kernel release.
1634 * You should have a pretty good reason to be extending these flags.
1636 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1637 * @IFF_EBRIDGE: Ethernet bridging device
1638 * @IFF_BONDING: bonding master or slave
1639 * @IFF_ISATAP: ISATAP interface (RFC4214)
1640 * @IFF_WAN_HDLC: WAN HDLC device
1641 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1643 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1644 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1645 * @IFF_MACVLAN_PORT: device used as macvlan port
1646 * @IFF_BRIDGE_PORT: device used as bridge port
1647 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1648 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1649 * @IFF_UNICAST_FLT: Supports unicast filtering
1650 * @IFF_TEAM_PORT: device used as team port
1651 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1652 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1653 * change when it's running
1654 * @IFF_MACVLAN: Macvlan device
1655 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1656 * underlying stacked devices
1657 * @IFF_L3MDEV_MASTER: device is an L3 master device
1658 * @IFF_NO_QUEUE: device can run without qdisc attached
1659 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1660 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1661 * @IFF_TEAM: device is a team device
1662 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1663 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1664 * entity (i.e. the master device for bridged veth)
1665 * @IFF_MACSEC: device is a MACsec device
1666 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1667 * @IFF_FAILOVER: device is a failover master device
1668 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1669 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1670 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1671 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1672 * skb_headlen(skb) == 0 (data starts from frag0)
1673 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1675 enum netdev_priv_flags {
1676 IFF_802_1Q_VLAN = 1<<0,
1680 IFF_WAN_HDLC = 1<<4,
1681 IFF_XMIT_DST_RELEASE = 1<<5,
1682 IFF_DONT_BRIDGE = 1<<6,
1683 IFF_DISABLE_NETPOLL = 1<<7,
1684 IFF_MACVLAN_PORT = 1<<8,
1685 IFF_BRIDGE_PORT = 1<<9,
1686 IFF_OVS_DATAPATH = 1<<10,
1687 IFF_TX_SKB_SHARING = 1<<11,
1688 IFF_UNICAST_FLT = 1<<12,
1689 IFF_TEAM_PORT = 1<<13,
1690 IFF_SUPP_NOFCS = 1<<14,
1691 IFF_LIVE_ADDR_CHANGE = 1<<15,
1692 IFF_MACVLAN = 1<<16,
1693 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1694 IFF_L3MDEV_MASTER = 1<<18,
1695 IFF_NO_QUEUE = 1<<19,
1696 IFF_OPENVSWITCH = 1<<20,
1697 IFF_L3MDEV_SLAVE = 1<<21,
1699 IFF_RXFH_CONFIGURED = 1<<23,
1700 IFF_PHONY_HEADROOM = 1<<24,
1702 IFF_NO_RX_HANDLER = 1<<26,
1703 IFF_FAILOVER = 1<<27,
1704 IFF_FAILOVER_SLAVE = 1<<28,
1705 IFF_L3MDEV_RX_HANDLER = 1<<29,
1706 IFF_LIVE_RENAME_OK = 1<<30,
1707 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1708 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1711 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1712 #define IFF_EBRIDGE IFF_EBRIDGE
1713 #define IFF_BONDING IFF_BONDING
1714 #define IFF_ISATAP IFF_ISATAP
1715 #define IFF_WAN_HDLC IFF_WAN_HDLC
1716 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1717 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1718 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1719 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1720 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1721 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1722 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1723 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1724 #define IFF_TEAM_PORT IFF_TEAM_PORT
1725 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1726 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1727 #define IFF_MACVLAN IFF_MACVLAN
1728 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1729 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1730 #define IFF_NO_QUEUE IFF_NO_QUEUE
1731 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1732 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1733 #define IFF_TEAM IFF_TEAM
1734 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1735 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1736 #define IFF_MACSEC IFF_MACSEC
1737 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1738 #define IFF_FAILOVER IFF_FAILOVER
1739 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1740 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1741 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1742 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1744 /* Specifies the type of the struct net_device::ml_priv pointer */
1745 enum netdev_ml_priv_type {
1751 * struct net_device - The DEVICE structure.
1753 * Actually, this whole structure is a big mistake. It mixes I/O
1754 * data with strictly "high-level" data, and it has to know about
1755 * almost every data structure used in the INET module.
1757 * @name: This is the first field of the "visible" part of this structure
1758 * (i.e. as seen by users in the "Space.c" file). It is the name
1761 * @name_node: Name hashlist node
1762 * @ifalias: SNMP alias
1763 * @mem_end: Shared memory end
1764 * @mem_start: Shared memory start
1765 * @base_addr: Device I/O address
1766 * @irq: Device IRQ number
1768 * @state: Generic network queuing layer state, see netdev_state_t
1769 * @dev_list: The global list of network devices
1770 * @napi_list: List entry used for polling NAPI devices
1771 * @unreg_list: List entry when we are unregistering the
1772 * device; see the function unregister_netdev
1773 * @close_list: List entry used when we are closing the device
1774 * @ptype_all: Device-specific packet handlers for all protocols
1775 * @ptype_specific: Device-specific, protocol-specific packet handlers
1777 * @adj_list: Directly linked devices, like slaves for bonding
1778 * @features: Currently active device features
1779 * @hw_features: User-changeable features
1781 * @wanted_features: User-requested features
1782 * @vlan_features: Mask of features inheritable by VLAN devices
1784 * @hw_enc_features: Mask of features inherited by encapsulating devices
1785 * This field indicates what encapsulation
1786 * offloads the hardware is capable of doing,
1787 * and drivers will need to set them appropriately.
1789 * @mpls_features: Mask of features inheritable by MPLS
1790 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1792 * @ifindex: interface index
1793 * @group: The group the device belongs to
1795 * @stats: Statistics struct, which was left as a legacy, use
1796 * rtnl_link_stats64 instead
1798 * @core_stats: core networking counters,
1799 * do not use this in drivers
1800 * @carrier_up_count: Number of times the carrier has been up
1801 * @carrier_down_count: Number of times the carrier has been down
1803 * @wireless_handlers: List of functions to handle Wireless Extensions,
1805 * see <net/iw_handler.h> for details.
1806 * @wireless_data: Instance data managed by the core of wireless extensions
1808 * @netdev_ops: Includes several pointers to callbacks,
1809 * if one wants to override the ndo_*() functions
1810 * @ethtool_ops: Management operations
1811 * @l3mdev_ops: Layer 3 master device operations
1812 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1813 * discovery handling. Necessary for e.g. 6LoWPAN.
1814 * @xfrmdev_ops: Transformation offload operations
1815 * @tlsdev_ops: Transport Layer Security offload operations
1816 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1817 * of Layer 2 headers.
1819 * @flags: Interface flags (a la BSD)
1820 * @priv_flags: Like 'flags' but invisible to userspace,
1821 * see if.h for the definitions
1822 * @gflags: Global flags ( kept as legacy )
1823 * @padded: How much padding added by alloc_netdev()
1824 * @operstate: RFC2863 operstate
1825 * @link_mode: Mapping policy to operstate
1826 * @if_port: Selectable AUI, TP, ...
1828 * @mtu: Interface MTU value
1829 * @min_mtu: Interface Minimum MTU value
1830 * @max_mtu: Interface Maximum MTU value
1831 * @type: Interface hardware type
1832 * @hard_header_len: Maximum hardware header length.
1833 * @min_header_len: Minimum hardware header length
1835 * @needed_headroom: Extra headroom the hardware may need, but not in all
1836 * cases can this be guaranteed
1837 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1838 * cases can this be guaranteed. Some cases also use
1839 * LL_MAX_HEADER instead to allocate the skb
1841 * interface address info:
1843 * @perm_addr: Permanent hw address
1844 * @addr_assign_type: Hw address assignment type
1845 * @addr_len: Hardware address length
1846 * @upper_level: Maximum depth level of upper devices.
1847 * @lower_level: Maximum depth level of lower devices.
1848 * @neigh_priv_len: Used in neigh_alloc()
1849 * @dev_id: Used to differentiate devices that share
1850 * the same link layer address
1851 * @dev_port: Used to differentiate devices that share
1853 * @addr_list_lock: XXX: need comments on this one
1854 * @name_assign_type: network interface name assignment type
1855 * @uc_promisc: Counter that indicates promiscuous mode
1856 * has been enabled due to the need to listen to
1857 * additional unicast addresses in a device that
1858 * does not implement ndo_set_rx_mode()
1859 * @uc: unicast mac addresses
1860 * @mc: multicast mac addresses
1861 * @dev_addrs: list of device hw addresses
1862 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1863 * @promiscuity: Number of times the NIC is told to work in
1864 * promiscuous mode; if it becomes 0 the NIC will
1865 * exit promiscuous mode
1866 * @allmulti: Counter, enables or disables allmulticast mode
1868 * @vlan_info: VLAN info
1869 * @dsa_ptr: dsa specific data
1870 * @tipc_ptr: TIPC specific data
1871 * @atalk_ptr: AppleTalk link
1872 * @ip_ptr: IPv4 specific data
1873 * @ip6_ptr: IPv6 specific data
1874 * @ax25_ptr: AX.25 specific data
1875 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1876 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1878 * @mpls_ptr: mpls_dev struct pointer
1879 * @mctp_ptr: MCTP specific data
1881 * @dev_addr: Hw address (before bcast,
1882 * because most packets are unicast)
1884 * @_rx: Array of RX queues
1885 * @num_rx_queues: Number of RX queues
1886 * allocated at register_netdev() time
1887 * @real_num_rx_queues: Number of RX queues currently active in device
1888 * @xdp_prog: XDP sockets filter program pointer
1889 * @gro_flush_timeout: timeout for GRO layer in NAPI
1890 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1891 * allow to avoid NIC hard IRQ, on busy queues.
1893 * @rx_handler: handler for received packets
1894 * @rx_handler_data: XXX: need comments on this one
1895 * @miniq_ingress: ingress/clsact qdisc specific data for
1896 * ingress processing
1897 * @ingress_queue: XXX: need comments on this one
1898 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1899 * @broadcast: hw bcast address
1901 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1902 * indexed by RX queue number. Assigned by driver.
1903 * This must only be set if the ndo_rx_flow_steer
1904 * operation is defined
1905 * @index_hlist: Device index hash chain
1907 * @_tx: Array of TX queues
1908 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1909 * @real_num_tx_queues: Number of TX queues currently active in device
1910 * @qdisc: Root qdisc from userspace point of view
1911 * @tx_queue_len: Max frames per queue allowed
1912 * @tx_global_lock: XXX: need comments on this one
1913 * @xdp_bulkq: XDP device bulk queue
1914 * @xps_maps: all CPUs/RXQs maps for XPS device
1916 * @xps_maps: XXX: need comments on this one
1917 * @miniq_egress: clsact qdisc specific data for
1919 * @nf_hooks_egress: netfilter hooks executed for egress packets
1920 * @qdisc_hash: qdisc hash table
1921 * @watchdog_timeo: Represents the timeout that is used by
1922 * the watchdog (see dev_watchdog())
1923 * @watchdog_timer: List of timers
1925 * @proto_down_reason: reason a netdev interface is held down
1926 * @pcpu_refcnt: Number of references to this device
1927 * @dev_refcnt: Number of references to this device
1928 * @refcnt_tracker: Tracker directory for tracked references to this device
1929 * @todo_list: Delayed register/unregister
1930 * @link_watch_list: XXX: need comments on this one
1932 * @reg_state: Register/unregister state machine
1933 * @dismantle: Device is going to be freed
1934 * @rtnl_link_state: This enum represents the phases of creating
1937 * @needs_free_netdev: Should unregister perform free_netdev?
1938 * @priv_destructor: Called from unregister
1939 * @npinfo: XXX: need comments on this one
1940 * @nd_net: Network namespace this network device is inside
1942 * @ml_priv: Mid-layer private
1943 * @ml_priv_type: Mid-layer private type
1944 * @lstats: Loopback statistics
1945 * @tstats: Tunnel statistics
1946 * @dstats: Dummy statistics
1947 * @vstats: Virtual ethernet statistics
1952 * @dm_private: Drop monitor private
1954 * @dev: Class/net/name entry
1955 * @sysfs_groups: Space for optional device, statistics and wireless
1958 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1959 * @rtnl_link_ops: Rtnl_link_ops
1961 * @gso_max_size: Maximum size of generic segmentation offload
1962 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1963 * @gso_max_segs: Maximum number of segments that can be passed to the
1965 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1967 * @dcbnl_ops: Data Center Bridging netlink ops
1968 * @num_tc: Number of traffic classes in the net device
1969 * @tc_to_txq: XXX: need comments on this one
1970 * @prio_tc_map: XXX: need comments on this one
1972 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1974 * @priomap: XXX: need comments on this one
1975 * @phydev: Physical device may attach itself
1976 * for hardware timestamping
1977 * @sfp_bus: attached &struct sfp_bus structure.
1979 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1981 * @proto_down: protocol port state information can be sent to the
1982 * switch driver and used to set the phys state of the
1985 * @wol_enabled: Wake-on-LAN is enabled
1987 * @threaded: napi threaded mode is enabled
1989 * @net_notifier_list: List of per-net netdev notifier block
1990 * that follow this device when it is moved
1991 * to another network namespace.
1993 * @macsec_ops: MACsec offloading ops
1995 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1996 * offload capabilities of the device
1997 * @udp_tunnel_nic: UDP tunnel offload state
1998 * @xdp_state: stores info on attached XDP BPF programs
2000 * @nested_level: Used as a parameter of spin_lock_nested() of
2001 * dev->addr_list_lock.
2002 * @unlink_list: As netif_addr_lock() can be called recursively,
2003 * keep a list of interfaces to be deleted.
2004 * @gro_max_size: Maximum size of aggregated packet in generic
2005 * receive offload (GRO)
2007 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2008 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2009 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2010 * @dev_registered_tracker: tracker for reference held while
2012 * @offload_xstats_l3: L3 HW stats for this netdevice.
2014 * FIXME: cleanup struct net_device such that network protocol info
2019 char name[IFNAMSIZ];
2020 struct netdev_name_node *name_node;
2021 struct dev_ifalias __rcu *ifalias;
2023 * I/O specific fields
2024 * FIXME: Merge these and struct ifmap into one
2026 unsigned long mem_end;
2027 unsigned long mem_start;
2028 unsigned long base_addr;
2031 * Some hardware also needs these fields (state,dev_list,
2032 * napi_list,unreg_list,close_list) but they are not
2033 * part of the usual set specified in Space.c.
2036 unsigned long state;
2038 struct list_head dev_list;
2039 struct list_head napi_list;
2040 struct list_head unreg_list;
2041 struct list_head close_list;
2042 struct list_head ptype_all;
2043 struct list_head ptype_specific;
2046 struct list_head upper;
2047 struct list_head lower;
2050 /* Read-mostly cache-line for fast-path access */
2052 unsigned long long priv_flags;
2053 const struct net_device_ops *netdev_ops;
2055 unsigned short gflags;
2056 unsigned short hard_header_len;
2058 /* Note : dev->mtu is often read without holding a lock.
2059 * Writers usually hold RTNL.
2060 * It is recommended to use READ_ONCE() to annotate the reads,
2061 * and to use WRITE_ONCE() to annotate the writes.
2064 unsigned short needed_headroom;
2065 unsigned short needed_tailroom;
2067 netdev_features_t features;
2068 netdev_features_t hw_features;
2069 netdev_features_t wanted_features;
2070 netdev_features_t vlan_features;
2071 netdev_features_t hw_enc_features;
2072 netdev_features_t mpls_features;
2073 netdev_features_t gso_partial_features;
2075 unsigned int min_mtu;
2076 unsigned int max_mtu;
2077 unsigned short type;
2078 unsigned char min_header_len;
2079 unsigned char name_assign_type;
2083 struct net_device_stats stats; /* not used by modern drivers */
2085 struct net_device_core_stats __percpu *core_stats;
2087 /* Stats to monitor link on/off, flapping */
2088 atomic_t carrier_up_count;
2089 atomic_t carrier_down_count;
2091 #ifdef CONFIG_WIRELESS_EXT
2092 const struct iw_handler_def *wireless_handlers;
2093 struct iw_public_data *wireless_data;
2095 const struct ethtool_ops *ethtool_ops;
2096 #ifdef CONFIG_NET_L3_MASTER_DEV
2097 const struct l3mdev_ops *l3mdev_ops;
2099 #if IS_ENABLED(CONFIG_IPV6)
2100 const struct ndisc_ops *ndisc_ops;
2103 #ifdef CONFIG_XFRM_OFFLOAD
2104 const struct xfrmdev_ops *xfrmdev_ops;
2107 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2108 const struct tlsdev_ops *tlsdev_ops;
2111 const struct header_ops *header_ops;
2113 unsigned char operstate;
2114 unsigned char link_mode;
2116 unsigned char if_port;
2119 /* Interface address info. */
2120 unsigned char perm_addr[MAX_ADDR_LEN];
2121 unsigned char addr_assign_type;
2122 unsigned char addr_len;
2123 unsigned char upper_level;
2124 unsigned char lower_level;
2126 unsigned short neigh_priv_len;
2127 unsigned short dev_id;
2128 unsigned short dev_port;
2129 unsigned short padded;
2131 spinlock_t addr_list_lock;
2134 struct netdev_hw_addr_list uc;
2135 struct netdev_hw_addr_list mc;
2136 struct netdev_hw_addr_list dev_addrs;
2139 struct kset *queues_kset;
2141 #ifdef CONFIG_LOCKDEP
2142 struct list_head unlink_list;
2144 unsigned int promiscuity;
2145 unsigned int allmulti;
2147 #ifdef CONFIG_LOCKDEP
2148 unsigned char nested_level;
2152 /* Protocol-specific pointers */
2154 struct in_device __rcu *ip_ptr;
2155 struct inet6_dev __rcu *ip6_ptr;
2156 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2157 struct vlan_info __rcu *vlan_info;
2159 #if IS_ENABLED(CONFIG_NET_DSA)
2160 struct dsa_port *dsa_ptr;
2162 #if IS_ENABLED(CONFIG_TIPC)
2163 struct tipc_bearer __rcu *tipc_ptr;
2165 #if IS_ENABLED(CONFIG_ATALK)
2168 #if IS_ENABLED(CONFIG_AX25)
2171 #if IS_ENABLED(CONFIG_CFG80211)
2172 struct wireless_dev *ieee80211_ptr;
2174 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2175 struct wpan_dev *ieee802154_ptr;
2177 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2178 struct mpls_dev __rcu *mpls_ptr;
2180 #if IS_ENABLED(CONFIG_MCTP)
2181 struct mctp_dev __rcu *mctp_ptr;
2185 * Cache lines mostly used on receive path (including eth_type_trans())
2187 /* Interface address info used in eth_type_trans() */
2188 const unsigned char *dev_addr;
2190 struct netdev_rx_queue *_rx;
2191 unsigned int num_rx_queues;
2192 unsigned int real_num_rx_queues;
2194 struct bpf_prog __rcu *xdp_prog;
2195 unsigned long gro_flush_timeout;
2196 int napi_defer_hard_irqs;
2197 #define GRO_LEGACY_MAX_SIZE 65536u
2198 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2199 * and shinfo->gso_segs is a 16bit field.
2201 #define GRO_MAX_SIZE (8 * 65535u)
2202 unsigned int gro_max_size;
2203 rx_handler_func_t __rcu *rx_handler;
2204 void __rcu *rx_handler_data;
2206 #ifdef CONFIG_NET_CLS_ACT
2207 struct mini_Qdisc __rcu *miniq_ingress;
2209 struct netdev_queue __rcu *ingress_queue;
2210 #ifdef CONFIG_NETFILTER_INGRESS
2211 struct nf_hook_entries __rcu *nf_hooks_ingress;
2214 unsigned char broadcast[MAX_ADDR_LEN];
2215 #ifdef CONFIG_RFS_ACCEL
2216 struct cpu_rmap *rx_cpu_rmap;
2218 struct hlist_node index_hlist;
2221 * Cache lines mostly used on transmit path
2223 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2224 unsigned int num_tx_queues;
2225 unsigned int real_num_tx_queues;
2226 struct Qdisc __rcu *qdisc;
2227 unsigned int tx_queue_len;
2228 spinlock_t tx_global_lock;
2230 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2233 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2235 #ifdef CONFIG_NET_CLS_ACT
2236 struct mini_Qdisc __rcu *miniq_egress;
2238 #ifdef CONFIG_NETFILTER_EGRESS
2239 struct nf_hook_entries __rcu *nf_hooks_egress;
2242 #ifdef CONFIG_NET_SCHED
2243 DECLARE_HASHTABLE (qdisc_hash, 4);
2245 /* These may be needed for future network-power-down code. */
2246 struct timer_list watchdog_timer;
2249 u32 proto_down_reason;
2251 struct list_head todo_list;
2253 #ifdef CONFIG_PCPU_DEV_REFCNT
2254 int __percpu *pcpu_refcnt;
2256 refcount_t dev_refcnt;
2258 struct ref_tracker_dir refcnt_tracker;
2260 struct list_head link_watch_list;
2262 enum { NETREG_UNINITIALIZED=0,
2263 NETREG_REGISTERED, /* completed register_netdevice */
2264 NETREG_UNREGISTERING, /* called unregister_netdevice */
2265 NETREG_UNREGISTERED, /* completed unregister todo */
2266 NETREG_RELEASED, /* called free_netdev */
2267 NETREG_DUMMY, /* dummy device for NAPI poll */
2273 RTNL_LINK_INITIALIZED,
2274 RTNL_LINK_INITIALIZING,
2275 } rtnl_link_state:16;
2277 bool needs_free_netdev;
2278 void (*priv_destructor)(struct net_device *dev);
2280 #ifdef CONFIG_NETPOLL
2281 struct netpoll_info __rcu *npinfo;
2284 possible_net_t nd_net;
2286 /* mid-layer private */
2288 enum netdev_ml_priv_type ml_priv_type;
2291 struct pcpu_lstats __percpu *lstats;
2292 struct pcpu_sw_netstats __percpu *tstats;
2293 struct pcpu_dstats __percpu *dstats;
2296 #if IS_ENABLED(CONFIG_GARP)
2297 struct garp_port __rcu *garp_port;
2299 #if IS_ENABLED(CONFIG_MRP)
2300 struct mrp_port __rcu *mrp_port;
2302 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2303 struct dm_hw_stat_delta __rcu *dm_private;
2306 const struct attribute_group *sysfs_groups[4];
2307 const struct attribute_group *sysfs_rx_queue_group;
2309 const struct rtnl_link_ops *rtnl_link_ops;
2311 /* for setting kernel sock attribute on TCP connection setup */
2312 #define GSO_MAX_SEGS 65535u
2313 #define GSO_LEGACY_MAX_SIZE 65536u
2314 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2315 * and shinfo->gso_segs is a 16bit field.
2317 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2319 unsigned int gso_max_size;
2320 #define TSO_LEGACY_MAX_SIZE 65536
2321 #define TSO_MAX_SIZE UINT_MAX
2322 unsigned int tso_max_size;
2324 #define TSO_MAX_SEGS U16_MAX
2328 const struct dcbnl_rtnl_ops *dcbnl_ops;
2331 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2332 u8 prio_tc_map[TC_BITMASK + 1];
2334 #if IS_ENABLED(CONFIG_FCOE)
2335 unsigned int fcoe_ddp_xid;
2337 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2338 struct netprio_map __rcu *priomap;
2340 struct phy_device *phydev;
2341 struct sfp_bus *sfp_bus;
2342 struct lock_class_key *qdisc_tx_busylock;
2344 unsigned wol_enabled:1;
2345 unsigned threaded:1;
2347 struct list_head net_notifier_list;
2349 #if IS_ENABLED(CONFIG_MACSEC)
2350 /* MACsec management functions */
2351 const struct macsec_ops *macsec_ops;
2353 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2354 struct udp_tunnel_nic *udp_tunnel_nic;
2356 /* protected by rtnl_lock */
2357 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2359 u8 dev_addr_shadow[MAX_ADDR_LEN];
2360 netdevice_tracker linkwatch_dev_tracker;
2361 netdevice_tracker watchdog_dev_tracker;
2362 netdevice_tracker dev_registered_tracker;
2363 struct rtnl_hw_stats64 *offload_xstats_l3;
2365 #define to_net_dev(d) container_of(d, struct net_device, dev)
2367 static inline bool netif_elide_gro(const struct net_device *dev)
2369 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2374 #define NETDEV_ALIGN 32
2377 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2379 return dev->prio_tc_map[prio & TC_BITMASK];
2383 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2385 if (tc >= dev->num_tc)
2388 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2392 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2393 void netdev_reset_tc(struct net_device *dev);
2394 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2395 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2398 int netdev_get_num_tc(struct net_device *dev)
2403 static inline void net_prefetch(void *p)
2406 #if L1_CACHE_BYTES < 128
2407 prefetch((u8 *)p + L1_CACHE_BYTES);
2411 static inline void net_prefetchw(void *p)
2414 #if L1_CACHE_BYTES < 128
2415 prefetchw((u8 *)p + L1_CACHE_BYTES);
2419 void netdev_unbind_sb_channel(struct net_device *dev,
2420 struct net_device *sb_dev);
2421 int netdev_bind_sb_channel_queue(struct net_device *dev,
2422 struct net_device *sb_dev,
2423 u8 tc, u16 count, u16 offset);
2424 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2425 static inline int netdev_get_sb_channel(struct net_device *dev)
2427 return max_t(int, -dev->num_tc, 0);
2431 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2434 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2435 return &dev->_tx[index];
2438 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2439 const struct sk_buff *skb)
2441 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2444 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2445 void (*f)(struct net_device *,
2446 struct netdev_queue *,
2452 for (i = 0; i < dev->num_tx_queues; i++)
2453 f(dev, &dev->_tx[i], arg);
2456 #define netdev_lockdep_set_classes(dev) \
2458 static struct lock_class_key qdisc_tx_busylock_key; \
2459 static struct lock_class_key qdisc_xmit_lock_key; \
2460 static struct lock_class_key dev_addr_list_lock_key; \
2463 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2464 lockdep_set_class(&(dev)->addr_list_lock, \
2465 &dev_addr_list_lock_key); \
2466 for (i = 0; i < (dev)->num_tx_queues; i++) \
2467 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2468 &qdisc_xmit_lock_key); \
2471 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2472 struct net_device *sb_dev);
2473 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2474 struct sk_buff *skb,
2475 struct net_device *sb_dev);
2477 /* returns the headroom that the master device needs to take in account
2478 * when forwarding to this dev
2480 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2482 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2485 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2487 if (dev->netdev_ops->ndo_set_rx_headroom)
2488 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2491 /* set the device rx headroom to the dev's default */
2492 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2494 netdev_set_rx_headroom(dev, -1);
2497 static inline void *netdev_get_ml_priv(struct net_device *dev,
2498 enum netdev_ml_priv_type type)
2500 if (dev->ml_priv_type != type)
2503 return dev->ml_priv;
2506 static inline void netdev_set_ml_priv(struct net_device *dev,
2508 enum netdev_ml_priv_type type)
2510 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2511 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2512 dev->ml_priv_type, type);
2513 WARN(!dev->ml_priv_type && dev->ml_priv,
2514 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2516 dev->ml_priv = ml_priv;
2517 dev->ml_priv_type = type;
2521 * Net namespace inlines
2524 struct net *dev_net(const struct net_device *dev)
2526 return read_pnet(&dev->nd_net);
2530 void dev_net_set(struct net_device *dev, struct net *net)
2532 write_pnet(&dev->nd_net, net);
2536 * netdev_priv - access network device private data
2537 * @dev: network device
2539 * Get network device private data
2541 static inline void *netdev_priv(const struct net_device *dev)
2543 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2546 /* Set the sysfs physical device reference for the network logical device
2547 * if set prior to registration will cause a symlink during initialization.
2549 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2551 /* Set the sysfs device type for the network logical device to allow
2552 * fine-grained identification of different network device types. For
2553 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2555 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2557 /* Default NAPI poll() weight
2558 * Device drivers are strongly advised to not use bigger value
2560 #define NAPI_POLL_WEIGHT 64
2562 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2563 int (*poll)(struct napi_struct *, int), int weight);
2566 * netif_napi_add() - initialize a NAPI context
2567 * @dev: network device
2568 * @napi: NAPI context
2569 * @poll: polling function
2571 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2572 * *any* of the other NAPI-related functions.
2575 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2576 int (*poll)(struct napi_struct *, int))
2578 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2582 netif_napi_add_tx_weight(struct net_device *dev,
2583 struct napi_struct *napi,
2584 int (*poll)(struct napi_struct *, int),
2587 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2588 netif_napi_add_weight(dev, napi, poll, weight);
2592 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2593 * @dev: network device
2594 * @napi: NAPI context
2595 * @poll: polling function
2597 * This variant of netif_napi_add() should be used from drivers using NAPI
2598 * to exclusively poll a TX queue.
2599 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2601 static inline void netif_napi_add_tx(struct net_device *dev,
2602 struct napi_struct *napi,
2603 int (*poll)(struct napi_struct *, int))
2605 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2609 * __netif_napi_del - remove a NAPI context
2610 * @napi: NAPI context
2612 * Warning: caller must observe RCU grace period before freeing memory
2613 * containing @napi. Drivers might want to call this helper to combine
2614 * all the needed RCU grace periods into a single one.
2616 void __netif_napi_del(struct napi_struct *napi);
2619 * netif_napi_del - remove a NAPI context
2620 * @napi: NAPI context
2622 * netif_napi_del() removes a NAPI context from the network device NAPI list
2624 static inline void netif_napi_del(struct napi_struct *napi)
2626 __netif_napi_del(napi);
2630 struct packet_type {
2631 __be16 type; /* This is really htons(ether_type). */
2632 bool ignore_outgoing;
2633 struct net_device *dev; /* NULL is wildcarded here */
2634 netdevice_tracker dev_tracker;
2635 int (*func) (struct sk_buff *,
2636 struct net_device *,
2637 struct packet_type *,
2638 struct net_device *);
2639 void (*list_func) (struct list_head *,
2640 struct packet_type *,
2641 struct net_device *);
2642 bool (*id_match)(struct packet_type *ptype,
2644 struct net *af_packet_net;
2645 void *af_packet_priv;
2646 struct list_head list;
2649 struct offload_callbacks {
2650 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2651 netdev_features_t features);
2652 struct sk_buff *(*gro_receive)(struct list_head *head,
2653 struct sk_buff *skb);
2654 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2657 struct packet_offload {
2658 __be16 type; /* This is really htons(ether_type). */
2660 struct offload_callbacks callbacks;
2661 struct list_head list;
2664 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2665 struct pcpu_sw_netstats {
2666 u64_stats_t rx_packets;
2667 u64_stats_t rx_bytes;
2668 u64_stats_t tx_packets;
2669 u64_stats_t tx_bytes;
2670 struct u64_stats_sync syncp;
2671 } __aligned(4 * sizeof(u64));
2673 struct pcpu_lstats {
2674 u64_stats_t packets;
2676 struct u64_stats_sync syncp;
2677 } __aligned(2 * sizeof(u64));
2679 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2681 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2683 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2685 u64_stats_update_begin(&tstats->syncp);
2686 u64_stats_add(&tstats->rx_bytes, len);
2687 u64_stats_inc(&tstats->rx_packets);
2688 u64_stats_update_end(&tstats->syncp);
2691 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2692 unsigned int packets,
2695 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2697 u64_stats_update_begin(&tstats->syncp);
2698 u64_stats_add(&tstats->tx_bytes, len);
2699 u64_stats_add(&tstats->tx_packets, packets);
2700 u64_stats_update_end(&tstats->syncp);
2703 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2705 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2707 u64_stats_update_begin(&lstats->syncp);
2708 u64_stats_add(&lstats->bytes, len);
2709 u64_stats_inc(&lstats->packets);
2710 u64_stats_update_end(&lstats->syncp);
2713 #define __netdev_alloc_pcpu_stats(type, gfp) \
2715 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2718 for_each_possible_cpu(__cpu) { \
2719 typeof(type) *stat; \
2720 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2721 u64_stats_init(&stat->syncp); \
2727 #define netdev_alloc_pcpu_stats(type) \
2728 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2730 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2732 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2735 for_each_possible_cpu(__cpu) { \
2736 typeof(type) *stat; \
2737 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2738 u64_stats_init(&stat->syncp); \
2744 enum netdev_lag_tx_type {
2745 NETDEV_LAG_TX_TYPE_UNKNOWN,
2746 NETDEV_LAG_TX_TYPE_RANDOM,
2747 NETDEV_LAG_TX_TYPE_BROADCAST,
2748 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2749 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2750 NETDEV_LAG_TX_TYPE_HASH,
2753 enum netdev_lag_hash {
2754 NETDEV_LAG_HASH_NONE,
2756 NETDEV_LAG_HASH_L34,
2757 NETDEV_LAG_HASH_L23,
2758 NETDEV_LAG_HASH_E23,
2759 NETDEV_LAG_HASH_E34,
2760 NETDEV_LAG_HASH_VLAN_SRCMAC,
2761 NETDEV_LAG_HASH_UNKNOWN,
2764 struct netdev_lag_upper_info {
2765 enum netdev_lag_tx_type tx_type;
2766 enum netdev_lag_hash hash_type;
2769 struct netdev_lag_lower_state_info {
2774 #include <linux/notifier.h>
2776 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2777 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2781 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2783 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2784 detected a hardware crash and restarted
2785 - we can use this eg to kick tcp sessions
2787 NETDEV_CHANGE, /* Notify device state change */
2790 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2791 NETDEV_CHANGEADDR, /* notify after the address change */
2792 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2796 NETDEV_BONDING_FAILOVER,
2798 NETDEV_PRE_TYPE_CHANGE,
2799 NETDEV_POST_TYPE_CHANGE,
2802 NETDEV_NOTIFY_PEERS,
2806 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2807 NETDEV_CHANGEINFODATA,
2808 NETDEV_BONDING_INFO,
2809 NETDEV_PRECHANGEUPPER,
2810 NETDEV_CHANGELOWERSTATE,
2811 NETDEV_UDP_TUNNEL_PUSH_INFO,
2812 NETDEV_UDP_TUNNEL_DROP_INFO,
2813 NETDEV_CHANGE_TX_QUEUE_LEN,
2814 NETDEV_CVLAN_FILTER_PUSH_INFO,
2815 NETDEV_CVLAN_FILTER_DROP_INFO,
2816 NETDEV_SVLAN_FILTER_PUSH_INFO,
2817 NETDEV_SVLAN_FILTER_DROP_INFO,
2818 NETDEV_OFFLOAD_XSTATS_ENABLE,
2819 NETDEV_OFFLOAD_XSTATS_DISABLE,
2820 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2821 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2823 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2825 int register_netdevice_notifier(struct notifier_block *nb);
2826 int unregister_netdevice_notifier(struct notifier_block *nb);
2827 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2828 int unregister_netdevice_notifier_net(struct net *net,
2829 struct notifier_block *nb);
2830 int register_netdevice_notifier_dev_net(struct net_device *dev,
2831 struct notifier_block *nb,
2832 struct netdev_net_notifier *nn);
2833 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2834 struct notifier_block *nb,
2835 struct netdev_net_notifier *nn);
2837 struct netdev_notifier_info {
2838 struct net_device *dev;
2839 struct netlink_ext_ack *extack;
2842 struct netdev_notifier_info_ext {
2843 struct netdev_notifier_info info; /* must be first */
2849 struct netdev_notifier_change_info {
2850 struct netdev_notifier_info info; /* must be first */
2851 unsigned int flags_changed;
2854 struct netdev_notifier_changeupper_info {
2855 struct netdev_notifier_info info; /* must be first */
2856 struct net_device *upper_dev; /* new upper dev */
2857 bool master; /* is upper dev master */
2858 bool linking; /* is the notification for link or unlink */
2859 void *upper_info; /* upper dev info */
2862 struct netdev_notifier_changelowerstate_info {
2863 struct netdev_notifier_info info; /* must be first */
2864 void *lower_state_info; /* is lower dev state */
2867 struct netdev_notifier_pre_changeaddr_info {
2868 struct netdev_notifier_info info; /* must be first */
2869 const unsigned char *dev_addr;
2872 enum netdev_offload_xstats_type {
2873 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2876 struct netdev_notifier_offload_xstats_info {
2877 struct netdev_notifier_info info; /* must be first */
2878 enum netdev_offload_xstats_type type;
2881 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2882 struct netdev_notifier_offload_xstats_rd *report_delta;
2883 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2884 struct netdev_notifier_offload_xstats_ru *report_used;
2888 int netdev_offload_xstats_enable(struct net_device *dev,
2889 enum netdev_offload_xstats_type type,
2890 struct netlink_ext_ack *extack);
2891 int netdev_offload_xstats_disable(struct net_device *dev,
2892 enum netdev_offload_xstats_type type);
2893 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2894 enum netdev_offload_xstats_type type);
2895 int netdev_offload_xstats_get(struct net_device *dev,
2896 enum netdev_offload_xstats_type type,
2897 struct rtnl_hw_stats64 *stats, bool *used,
2898 struct netlink_ext_ack *extack);
2900 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2901 const struct rtnl_hw_stats64 *stats);
2903 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2904 void netdev_offload_xstats_push_delta(struct net_device *dev,
2905 enum netdev_offload_xstats_type type,
2906 const struct rtnl_hw_stats64 *stats);
2908 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2909 struct net_device *dev)
2912 info->extack = NULL;
2915 static inline struct net_device *
2916 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2921 static inline struct netlink_ext_ack *
2922 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2924 return info->extack;
2927 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2930 extern rwlock_t dev_base_lock; /* Device list lock */
2932 #define for_each_netdev(net, d) \
2933 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2934 #define for_each_netdev_reverse(net, d) \
2935 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2936 #define for_each_netdev_rcu(net, d) \
2937 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2938 #define for_each_netdev_safe(net, d, n) \
2939 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2940 #define for_each_netdev_continue(net, d) \
2941 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2942 #define for_each_netdev_continue_reverse(net, d) \
2943 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2945 #define for_each_netdev_continue_rcu(net, d) \
2946 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2947 #define for_each_netdev_in_bond_rcu(bond, slave) \
2948 for_each_netdev_rcu(&init_net, slave) \
2949 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2950 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2952 static inline struct net_device *next_net_device(struct net_device *dev)
2954 struct list_head *lh;
2958 lh = dev->dev_list.next;
2959 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2962 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2964 struct list_head *lh;
2968 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2969 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2972 static inline struct net_device *first_net_device(struct net *net)
2974 return list_empty(&net->dev_base_head) ? NULL :
2975 net_device_entry(net->dev_base_head.next);
2978 static inline struct net_device *first_net_device_rcu(struct net *net)
2980 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2982 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2985 int netdev_boot_setup_check(struct net_device *dev);
2986 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2987 const char *hwaddr);
2988 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2989 void dev_add_pack(struct packet_type *pt);
2990 void dev_remove_pack(struct packet_type *pt);
2991 void __dev_remove_pack(struct packet_type *pt);
2992 void dev_add_offload(struct packet_offload *po);
2993 void dev_remove_offload(struct packet_offload *po);
2995 int dev_get_iflink(const struct net_device *dev);
2996 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2997 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2998 struct net_device_path_stack *stack);
2999 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3000 unsigned short mask);
3001 struct net_device *dev_get_by_name(struct net *net, const char *name);
3002 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3003 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3004 bool netdev_name_in_use(struct net *net, const char *name);
3005 int dev_alloc_name(struct net_device *dev, const char *name);
3006 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3007 void dev_close(struct net_device *dev);
3008 void dev_close_many(struct list_head *head, bool unlink);
3009 void dev_disable_lro(struct net_device *dev);
3010 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3011 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3012 struct net_device *sb_dev);
3013 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3014 struct net_device *sb_dev);
3016 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3017 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3019 static inline int dev_queue_xmit(struct sk_buff *skb)
3021 return __dev_queue_xmit(skb, NULL);
3024 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3025 struct net_device *sb_dev)
3027 return __dev_queue_xmit(skb, sb_dev);
3030 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3034 ret = __dev_direct_xmit(skb, queue_id);
3035 if (!dev_xmit_complete(ret))
3040 int register_netdevice(struct net_device *dev);
3041 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3042 void unregister_netdevice_many(struct list_head *head);
3043 static inline void unregister_netdevice(struct net_device *dev)
3045 unregister_netdevice_queue(dev, NULL);
3048 int netdev_refcnt_read(const struct net_device *dev);
3049 void free_netdev(struct net_device *dev);
3050 void netdev_freemem(struct net_device *dev);
3051 int init_dummy_netdev(struct net_device *dev);
3053 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3054 struct sk_buff *skb,
3056 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3058 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3059 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3060 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3061 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3062 int dev_restart(struct net_device *dev);
3065 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3066 unsigned short type,
3067 const void *daddr, const void *saddr,
3070 if (!dev->header_ops || !dev->header_ops->create)
3073 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3076 static inline int dev_parse_header(const struct sk_buff *skb,
3077 unsigned char *haddr)
3079 const struct net_device *dev = skb->dev;
3081 if (!dev->header_ops || !dev->header_ops->parse)
3083 return dev->header_ops->parse(skb, haddr);
3086 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3088 const struct net_device *dev = skb->dev;
3090 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3092 return dev->header_ops->parse_protocol(skb);
3095 /* ll_header must have at least hard_header_len allocated */
3096 static inline bool dev_validate_header(const struct net_device *dev,
3097 char *ll_header, int len)
3099 if (likely(len >= dev->hard_header_len))
3101 if (len < dev->min_header_len)
3104 if (capable(CAP_SYS_RAWIO)) {
3105 memset(ll_header + len, 0, dev->hard_header_len - len);
3109 if (dev->header_ops && dev->header_ops->validate)
3110 return dev->header_ops->validate(ll_header, len);
3115 static inline bool dev_has_header(const struct net_device *dev)
3117 return dev->header_ops && dev->header_ops->create;
3121 * Incoming packets are placed on per-CPU queues
3123 struct softnet_data {
3124 struct list_head poll_list;
3125 struct sk_buff_head process_queue;
3128 unsigned int processed;
3129 unsigned int time_squeeze;
3130 unsigned int received_rps;
3132 struct softnet_data *rps_ipi_list;
3134 #ifdef CONFIG_NET_FLOW_LIMIT
3135 struct sd_flow_limit __rcu *flow_limit;
3137 struct Qdisc *output_queue;
3138 struct Qdisc **output_queue_tailp;
3139 struct sk_buff *completion_queue;
3140 #ifdef CONFIG_XFRM_OFFLOAD
3141 struct sk_buff_head xfrm_backlog;
3143 /* written and read only by owning cpu: */
3147 #ifdef CONFIG_NET_EGRESS
3152 /* input_queue_head should be written by cpu owning this struct,
3153 * and only read by other cpus. Worth using a cache line.
3155 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3157 /* Elements below can be accessed between CPUs for RPS/RFS */
3158 call_single_data_t csd ____cacheline_aligned_in_smp;
3159 struct softnet_data *rps_ipi_next;
3161 unsigned int input_queue_tail;
3163 unsigned int dropped;
3164 struct sk_buff_head input_pkt_queue;
3165 struct napi_struct backlog;
3167 /* Another possibly contended cache line */
3168 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3170 int defer_ipi_scheduled;
3171 struct sk_buff *defer_list;
3172 call_single_data_t defer_csd;
3175 static inline void input_queue_head_incr(struct softnet_data *sd)
3178 sd->input_queue_head++;
3182 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3183 unsigned int *qtail)
3186 *qtail = ++sd->input_queue_tail;
3190 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3192 static inline int dev_recursion_level(void)
3194 return this_cpu_read(softnet_data.xmit.recursion);
3197 #define XMIT_RECURSION_LIMIT 8
3198 static inline bool dev_xmit_recursion(void)
3200 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3201 XMIT_RECURSION_LIMIT);
3204 static inline void dev_xmit_recursion_inc(void)
3206 __this_cpu_inc(softnet_data.xmit.recursion);
3209 static inline void dev_xmit_recursion_dec(void)
3211 __this_cpu_dec(softnet_data.xmit.recursion);
3214 void __netif_schedule(struct Qdisc *q);
3215 void netif_schedule_queue(struct netdev_queue *txq);
3217 static inline void netif_tx_schedule_all(struct net_device *dev)
3221 for (i = 0; i < dev->num_tx_queues; i++)
3222 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3225 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3227 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3231 * netif_start_queue - allow transmit
3232 * @dev: network device
3234 * Allow upper layers to call the device hard_start_xmit routine.
3236 static inline void netif_start_queue(struct net_device *dev)
3238 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3241 static inline void netif_tx_start_all_queues(struct net_device *dev)
3245 for (i = 0; i < dev->num_tx_queues; i++) {
3246 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3247 netif_tx_start_queue(txq);
3251 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3254 * netif_wake_queue - restart transmit
3255 * @dev: network device
3257 * Allow upper layers to call the device hard_start_xmit routine.
3258 * Used for flow control when transmit resources are available.
3260 static inline void netif_wake_queue(struct net_device *dev)
3262 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3265 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3269 for (i = 0; i < dev->num_tx_queues; i++) {
3270 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3271 netif_tx_wake_queue(txq);
3275 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3277 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3281 * netif_stop_queue - stop transmitted packets
3282 * @dev: network device
3284 * Stop upper layers calling the device hard_start_xmit routine.
3285 * Used for flow control when transmit resources are unavailable.
3287 static inline void netif_stop_queue(struct net_device *dev)
3289 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3292 void netif_tx_stop_all_queues(struct net_device *dev);
3294 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3296 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3300 * netif_queue_stopped - test if transmit queue is flowblocked
3301 * @dev: network device
3303 * Test if transmit queue on device is currently unable to send.
3305 static inline bool netif_queue_stopped(const struct net_device *dev)
3307 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3310 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3312 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3316 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3318 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3322 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3324 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3328 * netdev_queue_set_dql_min_limit - set dql minimum limit
3329 * @dev_queue: pointer to transmit queue
3330 * @min_limit: dql minimum limit
3332 * Forces xmit_more() to return true until the minimum threshold
3333 * defined by @min_limit is reached (or until the tx queue is
3334 * empty). Warning: to be use with care, misuse will impact the
3337 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3338 unsigned int min_limit)
3341 dev_queue->dql.min_limit = min_limit;
3346 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3347 * @dev_queue: pointer to transmit queue
3349 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3350 * to give appropriate hint to the CPU.
3352 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3355 prefetchw(&dev_queue->dql.num_queued);
3360 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3361 * @dev_queue: pointer to transmit queue
3363 * BQL enabled drivers might use this helper in their TX completion path,
3364 * to give appropriate hint to the CPU.
3366 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3369 prefetchw(&dev_queue->dql.limit);
3374 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3375 * @dev_queue: network device queue
3376 * @bytes: number of bytes queued to the device queue
3378 * Report the number of bytes queued for sending/completion to the network
3379 * device hardware queue. @bytes should be a good approximation and should
3380 * exactly match netdev_completed_queue() @bytes.
3381 * This is typically called once per packet, from ndo_start_xmit().
3383 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3387 dql_queued(&dev_queue->dql, bytes);
3389 if (likely(dql_avail(&dev_queue->dql) >= 0))
3392 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3395 * The XOFF flag must be set before checking the dql_avail below,
3396 * because in netdev_tx_completed_queue we update the dql_completed
3397 * before checking the XOFF flag.
3401 /* check again in case another CPU has just made room avail */
3402 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3403 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3407 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3408 * that they should not test BQL status themselves.
3409 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3411 * Returns true if the doorbell must be used to kick the NIC.
3413 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3419 dql_queued(&dev_queue->dql, bytes);
3421 return netif_tx_queue_stopped(dev_queue);
3423 netdev_tx_sent_queue(dev_queue, bytes);
3428 * netdev_sent_queue - report the number of bytes queued to hardware
3429 * @dev: network device
3430 * @bytes: number of bytes queued to the hardware device queue
3432 * Report the number of bytes queued for sending/completion to the network
3433 * device hardware queue#0. @bytes should be a good approximation and should
3434 * exactly match netdev_completed_queue() @bytes.
3435 * This is typically called once per packet, from ndo_start_xmit().
3437 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3439 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3442 static inline bool __netdev_sent_queue(struct net_device *dev,
3446 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3451 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3452 * @dev_queue: network device queue
3453 * @pkts: number of packets (currently ignored)
3454 * @bytes: number of bytes dequeued from the device queue
3456 * Must be called at most once per TX completion round (and not per
3457 * individual packet), so that BQL can adjust its limits appropriately.
3459 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3460 unsigned int pkts, unsigned int bytes)
3463 if (unlikely(!bytes))
3466 dql_completed(&dev_queue->dql, bytes);
3469 * Without the memory barrier there is a small possiblity that
3470 * netdev_tx_sent_queue will miss the update and cause the queue to
3471 * be stopped forever
3475 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3478 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3479 netif_schedule_queue(dev_queue);
3484 * netdev_completed_queue - report bytes and packets completed by device
3485 * @dev: network device
3486 * @pkts: actual number of packets sent over the medium
3487 * @bytes: actual number of bytes sent over the medium
3489 * Report the number of bytes and packets transmitted by the network device
3490 * hardware queue over the physical medium, @bytes must exactly match the
3491 * @bytes amount passed to netdev_sent_queue()
3493 static inline void netdev_completed_queue(struct net_device *dev,
3494 unsigned int pkts, unsigned int bytes)
3496 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3499 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3502 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3508 * netdev_reset_queue - reset the packets and bytes count of a network device
3509 * @dev_queue: network device
3511 * Reset the bytes and packet count of a network device and clear the
3512 * software flow control OFF bit for this network device
3514 static inline void netdev_reset_queue(struct net_device *dev_queue)
3516 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3520 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3521 * @dev: network device
3522 * @queue_index: given tx queue index
3524 * Returns 0 if given tx queue index >= number of device tx queues,
3525 * otherwise returns the originally passed tx queue index.
3527 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3529 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3530 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3531 dev->name, queue_index,
3532 dev->real_num_tx_queues);
3540 * netif_running - test if up
3541 * @dev: network device
3543 * Test if the device has been brought up.
3545 static inline bool netif_running(const struct net_device *dev)
3547 return test_bit(__LINK_STATE_START, &dev->state);
3551 * Routines to manage the subqueues on a device. We only need start,
3552 * stop, and a check if it's stopped. All other device management is
3553 * done at the overall netdevice level.
3554 * Also test the device if we're multiqueue.
3558 * netif_start_subqueue - allow sending packets on subqueue
3559 * @dev: network device
3560 * @queue_index: sub queue index
3562 * Start individual transmit queue of a device with multiple transmit queues.
3564 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3566 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3568 netif_tx_start_queue(txq);
3572 * netif_stop_subqueue - stop sending packets on subqueue
3573 * @dev: network device
3574 * @queue_index: sub queue index
3576 * Stop individual transmit queue of a device with multiple transmit queues.
3578 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3580 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3581 netif_tx_stop_queue(txq);
3585 * __netif_subqueue_stopped - test status of subqueue
3586 * @dev: network device
3587 * @queue_index: sub queue index
3589 * Check individual transmit queue of a device with multiple transmit queues.
3591 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3594 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3596 return netif_tx_queue_stopped(txq);
3600 * netif_subqueue_stopped - test status of subqueue
3601 * @dev: network device
3602 * @skb: sub queue buffer pointer
3604 * Check individual transmit queue of a device with multiple transmit queues.
3606 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3607 struct sk_buff *skb)
3609 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3613 * netif_wake_subqueue - allow sending packets on subqueue
3614 * @dev: network device
3615 * @queue_index: sub queue index
3617 * Resume individual transmit queue of a device with multiple transmit queues.
3619 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3621 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3623 netif_tx_wake_queue(txq);
3627 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3629 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3630 u16 index, enum xps_map_type type);
3633 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3634 * @j: CPU/Rx queue index
3635 * @mask: bitmask of all cpus/rx queues
3636 * @nr_bits: number of bits in the bitmask
3638 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3640 static inline bool netif_attr_test_mask(unsigned long j,
3641 const unsigned long *mask,
3642 unsigned int nr_bits)
3644 cpu_max_bits_warn(j, nr_bits);
3645 return test_bit(j, mask);
3649 * netif_attr_test_online - Test for online CPU/Rx queue
3650 * @j: CPU/Rx queue index
3651 * @online_mask: bitmask for CPUs/Rx queues that are online
3652 * @nr_bits: number of bits in the bitmask
3654 * Returns true if a CPU/Rx queue is online.
3656 static inline bool netif_attr_test_online(unsigned long j,
3657 const unsigned long *online_mask,
3658 unsigned int nr_bits)
3660 cpu_max_bits_warn(j, nr_bits);
3663 return test_bit(j, online_mask);
3665 return (j < nr_bits);
3669 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3670 * @n: CPU/Rx queue index
3671 * @srcp: the cpumask/Rx queue mask pointer
3672 * @nr_bits: number of bits in the bitmask
3674 * Returns >= nr_bits if no further CPUs/Rx queues set.
3676 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3677 unsigned int nr_bits)
3679 /* -1 is a legal arg here. */
3681 cpu_max_bits_warn(n, nr_bits);
3684 return find_next_bit(srcp, nr_bits, n + 1);
3690 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3691 * @n: CPU/Rx queue index
3692 * @src1p: the first CPUs/Rx queues mask pointer
3693 * @src2p: the second CPUs/Rx queues mask pointer
3694 * @nr_bits: number of bits in the bitmask
3696 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3698 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3699 const unsigned long *src2p,
3700 unsigned int nr_bits)
3702 /* -1 is a legal arg here. */
3704 cpu_max_bits_warn(n, nr_bits);
3707 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3709 return find_next_bit(src1p, nr_bits, n + 1);
3711 return find_next_bit(src2p, nr_bits, n + 1);
3716 static inline int netif_set_xps_queue(struct net_device *dev,
3717 const struct cpumask *mask,
3723 static inline int __netif_set_xps_queue(struct net_device *dev,
3724 const unsigned long *mask,
3725 u16 index, enum xps_map_type type)
3732 * netif_is_multiqueue - test if device has multiple transmit queues
3733 * @dev: network device
3735 * Check if device has multiple transmit queues
3737 static inline bool netif_is_multiqueue(const struct net_device *dev)
3739 return dev->num_tx_queues > 1;
3742 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3745 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3747 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3750 dev->real_num_rx_queues = rxqs;
3754 int netif_set_real_num_queues(struct net_device *dev,
3755 unsigned int txq, unsigned int rxq);
3757 static inline struct netdev_rx_queue *
3758 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3760 return dev->_rx + rxq;
3764 static inline unsigned int get_netdev_rx_queue_index(
3765 struct netdev_rx_queue *queue)
3767 struct net_device *dev = queue->dev;
3768 int index = queue - dev->_rx;
3770 BUG_ON(index >= dev->num_rx_queues);
3775 int netif_get_num_default_rss_queues(void);
3777 enum skb_free_reason {
3778 SKB_REASON_CONSUMED,
3782 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3783 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3786 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3787 * interrupt context or with hardware interrupts being disabled.
3788 * (in_hardirq() || irqs_disabled())
3790 * We provide four helpers that can be used in following contexts :
3792 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3793 * replacing kfree_skb(skb)
3795 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3796 * Typically used in place of consume_skb(skb) in TX completion path
3798 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3799 * replacing kfree_skb(skb)
3801 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3802 * and consumed a packet. Used in place of consume_skb(skb)
3804 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3806 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3809 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3811 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3814 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3816 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3819 static inline void dev_consume_skb_any(struct sk_buff *skb)
3821 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3824 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3825 struct bpf_prog *xdp_prog);
3826 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3827 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3828 int netif_rx(struct sk_buff *skb);
3829 int __netif_rx(struct sk_buff *skb);
3831 int netif_receive_skb(struct sk_buff *skb);
3832 int netif_receive_skb_core(struct sk_buff *skb);
3833 void netif_receive_skb_list_internal(struct list_head *head);
3834 void netif_receive_skb_list(struct list_head *head);
3835 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3836 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3837 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3838 void napi_get_frags_check(struct napi_struct *napi);
3839 gro_result_t napi_gro_frags(struct napi_struct *napi);
3840 struct packet_offload *gro_find_receive_by_type(__be16 type);
3841 struct packet_offload *gro_find_complete_by_type(__be16 type);
3843 static inline void napi_free_frags(struct napi_struct *napi)
3845 kfree_skb(napi->skb);
3849 bool netdev_is_rx_handler_busy(struct net_device *dev);
3850 int netdev_rx_handler_register(struct net_device *dev,
3851 rx_handler_func_t *rx_handler,
3852 void *rx_handler_data);
3853 void netdev_rx_handler_unregister(struct net_device *dev);
3855 bool dev_valid_name(const char *name);
3856 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3858 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3860 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3861 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3862 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3863 void __user *data, bool *need_copyout);
3864 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3865 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3866 unsigned int dev_get_flags(const struct net_device *);
3867 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3868 struct netlink_ext_ack *extack);
3869 int dev_change_flags(struct net_device *dev, unsigned int flags,
3870 struct netlink_ext_ack *extack);
3871 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3872 unsigned int gchanges);
3873 int dev_set_alias(struct net_device *, const char *, size_t);
3874 int dev_get_alias(const struct net_device *, char *, size_t);
3875 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3876 const char *pat, int new_ifindex);
3878 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3881 return __dev_change_net_namespace(dev, net, pat, 0);
3883 int __dev_set_mtu(struct net_device *, int);
3884 int dev_set_mtu(struct net_device *, int);
3885 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3886 struct netlink_ext_ack *extack);
3887 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3888 struct netlink_ext_ack *extack);
3889 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3890 struct netlink_ext_ack *extack);
3891 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3892 int dev_get_port_parent_id(struct net_device *dev,
3893 struct netdev_phys_item_id *ppid, bool recurse);
3894 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3895 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3896 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3897 struct netdev_queue *txq, int *ret);
3899 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3900 u8 dev_xdp_prog_count(struct net_device *dev);
3901 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3903 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3904 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3905 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3906 bool is_skb_forwardable(const struct net_device *dev,
3907 const struct sk_buff *skb);
3909 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3910 const struct sk_buff *skb,
3911 const bool check_mtu)
3913 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3916 if (!(dev->flags & IFF_UP))
3922 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3923 if (skb->len <= len)
3926 /* if TSO is enabled, we don't care about the length as the packet
3927 * could be forwarded without being segmented before
3929 if (skb_is_gso(skb))
3935 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3937 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3939 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3940 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3945 return netdev_core_stats_alloc(dev);
3948 #define DEV_CORE_STATS_INC(FIELD) \
3949 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3951 struct net_device_core_stats __percpu *p; \
3953 p = dev_core_stats(dev); \
3955 this_cpu_inc(p->FIELD); \
3957 DEV_CORE_STATS_INC(rx_dropped)
3958 DEV_CORE_STATS_INC(tx_dropped)
3959 DEV_CORE_STATS_INC(rx_nohandler)
3960 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3962 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3963 struct sk_buff *skb,
3964 const bool check_mtu)
3966 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3967 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3968 dev_core_stats_rx_dropped_inc(dev);
3973 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3978 bool dev_nit_active(struct net_device *dev);
3979 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3981 static inline void __dev_put(struct net_device *dev)
3984 #ifdef CONFIG_PCPU_DEV_REFCNT
3985 this_cpu_dec(*dev->pcpu_refcnt);
3987 refcount_dec(&dev->dev_refcnt);
3992 static inline void __dev_hold(struct net_device *dev)
3995 #ifdef CONFIG_PCPU_DEV_REFCNT
3996 this_cpu_inc(*dev->pcpu_refcnt);
3998 refcount_inc(&dev->dev_refcnt);
4003 static inline void __netdev_tracker_alloc(struct net_device *dev,
4004 netdevice_tracker *tracker,
4007 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4008 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4012 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4013 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4015 static inline void netdev_tracker_alloc(struct net_device *dev,
4016 netdevice_tracker *tracker, gfp_t gfp)
4018 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4019 refcount_dec(&dev->refcnt_tracker.no_tracker);
4020 __netdev_tracker_alloc(dev, tracker, gfp);
4024 static inline void netdev_tracker_free(struct net_device *dev,
4025 netdevice_tracker *tracker)
4027 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4028 ref_tracker_free(&dev->refcnt_tracker, tracker);
4032 static inline void netdev_hold(struct net_device *dev,
4033 netdevice_tracker *tracker, gfp_t gfp)
4037 __netdev_tracker_alloc(dev, tracker, gfp);
4041 static inline void netdev_put(struct net_device *dev,
4042 netdevice_tracker *tracker)
4045 netdev_tracker_free(dev, tracker);
4051 * dev_hold - get reference to device
4052 * @dev: network device
4054 * Hold reference to device to keep it from being freed.
4055 * Try using netdev_hold() instead.
4057 static inline void dev_hold(struct net_device *dev)
4059 netdev_hold(dev, NULL, GFP_ATOMIC);
4063 * dev_put - release reference to device
4064 * @dev: network device
4066 * Release reference to device to allow it to be freed.
4067 * Try using netdev_put() instead.
4069 static inline void dev_put(struct net_device *dev)
4071 netdev_put(dev, NULL);
4074 static inline void netdev_ref_replace(struct net_device *odev,
4075 struct net_device *ndev,
4076 netdevice_tracker *tracker,
4080 netdev_tracker_free(odev, tracker);
4086 __netdev_tracker_alloc(ndev, tracker, gfp);
4089 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4090 * and _off may be called from IRQ context, but it is caller
4091 * who is responsible for serialization of these calls.
4093 * The name carrier is inappropriate, these functions should really be
4094 * called netif_lowerlayer_*() because they represent the state of any
4095 * kind of lower layer not just hardware media.
4097 void linkwatch_fire_event(struct net_device *dev);
4100 * netif_carrier_ok - test if carrier present
4101 * @dev: network device
4103 * Check if carrier is present on device
4105 static inline bool netif_carrier_ok(const struct net_device *dev)
4107 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4110 unsigned long dev_trans_start(struct net_device *dev);
4112 void __netdev_watchdog_up(struct net_device *dev);
4114 void netif_carrier_on(struct net_device *dev);
4115 void netif_carrier_off(struct net_device *dev);
4116 void netif_carrier_event(struct net_device *dev);
4119 * netif_dormant_on - mark device as dormant.
4120 * @dev: network device
4122 * Mark device as dormant (as per RFC2863).
4124 * The dormant state indicates that the relevant interface is not
4125 * actually in a condition to pass packets (i.e., it is not 'up') but is
4126 * in a "pending" state, waiting for some external event. For "on-
4127 * demand" interfaces, this new state identifies the situation where the
4128 * interface is waiting for events to place it in the up state.
4130 static inline void netif_dormant_on(struct net_device *dev)
4132 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4133 linkwatch_fire_event(dev);
4137 * netif_dormant_off - set device as not dormant.
4138 * @dev: network device
4140 * Device is not in dormant state.
4142 static inline void netif_dormant_off(struct net_device *dev)
4144 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4145 linkwatch_fire_event(dev);
4149 * netif_dormant - test if device is dormant
4150 * @dev: network device
4152 * Check if device is dormant.
4154 static inline bool netif_dormant(const struct net_device *dev)
4156 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4161 * netif_testing_on - mark device as under test.
4162 * @dev: network device
4164 * Mark device as under test (as per RFC2863).
4166 * The testing state indicates that some test(s) must be performed on
4167 * the interface. After completion, of the test, the interface state
4168 * will change to up, dormant, or down, as appropriate.
4170 static inline void netif_testing_on(struct net_device *dev)
4172 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4173 linkwatch_fire_event(dev);
4177 * netif_testing_off - set device as not under test.
4178 * @dev: network device
4180 * Device is not in testing state.
4182 static inline void netif_testing_off(struct net_device *dev)
4184 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4185 linkwatch_fire_event(dev);
4189 * netif_testing - test if device is under test
4190 * @dev: network device
4192 * Check if device is under test
4194 static inline bool netif_testing(const struct net_device *dev)
4196 return test_bit(__LINK_STATE_TESTING, &dev->state);
4201 * netif_oper_up - test if device is operational
4202 * @dev: network device
4204 * Check if carrier is operational
4206 static inline bool netif_oper_up(const struct net_device *dev)
4208 return (dev->operstate == IF_OPER_UP ||
4209 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4213 * netif_device_present - is device available or removed
4214 * @dev: network device
4216 * Check if device has not been removed from system.
4218 static inline bool netif_device_present(const struct net_device *dev)
4220 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4223 void netif_device_detach(struct net_device *dev);
4225 void netif_device_attach(struct net_device *dev);
4228 * Network interface message level settings
4233 NETIF_MSG_PROBE_BIT,
4235 NETIF_MSG_TIMER_BIT,
4236 NETIF_MSG_IFDOWN_BIT,
4238 NETIF_MSG_RX_ERR_BIT,
4239 NETIF_MSG_TX_ERR_BIT,
4240 NETIF_MSG_TX_QUEUED_BIT,
4242 NETIF_MSG_TX_DONE_BIT,
4243 NETIF_MSG_RX_STATUS_BIT,
4244 NETIF_MSG_PKTDATA_BIT,
4248 /* When you add a new bit above, update netif_msg_class_names array
4249 * in net/ethtool/common.c
4251 NETIF_MSG_CLASS_COUNT,
4253 /* Both ethtool_ops interface and internal driver implementation use u32 */
4254 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4256 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4257 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4259 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4260 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4261 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4262 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4263 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4264 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4265 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4266 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4267 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4268 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4269 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4270 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4271 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4272 #define NETIF_MSG_HW __NETIF_MSG(HW)
4273 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4275 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4276 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4277 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4278 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4279 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4280 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4281 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4282 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4283 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4284 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4285 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4286 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4287 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4288 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4289 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4291 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4294 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4295 return default_msg_enable_bits;
4296 if (debug_value == 0) /* no output */
4298 /* set low N bits */
4299 return (1U << debug_value) - 1;
4302 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4304 spin_lock(&txq->_xmit_lock);
4305 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4306 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4309 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4311 __acquire(&txq->_xmit_lock);
4315 static inline void __netif_tx_release(struct netdev_queue *txq)
4317 __release(&txq->_xmit_lock);
4320 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4322 spin_lock_bh(&txq->_xmit_lock);
4323 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4324 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4327 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4329 bool ok = spin_trylock(&txq->_xmit_lock);
4332 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4333 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4338 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4340 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4341 WRITE_ONCE(txq->xmit_lock_owner, -1);
4342 spin_unlock(&txq->_xmit_lock);
4345 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4347 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4348 WRITE_ONCE(txq->xmit_lock_owner, -1);
4349 spin_unlock_bh(&txq->_xmit_lock);
4353 * txq->trans_start can be read locklessly from dev_watchdog()
4355 static inline void txq_trans_update(struct netdev_queue *txq)
4357 if (txq->xmit_lock_owner != -1)
4358 WRITE_ONCE(txq->trans_start, jiffies);
4361 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4363 unsigned long now = jiffies;
4365 if (READ_ONCE(txq->trans_start) != now)
4366 WRITE_ONCE(txq->trans_start, now);
4369 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4370 static inline void netif_trans_update(struct net_device *dev)
4372 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4374 txq_trans_cond_update(txq);
4378 * netif_tx_lock - grab network device transmit lock
4379 * @dev: network device
4381 * Get network device transmit lock
4383 void netif_tx_lock(struct net_device *dev);
4385 static inline void netif_tx_lock_bh(struct net_device *dev)
4391 void netif_tx_unlock(struct net_device *dev);
4393 static inline void netif_tx_unlock_bh(struct net_device *dev)
4395 netif_tx_unlock(dev);
4399 #define HARD_TX_LOCK(dev, txq, cpu) { \
4400 if ((dev->features & NETIF_F_LLTX) == 0) { \
4401 __netif_tx_lock(txq, cpu); \
4403 __netif_tx_acquire(txq); \
4407 #define HARD_TX_TRYLOCK(dev, txq) \
4408 (((dev->features & NETIF_F_LLTX) == 0) ? \
4409 __netif_tx_trylock(txq) : \
4410 __netif_tx_acquire(txq))
4412 #define HARD_TX_UNLOCK(dev, txq) { \
4413 if ((dev->features & NETIF_F_LLTX) == 0) { \
4414 __netif_tx_unlock(txq); \
4416 __netif_tx_release(txq); \
4420 static inline void netif_tx_disable(struct net_device *dev)
4426 cpu = smp_processor_id();
4427 spin_lock(&dev->tx_global_lock);
4428 for (i = 0; i < dev->num_tx_queues; i++) {
4429 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4431 __netif_tx_lock(txq, cpu);
4432 netif_tx_stop_queue(txq);
4433 __netif_tx_unlock(txq);
4435 spin_unlock(&dev->tx_global_lock);
4439 static inline void netif_addr_lock(struct net_device *dev)
4441 unsigned char nest_level = 0;
4443 #ifdef CONFIG_LOCKDEP
4444 nest_level = dev->nested_level;
4446 spin_lock_nested(&dev->addr_list_lock, nest_level);
4449 static inline void netif_addr_lock_bh(struct net_device *dev)
4451 unsigned char nest_level = 0;
4453 #ifdef CONFIG_LOCKDEP
4454 nest_level = dev->nested_level;
4457 spin_lock_nested(&dev->addr_list_lock, nest_level);
4460 static inline void netif_addr_unlock(struct net_device *dev)
4462 spin_unlock(&dev->addr_list_lock);
4465 static inline void netif_addr_unlock_bh(struct net_device *dev)
4467 spin_unlock_bh(&dev->addr_list_lock);
4471 * dev_addrs walker. Should be used only for read access. Call with
4472 * rcu_read_lock held.
4474 #define for_each_dev_addr(dev, ha) \
4475 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4477 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4479 void ether_setup(struct net_device *dev);
4481 /* Support for loadable net-drivers */
4482 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4483 unsigned char name_assign_type,
4484 void (*setup)(struct net_device *),
4485 unsigned int txqs, unsigned int rxqs);
4486 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4487 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4489 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4490 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4493 int register_netdev(struct net_device *dev);
4494 void unregister_netdev(struct net_device *dev);
4496 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4498 /* General hardware address lists handling functions */
4499 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4500 struct netdev_hw_addr_list *from_list, int addr_len);
4501 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4502 struct netdev_hw_addr_list *from_list, int addr_len);
4503 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4504 struct net_device *dev,
4505 int (*sync)(struct net_device *, const unsigned char *),
4506 int (*unsync)(struct net_device *,
4507 const unsigned char *));
4508 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4509 struct net_device *dev,
4510 int (*sync)(struct net_device *,
4511 const unsigned char *, int),
4512 int (*unsync)(struct net_device *,
4513 const unsigned char *, int));
4514 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4515 struct net_device *dev,
4516 int (*unsync)(struct net_device *,
4517 const unsigned char *, int));
4518 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4519 struct net_device *dev,
4520 int (*unsync)(struct net_device *,
4521 const unsigned char *));
4522 void __hw_addr_init(struct netdev_hw_addr_list *list);
4524 /* Functions used for device addresses handling */
4525 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4526 const void *addr, size_t len);
4529 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4531 dev_addr_mod(dev, 0, addr, len);
4534 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4536 __dev_addr_set(dev, addr, dev->addr_len);
4539 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4540 unsigned char addr_type);
4541 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4542 unsigned char addr_type);
4544 /* Functions used for unicast addresses handling */
4545 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4546 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4547 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4548 int dev_uc_sync(struct net_device *to, struct net_device *from);
4549 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4550 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4551 void dev_uc_flush(struct net_device *dev);
4552 void dev_uc_init(struct net_device *dev);
4555 * __dev_uc_sync - Synchonize device's unicast list
4556 * @dev: device to sync
4557 * @sync: function to call if address should be added
4558 * @unsync: function to call if address should be removed
4560 * Add newly added addresses to the interface, and release
4561 * addresses that have been deleted.
4563 static inline int __dev_uc_sync(struct net_device *dev,
4564 int (*sync)(struct net_device *,
4565 const unsigned char *),
4566 int (*unsync)(struct net_device *,
4567 const unsigned char *))
4569 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4573 * __dev_uc_unsync - Remove synchronized addresses from device
4574 * @dev: device to sync
4575 * @unsync: function to call if address should be removed
4577 * Remove all addresses that were added to the device by dev_uc_sync().
4579 static inline void __dev_uc_unsync(struct net_device *dev,
4580 int (*unsync)(struct net_device *,
4581 const unsigned char *))
4583 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4586 /* Functions used for multicast addresses handling */
4587 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4588 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4589 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4590 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4591 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4592 int dev_mc_sync(struct net_device *to, struct net_device *from);
4593 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4594 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4595 void dev_mc_flush(struct net_device *dev);
4596 void dev_mc_init(struct net_device *dev);
4599 * __dev_mc_sync - Synchonize device's multicast list
4600 * @dev: device to sync
4601 * @sync: function to call if address should be added
4602 * @unsync: function to call if address should be removed
4604 * Add newly added addresses to the interface, and release
4605 * addresses that have been deleted.
4607 static inline int __dev_mc_sync(struct net_device *dev,
4608 int (*sync)(struct net_device *,
4609 const unsigned char *),
4610 int (*unsync)(struct net_device *,
4611 const unsigned char *))
4613 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4617 * __dev_mc_unsync - Remove synchronized addresses from device
4618 * @dev: device to sync
4619 * @unsync: function to call if address should be removed
4621 * Remove all addresses that were added to the device by dev_mc_sync().
4623 static inline void __dev_mc_unsync(struct net_device *dev,
4624 int (*unsync)(struct net_device *,
4625 const unsigned char *))
4627 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4630 /* Functions used for secondary unicast and multicast support */
4631 void dev_set_rx_mode(struct net_device *dev);
4632 int dev_set_promiscuity(struct net_device *dev, int inc);
4633 int dev_set_allmulti(struct net_device *dev, int inc);
4634 void netdev_state_change(struct net_device *dev);
4635 void __netdev_notify_peers(struct net_device *dev);
4636 void netdev_notify_peers(struct net_device *dev);
4637 void netdev_features_change(struct net_device *dev);
4638 /* Load a device via the kmod */
4639 void dev_load(struct net *net, const char *name);
4640 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4641 struct rtnl_link_stats64 *storage);
4642 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4643 const struct net_device_stats *netdev_stats);
4644 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4645 const struct pcpu_sw_netstats __percpu *netstats);
4646 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4648 extern int netdev_max_backlog;
4649 extern int dev_rx_weight;
4650 extern int dev_tx_weight;
4651 extern int gro_normal_batch;
4654 NESTED_SYNC_IMM_BIT,
4655 NESTED_SYNC_TODO_BIT,
4658 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4659 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4661 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4662 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4664 struct netdev_nested_priv {
4665 unsigned char flags;
4669 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4670 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4671 struct list_head **iter);
4673 /* iterate through upper list, must be called under RCU read lock */
4674 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4675 for (iter = &(dev)->adj_list.upper, \
4676 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4678 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4680 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4681 int (*fn)(struct net_device *upper_dev,
4682 struct netdev_nested_priv *priv),
4683 struct netdev_nested_priv *priv);
4685 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4686 struct net_device *upper_dev);
4688 bool netdev_has_any_upper_dev(struct net_device *dev);
4690 void *netdev_lower_get_next_private(struct net_device *dev,
4691 struct list_head **iter);
4692 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4693 struct list_head **iter);
4695 #define netdev_for_each_lower_private(dev, priv, iter) \
4696 for (iter = (dev)->adj_list.lower.next, \
4697 priv = netdev_lower_get_next_private(dev, &(iter)); \
4699 priv = netdev_lower_get_next_private(dev, &(iter)))
4701 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4702 for (iter = &(dev)->adj_list.lower, \
4703 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4705 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4707 void *netdev_lower_get_next(struct net_device *dev,
4708 struct list_head **iter);
4710 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4711 for (iter = (dev)->adj_list.lower.next, \
4712 ldev = netdev_lower_get_next(dev, &(iter)); \
4714 ldev = netdev_lower_get_next(dev, &(iter)))
4716 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4717 struct list_head **iter);
4718 int netdev_walk_all_lower_dev(struct net_device *dev,
4719 int (*fn)(struct net_device *lower_dev,
4720 struct netdev_nested_priv *priv),
4721 struct netdev_nested_priv *priv);
4722 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4723 int (*fn)(struct net_device *lower_dev,
4724 struct netdev_nested_priv *priv),
4725 struct netdev_nested_priv *priv);
4727 void *netdev_adjacent_get_private(struct list_head *adj_list);
4728 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4729 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4730 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4731 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4732 struct netlink_ext_ack *extack);
4733 int netdev_master_upper_dev_link(struct net_device *dev,
4734 struct net_device *upper_dev,
4735 void *upper_priv, void *upper_info,
4736 struct netlink_ext_ack *extack);
4737 void netdev_upper_dev_unlink(struct net_device *dev,
4738 struct net_device *upper_dev);
4739 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4740 struct net_device *new_dev,
4741 struct net_device *dev,
4742 struct netlink_ext_ack *extack);
4743 void netdev_adjacent_change_commit(struct net_device *old_dev,
4744 struct net_device *new_dev,
4745 struct net_device *dev);
4746 void netdev_adjacent_change_abort(struct net_device *old_dev,
4747 struct net_device *new_dev,
4748 struct net_device *dev);
4749 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4750 void *netdev_lower_dev_get_private(struct net_device *dev,
4751 struct net_device *lower_dev);
4752 void netdev_lower_state_changed(struct net_device *lower_dev,
4753 void *lower_state_info);
4755 /* RSS keys are 40 or 52 bytes long */
4756 #define NETDEV_RSS_KEY_LEN 52
4757 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4758 void netdev_rss_key_fill(void *buffer, size_t len);
4760 int skb_checksum_help(struct sk_buff *skb);
4761 int skb_crc32c_csum_help(struct sk_buff *skb);
4762 int skb_csum_hwoffload_help(struct sk_buff *skb,
4763 const netdev_features_t features);
4765 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4766 netdev_features_t features, bool tx_path);
4767 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4768 netdev_features_t features, __be16 type);
4769 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4770 netdev_features_t features);
4772 struct netdev_bonding_info {
4777 struct netdev_notifier_bonding_info {
4778 struct netdev_notifier_info info; /* must be first */
4779 struct netdev_bonding_info bonding_info;
4782 void netdev_bonding_info_change(struct net_device *dev,
4783 struct netdev_bonding_info *bonding_info);
4785 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4786 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4788 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4795 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4797 return __skb_gso_segment(skb, features, true);
4799 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4801 static inline bool can_checksum_protocol(netdev_features_t features,
4804 if (protocol == htons(ETH_P_FCOE))
4805 return !!(features & NETIF_F_FCOE_CRC);
4807 /* Assume this is an IP checksum (not SCTP CRC) */
4809 if (features & NETIF_F_HW_CSUM) {
4810 /* Can checksum everything */
4815 case htons(ETH_P_IP):
4816 return !!(features & NETIF_F_IP_CSUM);
4817 case htons(ETH_P_IPV6):
4818 return !!(features & NETIF_F_IPV6_CSUM);
4825 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4827 static inline void netdev_rx_csum_fault(struct net_device *dev,
4828 struct sk_buff *skb)
4832 /* rx skb timestamps */
4833 void net_enable_timestamp(void);
4834 void net_disable_timestamp(void);
4836 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4837 const struct skb_shared_hwtstamps *hwtstamps,
4840 const struct net_device_ops *ops = dev->netdev_ops;
4842 if (ops->ndo_get_tstamp)
4843 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4845 return hwtstamps->hwtstamp;
4848 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4849 struct sk_buff *skb, struct net_device *dev,
4852 __this_cpu_write(softnet_data.xmit.more, more);
4853 return ops->ndo_start_xmit(skb, dev);
4856 static inline bool netdev_xmit_more(void)
4858 return __this_cpu_read(softnet_data.xmit.more);
4861 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4862 struct netdev_queue *txq, bool more)
4864 const struct net_device_ops *ops = dev->netdev_ops;
4867 rc = __netdev_start_xmit(ops, skb, dev, more);
4868 if (rc == NETDEV_TX_OK)
4869 txq_trans_update(txq);
4874 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4876 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4879 extern const struct kobj_ns_type_operations net_ns_type_operations;
4881 const char *netdev_drivername(const struct net_device *dev);
4883 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4884 netdev_features_t f2)
4886 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4887 if (f1 & NETIF_F_HW_CSUM)
4888 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4890 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4896 static inline netdev_features_t netdev_get_wanted_features(
4897 struct net_device *dev)
4899 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4901 netdev_features_t netdev_increment_features(netdev_features_t all,
4902 netdev_features_t one, netdev_features_t mask);
4904 /* Allow TSO being used on stacked device :
4905 * Performing the GSO segmentation before last device
4906 * is a performance improvement.
4908 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4909 netdev_features_t mask)
4911 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4914 int __netdev_update_features(struct net_device *dev);
4915 void netdev_update_features(struct net_device *dev);
4916 void netdev_change_features(struct net_device *dev);
4918 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4919 struct net_device *dev);
4921 netdev_features_t passthru_features_check(struct sk_buff *skb,
4922 struct net_device *dev,
4923 netdev_features_t features);
4924 netdev_features_t netif_skb_features(struct sk_buff *skb);
4926 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4928 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4930 /* check flags correspondence */
4931 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4932 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4933 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4934 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4935 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4936 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4937 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4938 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4939 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4940 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4941 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4942 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4943 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4944 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4945 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4946 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4947 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4948 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4949 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4951 return (features & feature) == feature;
4954 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4956 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4957 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4960 static inline bool netif_needs_gso(struct sk_buff *skb,
4961 netdev_features_t features)
4963 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4964 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4965 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4968 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4969 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
4970 void netif_inherit_tso_max(struct net_device *to,
4971 const struct net_device *from);
4973 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4974 int pulled_hlen, u16 mac_offset,
4977 skb->protocol = protocol;
4978 skb->encapsulation = 1;
4979 skb_push(skb, pulled_hlen);
4980 skb_reset_transport_header(skb);
4981 skb->mac_header = mac_offset;
4982 skb->network_header = skb->mac_header + mac_len;
4983 skb->mac_len = mac_len;
4986 static inline bool netif_is_macsec(const struct net_device *dev)
4988 return dev->priv_flags & IFF_MACSEC;
4991 static inline bool netif_is_macvlan(const struct net_device *dev)
4993 return dev->priv_flags & IFF_MACVLAN;
4996 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4998 return dev->priv_flags & IFF_MACVLAN_PORT;
5001 static inline bool netif_is_bond_master(const struct net_device *dev)
5003 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5006 static inline bool netif_is_bond_slave(const struct net_device *dev)
5008 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5011 static inline bool netif_supports_nofcs(struct net_device *dev)
5013 return dev->priv_flags & IFF_SUPP_NOFCS;
5016 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5018 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5021 static inline bool netif_is_l3_master(const struct net_device *dev)
5023 return dev->priv_flags & IFF_L3MDEV_MASTER;
5026 static inline bool netif_is_l3_slave(const struct net_device *dev)
5028 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5031 static inline bool netif_is_bridge_master(const struct net_device *dev)
5033 return dev->priv_flags & IFF_EBRIDGE;
5036 static inline bool netif_is_bridge_port(const struct net_device *dev)
5038 return dev->priv_flags & IFF_BRIDGE_PORT;
5041 static inline bool netif_is_ovs_master(const struct net_device *dev)
5043 return dev->priv_flags & IFF_OPENVSWITCH;
5046 static inline bool netif_is_ovs_port(const struct net_device *dev)
5048 return dev->priv_flags & IFF_OVS_DATAPATH;
5051 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5053 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5056 static inline bool netif_is_team_master(const struct net_device *dev)
5058 return dev->priv_flags & IFF_TEAM;
5061 static inline bool netif_is_team_port(const struct net_device *dev)
5063 return dev->priv_flags & IFF_TEAM_PORT;
5066 static inline bool netif_is_lag_master(const struct net_device *dev)
5068 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5071 static inline bool netif_is_lag_port(const struct net_device *dev)
5073 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5076 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5078 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5081 static inline bool netif_is_failover(const struct net_device *dev)
5083 return dev->priv_flags & IFF_FAILOVER;
5086 static inline bool netif_is_failover_slave(const struct net_device *dev)
5088 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5091 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5092 static inline void netif_keep_dst(struct net_device *dev)
5094 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5097 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5098 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5100 /* TODO: reserve and use an additional IFF bit, if we get more users */
5101 return netif_is_macsec(dev);
5104 extern struct pernet_operations __net_initdata loopback_net_ops;
5106 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5108 /* netdev_printk helpers, similar to dev_printk */
5110 static inline const char *netdev_name(const struct net_device *dev)
5112 if (!dev->name[0] || strchr(dev->name, '%'))
5113 return "(unnamed net_device)";
5117 static inline bool netdev_unregistering(const struct net_device *dev)
5119 return dev->reg_state == NETREG_UNREGISTERING;
5122 static inline const char *netdev_reg_state(const struct net_device *dev)
5124 switch (dev->reg_state) {
5125 case NETREG_UNINITIALIZED: return " (uninitialized)";
5126 case NETREG_REGISTERED: return "";
5127 case NETREG_UNREGISTERING: return " (unregistering)";
5128 case NETREG_UNREGISTERED: return " (unregistered)";
5129 case NETREG_RELEASED: return " (released)";
5130 case NETREG_DUMMY: return " (dummy)";
5133 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5134 return " (unknown)";
5137 #define MODULE_ALIAS_NETDEV(device) \
5138 MODULE_ALIAS("netdev-" device)
5141 * netdev_WARN() acts like dev_printk(), but with the key difference
5142 * of using a WARN/WARN_ON to get the message out, including the
5143 * file/line information and a backtrace.
5145 #define netdev_WARN(dev, format, args...) \
5146 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5147 netdev_reg_state(dev), ##args)
5149 #define netdev_WARN_ONCE(dev, format, args...) \
5150 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5151 netdev_reg_state(dev), ##args)
5154 * The list of packet types we will receive (as opposed to discard)
5155 * and the routines to invoke.
5157 * Why 16. Because with 16 the only overlap we get on a hash of the
5158 * low nibble of the protocol value is RARP/SNAP/X.25.
5172 #define PTYPE_HASH_SIZE (16)
5173 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5175 extern struct list_head ptype_all __read_mostly;
5176 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5178 extern struct net_device *blackhole_netdev;
5180 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5181 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5182 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5183 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5185 #endif /* _LINUX_NETDEVICE_H */