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>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <net/net_namespace.h>
39 #include <net/dcbnl.h>
41 #include <net/netprio_cgroup.h>
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <linux/hashtable.h>
50 #include <linux/rbtree.h>
57 struct ip_tunnel_parm;
58 struct macsec_context;
64 /* 802.15.4 specific */
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
69 struct udp_tunnel_nic_info;
70 struct udp_tunnel_nic;
74 void synchronize_net(void);
75 void netdev_set_default_ethtool_ops(struct net_device *dev,
76 const struct ethtool_ops *ops);
78 /* Backlog congestion levels */
79 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
80 #define NET_RX_DROP 1 /* packet dropped */
82 #define MAX_NEST_DEV 8
85 * Transmit return codes: transmit return codes originate from three different
88 * - qdisc return codes
89 * - driver transmit return codes
92 * Drivers are allowed to return any one of those in their hard_start_xmit()
93 * function. Real network devices commonly used with qdiscs should only return
94 * the driver transmit return codes though - when qdiscs are used, the actual
95 * transmission happens asynchronously, so the value is not propagated to
96 * higher layers. Virtual network devices transmit synchronously; in this case
97 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
98 * others are propagated to higher layers.
101 /* qdisc ->enqueue() return codes. */
102 #define NET_XMIT_SUCCESS 0x00
103 #define NET_XMIT_DROP 0x01 /* skb dropped */
104 #define NET_XMIT_CN 0x02 /* congestion notification */
105 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
107 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
108 * indicates that the device will soon be dropping packets, or already drops
109 * some packets of the same priority; prompting us to send less aggressively. */
110 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
111 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
113 /* Driver transmit return codes */
114 #define NETDEV_TX_MASK 0xf0
117 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
118 NETDEV_TX_OK = 0x00, /* driver took care of packet */
119 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
121 typedef enum netdev_tx netdev_tx_t;
124 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
125 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
127 static inline bool dev_xmit_complete(int rc)
130 * Positive cases with an skb consumed by a driver:
131 * - successful transmission (rc == NETDEV_TX_OK)
132 * - error while transmitting (rc < 0)
133 * - error while queueing to a different device (rc & NET_XMIT_MASK)
135 if (likely(rc < NET_XMIT_MASK))
142 * Compute the worst-case header length according to the protocols
146 #if defined(CONFIG_HYPERV_NET)
147 # define LL_MAX_HEADER 128
148 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
149 # if defined(CONFIG_MAC80211_MESH)
150 # define LL_MAX_HEADER 128
152 # define LL_MAX_HEADER 96
155 # define LL_MAX_HEADER 32
158 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
159 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
160 #define MAX_HEADER LL_MAX_HEADER
162 #define MAX_HEADER (LL_MAX_HEADER + 48)
166 * Old network device statistics. Fields are native words
167 * (unsigned long) so they can be read and written atomically.
170 struct net_device_stats {
171 unsigned long rx_packets;
172 unsigned long tx_packets;
173 unsigned long rx_bytes;
174 unsigned long tx_bytes;
175 unsigned long rx_errors;
176 unsigned long tx_errors;
177 unsigned long rx_dropped;
178 unsigned long tx_dropped;
179 unsigned long multicast;
180 unsigned long collisions;
181 unsigned long rx_length_errors;
182 unsigned long rx_over_errors;
183 unsigned long rx_crc_errors;
184 unsigned long rx_frame_errors;
185 unsigned long rx_fifo_errors;
186 unsigned long rx_missed_errors;
187 unsigned long tx_aborted_errors;
188 unsigned long tx_carrier_errors;
189 unsigned long tx_fifo_errors;
190 unsigned long tx_heartbeat_errors;
191 unsigned long tx_window_errors;
192 unsigned long rx_compressed;
193 unsigned long tx_compressed;
197 #include <linux/cache.h>
198 #include <linux/skbuff.h>
201 #include <linux/static_key.h>
202 extern struct static_key_false rps_needed;
203 extern struct static_key_false rfs_needed;
210 struct netdev_hw_addr {
211 struct list_head list;
213 unsigned char addr[MAX_ADDR_LEN];
215 #define NETDEV_HW_ADDR_T_LAN 1
216 #define NETDEV_HW_ADDR_T_SAN 2
217 #define NETDEV_HW_ADDR_T_UNICAST 3
218 #define NETDEV_HW_ADDR_T_MULTICAST 4
223 struct rcu_head rcu_head;
226 struct netdev_hw_addr_list {
227 struct list_head list;
230 /* Auxiliary tree for faster lookup on addition and deletion */
234 #define netdev_hw_addr_list_count(l) ((l)->count)
235 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
236 #define netdev_hw_addr_list_for_each(ha, l) \
237 list_for_each_entry(ha, &(l)->list, list)
239 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
240 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
241 #define netdev_for_each_uc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
244 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
245 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
246 #define netdev_for_each_mc_addr(ha, dev) \
247 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
253 /* cached hardware header; allow for machine alignment needs. */
254 #define HH_DATA_MOD 16
255 #define HH_DATA_OFF(__len) \
256 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
257 #define HH_DATA_ALIGN(__len) \
258 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
259 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
262 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
264 * dev->hard_header_len ? (dev->hard_header_len +
265 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
267 * We could use other alignment values, but we must maintain the
268 * relationship HH alignment <= LL alignment.
270 #define LL_RESERVED_SPACE(dev) \
271 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
272 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
273 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
276 int (*create) (struct sk_buff *skb, struct net_device *dev,
277 unsigned short type, const void *daddr,
278 const void *saddr, unsigned int len);
279 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
280 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
281 void (*cache_update)(struct hh_cache *hh,
282 const struct net_device *dev,
283 const unsigned char *haddr);
284 bool (*validate)(const char *ll_header, unsigned int len);
285 __be16 (*parse_protocol)(const struct sk_buff *skb);
288 /* These flag bits are private to the generic network queueing
289 * layer; they may not be explicitly referenced by any other
293 enum netdev_state_t {
295 __LINK_STATE_PRESENT,
296 __LINK_STATE_NOCARRIER,
297 __LINK_STATE_LINKWATCH_PENDING,
298 __LINK_STATE_DORMANT,
299 __LINK_STATE_TESTING,
304 struct list_head list;
309 * size of gro hash buckets, must less than bit number of
310 * napi_struct::gro_bitmask
312 #define GRO_HASH_BUCKETS 8
315 * Structure for NAPI scheduling similar to tasklet but with weighting
318 /* The poll_list must only be managed by the entity which
319 * changes the state of the NAPI_STATE_SCHED bit. This means
320 * whoever atomically sets that bit can add this napi_struct
321 * to the per-CPU poll_list, and whoever clears that bit
322 * can remove from the list right before clearing the bit.
324 struct list_head poll_list;
328 int defer_hard_irqs_count;
329 unsigned long gro_bitmask;
330 int (*poll)(struct napi_struct *, int);
331 #ifdef CONFIG_NETPOLL
334 struct net_device *dev;
335 struct gro_list gro_hash[GRO_HASH_BUCKETS];
337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
338 int rx_count; /* length of rx_list */
339 struct hrtimer timer;
340 struct list_head dev_list;
341 struct hlist_node napi_hash_node;
342 unsigned int napi_id;
343 struct task_struct *thread;
347 NAPI_STATE_SCHED, /* Poll is scheduled */
348 NAPI_STATE_MISSED, /* reschedule a napi */
349 NAPI_STATE_DISABLE, /* Disable pending */
350 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
351 NAPI_STATE_LISTED, /* NAPI added to system lists */
352 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
353 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
354 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
355 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
356 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
364 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
367 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
368 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
369 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
379 typedef enum gro_result gro_result_t;
382 * enum rx_handler_result - Possible return values for rx_handlers.
383 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
385 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
386 * case skb->dev was changed by rx_handler.
387 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
388 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
390 * rx_handlers are functions called from inside __netif_receive_skb(), to do
391 * special processing of the skb, prior to delivery to protocol handlers.
393 * Currently, a net_device can only have a single rx_handler registered. Trying
394 * to register a second rx_handler will return -EBUSY.
396 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
397 * To unregister a rx_handler on a net_device, use
398 * netdev_rx_handler_unregister().
400 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
403 * If the rx_handler consumed the skb in some way, it should return
404 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
405 * the skb to be delivered in some other way.
407 * If the rx_handler changed skb->dev, to divert the skb to another
408 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
409 * new device will be called if it exists.
411 * If the rx_handler decides the skb should be ignored, it should return
412 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
413 * are registered on exact device (ptype->dev == skb->dev).
415 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
416 * delivered, it should return RX_HANDLER_PASS.
418 * A device without a registered rx_handler will behave as if rx_handler
419 * returned RX_HANDLER_PASS.
422 enum rx_handler_result {
428 typedef enum rx_handler_result rx_handler_result_t;
429 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
431 void __napi_schedule(struct napi_struct *n);
432 void __napi_schedule_irqoff(struct napi_struct *n);
434 static inline bool napi_disable_pending(struct napi_struct *n)
436 return test_bit(NAPI_STATE_DISABLE, &n->state);
439 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
441 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
444 bool napi_schedule_prep(struct napi_struct *n);
447 * napi_schedule - schedule NAPI poll
450 * Schedule NAPI poll routine to be called if it is not already
453 static inline void napi_schedule(struct napi_struct *n)
455 if (napi_schedule_prep(n))
460 * napi_schedule_irqoff - schedule NAPI poll
463 * Variant of napi_schedule(), assuming hard irqs are masked.
465 static inline void napi_schedule_irqoff(struct napi_struct *n)
467 if (napi_schedule_prep(n))
468 __napi_schedule_irqoff(n);
471 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
472 static inline bool napi_reschedule(struct napi_struct *napi)
474 if (napi_schedule_prep(napi)) {
475 __napi_schedule(napi);
481 bool napi_complete_done(struct napi_struct *n, int work_done);
483 * napi_complete - NAPI processing complete
486 * Mark NAPI processing as complete.
487 * Consider using napi_complete_done() instead.
488 * Return false if device should avoid rearming interrupts.
490 static inline bool napi_complete(struct napi_struct *n)
492 return napi_complete_done(n, 0);
495 int dev_set_threaded(struct net_device *dev, bool threaded);
498 * napi_disable - prevent NAPI from scheduling
501 * Stop NAPI from being scheduled on this context.
502 * Waits till any outstanding processing completes.
504 void napi_disable(struct napi_struct *n);
506 void napi_enable(struct napi_struct *n);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct *n)
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
526 * napi_if_scheduled_mark_missed - if napi is running, set the
530 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
533 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
535 unsigned long val, new;
538 val = READ_ONCE(n->state);
539 if (val & NAPIF_STATE_DISABLE)
542 if (!(val & NAPIF_STATE_SCHED))
545 new = val | NAPIF_STATE_MISSED;
546 } while (cmpxchg(&n->state, val, new) != val);
551 enum netdev_queue_state_t {
552 __QUEUE_STATE_DRV_XOFF,
553 __QUEUE_STATE_STACK_XOFF,
554 __QUEUE_STATE_FROZEN,
557 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
558 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
559 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
561 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
562 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
564 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
568 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
569 * netif_tx_* functions below are used to manipulate this flag. The
570 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
571 * queue independently. The netif_xmit_*stopped functions below are called
572 * to check if the queue has been stopped by the driver or stack (either
573 * of the XOFF bits are set in the state). Drivers should not need to call
574 * netif_xmit*stopped functions, they should only be using netif_tx_*.
577 struct netdev_queue {
581 struct net_device *dev;
582 struct Qdisc __rcu *qdisc;
583 struct Qdisc *qdisc_sleeping;
587 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 unsigned long tx_maxrate;
592 * Number of TX timeouts for this queue
593 * (/sys/class/net/DEV/Q/trans_timeout)
595 unsigned long trans_timeout;
597 /* Subordinate device that the queue has been assigned to */
598 struct net_device *sb_dev;
599 #ifdef CONFIG_XDP_SOCKETS
600 struct xsk_buff_pool *pool;
605 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
608 * Time (in jiffies) of last Tx
610 unsigned long trans_start;
617 } ____cacheline_aligned_in_smp;
619 extern int sysctl_fb_tunnels_only_for_init_net;
620 extern int sysctl_devconf_inherit_init_net;
623 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
624 * == 1 : For initns only
627 static inline bool net_has_fallback_tunnels(const struct net *net)
629 #if IS_ENABLED(CONFIG_SYSCTL)
630 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
632 return !fb_tunnels_only_for_init_net ||
633 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
639 static inline int net_inherit_devconf(void)
641 #if IS_ENABLED(CONFIG_SYSCTL)
642 return READ_ONCE(sysctl_devconf_inherit_init_net);
648 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
659 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
666 * This structure holds an RPS map which can be of variable length. The
667 * map is an array of CPUs.
674 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
677 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
678 * tail pointer for that CPU's input queue at the time of last enqueue, and
679 * a hardware filter index.
681 struct rps_dev_flow {
684 unsigned int last_qtail;
686 #define RPS_NO_FILTER 0xffff
689 * The rps_dev_flow_table structure contains a table of flow mappings.
691 struct rps_dev_flow_table {
694 struct rps_dev_flow flows[];
696 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
697 ((_num) * sizeof(struct rps_dev_flow)))
700 * The rps_sock_flow_table contains mappings of flows to the last CPU
701 * on which they were processed by the application (set in recvmsg).
702 * Each entry is a 32bit value. Upper part is the high-order bits
703 * of flow hash, lower part is CPU number.
704 * rps_cpu_mask is used to partition the space, depending on number of
705 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
706 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
707 * meaning we use 32-6=26 bits for the hash.
709 struct rps_sock_flow_table {
712 u32 ents[] ____cacheline_aligned_in_smp;
714 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
716 #define RPS_NO_CPU 0xffff
718 extern u32 rps_cpu_mask;
719 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
721 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
725 unsigned int index = hash & table->mask;
726 u32 val = hash & ~rps_cpu_mask;
728 /* We only give a hint, preemption can change CPU under us */
729 val |= raw_smp_processor_id();
731 if (table->ents[index] != val)
732 table->ents[index] = val;
736 #ifdef CONFIG_RFS_ACCEL
737 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
740 #endif /* CONFIG_RPS */
742 /* This structure contains an instance of an RX queue. */
743 struct netdev_rx_queue {
744 struct xdp_rxq_info xdp_rxq;
746 struct rps_map __rcu *rps_map;
747 struct rps_dev_flow_table __rcu *rps_flow_table;
750 struct net_device *dev;
751 #ifdef CONFIG_XDP_SOCKETS
752 struct xsk_buff_pool *pool;
754 } ____cacheline_aligned_in_smp;
757 * RX queue sysfs structures and functions.
759 struct rx_queue_attribute {
760 struct attribute attr;
761 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
762 ssize_t (*store)(struct netdev_rx_queue *queue,
763 const char *buf, size_t len);
766 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
775 * This structure holds an XPS map which can be of variable length. The
776 * map is an array of queues.
780 unsigned int alloc_len;
784 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
785 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
786 - sizeof(struct xps_map)) / sizeof(u16))
789 * This structure holds all XPS maps for device. Maps are indexed by CPU.
791 * We keep track of the number of cpus/rxqs used when the struct is allocated,
792 * in nr_ids. This will help not accessing out-of-bound memory.
794 * We keep track of the number of traffic classes used when the struct is
795 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
796 * not crossing its upper bound, as the original dev->num_tc can be updated in
799 struct xps_dev_maps {
803 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
806 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
807 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
809 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
810 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
812 #endif /* CONFIG_XPS */
814 #define TC_MAX_QUEUE 16
815 #define TC_BITMASK 15
816 /* HW offloaded queuing disciplines txq count and offset maps */
817 struct netdev_tc_txq {
822 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
824 * This structure is to hold information about the device
825 * configured to run FCoE protocol stack.
827 struct netdev_fcoe_hbainfo {
828 char manufacturer[64];
829 char serial_number[64];
830 char hardware_version[64];
831 char driver_version[64];
832 char optionrom_version[64];
833 char firmware_version[64];
835 char model_description[256];
839 #define MAX_PHYS_ITEM_ID_LEN 32
841 /* This structure holds a unique identifier to identify some
842 * physical item (port for example) used by a netdevice.
844 struct netdev_phys_item_id {
845 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
846 unsigned char id_len;
849 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
850 struct netdev_phys_item_id *b)
852 return a->id_len == b->id_len &&
853 memcmp(a->id, b->id, a->id_len) == 0;
856 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
858 struct net_device *sb_dev);
860 enum net_device_path_type {
861 DEV_PATH_ETHERNET = 0,
868 struct net_device_path {
869 enum net_device_path_type type;
870 const struct net_device *dev;
879 DEV_PATH_BR_VLAN_KEEP,
880 DEV_PATH_BR_VLAN_TAG,
881 DEV_PATH_BR_VLAN_UNTAG,
882 DEV_PATH_BR_VLAN_UNTAG_HW,
894 #define NET_DEVICE_PATH_STACK_MAX 5
895 #define NET_DEVICE_PATH_VLAN_MAX 2
897 struct net_device_path_stack {
899 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
902 struct net_device_path_ctx {
903 const struct net_device *dev;
910 } vlan[NET_DEVICE_PATH_VLAN_MAX];
914 TC_SETUP_QDISC_MQPRIO,
917 TC_SETUP_CLSMATCHALL,
927 TC_SETUP_QDISC_TAPRIO,
935 /* These structures hold the attributes of bpf state that are being passed
936 * to the netdevice through the bpf op.
938 enum bpf_netdev_command {
939 /* Set or clear a bpf program used in the earliest stages of packet
940 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
941 * is responsible for calling bpf_prog_put on any old progs that are
942 * stored. In case of error, the callee need not release the new prog
943 * reference, but on success it takes ownership and must bpf_prog_put
944 * when it is no longer used.
948 /* BPF program for offload callbacks, invoked at program load time. */
949 BPF_OFFLOAD_MAP_ALLOC,
950 BPF_OFFLOAD_MAP_FREE,
954 struct bpf_prog_offload_ops;
955 struct netlink_ext_ack;
957 struct xdp_dev_bulk_queue;
967 struct bpf_xdp_entity {
968 struct bpf_prog *prog;
969 struct bpf_xdp_link *link;
973 enum bpf_netdev_command command;
978 struct bpf_prog *prog;
979 struct netlink_ext_ack *extack;
981 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
983 struct bpf_offloaded_map *offmap;
985 /* XDP_SETUP_XSK_POOL */
987 struct xsk_buff_pool *pool;
993 /* Flags for ndo_xsk_wakeup. */
994 #define XDP_WAKEUP_RX (1 << 0)
995 #define XDP_WAKEUP_TX (1 << 1)
997 #ifdef CONFIG_XFRM_OFFLOAD
999 int (*xdo_dev_state_add) (struct xfrm_state *x);
1000 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1001 void (*xdo_dev_state_free) (struct xfrm_state *x);
1002 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1003 struct xfrm_state *x);
1004 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1008 struct dev_ifalias {
1009 struct rcu_head rcuhead;
1016 struct netdev_name_node {
1017 struct hlist_node hlist;
1018 struct list_head list;
1019 struct net_device *dev;
1023 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1024 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1026 struct netdev_net_notifier {
1027 struct list_head list;
1028 struct notifier_block *nb;
1032 * This structure defines the management hooks for network devices.
1033 * The following hooks can be defined; unless noted otherwise, they are
1034 * optional and can be filled with a null pointer.
1036 * int (*ndo_init)(struct net_device *dev);
1037 * This function is called once when a network device is registered.
1038 * The network device can use this for any late stage initialization
1039 * or semantic validation. It can fail with an error code which will
1040 * be propagated back to register_netdev.
1042 * void (*ndo_uninit)(struct net_device *dev);
1043 * This function is called when device is unregistered or when registration
1044 * fails. It is not called if init fails.
1046 * int (*ndo_open)(struct net_device *dev);
1047 * This function is called when a network device transitions to the up
1050 * int (*ndo_stop)(struct net_device *dev);
1051 * This function is called when a network device transitions to the down
1054 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1055 * struct net_device *dev);
1056 * Called when a packet needs to be transmitted.
1057 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1058 * the queue before that can happen; it's for obsolete devices and weird
1059 * corner cases, but the stack really does a non-trivial amount
1060 * of useless work if you return NETDEV_TX_BUSY.
1061 * Required; cannot be NULL.
1063 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1064 * struct net_device *dev
1065 * netdev_features_t features);
1066 * Called by core transmit path to determine if device is capable of
1067 * performing offload operations on a given packet. This is to give
1068 * the device an opportunity to implement any restrictions that cannot
1069 * be otherwise expressed by feature flags. The check is called with
1070 * the set of features that the stack has calculated and it returns
1071 * those the driver believes to be appropriate.
1073 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1074 * struct net_device *sb_dev);
1075 * Called to decide which queue to use when device supports multiple
1078 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1079 * This function is called to allow device receiver to make
1080 * changes to configuration when multicast or promiscuous is enabled.
1082 * void (*ndo_set_rx_mode)(struct net_device *dev);
1083 * This function is called device changes address list filtering.
1084 * If driver handles unicast address filtering, it should set
1085 * IFF_UNICAST_FLT in its priv_flags.
1087 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1088 * This function is called when the Media Access Control address
1089 * needs to be changed. If this interface is not defined, the
1090 * MAC address can not be changed.
1092 * int (*ndo_validate_addr)(struct net_device *dev);
1093 * Test if Media Access Control address is valid for the device.
1095 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1096 * Old-style ioctl entry point. This is used internally by the
1097 * appletalk and ieee802154 subsystems but is no longer called by
1098 * the device ioctl handler.
1100 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1101 * Used by the bonding driver for its device specific ioctls:
1102 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1103 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1105 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1106 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1107 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1109 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1110 * Used to set network devices bus interface parameters. This interface
1111 * is retained for legacy reasons; new devices should use the bus
1112 * interface (PCI) for low level management.
1114 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1115 * Called when a user wants to change the Maximum Transfer Unit
1118 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1119 * Callback used when the transmitter has not made any progress
1120 * for dev->watchdog ticks.
1122 * void (*ndo_get_stats64)(struct net_device *dev,
1123 * struct rtnl_link_stats64 *storage);
1124 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1125 * Called when a user wants to get the network device usage
1126 * statistics. Drivers must do one of the following:
1127 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1128 * rtnl_link_stats64 structure passed by the caller.
1129 * 2. Define @ndo_get_stats to update a net_device_stats structure
1130 * (which should normally be dev->stats) and return a pointer to
1131 * it. The structure may be changed asynchronously only if each
1132 * field is written atomically.
1133 * 3. Update dev->stats asynchronously and atomically, and define
1134 * neither operation.
1136 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1137 * Return true if this device supports offload stats of this attr_id.
1139 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1141 * Get statistics for offload operations by attr_id. Write it into the
1142 * attr_data pointer.
1144 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1145 * If device supports VLAN filtering this function is called when a
1146 * VLAN id is registered.
1148 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1149 * If device supports VLAN filtering this function is called when a
1150 * VLAN id is unregistered.
1152 * void (*ndo_poll_controller)(struct net_device *dev);
1154 * SR-IOV management functions.
1155 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1156 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1157 * u8 qos, __be16 proto);
1158 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1160 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1161 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1162 * int (*ndo_get_vf_config)(struct net_device *dev,
1163 * int vf, struct ifla_vf_info *ivf);
1164 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1165 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1166 * struct nlattr *port[]);
1168 * Enable or disable the VF ability to query its RSS Redirection Table and
1169 * Hash Key. This is needed since on some devices VF share this information
1170 * with PF and querying it may introduce a theoretical security risk.
1171 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1172 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1173 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1175 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1176 * This is always called from the stack with the rtnl lock held and netif
1177 * tx queues stopped. This allows the netdevice to perform queue
1178 * management safely.
1180 * Fiber Channel over Ethernet (FCoE) offload functions.
1181 * int (*ndo_fcoe_enable)(struct net_device *dev);
1182 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1183 * so the underlying device can perform whatever needed configuration or
1184 * initialization to support acceleration of FCoE traffic.
1186 * int (*ndo_fcoe_disable)(struct net_device *dev);
1187 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1188 * so the underlying device can perform whatever needed clean-ups to
1189 * stop supporting acceleration of FCoE traffic.
1191 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1192 * struct scatterlist *sgl, unsigned int sgc);
1193 * Called when the FCoE Initiator wants to initialize an I/O that
1194 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1195 * perform necessary setup and returns 1 to indicate the device is set up
1196 * successfully to perform DDP on this I/O, otherwise this returns 0.
1198 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1199 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1200 * indicated by the FC exchange id 'xid', so the underlying device can
1201 * clean up and reuse resources for later DDP requests.
1203 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1204 * struct scatterlist *sgl, unsigned int sgc);
1205 * Called when the FCoE Target wants to initialize an I/O that
1206 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1207 * perform necessary setup and returns 1 to indicate the device is set up
1208 * successfully to perform DDP on this I/O, otherwise this returns 0.
1210 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1211 * struct netdev_fcoe_hbainfo *hbainfo);
1212 * Called when the FCoE Protocol stack wants information on the underlying
1213 * device. This information is utilized by the FCoE protocol stack to
1214 * register attributes with Fiber Channel management service as per the
1215 * FC-GS Fabric Device Management Information(FDMI) specification.
1217 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1218 * Called when the underlying device wants to override default World Wide
1219 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1220 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1221 * protocol stack to use.
1224 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1225 * u16 rxq_index, u32 flow_id);
1226 * Set hardware filter for RFS. rxq_index is the target queue index;
1227 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1228 * Return the filter ID on success, or a negative error code.
1230 * Slave management functions (for bridge, bonding, etc).
1231 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1232 * Called to make another netdev an underling.
1234 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1235 * Called to release previously enslaved netdev.
1237 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1238 * struct sk_buff *skb,
1240 * Get the xmit slave of master device. If all_slaves is true, function
1241 * assume all the slaves can transmit.
1243 * Feature/offload setting functions.
1244 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1245 * netdev_features_t features);
1246 * Adjusts the requested feature flags according to device-specific
1247 * constraints, and returns the resulting flags. Must not modify
1250 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1251 * Called to update device configuration to new features. Passed
1252 * feature set might be less than what was returned by ndo_fix_features()).
1253 * Must return >0 or -errno if it changed dev->features itself.
1255 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1256 * struct net_device *dev,
1257 * const unsigned char *addr, u16 vid, u16 flags,
1258 * struct netlink_ext_ack *extack);
1259 * Adds an FDB entry to dev for addr.
1260 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1261 * struct net_device *dev,
1262 * const unsigned char *addr, u16 vid)
1263 * Deletes the FDB entry from dev coresponding to addr.
1264 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1265 * struct net_device *dev, struct net_device *filter_dev,
1267 * Used to add FDB entries to dump requests. Implementers should add
1268 * entries to skb and update idx with the number of entries.
1270 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1271 * u16 flags, struct netlink_ext_ack *extack)
1272 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1273 * struct net_device *dev, u32 filter_mask,
1275 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1278 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1279 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1280 * which do not represent real hardware may define this to allow their
1281 * userspace components to manage their virtual carrier state. Devices
1282 * that determine carrier state from physical hardware properties (eg
1283 * network cables) or protocol-dependent mechanisms (eg
1284 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1286 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1287 * struct netdev_phys_item_id *ppid);
1288 * Called to get ID of physical port of this device. If driver does
1289 * not implement this, it is assumed that the hw is not able to have
1290 * multiple net devices on single physical port.
1292 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1293 * struct netdev_phys_item_id *ppid)
1294 * Called to get the parent ID of the physical port of this device.
1296 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1297 * struct net_device *dev)
1298 * Called by upper layer devices to accelerate switching or other
1299 * station functionality into hardware. 'pdev is the lowerdev
1300 * to use for the offload and 'dev' is the net device that will
1301 * back the offload. Returns a pointer to the private structure
1302 * the upper layer will maintain.
1303 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1304 * Called by upper layer device to delete the station created
1305 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1306 * the station and priv is the structure returned by the add
1308 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1309 * int queue_index, u32 maxrate);
1310 * Called when a user wants to set a max-rate limitation of specific
1312 * int (*ndo_get_iflink)(const struct net_device *dev);
1313 * Called to get the iflink value of this device.
1314 * void (*ndo_change_proto_down)(struct net_device *dev,
1316 * This function is used to pass protocol port error state information
1317 * to the switch driver. The switch driver can react to the proto_down
1318 * by doing a phys down on the associated switch port.
1319 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1320 * This function is used to get egress tunnel information for given skb.
1321 * This is useful for retrieving outer tunnel header parameters while
1323 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1324 * This function is used to specify the headroom that the skb must
1325 * consider when allocation skb during packet reception. Setting
1326 * appropriate rx headroom value allows avoiding skb head copy on
1327 * forward. Setting a negative value resets the rx headroom to the
1329 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1330 * This function is used to set or query state related to XDP on the
1331 * netdevice and manage BPF offload. See definition of
1332 * enum bpf_netdev_command for details.
1333 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1335 * This function is used to submit @n XDP packets for transmit on a
1336 * netdevice. Returns number of frames successfully transmitted, frames
1337 * that got dropped are freed/returned via xdp_return_frame().
1338 * Returns negative number, means general error invoking ndo, meaning
1339 * no frames were xmit'ed and core-caller will free all frames.
1340 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1341 * struct xdp_buff *xdp);
1342 * Get the xmit slave of master device based on the xdp_buff.
1343 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1344 * This function is used to wake up the softirq, ksoftirqd or kthread
1345 * responsible for sending and/or receiving packets on a specific
1346 * queue id bound to an AF_XDP socket. The flags field specifies if
1347 * only RX, only Tx, or both should be woken up using the flags
1348 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1349 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1350 * Get devlink port instance associated with a given netdev.
1351 * Called with a reference on the netdevice and devlink locks only,
1352 * rtnl_lock is not held.
1353 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1355 * Add, change, delete or get information on an IPv4 tunnel.
1356 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1357 * If a device is paired with a peer device, return the peer instance.
1358 * The caller must be under RCU read context.
1359 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1360 * Get the forwarding path to reach the real device from the HW destination address
1362 struct net_device_ops {
1363 int (*ndo_init)(struct net_device *dev);
1364 void (*ndo_uninit)(struct net_device *dev);
1365 int (*ndo_open)(struct net_device *dev);
1366 int (*ndo_stop)(struct net_device *dev);
1367 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1368 struct net_device *dev);
1369 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1370 struct net_device *dev,
1371 netdev_features_t features);
1372 u16 (*ndo_select_queue)(struct net_device *dev,
1373 struct sk_buff *skb,
1374 struct net_device *sb_dev);
1375 void (*ndo_change_rx_flags)(struct net_device *dev,
1377 void (*ndo_set_rx_mode)(struct net_device *dev);
1378 int (*ndo_set_mac_address)(struct net_device *dev,
1380 int (*ndo_validate_addr)(struct net_device *dev);
1381 int (*ndo_do_ioctl)(struct net_device *dev,
1382 struct ifreq *ifr, int cmd);
1383 int (*ndo_eth_ioctl)(struct net_device *dev,
1384 struct ifreq *ifr, int cmd);
1385 int (*ndo_siocbond)(struct net_device *dev,
1386 struct ifreq *ifr, int cmd);
1387 int (*ndo_siocwandev)(struct net_device *dev,
1388 struct if_settings *ifs);
1389 int (*ndo_siocdevprivate)(struct net_device *dev,
1391 void __user *data, int cmd);
1392 int (*ndo_set_config)(struct net_device *dev,
1394 int (*ndo_change_mtu)(struct net_device *dev,
1396 int (*ndo_neigh_setup)(struct net_device *dev,
1397 struct neigh_parms *);
1398 void (*ndo_tx_timeout) (struct net_device *dev,
1399 unsigned int txqueue);
1401 void (*ndo_get_stats64)(struct net_device *dev,
1402 struct rtnl_link_stats64 *storage);
1403 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1404 int (*ndo_get_offload_stats)(int attr_id,
1405 const struct net_device *dev,
1407 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1409 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1410 __be16 proto, u16 vid);
1411 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1412 __be16 proto, u16 vid);
1413 #ifdef CONFIG_NET_POLL_CONTROLLER
1414 void (*ndo_poll_controller)(struct net_device *dev);
1415 int (*ndo_netpoll_setup)(struct net_device *dev,
1416 struct netpoll_info *info);
1417 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1419 int (*ndo_set_vf_mac)(struct net_device *dev,
1420 int queue, u8 *mac);
1421 int (*ndo_set_vf_vlan)(struct net_device *dev,
1422 int queue, u16 vlan,
1423 u8 qos, __be16 proto);
1424 int (*ndo_set_vf_rate)(struct net_device *dev,
1425 int vf, int min_tx_rate,
1427 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1428 int vf, bool setting);
1429 int (*ndo_set_vf_trust)(struct net_device *dev,
1430 int vf, bool setting);
1431 int (*ndo_get_vf_config)(struct net_device *dev,
1433 struct ifla_vf_info *ivf);
1434 int (*ndo_set_vf_link_state)(struct net_device *dev,
1435 int vf, int link_state);
1436 int (*ndo_get_vf_stats)(struct net_device *dev,
1438 struct ifla_vf_stats
1440 int (*ndo_set_vf_port)(struct net_device *dev,
1442 struct nlattr *port[]);
1443 int (*ndo_get_vf_port)(struct net_device *dev,
1444 int vf, struct sk_buff *skb);
1445 int (*ndo_get_vf_guid)(struct net_device *dev,
1447 struct ifla_vf_guid *node_guid,
1448 struct ifla_vf_guid *port_guid);
1449 int (*ndo_set_vf_guid)(struct net_device *dev,
1452 int (*ndo_set_vf_rss_query_en)(
1453 struct net_device *dev,
1454 int vf, bool setting);
1455 int (*ndo_setup_tc)(struct net_device *dev,
1456 enum tc_setup_type type,
1458 #if IS_ENABLED(CONFIG_FCOE)
1459 int (*ndo_fcoe_enable)(struct net_device *dev);
1460 int (*ndo_fcoe_disable)(struct net_device *dev);
1461 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1463 struct scatterlist *sgl,
1465 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1467 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1469 struct scatterlist *sgl,
1471 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1472 struct netdev_fcoe_hbainfo *hbainfo);
1475 #if IS_ENABLED(CONFIG_LIBFCOE)
1476 #define NETDEV_FCOE_WWNN 0
1477 #define NETDEV_FCOE_WWPN 1
1478 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1479 u64 *wwn, int type);
1482 #ifdef CONFIG_RFS_ACCEL
1483 int (*ndo_rx_flow_steer)(struct net_device *dev,
1484 const struct sk_buff *skb,
1488 int (*ndo_add_slave)(struct net_device *dev,
1489 struct net_device *slave_dev,
1490 struct netlink_ext_ack *extack);
1491 int (*ndo_del_slave)(struct net_device *dev,
1492 struct net_device *slave_dev);
1493 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1494 struct sk_buff *skb,
1496 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1498 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1499 netdev_features_t features);
1500 int (*ndo_set_features)(struct net_device *dev,
1501 netdev_features_t features);
1502 int (*ndo_neigh_construct)(struct net_device *dev,
1503 struct neighbour *n);
1504 void (*ndo_neigh_destroy)(struct net_device *dev,
1505 struct neighbour *n);
1507 int (*ndo_fdb_add)(struct ndmsg *ndm,
1508 struct nlattr *tb[],
1509 struct net_device *dev,
1510 const unsigned char *addr,
1513 struct netlink_ext_ack *extack);
1514 int (*ndo_fdb_del)(struct ndmsg *ndm,
1515 struct nlattr *tb[],
1516 struct net_device *dev,
1517 const unsigned char *addr,
1519 int (*ndo_fdb_dump)(struct sk_buff *skb,
1520 struct netlink_callback *cb,
1521 struct net_device *dev,
1522 struct net_device *filter_dev,
1524 int (*ndo_fdb_get)(struct sk_buff *skb,
1525 struct nlattr *tb[],
1526 struct net_device *dev,
1527 const unsigned char *addr,
1528 u16 vid, u32 portid, u32 seq,
1529 struct netlink_ext_ack *extack);
1530 int (*ndo_bridge_setlink)(struct net_device *dev,
1531 struct nlmsghdr *nlh,
1533 struct netlink_ext_ack *extack);
1534 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1536 struct net_device *dev,
1539 int (*ndo_bridge_dellink)(struct net_device *dev,
1540 struct nlmsghdr *nlh,
1542 int (*ndo_change_carrier)(struct net_device *dev,
1544 int (*ndo_get_phys_port_id)(struct net_device *dev,
1545 struct netdev_phys_item_id *ppid);
1546 int (*ndo_get_port_parent_id)(struct net_device *dev,
1547 struct netdev_phys_item_id *ppid);
1548 int (*ndo_get_phys_port_name)(struct net_device *dev,
1549 char *name, size_t len);
1550 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1551 struct net_device *dev);
1552 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1555 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1558 int (*ndo_get_iflink)(const struct net_device *dev);
1559 int (*ndo_change_proto_down)(struct net_device *dev,
1561 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1562 struct sk_buff *skb);
1563 void (*ndo_set_rx_headroom)(struct net_device *dev,
1564 int needed_headroom);
1565 int (*ndo_bpf)(struct net_device *dev,
1566 struct netdev_bpf *bpf);
1567 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1568 struct xdp_frame **xdp,
1570 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1571 struct xdp_buff *xdp);
1572 int (*ndo_xsk_wakeup)(struct net_device *dev,
1573 u32 queue_id, u32 flags);
1574 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1575 int (*ndo_tunnel_ctl)(struct net_device *dev,
1576 struct ip_tunnel_parm *p, int cmd);
1577 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1578 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1579 struct net_device_path *path);
1583 * enum netdev_priv_flags - &struct net_device priv_flags
1585 * These are the &struct net_device, they are only set internally
1586 * by drivers and used in the kernel. These flags are invisible to
1587 * userspace; this means that the order of these flags can change
1588 * during any kernel release.
1590 * You should have a pretty good reason to be extending these flags.
1592 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1593 * @IFF_EBRIDGE: Ethernet bridging device
1594 * @IFF_BONDING: bonding master or slave
1595 * @IFF_ISATAP: ISATAP interface (RFC4214)
1596 * @IFF_WAN_HDLC: WAN HDLC device
1597 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1599 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1600 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1601 * @IFF_MACVLAN_PORT: device used as macvlan port
1602 * @IFF_BRIDGE_PORT: device used as bridge port
1603 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1604 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1605 * @IFF_UNICAST_FLT: Supports unicast filtering
1606 * @IFF_TEAM_PORT: device used as team port
1607 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1608 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1609 * change when it's running
1610 * @IFF_MACVLAN: Macvlan device
1611 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1612 * underlying stacked devices
1613 * @IFF_L3MDEV_MASTER: device is an L3 master device
1614 * @IFF_NO_QUEUE: device can run without qdisc attached
1615 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1616 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1617 * @IFF_TEAM: device is a team device
1618 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1619 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1620 * entity (i.e. the master device for bridged veth)
1621 * @IFF_MACSEC: device is a MACsec device
1622 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1623 * @IFF_FAILOVER: device is a failover master device
1624 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1625 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1626 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1627 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1628 * skb_headlen(skb) == 0 (data starts from frag0)
1630 enum netdev_priv_flags {
1631 IFF_802_1Q_VLAN = 1<<0,
1635 IFF_WAN_HDLC = 1<<4,
1636 IFF_XMIT_DST_RELEASE = 1<<5,
1637 IFF_DONT_BRIDGE = 1<<6,
1638 IFF_DISABLE_NETPOLL = 1<<7,
1639 IFF_MACVLAN_PORT = 1<<8,
1640 IFF_BRIDGE_PORT = 1<<9,
1641 IFF_OVS_DATAPATH = 1<<10,
1642 IFF_TX_SKB_SHARING = 1<<11,
1643 IFF_UNICAST_FLT = 1<<12,
1644 IFF_TEAM_PORT = 1<<13,
1645 IFF_SUPP_NOFCS = 1<<14,
1646 IFF_LIVE_ADDR_CHANGE = 1<<15,
1647 IFF_MACVLAN = 1<<16,
1648 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1649 IFF_L3MDEV_MASTER = 1<<18,
1650 IFF_NO_QUEUE = 1<<19,
1651 IFF_OPENVSWITCH = 1<<20,
1652 IFF_L3MDEV_SLAVE = 1<<21,
1654 IFF_RXFH_CONFIGURED = 1<<23,
1655 IFF_PHONY_HEADROOM = 1<<24,
1657 IFF_NO_RX_HANDLER = 1<<26,
1658 IFF_FAILOVER = 1<<27,
1659 IFF_FAILOVER_SLAVE = 1<<28,
1660 IFF_L3MDEV_RX_HANDLER = 1<<29,
1661 IFF_LIVE_RENAME_OK = 1<<30,
1662 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1665 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1666 #define IFF_EBRIDGE IFF_EBRIDGE
1667 #define IFF_BONDING IFF_BONDING
1668 #define IFF_ISATAP IFF_ISATAP
1669 #define IFF_WAN_HDLC IFF_WAN_HDLC
1670 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1671 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1672 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1673 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1674 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1675 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1676 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1677 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1678 #define IFF_TEAM_PORT IFF_TEAM_PORT
1679 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1680 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1681 #define IFF_MACVLAN IFF_MACVLAN
1682 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1683 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1684 #define IFF_NO_QUEUE IFF_NO_QUEUE
1685 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1686 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1687 #define IFF_TEAM IFF_TEAM
1688 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1689 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1690 #define IFF_MACSEC IFF_MACSEC
1691 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1692 #define IFF_FAILOVER IFF_FAILOVER
1693 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1694 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1695 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1696 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1698 /* Specifies the type of the struct net_device::ml_priv pointer */
1699 enum netdev_ml_priv_type {
1705 * struct net_device - The DEVICE structure.
1707 * Actually, this whole structure is a big mistake. It mixes I/O
1708 * data with strictly "high-level" data, and it has to know about
1709 * almost every data structure used in the INET module.
1711 * @name: This is the first field of the "visible" part of this structure
1712 * (i.e. as seen by users in the "Space.c" file). It is the name
1715 * @name_node: Name hashlist node
1716 * @ifalias: SNMP alias
1717 * @mem_end: Shared memory end
1718 * @mem_start: Shared memory start
1719 * @base_addr: Device I/O address
1720 * @irq: Device IRQ number
1722 * @state: Generic network queuing layer state, see netdev_state_t
1723 * @dev_list: The global list of network devices
1724 * @napi_list: List entry used for polling NAPI devices
1725 * @unreg_list: List entry when we are unregistering the
1726 * device; see the function unregister_netdev
1727 * @close_list: List entry used when we are closing the device
1728 * @ptype_all: Device-specific packet handlers for all protocols
1729 * @ptype_specific: Device-specific, protocol-specific packet handlers
1731 * @adj_list: Directly linked devices, like slaves for bonding
1732 * @features: Currently active device features
1733 * @hw_features: User-changeable features
1735 * @wanted_features: User-requested features
1736 * @vlan_features: Mask of features inheritable by VLAN devices
1738 * @hw_enc_features: Mask of features inherited by encapsulating devices
1739 * This field indicates what encapsulation
1740 * offloads the hardware is capable of doing,
1741 * and drivers will need to set them appropriately.
1743 * @mpls_features: Mask of features inheritable by MPLS
1744 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1746 * @ifindex: interface index
1747 * @group: The group the device belongs to
1749 * @stats: Statistics struct, which was left as a legacy, use
1750 * rtnl_link_stats64 instead
1752 * @rx_dropped: Dropped packets by core network,
1753 * do not use this in drivers
1754 * @tx_dropped: Dropped packets by core network,
1755 * do not use this in drivers
1756 * @rx_nohandler: nohandler dropped packets by core network on
1757 * inactive devices, do not use this in drivers
1758 * @carrier_up_count: Number of times the carrier has been up
1759 * @carrier_down_count: Number of times the carrier has been down
1761 * @wireless_handlers: List of functions to handle Wireless Extensions,
1763 * see <net/iw_handler.h> for details.
1764 * @wireless_data: Instance data managed by the core of wireless extensions
1766 * @netdev_ops: Includes several pointers to callbacks,
1767 * if one wants to override the ndo_*() functions
1768 * @ethtool_ops: Management operations
1769 * @l3mdev_ops: Layer 3 master device operations
1770 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1771 * discovery handling. Necessary for e.g. 6LoWPAN.
1772 * @xfrmdev_ops: Transformation offload operations
1773 * @tlsdev_ops: Transport Layer Security offload operations
1774 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1775 * of Layer 2 headers.
1777 * @flags: Interface flags (a la BSD)
1778 * @priv_flags: Like 'flags' but invisible to userspace,
1779 * see if.h for the definitions
1780 * @gflags: Global flags ( kept as legacy )
1781 * @padded: How much padding added by alloc_netdev()
1782 * @operstate: RFC2863 operstate
1783 * @link_mode: Mapping policy to operstate
1784 * @if_port: Selectable AUI, TP, ...
1786 * @mtu: Interface MTU value
1787 * @min_mtu: Interface Minimum MTU value
1788 * @max_mtu: Interface Maximum MTU value
1789 * @type: Interface hardware type
1790 * @hard_header_len: Maximum hardware header length.
1791 * @min_header_len: Minimum hardware header length
1793 * @needed_headroom: Extra headroom the hardware may need, but not in all
1794 * cases can this be guaranteed
1795 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1796 * cases can this be guaranteed. Some cases also use
1797 * LL_MAX_HEADER instead to allocate the skb
1799 * interface address info:
1801 * @perm_addr: Permanent hw address
1802 * @addr_assign_type: Hw address assignment type
1803 * @addr_len: Hardware address length
1804 * @upper_level: Maximum depth level of upper devices.
1805 * @lower_level: Maximum depth level of lower devices.
1806 * @neigh_priv_len: Used in neigh_alloc()
1807 * @dev_id: Used to differentiate devices that share
1808 * the same link layer address
1809 * @dev_port: Used to differentiate devices that share
1811 * @addr_list_lock: XXX: need comments on this one
1812 * @name_assign_type: network interface name assignment type
1813 * @uc_promisc: Counter that indicates promiscuous mode
1814 * has been enabled due to the need to listen to
1815 * additional unicast addresses in a device that
1816 * does not implement ndo_set_rx_mode()
1817 * @uc: unicast mac addresses
1818 * @mc: multicast mac addresses
1819 * @dev_addrs: list of device hw addresses
1820 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1821 * @promiscuity: Number of times the NIC is told to work in
1822 * promiscuous mode; if it becomes 0 the NIC will
1823 * exit promiscuous mode
1824 * @allmulti: Counter, enables or disables allmulticast mode
1826 * @vlan_info: VLAN info
1827 * @dsa_ptr: dsa specific data
1828 * @tipc_ptr: TIPC specific data
1829 * @atalk_ptr: AppleTalk link
1830 * @ip_ptr: IPv4 specific data
1831 * @dn_ptr: DECnet specific data
1832 * @ip6_ptr: IPv6 specific data
1833 * @ax25_ptr: AX.25 specific data
1834 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1835 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1837 * @mpls_ptr: mpls_dev struct pointer
1838 * @mctp_ptr: MCTP specific data
1840 * @dev_addr: Hw address (before bcast,
1841 * because most packets are unicast)
1843 * @_rx: Array of RX queues
1844 * @num_rx_queues: Number of RX queues
1845 * allocated at register_netdev() time
1846 * @real_num_rx_queues: Number of RX queues currently active in device
1847 * @xdp_prog: XDP sockets filter program pointer
1848 * @gro_flush_timeout: timeout for GRO layer in NAPI
1849 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1850 * allow to avoid NIC hard IRQ, on busy queues.
1852 * @rx_handler: handler for received packets
1853 * @rx_handler_data: XXX: need comments on this one
1854 * @miniq_ingress: ingress/clsact qdisc specific data for
1855 * ingress processing
1856 * @ingress_queue: XXX: need comments on this one
1857 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1858 * @broadcast: hw bcast address
1860 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1861 * indexed by RX queue number. Assigned by driver.
1862 * This must only be set if the ndo_rx_flow_steer
1863 * operation is defined
1864 * @index_hlist: Device index hash chain
1866 * @_tx: Array of TX queues
1867 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1868 * @real_num_tx_queues: Number of TX queues currently active in device
1869 * @qdisc: Root qdisc from userspace point of view
1870 * @tx_queue_len: Max frames per queue allowed
1871 * @tx_global_lock: XXX: need comments on this one
1872 * @xdp_bulkq: XDP device bulk queue
1873 * @xps_maps: all CPUs/RXQs maps for XPS device
1875 * @xps_maps: XXX: need comments on this one
1876 * @miniq_egress: clsact qdisc specific data for
1878 * @qdisc_hash: qdisc hash table
1879 * @watchdog_timeo: Represents the timeout that is used by
1880 * the watchdog (see dev_watchdog())
1881 * @watchdog_timer: List of timers
1883 * @proto_down_reason: reason a netdev interface is held down
1884 * @pcpu_refcnt: Number of references to this device
1885 * @dev_refcnt: Number of references to this device
1886 * @todo_list: Delayed register/unregister
1887 * @link_watch_list: XXX: need comments on this one
1889 * @reg_state: Register/unregister state machine
1890 * @dismantle: Device is going to be freed
1891 * @rtnl_link_state: This enum represents the phases of creating
1894 * @needs_free_netdev: Should unregister perform free_netdev?
1895 * @priv_destructor: Called from unregister
1896 * @npinfo: XXX: need comments on this one
1897 * @nd_net: Network namespace this network device is inside
1899 * @ml_priv: Mid-layer private
1900 * @ml_priv_type: Mid-layer private type
1901 * @lstats: Loopback statistics
1902 * @tstats: Tunnel statistics
1903 * @dstats: Dummy statistics
1904 * @vstats: Virtual ethernet statistics
1909 * @dev: Class/net/name entry
1910 * @sysfs_groups: Space for optional device, statistics and wireless
1913 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1914 * @rtnl_link_ops: Rtnl_link_ops
1916 * @gso_max_size: Maximum size of generic segmentation offload
1917 * @gso_max_segs: Maximum number of segments that can be passed to the
1920 * @dcbnl_ops: Data Center Bridging netlink ops
1921 * @num_tc: Number of traffic classes in the net device
1922 * @tc_to_txq: XXX: need comments on this one
1923 * @prio_tc_map: XXX: need comments on this one
1925 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1927 * @priomap: XXX: need comments on this one
1928 * @phydev: Physical device may attach itself
1929 * for hardware timestamping
1930 * @sfp_bus: attached &struct sfp_bus structure.
1932 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1934 * @proto_down: protocol port state information can be sent to the
1935 * switch driver and used to set the phys state of the
1938 * @wol_enabled: Wake-on-LAN is enabled
1940 * @threaded: napi threaded mode is enabled
1942 * @net_notifier_list: List of per-net netdev notifier block
1943 * that follow this device when it is moved
1944 * to another network namespace.
1946 * @macsec_ops: MACsec offloading ops
1948 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1949 * offload capabilities of the device
1950 * @udp_tunnel_nic: UDP tunnel offload state
1951 * @xdp_state: stores info on attached XDP BPF programs
1953 * @nested_level: Used as as a parameter of spin_lock_nested() of
1954 * dev->addr_list_lock.
1955 * @unlink_list: As netif_addr_lock() can be called recursively,
1956 * keep a list of interfaces to be deleted.
1958 * FIXME: cleanup struct net_device such that network protocol info
1963 char name[IFNAMSIZ];
1964 struct netdev_name_node *name_node;
1965 struct dev_ifalias __rcu *ifalias;
1967 * I/O specific fields
1968 * FIXME: Merge these and struct ifmap into one
1970 unsigned long mem_end;
1971 unsigned long mem_start;
1972 unsigned long base_addr;
1975 * Some hardware also needs these fields (state,dev_list,
1976 * napi_list,unreg_list,close_list) but they are not
1977 * part of the usual set specified in Space.c.
1980 unsigned long state;
1982 struct list_head dev_list;
1983 struct list_head napi_list;
1984 struct list_head unreg_list;
1985 struct list_head close_list;
1986 struct list_head ptype_all;
1987 struct list_head ptype_specific;
1990 struct list_head upper;
1991 struct list_head lower;
1994 /* Read-mostly cache-line for fast-path access */
1996 unsigned int priv_flags;
1997 const struct net_device_ops *netdev_ops;
1999 unsigned short gflags;
2000 unsigned short hard_header_len;
2002 /* Note : dev->mtu is often read without holding a lock.
2003 * Writers usually hold RTNL.
2004 * It is recommended to use READ_ONCE() to annotate the reads,
2005 * and to use WRITE_ONCE() to annotate the writes.
2008 unsigned short needed_headroom;
2009 unsigned short needed_tailroom;
2011 netdev_features_t features;
2012 netdev_features_t hw_features;
2013 netdev_features_t wanted_features;
2014 netdev_features_t vlan_features;
2015 netdev_features_t hw_enc_features;
2016 netdev_features_t mpls_features;
2017 netdev_features_t gso_partial_features;
2019 unsigned int min_mtu;
2020 unsigned int max_mtu;
2021 unsigned short type;
2022 unsigned char min_header_len;
2023 unsigned char name_assign_type;
2027 struct net_device_stats stats; /* not used by modern drivers */
2029 atomic_long_t rx_dropped;
2030 atomic_long_t tx_dropped;
2031 atomic_long_t rx_nohandler;
2033 /* Stats to monitor link on/off, flapping */
2034 atomic_t carrier_up_count;
2035 atomic_t carrier_down_count;
2037 #ifdef CONFIG_WIRELESS_EXT
2038 const struct iw_handler_def *wireless_handlers;
2039 struct iw_public_data *wireless_data;
2041 const struct ethtool_ops *ethtool_ops;
2042 #ifdef CONFIG_NET_L3_MASTER_DEV
2043 const struct l3mdev_ops *l3mdev_ops;
2045 #if IS_ENABLED(CONFIG_IPV6)
2046 const struct ndisc_ops *ndisc_ops;
2049 #ifdef CONFIG_XFRM_OFFLOAD
2050 const struct xfrmdev_ops *xfrmdev_ops;
2053 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2054 const struct tlsdev_ops *tlsdev_ops;
2057 const struct header_ops *header_ops;
2059 unsigned char operstate;
2060 unsigned char link_mode;
2062 unsigned char if_port;
2065 /* Interface address info. */
2066 unsigned char perm_addr[MAX_ADDR_LEN];
2067 unsigned char addr_assign_type;
2068 unsigned char addr_len;
2069 unsigned char upper_level;
2070 unsigned char lower_level;
2072 unsigned short neigh_priv_len;
2073 unsigned short dev_id;
2074 unsigned short dev_port;
2075 unsigned short padded;
2077 spinlock_t addr_list_lock;
2080 struct netdev_hw_addr_list uc;
2081 struct netdev_hw_addr_list mc;
2082 struct netdev_hw_addr_list dev_addrs;
2085 struct kset *queues_kset;
2087 #ifdef CONFIG_LOCKDEP
2088 struct list_head unlink_list;
2090 unsigned int promiscuity;
2091 unsigned int allmulti;
2093 #ifdef CONFIG_LOCKDEP
2094 unsigned char nested_level;
2098 /* Protocol-specific pointers */
2100 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2101 struct vlan_info __rcu *vlan_info;
2103 #if IS_ENABLED(CONFIG_NET_DSA)
2104 struct dsa_port *dsa_ptr;
2106 #if IS_ENABLED(CONFIG_TIPC)
2107 struct tipc_bearer __rcu *tipc_ptr;
2109 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2112 struct in_device __rcu *ip_ptr;
2113 #if IS_ENABLED(CONFIG_DECNET)
2114 struct dn_dev __rcu *dn_ptr;
2116 struct inet6_dev __rcu *ip6_ptr;
2117 #if IS_ENABLED(CONFIG_AX25)
2120 struct wireless_dev *ieee80211_ptr;
2121 struct wpan_dev *ieee802154_ptr;
2122 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2123 struct mpls_dev __rcu *mpls_ptr;
2125 #if IS_ENABLED(CONFIG_MCTP)
2126 struct mctp_dev __rcu *mctp_ptr;
2130 * Cache lines mostly used on receive path (including eth_type_trans())
2132 /* Interface address info used in eth_type_trans() */
2133 unsigned char *dev_addr;
2135 struct netdev_rx_queue *_rx;
2136 unsigned int num_rx_queues;
2137 unsigned int real_num_rx_queues;
2139 struct bpf_prog __rcu *xdp_prog;
2140 unsigned long gro_flush_timeout;
2141 int napi_defer_hard_irqs;
2142 rx_handler_func_t __rcu *rx_handler;
2143 void __rcu *rx_handler_data;
2145 #ifdef CONFIG_NET_CLS_ACT
2146 struct mini_Qdisc __rcu *miniq_ingress;
2148 struct netdev_queue __rcu *ingress_queue;
2149 #ifdef CONFIG_NETFILTER_INGRESS
2150 struct nf_hook_entries __rcu *nf_hooks_ingress;
2153 unsigned char broadcast[MAX_ADDR_LEN];
2154 #ifdef CONFIG_RFS_ACCEL
2155 struct cpu_rmap *rx_cpu_rmap;
2157 struct hlist_node index_hlist;
2160 * Cache lines mostly used on transmit path
2162 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2163 unsigned int num_tx_queues;
2164 unsigned int real_num_tx_queues;
2165 struct Qdisc __rcu *qdisc;
2166 unsigned int tx_queue_len;
2167 spinlock_t tx_global_lock;
2169 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2172 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2174 #ifdef CONFIG_NET_CLS_ACT
2175 struct mini_Qdisc __rcu *miniq_egress;
2178 #ifdef CONFIG_NET_SCHED
2179 DECLARE_HASHTABLE (qdisc_hash, 4);
2181 /* These may be needed for future network-power-down code. */
2182 struct timer_list watchdog_timer;
2185 u32 proto_down_reason;
2187 struct list_head todo_list;
2189 #ifdef CONFIG_PCPU_DEV_REFCNT
2190 int __percpu *pcpu_refcnt;
2192 refcount_t dev_refcnt;
2195 struct list_head link_watch_list;
2197 enum { NETREG_UNINITIALIZED=0,
2198 NETREG_REGISTERED, /* completed register_netdevice */
2199 NETREG_UNREGISTERING, /* called unregister_netdevice */
2200 NETREG_UNREGISTERED, /* completed unregister todo */
2201 NETREG_RELEASED, /* called free_netdev */
2202 NETREG_DUMMY, /* dummy device for NAPI poll */
2208 RTNL_LINK_INITIALIZED,
2209 RTNL_LINK_INITIALIZING,
2210 } rtnl_link_state:16;
2212 bool needs_free_netdev;
2213 void (*priv_destructor)(struct net_device *dev);
2215 #ifdef CONFIG_NETPOLL
2216 struct netpoll_info __rcu *npinfo;
2219 possible_net_t nd_net;
2221 /* mid-layer private */
2223 enum netdev_ml_priv_type ml_priv_type;
2226 struct pcpu_lstats __percpu *lstats;
2227 struct pcpu_sw_netstats __percpu *tstats;
2228 struct pcpu_dstats __percpu *dstats;
2231 #if IS_ENABLED(CONFIG_GARP)
2232 struct garp_port __rcu *garp_port;
2234 #if IS_ENABLED(CONFIG_MRP)
2235 struct mrp_port __rcu *mrp_port;
2239 const struct attribute_group *sysfs_groups[4];
2240 const struct attribute_group *sysfs_rx_queue_group;
2242 const struct rtnl_link_ops *rtnl_link_ops;
2244 /* for setting kernel sock attribute on TCP connection setup */
2245 #define GSO_MAX_SIZE 65536
2246 unsigned int gso_max_size;
2247 #define GSO_MAX_SEGS 65535
2251 const struct dcbnl_rtnl_ops *dcbnl_ops;
2254 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2255 u8 prio_tc_map[TC_BITMASK + 1];
2257 #if IS_ENABLED(CONFIG_FCOE)
2258 unsigned int fcoe_ddp_xid;
2260 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2261 struct netprio_map __rcu *priomap;
2263 struct phy_device *phydev;
2264 struct sfp_bus *sfp_bus;
2265 struct lock_class_key *qdisc_tx_busylock;
2267 unsigned wol_enabled:1;
2268 unsigned threaded:1;
2270 struct list_head net_notifier_list;
2272 #if IS_ENABLED(CONFIG_MACSEC)
2273 /* MACsec management functions */
2274 const struct macsec_ops *macsec_ops;
2276 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2277 struct udp_tunnel_nic *udp_tunnel_nic;
2279 /* protected by rtnl_lock */
2280 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2282 #define to_net_dev(d) container_of(d, struct net_device, dev)
2284 static inline bool netif_elide_gro(const struct net_device *dev)
2286 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2291 #define NETDEV_ALIGN 32
2294 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2296 return dev->prio_tc_map[prio & TC_BITMASK];
2300 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2302 if (tc >= dev->num_tc)
2305 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2309 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2310 void netdev_reset_tc(struct net_device *dev);
2311 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2312 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2315 int netdev_get_num_tc(struct net_device *dev)
2320 static inline void net_prefetch(void *p)
2323 #if L1_CACHE_BYTES < 128
2324 prefetch((u8 *)p + L1_CACHE_BYTES);
2328 static inline void net_prefetchw(void *p)
2331 #if L1_CACHE_BYTES < 128
2332 prefetchw((u8 *)p + L1_CACHE_BYTES);
2336 void netdev_unbind_sb_channel(struct net_device *dev,
2337 struct net_device *sb_dev);
2338 int netdev_bind_sb_channel_queue(struct net_device *dev,
2339 struct net_device *sb_dev,
2340 u8 tc, u16 count, u16 offset);
2341 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2342 static inline int netdev_get_sb_channel(struct net_device *dev)
2344 return max_t(int, -dev->num_tc, 0);
2348 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2351 return &dev->_tx[index];
2354 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2355 const struct sk_buff *skb)
2357 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2360 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2361 void (*f)(struct net_device *,
2362 struct netdev_queue *,
2368 for (i = 0; i < dev->num_tx_queues; i++)
2369 f(dev, &dev->_tx[i], arg);
2372 #define netdev_lockdep_set_classes(dev) \
2374 static struct lock_class_key qdisc_tx_busylock_key; \
2375 static struct lock_class_key qdisc_xmit_lock_key; \
2376 static struct lock_class_key dev_addr_list_lock_key; \
2379 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2380 lockdep_set_class(&(dev)->addr_list_lock, \
2381 &dev_addr_list_lock_key); \
2382 for (i = 0; i < (dev)->num_tx_queues; i++) \
2383 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2384 &qdisc_xmit_lock_key); \
2387 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2388 struct net_device *sb_dev);
2389 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2390 struct sk_buff *skb,
2391 struct net_device *sb_dev);
2393 /* returns the headroom that the master device needs to take in account
2394 * when forwarding to this dev
2396 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2398 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2401 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2403 if (dev->netdev_ops->ndo_set_rx_headroom)
2404 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2407 /* set the device rx headroom to the dev's default */
2408 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2410 netdev_set_rx_headroom(dev, -1);
2413 static inline void *netdev_get_ml_priv(struct net_device *dev,
2414 enum netdev_ml_priv_type type)
2416 if (dev->ml_priv_type != type)
2419 return dev->ml_priv;
2422 static inline void netdev_set_ml_priv(struct net_device *dev,
2424 enum netdev_ml_priv_type type)
2426 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2427 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2428 dev->ml_priv_type, type);
2429 WARN(!dev->ml_priv_type && dev->ml_priv,
2430 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2432 dev->ml_priv = ml_priv;
2433 dev->ml_priv_type = type;
2437 * Net namespace inlines
2440 struct net *dev_net(const struct net_device *dev)
2442 return read_pnet(&dev->nd_net);
2446 void dev_net_set(struct net_device *dev, struct net *net)
2448 write_pnet(&dev->nd_net, net);
2452 * netdev_priv - access network device private data
2453 * @dev: network device
2455 * Get network device private data
2457 static inline void *netdev_priv(const struct net_device *dev)
2459 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2462 /* Set the sysfs physical device reference for the network logical device
2463 * if set prior to registration will cause a symlink during initialization.
2465 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2467 /* Set the sysfs device type for the network logical device to allow
2468 * fine-grained identification of different network device types. For
2469 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2471 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2473 /* Default NAPI poll() weight
2474 * Device drivers are strongly advised to not use bigger value
2476 #define NAPI_POLL_WEIGHT 64
2479 * netif_napi_add - initialize a NAPI context
2480 * @dev: network device
2481 * @napi: NAPI context
2482 * @poll: polling function
2483 * @weight: default weight
2485 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2486 * *any* of the other NAPI-related functions.
2488 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2489 int (*poll)(struct napi_struct *, int), int weight);
2492 * netif_tx_napi_add - initialize a NAPI context
2493 * @dev: network device
2494 * @napi: NAPI context
2495 * @poll: polling function
2496 * @weight: default weight
2498 * This variant of netif_napi_add() should be used from drivers using NAPI
2499 * to exclusively poll a TX queue.
2500 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2502 static inline void netif_tx_napi_add(struct net_device *dev,
2503 struct napi_struct *napi,
2504 int (*poll)(struct napi_struct *, int),
2507 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2508 netif_napi_add(dev, napi, poll, weight);
2512 * __netif_napi_del - remove a NAPI context
2513 * @napi: NAPI context
2515 * Warning: caller must observe RCU grace period before freeing memory
2516 * containing @napi. Drivers might want to call this helper to combine
2517 * all the needed RCU grace periods into a single one.
2519 void __netif_napi_del(struct napi_struct *napi);
2522 * netif_napi_del - remove a NAPI context
2523 * @napi: NAPI context
2525 * netif_napi_del() removes a NAPI context from the network device NAPI list
2527 static inline void netif_napi_del(struct napi_struct *napi)
2529 __netif_napi_del(napi);
2533 struct napi_gro_cb {
2534 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2537 /* Length of frag0. */
2538 unsigned int frag0_len;
2540 /* This indicates where we are processing relative to skb->data. */
2543 /* This is non-zero if the packet cannot be merged with the new skb. */
2546 /* Save the IP ID here and check when we get to the transport layer */
2549 /* Number of segments aggregated. */
2552 /* Start offset for remote checksum offload */
2553 u16 gro_remcsum_start;
2555 /* jiffies when first packet was created/queued */
2558 /* Used in ipv6_gro_receive() and foo-over-udp */
2561 /* This is non-zero if the packet may be of the same flow. */
2564 /* Used in tunnel GRO receive */
2567 /* GRO checksum is valid */
2570 /* Number of checksums via CHECKSUM_UNNECESSARY */
2575 #define NAPI_GRO_FREE 1
2576 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2578 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2581 /* Used in GRE, set in fou/gue_gro_receive */
2584 /* Used to determine if flush_id can be ignored */
2587 /* Number of gro_receive callbacks this packet already went through */
2588 u8 recursion_counter:4;
2590 /* GRO is done by frag_list pointer chaining. */
2593 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2596 /* used in skb_gro_receive() slow path */
2597 struct sk_buff *last;
2600 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2602 #define GRO_RECURSION_LIMIT 15
2603 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2605 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2608 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2609 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2610 struct list_head *head,
2611 struct sk_buff *skb)
2613 if (unlikely(gro_recursion_inc_test(skb))) {
2614 NAPI_GRO_CB(skb)->flush |= 1;
2618 return cb(head, skb);
2621 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2623 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2625 struct list_head *head,
2626 struct sk_buff *skb)
2628 if (unlikely(gro_recursion_inc_test(skb))) {
2629 NAPI_GRO_CB(skb)->flush |= 1;
2633 return cb(sk, head, skb);
2636 struct packet_type {
2637 __be16 type; /* This is really htons(ether_type). */
2638 bool ignore_outgoing;
2639 struct net_device *dev; /* NULL is wildcarded here */
2640 int (*func) (struct sk_buff *,
2641 struct net_device *,
2642 struct packet_type *,
2643 struct net_device *);
2644 void (*list_func) (struct list_head *,
2645 struct packet_type *,
2646 struct net_device *);
2647 bool (*id_match)(struct packet_type *ptype,
2649 struct net *af_packet_net;
2650 void *af_packet_priv;
2651 struct list_head list;
2654 struct offload_callbacks {
2655 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2656 netdev_features_t features);
2657 struct sk_buff *(*gro_receive)(struct list_head *head,
2658 struct sk_buff *skb);
2659 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2662 struct packet_offload {
2663 __be16 type; /* This is really htons(ether_type). */
2665 struct offload_callbacks callbacks;
2666 struct list_head list;
2669 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2670 struct pcpu_sw_netstats {
2675 struct u64_stats_sync syncp;
2676 } __aligned(4 * sizeof(u64));
2678 struct pcpu_lstats {
2679 u64_stats_t packets;
2681 struct u64_stats_sync syncp;
2682 } __aligned(2 * sizeof(u64));
2684 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2686 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2688 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2690 u64_stats_update_begin(&tstats->syncp);
2691 tstats->rx_bytes += len;
2692 tstats->rx_packets++;
2693 u64_stats_update_end(&tstats->syncp);
2696 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2697 unsigned int packets,
2700 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2702 u64_stats_update_begin(&tstats->syncp);
2703 tstats->tx_bytes += len;
2704 tstats->tx_packets += packets;
2705 u64_stats_update_end(&tstats->syncp);
2708 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2710 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2712 u64_stats_update_begin(&lstats->syncp);
2713 u64_stats_add(&lstats->bytes, len);
2714 u64_stats_inc(&lstats->packets);
2715 u64_stats_update_end(&lstats->syncp);
2718 #define __netdev_alloc_pcpu_stats(type, gfp) \
2720 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2723 for_each_possible_cpu(__cpu) { \
2724 typeof(type) *stat; \
2725 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2726 u64_stats_init(&stat->syncp); \
2732 #define netdev_alloc_pcpu_stats(type) \
2733 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2735 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2737 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2740 for_each_possible_cpu(__cpu) { \
2741 typeof(type) *stat; \
2742 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2743 u64_stats_init(&stat->syncp); \
2749 enum netdev_lag_tx_type {
2750 NETDEV_LAG_TX_TYPE_UNKNOWN,
2751 NETDEV_LAG_TX_TYPE_RANDOM,
2752 NETDEV_LAG_TX_TYPE_BROADCAST,
2753 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2754 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2755 NETDEV_LAG_TX_TYPE_HASH,
2758 enum netdev_lag_hash {
2759 NETDEV_LAG_HASH_NONE,
2761 NETDEV_LAG_HASH_L34,
2762 NETDEV_LAG_HASH_L23,
2763 NETDEV_LAG_HASH_E23,
2764 NETDEV_LAG_HASH_E34,
2765 NETDEV_LAG_HASH_VLAN_SRCMAC,
2766 NETDEV_LAG_HASH_UNKNOWN,
2769 struct netdev_lag_upper_info {
2770 enum netdev_lag_tx_type tx_type;
2771 enum netdev_lag_hash hash_type;
2774 struct netdev_lag_lower_state_info {
2779 #include <linux/notifier.h>
2781 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2782 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2786 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2788 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2789 detected a hardware crash and restarted
2790 - we can use this eg to kick tcp sessions
2792 NETDEV_CHANGE, /* Notify device state change */
2795 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2796 NETDEV_CHANGEADDR, /* notify after the address change */
2797 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2801 NETDEV_BONDING_FAILOVER,
2803 NETDEV_PRE_TYPE_CHANGE,
2804 NETDEV_POST_TYPE_CHANGE,
2807 NETDEV_NOTIFY_PEERS,
2811 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2812 NETDEV_CHANGEINFODATA,
2813 NETDEV_BONDING_INFO,
2814 NETDEV_PRECHANGEUPPER,
2815 NETDEV_CHANGELOWERSTATE,
2816 NETDEV_UDP_TUNNEL_PUSH_INFO,
2817 NETDEV_UDP_TUNNEL_DROP_INFO,
2818 NETDEV_CHANGE_TX_QUEUE_LEN,
2819 NETDEV_CVLAN_FILTER_PUSH_INFO,
2820 NETDEV_CVLAN_FILTER_DROP_INFO,
2821 NETDEV_SVLAN_FILTER_PUSH_INFO,
2822 NETDEV_SVLAN_FILTER_DROP_INFO,
2824 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2826 int register_netdevice_notifier(struct notifier_block *nb);
2827 int unregister_netdevice_notifier(struct notifier_block *nb);
2828 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2829 int unregister_netdevice_notifier_net(struct net *net,
2830 struct notifier_block *nb);
2831 int register_netdevice_notifier_dev_net(struct net_device *dev,
2832 struct notifier_block *nb,
2833 struct netdev_net_notifier *nn);
2834 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2835 struct notifier_block *nb,
2836 struct netdev_net_notifier *nn);
2838 struct netdev_notifier_info {
2839 struct net_device *dev;
2840 struct netlink_ext_ack *extack;
2843 struct netdev_notifier_info_ext {
2844 struct netdev_notifier_info info; /* must be first */
2850 struct netdev_notifier_change_info {
2851 struct netdev_notifier_info info; /* must be first */
2852 unsigned int flags_changed;
2855 struct netdev_notifier_changeupper_info {
2856 struct netdev_notifier_info info; /* must be first */
2857 struct net_device *upper_dev; /* new upper dev */
2858 bool master; /* is upper dev master */
2859 bool linking; /* is the notification for link or unlink */
2860 void *upper_info; /* upper dev info */
2863 struct netdev_notifier_changelowerstate_info {
2864 struct netdev_notifier_info info; /* must be first */
2865 void *lower_state_info; /* is lower dev state */
2868 struct netdev_notifier_pre_changeaddr_info {
2869 struct netdev_notifier_info info; /* must be first */
2870 const unsigned char *dev_addr;
2873 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2874 struct net_device *dev)
2877 info->extack = NULL;
2880 static inline struct net_device *
2881 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2886 static inline struct netlink_ext_ack *
2887 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2889 return info->extack;
2892 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2895 extern rwlock_t dev_base_lock; /* Device list lock */
2897 #define for_each_netdev(net, d) \
2898 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2899 #define for_each_netdev_reverse(net, d) \
2900 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2901 #define for_each_netdev_rcu(net, d) \
2902 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2903 #define for_each_netdev_safe(net, d, n) \
2904 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2905 #define for_each_netdev_continue(net, d) \
2906 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2907 #define for_each_netdev_continue_reverse(net, d) \
2908 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2910 #define for_each_netdev_continue_rcu(net, d) \
2911 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2912 #define for_each_netdev_in_bond_rcu(bond, slave) \
2913 for_each_netdev_rcu(&init_net, slave) \
2914 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2915 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2917 static inline struct net_device *next_net_device(struct net_device *dev)
2919 struct list_head *lh;
2923 lh = dev->dev_list.next;
2924 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2927 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2929 struct list_head *lh;
2933 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2934 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2937 static inline struct net_device *first_net_device(struct net *net)
2939 return list_empty(&net->dev_base_head) ? NULL :
2940 net_device_entry(net->dev_base_head.next);
2943 static inline struct net_device *first_net_device_rcu(struct net *net)
2945 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2947 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2950 int netdev_boot_setup_check(struct net_device *dev);
2951 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2952 const char *hwaddr);
2953 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2954 void dev_add_pack(struct packet_type *pt);
2955 void dev_remove_pack(struct packet_type *pt);
2956 void __dev_remove_pack(struct packet_type *pt);
2957 void dev_add_offload(struct packet_offload *po);
2958 void dev_remove_offload(struct packet_offload *po);
2960 int dev_get_iflink(const struct net_device *dev);
2961 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2962 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2963 struct net_device_path_stack *stack);
2964 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2965 unsigned short mask);
2966 struct net_device *dev_get_by_name(struct net *net, const char *name);
2967 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2968 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2969 int dev_alloc_name(struct net_device *dev, const char *name);
2970 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2971 void dev_close(struct net_device *dev);
2972 void dev_close_many(struct list_head *head, bool unlink);
2973 void dev_disable_lro(struct net_device *dev);
2974 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2975 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2976 struct net_device *sb_dev);
2977 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2978 struct net_device *sb_dev);
2980 int dev_queue_xmit(struct sk_buff *skb);
2981 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2982 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2984 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2988 ret = __dev_direct_xmit(skb, queue_id);
2989 if (!dev_xmit_complete(ret))
2994 int register_netdevice(struct net_device *dev);
2995 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2996 void unregister_netdevice_many(struct list_head *head);
2997 static inline void unregister_netdevice(struct net_device *dev)
2999 unregister_netdevice_queue(dev, NULL);
3002 int netdev_refcnt_read(const struct net_device *dev);
3003 void free_netdev(struct net_device *dev);
3004 void netdev_freemem(struct net_device *dev);
3005 int init_dummy_netdev(struct net_device *dev);
3007 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3008 struct sk_buff *skb,
3010 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3012 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3013 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3014 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3015 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3016 int netdev_get_name(struct net *net, char *name, int ifindex);
3017 int dev_restart(struct net_device *dev);
3018 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3019 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3021 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3023 return NAPI_GRO_CB(skb)->data_offset;
3026 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3028 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3031 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3033 NAPI_GRO_CB(skb)->data_offset += len;
3036 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3037 unsigned int offset)
3039 return NAPI_GRO_CB(skb)->frag0 + offset;
3042 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3044 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3047 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3049 NAPI_GRO_CB(skb)->frag0 = NULL;
3050 NAPI_GRO_CB(skb)->frag0_len = 0;
3053 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3054 unsigned int offset)
3056 if (!pskb_may_pull(skb, hlen))
3059 skb_gro_frag0_invalidate(skb);
3060 return skb->data + offset;
3063 static inline void *skb_gro_network_header(struct sk_buff *skb)
3065 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3066 skb_network_offset(skb);
3069 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3070 const void *start, unsigned int len)
3072 if (NAPI_GRO_CB(skb)->csum_valid)
3073 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3074 csum_partial(start, len, 0));
3077 /* GRO checksum functions. These are logical equivalents of the normal
3078 * checksum functions (in skbuff.h) except that they operate on the GRO
3079 * offsets and fields in sk_buff.
3082 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3084 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3086 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3089 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3093 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3094 skb_checksum_start_offset(skb) <
3095 skb_gro_offset(skb)) &&
3096 !skb_at_gro_remcsum_start(skb) &&
3097 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3098 (!zero_okay || check));
3101 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3104 if (NAPI_GRO_CB(skb)->csum_valid &&
3105 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3108 NAPI_GRO_CB(skb)->csum = psum;
3110 return __skb_gro_checksum_complete(skb);
3113 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3115 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3116 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3117 NAPI_GRO_CB(skb)->csum_cnt--;
3119 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3120 * verified a new top level checksum or an encapsulated one
3121 * during GRO. This saves work if we fallback to normal path.
3123 __skb_incr_checksum_unnecessary(skb);
3127 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3130 __sum16 __ret = 0; \
3131 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3132 __ret = __skb_gro_checksum_validate_complete(skb, \
3133 compute_pseudo(skb, proto)); \
3135 skb_gro_incr_csum_unnecessary(skb); \
3139 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3140 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3142 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3144 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3146 #define skb_gro_checksum_simple_validate(skb) \
3147 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3149 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3151 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3152 !NAPI_GRO_CB(skb)->csum_valid);
3155 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3158 NAPI_GRO_CB(skb)->csum = ~pseudo;
3159 NAPI_GRO_CB(skb)->csum_valid = 1;
3162 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3164 if (__skb_gro_checksum_convert_check(skb)) \
3165 __skb_gro_checksum_convert(skb, \
3166 compute_pseudo(skb, proto)); \
3169 struct gro_remcsum {
3174 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3180 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3181 unsigned int off, size_t hdrlen,
3182 int start, int offset,
3183 struct gro_remcsum *grc,
3187 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3189 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3192 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3196 ptr = skb_gro_header_fast(skb, off);
3197 if (skb_gro_header_hard(skb, off + plen)) {
3198 ptr = skb_gro_header_slow(skb, off + plen, off);
3203 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3206 /* Adjust skb->csum since we changed the packet */
3207 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3209 grc->offset = off + hdrlen + offset;
3215 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3216 struct gro_remcsum *grc)
3219 size_t plen = grc->offset + sizeof(u16);
3224 ptr = skb_gro_header_fast(skb, grc->offset);
3225 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3226 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3231 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3234 #ifdef CONFIG_XFRM_OFFLOAD
3235 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3237 if (PTR_ERR(pp) != -EINPROGRESS)
3238 NAPI_GRO_CB(skb)->flush |= flush;
3240 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3243 struct gro_remcsum *grc)
3245 if (PTR_ERR(pp) != -EINPROGRESS) {
3246 NAPI_GRO_CB(skb)->flush |= flush;
3247 skb_gro_remcsum_cleanup(skb, grc);
3248 skb->remcsum_offload = 0;
3252 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3254 NAPI_GRO_CB(skb)->flush |= flush;
3256 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3259 struct gro_remcsum *grc)
3261 NAPI_GRO_CB(skb)->flush |= flush;
3262 skb_gro_remcsum_cleanup(skb, grc);
3263 skb->remcsum_offload = 0;
3267 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3268 unsigned short type,
3269 const void *daddr, const void *saddr,
3272 if (!dev->header_ops || !dev->header_ops->create)
3275 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3278 static inline int dev_parse_header(const struct sk_buff *skb,
3279 unsigned char *haddr)
3281 const struct net_device *dev = skb->dev;
3283 if (!dev->header_ops || !dev->header_ops->parse)
3285 return dev->header_ops->parse(skb, haddr);
3288 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3290 const struct net_device *dev = skb->dev;
3292 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3294 return dev->header_ops->parse_protocol(skb);
3297 /* ll_header must have at least hard_header_len allocated */
3298 static inline bool dev_validate_header(const struct net_device *dev,
3299 char *ll_header, int len)
3301 if (likely(len >= dev->hard_header_len))
3303 if (len < dev->min_header_len)
3306 if (capable(CAP_SYS_RAWIO)) {
3307 memset(ll_header + len, 0, dev->hard_header_len - len);
3311 if (dev->header_ops && dev->header_ops->validate)
3312 return dev->header_ops->validate(ll_header, len);
3317 static inline bool dev_has_header(const struct net_device *dev)
3319 return dev->header_ops && dev->header_ops->create;
3322 #ifdef CONFIG_NET_FLOW_LIMIT
3323 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3324 struct sd_flow_limit {
3326 unsigned int num_buckets;
3327 unsigned int history_head;
3328 u16 history[FLOW_LIMIT_HISTORY];
3332 extern int netdev_flow_limit_table_len;
3333 #endif /* CONFIG_NET_FLOW_LIMIT */
3336 * Incoming packets are placed on per-CPU queues
3338 struct softnet_data {
3339 struct list_head poll_list;
3340 struct sk_buff_head process_queue;
3343 unsigned int processed;
3344 unsigned int time_squeeze;
3345 unsigned int received_rps;
3347 struct softnet_data *rps_ipi_list;
3349 #ifdef CONFIG_NET_FLOW_LIMIT
3350 struct sd_flow_limit __rcu *flow_limit;
3352 struct Qdisc *output_queue;
3353 struct Qdisc **output_queue_tailp;
3354 struct sk_buff *completion_queue;
3355 #ifdef CONFIG_XFRM_OFFLOAD
3356 struct sk_buff_head xfrm_backlog;
3358 /* written and read only by owning cpu: */
3364 /* input_queue_head should be written by cpu owning this struct,
3365 * and only read by other cpus. Worth using a cache line.
3367 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3369 /* Elements below can be accessed between CPUs for RPS/RFS */
3370 call_single_data_t csd ____cacheline_aligned_in_smp;
3371 struct softnet_data *rps_ipi_next;
3373 unsigned int input_queue_tail;
3375 unsigned int dropped;
3376 struct sk_buff_head input_pkt_queue;
3377 struct napi_struct backlog;
3381 static inline void input_queue_head_incr(struct softnet_data *sd)
3384 sd->input_queue_head++;
3388 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3389 unsigned int *qtail)
3392 *qtail = ++sd->input_queue_tail;
3396 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3398 static inline int dev_recursion_level(void)
3400 return this_cpu_read(softnet_data.xmit.recursion);
3403 #define XMIT_RECURSION_LIMIT 8
3404 static inline bool dev_xmit_recursion(void)
3406 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3407 XMIT_RECURSION_LIMIT);
3410 static inline void dev_xmit_recursion_inc(void)
3412 __this_cpu_inc(softnet_data.xmit.recursion);
3415 static inline void dev_xmit_recursion_dec(void)
3417 __this_cpu_dec(softnet_data.xmit.recursion);
3420 void __netif_schedule(struct Qdisc *q);
3421 void netif_schedule_queue(struct netdev_queue *txq);
3423 static inline void netif_tx_schedule_all(struct net_device *dev)
3427 for (i = 0; i < dev->num_tx_queues; i++)
3428 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3431 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3433 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3437 * netif_start_queue - allow transmit
3438 * @dev: network device
3440 * Allow upper layers to call the device hard_start_xmit routine.
3442 static inline void netif_start_queue(struct net_device *dev)
3444 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3447 static inline void netif_tx_start_all_queues(struct net_device *dev)
3451 for (i = 0; i < dev->num_tx_queues; i++) {
3452 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3453 netif_tx_start_queue(txq);
3457 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3460 * netif_wake_queue - restart transmit
3461 * @dev: network device
3463 * Allow upper layers to call the device hard_start_xmit routine.
3464 * Used for flow control when transmit resources are available.
3466 static inline void netif_wake_queue(struct net_device *dev)
3468 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3471 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3475 for (i = 0; i < dev->num_tx_queues; i++) {
3476 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3477 netif_tx_wake_queue(txq);
3481 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3483 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3487 * netif_stop_queue - stop transmitted packets
3488 * @dev: network device
3490 * Stop upper layers calling the device hard_start_xmit routine.
3491 * Used for flow control when transmit resources are unavailable.
3493 static inline void netif_stop_queue(struct net_device *dev)
3495 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3498 void netif_tx_stop_all_queues(struct net_device *dev);
3500 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3502 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3506 * netif_queue_stopped - test if transmit queue is flowblocked
3507 * @dev: network device
3509 * Test if transmit queue on device is currently unable to send.
3511 static inline bool netif_queue_stopped(const struct net_device *dev)
3513 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3516 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3518 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3522 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3524 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3528 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3530 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3534 * netdev_queue_set_dql_min_limit - set dql minimum limit
3535 * @dev_queue: pointer to transmit queue
3536 * @min_limit: dql minimum limit
3538 * Forces xmit_more() to return true until the minimum threshold
3539 * defined by @min_limit is reached (or until the tx queue is
3540 * empty). Warning: to be use with care, misuse will impact the
3543 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3544 unsigned int min_limit)
3547 dev_queue->dql.min_limit = min_limit;
3552 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3553 * @dev_queue: pointer to transmit queue
3555 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3556 * to give appropriate hint to the CPU.
3558 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3561 prefetchw(&dev_queue->dql.num_queued);
3566 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3567 * @dev_queue: pointer to transmit queue
3569 * BQL enabled drivers might use this helper in their TX completion path,
3570 * to give appropriate hint to the CPU.
3572 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3575 prefetchw(&dev_queue->dql.limit);
3579 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3583 dql_queued(&dev_queue->dql, bytes);
3585 if (likely(dql_avail(&dev_queue->dql) >= 0))
3588 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3591 * The XOFF flag must be set before checking the dql_avail below,
3592 * because in netdev_tx_completed_queue we update the dql_completed
3593 * before checking the XOFF flag.
3597 /* check again in case another CPU has just made room avail */
3598 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3599 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3603 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3604 * that they should not test BQL status themselves.
3605 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3607 * Returns true if the doorbell must be used to kick the NIC.
3609 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3615 dql_queued(&dev_queue->dql, bytes);
3617 return netif_tx_queue_stopped(dev_queue);
3619 netdev_tx_sent_queue(dev_queue, bytes);
3624 * netdev_sent_queue - report the number of bytes queued to hardware
3625 * @dev: network device
3626 * @bytes: number of bytes queued to the hardware device queue
3628 * Report the number of bytes queued for sending/completion to the network
3629 * device hardware queue. @bytes should be a good approximation and should
3630 * exactly match netdev_completed_queue() @bytes
3632 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3634 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3637 static inline bool __netdev_sent_queue(struct net_device *dev,
3641 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3645 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3646 unsigned int pkts, unsigned int bytes)
3649 if (unlikely(!bytes))
3652 dql_completed(&dev_queue->dql, bytes);
3655 * Without the memory barrier there is a small possiblity that
3656 * netdev_tx_sent_queue will miss the update and cause the queue to
3657 * be stopped forever
3661 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3664 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3665 netif_schedule_queue(dev_queue);
3670 * netdev_completed_queue - report bytes and packets completed by device
3671 * @dev: network device
3672 * @pkts: actual number of packets sent over the medium
3673 * @bytes: actual number of bytes sent over the medium
3675 * Report the number of bytes and packets transmitted by the network device
3676 * hardware queue over the physical medium, @bytes must exactly match the
3677 * @bytes amount passed to netdev_sent_queue()
3679 static inline void netdev_completed_queue(struct net_device *dev,
3680 unsigned int pkts, unsigned int bytes)
3682 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3685 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3688 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3694 * netdev_reset_queue - reset the packets and bytes count of a network device
3695 * @dev_queue: network device
3697 * Reset the bytes and packet count of a network device and clear the
3698 * software flow control OFF bit for this network device
3700 static inline void netdev_reset_queue(struct net_device *dev_queue)
3702 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3706 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3707 * @dev: network device
3708 * @queue_index: given tx queue index
3710 * Returns 0 if given tx queue index >= number of device tx queues,
3711 * otherwise returns the originally passed tx queue index.
3713 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3715 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3716 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3717 dev->name, queue_index,
3718 dev->real_num_tx_queues);
3726 * netif_running - test if up
3727 * @dev: network device
3729 * Test if the device has been brought up.
3731 static inline bool netif_running(const struct net_device *dev)
3733 return test_bit(__LINK_STATE_START, &dev->state);
3737 * Routines to manage the subqueues on a device. We only need start,
3738 * stop, and a check if it's stopped. All other device management is
3739 * done at the overall netdevice level.
3740 * Also test the device if we're multiqueue.
3744 * netif_start_subqueue - allow sending packets on subqueue
3745 * @dev: network device
3746 * @queue_index: sub queue index
3748 * Start individual transmit queue of a device with multiple transmit queues.
3750 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3752 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3754 netif_tx_start_queue(txq);
3758 * netif_stop_subqueue - stop sending packets on subqueue
3759 * @dev: network device
3760 * @queue_index: sub queue index
3762 * Stop individual transmit queue of a device with multiple transmit queues.
3764 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3766 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3767 netif_tx_stop_queue(txq);
3771 * __netif_subqueue_stopped - test status of subqueue
3772 * @dev: network device
3773 * @queue_index: sub queue index
3775 * Check individual transmit queue of a device with multiple transmit queues.
3777 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3780 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3782 return netif_tx_queue_stopped(txq);
3786 * netif_subqueue_stopped - test status of subqueue
3787 * @dev: network device
3788 * @skb: sub queue buffer pointer
3790 * Check individual transmit queue of a device with multiple transmit queues.
3792 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3793 struct sk_buff *skb)
3795 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3799 * netif_wake_subqueue - allow sending packets on subqueue
3800 * @dev: network device
3801 * @queue_index: sub queue index
3803 * Resume individual transmit queue of a device with multiple transmit queues.
3805 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3807 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3809 netif_tx_wake_queue(txq);
3813 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3815 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3816 u16 index, enum xps_map_type type);
3819 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3820 * @j: CPU/Rx queue index
3821 * @mask: bitmask of all cpus/rx queues
3822 * @nr_bits: number of bits in the bitmask
3824 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3826 static inline bool netif_attr_test_mask(unsigned long j,
3827 const unsigned long *mask,
3828 unsigned int nr_bits)
3830 cpu_max_bits_warn(j, nr_bits);
3831 return test_bit(j, mask);
3835 * netif_attr_test_online - Test for online CPU/Rx queue
3836 * @j: CPU/Rx queue index
3837 * @online_mask: bitmask for CPUs/Rx queues that are online
3838 * @nr_bits: number of bits in the bitmask
3840 * Returns true if a CPU/Rx queue is online.
3842 static inline bool netif_attr_test_online(unsigned long j,
3843 const unsigned long *online_mask,
3844 unsigned int nr_bits)
3846 cpu_max_bits_warn(j, nr_bits);
3849 return test_bit(j, online_mask);
3851 return (j < nr_bits);
3855 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3856 * @n: CPU/Rx queue index
3857 * @srcp: the cpumask/Rx queue mask pointer
3858 * @nr_bits: number of bits in the bitmask
3860 * Returns >= nr_bits if no further CPUs/Rx queues set.
3862 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3863 unsigned int nr_bits)
3865 /* -1 is a legal arg here. */
3867 cpu_max_bits_warn(n, nr_bits);
3870 return find_next_bit(srcp, nr_bits, n + 1);
3876 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3877 * @n: CPU/Rx queue index
3878 * @src1p: the first CPUs/Rx queues mask pointer
3879 * @src2p: the second CPUs/Rx queues mask pointer
3880 * @nr_bits: number of bits in the bitmask
3882 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3884 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3885 const unsigned long *src2p,
3886 unsigned int nr_bits)
3888 /* -1 is a legal arg here. */
3890 cpu_max_bits_warn(n, nr_bits);
3893 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3895 return find_next_bit(src1p, nr_bits, n + 1);
3897 return find_next_bit(src2p, nr_bits, n + 1);
3902 static inline int netif_set_xps_queue(struct net_device *dev,
3903 const struct cpumask *mask,
3909 static inline int __netif_set_xps_queue(struct net_device *dev,
3910 const unsigned long *mask,
3911 u16 index, enum xps_map_type type)
3918 * netif_is_multiqueue - test if device has multiple transmit queues
3919 * @dev: network device
3921 * Check if device has multiple transmit queues
3923 static inline bool netif_is_multiqueue(const struct net_device *dev)
3925 return dev->num_tx_queues > 1;
3928 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3931 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3933 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3936 dev->real_num_rx_queues = rxqs;
3940 int netif_set_real_num_queues(struct net_device *dev,
3941 unsigned int txq, unsigned int rxq);
3943 static inline struct netdev_rx_queue *
3944 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3946 return dev->_rx + rxq;
3950 static inline unsigned int get_netdev_rx_queue_index(
3951 struct netdev_rx_queue *queue)
3953 struct net_device *dev = queue->dev;
3954 int index = queue - dev->_rx;
3956 BUG_ON(index >= dev->num_rx_queues);
3961 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3962 int netif_get_num_default_rss_queues(void);
3964 enum skb_free_reason {
3965 SKB_REASON_CONSUMED,
3969 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3970 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3973 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3974 * interrupt context or with hardware interrupts being disabled.
3975 * (in_hardirq() || irqs_disabled())
3977 * We provide four helpers that can be used in following contexts :
3979 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3980 * replacing kfree_skb(skb)
3982 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3983 * Typically used in place of consume_skb(skb) in TX completion path
3985 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3986 * replacing kfree_skb(skb)
3988 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3989 * and consumed a packet. Used in place of consume_skb(skb)
3991 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3993 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3996 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3998 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
4001 static inline void dev_kfree_skb_any(struct sk_buff *skb)
4003 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
4006 static inline void dev_consume_skb_any(struct sk_buff *skb)
4008 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
4011 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
4012 struct bpf_prog *xdp_prog);
4013 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
4014 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
4015 int netif_rx(struct sk_buff *skb);
4016 int netif_rx_ni(struct sk_buff *skb);
4017 int netif_rx_any_context(struct sk_buff *skb);
4018 int netif_receive_skb(struct sk_buff *skb);
4019 int netif_receive_skb_core(struct sk_buff *skb);
4020 void netif_receive_skb_list(struct list_head *head);
4021 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4022 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4023 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4024 gro_result_t napi_gro_frags(struct napi_struct *napi);
4025 struct packet_offload *gro_find_receive_by_type(__be16 type);
4026 struct packet_offload *gro_find_complete_by_type(__be16 type);
4028 static inline void napi_free_frags(struct napi_struct *napi)
4030 kfree_skb(napi->skb);
4034 bool netdev_is_rx_handler_busy(struct net_device *dev);
4035 int netdev_rx_handler_register(struct net_device *dev,
4036 rx_handler_func_t *rx_handler,
4037 void *rx_handler_data);
4038 void netdev_rx_handler_unregister(struct net_device *dev);
4040 bool dev_valid_name(const char *name);
4041 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
4043 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4045 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4046 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4047 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4048 void __user *data, bool *need_copyout);
4049 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4050 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4051 unsigned int dev_get_flags(const struct net_device *);
4052 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4053 struct netlink_ext_ack *extack);
4054 int dev_change_flags(struct net_device *dev, unsigned int flags,
4055 struct netlink_ext_ack *extack);
4056 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4057 unsigned int gchanges);
4058 int dev_change_name(struct net_device *, const char *);
4059 int dev_set_alias(struct net_device *, const char *, size_t);
4060 int dev_get_alias(const struct net_device *, char *, size_t);
4061 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4062 const char *pat, int new_ifindex);
4064 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4067 return __dev_change_net_namespace(dev, net, pat, 0);
4069 int __dev_set_mtu(struct net_device *, int);
4070 int dev_validate_mtu(struct net_device *dev, int mtu,
4071 struct netlink_ext_ack *extack);
4072 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4073 struct netlink_ext_ack *extack);
4074 int dev_set_mtu(struct net_device *, int);
4075 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4076 void dev_set_group(struct net_device *, int);
4077 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4078 struct netlink_ext_ack *extack);
4079 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4080 struct netlink_ext_ack *extack);
4081 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4082 struct netlink_ext_ack *extack);
4083 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4084 int dev_change_carrier(struct net_device *, bool new_carrier);
4085 int dev_get_phys_port_id(struct net_device *dev,
4086 struct netdev_phys_item_id *ppid);
4087 int dev_get_phys_port_name(struct net_device *dev,
4088 char *name, size_t len);
4089 int dev_get_port_parent_id(struct net_device *dev,
4090 struct netdev_phys_item_id *ppid, bool recurse);
4091 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4092 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4093 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4094 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4096 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4097 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4098 struct netdev_queue *txq, int *ret);
4100 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4101 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4102 int fd, int expected_fd, u32 flags);
4103 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4104 u8 dev_xdp_prog_count(struct net_device *dev);
4105 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4107 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4108 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4109 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4110 bool is_skb_forwardable(const struct net_device *dev,
4111 const struct sk_buff *skb);
4113 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4114 const struct sk_buff *skb,
4115 const bool check_mtu)
4117 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4120 if (!(dev->flags & IFF_UP))
4126 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4127 if (skb->len <= len)
4130 /* if TSO is enabled, we don't care about the length as the packet
4131 * could be forwarded without being segmented before
4133 if (skb_is_gso(skb))
4139 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4140 struct sk_buff *skb,
4141 const bool check_mtu)
4143 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4144 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4145 atomic_long_inc(&dev->rx_dropped);
4150 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4155 bool dev_nit_active(struct net_device *dev);
4156 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4158 extern int netdev_budget;
4159 extern unsigned int netdev_budget_usecs;
4161 /* Called by rtnetlink.c:rtnl_unlock() */
4162 void netdev_run_todo(void);
4165 * dev_put - release reference to device
4166 * @dev: network device
4168 * Release reference to device to allow it to be freed.
4170 static inline void dev_put(struct net_device *dev)
4173 #ifdef CONFIG_PCPU_DEV_REFCNT
4174 this_cpu_dec(*dev->pcpu_refcnt);
4176 refcount_dec(&dev->dev_refcnt);
4182 * dev_hold - get reference to device
4183 * @dev: network device
4185 * Hold reference to device to keep it from being freed.
4187 static inline void dev_hold(struct net_device *dev)
4190 #ifdef CONFIG_PCPU_DEV_REFCNT
4191 this_cpu_inc(*dev->pcpu_refcnt);
4193 refcount_inc(&dev->dev_refcnt);
4198 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4199 * and _off may be called from IRQ context, but it is caller
4200 * who is responsible for serialization of these calls.
4202 * The name carrier is inappropriate, these functions should really be
4203 * called netif_lowerlayer_*() because they represent the state of any
4204 * kind of lower layer not just hardware media.
4207 void linkwatch_init_dev(struct net_device *dev);
4208 void linkwatch_fire_event(struct net_device *dev);
4209 void linkwatch_forget_dev(struct net_device *dev);
4212 * netif_carrier_ok - test if carrier present
4213 * @dev: network device
4215 * Check if carrier is present on device
4217 static inline bool netif_carrier_ok(const struct net_device *dev)
4219 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4222 unsigned long dev_trans_start(struct net_device *dev);
4224 void __netdev_watchdog_up(struct net_device *dev);
4226 void netif_carrier_on(struct net_device *dev);
4227 void netif_carrier_off(struct net_device *dev);
4228 void netif_carrier_event(struct net_device *dev);
4231 * netif_dormant_on - mark device as dormant.
4232 * @dev: network device
4234 * Mark device as dormant (as per RFC2863).
4236 * The dormant state indicates that the relevant interface is not
4237 * actually in a condition to pass packets (i.e., it is not 'up') but is
4238 * in a "pending" state, waiting for some external event. For "on-
4239 * demand" interfaces, this new state identifies the situation where the
4240 * interface is waiting for events to place it in the up state.
4242 static inline void netif_dormant_on(struct net_device *dev)
4244 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4245 linkwatch_fire_event(dev);
4249 * netif_dormant_off - set device as not dormant.
4250 * @dev: network device
4252 * Device is not in dormant state.
4254 static inline void netif_dormant_off(struct net_device *dev)
4256 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4257 linkwatch_fire_event(dev);
4261 * netif_dormant - test if device is dormant
4262 * @dev: network device
4264 * Check if device is dormant.
4266 static inline bool netif_dormant(const struct net_device *dev)
4268 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4273 * netif_testing_on - mark device as under test.
4274 * @dev: network device
4276 * Mark device as under test (as per RFC2863).
4278 * The testing state indicates that some test(s) must be performed on
4279 * the interface. After completion, of the test, the interface state
4280 * will change to up, dormant, or down, as appropriate.
4282 static inline void netif_testing_on(struct net_device *dev)
4284 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4285 linkwatch_fire_event(dev);
4289 * netif_testing_off - set device as not under test.
4290 * @dev: network device
4292 * Device is not in testing state.
4294 static inline void netif_testing_off(struct net_device *dev)
4296 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4297 linkwatch_fire_event(dev);
4301 * netif_testing - test if device is under test
4302 * @dev: network device
4304 * Check if device is under test
4306 static inline bool netif_testing(const struct net_device *dev)
4308 return test_bit(__LINK_STATE_TESTING, &dev->state);
4313 * netif_oper_up - test if device is operational
4314 * @dev: network device
4316 * Check if carrier is operational
4318 static inline bool netif_oper_up(const struct net_device *dev)
4320 return (dev->operstate == IF_OPER_UP ||
4321 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4325 * netif_device_present - is device available or removed
4326 * @dev: network device
4328 * Check if device has not been removed from system.
4330 static inline bool netif_device_present(const struct net_device *dev)
4332 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4335 void netif_device_detach(struct net_device *dev);
4337 void netif_device_attach(struct net_device *dev);
4340 * Network interface message level settings
4345 NETIF_MSG_PROBE_BIT,
4347 NETIF_MSG_TIMER_BIT,
4348 NETIF_MSG_IFDOWN_BIT,
4350 NETIF_MSG_RX_ERR_BIT,
4351 NETIF_MSG_TX_ERR_BIT,
4352 NETIF_MSG_TX_QUEUED_BIT,
4354 NETIF_MSG_TX_DONE_BIT,
4355 NETIF_MSG_RX_STATUS_BIT,
4356 NETIF_MSG_PKTDATA_BIT,
4360 /* When you add a new bit above, update netif_msg_class_names array
4361 * in net/ethtool/common.c
4363 NETIF_MSG_CLASS_COUNT,
4365 /* Both ethtool_ops interface and internal driver implementation use u32 */
4366 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4368 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4369 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4371 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4372 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4373 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4374 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4375 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4376 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4377 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4378 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4379 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4380 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4381 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4382 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4383 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4384 #define NETIF_MSG_HW __NETIF_MSG(HW)
4385 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4387 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4388 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4389 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4390 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4391 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4392 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4393 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4394 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4395 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4396 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4397 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4398 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4399 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4400 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4401 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4403 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4406 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4407 return default_msg_enable_bits;
4408 if (debug_value == 0) /* no output */
4410 /* set low N bits */
4411 return (1U << debug_value) - 1;
4414 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4416 spin_lock(&txq->_xmit_lock);
4417 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4418 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4421 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4423 __acquire(&txq->_xmit_lock);
4427 static inline void __netif_tx_release(struct netdev_queue *txq)
4429 __release(&txq->_xmit_lock);
4432 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4434 spin_lock_bh(&txq->_xmit_lock);
4435 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4436 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4439 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4441 bool ok = spin_trylock(&txq->_xmit_lock);
4444 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4445 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4450 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4452 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4453 WRITE_ONCE(txq->xmit_lock_owner, -1);
4454 spin_unlock(&txq->_xmit_lock);
4457 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4459 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4460 WRITE_ONCE(txq->xmit_lock_owner, -1);
4461 spin_unlock_bh(&txq->_xmit_lock);
4464 static inline void txq_trans_update(struct netdev_queue *txq)
4466 if (txq->xmit_lock_owner != -1)
4467 txq->trans_start = jiffies;
4470 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4471 static inline void netif_trans_update(struct net_device *dev)
4473 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4475 if (txq->trans_start != jiffies)
4476 txq->trans_start = jiffies;
4480 * netif_tx_lock - grab network device transmit lock
4481 * @dev: network device
4483 * Get network device transmit lock
4485 static inline void netif_tx_lock(struct net_device *dev)
4490 spin_lock(&dev->tx_global_lock);
4491 cpu = smp_processor_id();
4492 for (i = 0; i < dev->num_tx_queues; i++) {
4493 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4495 /* We are the only thread of execution doing a
4496 * freeze, but we have to grab the _xmit_lock in
4497 * order to synchronize with threads which are in
4498 * the ->hard_start_xmit() handler and already
4499 * checked the frozen bit.
4501 __netif_tx_lock(txq, cpu);
4502 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4503 __netif_tx_unlock(txq);
4507 static inline void netif_tx_lock_bh(struct net_device *dev)
4513 static inline void netif_tx_unlock(struct net_device *dev)
4517 for (i = 0; i < dev->num_tx_queues; i++) {
4518 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4520 /* No need to grab the _xmit_lock here. If the
4521 * queue is not stopped for another reason, we
4524 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4525 netif_schedule_queue(txq);
4527 spin_unlock(&dev->tx_global_lock);
4530 static inline void netif_tx_unlock_bh(struct net_device *dev)
4532 netif_tx_unlock(dev);
4536 #define HARD_TX_LOCK(dev, txq, cpu) { \
4537 if ((dev->features & NETIF_F_LLTX) == 0) { \
4538 __netif_tx_lock(txq, cpu); \
4540 __netif_tx_acquire(txq); \
4544 #define HARD_TX_TRYLOCK(dev, txq) \
4545 (((dev->features & NETIF_F_LLTX) == 0) ? \
4546 __netif_tx_trylock(txq) : \
4547 __netif_tx_acquire(txq))
4549 #define HARD_TX_UNLOCK(dev, txq) { \
4550 if ((dev->features & NETIF_F_LLTX) == 0) { \
4551 __netif_tx_unlock(txq); \
4553 __netif_tx_release(txq); \
4557 static inline void netif_tx_disable(struct net_device *dev)
4563 cpu = smp_processor_id();
4564 spin_lock(&dev->tx_global_lock);
4565 for (i = 0; i < dev->num_tx_queues; i++) {
4566 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4568 __netif_tx_lock(txq, cpu);
4569 netif_tx_stop_queue(txq);
4570 __netif_tx_unlock(txq);
4572 spin_unlock(&dev->tx_global_lock);
4576 static inline void netif_addr_lock(struct net_device *dev)
4578 unsigned char nest_level = 0;
4580 #ifdef CONFIG_LOCKDEP
4581 nest_level = dev->nested_level;
4583 spin_lock_nested(&dev->addr_list_lock, nest_level);
4586 static inline void netif_addr_lock_bh(struct net_device *dev)
4588 unsigned char nest_level = 0;
4590 #ifdef CONFIG_LOCKDEP
4591 nest_level = dev->nested_level;
4594 spin_lock_nested(&dev->addr_list_lock, nest_level);
4597 static inline void netif_addr_unlock(struct net_device *dev)
4599 spin_unlock(&dev->addr_list_lock);
4602 static inline void netif_addr_unlock_bh(struct net_device *dev)
4604 spin_unlock_bh(&dev->addr_list_lock);
4608 * dev_addrs walker. Should be used only for read access. Call with
4609 * rcu_read_lock held.
4611 #define for_each_dev_addr(dev, ha) \
4612 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4614 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4616 void ether_setup(struct net_device *dev);
4618 /* Support for loadable net-drivers */
4619 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4620 unsigned char name_assign_type,
4621 void (*setup)(struct net_device *),
4622 unsigned int txqs, unsigned int rxqs);
4623 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4624 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4626 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4627 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4630 int register_netdev(struct net_device *dev);
4631 void unregister_netdev(struct net_device *dev);
4633 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4635 /* General hardware address lists handling functions */
4636 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4637 struct netdev_hw_addr_list *from_list, int addr_len);
4638 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4639 struct netdev_hw_addr_list *from_list, int addr_len);
4640 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4641 struct net_device *dev,
4642 int (*sync)(struct net_device *, const unsigned char *),
4643 int (*unsync)(struct net_device *,
4644 const unsigned char *));
4645 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4646 struct net_device *dev,
4647 int (*sync)(struct net_device *,
4648 const unsigned char *, int),
4649 int (*unsync)(struct net_device *,
4650 const unsigned char *, int));
4651 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4652 struct net_device *dev,
4653 int (*unsync)(struct net_device *,
4654 const unsigned char *, int));
4655 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4656 struct net_device *dev,
4657 int (*unsync)(struct net_device *,
4658 const unsigned char *));
4659 void __hw_addr_init(struct netdev_hw_addr_list *list);
4661 /* Functions used for device addresses handling */
4663 __dev_addr_set(struct net_device *dev, const u8 *addr, size_t len)
4665 memcpy(dev->dev_addr, addr, len);
4668 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4670 __dev_addr_set(dev, addr, dev->addr_len);
4674 dev_addr_mod(struct net_device *dev, unsigned int offset,
4675 const u8 *addr, size_t len)
4677 memcpy(&dev->dev_addr[offset], addr, len);
4680 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4681 unsigned char addr_type);
4682 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4683 unsigned char addr_type);
4684 void dev_addr_flush(struct net_device *dev);
4685 int dev_addr_init(struct net_device *dev);
4687 /* Functions used for unicast addresses handling */
4688 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4689 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4690 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4691 int dev_uc_sync(struct net_device *to, struct net_device *from);
4692 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4693 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4694 void dev_uc_flush(struct net_device *dev);
4695 void dev_uc_init(struct net_device *dev);
4698 * __dev_uc_sync - Synchonize device's unicast list
4699 * @dev: device to sync
4700 * @sync: function to call if address should be added
4701 * @unsync: function to call if address should be removed
4703 * Add newly added addresses to the interface, and release
4704 * addresses that have been deleted.
4706 static inline int __dev_uc_sync(struct net_device *dev,
4707 int (*sync)(struct net_device *,
4708 const unsigned char *),
4709 int (*unsync)(struct net_device *,
4710 const unsigned char *))
4712 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4716 * __dev_uc_unsync - Remove synchronized addresses from device
4717 * @dev: device to sync
4718 * @unsync: function to call if address should be removed
4720 * Remove all addresses that were added to the device by dev_uc_sync().
4722 static inline void __dev_uc_unsync(struct net_device *dev,
4723 int (*unsync)(struct net_device *,
4724 const unsigned char *))
4726 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4729 /* Functions used for multicast addresses handling */
4730 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4731 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4732 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4733 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4734 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4735 int dev_mc_sync(struct net_device *to, struct net_device *from);
4736 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4737 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4738 void dev_mc_flush(struct net_device *dev);
4739 void dev_mc_init(struct net_device *dev);
4742 * __dev_mc_sync - Synchonize device's multicast list
4743 * @dev: device to sync
4744 * @sync: function to call if address should be added
4745 * @unsync: function to call if address should be removed
4747 * Add newly added addresses to the interface, and release
4748 * addresses that have been deleted.
4750 static inline int __dev_mc_sync(struct net_device *dev,
4751 int (*sync)(struct net_device *,
4752 const unsigned char *),
4753 int (*unsync)(struct net_device *,
4754 const unsigned char *))
4756 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4760 * __dev_mc_unsync - Remove synchronized addresses from device
4761 * @dev: device to sync
4762 * @unsync: function to call if address should be removed
4764 * Remove all addresses that were added to the device by dev_mc_sync().
4766 static inline void __dev_mc_unsync(struct net_device *dev,
4767 int (*unsync)(struct net_device *,
4768 const unsigned char *))
4770 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4773 /* Functions used for secondary unicast and multicast support */
4774 void dev_set_rx_mode(struct net_device *dev);
4775 void __dev_set_rx_mode(struct net_device *dev);
4776 int dev_set_promiscuity(struct net_device *dev, int inc);
4777 int dev_set_allmulti(struct net_device *dev, int inc);
4778 void netdev_state_change(struct net_device *dev);
4779 void __netdev_notify_peers(struct net_device *dev);
4780 void netdev_notify_peers(struct net_device *dev);
4781 void netdev_features_change(struct net_device *dev);
4782 /* Load a device via the kmod */
4783 void dev_load(struct net *net, const char *name);
4784 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4785 struct rtnl_link_stats64 *storage);
4786 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4787 const struct net_device_stats *netdev_stats);
4788 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4789 const struct pcpu_sw_netstats __percpu *netstats);
4790 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4792 extern int netdev_max_backlog;
4793 extern int netdev_tstamp_prequeue;
4794 extern int netdev_unregister_timeout_secs;
4795 extern int weight_p;
4796 extern int dev_weight_rx_bias;
4797 extern int dev_weight_tx_bias;
4798 extern int dev_rx_weight;
4799 extern int dev_tx_weight;
4800 extern int gro_normal_batch;
4803 NESTED_SYNC_IMM_BIT,
4804 NESTED_SYNC_TODO_BIT,
4807 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4808 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4810 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4811 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4813 struct netdev_nested_priv {
4814 unsigned char flags;
4818 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4819 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4820 struct list_head **iter);
4821 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4822 struct list_head **iter);
4824 #ifdef CONFIG_LOCKDEP
4825 static LIST_HEAD(net_unlink_list);
4827 static inline void net_unlink_todo(struct net_device *dev)
4829 if (list_empty(&dev->unlink_list))
4830 list_add_tail(&dev->unlink_list, &net_unlink_list);
4834 /* iterate through upper list, must be called under RCU read lock */
4835 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4836 for (iter = &(dev)->adj_list.upper, \
4837 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4839 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4841 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4842 int (*fn)(struct net_device *upper_dev,
4843 struct netdev_nested_priv *priv),
4844 struct netdev_nested_priv *priv);
4846 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4847 struct net_device *upper_dev);
4849 bool netdev_has_any_upper_dev(struct net_device *dev);
4851 void *netdev_lower_get_next_private(struct net_device *dev,
4852 struct list_head **iter);
4853 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4854 struct list_head **iter);
4856 #define netdev_for_each_lower_private(dev, priv, iter) \
4857 for (iter = (dev)->adj_list.lower.next, \
4858 priv = netdev_lower_get_next_private(dev, &(iter)); \
4860 priv = netdev_lower_get_next_private(dev, &(iter)))
4862 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4863 for (iter = &(dev)->adj_list.lower, \
4864 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4866 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4868 void *netdev_lower_get_next(struct net_device *dev,
4869 struct list_head **iter);
4871 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4872 for (iter = (dev)->adj_list.lower.next, \
4873 ldev = netdev_lower_get_next(dev, &(iter)); \
4875 ldev = netdev_lower_get_next(dev, &(iter)))
4877 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4878 struct list_head **iter);
4879 int netdev_walk_all_lower_dev(struct net_device *dev,
4880 int (*fn)(struct net_device *lower_dev,
4881 struct netdev_nested_priv *priv),
4882 struct netdev_nested_priv *priv);
4883 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4884 int (*fn)(struct net_device *lower_dev,
4885 struct netdev_nested_priv *priv),
4886 struct netdev_nested_priv *priv);
4888 void *netdev_adjacent_get_private(struct list_head *adj_list);
4889 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4890 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4891 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4892 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4893 struct netlink_ext_ack *extack);
4894 int netdev_master_upper_dev_link(struct net_device *dev,
4895 struct net_device *upper_dev,
4896 void *upper_priv, void *upper_info,
4897 struct netlink_ext_ack *extack);
4898 void netdev_upper_dev_unlink(struct net_device *dev,
4899 struct net_device *upper_dev);
4900 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4901 struct net_device *new_dev,
4902 struct net_device *dev,
4903 struct netlink_ext_ack *extack);
4904 void netdev_adjacent_change_commit(struct net_device *old_dev,
4905 struct net_device *new_dev,
4906 struct net_device *dev);
4907 void netdev_adjacent_change_abort(struct net_device *old_dev,
4908 struct net_device *new_dev,
4909 struct net_device *dev);
4910 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4911 void *netdev_lower_dev_get_private(struct net_device *dev,
4912 struct net_device *lower_dev);
4913 void netdev_lower_state_changed(struct net_device *lower_dev,
4914 void *lower_state_info);
4916 /* RSS keys are 40 or 52 bytes long */
4917 #define NETDEV_RSS_KEY_LEN 52
4918 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4919 void netdev_rss_key_fill(void *buffer, size_t len);
4921 int skb_checksum_help(struct sk_buff *skb);
4922 int skb_crc32c_csum_help(struct sk_buff *skb);
4923 int skb_csum_hwoffload_help(struct sk_buff *skb,
4924 const netdev_features_t features);
4926 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4927 netdev_features_t features, bool tx_path);
4928 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4929 netdev_features_t features);
4931 struct netdev_bonding_info {
4936 struct netdev_notifier_bonding_info {
4937 struct netdev_notifier_info info; /* must be first */
4938 struct netdev_bonding_info bonding_info;
4941 void netdev_bonding_info_change(struct net_device *dev,
4942 struct netdev_bonding_info *bonding_info);
4944 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4945 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4947 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4954 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4956 return __skb_gso_segment(skb, features, true);
4958 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4960 static inline bool can_checksum_protocol(netdev_features_t features,
4963 if (protocol == htons(ETH_P_FCOE))
4964 return !!(features & NETIF_F_FCOE_CRC);
4966 /* Assume this is an IP checksum (not SCTP CRC) */
4968 if (features & NETIF_F_HW_CSUM) {
4969 /* Can checksum everything */
4974 case htons(ETH_P_IP):
4975 return !!(features & NETIF_F_IP_CSUM);
4976 case htons(ETH_P_IPV6):
4977 return !!(features & NETIF_F_IPV6_CSUM);
4984 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4986 static inline void netdev_rx_csum_fault(struct net_device *dev,
4987 struct sk_buff *skb)
4991 /* rx skb timestamps */
4992 void net_enable_timestamp(void);
4993 void net_disable_timestamp(void);
4995 #ifdef CONFIG_PROC_FS
4996 int __init dev_proc_init(void);
4998 #define dev_proc_init() 0
5001 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
5002 struct sk_buff *skb, struct net_device *dev,
5005 __this_cpu_write(softnet_data.xmit.more, more);
5006 return ops->ndo_start_xmit(skb, dev);
5009 static inline bool netdev_xmit_more(void)
5011 return __this_cpu_read(softnet_data.xmit.more);
5014 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
5015 struct netdev_queue *txq, bool more)
5017 const struct net_device_ops *ops = dev->netdev_ops;
5020 rc = __netdev_start_xmit(ops, skb, dev, more);
5021 if (rc == NETDEV_TX_OK)
5022 txq_trans_update(txq);
5027 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5029 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5032 extern const struct kobj_ns_type_operations net_ns_type_operations;
5034 const char *netdev_drivername(const struct net_device *dev);
5036 void linkwatch_run_queue(void);
5038 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5039 netdev_features_t f2)
5041 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5042 if (f1 & NETIF_F_HW_CSUM)
5043 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5045 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5051 static inline netdev_features_t netdev_get_wanted_features(
5052 struct net_device *dev)
5054 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5056 netdev_features_t netdev_increment_features(netdev_features_t all,
5057 netdev_features_t one, netdev_features_t mask);
5059 /* Allow TSO being used on stacked device :
5060 * Performing the GSO segmentation before last device
5061 * is a performance improvement.
5063 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5064 netdev_features_t mask)
5066 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5069 int __netdev_update_features(struct net_device *dev);
5070 void netdev_update_features(struct net_device *dev);
5071 void netdev_change_features(struct net_device *dev);
5073 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5074 struct net_device *dev);
5076 netdev_features_t passthru_features_check(struct sk_buff *skb,
5077 struct net_device *dev,
5078 netdev_features_t features);
5079 netdev_features_t netif_skb_features(struct sk_buff *skb);
5081 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5083 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5085 /* check flags correspondence */
5086 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5087 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5088 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5089 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5090 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5091 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5092 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5093 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5094 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5095 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5096 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5097 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5098 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5099 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5100 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5101 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5102 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5103 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5104 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5106 return (features & feature) == feature;
5109 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5111 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5112 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5115 static inline bool netif_needs_gso(struct sk_buff *skb,
5116 netdev_features_t features)
5118 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5119 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5120 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5123 static inline void netif_set_gso_max_size(struct net_device *dev,
5126 dev->gso_max_size = size;
5129 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5130 int pulled_hlen, u16 mac_offset,
5133 skb->protocol = protocol;
5134 skb->encapsulation = 1;
5135 skb_push(skb, pulled_hlen);
5136 skb_reset_transport_header(skb);
5137 skb->mac_header = mac_offset;
5138 skb->network_header = skb->mac_header + mac_len;
5139 skb->mac_len = mac_len;
5142 static inline bool netif_is_macsec(const struct net_device *dev)
5144 return dev->priv_flags & IFF_MACSEC;
5147 static inline bool netif_is_macvlan(const struct net_device *dev)
5149 return dev->priv_flags & IFF_MACVLAN;
5152 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5154 return dev->priv_flags & IFF_MACVLAN_PORT;
5157 static inline bool netif_is_bond_master(const struct net_device *dev)
5159 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5162 static inline bool netif_is_bond_slave(const struct net_device *dev)
5164 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5167 static inline bool netif_supports_nofcs(struct net_device *dev)
5169 return dev->priv_flags & IFF_SUPP_NOFCS;
5172 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5174 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5177 static inline bool netif_is_l3_master(const struct net_device *dev)
5179 return dev->priv_flags & IFF_L3MDEV_MASTER;
5182 static inline bool netif_is_l3_slave(const struct net_device *dev)
5184 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5187 static inline bool netif_is_bridge_master(const struct net_device *dev)
5189 return dev->priv_flags & IFF_EBRIDGE;
5192 static inline bool netif_is_bridge_port(const struct net_device *dev)
5194 return dev->priv_flags & IFF_BRIDGE_PORT;
5197 static inline bool netif_is_ovs_master(const struct net_device *dev)
5199 return dev->priv_flags & IFF_OPENVSWITCH;
5202 static inline bool netif_is_ovs_port(const struct net_device *dev)
5204 return dev->priv_flags & IFF_OVS_DATAPATH;
5207 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5209 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5212 static inline bool netif_is_team_master(const struct net_device *dev)
5214 return dev->priv_flags & IFF_TEAM;
5217 static inline bool netif_is_team_port(const struct net_device *dev)
5219 return dev->priv_flags & IFF_TEAM_PORT;
5222 static inline bool netif_is_lag_master(const struct net_device *dev)
5224 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5227 static inline bool netif_is_lag_port(const struct net_device *dev)
5229 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5232 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5234 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5237 static inline bool netif_is_failover(const struct net_device *dev)
5239 return dev->priv_flags & IFF_FAILOVER;
5242 static inline bool netif_is_failover_slave(const struct net_device *dev)
5244 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5247 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5248 static inline void netif_keep_dst(struct net_device *dev)
5250 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5253 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5254 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5256 /* TODO: reserve and use an additional IFF bit, if we get more users */
5257 return dev->priv_flags & IFF_MACSEC;
5260 extern struct pernet_operations __net_initdata loopback_net_ops;
5262 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5264 /* netdev_printk helpers, similar to dev_printk */
5266 static inline const char *netdev_name(const struct net_device *dev)
5268 if (!dev->name[0] || strchr(dev->name, '%'))
5269 return "(unnamed net_device)";
5273 static inline bool netdev_unregistering(const struct net_device *dev)
5275 return dev->reg_state == NETREG_UNREGISTERING;
5278 static inline const char *netdev_reg_state(const struct net_device *dev)
5280 switch (dev->reg_state) {
5281 case NETREG_UNINITIALIZED: return " (uninitialized)";
5282 case NETREG_REGISTERED: return "";
5283 case NETREG_UNREGISTERING: return " (unregistering)";
5284 case NETREG_UNREGISTERED: return " (unregistered)";
5285 case NETREG_RELEASED: return " (released)";
5286 case NETREG_DUMMY: return " (dummy)";
5289 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5290 return " (unknown)";
5293 __printf(3, 4) __cold
5294 void netdev_printk(const char *level, const struct net_device *dev,
5295 const char *format, ...);
5296 __printf(2, 3) __cold
5297 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5298 __printf(2, 3) __cold
5299 void netdev_alert(const struct net_device *dev, const char *format, ...);
5300 __printf(2, 3) __cold
5301 void netdev_crit(const struct net_device *dev, const char *format, ...);
5302 __printf(2, 3) __cold
5303 void netdev_err(const struct net_device *dev, const char *format, ...);
5304 __printf(2, 3) __cold
5305 void netdev_warn(const struct net_device *dev, const char *format, ...);
5306 __printf(2, 3) __cold
5307 void netdev_notice(const struct net_device *dev, const char *format, ...);
5308 __printf(2, 3) __cold
5309 void netdev_info(const struct net_device *dev, const char *format, ...);
5311 #define netdev_level_once(level, dev, fmt, ...) \
5313 static bool __print_once __read_mostly; \
5315 if (!__print_once) { \
5316 __print_once = true; \
5317 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5321 #define netdev_emerg_once(dev, fmt, ...) \
5322 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5323 #define netdev_alert_once(dev, fmt, ...) \
5324 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5325 #define netdev_crit_once(dev, fmt, ...) \
5326 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5327 #define netdev_err_once(dev, fmt, ...) \
5328 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5329 #define netdev_warn_once(dev, fmt, ...) \
5330 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5331 #define netdev_notice_once(dev, fmt, ...) \
5332 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5333 #define netdev_info_once(dev, fmt, ...) \
5334 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5336 #define MODULE_ALIAS_NETDEV(device) \
5337 MODULE_ALIAS("netdev-" device)
5339 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5340 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5341 #define netdev_dbg(__dev, format, args...) \
5343 dynamic_netdev_dbg(__dev, format, ##args); \
5345 #elif defined(DEBUG)
5346 #define netdev_dbg(__dev, format, args...) \
5347 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5349 #define netdev_dbg(__dev, format, args...) \
5352 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5356 #if defined(VERBOSE_DEBUG)
5357 #define netdev_vdbg netdev_dbg
5360 #define netdev_vdbg(dev, format, args...) \
5363 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5369 * netdev_WARN() acts like dev_printk(), but with the key difference
5370 * of using a WARN/WARN_ON to get the message out, including the
5371 * file/line information and a backtrace.
5373 #define netdev_WARN(dev, format, args...) \
5374 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5375 netdev_reg_state(dev), ##args)
5377 #define netdev_WARN_ONCE(dev, format, args...) \
5378 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5379 netdev_reg_state(dev), ##args)
5381 /* netif printk helpers, similar to netdev_printk */
5383 #define netif_printk(priv, type, level, dev, fmt, args...) \
5385 if (netif_msg_##type(priv)) \
5386 netdev_printk(level, (dev), fmt, ##args); \
5389 #define netif_level(level, priv, type, dev, fmt, args...) \
5391 if (netif_msg_##type(priv)) \
5392 netdev_##level(dev, fmt, ##args); \
5395 #define netif_emerg(priv, type, dev, fmt, args...) \
5396 netif_level(emerg, priv, type, dev, fmt, ##args)
5397 #define netif_alert(priv, type, dev, fmt, args...) \
5398 netif_level(alert, priv, type, dev, fmt, ##args)
5399 #define netif_crit(priv, type, dev, fmt, args...) \
5400 netif_level(crit, priv, type, dev, fmt, ##args)
5401 #define netif_err(priv, type, dev, fmt, args...) \
5402 netif_level(err, priv, type, dev, fmt, ##args)
5403 #define netif_warn(priv, type, dev, fmt, args...) \
5404 netif_level(warn, priv, type, dev, fmt, ##args)
5405 #define netif_notice(priv, type, dev, fmt, args...) \
5406 netif_level(notice, priv, type, dev, fmt, ##args)
5407 #define netif_info(priv, type, dev, fmt, args...) \
5408 netif_level(info, priv, type, dev, fmt, ##args)
5410 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5411 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5412 #define netif_dbg(priv, type, netdev, format, args...) \
5414 if (netif_msg_##type(priv)) \
5415 dynamic_netdev_dbg(netdev, format, ##args); \
5417 #elif defined(DEBUG)
5418 #define netif_dbg(priv, type, dev, format, args...) \
5419 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5421 #define netif_dbg(priv, type, dev, format, args...) \
5424 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5429 /* if @cond then downgrade to debug, else print at @level */
5430 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5433 netif_dbg(priv, type, netdev, fmt, ##args); \
5435 netif_ ## level(priv, type, netdev, fmt, ##args); \
5438 #if defined(VERBOSE_DEBUG)
5439 #define netif_vdbg netif_dbg
5441 #define netif_vdbg(priv, type, dev, format, args...) \
5444 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5450 * The list of packet types we will receive (as opposed to discard)
5451 * and the routines to invoke.
5453 * Why 16. Because with 16 the only overlap we get on a hash of the
5454 * low nibble of the protocol value is RARP/SNAP/X.25.
5468 #define PTYPE_HASH_SIZE (16)
5469 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5471 extern struct list_head ptype_all __read_mostly;
5472 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5474 extern struct net_device *blackhole_netdev;
5476 #endif /* _LINUX_NETDEVICE_H */