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 <linux/ethtool.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
56 struct ip_tunnel_parm;
57 struct macsec_context;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct udp_tunnel_nic_info;
69 struct udp_tunnel_nic;
73 void synchronize_net(void);
74 void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
77 /* Backlog congestion levels */
78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79 #define NET_RX_DROP 1 /* packet dropped */
81 #define MAX_NEST_DEV 8
84 * Transmit return codes: transmit return codes originate from three different
87 * - qdisc return codes
88 * - driver transmit return codes
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
100 /* qdisc ->enqueue() return codes. */
101 #define NET_XMIT_SUCCESS 0x00
102 #define NET_XMIT_DROP 0x01 /* skb dropped */
103 #define NET_XMIT_CN 0x02 /* congestion notification */
104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
112 /* Driver transmit return codes */
113 #define NETDEV_TX_MASK 0xf0
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
120 typedef enum netdev_tx netdev_tx_t;
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
126 static inline bool dev_xmit_complete(int rc)
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
134 if (likely(rc < NET_XMIT_MASK))
141 * Compute the worst-case header length according to the protocols
145 #if defined(CONFIG_HYPERV_NET)
146 # define LL_MAX_HEADER 128
147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
151 # define LL_MAX_HEADER 96
154 # define LL_MAX_HEADER 32
157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159 #define MAX_HEADER LL_MAX_HEADER
161 #define MAX_HEADER (LL_MAX_HEADER + 48)
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
169 #define NET_DEV_STAT(FIELD) \
171 unsigned long FIELD; \
172 atomic_long_t __##FIELD; \
175 struct net_device_stats {
176 NET_DEV_STAT(rx_packets);
177 NET_DEV_STAT(tx_packets);
178 NET_DEV_STAT(rx_bytes);
179 NET_DEV_STAT(tx_bytes);
180 NET_DEV_STAT(rx_errors);
181 NET_DEV_STAT(tx_errors);
182 NET_DEV_STAT(rx_dropped);
183 NET_DEV_STAT(tx_dropped);
184 NET_DEV_STAT(multicast);
185 NET_DEV_STAT(collisions);
186 NET_DEV_STAT(rx_length_errors);
187 NET_DEV_STAT(rx_over_errors);
188 NET_DEV_STAT(rx_crc_errors);
189 NET_DEV_STAT(rx_frame_errors);
190 NET_DEV_STAT(rx_fifo_errors);
191 NET_DEV_STAT(rx_missed_errors);
192 NET_DEV_STAT(tx_aborted_errors);
193 NET_DEV_STAT(tx_carrier_errors);
194 NET_DEV_STAT(tx_fifo_errors);
195 NET_DEV_STAT(tx_heartbeat_errors);
196 NET_DEV_STAT(tx_window_errors);
197 NET_DEV_STAT(rx_compressed);
198 NET_DEV_STAT(tx_compressed);
203 #include <linux/cache.h>
204 #include <linux/skbuff.h>
207 #include <linux/static_key.h>
208 extern struct static_key_false rps_needed;
209 extern struct static_key_false rfs_needed;
216 struct netdev_hw_addr {
217 struct list_head list;
218 unsigned char addr[MAX_ADDR_LEN];
220 #define NETDEV_HW_ADDR_T_LAN 1
221 #define NETDEV_HW_ADDR_T_SAN 2
222 #define NETDEV_HW_ADDR_T_UNICAST 3
223 #define NETDEV_HW_ADDR_T_MULTICAST 4
228 struct rcu_head rcu_head;
231 struct netdev_hw_addr_list {
232 struct list_head list;
236 #define netdev_hw_addr_list_count(l) ((l)->count)
237 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
238 #define netdev_hw_addr_list_for_each(ha, l) \
239 list_for_each_entry(ha, &(l)->list, list)
241 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
242 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
243 #define netdev_for_each_uc_addr(ha, dev) \
244 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
246 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
247 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
248 #define netdev_for_each_mc_addr(ha, dev) \
249 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
255 /* cached hardware header; allow for machine alignment needs. */
256 #define HH_DATA_MOD 16
257 #define HH_DATA_OFF(__len) \
258 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
259 #define HH_DATA_ALIGN(__len) \
260 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
261 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
264 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
266 * dev->hard_header_len ? (dev->hard_header_len +
267 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
269 * We could use other alignment values, but we must maintain the
270 * relationship HH alignment <= LL alignment.
272 #define LL_RESERVED_SPACE(dev) \
273 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
274 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
275 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
276 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
277 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
280 int (*create) (struct sk_buff *skb, struct net_device *dev,
281 unsigned short type, const void *daddr,
282 const void *saddr, unsigned int len);
283 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
284 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
285 void (*cache_update)(struct hh_cache *hh,
286 const struct net_device *dev,
287 const unsigned char *haddr);
288 bool (*validate)(const char *ll_header, unsigned int len);
289 __be16 (*parse_protocol)(const struct sk_buff *skb);
292 /* These flag bits are private to the generic network queueing
293 * layer; they may not be explicitly referenced by any other
297 enum netdev_state_t {
299 __LINK_STATE_PRESENT,
300 __LINK_STATE_NOCARRIER,
301 __LINK_STATE_LINKWATCH_PENDING,
302 __LINK_STATE_DORMANT,
303 __LINK_STATE_TESTING,
308 * This structure holds boot-time configured netdevice settings. They
309 * are then used in the device probing.
311 struct netdev_boot_setup {
315 #define NETDEV_BOOT_SETUP_MAX 8
317 int __init netdev_boot_setup(char *str);
320 struct list_head list;
325 * size of gro hash buckets, must less than bit number of
326 * napi_struct::gro_bitmask
328 #define GRO_HASH_BUCKETS 8
331 * Structure for NAPI scheduling similar to tasklet but with weighting
334 /* The poll_list must only be managed by the entity which
335 * changes the state of the NAPI_STATE_SCHED bit. This means
336 * whoever atomically sets that bit can add this napi_struct
337 * to the per-CPU poll_list, and whoever clears that bit
338 * can remove from the list right before clearing the bit.
340 struct list_head poll_list;
344 int defer_hard_irqs_count;
345 unsigned long gro_bitmask;
346 int (*poll)(struct napi_struct *, int);
347 #ifdef CONFIG_NETPOLL
350 struct net_device *dev;
351 struct gro_list gro_hash[GRO_HASH_BUCKETS];
353 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
354 int rx_count; /* length of rx_list */
355 struct hrtimer timer;
356 struct list_head dev_list;
357 struct hlist_node napi_hash_node;
358 unsigned int napi_id;
362 NAPI_STATE_SCHED, /* Poll is scheduled */
363 NAPI_STATE_MISSED, /* reschedule a napi */
364 NAPI_STATE_DISABLE, /* Disable pending */
365 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
366 NAPI_STATE_LISTED, /* NAPI added to system lists */
367 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
368 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
372 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
373 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
374 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
375 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
376 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
377 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
378 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
389 typedef enum gro_result gro_result_t;
392 * enum rx_handler_result - Possible return values for rx_handlers.
393 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
395 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
396 * case skb->dev was changed by rx_handler.
397 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
398 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
400 * rx_handlers are functions called from inside __netif_receive_skb(), to do
401 * special processing of the skb, prior to delivery to protocol handlers.
403 * Currently, a net_device can only have a single rx_handler registered. Trying
404 * to register a second rx_handler will return -EBUSY.
406 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
407 * To unregister a rx_handler on a net_device, use
408 * netdev_rx_handler_unregister().
410 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
413 * If the rx_handler consumed the skb in some way, it should return
414 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
415 * the skb to be delivered in some other way.
417 * If the rx_handler changed skb->dev, to divert the skb to another
418 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
419 * new device will be called if it exists.
421 * If the rx_handler decides the skb should be ignored, it should return
422 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
423 * are registered on exact device (ptype->dev == skb->dev).
425 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
426 * delivered, it should return RX_HANDLER_PASS.
428 * A device without a registered rx_handler will behave as if rx_handler
429 * returned RX_HANDLER_PASS.
432 enum rx_handler_result {
438 typedef enum rx_handler_result rx_handler_result_t;
439 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
441 void __napi_schedule(struct napi_struct *n);
442 void __napi_schedule_irqoff(struct napi_struct *n);
444 static inline bool napi_disable_pending(struct napi_struct *n)
446 return test_bit(NAPI_STATE_DISABLE, &n->state);
449 bool napi_schedule_prep(struct napi_struct *n);
452 * napi_schedule - schedule NAPI poll
455 * Schedule NAPI poll routine to be called if it is not already
458 static inline void napi_schedule(struct napi_struct *n)
460 if (napi_schedule_prep(n))
465 * napi_schedule_irqoff - schedule NAPI poll
468 * Variant of napi_schedule(), assuming hard irqs are masked.
470 static inline void napi_schedule_irqoff(struct napi_struct *n)
472 if (napi_schedule_prep(n))
473 __napi_schedule_irqoff(n);
476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
477 static inline bool napi_reschedule(struct napi_struct *napi)
479 if (napi_schedule_prep(napi)) {
480 __napi_schedule(napi);
486 bool napi_complete_done(struct napi_struct *n, int work_done);
488 * napi_complete - NAPI processing complete
491 * Mark NAPI processing as complete.
492 * Consider using napi_complete_done() instead.
493 * Return false if device should avoid rearming interrupts.
495 static inline bool napi_complete(struct napi_struct *n)
497 return napi_complete_done(n, 0);
501 * napi_disable - prevent NAPI from scheduling
504 * Stop NAPI from being scheduled on this context.
505 * Waits till any outstanding processing completes.
507 void napi_disable(struct napi_struct *n);
510 * napi_enable - enable NAPI scheduling
513 * Resume NAPI from being scheduled on this context.
514 * Must be paired with napi_disable.
516 static inline void napi_enable(struct napi_struct *n)
518 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
519 smp_mb__before_atomic();
520 clear_bit(NAPI_STATE_SCHED, &n->state);
521 clear_bit(NAPI_STATE_NPSVC, &n->state);
525 * napi_synchronize - wait until NAPI is not running
528 * Wait until NAPI is done being scheduled on this context.
529 * Waits till any outstanding processing completes but
530 * does not disable future activations.
532 static inline void napi_synchronize(const struct napi_struct *n)
534 if (IS_ENABLED(CONFIG_SMP))
535 while (test_bit(NAPI_STATE_SCHED, &n->state))
542 * napi_if_scheduled_mark_missed - if napi is running, set the
546 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
549 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
551 unsigned long val, new;
554 val = READ_ONCE(n->state);
555 if (val & NAPIF_STATE_DISABLE)
558 if (!(val & NAPIF_STATE_SCHED))
561 new = val | NAPIF_STATE_MISSED;
562 } while (cmpxchg(&n->state, val, new) != val);
567 enum netdev_queue_state_t {
568 __QUEUE_STATE_DRV_XOFF,
569 __QUEUE_STATE_STACK_XOFF,
570 __QUEUE_STATE_FROZEN,
573 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
574 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
575 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
577 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
578 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
580 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
584 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
585 * netif_tx_* functions below are used to manipulate this flag. The
586 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
587 * queue independently. The netif_xmit_*stopped functions below are called
588 * to check if the queue has been stopped by the driver or stack (either
589 * of the XOFF bits are set in the state). Drivers should not need to call
590 * netif_xmit*stopped functions, they should only be using netif_tx_*.
593 struct netdev_queue {
597 struct net_device *dev;
598 struct Qdisc __rcu *qdisc;
599 struct Qdisc *qdisc_sleeping;
603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 unsigned long tx_maxrate;
608 * Number of TX timeouts for this queue
609 * (/sys/class/net/DEV/Q/trans_timeout)
611 unsigned long trans_timeout;
613 /* Subordinate device that the queue has been assigned to */
614 struct net_device *sb_dev;
615 #ifdef CONFIG_XDP_SOCKETS
616 struct xsk_buff_pool *pool;
621 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
624 * Time (in jiffies) of last Tx
626 unsigned long trans_start;
633 } ____cacheline_aligned_in_smp;
635 extern int sysctl_fb_tunnels_only_for_init_net;
636 extern int sysctl_devconf_inherit_init_net;
639 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
640 * == 1 : For initns only
643 static inline bool net_has_fallback_tunnels(const struct net *net)
645 #if IS_ENABLED(CONFIG_SYSCTL)
646 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
648 return !fb_tunnels_only_for_init_net ||
649 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
655 static inline int net_inherit_devconf(void)
657 #if IS_ENABLED(CONFIG_SYSCTL)
658 return READ_ONCE(sysctl_devconf_inherit_init_net);
664 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
666 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
673 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
675 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
682 * This structure holds an RPS map which can be of variable length. The
683 * map is an array of CPUs.
690 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
693 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
694 * tail pointer for that CPU's input queue at the time of last enqueue, and
695 * a hardware filter index.
697 struct rps_dev_flow {
700 unsigned int last_qtail;
702 #define RPS_NO_FILTER 0xffff
705 * The rps_dev_flow_table structure contains a table of flow mappings.
707 struct rps_dev_flow_table {
710 struct rps_dev_flow flows[];
712 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
713 ((_num) * sizeof(struct rps_dev_flow)))
716 * The rps_sock_flow_table contains mappings of flows to the last CPU
717 * on which they were processed by the application (set in recvmsg).
718 * Each entry is a 32bit value. Upper part is the high-order bits
719 * of flow hash, lower part is CPU number.
720 * rps_cpu_mask is used to partition the space, depending on number of
721 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
722 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
723 * meaning we use 32-6=26 bits for the hash.
725 struct rps_sock_flow_table {
728 u32 ents[] ____cacheline_aligned_in_smp;
730 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
732 #define RPS_NO_CPU 0xffff
734 extern u32 rps_cpu_mask;
735 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
737 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
741 unsigned int index = hash & table->mask;
742 u32 val = hash & ~rps_cpu_mask;
744 /* We only give a hint, preemption can change CPU under us */
745 val |= raw_smp_processor_id();
747 /* The following WRITE_ONCE() is paired with the READ_ONCE()
748 * here, and another one in get_rps_cpu().
750 if (READ_ONCE(table->ents[index]) != val)
751 WRITE_ONCE(table->ents[index], val);
755 #ifdef CONFIG_RFS_ACCEL
756 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
759 #endif /* CONFIG_RPS */
761 /* This structure contains an instance of an RX queue. */
762 struct netdev_rx_queue {
764 struct rps_map __rcu *rps_map;
765 struct rps_dev_flow_table __rcu *rps_flow_table;
768 struct net_device *dev;
769 struct xdp_rxq_info xdp_rxq;
770 #ifdef CONFIG_XDP_SOCKETS
771 struct xsk_buff_pool *pool;
773 } ____cacheline_aligned_in_smp;
776 * RX queue sysfs structures and functions.
778 struct rx_queue_attribute {
779 struct attribute attr;
780 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
781 ssize_t (*store)(struct netdev_rx_queue *queue,
782 const char *buf, size_t len);
787 * This structure holds an XPS map which can be of variable length. The
788 * map is an array of queues.
792 unsigned int alloc_len;
796 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
797 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
798 - sizeof(struct xps_map)) / sizeof(u16))
801 * This structure holds all XPS maps for device. Maps are indexed by CPU.
803 struct xps_dev_maps {
805 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
808 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
809 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
811 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
812 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
814 #endif /* CONFIG_XPS */
816 #define TC_MAX_QUEUE 16
817 #define TC_BITMASK 15
818 /* HW offloaded queuing disciplines txq count and offset maps */
819 struct netdev_tc_txq {
824 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
826 * This structure is to hold information about the device
827 * configured to run FCoE protocol stack.
829 struct netdev_fcoe_hbainfo {
830 char manufacturer[64];
831 char serial_number[64];
832 char hardware_version[64];
833 char driver_version[64];
834 char optionrom_version[64];
835 char firmware_version[64];
837 char model_description[256];
841 #define MAX_PHYS_ITEM_ID_LEN 32
843 /* This structure holds a unique identifier to identify some
844 * physical item (port for example) used by a netdevice.
846 struct netdev_phys_item_id {
847 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
848 unsigned char id_len;
851 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
852 struct netdev_phys_item_id *b)
854 return a->id_len == b->id_len &&
855 memcmp(a->id, b->id, a->id_len) == 0;
858 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
860 struct net_device *sb_dev);
863 TC_SETUP_QDISC_MQPRIO,
866 TC_SETUP_CLSMATCHALL,
876 TC_SETUP_QDISC_TAPRIO,
883 /* These structures hold the attributes of bpf state that are being passed
884 * to the netdevice through the bpf op.
886 enum bpf_netdev_command {
887 /* Set or clear a bpf program used in the earliest stages of packet
888 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
889 * is responsible for calling bpf_prog_put on any old progs that are
890 * stored. In case of error, the callee need not release the new prog
891 * reference, but on success it takes ownership and must bpf_prog_put
892 * when it is no longer used.
896 /* BPF program for offload callbacks, invoked at program load time. */
897 BPF_OFFLOAD_MAP_ALLOC,
898 BPF_OFFLOAD_MAP_FREE,
902 struct bpf_prog_offload_ops;
903 struct netlink_ext_ack;
905 struct xdp_dev_bulk_queue;
915 struct bpf_xdp_entity {
916 struct bpf_prog *prog;
917 struct bpf_xdp_link *link;
921 enum bpf_netdev_command command;
926 struct bpf_prog *prog;
927 struct netlink_ext_ack *extack;
929 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
931 struct bpf_offloaded_map *offmap;
933 /* XDP_SETUP_XSK_POOL */
935 struct xsk_buff_pool *pool;
941 /* Flags for ndo_xsk_wakeup. */
942 #define XDP_WAKEUP_RX (1 << 0)
943 #define XDP_WAKEUP_TX (1 << 1)
945 #ifdef CONFIG_XFRM_OFFLOAD
947 int (*xdo_dev_state_add) (struct xfrm_state *x);
948 void (*xdo_dev_state_delete) (struct xfrm_state *x);
949 void (*xdo_dev_state_free) (struct xfrm_state *x);
950 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
951 struct xfrm_state *x);
952 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
957 struct rcu_head rcuhead;
964 struct netdev_name_node {
965 struct hlist_node hlist;
966 struct list_head list;
967 struct net_device *dev;
971 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
972 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
974 struct netdev_net_notifier {
975 struct list_head list;
976 struct notifier_block *nb;
980 * This structure defines the management hooks for network devices.
981 * The following hooks can be defined; unless noted otherwise, they are
982 * optional and can be filled with a null pointer.
984 * int (*ndo_init)(struct net_device *dev);
985 * This function is called once when a network device is registered.
986 * The network device can use this for any late stage initialization
987 * or semantic validation. It can fail with an error code which will
988 * be propagated back to register_netdev.
990 * void (*ndo_uninit)(struct net_device *dev);
991 * This function is called when device is unregistered or when registration
992 * fails. It is not called if init fails.
994 * int (*ndo_open)(struct net_device *dev);
995 * This function is called when a network device transitions to the up
998 * int (*ndo_stop)(struct net_device *dev);
999 * This function is called when a network device transitions to the down
1002 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1003 * struct net_device *dev);
1004 * Called when a packet needs to be transmitted.
1005 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1006 * the queue before that can happen; it's for obsolete devices and weird
1007 * corner cases, but the stack really does a non-trivial amount
1008 * of useless work if you return NETDEV_TX_BUSY.
1009 * Required; cannot be NULL.
1011 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1012 * struct net_device *dev
1013 * netdev_features_t features);
1014 * Called by core transmit path to determine if device is capable of
1015 * performing offload operations on a given packet. This is to give
1016 * the device an opportunity to implement any restrictions that cannot
1017 * be otherwise expressed by feature flags. The check is called with
1018 * the set of features that the stack has calculated and it returns
1019 * those the driver believes to be appropriate.
1021 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1022 * struct net_device *sb_dev);
1023 * Called to decide which queue to use when device supports multiple
1026 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1027 * This function is called to allow device receiver to make
1028 * changes to configuration when multicast or promiscuous is enabled.
1030 * void (*ndo_set_rx_mode)(struct net_device *dev);
1031 * This function is called device changes address list filtering.
1032 * If driver handles unicast address filtering, it should set
1033 * IFF_UNICAST_FLT in its priv_flags.
1035 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1036 * This function is called when the Media Access Control address
1037 * needs to be changed. If this interface is not defined, the
1038 * MAC address can not be changed.
1040 * int (*ndo_validate_addr)(struct net_device *dev);
1041 * Test if Media Access Control address is valid for the device.
1043 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1044 * Called when a user requests an ioctl which can't be handled by
1045 * the generic interface code. If not defined ioctls return
1046 * not supported error code.
1048 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1049 * Used to set network devices bus interface parameters. This interface
1050 * is retained for legacy reasons; new devices should use the bus
1051 * interface (PCI) for low level management.
1053 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1054 * Called when a user wants to change the Maximum Transfer Unit
1057 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1058 * Callback used when the transmitter has not made any progress
1059 * for dev->watchdog ticks.
1061 * void (*ndo_get_stats64)(struct net_device *dev,
1062 * struct rtnl_link_stats64 *storage);
1063 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1064 * Called when a user wants to get the network device usage
1065 * statistics. Drivers must do one of the following:
1066 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1067 * rtnl_link_stats64 structure passed by the caller.
1068 * 2. Define @ndo_get_stats to update a net_device_stats structure
1069 * (which should normally be dev->stats) and return a pointer to
1070 * it. The structure may be changed asynchronously only if each
1071 * field is written atomically.
1072 * 3. Update dev->stats asynchronously and atomically, and define
1073 * neither operation.
1075 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1076 * Return true if this device supports offload stats of this attr_id.
1078 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1080 * Get statistics for offload operations by attr_id. Write it into the
1081 * attr_data pointer.
1083 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1084 * If device supports VLAN filtering this function is called when a
1085 * VLAN id is registered.
1087 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1088 * If device supports VLAN filtering this function is called when a
1089 * VLAN id is unregistered.
1091 * void (*ndo_poll_controller)(struct net_device *dev);
1093 * SR-IOV management functions.
1094 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1095 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1096 * u8 qos, __be16 proto);
1097 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1099 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1100 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1101 * int (*ndo_get_vf_config)(struct net_device *dev,
1102 * int vf, struct ifla_vf_info *ivf);
1103 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1104 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1105 * struct nlattr *port[]);
1107 * Enable or disable the VF ability to query its RSS Redirection Table and
1108 * Hash Key. This is needed since on some devices VF share this information
1109 * with PF and querying it may introduce a theoretical security risk.
1110 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1111 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1112 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1114 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1115 * This is always called from the stack with the rtnl lock held and netif
1116 * tx queues stopped. This allows the netdevice to perform queue
1117 * management safely.
1119 * Fiber Channel over Ethernet (FCoE) offload functions.
1120 * int (*ndo_fcoe_enable)(struct net_device *dev);
1121 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1122 * so the underlying device can perform whatever needed configuration or
1123 * initialization to support acceleration of FCoE traffic.
1125 * int (*ndo_fcoe_disable)(struct net_device *dev);
1126 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1127 * so the underlying device can perform whatever needed clean-ups to
1128 * stop supporting acceleration of FCoE traffic.
1130 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1131 * struct scatterlist *sgl, unsigned int sgc);
1132 * Called when the FCoE Initiator wants to initialize an I/O that
1133 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1134 * perform necessary setup and returns 1 to indicate the device is set up
1135 * successfully to perform DDP on this I/O, otherwise this returns 0.
1137 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1138 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1139 * indicated by the FC exchange id 'xid', so the underlying device can
1140 * clean up and reuse resources for later DDP requests.
1142 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1143 * struct scatterlist *sgl, unsigned int sgc);
1144 * Called when the FCoE Target wants to initialize an I/O that
1145 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1146 * perform necessary setup and returns 1 to indicate the device is set up
1147 * successfully to perform DDP on this I/O, otherwise this returns 0.
1149 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1150 * struct netdev_fcoe_hbainfo *hbainfo);
1151 * Called when the FCoE Protocol stack wants information on the underlying
1152 * device. This information is utilized by the FCoE protocol stack to
1153 * register attributes with Fiber Channel management service as per the
1154 * FC-GS Fabric Device Management Information(FDMI) specification.
1156 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1157 * Called when the underlying device wants to override default World Wide
1158 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1159 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1160 * protocol stack to use.
1163 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1164 * u16 rxq_index, u32 flow_id);
1165 * Set hardware filter for RFS. rxq_index is the target queue index;
1166 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1167 * Return the filter ID on success, or a negative error code.
1169 * Slave management functions (for bridge, bonding, etc).
1170 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1171 * Called to make another netdev an underling.
1173 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1174 * Called to release previously enslaved netdev.
1176 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1177 * struct sk_buff *skb,
1179 * Get the xmit slave of master device. If all_slaves is true, function
1180 * assume all the slaves can transmit.
1182 * Feature/offload setting functions.
1183 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1184 * netdev_features_t features);
1185 * Adjusts the requested feature flags according to device-specific
1186 * constraints, and returns the resulting flags. Must not modify
1189 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1190 * Called to update device configuration to new features. Passed
1191 * feature set might be less than what was returned by ndo_fix_features()).
1192 * Must return >0 or -errno if it changed dev->features itself.
1194 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1195 * struct net_device *dev,
1196 * const unsigned char *addr, u16 vid, u16 flags,
1197 * struct netlink_ext_ack *extack);
1198 * Adds an FDB entry to dev for addr.
1199 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1200 * struct net_device *dev,
1201 * const unsigned char *addr, u16 vid)
1202 * Deletes the FDB entry from dev coresponding to addr.
1203 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1204 * struct net_device *dev, struct net_device *filter_dev,
1206 * Used to add FDB entries to dump requests. Implementers should add
1207 * entries to skb and update idx with the number of entries.
1209 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1210 * u16 flags, struct netlink_ext_ack *extack)
1211 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1212 * struct net_device *dev, u32 filter_mask,
1214 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1217 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1218 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1219 * which do not represent real hardware may define this to allow their
1220 * userspace components to manage their virtual carrier state. Devices
1221 * that determine carrier state from physical hardware properties (eg
1222 * network cables) or protocol-dependent mechanisms (eg
1223 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1225 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1226 * struct netdev_phys_item_id *ppid);
1227 * Called to get ID of physical port of this device. If driver does
1228 * not implement this, it is assumed that the hw is not able to have
1229 * multiple net devices on single physical port.
1231 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1232 * struct netdev_phys_item_id *ppid)
1233 * Called to get the parent ID of the physical port of this device.
1235 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1236 * struct udp_tunnel_info *ti);
1237 * Called by UDP tunnel to notify a driver about the UDP port and socket
1238 * address family that a UDP tunnel is listnening to. It is called only
1239 * when a new port starts listening. The operation is protected by the
1242 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1243 * struct udp_tunnel_info *ti);
1244 * Called by UDP tunnel to notify the driver about a UDP port and socket
1245 * address family that the UDP tunnel is not listening to anymore. The
1246 * operation is protected by the RTNL.
1248 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1249 * struct net_device *dev)
1250 * Called by upper layer devices to accelerate switching or other
1251 * station functionality into hardware. 'pdev is the lowerdev
1252 * to use for the offload and 'dev' is the net device that will
1253 * back the offload. Returns a pointer to the private structure
1254 * the upper layer will maintain.
1255 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1256 * Called by upper layer device to delete the station created
1257 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1258 * the station and priv is the structure returned by the add
1260 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1261 * int queue_index, u32 maxrate);
1262 * Called when a user wants to set a max-rate limitation of specific
1264 * int (*ndo_get_iflink)(const struct net_device *dev);
1265 * Called to get the iflink value of this device.
1266 * void (*ndo_change_proto_down)(struct net_device *dev,
1268 * This function is used to pass protocol port error state information
1269 * to the switch driver. The switch driver can react to the proto_down
1270 * by doing a phys down on the associated switch port.
1271 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1272 * This function is used to get egress tunnel information for given skb.
1273 * This is useful for retrieving outer tunnel header parameters while
1275 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1276 * This function is used to specify the headroom that the skb must
1277 * consider when allocation skb during packet reception. Setting
1278 * appropriate rx headroom value allows avoiding skb head copy on
1279 * forward. Setting a negative value resets the rx headroom to the
1281 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1282 * This function is used to set or query state related to XDP on the
1283 * netdevice and manage BPF offload. See definition of
1284 * enum bpf_netdev_command for details.
1285 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1287 * This function is used to submit @n XDP packets for transmit on a
1288 * netdevice. Returns number of frames successfully transmitted, frames
1289 * that got dropped are freed/returned via xdp_return_frame().
1290 * Returns negative number, means general error invoking ndo, meaning
1291 * no frames were xmit'ed and core-caller will free all frames.
1292 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1293 * This function is used to wake up the softirq, ksoftirqd or kthread
1294 * responsible for sending and/or receiving packets on a specific
1295 * queue id bound to an AF_XDP socket. The flags field specifies if
1296 * only RX, only Tx, or both should be woken up using the flags
1297 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1298 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1299 * Get devlink port instance associated with a given netdev.
1300 * Called with a reference on the netdevice and devlink locks only,
1301 * rtnl_lock is not held.
1302 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1304 * Add, change, delete or get information on an IPv4 tunnel.
1305 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1306 * If a device is paired with a peer device, return the peer instance.
1307 * The caller must be under RCU read context.
1309 struct net_device_ops {
1310 int (*ndo_init)(struct net_device *dev);
1311 void (*ndo_uninit)(struct net_device *dev);
1312 int (*ndo_open)(struct net_device *dev);
1313 int (*ndo_stop)(struct net_device *dev);
1314 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1315 struct net_device *dev);
1316 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1317 struct net_device *dev,
1318 netdev_features_t features);
1319 u16 (*ndo_select_queue)(struct net_device *dev,
1320 struct sk_buff *skb,
1321 struct net_device *sb_dev);
1322 void (*ndo_change_rx_flags)(struct net_device *dev,
1324 void (*ndo_set_rx_mode)(struct net_device *dev);
1325 int (*ndo_set_mac_address)(struct net_device *dev,
1327 int (*ndo_validate_addr)(struct net_device *dev);
1328 int (*ndo_do_ioctl)(struct net_device *dev,
1329 struct ifreq *ifr, int cmd);
1330 int (*ndo_set_config)(struct net_device *dev,
1332 int (*ndo_change_mtu)(struct net_device *dev,
1334 int (*ndo_neigh_setup)(struct net_device *dev,
1335 struct neigh_parms *);
1336 void (*ndo_tx_timeout) (struct net_device *dev,
1337 unsigned int txqueue);
1339 void (*ndo_get_stats64)(struct net_device *dev,
1340 struct rtnl_link_stats64 *storage);
1341 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1342 int (*ndo_get_offload_stats)(int attr_id,
1343 const struct net_device *dev,
1345 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1347 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1348 __be16 proto, u16 vid);
1349 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1350 __be16 proto, u16 vid);
1351 #ifdef CONFIG_NET_POLL_CONTROLLER
1352 void (*ndo_poll_controller)(struct net_device *dev);
1353 int (*ndo_netpoll_setup)(struct net_device *dev,
1354 struct netpoll_info *info);
1355 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1357 int (*ndo_set_vf_mac)(struct net_device *dev,
1358 int queue, u8 *mac);
1359 int (*ndo_set_vf_vlan)(struct net_device *dev,
1360 int queue, u16 vlan,
1361 u8 qos, __be16 proto);
1362 int (*ndo_set_vf_rate)(struct net_device *dev,
1363 int vf, int min_tx_rate,
1365 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1366 int vf, bool setting);
1367 int (*ndo_set_vf_trust)(struct net_device *dev,
1368 int vf, bool setting);
1369 int (*ndo_get_vf_config)(struct net_device *dev,
1371 struct ifla_vf_info *ivf);
1372 int (*ndo_set_vf_link_state)(struct net_device *dev,
1373 int vf, int link_state);
1374 int (*ndo_get_vf_stats)(struct net_device *dev,
1376 struct ifla_vf_stats
1378 int (*ndo_set_vf_port)(struct net_device *dev,
1380 struct nlattr *port[]);
1381 int (*ndo_get_vf_port)(struct net_device *dev,
1382 int vf, struct sk_buff *skb);
1383 int (*ndo_get_vf_guid)(struct net_device *dev,
1385 struct ifla_vf_guid *node_guid,
1386 struct ifla_vf_guid *port_guid);
1387 int (*ndo_set_vf_guid)(struct net_device *dev,
1390 int (*ndo_set_vf_rss_query_en)(
1391 struct net_device *dev,
1392 int vf, bool setting);
1393 int (*ndo_setup_tc)(struct net_device *dev,
1394 enum tc_setup_type type,
1396 #if IS_ENABLED(CONFIG_FCOE)
1397 int (*ndo_fcoe_enable)(struct net_device *dev);
1398 int (*ndo_fcoe_disable)(struct net_device *dev);
1399 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1401 struct scatterlist *sgl,
1403 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1405 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1407 struct scatterlist *sgl,
1409 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1410 struct netdev_fcoe_hbainfo *hbainfo);
1413 #if IS_ENABLED(CONFIG_LIBFCOE)
1414 #define NETDEV_FCOE_WWNN 0
1415 #define NETDEV_FCOE_WWPN 1
1416 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1417 u64 *wwn, int type);
1420 #ifdef CONFIG_RFS_ACCEL
1421 int (*ndo_rx_flow_steer)(struct net_device *dev,
1422 const struct sk_buff *skb,
1426 int (*ndo_add_slave)(struct net_device *dev,
1427 struct net_device *slave_dev,
1428 struct netlink_ext_ack *extack);
1429 int (*ndo_del_slave)(struct net_device *dev,
1430 struct net_device *slave_dev);
1431 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1432 struct sk_buff *skb,
1434 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1435 netdev_features_t features);
1436 int (*ndo_set_features)(struct net_device *dev,
1437 netdev_features_t features);
1438 int (*ndo_neigh_construct)(struct net_device *dev,
1439 struct neighbour *n);
1440 void (*ndo_neigh_destroy)(struct net_device *dev,
1441 struct neighbour *n);
1443 int (*ndo_fdb_add)(struct ndmsg *ndm,
1444 struct nlattr *tb[],
1445 struct net_device *dev,
1446 const unsigned char *addr,
1449 struct netlink_ext_ack *extack);
1450 int (*ndo_fdb_del)(struct ndmsg *ndm,
1451 struct nlattr *tb[],
1452 struct net_device *dev,
1453 const unsigned char *addr,
1455 int (*ndo_fdb_dump)(struct sk_buff *skb,
1456 struct netlink_callback *cb,
1457 struct net_device *dev,
1458 struct net_device *filter_dev,
1460 int (*ndo_fdb_get)(struct sk_buff *skb,
1461 struct nlattr *tb[],
1462 struct net_device *dev,
1463 const unsigned char *addr,
1464 u16 vid, u32 portid, u32 seq,
1465 struct netlink_ext_ack *extack);
1466 int (*ndo_bridge_setlink)(struct net_device *dev,
1467 struct nlmsghdr *nlh,
1469 struct netlink_ext_ack *extack);
1470 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1472 struct net_device *dev,
1475 int (*ndo_bridge_dellink)(struct net_device *dev,
1476 struct nlmsghdr *nlh,
1478 int (*ndo_change_carrier)(struct net_device *dev,
1480 int (*ndo_get_phys_port_id)(struct net_device *dev,
1481 struct netdev_phys_item_id *ppid);
1482 int (*ndo_get_port_parent_id)(struct net_device *dev,
1483 struct netdev_phys_item_id *ppid);
1484 int (*ndo_get_phys_port_name)(struct net_device *dev,
1485 char *name, size_t len);
1486 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1487 struct udp_tunnel_info *ti);
1488 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1489 struct udp_tunnel_info *ti);
1490 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1491 struct net_device *dev);
1492 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1495 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1498 int (*ndo_get_iflink)(const struct net_device *dev);
1499 int (*ndo_change_proto_down)(struct net_device *dev,
1501 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1502 struct sk_buff *skb);
1503 void (*ndo_set_rx_headroom)(struct net_device *dev,
1504 int needed_headroom);
1505 int (*ndo_bpf)(struct net_device *dev,
1506 struct netdev_bpf *bpf);
1507 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1508 struct xdp_frame **xdp,
1510 int (*ndo_xsk_wakeup)(struct net_device *dev,
1511 u32 queue_id, u32 flags);
1512 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1513 int (*ndo_tunnel_ctl)(struct net_device *dev,
1514 struct ip_tunnel_parm *p, int cmd);
1515 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1519 * enum net_device_priv_flags - &struct net_device priv_flags
1521 * These are the &struct net_device, they are only set internally
1522 * by drivers and used in the kernel. These flags are invisible to
1523 * userspace; this means that the order of these flags can change
1524 * during any kernel release.
1526 * You should have a pretty good reason to be extending these flags.
1528 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1529 * @IFF_EBRIDGE: Ethernet bridging device
1530 * @IFF_BONDING: bonding master or slave
1531 * @IFF_ISATAP: ISATAP interface (RFC4214)
1532 * @IFF_WAN_HDLC: WAN HDLC device
1533 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1535 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1536 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1537 * @IFF_MACVLAN_PORT: device used as macvlan port
1538 * @IFF_BRIDGE_PORT: device used as bridge port
1539 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1540 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1541 * @IFF_UNICAST_FLT: Supports unicast filtering
1542 * @IFF_TEAM_PORT: device used as team port
1543 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1544 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1545 * change when it's running
1546 * @IFF_MACVLAN: Macvlan device
1547 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1548 * underlying stacked devices
1549 * @IFF_L3MDEV_MASTER: device is an L3 master device
1550 * @IFF_NO_QUEUE: device can run without qdisc attached
1551 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1552 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1553 * @IFF_TEAM: device is a team device
1554 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1555 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1556 * entity (i.e. the master device for bridged veth)
1557 * @IFF_MACSEC: device is a MACsec device
1558 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1559 * @IFF_FAILOVER: device is a failover master device
1560 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1561 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1562 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1564 enum netdev_priv_flags {
1565 IFF_802_1Q_VLAN = 1<<0,
1569 IFF_WAN_HDLC = 1<<4,
1570 IFF_XMIT_DST_RELEASE = 1<<5,
1571 IFF_DONT_BRIDGE = 1<<6,
1572 IFF_DISABLE_NETPOLL = 1<<7,
1573 IFF_MACVLAN_PORT = 1<<8,
1574 IFF_BRIDGE_PORT = 1<<9,
1575 IFF_OVS_DATAPATH = 1<<10,
1576 IFF_TX_SKB_SHARING = 1<<11,
1577 IFF_UNICAST_FLT = 1<<12,
1578 IFF_TEAM_PORT = 1<<13,
1579 IFF_SUPP_NOFCS = 1<<14,
1580 IFF_LIVE_ADDR_CHANGE = 1<<15,
1581 IFF_MACVLAN = 1<<16,
1582 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1583 IFF_L3MDEV_MASTER = 1<<18,
1584 IFF_NO_QUEUE = 1<<19,
1585 IFF_OPENVSWITCH = 1<<20,
1586 IFF_L3MDEV_SLAVE = 1<<21,
1588 IFF_RXFH_CONFIGURED = 1<<23,
1589 IFF_PHONY_HEADROOM = 1<<24,
1591 IFF_NO_RX_HANDLER = 1<<26,
1592 IFF_FAILOVER = 1<<27,
1593 IFF_FAILOVER_SLAVE = 1<<28,
1594 IFF_L3MDEV_RX_HANDLER = 1<<29,
1595 IFF_LIVE_RENAME_OK = 1<<30,
1598 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1599 #define IFF_EBRIDGE IFF_EBRIDGE
1600 #define IFF_BONDING IFF_BONDING
1601 #define IFF_ISATAP IFF_ISATAP
1602 #define IFF_WAN_HDLC IFF_WAN_HDLC
1603 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1604 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1605 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1606 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1607 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1608 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1609 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1610 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1611 #define IFF_TEAM_PORT IFF_TEAM_PORT
1612 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1613 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1614 #define IFF_MACVLAN IFF_MACVLAN
1615 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1616 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1617 #define IFF_NO_QUEUE IFF_NO_QUEUE
1618 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1619 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1620 #define IFF_TEAM IFF_TEAM
1621 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1622 #define IFF_MACSEC IFF_MACSEC
1623 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1624 #define IFF_FAILOVER IFF_FAILOVER
1625 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1626 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1627 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1629 /* Specifies the type of the struct net_device::ml_priv pointer */
1630 enum netdev_ml_priv_type {
1636 * struct net_device - The DEVICE structure.
1638 * Actually, this whole structure is a big mistake. It mixes I/O
1639 * data with strictly "high-level" data, and it has to know about
1640 * almost every data structure used in the INET module.
1642 * @name: This is the first field of the "visible" part of this structure
1643 * (i.e. as seen by users in the "Space.c" file). It is the name
1646 * @name_node: Name hashlist node
1647 * @ifalias: SNMP alias
1648 * @mem_end: Shared memory end
1649 * @mem_start: Shared memory start
1650 * @base_addr: Device I/O address
1651 * @irq: Device IRQ number
1653 * @state: Generic network queuing layer state, see netdev_state_t
1654 * @dev_list: The global list of network devices
1655 * @napi_list: List entry used for polling NAPI devices
1656 * @unreg_list: List entry when we are unregistering the
1657 * device; see the function unregister_netdev
1658 * @close_list: List entry used when we are closing the device
1659 * @ptype_all: Device-specific packet handlers for all protocols
1660 * @ptype_specific: Device-specific, protocol-specific packet handlers
1662 * @adj_list: Directly linked devices, like slaves for bonding
1663 * @features: Currently active device features
1664 * @hw_features: User-changeable features
1666 * @wanted_features: User-requested features
1667 * @vlan_features: Mask of features inheritable by VLAN devices
1669 * @hw_enc_features: Mask of features inherited by encapsulating devices
1670 * This field indicates what encapsulation
1671 * offloads the hardware is capable of doing,
1672 * and drivers will need to set them appropriately.
1674 * @mpls_features: Mask of features inheritable by MPLS
1675 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1677 * @ifindex: interface index
1678 * @group: The group the device belongs to
1680 * @stats: Statistics struct, which was left as a legacy, use
1681 * rtnl_link_stats64 instead
1683 * @rx_dropped: Dropped packets by core network,
1684 * do not use this in drivers
1685 * @tx_dropped: Dropped packets by core network,
1686 * do not use this in drivers
1687 * @rx_nohandler: nohandler dropped packets by core network on
1688 * inactive devices, do not use this in drivers
1689 * @carrier_up_count: Number of times the carrier has been up
1690 * @carrier_down_count: Number of times the carrier has been down
1692 * @wireless_handlers: List of functions to handle Wireless Extensions,
1694 * see <net/iw_handler.h> for details.
1695 * @wireless_data: Instance data managed by the core of wireless extensions
1697 * @netdev_ops: Includes several pointers to callbacks,
1698 * if one wants to override the ndo_*() functions
1699 * @ethtool_ops: Management operations
1700 * @l3mdev_ops: Layer 3 master device operations
1701 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1702 * discovery handling. Necessary for e.g. 6LoWPAN.
1703 * @xfrmdev_ops: Transformation offload operations
1704 * @tlsdev_ops: Transport Layer Security offload operations
1705 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1706 * of Layer 2 headers.
1708 * @flags: Interface flags (a la BSD)
1709 * @priv_flags: Like 'flags' but invisible to userspace,
1710 * see if.h for the definitions
1711 * @gflags: Global flags ( kept as legacy )
1712 * @padded: How much padding added by alloc_netdev()
1713 * @operstate: RFC2863 operstate
1714 * @link_mode: Mapping policy to operstate
1715 * @if_port: Selectable AUI, TP, ...
1717 * @mtu: Interface MTU value
1718 * @min_mtu: Interface Minimum MTU value
1719 * @max_mtu: Interface Maximum MTU value
1720 * @type: Interface hardware type
1721 * @hard_header_len: Maximum hardware header length.
1722 * @min_header_len: Minimum hardware header length
1724 * @needed_headroom: Extra headroom the hardware may need, but not in all
1725 * cases can this be guaranteed
1726 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1727 * cases can this be guaranteed. Some cases also use
1728 * LL_MAX_HEADER instead to allocate the skb
1730 * interface address info:
1732 * @perm_addr: Permanent hw address
1733 * @addr_assign_type: Hw address assignment type
1734 * @addr_len: Hardware address length
1735 * @upper_level: Maximum depth level of upper devices.
1736 * @lower_level: Maximum depth level of lower devices.
1737 * @neigh_priv_len: Used in neigh_alloc()
1738 * @dev_id: Used to differentiate devices that share
1739 * the same link layer address
1740 * @dev_port: Used to differentiate devices that share
1742 * @addr_list_lock: XXX: need comments on this one
1743 * @name_assign_type: network interface name assignment type
1744 * @uc_promisc: Counter that indicates promiscuous mode
1745 * has been enabled due to the need to listen to
1746 * additional unicast addresses in a device that
1747 * does not implement ndo_set_rx_mode()
1748 * @uc: unicast mac addresses
1749 * @mc: multicast mac addresses
1750 * @dev_addrs: list of device hw addresses
1751 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1752 * @promiscuity: Number of times the NIC is told to work in
1753 * promiscuous mode; if it becomes 0 the NIC will
1754 * exit promiscuous mode
1755 * @allmulti: Counter, enables or disables allmulticast mode
1757 * @vlan_info: VLAN info
1758 * @dsa_ptr: dsa specific data
1759 * @tipc_ptr: TIPC specific data
1760 * @atalk_ptr: AppleTalk link
1761 * @ip_ptr: IPv4 specific data
1762 * @ip6_ptr: IPv6 specific data
1763 * @ax25_ptr: AX.25 specific data
1764 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1765 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1767 * @mpls_ptr: mpls_dev struct pointer
1769 * @dev_addr: Hw address (before bcast,
1770 * because most packets are unicast)
1772 * @_rx: Array of RX queues
1773 * @num_rx_queues: Number of RX queues
1774 * allocated at register_netdev() time
1775 * @real_num_rx_queues: Number of RX queues currently active in device
1776 * @xdp_prog: XDP sockets filter program pointer
1777 * @gro_flush_timeout: timeout for GRO layer in NAPI
1778 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1779 * allow to avoid NIC hard IRQ, on busy queues.
1781 * @rx_handler: handler for received packets
1782 * @rx_handler_data: XXX: need comments on this one
1783 * @miniq_ingress: ingress/clsact qdisc specific data for
1784 * ingress processing
1785 * @ingress_queue: XXX: need comments on this one
1786 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1787 * @broadcast: hw bcast address
1789 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1790 * indexed by RX queue number. Assigned by driver.
1791 * This must only be set if the ndo_rx_flow_steer
1792 * operation is defined
1793 * @index_hlist: Device index hash chain
1795 * @_tx: Array of TX queues
1796 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1797 * @real_num_tx_queues: Number of TX queues currently active in device
1798 * @qdisc: Root qdisc from userspace point of view
1799 * @tx_queue_len: Max frames per queue allowed
1800 * @tx_global_lock: XXX: need comments on this one
1801 * @xdp_bulkq: XDP device bulk queue
1802 * @xps_cpus_map: all CPUs map for XPS device
1803 * @xps_rxqs_map: all RXQs map for XPS device
1805 * @xps_maps: XXX: need comments on this one
1806 * @miniq_egress: clsact qdisc specific data for
1808 * @qdisc_hash: qdisc hash table
1809 * @watchdog_timeo: Represents the timeout that is used by
1810 * the watchdog (see dev_watchdog())
1811 * @watchdog_timer: List of timers
1813 * @proto_down_reason: reason a netdev interface is held down
1814 * @pcpu_refcnt: Number of references to this device
1815 * @todo_list: Delayed register/unregister
1816 * @link_watch_list: XXX: need comments on this one
1818 * @reg_state: Register/unregister state machine
1819 * @dismantle: Device is going to be freed
1820 * @rtnl_link_state: This enum represents the phases of creating
1823 * @needs_free_netdev: Should unregister perform free_netdev?
1824 * @priv_destructor: Called from unregister
1825 * @npinfo: XXX: need comments on this one
1826 * @nd_net: Network namespace this network device is inside
1828 * @ml_priv: Mid-layer private
1829 * @ml_priv_type: Mid-layer private type
1830 * @lstats: Loopback statistics
1831 * @tstats: Tunnel statistics
1832 * @dstats: Dummy statistics
1833 * @vstats: Virtual ethernet statistics
1838 * @dev: Class/net/name entry
1839 * @sysfs_groups: Space for optional device, statistics and wireless
1842 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1843 * @rtnl_link_ops: Rtnl_link_ops
1845 * @gso_max_size: Maximum size of generic segmentation offload
1846 * @gso_max_segs: Maximum number of segments that can be passed to the
1849 * @dcbnl_ops: Data Center Bridging netlink ops
1850 * @num_tc: Number of traffic classes in the net device
1851 * @tc_to_txq: XXX: need comments on this one
1852 * @prio_tc_map: XXX: need comments on this one
1854 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1856 * @priomap: XXX: need comments on this one
1857 * @phydev: Physical device may attach itself
1858 * for hardware timestamping
1859 * @sfp_bus: attached &struct sfp_bus structure.
1861 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1862 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1864 * @proto_down: protocol port state information can be sent to the
1865 * switch driver and used to set the phys state of the
1868 * @wol_enabled: Wake-on-LAN is enabled
1870 * @net_notifier_list: List of per-net netdev notifier block
1871 * that follow this device when it is moved
1872 * to another network namespace.
1874 * @macsec_ops: MACsec offloading ops
1876 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1877 * offload capabilities of the device
1878 * @udp_tunnel_nic: UDP tunnel offload state
1879 * @xdp_state: stores info on attached XDP BPF programs
1881 * @nested_level: Used as as a parameter of spin_lock_nested() of
1882 * dev->addr_list_lock.
1883 * @unlink_list: As netif_addr_lock() can be called recursively,
1884 * keep a list of interfaces to be deleted.
1886 * FIXME: cleanup struct net_device such that network protocol info
1891 char name[IFNAMSIZ];
1892 struct netdev_name_node *name_node;
1893 struct dev_ifalias __rcu *ifalias;
1895 * I/O specific fields
1896 * FIXME: Merge these and struct ifmap into one
1898 unsigned long mem_end;
1899 unsigned long mem_start;
1900 unsigned long base_addr;
1904 * Some hardware also needs these fields (state,dev_list,
1905 * napi_list,unreg_list,close_list) but they are not
1906 * part of the usual set specified in Space.c.
1909 unsigned long state;
1911 struct list_head dev_list;
1912 struct list_head napi_list;
1913 struct list_head unreg_list;
1914 struct list_head close_list;
1915 struct list_head ptype_all;
1916 struct list_head ptype_specific;
1919 struct list_head upper;
1920 struct list_head lower;
1923 netdev_features_t features;
1924 netdev_features_t hw_features;
1925 netdev_features_t wanted_features;
1926 netdev_features_t vlan_features;
1927 netdev_features_t hw_enc_features;
1928 netdev_features_t mpls_features;
1929 netdev_features_t gso_partial_features;
1934 struct net_device_stats stats;
1936 atomic_long_t rx_dropped;
1937 atomic_long_t tx_dropped;
1938 atomic_long_t rx_nohandler;
1940 /* Stats to monitor link on/off, flapping */
1941 atomic_t carrier_up_count;
1942 atomic_t carrier_down_count;
1944 #ifdef CONFIG_WIRELESS_EXT
1945 const struct iw_handler_def *wireless_handlers;
1946 struct iw_public_data *wireless_data;
1948 const struct net_device_ops *netdev_ops;
1949 const struct ethtool_ops *ethtool_ops;
1950 #ifdef CONFIG_NET_L3_MASTER_DEV
1951 const struct l3mdev_ops *l3mdev_ops;
1953 #if IS_ENABLED(CONFIG_IPV6)
1954 const struct ndisc_ops *ndisc_ops;
1957 #ifdef CONFIG_XFRM_OFFLOAD
1958 const struct xfrmdev_ops *xfrmdev_ops;
1961 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1962 const struct tlsdev_ops *tlsdev_ops;
1965 const struct header_ops *header_ops;
1968 unsigned int priv_flags;
1970 unsigned short gflags;
1971 unsigned short padded;
1973 unsigned char operstate;
1974 unsigned char link_mode;
1976 unsigned char if_port;
1979 /* Note : dev->mtu is often read without holding a lock.
1980 * Writers usually hold RTNL.
1981 * It is recommended to use READ_ONCE() to annotate the reads,
1982 * and to use WRITE_ONCE() to annotate the writes.
1985 unsigned int min_mtu;
1986 unsigned int max_mtu;
1987 unsigned short type;
1988 unsigned short hard_header_len;
1989 unsigned char min_header_len;
1990 unsigned char name_assign_type;
1992 unsigned short needed_headroom;
1993 unsigned short needed_tailroom;
1995 /* Interface address info. */
1996 unsigned char perm_addr[MAX_ADDR_LEN];
1997 unsigned char addr_assign_type;
1998 unsigned char addr_len;
1999 unsigned char upper_level;
2000 unsigned char lower_level;
2002 unsigned short neigh_priv_len;
2003 unsigned short dev_id;
2004 unsigned short dev_port;
2005 spinlock_t addr_list_lock;
2007 struct netdev_hw_addr_list uc;
2008 struct netdev_hw_addr_list mc;
2009 struct netdev_hw_addr_list dev_addrs;
2012 struct kset *queues_kset;
2014 #ifdef CONFIG_LOCKDEP
2015 struct list_head unlink_list;
2017 unsigned int promiscuity;
2018 unsigned int allmulti;
2020 #ifdef CONFIG_LOCKDEP
2021 unsigned char nested_level;
2025 /* Protocol-specific pointers */
2027 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2028 struct vlan_info __rcu *vlan_info;
2030 #if IS_ENABLED(CONFIG_NET_DSA)
2031 struct dsa_port *dsa_ptr;
2033 #if IS_ENABLED(CONFIG_TIPC)
2034 struct tipc_bearer __rcu *tipc_ptr;
2036 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2039 struct in_device __rcu *ip_ptr;
2040 struct inet6_dev __rcu *ip6_ptr;
2041 #if IS_ENABLED(CONFIG_AX25)
2044 struct wireless_dev *ieee80211_ptr;
2045 struct wpan_dev *ieee802154_ptr;
2046 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2047 struct mpls_dev __rcu *mpls_ptr;
2051 * Cache lines mostly used on receive path (including eth_type_trans())
2053 /* Interface address info used in eth_type_trans() */
2054 unsigned char *dev_addr;
2056 struct netdev_rx_queue *_rx;
2057 unsigned int num_rx_queues;
2058 unsigned int real_num_rx_queues;
2060 struct bpf_prog __rcu *xdp_prog;
2061 unsigned long gro_flush_timeout;
2062 int napi_defer_hard_irqs;
2063 rx_handler_func_t __rcu *rx_handler;
2064 void __rcu *rx_handler_data;
2066 #ifdef CONFIG_NET_CLS_ACT
2067 struct mini_Qdisc __rcu *miniq_ingress;
2069 struct netdev_queue __rcu *ingress_queue;
2070 #ifdef CONFIG_NETFILTER_INGRESS
2071 struct nf_hook_entries __rcu *nf_hooks_ingress;
2074 unsigned char broadcast[MAX_ADDR_LEN];
2075 #ifdef CONFIG_RFS_ACCEL
2076 struct cpu_rmap *rx_cpu_rmap;
2078 struct hlist_node index_hlist;
2081 * Cache lines mostly used on transmit path
2083 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2084 unsigned int num_tx_queues;
2085 unsigned int real_num_tx_queues;
2086 struct Qdisc __rcu *qdisc;
2087 unsigned int tx_queue_len;
2088 spinlock_t tx_global_lock;
2090 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2093 struct xps_dev_maps __rcu *xps_cpus_map;
2094 struct xps_dev_maps __rcu *xps_rxqs_map;
2096 #ifdef CONFIG_NET_CLS_ACT
2097 struct mini_Qdisc __rcu *miniq_egress;
2100 #ifdef CONFIG_NET_SCHED
2101 DECLARE_HASHTABLE (qdisc_hash, 4);
2103 /* These may be needed for future network-power-down code. */
2104 struct timer_list watchdog_timer;
2107 u32 proto_down_reason;
2109 struct list_head todo_list;
2110 int __percpu *pcpu_refcnt;
2112 struct list_head link_watch_list;
2114 enum { NETREG_UNINITIALIZED=0,
2115 NETREG_REGISTERED, /* completed register_netdevice */
2116 NETREG_UNREGISTERING, /* called unregister_netdevice */
2117 NETREG_UNREGISTERED, /* completed unregister todo */
2118 NETREG_RELEASED, /* called free_netdev */
2119 NETREG_DUMMY, /* dummy device for NAPI poll */
2125 RTNL_LINK_INITIALIZED,
2126 RTNL_LINK_INITIALIZING,
2127 } rtnl_link_state:16;
2129 bool needs_free_netdev;
2130 void (*priv_destructor)(struct net_device *dev);
2132 #ifdef CONFIG_NETPOLL
2133 struct netpoll_info __rcu *npinfo;
2136 possible_net_t nd_net;
2138 /* mid-layer private */
2140 enum netdev_ml_priv_type ml_priv_type;
2143 struct pcpu_lstats __percpu *lstats;
2144 struct pcpu_sw_netstats __percpu *tstats;
2145 struct pcpu_dstats __percpu *dstats;
2148 #if IS_ENABLED(CONFIG_GARP)
2149 struct garp_port __rcu *garp_port;
2151 #if IS_ENABLED(CONFIG_MRP)
2152 struct mrp_port __rcu *mrp_port;
2156 const struct attribute_group *sysfs_groups[4];
2157 const struct attribute_group *sysfs_rx_queue_group;
2159 const struct rtnl_link_ops *rtnl_link_ops;
2161 /* for setting kernel sock attribute on TCP connection setup */
2162 #define GSO_MAX_SIZE 65536
2163 unsigned int gso_max_size;
2164 #define GSO_MAX_SEGS 65535
2168 const struct dcbnl_rtnl_ops *dcbnl_ops;
2171 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2172 u8 prio_tc_map[TC_BITMASK + 1];
2174 #if IS_ENABLED(CONFIG_FCOE)
2175 unsigned int fcoe_ddp_xid;
2177 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2178 struct netprio_map __rcu *priomap;
2180 struct phy_device *phydev;
2181 struct sfp_bus *sfp_bus;
2182 struct lock_class_key *qdisc_tx_busylock;
2183 struct lock_class_key *qdisc_running_key;
2185 unsigned wol_enabled:1;
2187 struct list_head net_notifier_list;
2189 #if IS_ENABLED(CONFIG_MACSEC)
2190 /* MACsec management functions */
2191 const struct macsec_ops *macsec_ops;
2193 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2194 struct udp_tunnel_nic *udp_tunnel_nic;
2196 /* protected by rtnl_lock */
2197 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2199 #define to_net_dev(d) container_of(d, struct net_device, dev)
2201 static inline bool netif_elide_gro(const struct net_device *dev)
2203 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2208 #define NETDEV_ALIGN 32
2211 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2213 return dev->prio_tc_map[prio & TC_BITMASK];
2217 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2219 if (tc >= dev->num_tc)
2222 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2226 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2227 void netdev_reset_tc(struct net_device *dev);
2228 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2229 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2232 int netdev_get_num_tc(struct net_device *dev)
2237 static inline void net_prefetch(void *p)
2240 #if L1_CACHE_BYTES < 128
2241 prefetch((u8 *)p + L1_CACHE_BYTES);
2245 static inline void net_prefetchw(void *p)
2248 #if L1_CACHE_BYTES < 128
2249 prefetchw((u8 *)p + L1_CACHE_BYTES);
2253 void netdev_unbind_sb_channel(struct net_device *dev,
2254 struct net_device *sb_dev);
2255 int netdev_bind_sb_channel_queue(struct net_device *dev,
2256 struct net_device *sb_dev,
2257 u8 tc, u16 count, u16 offset);
2258 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2259 static inline int netdev_get_sb_channel(struct net_device *dev)
2261 return max_t(int, -dev->num_tc, 0);
2265 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2268 return &dev->_tx[index];
2271 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2272 const struct sk_buff *skb)
2274 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2277 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2278 void (*f)(struct net_device *,
2279 struct netdev_queue *,
2285 for (i = 0; i < dev->num_tx_queues; i++)
2286 f(dev, &dev->_tx[i], arg);
2289 #define netdev_lockdep_set_classes(dev) \
2291 static struct lock_class_key qdisc_tx_busylock_key; \
2292 static struct lock_class_key qdisc_running_key; \
2293 static struct lock_class_key qdisc_xmit_lock_key; \
2294 static struct lock_class_key dev_addr_list_lock_key; \
2297 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2298 (dev)->qdisc_running_key = &qdisc_running_key; \
2299 lockdep_set_class(&(dev)->addr_list_lock, \
2300 &dev_addr_list_lock_key); \
2301 for (i = 0; i < (dev)->num_tx_queues; i++) \
2302 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2303 &qdisc_xmit_lock_key); \
2306 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2307 struct net_device *sb_dev);
2308 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2309 struct sk_buff *skb,
2310 struct net_device *sb_dev);
2312 /* returns the headroom that the master device needs to take in account
2313 * when forwarding to this dev
2315 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2317 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2320 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2322 if (dev->netdev_ops->ndo_set_rx_headroom)
2323 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2326 /* set the device rx headroom to the dev's default */
2327 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2329 netdev_set_rx_headroom(dev, -1);
2332 static inline void *netdev_get_ml_priv(struct net_device *dev,
2333 enum netdev_ml_priv_type type)
2335 if (dev->ml_priv_type != type)
2338 return dev->ml_priv;
2341 static inline void netdev_set_ml_priv(struct net_device *dev,
2343 enum netdev_ml_priv_type type)
2345 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2346 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2347 dev->ml_priv_type, type);
2348 WARN(!dev->ml_priv_type && dev->ml_priv,
2349 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2351 dev->ml_priv = ml_priv;
2352 dev->ml_priv_type = type;
2356 * Net namespace inlines
2359 struct net *dev_net(const struct net_device *dev)
2361 return read_pnet(&dev->nd_net);
2365 void dev_net_set(struct net_device *dev, struct net *net)
2367 write_pnet(&dev->nd_net, net);
2371 * netdev_priv - access network device private data
2372 * @dev: network device
2374 * Get network device private data
2376 static inline void *netdev_priv(const struct net_device *dev)
2378 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2381 /* Set the sysfs physical device reference for the network logical device
2382 * if set prior to registration will cause a symlink during initialization.
2384 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2386 /* Set the sysfs device type for the network logical device to allow
2387 * fine-grained identification of different network device types. For
2388 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2390 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2392 /* Default NAPI poll() weight
2393 * Device drivers are strongly advised to not use bigger value
2395 #define NAPI_POLL_WEIGHT 64
2398 * netif_napi_add - initialize a NAPI context
2399 * @dev: network device
2400 * @napi: NAPI context
2401 * @poll: polling function
2402 * @weight: default weight
2404 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2405 * *any* of the other NAPI-related functions.
2407 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2408 int (*poll)(struct napi_struct *, int), int weight);
2411 * netif_tx_napi_add - initialize a NAPI context
2412 * @dev: network device
2413 * @napi: NAPI context
2414 * @poll: polling function
2415 * @weight: default weight
2417 * This variant of netif_napi_add() should be used from drivers using NAPI
2418 * to exclusively poll a TX queue.
2419 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2421 static inline void netif_tx_napi_add(struct net_device *dev,
2422 struct napi_struct *napi,
2423 int (*poll)(struct napi_struct *, int),
2426 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2427 netif_napi_add(dev, napi, poll, weight);
2431 * __netif_napi_del - remove a NAPI context
2432 * @napi: NAPI context
2434 * Warning: caller must observe RCU grace period before freeing memory
2435 * containing @napi. Drivers might want to call this helper to combine
2436 * all the needed RCU grace periods into a single one.
2438 void __netif_napi_del(struct napi_struct *napi);
2441 * netif_napi_del - remove a NAPI context
2442 * @napi: NAPI context
2444 * netif_napi_del() removes a NAPI context from the network device NAPI list
2446 static inline void netif_napi_del(struct napi_struct *napi)
2448 __netif_napi_del(napi);
2452 struct napi_gro_cb {
2453 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2456 /* Length of frag0. */
2457 unsigned int frag0_len;
2459 /* This indicates where we are processing relative to skb->data. */
2462 /* This is non-zero if the packet cannot be merged with the new skb. */
2465 /* Save the IP ID here and check when we get to the transport layer */
2468 /* Number of segments aggregated. */
2471 /* Start offset for remote checksum offload */
2472 u16 gro_remcsum_start;
2474 /* jiffies when first packet was created/queued */
2477 /* Used in ipv6_gro_receive() and foo-over-udp */
2480 /* This is non-zero if the packet may be of the same flow. */
2483 /* Used in tunnel GRO receive */
2486 /* GRO checksum is valid */
2489 /* Number of checksums via CHECKSUM_UNNECESSARY */
2494 #define NAPI_GRO_FREE 1
2495 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2497 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2500 /* Used in GRE, set in fou/gue_gro_receive */
2503 /* Used to determine if flush_id can be ignored */
2506 /* Number of gro_receive callbacks this packet already went through */
2507 u8 recursion_counter:4;
2509 /* GRO is done by frag_list pointer chaining. */
2512 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2515 /* used in skb_gro_receive() slow path */
2516 struct sk_buff *last;
2519 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2521 #define GRO_RECURSION_LIMIT 15
2522 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2524 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2527 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2528 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2529 struct list_head *head,
2530 struct sk_buff *skb)
2532 if (unlikely(gro_recursion_inc_test(skb))) {
2533 NAPI_GRO_CB(skb)->flush |= 1;
2537 return cb(head, skb);
2540 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2542 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2544 struct list_head *head,
2545 struct sk_buff *skb)
2547 if (unlikely(gro_recursion_inc_test(skb))) {
2548 NAPI_GRO_CB(skb)->flush |= 1;
2552 return cb(sk, head, skb);
2555 struct packet_type {
2556 __be16 type; /* This is really htons(ether_type). */
2557 bool ignore_outgoing;
2558 struct net_device *dev; /* NULL is wildcarded here */
2559 int (*func) (struct sk_buff *,
2560 struct net_device *,
2561 struct packet_type *,
2562 struct net_device *);
2563 void (*list_func) (struct list_head *,
2564 struct packet_type *,
2565 struct net_device *);
2566 bool (*id_match)(struct packet_type *ptype,
2568 struct net *af_packet_net;
2569 void *af_packet_priv;
2570 struct list_head list;
2573 struct offload_callbacks {
2574 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2575 netdev_features_t features);
2576 struct sk_buff *(*gro_receive)(struct list_head *head,
2577 struct sk_buff *skb);
2578 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2581 struct packet_offload {
2582 __be16 type; /* This is really htons(ether_type). */
2584 struct offload_callbacks callbacks;
2585 struct list_head list;
2588 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2589 struct pcpu_sw_netstats {
2594 struct u64_stats_sync syncp;
2595 } __aligned(4 * sizeof(u64));
2597 struct pcpu_lstats {
2598 u64_stats_t packets;
2600 struct u64_stats_sync syncp;
2601 } __aligned(2 * sizeof(u64));
2603 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2605 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2607 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2609 u64_stats_update_begin(&tstats->syncp);
2610 tstats->rx_bytes += len;
2611 tstats->rx_packets++;
2612 u64_stats_update_end(&tstats->syncp);
2615 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2617 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2619 u64_stats_update_begin(&lstats->syncp);
2620 u64_stats_add(&lstats->bytes, len);
2621 u64_stats_inc(&lstats->packets);
2622 u64_stats_update_end(&lstats->syncp);
2625 #define __netdev_alloc_pcpu_stats(type, gfp) \
2627 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2630 for_each_possible_cpu(__cpu) { \
2631 typeof(type) *stat; \
2632 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2633 u64_stats_init(&stat->syncp); \
2639 #define netdev_alloc_pcpu_stats(type) \
2640 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2642 enum netdev_lag_tx_type {
2643 NETDEV_LAG_TX_TYPE_UNKNOWN,
2644 NETDEV_LAG_TX_TYPE_RANDOM,
2645 NETDEV_LAG_TX_TYPE_BROADCAST,
2646 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2647 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2648 NETDEV_LAG_TX_TYPE_HASH,
2651 enum netdev_lag_hash {
2652 NETDEV_LAG_HASH_NONE,
2654 NETDEV_LAG_HASH_L34,
2655 NETDEV_LAG_HASH_L23,
2656 NETDEV_LAG_HASH_E23,
2657 NETDEV_LAG_HASH_E34,
2658 NETDEV_LAG_HASH_UNKNOWN,
2661 struct netdev_lag_upper_info {
2662 enum netdev_lag_tx_type tx_type;
2663 enum netdev_lag_hash hash_type;
2666 struct netdev_lag_lower_state_info {
2671 #include <linux/notifier.h>
2673 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2674 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2678 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2680 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2681 detected a hardware crash and restarted
2682 - we can use this eg to kick tcp sessions
2684 NETDEV_CHANGE, /* Notify device state change */
2687 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2688 NETDEV_CHANGEADDR, /* notify after the address change */
2689 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2693 NETDEV_BONDING_FAILOVER,
2695 NETDEV_PRE_TYPE_CHANGE,
2696 NETDEV_POST_TYPE_CHANGE,
2699 NETDEV_NOTIFY_PEERS,
2703 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2704 NETDEV_CHANGEINFODATA,
2705 NETDEV_BONDING_INFO,
2706 NETDEV_PRECHANGEUPPER,
2707 NETDEV_CHANGELOWERSTATE,
2708 NETDEV_UDP_TUNNEL_PUSH_INFO,
2709 NETDEV_UDP_TUNNEL_DROP_INFO,
2710 NETDEV_CHANGE_TX_QUEUE_LEN,
2711 NETDEV_CVLAN_FILTER_PUSH_INFO,
2712 NETDEV_CVLAN_FILTER_DROP_INFO,
2713 NETDEV_SVLAN_FILTER_PUSH_INFO,
2714 NETDEV_SVLAN_FILTER_DROP_INFO,
2716 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2718 int register_netdevice_notifier(struct notifier_block *nb);
2719 int unregister_netdevice_notifier(struct notifier_block *nb);
2720 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2721 int unregister_netdevice_notifier_net(struct net *net,
2722 struct notifier_block *nb);
2723 int register_netdevice_notifier_dev_net(struct net_device *dev,
2724 struct notifier_block *nb,
2725 struct netdev_net_notifier *nn);
2726 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2727 struct notifier_block *nb,
2728 struct netdev_net_notifier *nn);
2730 struct netdev_notifier_info {
2731 struct net_device *dev;
2732 struct netlink_ext_ack *extack;
2735 struct netdev_notifier_info_ext {
2736 struct netdev_notifier_info info; /* must be first */
2742 struct netdev_notifier_change_info {
2743 struct netdev_notifier_info info; /* must be first */
2744 unsigned int flags_changed;
2747 struct netdev_notifier_changeupper_info {
2748 struct netdev_notifier_info info; /* must be first */
2749 struct net_device *upper_dev; /* new upper dev */
2750 bool master; /* is upper dev master */
2751 bool linking; /* is the notification for link or unlink */
2752 void *upper_info; /* upper dev info */
2755 struct netdev_notifier_changelowerstate_info {
2756 struct netdev_notifier_info info; /* must be first */
2757 void *lower_state_info; /* is lower dev state */
2760 struct netdev_notifier_pre_changeaddr_info {
2761 struct netdev_notifier_info info; /* must be first */
2762 const unsigned char *dev_addr;
2765 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2766 struct net_device *dev)
2769 info->extack = NULL;
2772 static inline struct net_device *
2773 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2778 static inline struct netlink_ext_ack *
2779 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2781 return info->extack;
2784 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2787 extern rwlock_t dev_base_lock; /* Device list lock */
2789 #define for_each_netdev(net, d) \
2790 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2791 #define for_each_netdev_reverse(net, d) \
2792 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2793 #define for_each_netdev_rcu(net, d) \
2794 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2795 #define for_each_netdev_safe(net, d, n) \
2796 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2797 #define for_each_netdev_continue(net, d) \
2798 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2799 #define for_each_netdev_continue_reverse(net, d) \
2800 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2802 #define for_each_netdev_continue_rcu(net, d) \
2803 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2804 #define for_each_netdev_in_bond_rcu(bond, slave) \
2805 for_each_netdev_rcu(&init_net, slave) \
2806 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2807 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2809 static inline struct net_device *next_net_device(struct net_device *dev)
2811 struct list_head *lh;
2815 lh = dev->dev_list.next;
2816 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2819 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2821 struct list_head *lh;
2825 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2826 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2829 static inline struct net_device *first_net_device(struct net *net)
2831 return list_empty(&net->dev_base_head) ? NULL :
2832 net_device_entry(net->dev_base_head.next);
2835 static inline struct net_device *first_net_device_rcu(struct net *net)
2837 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2839 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2842 int netdev_boot_setup_check(struct net_device *dev);
2843 unsigned long netdev_boot_base(const char *prefix, int unit);
2844 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2845 const char *hwaddr);
2846 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2847 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2848 void dev_add_pack(struct packet_type *pt);
2849 void dev_remove_pack(struct packet_type *pt);
2850 void __dev_remove_pack(struct packet_type *pt);
2851 void dev_add_offload(struct packet_offload *po);
2852 void dev_remove_offload(struct packet_offload *po);
2854 int dev_get_iflink(const struct net_device *dev);
2855 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2856 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2857 unsigned short mask);
2858 struct net_device *dev_get_by_name(struct net *net, const char *name);
2859 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2860 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2861 int dev_alloc_name(struct net_device *dev, const char *name);
2862 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2863 void dev_close(struct net_device *dev);
2864 void dev_close_many(struct list_head *head, bool unlink);
2865 void dev_disable_lro(struct net_device *dev);
2866 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2867 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2868 struct net_device *sb_dev);
2869 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2870 struct net_device *sb_dev);
2872 int dev_queue_xmit(struct sk_buff *skb);
2873 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2874 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2876 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2880 ret = __dev_direct_xmit(skb, queue_id);
2881 if (!dev_xmit_complete(ret))
2886 int register_netdevice(struct net_device *dev);
2887 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2888 void unregister_netdevice_many(struct list_head *head);
2889 static inline void unregister_netdevice(struct net_device *dev)
2891 unregister_netdevice_queue(dev, NULL);
2894 int netdev_refcnt_read(const struct net_device *dev);
2895 void free_netdev(struct net_device *dev);
2896 void netdev_freemem(struct net_device *dev);
2897 int init_dummy_netdev(struct net_device *dev);
2899 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2900 struct sk_buff *skb,
2902 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2903 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2904 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2905 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2906 int netdev_get_name(struct net *net, char *name, int ifindex);
2907 int dev_restart(struct net_device *dev);
2908 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2909 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2911 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2913 return NAPI_GRO_CB(skb)->data_offset;
2916 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2918 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2921 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2923 NAPI_GRO_CB(skb)->data_offset += len;
2926 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2927 unsigned int offset)
2929 return NAPI_GRO_CB(skb)->frag0 + offset;
2932 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2934 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2937 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2939 NAPI_GRO_CB(skb)->frag0 = NULL;
2940 NAPI_GRO_CB(skb)->frag0_len = 0;
2943 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2944 unsigned int offset)
2946 if (!pskb_may_pull(skb, hlen))
2949 skb_gro_frag0_invalidate(skb);
2950 return skb->data + offset;
2953 static inline void *skb_gro_network_header(struct sk_buff *skb)
2955 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2956 skb_network_offset(skb);
2959 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2960 const void *start, unsigned int len)
2962 if (NAPI_GRO_CB(skb)->csum_valid)
2963 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2964 csum_partial(start, len, 0));
2967 /* GRO checksum functions. These are logical equivalents of the normal
2968 * checksum functions (in skbuff.h) except that they operate on the GRO
2969 * offsets and fields in sk_buff.
2972 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2974 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2976 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2979 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2983 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2984 skb_checksum_start_offset(skb) <
2985 skb_gro_offset(skb)) &&
2986 !skb_at_gro_remcsum_start(skb) &&
2987 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2988 (!zero_okay || check));
2991 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2994 if (NAPI_GRO_CB(skb)->csum_valid &&
2995 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2998 NAPI_GRO_CB(skb)->csum = psum;
3000 return __skb_gro_checksum_complete(skb);
3003 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3005 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3006 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3007 NAPI_GRO_CB(skb)->csum_cnt--;
3009 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3010 * verified a new top level checksum or an encapsulated one
3011 * during GRO. This saves work if we fallback to normal path.
3013 __skb_incr_checksum_unnecessary(skb);
3017 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3020 __sum16 __ret = 0; \
3021 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3022 __ret = __skb_gro_checksum_validate_complete(skb, \
3023 compute_pseudo(skb, proto)); \
3025 skb_gro_incr_csum_unnecessary(skb); \
3029 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3030 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3032 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3034 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3036 #define skb_gro_checksum_simple_validate(skb) \
3037 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3039 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3041 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3042 !NAPI_GRO_CB(skb)->csum_valid);
3045 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3048 NAPI_GRO_CB(skb)->csum = ~pseudo;
3049 NAPI_GRO_CB(skb)->csum_valid = 1;
3052 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3054 if (__skb_gro_checksum_convert_check(skb)) \
3055 __skb_gro_checksum_convert(skb, \
3056 compute_pseudo(skb, proto)); \
3059 struct gro_remcsum {
3064 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3070 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3071 unsigned int off, size_t hdrlen,
3072 int start, int offset,
3073 struct gro_remcsum *grc,
3077 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3079 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3082 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3086 ptr = skb_gro_header_fast(skb, off);
3087 if (skb_gro_header_hard(skb, off + plen)) {
3088 ptr = skb_gro_header_slow(skb, off + plen, off);
3093 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3096 /* Adjust skb->csum since we changed the packet */
3097 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3099 grc->offset = off + hdrlen + offset;
3105 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3106 struct gro_remcsum *grc)
3109 size_t plen = grc->offset + sizeof(u16);
3114 ptr = skb_gro_header_fast(skb, grc->offset);
3115 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3116 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3121 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3124 #ifdef CONFIG_XFRM_OFFLOAD
3125 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3127 if (PTR_ERR(pp) != -EINPROGRESS)
3128 NAPI_GRO_CB(skb)->flush |= flush;
3130 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3133 struct gro_remcsum *grc)
3135 if (PTR_ERR(pp) != -EINPROGRESS) {
3136 NAPI_GRO_CB(skb)->flush |= flush;
3137 skb_gro_remcsum_cleanup(skb, grc);
3138 skb->remcsum_offload = 0;
3142 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3144 NAPI_GRO_CB(skb)->flush |= flush;
3146 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3149 struct gro_remcsum *grc)
3151 NAPI_GRO_CB(skb)->flush |= flush;
3152 skb_gro_remcsum_cleanup(skb, grc);
3153 skb->remcsum_offload = 0;
3157 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3158 unsigned short type,
3159 const void *daddr, const void *saddr,
3162 if (!dev->header_ops || !dev->header_ops->create)
3165 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3168 static inline int dev_parse_header(const struct sk_buff *skb,
3169 unsigned char *haddr)
3171 const struct net_device *dev = skb->dev;
3173 if (!dev->header_ops || !dev->header_ops->parse)
3175 return dev->header_ops->parse(skb, haddr);
3178 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3180 const struct net_device *dev = skb->dev;
3182 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3184 return dev->header_ops->parse_protocol(skb);
3187 /* ll_header must have at least hard_header_len allocated */
3188 static inline bool dev_validate_header(const struct net_device *dev,
3189 char *ll_header, int len)
3191 if (likely(len >= dev->hard_header_len))
3193 if (len < dev->min_header_len)
3196 if (capable(CAP_SYS_RAWIO)) {
3197 memset(ll_header + len, 0, dev->hard_header_len - len);
3201 if (dev->header_ops && dev->header_ops->validate)
3202 return dev->header_ops->validate(ll_header, len);
3207 static inline bool dev_has_header(const struct net_device *dev)
3209 return dev->header_ops && dev->header_ops->create;
3212 #ifdef CONFIG_NET_FLOW_LIMIT
3213 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3214 struct sd_flow_limit {
3216 unsigned int num_buckets;
3217 unsigned int history_head;
3218 u16 history[FLOW_LIMIT_HISTORY];
3222 extern int netdev_flow_limit_table_len;
3223 #endif /* CONFIG_NET_FLOW_LIMIT */
3226 * Incoming packets are placed on per-CPU queues
3228 struct softnet_data {
3229 struct list_head poll_list;
3230 struct sk_buff_head process_queue;
3233 unsigned int processed;
3234 unsigned int time_squeeze;
3235 unsigned int received_rps;
3237 struct softnet_data *rps_ipi_list;
3239 #ifdef CONFIG_NET_FLOW_LIMIT
3240 struct sd_flow_limit __rcu *flow_limit;
3242 struct Qdisc *output_queue;
3243 struct Qdisc **output_queue_tailp;
3244 struct sk_buff *completion_queue;
3245 #ifdef CONFIG_XFRM_OFFLOAD
3246 struct sk_buff_head xfrm_backlog;
3248 /* written and read only by owning cpu: */
3254 /* input_queue_head should be written by cpu owning this struct,
3255 * and only read by other cpus. Worth using a cache line.
3257 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3259 /* Elements below can be accessed between CPUs for RPS/RFS */
3260 call_single_data_t csd ____cacheline_aligned_in_smp;
3261 struct softnet_data *rps_ipi_next;
3263 unsigned int input_queue_tail;
3265 unsigned int dropped;
3266 struct sk_buff_head input_pkt_queue;
3267 struct napi_struct backlog;
3271 static inline void input_queue_head_incr(struct softnet_data *sd)
3274 sd->input_queue_head++;
3278 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3279 unsigned int *qtail)
3282 *qtail = ++sd->input_queue_tail;
3286 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3288 static inline int dev_recursion_level(void)
3290 return this_cpu_read(softnet_data.xmit.recursion);
3293 #define XMIT_RECURSION_LIMIT 8
3294 static inline bool dev_xmit_recursion(void)
3296 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3297 XMIT_RECURSION_LIMIT);
3300 static inline void dev_xmit_recursion_inc(void)
3302 __this_cpu_inc(softnet_data.xmit.recursion);
3305 static inline void dev_xmit_recursion_dec(void)
3307 __this_cpu_dec(softnet_data.xmit.recursion);
3310 void __netif_schedule(struct Qdisc *q);
3311 void netif_schedule_queue(struct netdev_queue *txq);
3313 static inline void netif_tx_schedule_all(struct net_device *dev)
3317 for (i = 0; i < dev->num_tx_queues; i++)
3318 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3321 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3323 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3327 * netif_start_queue - allow transmit
3328 * @dev: network device
3330 * Allow upper layers to call the device hard_start_xmit routine.
3332 static inline void netif_start_queue(struct net_device *dev)
3334 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3337 static inline void netif_tx_start_all_queues(struct net_device *dev)
3341 for (i = 0; i < dev->num_tx_queues; i++) {
3342 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3343 netif_tx_start_queue(txq);
3347 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3350 * netif_wake_queue - restart transmit
3351 * @dev: network device
3353 * Allow upper layers to call the device hard_start_xmit routine.
3354 * Used for flow control when transmit resources are available.
3356 static inline void netif_wake_queue(struct net_device *dev)
3358 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3361 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3365 for (i = 0; i < dev->num_tx_queues; i++) {
3366 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3367 netif_tx_wake_queue(txq);
3371 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3373 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3377 * netif_stop_queue - stop transmitted packets
3378 * @dev: network device
3380 * Stop upper layers calling the device hard_start_xmit routine.
3381 * Used for flow control when transmit resources are unavailable.
3383 static inline void netif_stop_queue(struct net_device *dev)
3385 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3388 void netif_tx_stop_all_queues(struct net_device *dev);
3390 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3392 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3396 * netif_queue_stopped - test if transmit queue is flowblocked
3397 * @dev: network device
3399 * Test if transmit queue on device is currently unable to send.
3401 static inline bool netif_queue_stopped(const struct net_device *dev)
3403 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3406 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3408 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3412 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3414 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3418 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3420 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3424 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3425 * @dev_queue: pointer to transmit queue
3427 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3428 * to give appropriate hint to the CPU.
3430 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3433 prefetchw(&dev_queue->dql.num_queued);
3438 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3439 * @dev_queue: pointer to transmit queue
3441 * BQL enabled drivers might use this helper in their TX completion path,
3442 * to give appropriate hint to the CPU.
3444 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3447 prefetchw(&dev_queue->dql.limit);
3451 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3455 dql_queued(&dev_queue->dql, bytes);
3457 if (likely(dql_avail(&dev_queue->dql) >= 0))
3460 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3463 * The XOFF flag must be set before checking the dql_avail below,
3464 * because in netdev_tx_completed_queue we update the dql_completed
3465 * before checking the XOFF flag.
3469 /* check again in case another CPU has just made room avail */
3470 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3471 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3475 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3476 * that they should not test BQL status themselves.
3477 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3479 * Returns true if the doorbell must be used to kick the NIC.
3481 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3487 dql_queued(&dev_queue->dql, bytes);
3489 return netif_tx_queue_stopped(dev_queue);
3491 netdev_tx_sent_queue(dev_queue, bytes);
3496 * netdev_sent_queue - report the number of bytes queued to hardware
3497 * @dev: network device
3498 * @bytes: number of bytes queued to the hardware device queue
3500 * Report the number of bytes queued for sending/completion to the network
3501 * device hardware queue. @bytes should be a good approximation and should
3502 * exactly match netdev_completed_queue() @bytes
3504 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3506 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3509 static inline bool __netdev_sent_queue(struct net_device *dev,
3513 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3517 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3518 unsigned int pkts, unsigned int bytes)
3521 if (unlikely(!bytes))
3524 dql_completed(&dev_queue->dql, bytes);
3527 * Without the memory barrier there is a small possiblity that
3528 * netdev_tx_sent_queue will miss the update and cause the queue to
3529 * be stopped forever
3533 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3536 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3537 netif_schedule_queue(dev_queue);
3542 * netdev_completed_queue - report bytes and packets completed by device
3543 * @dev: network device
3544 * @pkts: actual number of packets sent over the medium
3545 * @bytes: actual number of bytes sent over the medium
3547 * Report the number of bytes and packets transmitted by the network device
3548 * hardware queue over the physical medium, @bytes must exactly match the
3549 * @bytes amount passed to netdev_sent_queue()
3551 static inline void netdev_completed_queue(struct net_device *dev,
3552 unsigned int pkts, unsigned int bytes)
3554 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3557 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3560 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3566 * netdev_reset_queue - reset the packets and bytes count of a network device
3567 * @dev_queue: network device
3569 * Reset the bytes and packet count of a network device and clear the
3570 * software flow control OFF bit for this network device
3572 static inline void netdev_reset_queue(struct net_device *dev_queue)
3574 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3578 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3579 * @dev: network device
3580 * @queue_index: given tx queue index
3582 * Returns 0 if given tx queue index >= number of device tx queues,
3583 * otherwise returns the originally passed tx queue index.
3585 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3587 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3588 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3589 dev->name, queue_index,
3590 dev->real_num_tx_queues);
3598 * netif_running - test if up
3599 * @dev: network device
3601 * Test if the device has been brought up.
3603 static inline bool netif_running(const struct net_device *dev)
3605 return test_bit(__LINK_STATE_START, &dev->state);
3609 * Routines to manage the subqueues on a device. We only need start,
3610 * stop, and a check if it's stopped. All other device management is
3611 * done at the overall netdevice level.
3612 * Also test the device if we're multiqueue.
3616 * netif_start_subqueue - allow sending packets on subqueue
3617 * @dev: network device
3618 * @queue_index: sub queue index
3620 * Start individual transmit queue of a device with multiple transmit queues.
3622 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3624 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3626 netif_tx_start_queue(txq);
3630 * netif_stop_subqueue - stop sending packets on subqueue
3631 * @dev: network device
3632 * @queue_index: sub queue index
3634 * Stop individual transmit queue of a device with multiple transmit queues.
3636 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3638 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3639 netif_tx_stop_queue(txq);
3643 * netif_subqueue_stopped - test status of subqueue
3644 * @dev: network device
3645 * @queue_index: sub queue index
3647 * Check individual transmit queue of a device with multiple transmit queues.
3649 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3652 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3654 return netif_tx_queue_stopped(txq);
3657 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3658 struct sk_buff *skb)
3660 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3664 * netif_wake_subqueue - allow sending packets on subqueue
3665 * @dev: network device
3666 * @queue_index: sub queue index
3668 * Resume individual transmit queue of a device with multiple transmit queues.
3670 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3672 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3674 netif_tx_wake_queue(txq);
3678 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3680 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3681 u16 index, bool is_rxqs_map);
3684 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3685 * @j: CPU/Rx queue index
3686 * @mask: bitmask of all cpus/rx queues
3687 * @nr_bits: number of bits in the bitmask
3689 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3691 static inline bool netif_attr_test_mask(unsigned long j,
3692 const unsigned long *mask,
3693 unsigned int nr_bits)
3695 cpu_max_bits_warn(j, nr_bits);
3696 return test_bit(j, mask);
3700 * netif_attr_test_online - Test for online CPU/Rx queue
3701 * @j: CPU/Rx queue index
3702 * @online_mask: bitmask for CPUs/Rx queues that are online
3703 * @nr_bits: number of bits in the bitmask
3705 * Returns true if a CPU/Rx queue is online.
3707 static inline bool netif_attr_test_online(unsigned long j,
3708 const unsigned long *online_mask,
3709 unsigned int nr_bits)
3711 cpu_max_bits_warn(j, nr_bits);
3714 return test_bit(j, online_mask);
3716 return (j < nr_bits);
3720 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3721 * @n: CPU/Rx queue index
3722 * @srcp: the cpumask/Rx queue mask pointer
3723 * @nr_bits: number of bits in the bitmask
3725 * Returns >= nr_bits if no further CPUs/Rx queues set.
3727 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3728 unsigned int nr_bits)
3730 /* -1 is a legal arg here. */
3732 cpu_max_bits_warn(n, nr_bits);
3735 return find_next_bit(srcp, nr_bits, n + 1);
3741 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3742 * @n: CPU/Rx queue index
3743 * @src1p: the first CPUs/Rx queues mask pointer
3744 * @src2p: the second CPUs/Rx queues mask pointer
3745 * @nr_bits: number of bits in the bitmask
3747 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3749 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3750 const unsigned long *src2p,
3751 unsigned int nr_bits)
3753 /* -1 is a legal arg here. */
3755 cpu_max_bits_warn(n, nr_bits);
3758 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3760 return find_next_bit(src1p, nr_bits, n + 1);
3762 return find_next_bit(src2p, nr_bits, n + 1);
3767 static inline int netif_set_xps_queue(struct net_device *dev,
3768 const struct cpumask *mask,
3774 static inline int __netif_set_xps_queue(struct net_device *dev,
3775 const unsigned long *mask,
3776 u16 index, bool is_rxqs_map)
3783 * netif_is_multiqueue - test if device has multiple transmit queues
3784 * @dev: network device
3786 * Check if device has multiple transmit queues
3788 static inline bool netif_is_multiqueue(const struct net_device *dev)
3790 return dev->num_tx_queues > 1;
3793 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3796 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3798 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3801 dev->real_num_rx_queues = rxqs;
3806 static inline struct netdev_rx_queue *
3807 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3809 return dev->_rx + rxq;
3813 static inline unsigned int get_netdev_rx_queue_index(
3814 struct netdev_rx_queue *queue)
3816 struct net_device *dev = queue->dev;
3817 int index = queue - dev->_rx;
3819 BUG_ON(index >= dev->num_rx_queues);
3824 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3825 int netif_get_num_default_rss_queues(void);
3827 enum skb_free_reason {
3828 SKB_REASON_CONSUMED,
3832 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3833 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3836 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3837 * interrupt context or with hardware interrupts being disabled.
3838 * (in_irq() || irqs_disabled())
3840 * We provide four helpers that can be used in following contexts :
3842 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3843 * replacing kfree_skb(skb)
3845 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3846 * Typically used in place of consume_skb(skb) in TX completion path
3848 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3849 * replacing kfree_skb(skb)
3851 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3852 * and consumed a packet. Used in place of consume_skb(skb)
3854 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3856 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3859 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3861 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3864 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3866 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3869 static inline void dev_consume_skb_any(struct sk_buff *skb)
3871 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3874 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3875 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3876 int netif_rx(struct sk_buff *skb);
3877 int netif_rx_ni(struct sk_buff *skb);
3878 int netif_rx_any_context(struct sk_buff *skb);
3879 int netif_receive_skb(struct sk_buff *skb);
3880 int netif_receive_skb_core(struct sk_buff *skb);
3881 void netif_receive_skb_list(struct list_head *head);
3882 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3883 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3884 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3885 gro_result_t napi_gro_frags(struct napi_struct *napi);
3886 struct packet_offload *gro_find_receive_by_type(__be16 type);
3887 struct packet_offload *gro_find_complete_by_type(__be16 type);
3889 static inline void napi_free_frags(struct napi_struct *napi)
3891 kfree_skb(napi->skb);
3895 bool netdev_is_rx_handler_busy(struct net_device *dev);
3896 int netdev_rx_handler_register(struct net_device *dev,
3897 rx_handler_func_t *rx_handler,
3898 void *rx_handler_data);
3899 void netdev_rx_handler_unregister(struct net_device *dev);
3901 bool dev_valid_name(const char *name);
3902 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3904 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3906 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3907 bool *need_copyout);
3908 int dev_ifconf(struct net *net, struct ifconf *, int);
3909 int dev_ethtool(struct net *net, struct ifreq *);
3910 unsigned int dev_get_flags(const struct net_device *);
3911 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3912 struct netlink_ext_ack *extack);
3913 int dev_change_flags(struct net_device *dev, unsigned int flags,
3914 struct netlink_ext_ack *extack);
3915 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3916 unsigned int gchanges);
3917 int dev_change_name(struct net_device *, const char *);
3918 int dev_set_alias(struct net_device *, const char *, size_t);
3919 int dev_get_alias(const struct net_device *, char *, size_t);
3920 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3921 int __dev_set_mtu(struct net_device *, int);
3922 int dev_validate_mtu(struct net_device *dev, int mtu,
3923 struct netlink_ext_ack *extack);
3924 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3925 struct netlink_ext_ack *extack);
3926 int dev_set_mtu(struct net_device *, int);
3927 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3928 void dev_set_group(struct net_device *, int);
3929 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3930 struct netlink_ext_ack *extack);
3931 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3932 struct netlink_ext_ack *extack);
3933 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3934 struct netlink_ext_ack *extack);
3935 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3936 int dev_change_carrier(struct net_device *, bool new_carrier);
3937 int dev_get_phys_port_id(struct net_device *dev,
3938 struct netdev_phys_item_id *ppid);
3939 int dev_get_phys_port_name(struct net_device *dev,
3940 char *name, size_t len);
3941 int dev_get_port_parent_id(struct net_device *dev,
3942 struct netdev_phys_item_id *ppid, bool recurse);
3943 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3944 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3945 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3946 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3948 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3949 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3950 struct netdev_queue *txq, int *ret);
3952 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3953 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3954 int fd, int expected_fd, u32 flags);
3955 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3956 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3958 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3960 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3961 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3962 bool is_skb_forwardable(const struct net_device *dev,
3963 const struct sk_buff *skb);
3965 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3966 struct sk_buff *skb)
3968 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3969 unlikely(!is_skb_forwardable(dev, skb))) {
3970 atomic_long_inc(&dev->rx_dropped);
3975 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3980 bool dev_nit_active(struct net_device *dev);
3981 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3983 extern int netdev_budget;
3984 extern unsigned int netdev_budget_usecs;
3986 /* Called by rtnetlink.c:rtnl_unlock() */
3987 void netdev_run_todo(void);
3990 * dev_put - release reference to device
3991 * @dev: network device
3993 * Release reference to device to allow it to be freed.
3995 static inline void dev_put(struct net_device *dev)
3998 this_cpu_dec(*dev->pcpu_refcnt);
4002 * dev_hold - get reference to device
4003 * @dev: network device
4005 * Hold reference to device to keep it from being freed.
4007 static inline void dev_hold(struct net_device *dev)
4010 this_cpu_inc(*dev->pcpu_refcnt);
4013 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4014 * and _off may be called from IRQ context, but it is caller
4015 * who is responsible for serialization of these calls.
4017 * The name carrier is inappropriate, these functions should really be
4018 * called netif_lowerlayer_*() because they represent the state of any
4019 * kind of lower layer not just hardware media.
4022 void linkwatch_init_dev(struct net_device *dev);
4023 void linkwatch_fire_event(struct net_device *dev);
4024 void linkwatch_forget_dev(struct net_device *dev);
4027 * netif_carrier_ok - test if carrier present
4028 * @dev: network device
4030 * Check if carrier is present on device
4032 static inline bool netif_carrier_ok(const struct net_device *dev)
4034 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4037 unsigned long dev_trans_start(struct net_device *dev);
4039 void __netdev_watchdog_up(struct net_device *dev);
4041 void netif_carrier_on(struct net_device *dev);
4043 void netif_carrier_off(struct net_device *dev);
4046 * netif_dormant_on - mark device as dormant.
4047 * @dev: network device
4049 * Mark device as dormant (as per RFC2863).
4051 * The dormant state indicates that the relevant interface is not
4052 * actually in a condition to pass packets (i.e., it is not 'up') but is
4053 * in a "pending" state, waiting for some external event. For "on-
4054 * demand" interfaces, this new state identifies the situation where the
4055 * interface is waiting for events to place it in the up state.
4057 static inline void netif_dormant_on(struct net_device *dev)
4059 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4060 linkwatch_fire_event(dev);
4064 * netif_dormant_off - set device as not dormant.
4065 * @dev: network device
4067 * Device is not in dormant state.
4069 static inline void netif_dormant_off(struct net_device *dev)
4071 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4072 linkwatch_fire_event(dev);
4076 * netif_dormant - test if device is dormant
4077 * @dev: network device
4079 * Check if device is dormant.
4081 static inline bool netif_dormant(const struct net_device *dev)
4083 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4088 * netif_testing_on - mark device as under test.
4089 * @dev: network device
4091 * Mark device as under test (as per RFC2863).
4093 * The testing state indicates that some test(s) must be performed on
4094 * the interface. After completion, of the test, the interface state
4095 * will change to up, dormant, or down, as appropriate.
4097 static inline void netif_testing_on(struct net_device *dev)
4099 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4100 linkwatch_fire_event(dev);
4104 * netif_testing_off - set device as not under test.
4105 * @dev: network device
4107 * Device is not in testing state.
4109 static inline void netif_testing_off(struct net_device *dev)
4111 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4112 linkwatch_fire_event(dev);
4116 * netif_testing - test if device is under test
4117 * @dev: network device
4119 * Check if device is under test
4121 static inline bool netif_testing(const struct net_device *dev)
4123 return test_bit(__LINK_STATE_TESTING, &dev->state);
4128 * netif_oper_up - test if device is operational
4129 * @dev: network device
4131 * Check if carrier is operational
4133 static inline bool netif_oper_up(const struct net_device *dev)
4135 return (dev->operstate == IF_OPER_UP ||
4136 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4140 * netif_device_present - is device available or removed
4141 * @dev: network device
4143 * Check if device has not been removed from system.
4145 static inline bool netif_device_present(struct net_device *dev)
4147 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4150 void netif_device_detach(struct net_device *dev);
4152 void netif_device_attach(struct net_device *dev);
4155 * Network interface message level settings
4160 NETIF_MSG_PROBE_BIT,
4162 NETIF_MSG_TIMER_BIT,
4163 NETIF_MSG_IFDOWN_BIT,
4165 NETIF_MSG_RX_ERR_BIT,
4166 NETIF_MSG_TX_ERR_BIT,
4167 NETIF_MSG_TX_QUEUED_BIT,
4169 NETIF_MSG_TX_DONE_BIT,
4170 NETIF_MSG_RX_STATUS_BIT,
4171 NETIF_MSG_PKTDATA_BIT,
4175 /* When you add a new bit above, update netif_msg_class_names array
4176 * in net/ethtool/common.c
4178 NETIF_MSG_CLASS_COUNT,
4180 /* Both ethtool_ops interface and internal driver implementation use u32 */
4181 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4183 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4184 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4186 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4187 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4188 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4189 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4190 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4191 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4192 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4193 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4194 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4195 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4196 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4197 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4198 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4199 #define NETIF_MSG_HW __NETIF_MSG(HW)
4200 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4202 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4203 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4204 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4205 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4206 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4207 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4208 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4209 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4210 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4211 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4212 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4213 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4214 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4215 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4216 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4218 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4221 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4222 return default_msg_enable_bits;
4223 if (debug_value == 0) /* no output */
4225 /* set low N bits */
4226 return (1U << debug_value) - 1;
4229 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4231 spin_lock(&txq->_xmit_lock);
4232 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4233 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4236 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4238 __acquire(&txq->_xmit_lock);
4242 static inline void __netif_tx_release(struct netdev_queue *txq)
4244 __release(&txq->_xmit_lock);
4247 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4249 spin_lock_bh(&txq->_xmit_lock);
4250 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4251 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4254 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4256 bool ok = spin_trylock(&txq->_xmit_lock);
4259 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4260 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4265 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4267 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4268 WRITE_ONCE(txq->xmit_lock_owner, -1);
4269 spin_unlock(&txq->_xmit_lock);
4272 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4274 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4275 WRITE_ONCE(txq->xmit_lock_owner, -1);
4276 spin_unlock_bh(&txq->_xmit_lock);
4279 static inline void txq_trans_update(struct netdev_queue *txq)
4281 if (txq->xmit_lock_owner != -1)
4282 txq->trans_start = jiffies;
4285 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4286 static inline void netif_trans_update(struct net_device *dev)
4288 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4290 if (txq->trans_start != jiffies)
4291 txq->trans_start = jiffies;
4295 * netif_tx_lock - grab network device transmit lock
4296 * @dev: network device
4298 * Get network device transmit lock
4300 static inline void netif_tx_lock(struct net_device *dev)
4305 spin_lock(&dev->tx_global_lock);
4306 cpu = smp_processor_id();
4307 for (i = 0; i < dev->num_tx_queues; i++) {
4308 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4310 /* We are the only thread of execution doing a
4311 * freeze, but we have to grab the _xmit_lock in
4312 * order to synchronize with threads which are in
4313 * the ->hard_start_xmit() handler and already
4314 * checked the frozen bit.
4316 __netif_tx_lock(txq, cpu);
4317 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4318 __netif_tx_unlock(txq);
4322 static inline void netif_tx_lock_bh(struct net_device *dev)
4328 static inline void netif_tx_unlock(struct net_device *dev)
4332 for (i = 0; i < dev->num_tx_queues; i++) {
4333 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4335 /* No need to grab the _xmit_lock here. If the
4336 * queue is not stopped for another reason, we
4339 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4340 netif_schedule_queue(txq);
4342 spin_unlock(&dev->tx_global_lock);
4345 static inline void netif_tx_unlock_bh(struct net_device *dev)
4347 netif_tx_unlock(dev);
4351 #define HARD_TX_LOCK(dev, txq, cpu) { \
4352 if ((dev->features & NETIF_F_LLTX) == 0) { \
4353 __netif_tx_lock(txq, cpu); \
4355 __netif_tx_acquire(txq); \
4359 #define HARD_TX_TRYLOCK(dev, txq) \
4360 (((dev->features & NETIF_F_LLTX) == 0) ? \
4361 __netif_tx_trylock(txq) : \
4362 __netif_tx_acquire(txq))
4364 #define HARD_TX_UNLOCK(dev, txq) { \
4365 if ((dev->features & NETIF_F_LLTX) == 0) { \
4366 __netif_tx_unlock(txq); \
4368 __netif_tx_release(txq); \
4372 static inline void netif_tx_disable(struct net_device *dev)
4378 cpu = smp_processor_id();
4379 spin_lock(&dev->tx_global_lock);
4380 for (i = 0; i < dev->num_tx_queues; i++) {
4381 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4383 __netif_tx_lock(txq, cpu);
4384 netif_tx_stop_queue(txq);
4385 __netif_tx_unlock(txq);
4387 spin_unlock(&dev->tx_global_lock);
4391 static inline void netif_addr_lock(struct net_device *dev)
4393 unsigned char nest_level = 0;
4395 #ifdef CONFIG_LOCKDEP
4396 nest_level = dev->nested_level;
4398 spin_lock_nested(&dev->addr_list_lock, nest_level);
4401 static inline void netif_addr_lock_bh(struct net_device *dev)
4403 unsigned char nest_level = 0;
4405 #ifdef CONFIG_LOCKDEP
4406 nest_level = dev->nested_level;
4409 spin_lock_nested(&dev->addr_list_lock, nest_level);
4412 static inline void netif_addr_unlock(struct net_device *dev)
4414 spin_unlock(&dev->addr_list_lock);
4417 static inline void netif_addr_unlock_bh(struct net_device *dev)
4419 spin_unlock_bh(&dev->addr_list_lock);
4423 * dev_addrs walker. Should be used only for read access. Call with
4424 * rcu_read_lock held.
4426 #define for_each_dev_addr(dev, ha) \
4427 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4429 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4431 void ether_setup(struct net_device *dev);
4433 /* Support for loadable net-drivers */
4434 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4435 unsigned char name_assign_type,
4436 void (*setup)(struct net_device *),
4437 unsigned int txqs, unsigned int rxqs);
4438 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4439 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4441 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4442 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4445 int register_netdev(struct net_device *dev);
4446 void unregister_netdev(struct net_device *dev);
4448 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4450 /* General hardware address lists handling functions */
4451 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4452 struct netdev_hw_addr_list *from_list, int addr_len);
4453 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4454 struct netdev_hw_addr_list *from_list, int addr_len);
4455 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4456 struct net_device *dev,
4457 int (*sync)(struct net_device *, const unsigned char *),
4458 int (*unsync)(struct net_device *,
4459 const unsigned char *));
4460 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4461 struct net_device *dev,
4462 int (*sync)(struct net_device *,
4463 const unsigned char *, int),
4464 int (*unsync)(struct net_device *,
4465 const unsigned char *, int));
4466 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4467 struct net_device *dev,
4468 int (*unsync)(struct net_device *,
4469 const unsigned char *, int));
4470 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4471 struct net_device *dev,
4472 int (*unsync)(struct net_device *,
4473 const unsigned char *));
4474 void __hw_addr_init(struct netdev_hw_addr_list *list);
4476 /* Functions used for device addresses handling */
4478 __dev_addr_set(struct net_device *dev, const u8 *addr, size_t len)
4480 memcpy(dev->dev_addr, addr, len);
4483 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4485 __dev_addr_set(dev, addr, dev->addr_len);
4489 dev_addr_mod(struct net_device *dev, unsigned int offset,
4490 const u8 *addr, size_t len)
4492 memcpy(&dev->dev_addr[offset], addr, len);
4495 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4496 unsigned char addr_type);
4497 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4498 unsigned char addr_type);
4499 void dev_addr_flush(struct net_device *dev);
4500 int dev_addr_init(struct net_device *dev);
4502 /* Functions used for unicast addresses handling */
4503 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4504 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4505 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4506 int dev_uc_sync(struct net_device *to, struct net_device *from);
4507 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4508 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4509 void dev_uc_flush(struct net_device *dev);
4510 void dev_uc_init(struct net_device *dev);
4513 * __dev_uc_sync - Synchonize device's unicast list
4514 * @dev: device to sync
4515 * @sync: function to call if address should be added
4516 * @unsync: function to call if address should be removed
4518 * Add newly added addresses to the interface, and release
4519 * addresses that have been deleted.
4521 static inline int __dev_uc_sync(struct net_device *dev,
4522 int (*sync)(struct net_device *,
4523 const unsigned char *),
4524 int (*unsync)(struct net_device *,
4525 const unsigned char *))
4527 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4531 * __dev_uc_unsync - Remove synchronized addresses from device
4532 * @dev: device to sync
4533 * @unsync: function to call if address should be removed
4535 * Remove all addresses that were added to the device by dev_uc_sync().
4537 static inline void __dev_uc_unsync(struct net_device *dev,
4538 int (*unsync)(struct net_device *,
4539 const unsigned char *))
4541 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4544 /* Functions used for multicast addresses handling */
4545 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4546 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4547 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4548 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4549 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4550 int dev_mc_sync(struct net_device *to, struct net_device *from);
4551 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4552 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4553 void dev_mc_flush(struct net_device *dev);
4554 void dev_mc_init(struct net_device *dev);
4557 * __dev_mc_sync - Synchonize device's multicast list
4558 * @dev: device to sync
4559 * @sync: function to call if address should be added
4560 * @unsync: function to call if address should be removed
4562 * Add newly added addresses to the interface, and release
4563 * addresses that have been deleted.
4565 static inline int __dev_mc_sync(struct net_device *dev,
4566 int (*sync)(struct net_device *,
4567 const unsigned char *),
4568 int (*unsync)(struct net_device *,
4569 const unsigned char *))
4571 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4575 * __dev_mc_unsync - Remove synchronized addresses from device
4576 * @dev: device to sync
4577 * @unsync: function to call if address should be removed
4579 * Remove all addresses that were added to the device by dev_mc_sync().
4581 static inline void __dev_mc_unsync(struct net_device *dev,
4582 int (*unsync)(struct net_device *,
4583 const unsigned char *))
4585 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4588 /* Functions used for secondary unicast and multicast support */
4589 void dev_set_rx_mode(struct net_device *dev);
4590 void __dev_set_rx_mode(struct net_device *dev);
4591 int dev_set_promiscuity(struct net_device *dev, int inc);
4592 int dev_set_allmulti(struct net_device *dev, int inc);
4593 void netdev_state_change(struct net_device *dev);
4594 void netdev_notify_peers(struct net_device *dev);
4595 void netdev_features_change(struct net_device *dev);
4596 /* Load a device via the kmod */
4597 void dev_load(struct net *net, const char *name);
4598 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4599 struct rtnl_link_stats64 *storage);
4600 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4601 const struct net_device_stats *netdev_stats);
4602 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4603 const struct pcpu_sw_netstats __percpu *netstats);
4605 extern int netdev_max_backlog;
4606 extern int netdev_tstamp_prequeue;
4607 extern int weight_p;
4608 extern int dev_weight_rx_bias;
4609 extern int dev_weight_tx_bias;
4610 extern int dev_rx_weight;
4611 extern int dev_tx_weight;
4612 extern int gro_normal_batch;
4615 NESTED_SYNC_IMM_BIT,
4616 NESTED_SYNC_TODO_BIT,
4619 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4620 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4622 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4623 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4625 struct netdev_nested_priv {
4626 unsigned char flags;
4630 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4631 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4632 struct list_head **iter);
4633 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4634 struct list_head **iter);
4636 #ifdef CONFIG_LOCKDEP
4637 static LIST_HEAD(net_unlink_list);
4639 static inline void net_unlink_todo(struct net_device *dev)
4641 if (list_empty(&dev->unlink_list))
4642 list_add_tail(&dev->unlink_list, &net_unlink_list);
4646 /* iterate through upper list, must be called under RCU read lock */
4647 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4648 for (iter = &(dev)->adj_list.upper, \
4649 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4651 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4653 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4654 int (*fn)(struct net_device *upper_dev,
4655 struct netdev_nested_priv *priv),
4656 struct netdev_nested_priv *priv);
4658 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4659 struct net_device *upper_dev);
4661 bool netdev_has_any_upper_dev(struct net_device *dev);
4663 void *netdev_lower_get_next_private(struct net_device *dev,
4664 struct list_head **iter);
4665 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4666 struct list_head **iter);
4668 #define netdev_for_each_lower_private(dev, priv, iter) \
4669 for (iter = (dev)->adj_list.lower.next, \
4670 priv = netdev_lower_get_next_private(dev, &(iter)); \
4672 priv = netdev_lower_get_next_private(dev, &(iter)))
4674 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4675 for (iter = &(dev)->adj_list.lower, \
4676 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4678 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4680 void *netdev_lower_get_next(struct net_device *dev,
4681 struct list_head **iter);
4683 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4684 for (iter = (dev)->adj_list.lower.next, \
4685 ldev = netdev_lower_get_next(dev, &(iter)); \
4687 ldev = netdev_lower_get_next(dev, &(iter)))
4689 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4690 struct list_head **iter);
4691 int netdev_walk_all_lower_dev(struct net_device *dev,
4692 int (*fn)(struct net_device *lower_dev,
4693 struct netdev_nested_priv *priv),
4694 struct netdev_nested_priv *priv);
4695 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4696 int (*fn)(struct net_device *lower_dev,
4697 struct netdev_nested_priv *priv),
4698 struct netdev_nested_priv *priv);
4700 void *netdev_adjacent_get_private(struct list_head *adj_list);
4701 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4702 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4703 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4704 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4705 struct netlink_ext_ack *extack);
4706 int netdev_master_upper_dev_link(struct net_device *dev,
4707 struct net_device *upper_dev,
4708 void *upper_priv, void *upper_info,
4709 struct netlink_ext_ack *extack);
4710 void netdev_upper_dev_unlink(struct net_device *dev,
4711 struct net_device *upper_dev);
4712 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4713 struct net_device *new_dev,
4714 struct net_device *dev,
4715 struct netlink_ext_ack *extack);
4716 void netdev_adjacent_change_commit(struct net_device *old_dev,
4717 struct net_device *new_dev,
4718 struct net_device *dev);
4719 void netdev_adjacent_change_abort(struct net_device *old_dev,
4720 struct net_device *new_dev,
4721 struct net_device *dev);
4722 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4723 void *netdev_lower_dev_get_private(struct net_device *dev,
4724 struct net_device *lower_dev);
4725 void netdev_lower_state_changed(struct net_device *lower_dev,
4726 void *lower_state_info);
4728 /* RSS keys are 40 or 52 bytes long */
4729 #define NETDEV_RSS_KEY_LEN 52
4730 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4731 void netdev_rss_key_fill(void *buffer, size_t len);
4733 int skb_checksum_help(struct sk_buff *skb);
4734 int skb_crc32c_csum_help(struct sk_buff *skb);
4735 int skb_csum_hwoffload_help(struct sk_buff *skb,
4736 const netdev_features_t features);
4738 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4739 netdev_features_t features, bool tx_path);
4740 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4741 netdev_features_t features);
4743 struct netdev_bonding_info {
4748 struct netdev_notifier_bonding_info {
4749 struct netdev_notifier_info info; /* must be first */
4750 struct netdev_bonding_info bonding_info;
4753 void netdev_bonding_info_change(struct net_device *dev,
4754 struct netdev_bonding_info *bonding_info);
4756 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4757 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4759 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4766 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4768 return __skb_gso_segment(skb, features, true);
4770 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4772 static inline bool can_checksum_protocol(netdev_features_t features,
4775 if (protocol == htons(ETH_P_FCOE))
4776 return !!(features & NETIF_F_FCOE_CRC);
4778 /* Assume this is an IP checksum (not SCTP CRC) */
4780 if (features & NETIF_F_HW_CSUM) {
4781 /* Can checksum everything */
4786 case htons(ETH_P_IP):
4787 return !!(features & NETIF_F_IP_CSUM);
4788 case htons(ETH_P_IPV6):
4789 return !!(features & NETIF_F_IPV6_CSUM);
4796 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4798 static inline void netdev_rx_csum_fault(struct net_device *dev,
4799 struct sk_buff *skb)
4803 /* rx skb timestamps */
4804 void net_enable_timestamp(void);
4805 void net_disable_timestamp(void);
4807 #ifdef CONFIG_PROC_FS
4808 int __init dev_proc_init(void);
4810 #define dev_proc_init() 0
4813 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4814 struct sk_buff *skb, struct net_device *dev,
4817 __this_cpu_write(softnet_data.xmit.more, more);
4818 return ops->ndo_start_xmit(skb, dev);
4821 static inline bool netdev_xmit_more(void)
4823 return __this_cpu_read(softnet_data.xmit.more);
4826 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4827 struct netdev_queue *txq, bool more)
4829 const struct net_device_ops *ops = dev->netdev_ops;
4832 rc = __netdev_start_xmit(ops, skb, dev, more);
4833 if (rc == NETDEV_TX_OK)
4834 txq_trans_update(txq);
4839 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4841 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4844 extern const struct kobj_ns_type_operations net_ns_type_operations;
4846 const char *netdev_drivername(const struct net_device *dev);
4848 void linkwatch_run_queue(void);
4850 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4851 netdev_features_t f2)
4853 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4854 if (f1 & NETIF_F_HW_CSUM)
4855 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4857 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4863 static inline netdev_features_t netdev_get_wanted_features(
4864 struct net_device *dev)
4866 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4868 netdev_features_t netdev_increment_features(netdev_features_t all,
4869 netdev_features_t one, netdev_features_t mask);
4871 /* Allow TSO being used on stacked device :
4872 * Performing the GSO segmentation before last device
4873 * is a performance improvement.
4875 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4876 netdev_features_t mask)
4878 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4881 int __netdev_update_features(struct net_device *dev);
4882 void netdev_update_features(struct net_device *dev);
4883 void netdev_change_features(struct net_device *dev);
4885 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4886 struct net_device *dev);
4888 netdev_features_t passthru_features_check(struct sk_buff *skb,
4889 struct net_device *dev,
4890 netdev_features_t features);
4891 netdev_features_t netif_skb_features(struct sk_buff *skb);
4893 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4895 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4897 /* check flags correspondence */
4898 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4899 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4900 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4901 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4902 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4903 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4904 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4905 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4906 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4907 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4908 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4909 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4910 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4911 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4912 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4913 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4914 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4915 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4916 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4918 return (features & feature) == feature;
4921 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4923 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4924 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4927 static inline bool netif_needs_gso(struct sk_buff *skb,
4928 netdev_features_t features)
4930 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4931 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4932 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4935 static inline void netif_set_gso_max_size(struct net_device *dev,
4938 dev->gso_max_size = size;
4941 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4942 int pulled_hlen, u16 mac_offset,
4945 skb->protocol = protocol;
4946 skb->encapsulation = 1;
4947 skb_push(skb, pulled_hlen);
4948 skb_reset_transport_header(skb);
4949 skb->mac_header = mac_offset;
4950 skb->network_header = skb->mac_header + mac_len;
4951 skb->mac_len = mac_len;
4954 static inline bool netif_is_macsec(const struct net_device *dev)
4956 return dev->priv_flags & IFF_MACSEC;
4959 static inline bool netif_is_macvlan(const struct net_device *dev)
4961 return dev->priv_flags & IFF_MACVLAN;
4964 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4966 return dev->priv_flags & IFF_MACVLAN_PORT;
4969 static inline bool netif_is_bond_master(const struct net_device *dev)
4971 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4974 static inline bool netif_is_bond_slave(const struct net_device *dev)
4976 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4979 static inline bool netif_supports_nofcs(struct net_device *dev)
4981 return dev->priv_flags & IFF_SUPP_NOFCS;
4984 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4986 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4989 static inline bool netif_is_l3_master(const struct net_device *dev)
4991 return dev->priv_flags & IFF_L3MDEV_MASTER;
4994 static inline bool netif_is_l3_slave(const struct net_device *dev)
4996 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4999 static inline bool netif_is_bridge_master(const struct net_device *dev)
5001 return dev->priv_flags & IFF_EBRIDGE;
5004 static inline bool netif_is_bridge_port(const struct net_device *dev)
5006 return dev->priv_flags & IFF_BRIDGE_PORT;
5009 static inline bool netif_is_ovs_master(const struct net_device *dev)
5011 return dev->priv_flags & IFF_OPENVSWITCH;
5014 static inline bool netif_is_ovs_port(const struct net_device *dev)
5016 return dev->priv_flags & IFF_OVS_DATAPATH;
5019 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5021 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5024 static inline bool netif_is_team_master(const struct net_device *dev)
5026 return dev->priv_flags & IFF_TEAM;
5029 static inline bool netif_is_team_port(const struct net_device *dev)
5031 return dev->priv_flags & IFF_TEAM_PORT;
5034 static inline bool netif_is_lag_master(const struct net_device *dev)
5036 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5039 static inline bool netif_is_lag_port(const struct net_device *dev)
5041 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5044 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5046 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5049 static inline bool netif_is_failover(const struct net_device *dev)
5051 return dev->priv_flags & IFF_FAILOVER;
5054 static inline bool netif_is_failover_slave(const struct net_device *dev)
5056 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5059 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5060 static inline void netif_keep_dst(struct net_device *dev)
5062 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5065 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5066 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5068 /* TODO: reserve and use an additional IFF bit, if we get more users */
5069 return dev->priv_flags & IFF_MACSEC;
5072 extern struct pernet_operations __net_initdata loopback_net_ops;
5074 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5076 /* netdev_printk helpers, similar to dev_printk */
5078 static inline const char *netdev_name(const struct net_device *dev)
5080 if (!dev->name[0] || strchr(dev->name, '%'))
5081 return "(unnamed net_device)";
5085 static inline bool netdev_unregistering(const struct net_device *dev)
5087 return dev->reg_state == NETREG_UNREGISTERING;
5090 static inline const char *netdev_reg_state(const struct net_device *dev)
5092 switch (dev->reg_state) {
5093 case NETREG_UNINITIALIZED: return " (uninitialized)";
5094 case NETREG_REGISTERED: return "";
5095 case NETREG_UNREGISTERING: return " (unregistering)";
5096 case NETREG_UNREGISTERED: return " (unregistered)";
5097 case NETREG_RELEASED: return " (released)";
5098 case NETREG_DUMMY: return " (dummy)";
5101 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5102 return " (unknown)";
5105 __printf(3, 4) __cold
5106 void netdev_printk(const char *level, const struct net_device *dev,
5107 const char *format, ...);
5108 __printf(2, 3) __cold
5109 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5110 __printf(2, 3) __cold
5111 void netdev_alert(const struct net_device *dev, const char *format, ...);
5112 __printf(2, 3) __cold
5113 void netdev_crit(const struct net_device *dev, const char *format, ...);
5114 __printf(2, 3) __cold
5115 void netdev_err(const struct net_device *dev, const char *format, ...);
5116 __printf(2, 3) __cold
5117 void netdev_warn(const struct net_device *dev, const char *format, ...);
5118 __printf(2, 3) __cold
5119 void netdev_notice(const struct net_device *dev, const char *format, ...);
5120 __printf(2, 3) __cold
5121 void netdev_info(const struct net_device *dev, const char *format, ...);
5123 #define netdev_level_once(level, dev, fmt, ...) \
5125 static bool __print_once __read_mostly; \
5127 if (!__print_once) { \
5128 __print_once = true; \
5129 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5133 #define netdev_emerg_once(dev, fmt, ...) \
5134 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5135 #define netdev_alert_once(dev, fmt, ...) \
5136 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5137 #define netdev_crit_once(dev, fmt, ...) \
5138 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5139 #define netdev_err_once(dev, fmt, ...) \
5140 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5141 #define netdev_warn_once(dev, fmt, ...) \
5142 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5143 #define netdev_notice_once(dev, fmt, ...) \
5144 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5145 #define netdev_info_once(dev, fmt, ...) \
5146 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5148 #define MODULE_ALIAS_NETDEV(device) \
5149 MODULE_ALIAS("netdev-" device)
5151 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5152 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5153 #define netdev_dbg(__dev, format, args...) \
5155 dynamic_netdev_dbg(__dev, format, ##args); \
5157 #elif defined(DEBUG)
5158 #define netdev_dbg(__dev, format, args...) \
5159 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5161 #define netdev_dbg(__dev, format, args...) \
5164 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5168 #if defined(VERBOSE_DEBUG)
5169 #define netdev_vdbg netdev_dbg
5172 #define netdev_vdbg(dev, format, args...) \
5175 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5181 * netdev_WARN() acts like dev_printk(), but with the key difference
5182 * of using a WARN/WARN_ON to get the message out, including the
5183 * file/line information and a backtrace.
5185 #define netdev_WARN(dev, format, args...) \
5186 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5187 netdev_reg_state(dev), ##args)
5189 #define netdev_WARN_ONCE(dev, format, args...) \
5190 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5191 netdev_reg_state(dev), ##args)
5193 /* netif printk helpers, similar to netdev_printk */
5195 #define netif_printk(priv, type, level, dev, fmt, args...) \
5197 if (netif_msg_##type(priv)) \
5198 netdev_printk(level, (dev), fmt, ##args); \
5201 #define netif_level(level, priv, type, dev, fmt, args...) \
5203 if (netif_msg_##type(priv)) \
5204 netdev_##level(dev, fmt, ##args); \
5207 #define netif_emerg(priv, type, dev, fmt, args...) \
5208 netif_level(emerg, priv, type, dev, fmt, ##args)
5209 #define netif_alert(priv, type, dev, fmt, args...) \
5210 netif_level(alert, priv, type, dev, fmt, ##args)
5211 #define netif_crit(priv, type, dev, fmt, args...) \
5212 netif_level(crit, priv, type, dev, fmt, ##args)
5213 #define netif_err(priv, type, dev, fmt, args...) \
5214 netif_level(err, priv, type, dev, fmt, ##args)
5215 #define netif_warn(priv, type, dev, fmt, args...) \
5216 netif_level(warn, priv, type, dev, fmt, ##args)
5217 #define netif_notice(priv, type, dev, fmt, args...) \
5218 netif_level(notice, priv, type, dev, fmt, ##args)
5219 #define netif_info(priv, type, dev, fmt, args...) \
5220 netif_level(info, priv, type, dev, fmt, ##args)
5222 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5223 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5224 #define netif_dbg(priv, type, netdev, format, args...) \
5226 if (netif_msg_##type(priv)) \
5227 dynamic_netdev_dbg(netdev, format, ##args); \
5229 #elif defined(DEBUG)
5230 #define netif_dbg(priv, type, dev, format, args...) \
5231 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5233 #define netif_dbg(priv, type, dev, format, args...) \
5236 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5241 /* if @cond then downgrade to debug, else print at @level */
5242 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5245 netif_dbg(priv, type, netdev, fmt, ##args); \
5247 netif_ ## level(priv, type, netdev, fmt, ##args); \
5250 #if defined(VERBOSE_DEBUG)
5251 #define netif_vdbg netif_dbg
5253 #define netif_vdbg(priv, type, dev, format, args...) \
5256 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5262 * The list of packet types we will receive (as opposed to discard)
5263 * and the routines to invoke.
5265 * Why 16. Because with 16 the only overlap we get on a hash of the
5266 * low nibble of the protocol value is RARP/SNAP/X.25.
5280 #define PTYPE_HASH_SIZE (16)
5281 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5283 extern struct net_device *blackhole_netdev;
5285 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5286 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5287 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5288 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5289 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5291 #endif /* _LINUX_NETDEVICE_H */