1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <net/net_namespace.h>
39 #include <net/dcbnl.h>
41 #include <net/netprio_cgroup.h>
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <linux/hashtable.h>
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 struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
196 #include <linux/cache.h>
197 #include <linux/skbuff.h>
200 #include <linux/static_key.h>
201 extern struct static_key_false rps_needed;
202 extern struct static_key_false rfs_needed;
209 struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
213 #define NETDEV_HW_ADDR_T_LAN 1
214 #define NETDEV_HW_ADDR_T_SAN 2
215 #define NETDEV_HW_ADDR_T_UNICAST 3
216 #define NETDEV_HW_ADDR_T_MULTICAST 4
221 struct rcu_head rcu_head;
224 struct netdev_hw_addr_list {
225 struct list_head list;
229 #define netdev_hw_addr_list_count(l) ((l)->count)
230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231 #define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236 #define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241 #define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
248 /* cached hardware header; allow for machine alignment needs. */
249 #define HH_DATA_MOD 16
250 #define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252 #define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
265 #define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
283 /* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
288 enum netdev_state_t {
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
302 struct netdev_boot_setup {
306 #define NETDEV_BOOT_SETUP_MAX 8
308 int __init netdev_boot_setup(char *str);
311 struct list_head list;
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
319 #define GRO_HASH_BUCKETS 8
322 * Structure for NAPI scheduling similar to tasklet but with weighting
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
331 struct list_head poll_list;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338 #ifdef CONFIG_NETPOLL
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
350 struct task_struct *thread;
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_MISSED, /* reschedule a napi */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
358 NAPI_STATE_LISTED, /* NAPI added to system lists */
359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
363 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
367 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
368 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
369 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
370 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
371 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
372 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
373 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
374 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
375 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
376 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
386 typedef enum gro_result gro_result_t;
389 * enum rx_handler_result - Possible return values for rx_handlers.
390 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
392 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
393 * case skb->dev was changed by rx_handler.
394 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
395 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
397 * rx_handlers are functions called from inside __netif_receive_skb(), to do
398 * special processing of the skb, prior to delivery to protocol handlers.
400 * Currently, a net_device can only have a single rx_handler registered. Trying
401 * to register a second rx_handler will return -EBUSY.
403 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
404 * To unregister a rx_handler on a net_device, use
405 * netdev_rx_handler_unregister().
407 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
410 * If the rx_handler consumed the skb in some way, it should return
411 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
412 * the skb to be delivered in some other way.
414 * If the rx_handler changed skb->dev, to divert the skb to another
415 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
416 * new device will be called if it exists.
418 * If the rx_handler decides the skb should be ignored, it should return
419 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
420 * are registered on exact device (ptype->dev == skb->dev).
422 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
423 * delivered, it should return RX_HANDLER_PASS.
425 * A device without a registered rx_handler will behave as if rx_handler
426 * returned RX_HANDLER_PASS.
429 enum rx_handler_result {
435 typedef enum rx_handler_result rx_handler_result_t;
436 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
438 void __napi_schedule(struct napi_struct *n);
439 void __napi_schedule_irqoff(struct napi_struct *n);
441 static inline bool napi_disable_pending(struct napi_struct *n)
443 return test_bit(NAPI_STATE_DISABLE, &n->state);
446 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
448 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
451 bool napi_schedule_prep(struct napi_struct *n);
454 * napi_schedule - schedule NAPI poll
457 * Schedule NAPI poll routine to be called if it is not already
460 static inline void napi_schedule(struct napi_struct *n)
462 if (napi_schedule_prep(n))
467 * napi_schedule_irqoff - schedule NAPI poll
470 * Variant of napi_schedule(), assuming hard irqs are masked.
472 static inline void napi_schedule_irqoff(struct napi_struct *n)
474 if (napi_schedule_prep(n))
475 __napi_schedule_irqoff(n);
478 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
479 static inline bool napi_reschedule(struct napi_struct *napi)
481 if (napi_schedule_prep(napi)) {
482 __napi_schedule(napi);
488 bool napi_complete_done(struct napi_struct *n, int work_done);
490 * napi_complete - NAPI processing complete
493 * Mark NAPI processing as complete.
494 * Consider using napi_complete_done() instead.
495 * Return false if device should avoid rearming interrupts.
497 static inline bool napi_complete(struct napi_struct *n)
499 return napi_complete_done(n, 0);
502 int dev_set_threaded(struct net_device *dev, bool threaded);
505 * napi_disable - prevent NAPI from scheduling
508 * Stop NAPI from being scheduled on this context.
509 * Waits till any outstanding processing completes.
511 void napi_disable(struct napi_struct *n);
513 void napi_enable(struct napi_struct *n);
516 * napi_synchronize - wait until NAPI is not running
519 * Wait until NAPI is done being scheduled on this context.
520 * Waits till any outstanding processing completes but
521 * does not disable future activations.
523 static inline void napi_synchronize(const struct napi_struct *n)
525 if (IS_ENABLED(CONFIG_SMP))
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
533 * napi_if_scheduled_mark_missed - if napi is running, set the
537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
540 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
542 unsigned long val, new;
545 val = READ_ONCE(n->state);
546 if (val & NAPIF_STATE_DISABLE)
549 if (!(val & NAPIF_STATE_SCHED))
552 new = val | NAPIF_STATE_MISSED;
553 } while (cmpxchg(&n->state, val, new) != val);
558 enum netdev_queue_state_t {
559 __QUEUE_STATE_DRV_XOFF,
560 __QUEUE_STATE_STACK_XOFF,
561 __QUEUE_STATE_FROZEN,
564 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
565 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
566 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
568 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
569 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
571 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
576 * netif_tx_* functions below are used to manipulate this flag. The
577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
578 * queue independently. The netif_xmit_*stopped functions below are called
579 * to check if the queue has been stopped by the driver or stack (either
580 * of the XOFF bits are set in the state). Drivers should not need to call
581 * netif_xmit*stopped functions, they should only be using netif_tx_*.
584 struct netdev_queue {
588 struct net_device *dev;
589 struct Qdisc __rcu *qdisc;
590 struct Qdisc *qdisc_sleeping;
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 unsigned long tx_maxrate;
599 * Number of TX timeouts for this queue
600 * (/sys/class/net/DEV/Q/trans_timeout)
602 unsigned long trans_timeout;
604 /* Subordinate device that the queue has been assigned to */
605 struct net_device *sb_dev;
606 #ifdef CONFIG_XDP_SOCKETS
607 struct xsk_buff_pool *pool;
612 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
615 * Time (in jiffies) of last Tx
617 unsigned long trans_start;
624 } ____cacheline_aligned_in_smp;
626 extern int sysctl_fb_tunnels_only_for_init_net;
627 extern int sysctl_devconf_inherit_init_net;
630 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
631 * == 1 : For initns only
634 static inline bool net_has_fallback_tunnels(const struct net *net)
636 return !IS_ENABLED(CONFIG_SYSCTL) ||
637 !sysctl_fb_tunnels_only_for_init_net ||
638 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
674 struct rps_dev_flow {
677 unsigned int last_qtail;
679 #define RPS_NO_FILTER 0xffff
682 * The rps_dev_flow_table structure contains a table of flow mappings.
684 struct rps_dev_flow_table {
687 struct rps_dev_flow flows[];
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
702 struct rps_sock_flow_table {
705 u32 ents[] ____cacheline_aligned_in_smp;
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
709 #define RPS_NO_CPU 0xffff
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
724 if (table->ents[index] != val)
725 table->ents[index] = val;
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 #endif /* CONFIG_RPS */
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xsk_buff_pool *pool;
747 } ____cacheline_aligned_in_smp;
750 * RX queue sysfs structures and functions.
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
759 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
768 * This structure holds an XPS map which can be of variable length. The
769 * map is an array of queues.
773 unsigned int alloc_len;
777 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
778 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
779 - sizeof(struct xps_map)) / sizeof(u16))
782 * This structure holds all XPS maps for device. Maps are indexed by CPU.
784 * We keep track of the number of cpus/rxqs used when the struct is allocated,
785 * in nr_ids. This will help not accessing out-of-bound memory.
787 * We keep track of the number of traffic classes used when the struct is
788 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
789 * not crossing its upper bound, as the original dev->num_tc can be updated in
792 struct xps_dev_maps {
796 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
799 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
800 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
802 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
803 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
805 #endif /* CONFIG_XPS */
807 #define TC_MAX_QUEUE 16
808 #define TC_BITMASK 15
809 /* HW offloaded queuing disciplines txq count and offset maps */
810 struct netdev_tc_txq {
815 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
817 * This structure is to hold information about the device
818 * configured to run FCoE protocol stack.
820 struct netdev_fcoe_hbainfo {
821 char manufacturer[64];
822 char serial_number[64];
823 char hardware_version[64];
824 char driver_version[64];
825 char optionrom_version[64];
826 char firmware_version[64];
828 char model_description[256];
832 #define MAX_PHYS_ITEM_ID_LEN 32
834 /* This structure holds a unique identifier to identify some
835 * physical item (port for example) used by a netdevice.
837 struct netdev_phys_item_id {
838 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
839 unsigned char id_len;
842 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
843 struct netdev_phys_item_id *b)
845 return a->id_len == b->id_len &&
846 memcmp(a->id, b->id, a->id_len) == 0;
849 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
851 struct net_device *sb_dev);
853 enum net_device_path_type {
854 DEV_PATH_ETHERNET = 0,
861 struct net_device_path {
862 enum net_device_path_type type;
863 const struct net_device *dev;
872 DEV_PATH_BR_VLAN_KEEP,
873 DEV_PATH_BR_VLAN_TAG,
874 DEV_PATH_BR_VLAN_UNTAG,
875 DEV_PATH_BR_VLAN_UNTAG_HW,
887 #define NET_DEVICE_PATH_STACK_MAX 5
888 #define NET_DEVICE_PATH_VLAN_MAX 2
890 struct net_device_path_stack {
892 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
895 struct net_device_path_ctx {
896 const struct net_device *dev;
903 } vlan[NET_DEVICE_PATH_VLAN_MAX];
907 TC_SETUP_QDISC_MQPRIO,
910 TC_SETUP_CLSMATCHALL,
920 TC_SETUP_QDISC_TAPRIO,
928 /* These structures hold the attributes of bpf state that are being passed
929 * to the netdevice through the bpf op.
931 enum bpf_netdev_command {
932 /* Set or clear a bpf program used in the earliest stages of packet
933 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
934 * is responsible for calling bpf_prog_put on any old progs that are
935 * stored. In case of error, the callee need not release the new prog
936 * reference, but on success it takes ownership and must bpf_prog_put
937 * when it is no longer used.
941 /* BPF program for offload callbacks, invoked at program load time. */
942 BPF_OFFLOAD_MAP_ALLOC,
943 BPF_OFFLOAD_MAP_FREE,
947 struct bpf_prog_offload_ops;
948 struct netlink_ext_ack;
950 struct xdp_dev_bulk_queue;
960 struct bpf_xdp_entity {
961 struct bpf_prog *prog;
962 struct bpf_xdp_link *link;
966 enum bpf_netdev_command command;
971 struct bpf_prog *prog;
972 struct netlink_ext_ack *extack;
974 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
976 struct bpf_offloaded_map *offmap;
978 /* XDP_SETUP_XSK_POOL */
980 struct xsk_buff_pool *pool;
986 /* Flags for ndo_xsk_wakeup. */
987 #define XDP_WAKEUP_RX (1 << 0)
988 #define XDP_WAKEUP_TX (1 << 1)
990 #ifdef CONFIG_XFRM_OFFLOAD
992 int (*xdo_dev_state_add) (struct xfrm_state *x);
993 void (*xdo_dev_state_delete) (struct xfrm_state *x);
994 void (*xdo_dev_state_free) (struct xfrm_state *x);
995 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
996 struct xfrm_state *x);
997 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1001 struct dev_ifalias {
1002 struct rcu_head rcuhead;
1009 struct netdev_name_node {
1010 struct hlist_node hlist;
1011 struct list_head list;
1012 struct net_device *dev;
1016 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1017 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1019 struct netdev_net_notifier {
1020 struct list_head list;
1021 struct notifier_block *nb;
1025 * This structure defines the management hooks for network devices.
1026 * The following hooks can be defined; unless noted otherwise, they are
1027 * optional and can be filled with a null pointer.
1029 * int (*ndo_init)(struct net_device *dev);
1030 * This function is called once when a network device is registered.
1031 * The network device can use this for any late stage initialization
1032 * or semantic validation. It can fail with an error code which will
1033 * be propagated back to register_netdev.
1035 * void (*ndo_uninit)(struct net_device *dev);
1036 * This function is called when device is unregistered or when registration
1037 * fails. It is not called if init fails.
1039 * int (*ndo_open)(struct net_device *dev);
1040 * This function is called when a network device transitions to the up
1043 * int (*ndo_stop)(struct net_device *dev);
1044 * This function is called when a network device transitions to the down
1047 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1048 * struct net_device *dev);
1049 * Called when a packet needs to be transmitted.
1050 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1051 * the queue before that can happen; it's for obsolete devices and weird
1052 * corner cases, but the stack really does a non-trivial amount
1053 * of useless work if you return NETDEV_TX_BUSY.
1054 * Required; cannot be NULL.
1056 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1057 * struct net_device *dev
1058 * netdev_features_t features);
1059 * Called by core transmit path to determine if device is capable of
1060 * performing offload operations on a given packet. This is to give
1061 * the device an opportunity to implement any restrictions that cannot
1062 * be otherwise expressed by feature flags. The check is called with
1063 * the set of features that the stack has calculated and it returns
1064 * those the driver believes to be appropriate.
1066 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1067 * struct net_device *sb_dev);
1068 * Called to decide which queue to use when device supports multiple
1071 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1072 * This function is called to allow device receiver to make
1073 * changes to configuration when multicast or promiscuous is enabled.
1075 * void (*ndo_set_rx_mode)(struct net_device *dev);
1076 * This function is called device changes address list filtering.
1077 * If driver handles unicast address filtering, it should set
1078 * IFF_UNICAST_FLT in its priv_flags.
1080 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1081 * This function is called when the Media Access Control address
1082 * needs to be changed. If this interface is not defined, the
1083 * MAC address can not be changed.
1085 * int (*ndo_validate_addr)(struct net_device *dev);
1086 * Test if Media Access Control address is valid for the device.
1088 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1089 * Called when a user requests an ioctl which can't be handled by
1090 * the generic interface code. If not defined ioctls return
1091 * not supported error code.
1093 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1094 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1095 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1097 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1098 * Used to set network devices bus interface parameters. This interface
1099 * is retained for legacy reasons; new devices should use the bus
1100 * interface (PCI) for low level management.
1102 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1103 * Called when a user wants to change the Maximum Transfer Unit
1106 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1107 * Callback used when the transmitter has not made any progress
1108 * for dev->watchdog ticks.
1110 * void (*ndo_get_stats64)(struct net_device *dev,
1111 * struct rtnl_link_stats64 *storage);
1112 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1113 * Called when a user wants to get the network device usage
1114 * statistics. Drivers must do one of the following:
1115 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1116 * rtnl_link_stats64 structure passed by the caller.
1117 * 2. Define @ndo_get_stats to update a net_device_stats structure
1118 * (which should normally be dev->stats) and return a pointer to
1119 * it. The structure may be changed asynchronously only if each
1120 * field is written atomically.
1121 * 3. Update dev->stats asynchronously and atomically, and define
1122 * neither operation.
1124 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1125 * Return true if this device supports offload stats of this attr_id.
1127 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1129 * Get statistics for offload operations by attr_id. Write it into the
1130 * attr_data pointer.
1132 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1133 * If device supports VLAN filtering this function is called when a
1134 * VLAN id is registered.
1136 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1137 * If device supports VLAN filtering this function is called when a
1138 * VLAN id is unregistered.
1140 * void (*ndo_poll_controller)(struct net_device *dev);
1142 * SR-IOV management functions.
1143 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1144 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1145 * u8 qos, __be16 proto);
1146 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1148 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1149 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1150 * int (*ndo_get_vf_config)(struct net_device *dev,
1151 * int vf, struct ifla_vf_info *ivf);
1152 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1153 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1154 * struct nlattr *port[]);
1156 * Enable or disable the VF ability to query its RSS Redirection Table and
1157 * Hash Key. This is needed since on some devices VF share this information
1158 * with PF and querying it may introduce a theoretical security risk.
1159 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1160 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1161 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1163 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1164 * This is always called from the stack with the rtnl lock held and netif
1165 * tx queues stopped. This allows the netdevice to perform queue
1166 * management safely.
1168 * Fiber Channel over Ethernet (FCoE) offload functions.
1169 * int (*ndo_fcoe_enable)(struct net_device *dev);
1170 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1171 * so the underlying device can perform whatever needed configuration or
1172 * initialization to support acceleration of FCoE traffic.
1174 * int (*ndo_fcoe_disable)(struct net_device *dev);
1175 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1176 * so the underlying device can perform whatever needed clean-ups to
1177 * stop supporting acceleration of FCoE traffic.
1179 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1180 * struct scatterlist *sgl, unsigned int sgc);
1181 * Called when the FCoE Initiator wants to initialize an I/O that
1182 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1183 * perform necessary setup and returns 1 to indicate the device is set up
1184 * successfully to perform DDP on this I/O, otherwise this returns 0.
1186 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1187 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1188 * indicated by the FC exchange id 'xid', so the underlying device can
1189 * clean up and reuse resources for later DDP requests.
1191 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1192 * struct scatterlist *sgl, unsigned int sgc);
1193 * Called when the FCoE Target wants to initialize an I/O that
1194 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1195 * perform necessary setup and returns 1 to indicate the device is set up
1196 * successfully to perform DDP on this I/O, otherwise this returns 0.
1198 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1199 * struct netdev_fcoe_hbainfo *hbainfo);
1200 * Called when the FCoE Protocol stack wants information on the underlying
1201 * device. This information is utilized by the FCoE protocol stack to
1202 * register attributes with Fiber Channel management service as per the
1203 * FC-GS Fabric Device Management Information(FDMI) specification.
1205 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1206 * Called when the underlying device wants to override default World Wide
1207 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1208 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1209 * protocol stack to use.
1212 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1213 * u16 rxq_index, u32 flow_id);
1214 * Set hardware filter for RFS. rxq_index is the target queue index;
1215 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1216 * Return the filter ID on success, or a negative error code.
1218 * Slave management functions (for bridge, bonding, etc).
1219 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1220 * Called to make another netdev an underling.
1222 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1223 * Called to release previously enslaved netdev.
1225 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1226 * struct sk_buff *skb,
1228 * Get the xmit slave of master device. If all_slaves is true, function
1229 * assume all the slaves can transmit.
1231 * Feature/offload setting functions.
1232 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1233 * netdev_features_t features);
1234 * Adjusts the requested feature flags according to device-specific
1235 * constraints, and returns the resulting flags. Must not modify
1238 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1239 * Called to update device configuration to new features. Passed
1240 * feature set might be less than what was returned by ndo_fix_features()).
1241 * Must return >0 or -errno if it changed dev->features itself.
1243 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1244 * struct net_device *dev,
1245 * const unsigned char *addr, u16 vid, u16 flags,
1246 * struct netlink_ext_ack *extack);
1247 * Adds an FDB entry to dev for addr.
1248 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1249 * struct net_device *dev,
1250 * const unsigned char *addr, u16 vid)
1251 * Deletes the FDB entry from dev coresponding to addr.
1252 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1253 * struct net_device *dev, struct net_device *filter_dev,
1255 * Used to add FDB entries to dump requests. Implementers should add
1256 * entries to skb and update idx with the number of entries.
1258 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1259 * u16 flags, struct netlink_ext_ack *extack)
1260 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1261 * struct net_device *dev, u32 filter_mask,
1263 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1266 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1267 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1268 * which do not represent real hardware may define this to allow their
1269 * userspace components to manage their virtual carrier state. Devices
1270 * that determine carrier state from physical hardware properties (eg
1271 * network cables) or protocol-dependent mechanisms (eg
1272 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1274 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1275 * struct netdev_phys_item_id *ppid);
1276 * Called to get ID of physical port of this device. If driver does
1277 * not implement this, it is assumed that the hw is not able to have
1278 * multiple net devices on single physical port.
1280 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1281 * struct netdev_phys_item_id *ppid)
1282 * Called to get the parent ID of the physical port of this device.
1284 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1285 * struct net_device *dev)
1286 * Called by upper layer devices to accelerate switching or other
1287 * station functionality into hardware. 'pdev is the lowerdev
1288 * to use for the offload and 'dev' is the net device that will
1289 * back the offload. Returns a pointer to the private structure
1290 * the upper layer will maintain.
1291 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1292 * Called by upper layer device to delete the station created
1293 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1294 * the station and priv is the structure returned by the add
1296 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1297 * int queue_index, u32 maxrate);
1298 * Called when a user wants to set a max-rate limitation of specific
1300 * int (*ndo_get_iflink)(const struct net_device *dev);
1301 * Called to get the iflink value of this device.
1302 * void (*ndo_change_proto_down)(struct net_device *dev,
1304 * This function is used to pass protocol port error state information
1305 * to the switch driver. The switch driver can react to the proto_down
1306 * by doing a phys down on the associated switch port.
1307 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1308 * This function is used to get egress tunnel information for given skb.
1309 * This is useful for retrieving outer tunnel header parameters while
1311 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1312 * This function is used to specify the headroom that the skb must
1313 * consider when allocation skb during packet reception. Setting
1314 * appropriate rx headroom value allows avoiding skb head copy on
1315 * forward. Setting a negative value resets the rx headroom to the
1317 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1318 * This function is used to set or query state related to XDP on the
1319 * netdevice and manage BPF offload. See definition of
1320 * enum bpf_netdev_command for details.
1321 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1323 * This function is used to submit @n XDP packets for transmit on a
1324 * netdevice. Returns number of frames successfully transmitted, frames
1325 * that got dropped are freed/returned via xdp_return_frame().
1326 * Returns negative number, means general error invoking ndo, meaning
1327 * no frames were xmit'ed and core-caller will free all frames.
1328 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1329 * This function is used to wake up the softirq, ksoftirqd or kthread
1330 * responsible for sending and/or receiving packets on a specific
1331 * queue id bound to an AF_XDP socket. The flags field specifies if
1332 * only RX, only Tx, or both should be woken up using the flags
1333 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1334 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1335 * Get devlink port instance associated with a given netdev.
1336 * Called with a reference on the netdevice and devlink locks only,
1337 * rtnl_lock is not held.
1338 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1340 * Add, change, delete or get information on an IPv4 tunnel.
1341 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1342 * If a device is paired with a peer device, return the peer instance.
1343 * The caller must be under RCU read context.
1344 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1345 * Get the forwarding path to reach the real device from the HW destination address
1347 struct net_device_ops {
1348 int (*ndo_init)(struct net_device *dev);
1349 void (*ndo_uninit)(struct net_device *dev);
1350 int (*ndo_open)(struct net_device *dev);
1351 int (*ndo_stop)(struct net_device *dev);
1352 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1353 struct net_device *dev);
1354 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1355 struct net_device *dev,
1356 netdev_features_t features);
1357 u16 (*ndo_select_queue)(struct net_device *dev,
1358 struct sk_buff *skb,
1359 struct net_device *sb_dev);
1360 void (*ndo_change_rx_flags)(struct net_device *dev,
1362 void (*ndo_set_rx_mode)(struct net_device *dev);
1363 int (*ndo_set_mac_address)(struct net_device *dev,
1365 int (*ndo_validate_addr)(struct net_device *dev);
1366 int (*ndo_do_ioctl)(struct net_device *dev,
1367 struct ifreq *ifr, int cmd);
1368 int (*ndo_eth_ioctl)(struct net_device *dev,
1369 struct ifreq *ifr, int cmd);
1370 int (*ndo_siocdevprivate)(struct net_device *dev,
1372 void __user *data, int cmd);
1373 int (*ndo_set_config)(struct net_device *dev,
1375 int (*ndo_change_mtu)(struct net_device *dev,
1377 int (*ndo_neigh_setup)(struct net_device *dev,
1378 struct neigh_parms *);
1379 void (*ndo_tx_timeout) (struct net_device *dev,
1380 unsigned int txqueue);
1382 void (*ndo_get_stats64)(struct net_device *dev,
1383 struct rtnl_link_stats64 *storage);
1384 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1385 int (*ndo_get_offload_stats)(int attr_id,
1386 const struct net_device *dev,
1388 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1390 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1391 __be16 proto, u16 vid);
1392 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1393 __be16 proto, u16 vid);
1394 #ifdef CONFIG_NET_POLL_CONTROLLER
1395 void (*ndo_poll_controller)(struct net_device *dev);
1396 int (*ndo_netpoll_setup)(struct net_device *dev,
1397 struct netpoll_info *info);
1398 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1400 int (*ndo_set_vf_mac)(struct net_device *dev,
1401 int queue, u8 *mac);
1402 int (*ndo_set_vf_vlan)(struct net_device *dev,
1403 int queue, u16 vlan,
1404 u8 qos, __be16 proto);
1405 int (*ndo_set_vf_rate)(struct net_device *dev,
1406 int vf, int min_tx_rate,
1408 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1409 int vf, bool setting);
1410 int (*ndo_set_vf_trust)(struct net_device *dev,
1411 int vf, bool setting);
1412 int (*ndo_get_vf_config)(struct net_device *dev,
1414 struct ifla_vf_info *ivf);
1415 int (*ndo_set_vf_link_state)(struct net_device *dev,
1416 int vf, int link_state);
1417 int (*ndo_get_vf_stats)(struct net_device *dev,
1419 struct ifla_vf_stats
1421 int (*ndo_set_vf_port)(struct net_device *dev,
1423 struct nlattr *port[]);
1424 int (*ndo_get_vf_port)(struct net_device *dev,
1425 int vf, struct sk_buff *skb);
1426 int (*ndo_get_vf_guid)(struct net_device *dev,
1428 struct ifla_vf_guid *node_guid,
1429 struct ifla_vf_guid *port_guid);
1430 int (*ndo_set_vf_guid)(struct net_device *dev,
1433 int (*ndo_set_vf_rss_query_en)(
1434 struct net_device *dev,
1435 int vf, bool setting);
1436 int (*ndo_setup_tc)(struct net_device *dev,
1437 enum tc_setup_type type,
1439 #if IS_ENABLED(CONFIG_FCOE)
1440 int (*ndo_fcoe_enable)(struct net_device *dev);
1441 int (*ndo_fcoe_disable)(struct net_device *dev);
1442 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1444 struct scatterlist *sgl,
1446 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1448 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1450 struct scatterlist *sgl,
1452 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1453 struct netdev_fcoe_hbainfo *hbainfo);
1456 #if IS_ENABLED(CONFIG_LIBFCOE)
1457 #define NETDEV_FCOE_WWNN 0
1458 #define NETDEV_FCOE_WWPN 1
1459 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1460 u64 *wwn, int type);
1463 #ifdef CONFIG_RFS_ACCEL
1464 int (*ndo_rx_flow_steer)(struct net_device *dev,
1465 const struct sk_buff *skb,
1469 int (*ndo_add_slave)(struct net_device *dev,
1470 struct net_device *slave_dev,
1471 struct netlink_ext_ack *extack);
1472 int (*ndo_del_slave)(struct net_device *dev,
1473 struct net_device *slave_dev);
1474 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1475 struct sk_buff *skb,
1477 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1479 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1480 netdev_features_t features);
1481 int (*ndo_set_features)(struct net_device *dev,
1482 netdev_features_t features);
1483 int (*ndo_neigh_construct)(struct net_device *dev,
1484 struct neighbour *n);
1485 void (*ndo_neigh_destroy)(struct net_device *dev,
1486 struct neighbour *n);
1488 int (*ndo_fdb_add)(struct ndmsg *ndm,
1489 struct nlattr *tb[],
1490 struct net_device *dev,
1491 const unsigned char *addr,
1494 struct netlink_ext_ack *extack);
1495 int (*ndo_fdb_del)(struct ndmsg *ndm,
1496 struct nlattr *tb[],
1497 struct net_device *dev,
1498 const unsigned char *addr,
1500 int (*ndo_fdb_dump)(struct sk_buff *skb,
1501 struct netlink_callback *cb,
1502 struct net_device *dev,
1503 struct net_device *filter_dev,
1505 int (*ndo_fdb_get)(struct sk_buff *skb,
1506 struct nlattr *tb[],
1507 struct net_device *dev,
1508 const unsigned char *addr,
1509 u16 vid, u32 portid, u32 seq,
1510 struct netlink_ext_ack *extack);
1511 int (*ndo_bridge_setlink)(struct net_device *dev,
1512 struct nlmsghdr *nlh,
1514 struct netlink_ext_ack *extack);
1515 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1517 struct net_device *dev,
1520 int (*ndo_bridge_dellink)(struct net_device *dev,
1521 struct nlmsghdr *nlh,
1523 int (*ndo_change_carrier)(struct net_device *dev,
1525 int (*ndo_get_phys_port_id)(struct net_device *dev,
1526 struct netdev_phys_item_id *ppid);
1527 int (*ndo_get_port_parent_id)(struct net_device *dev,
1528 struct netdev_phys_item_id *ppid);
1529 int (*ndo_get_phys_port_name)(struct net_device *dev,
1530 char *name, size_t len);
1531 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1532 struct net_device *dev);
1533 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1536 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1539 int (*ndo_get_iflink)(const struct net_device *dev);
1540 int (*ndo_change_proto_down)(struct net_device *dev,
1542 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1543 struct sk_buff *skb);
1544 void (*ndo_set_rx_headroom)(struct net_device *dev,
1545 int needed_headroom);
1546 int (*ndo_bpf)(struct net_device *dev,
1547 struct netdev_bpf *bpf);
1548 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1549 struct xdp_frame **xdp,
1551 int (*ndo_xsk_wakeup)(struct net_device *dev,
1552 u32 queue_id, u32 flags);
1553 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1554 int (*ndo_tunnel_ctl)(struct net_device *dev,
1555 struct ip_tunnel_parm *p, int cmd);
1556 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1557 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1558 struct net_device_path *path);
1562 * enum netdev_priv_flags - &struct net_device priv_flags
1564 * These are the &struct net_device, they are only set internally
1565 * by drivers and used in the kernel. These flags are invisible to
1566 * userspace; this means that the order of these flags can change
1567 * during any kernel release.
1569 * You should have a pretty good reason to be extending these flags.
1571 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1572 * @IFF_EBRIDGE: Ethernet bridging device
1573 * @IFF_BONDING: bonding master or slave
1574 * @IFF_ISATAP: ISATAP interface (RFC4214)
1575 * @IFF_WAN_HDLC: WAN HDLC device
1576 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1578 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1579 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1580 * @IFF_MACVLAN_PORT: device used as macvlan port
1581 * @IFF_BRIDGE_PORT: device used as bridge port
1582 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1583 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1584 * @IFF_UNICAST_FLT: Supports unicast filtering
1585 * @IFF_TEAM_PORT: device used as team port
1586 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1587 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1588 * change when it's running
1589 * @IFF_MACVLAN: Macvlan device
1590 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1591 * underlying stacked devices
1592 * @IFF_L3MDEV_MASTER: device is an L3 master device
1593 * @IFF_NO_QUEUE: device can run without qdisc attached
1594 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1595 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1596 * @IFF_TEAM: device is a team device
1597 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1598 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1599 * entity (i.e. the master device for bridged veth)
1600 * @IFF_MACSEC: device is a MACsec device
1601 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1602 * @IFF_FAILOVER: device is a failover master device
1603 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1604 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1605 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1606 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1607 * skb_headlen(skb) == 0 (data starts from frag0)
1609 enum netdev_priv_flags {
1610 IFF_802_1Q_VLAN = 1<<0,
1614 IFF_WAN_HDLC = 1<<4,
1615 IFF_XMIT_DST_RELEASE = 1<<5,
1616 IFF_DONT_BRIDGE = 1<<6,
1617 IFF_DISABLE_NETPOLL = 1<<7,
1618 IFF_MACVLAN_PORT = 1<<8,
1619 IFF_BRIDGE_PORT = 1<<9,
1620 IFF_OVS_DATAPATH = 1<<10,
1621 IFF_TX_SKB_SHARING = 1<<11,
1622 IFF_UNICAST_FLT = 1<<12,
1623 IFF_TEAM_PORT = 1<<13,
1624 IFF_SUPP_NOFCS = 1<<14,
1625 IFF_LIVE_ADDR_CHANGE = 1<<15,
1626 IFF_MACVLAN = 1<<16,
1627 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1628 IFF_L3MDEV_MASTER = 1<<18,
1629 IFF_NO_QUEUE = 1<<19,
1630 IFF_OPENVSWITCH = 1<<20,
1631 IFF_L3MDEV_SLAVE = 1<<21,
1633 IFF_RXFH_CONFIGURED = 1<<23,
1634 IFF_PHONY_HEADROOM = 1<<24,
1636 IFF_NO_RX_HANDLER = 1<<26,
1637 IFF_FAILOVER = 1<<27,
1638 IFF_FAILOVER_SLAVE = 1<<28,
1639 IFF_L3MDEV_RX_HANDLER = 1<<29,
1640 IFF_LIVE_RENAME_OK = 1<<30,
1641 IFF_TX_SKB_NO_LINEAR = 1<<31,
1644 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1645 #define IFF_EBRIDGE IFF_EBRIDGE
1646 #define IFF_BONDING IFF_BONDING
1647 #define IFF_ISATAP IFF_ISATAP
1648 #define IFF_WAN_HDLC IFF_WAN_HDLC
1649 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1650 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1651 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1652 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1653 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1654 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1655 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1656 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1657 #define IFF_TEAM_PORT IFF_TEAM_PORT
1658 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1659 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1660 #define IFF_MACVLAN IFF_MACVLAN
1661 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1662 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1663 #define IFF_NO_QUEUE IFF_NO_QUEUE
1664 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1665 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1666 #define IFF_TEAM IFF_TEAM
1667 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1668 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1669 #define IFF_MACSEC IFF_MACSEC
1670 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1671 #define IFF_FAILOVER IFF_FAILOVER
1672 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1673 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1674 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1675 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1677 /* Specifies the type of the struct net_device::ml_priv pointer */
1678 enum netdev_ml_priv_type {
1684 * struct net_device - The DEVICE structure.
1686 * Actually, this whole structure is a big mistake. It mixes I/O
1687 * data with strictly "high-level" data, and it has to know about
1688 * almost every data structure used in the INET module.
1690 * @name: This is the first field of the "visible" part of this structure
1691 * (i.e. as seen by users in the "Space.c" file). It is the name
1694 * @name_node: Name hashlist node
1695 * @ifalias: SNMP alias
1696 * @mem_end: Shared memory end
1697 * @mem_start: Shared memory start
1698 * @base_addr: Device I/O address
1699 * @irq: Device IRQ number
1701 * @state: Generic network queuing layer state, see netdev_state_t
1702 * @dev_list: The global list of network devices
1703 * @napi_list: List entry used for polling NAPI devices
1704 * @unreg_list: List entry when we are unregistering the
1705 * device; see the function unregister_netdev
1706 * @close_list: List entry used when we are closing the device
1707 * @ptype_all: Device-specific packet handlers for all protocols
1708 * @ptype_specific: Device-specific, protocol-specific packet handlers
1710 * @adj_list: Directly linked devices, like slaves for bonding
1711 * @features: Currently active device features
1712 * @hw_features: User-changeable features
1714 * @wanted_features: User-requested features
1715 * @vlan_features: Mask of features inheritable by VLAN devices
1717 * @hw_enc_features: Mask of features inherited by encapsulating devices
1718 * This field indicates what encapsulation
1719 * offloads the hardware is capable of doing,
1720 * and drivers will need to set them appropriately.
1722 * @mpls_features: Mask of features inheritable by MPLS
1723 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1725 * @ifindex: interface index
1726 * @group: The group the device belongs to
1728 * @stats: Statistics struct, which was left as a legacy, use
1729 * rtnl_link_stats64 instead
1731 * @rx_dropped: Dropped packets by core network,
1732 * do not use this in drivers
1733 * @tx_dropped: Dropped packets by core network,
1734 * do not use this in drivers
1735 * @rx_nohandler: nohandler dropped packets by core network on
1736 * inactive devices, do not use this in drivers
1737 * @carrier_up_count: Number of times the carrier has been up
1738 * @carrier_down_count: Number of times the carrier has been down
1740 * @wireless_handlers: List of functions to handle Wireless Extensions,
1742 * see <net/iw_handler.h> for details.
1743 * @wireless_data: Instance data managed by the core of wireless extensions
1745 * @netdev_ops: Includes several pointers to callbacks,
1746 * if one wants to override the ndo_*() functions
1747 * @ethtool_ops: Management operations
1748 * @l3mdev_ops: Layer 3 master device operations
1749 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1750 * discovery handling. Necessary for e.g. 6LoWPAN.
1751 * @xfrmdev_ops: Transformation offload operations
1752 * @tlsdev_ops: Transport Layer Security offload operations
1753 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1754 * of Layer 2 headers.
1756 * @flags: Interface flags (a la BSD)
1757 * @priv_flags: Like 'flags' but invisible to userspace,
1758 * see if.h for the definitions
1759 * @gflags: Global flags ( kept as legacy )
1760 * @padded: How much padding added by alloc_netdev()
1761 * @operstate: RFC2863 operstate
1762 * @link_mode: Mapping policy to operstate
1763 * @if_port: Selectable AUI, TP, ...
1765 * @mtu: Interface MTU value
1766 * @min_mtu: Interface Minimum MTU value
1767 * @max_mtu: Interface Maximum MTU value
1768 * @type: Interface hardware type
1769 * @hard_header_len: Maximum hardware header length.
1770 * @min_header_len: Minimum hardware header length
1772 * @needed_headroom: Extra headroom the hardware may need, but not in all
1773 * cases can this be guaranteed
1774 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1775 * cases can this be guaranteed. Some cases also use
1776 * LL_MAX_HEADER instead to allocate the skb
1778 * interface address info:
1780 * @perm_addr: Permanent hw address
1781 * @addr_assign_type: Hw address assignment type
1782 * @addr_len: Hardware address length
1783 * @upper_level: Maximum depth level of upper devices.
1784 * @lower_level: Maximum depth level of lower devices.
1785 * @neigh_priv_len: Used in neigh_alloc()
1786 * @dev_id: Used to differentiate devices that share
1787 * the same link layer address
1788 * @dev_port: Used to differentiate devices that share
1790 * @addr_list_lock: XXX: need comments on this one
1791 * @name_assign_type: network interface name assignment type
1792 * @uc_promisc: Counter that indicates promiscuous mode
1793 * has been enabled due to the need to listen to
1794 * additional unicast addresses in a device that
1795 * does not implement ndo_set_rx_mode()
1796 * @uc: unicast mac addresses
1797 * @mc: multicast mac addresses
1798 * @dev_addrs: list of device hw addresses
1799 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1800 * @promiscuity: Number of times the NIC is told to work in
1801 * promiscuous mode; if it becomes 0 the NIC will
1802 * exit promiscuous mode
1803 * @allmulti: Counter, enables or disables allmulticast mode
1805 * @vlan_info: VLAN info
1806 * @dsa_ptr: dsa specific data
1807 * @tipc_ptr: TIPC specific data
1808 * @atalk_ptr: AppleTalk link
1809 * @ip_ptr: IPv4 specific data
1810 * @dn_ptr: DECnet specific data
1811 * @ip6_ptr: IPv6 specific data
1812 * @ax25_ptr: AX.25 specific data
1813 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1814 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1816 * @mpls_ptr: mpls_dev struct pointer
1818 * @dev_addr: Hw address (before bcast,
1819 * because most packets are unicast)
1821 * @_rx: Array of RX queues
1822 * @num_rx_queues: Number of RX queues
1823 * allocated at register_netdev() time
1824 * @real_num_rx_queues: Number of RX queues currently active in device
1825 * @xdp_prog: XDP sockets filter program pointer
1826 * @gro_flush_timeout: timeout for GRO layer in NAPI
1827 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1828 * allow to avoid NIC hard IRQ, on busy queues.
1830 * @rx_handler: handler for received packets
1831 * @rx_handler_data: XXX: need comments on this one
1832 * @miniq_ingress: ingress/clsact qdisc specific data for
1833 * ingress processing
1834 * @ingress_queue: XXX: need comments on this one
1835 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1836 * @broadcast: hw bcast address
1838 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1839 * indexed by RX queue number. Assigned by driver.
1840 * This must only be set if the ndo_rx_flow_steer
1841 * operation is defined
1842 * @index_hlist: Device index hash chain
1844 * @_tx: Array of TX queues
1845 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1846 * @real_num_tx_queues: Number of TX queues currently active in device
1847 * @qdisc: Root qdisc from userspace point of view
1848 * @tx_queue_len: Max frames per queue allowed
1849 * @tx_global_lock: XXX: need comments on this one
1850 * @xdp_bulkq: XDP device bulk queue
1851 * @xps_maps: all CPUs/RXQs maps for XPS device
1853 * @xps_maps: XXX: need comments on this one
1854 * @miniq_egress: clsact qdisc specific data for
1856 * @qdisc_hash: qdisc hash table
1857 * @watchdog_timeo: Represents the timeout that is used by
1858 * the watchdog (see dev_watchdog())
1859 * @watchdog_timer: List of timers
1861 * @proto_down_reason: reason a netdev interface is held down
1862 * @pcpu_refcnt: Number of references to this device
1863 * @dev_refcnt: Number of references to this device
1864 * @todo_list: Delayed register/unregister
1865 * @link_watch_list: XXX: need comments on this one
1867 * @reg_state: Register/unregister state machine
1868 * @dismantle: Device is going to be freed
1869 * @rtnl_link_state: This enum represents the phases of creating
1872 * @needs_free_netdev: Should unregister perform free_netdev?
1873 * @priv_destructor: Called from unregister
1874 * @npinfo: XXX: need comments on this one
1875 * @nd_net: Network namespace this network device is inside
1877 * @ml_priv: Mid-layer private
1878 * @ml_priv_type: Mid-layer private type
1879 * @lstats: Loopback statistics
1880 * @tstats: Tunnel statistics
1881 * @dstats: Dummy statistics
1882 * @vstats: Virtual ethernet statistics
1887 * @dev: Class/net/name entry
1888 * @sysfs_groups: Space for optional device, statistics and wireless
1891 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1892 * @rtnl_link_ops: Rtnl_link_ops
1894 * @gso_max_size: Maximum size of generic segmentation offload
1895 * @gso_max_segs: Maximum number of segments that can be passed to the
1898 * @dcbnl_ops: Data Center Bridging netlink ops
1899 * @num_tc: Number of traffic classes in the net device
1900 * @tc_to_txq: XXX: need comments on this one
1901 * @prio_tc_map: XXX: need comments on this one
1903 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1905 * @priomap: XXX: need comments on this one
1906 * @phydev: Physical device may attach itself
1907 * for hardware timestamping
1908 * @sfp_bus: attached &struct sfp_bus structure.
1910 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1911 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1913 * @proto_down: protocol port state information can be sent to the
1914 * switch driver and used to set the phys state of the
1917 * @wol_enabled: Wake-on-LAN is enabled
1919 * @threaded: napi threaded mode is enabled
1921 * @net_notifier_list: List of per-net netdev notifier block
1922 * that follow this device when it is moved
1923 * to another network namespace.
1925 * @macsec_ops: MACsec offloading ops
1927 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1928 * offload capabilities of the device
1929 * @udp_tunnel_nic: UDP tunnel offload state
1930 * @xdp_state: stores info on attached XDP BPF programs
1932 * @nested_level: Used as as a parameter of spin_lock_nested() of
1933 * dev->addr_list_lock.
1934 * @unlink_list: As netif_addr_lock() can be called recursively,
1935 * keep a list of interfaces to be deleted.
1937 * FIXME: cleanup struct net_device such that network protocol info
1942 char name[IFNAMSIZ];
1943 struct netdev_name_node *name_node;
1944 struct dev_ifalias __rcu *ifalias;
1946 * I/O specific fields
1947 * FIXME: Merge these and struct ifmap into one
1949 unsigned long mem_end;
1950 unsigned long mem_start;
1951 unsigned long base_addr;
1954 * Some hardware also needs these fields (state,dev_list,
1955 * napi_list,unreg_list,close_list) but they are not
1956 * part of the usual set specified in Space.c.
1959 unsigned long state;
1961 struct list_head dev_list;
1962 struct list_head napi_list;
1963 struct list_head unreg_list;
1964 struct list_head close_list;
1965 struct list_head ptype_all;
1966 struct list_head ptype_specific;
1969 struct list_head upper;
1970 struct list_head lower;
1973 /* Read-mostly cache-line for fast-path access */
1975 unsigned int priv_flags;
1976 const struct net_device_ops *netdev_ops;
1978 unsigned short gflags;
1979 unsigned short hard_header_len;
1981 /* Note : dev->mtu is often read without holding a lock.
1982 * Writers usually hold RTNL.
1983 * It is recommended to use READ_ONCE() to annotate the reads,
1984 * and to use WRITE_ONCE() to annotate the writes.
1987 unsigned short needed_headroom;
1988 unsigned short needed_tailroom;
1990 netdev_features_t features;
1991 netdev_features_t hw_features;
1992 netdev_features_t wanted_features;
1993 netdev_features_t vlan_features;
1994 netdev_features_t hw_enc_features;
1995 netdev_features_t mpls_features;
1996 netdev_features_t gso_partial_features;
1998 unsigned int min_mtu;
1999 unsigned int max_mtu;
2000 unsigned short type;
2001 unsigned char min_header_len;
2002 unsigned char name_assign_type;
2006 struct net_device_stats stats; /* not used by modern drivers */
2008 atomic_long_t rx_dropped;
2009 atomic_long_t tx_dropped;
2010 atomic_long_t rx_nohandler;
2012 /* Stats to monitor link on/off, flapping */
2013 atomic_t carrier_up_count;
2014 atomic_t carrier_down_count;
2016 #ifdef CONFIG_WIRELESS_EXT
2017 const struct iw_handler_def *wireless_handlers;
2018 struct iw_public_data *wireless_data;
2020 const struct ethtool_ops *ethtool_ops;
2021 #ifdef CONFIG_NET_L3_MASTER_DEV
2022 const struct l3mdev_ops *l3mdev_ops;
2024 #if IS_ENABLED(CONFIG_IPV6)
2025 const struct ndisc_ops *ndisc_ops;
2028 #ifdef CONFIG_XFRM_OFFLOAD
2029 const struct xfrmdev_ops *xfrmdev_ops;
2032 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2033 const struct tlsdev_ops *tlsdev_ops;
2036 const struct header_ops *header_ops;
2038 unsigned char operstate;
2039 unsigned char link_mode;
2041 unsigned char if_port;
2044 /* Interface address info. */
2045 unsigned char perm_addr[MAX_ADDR_LEN];
2046 unsigned char addr_assign_type;
2047 unsigned char addr_len;
2048 unsigned char upper_level;
2049 unsigned char lower_level;
2051 unsigned short neigh_priv_len;
2052 unsigned short dev_id;
2053 unsigned short dev_port;
2054 unsigned short padded;
2056 spinlock_t addr_list_lock;
2059 struct netdev_hw_addr_list uc;
2060 struct netdev_hw_addr_list mc;
2061 struct netdev_hw_addr_list dev_addrs;
2064 struct kset *queues_kset;
2066 #ifdef CONFIG_LOCKDEP
2067 struct list_head unlink_list;
2069 unsigned int promiscuity;
2070 unsigned int allmulti;
2072 #ifdef CONFIG_LOCKDEP
2073 unsigned char nested_level;
2077 /* Protocol-specific pointers */
2079 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2080 struct vlan_info __rcu *vlan_info;
2082 #if IS_ENABLED(CONFIG_NET_DSA)
2083 struct dsa_port *dsa_ptr;
2085 #if IS_ENABLED(CONFIG_TIPC)
2086 struct tipc_bearer __rcu *tipc_ptr;
2088 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2091 struct in_device __rcu *ip_ptr;
2092 #if IS_ENABLED(CONFIG_DECNET)
2093 struct dn_dev __rcu *dn_ptr;
2095 struct inet6_dev __rcu *ip6_ptr;
2096 #if IS_ENABLED(CONFIG_AX25)
2099 struct wireless_dev *ieee80211_ptr;
2100 struct wpan_dev *ieee802154_ptr;
2101 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2102 struct mpls_dev __rcu *mpls_ptr;
2106 * Cache lines mostly used on receive path (including eth_type_trans())
2108 /* Interface address info used in eth_type_trans() */
2109 unsigned char *dev_addr;
2111 struct netdev_rx_queue *_rx;
2112 unsigned int num_rx_queues;
2113 unsigned int real_num_rx_queues;
2115 struct bpf_prog __rcu *xdp_prog;
2116 unsigned long gro_flush_timeout;
2117 int napi_defer_hard_irqs;
2118 rx_handler_func_t __rcu *rx_handler;
2119 void __rcu *rx_handler_data;
2121 #ifdef CONFIG_NET_CLS_ACT
2122 struct mini_Qdisc __rcu *miniq_ingress;
2124 struct netdev_queue __rcu *ingress_queue;
2125 #ifdef CONFIG_NETFILTER_INGRESS
2126 struct nf_hook_entries __rcu *nf_hooks_ingress;
2129 unsigned char broadcast[MAX_ADDR_LEN];
2130 #ifdef CONFIG_RFS_ACCEL
2131 struct cpu_rmap *rx_cpu_rmap;
2133 struct hlist_node index_hlist;
2136 * Cache lines mostly used on transmit path
2138 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2139 unsigned int num_tx_queues;
2140 unsigned int real_num_tx_queues;
2141 struct Qdisc *qdisc;
2142 unsigned int tx_queue_len;
2143 spinlock_t tx_global_lock;
2145 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2148 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2150 #ifdef CONFIG_NET_CLS_ACT
2151 struct mini_Qdisc __rcu *miniq_egress;
2154 #ifdef CONFIG_NET_SCHED
2155 DECLARE_HASHTABLE (qdisc_hash, 4);
2157 /* These may be needed for future network-power-down code. */
2158 struct timer_list watchdog_timer;
2161 u32 proto_down_reason;
2163 struct list_head todo_list;
2165 #ifdef CONFIG_PCPU_DEV_REFCNT
2166 int __percpu *pcpu_refcnt;
2168 refcount_t dev_refcnt;
2171 struct list_head link_watch_list;
2173 enum { NETREG_UNINITIALIZED=0,
2174 NETREG_REGISTERED, /* completed register_netdevice */
2175 NETREG_UNREGISTERING, /* called unregister_netdevice */
2176 NETREG_UNREGISTERED, /* completed unregister todo */
2177 NETREG_RELEASED, /* called free_netdev */
2178 NETREG_DUMMY, /* dummy device for NAPI poll */
2184 RTNL_LINK_INITIALIZED,
2185 RTNL_LINK_INITIALIZING,
2186 } rtnl_link_state:16;
2188 bool needs_free_netdev;
2189 void (*priv_destructor)(struct net_device *dev);
2191 #ifdef CONFIG_NETPOLL
2192 struct netpoll_info __rcu *npinfo;
2195 possible_net_t nd_net;
2197 /* mid-layer private */
2199 enum netdev_ml_priv_type ml_priv_type;
2202 struct pcpu_lstats __percpu *lstats;
2203 struct pcpu_sw_netstats __percpu *tstats;
2204 struct pcpu_dstats __percpu *dstats;
2207 #if IS_ENABLED(CONFIG_GARP)
2208 struct garp_port __rcu *garp_port;
2210 #if IS_ENABLED(CONFIG_MRP)
2211 struct mrp_port __rcu *mrp_port;
2215 const struct attribute_group *sysfs_groups[4];
2216 const struct attribute_group *sysfs_rx_queue_group;
2218 const struct rtnl_link_ops *rtnl_link_ops;
2220 /* for setting kernel sock attribute on TCP connection setup */
2221 #define GSO_MAX_SIZE 65536
2222 unsigned int gso_max_size;
2223 #define GSO_MAX_SEGS 65535
2227 const struct dcbnl_rtnl_ops *dcbnl_ops;
2230 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2231 u8 prio_tc_map[TC_BITMASK + 1];
2233 #if IS_ENABLED(CONFIG_FCOE)
2234 unsigned int fcoe_ddp_xid;
2236 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2237 struct netprio_map __rcu *priomap;
2239 struct phy_device *phydev;
2240 struct sfp_bus *sfp_bus;
2241 struct lock_class_key *qdisc_tx_busylock;
2242 struct lock_class_key *qdisc_running_key;
2244 unsigned wol_enabled:1;
2245 unsigned threaded:1;
2247 struct list_head net_notifier_list;
2249 #if IS_ENABLED(CONFIG_MACSEC)
2250 /* MACsec management functions */
2251 const struct macsec_ops *macsec_ops;
2253 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2254 struct udp_tunnel_nic *udp_tunnel_nic;
2256 /* protected by rtnl_lock */
2257 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2259 #define to_net_dev(d) container_of(d, struct net_device, dev)
2261 static inline bool netif_elide_gro(const struct net_device *dev)
2263 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2268 #define NETDEV_ALIGN 32
2271 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2273 return dev->prio_tc_map[prio & TC_BITMASK];
2277 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2279 if (tc >= dev->num_tc)
2282 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2286 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2287 void netdev_reset_tc(struct net_device *dev);
2288 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2289 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2292 int netdev_get_num_tc(struct net_device *dev)
2297 static inline void net_prefetch(void *p)
2300 #if L1_CACHE_BYTES < 128
2301 prefetch((u8 *)p + L1_CACHE_BYTES);
2305 static inline void net_prefetchw(void *p)
2308 #if L1_CACHE_BYTES < 128
2309 prefetchw((u8 *)p + L1_CACHE_BYTES);
2313 void netdev_unbind_sb_channel(struct net_device *dev,
2314 struct net_device *sb_dev);
2315 int netdev_bind_sb_channel_queue(struct net_device *dev,
2316 struct net_device *sb_dev,
2317 u8 tc, u16 count, u16 offset);
2318 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2319 static inline int netdev_get_sb_channel(struct net_device *dev)
2321 return max_t(int, -dev->num_tc, 0);
2325 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2328 return &dev->_tx[index];
2331 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2332 const struct sk_buff *skb)
2334 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2337 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2338 void (*f)(struct net_device *,
2339 struct netdev_queue *,
2345 for (i = 0; i < dev->num_tx_queues; i++)
2346 f(dev, &dev->_tx[i], arg);
2349 #define netdev_lockdep_set_classes(dev) \
2351 static struct lock_class_key qdisc_tx_busylock_key; \
2352 static struct lock_class_key qdisc_running_key; \
2353 static struct lock_class_key qdisc_xmit_lock_key; \
2354 static struct lock_class_key dev_addr_list_lock_key; \
2357 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2358 (dev)->qdisc_running_key = &qdisc_running_key; \
2359 lockdep_set_class(&(dev)->addr_list_lock, \
2360 &dev_addr_list_lock_key); \
2361 for (i = 0; i < (dev)->num_tx_queues; i++) \
2362 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2363 &qdisc_xmit_lock_key); \
2366 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2367 struct net_device *sb_dev);
2368 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2369 struct sk_buff *skb,
2370 struct net_device *sb_dev);
2372 /* returns the headroom that the master device needs to take in account
2373 * when forwarding to this dev
2375 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2377 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2380 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2382 if (dev->netdev_ops->ndo_set_rx_headroom)
2383 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2386 /* set the device rx headroom to the dev's default */
2387 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2389 netdev_set_rx_headroom(dev, -1);
2392 static inline void *netdev_get_ml_priv(struct net_device *dev,
2393 enum netdev_ml_priv_type type)
2395 if (dev->ml_priv_type != type)
2398 return dev->ml_priv;
2401 static inline void netdev_set_ml_priv(struct net_device *dev,
2403 enum netdev_ml_priv_type type)
2405 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2406 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2407 dev->ml_priv_type, type);
2408 WARN(!dev->ml_priv_type && dev->ml_priv,
2409 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2411 dev->ml_priv = ml_priv;
2412 dev->ml_priv_type = type;
2416 * Net namespace inlines
2419 struct net *dev_net(const struct net_device *dev)
2421 return read_pnet(&dev->nd_net);
2425 void dev_net_set(struct net_device *dev, struct net *net)
2427 write_pnet(&dev->nd_net, net);
2431 * netdev_priv - access network device private data
2432 * @dev: network device
2434 * Get network device private data
2436 static inline void *netdev_priv(const struct net_device *dev)
2438 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2441 /* Set the sysfs physical device reference for the network logical device
2442 * if set prior to registration will cause a symlink during initialization.
2444 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2446 /* Set the sysfs device type for the network logical device to allow
2447 * fine-grained identification of different network device types. For
2448 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2450 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2452 /* Default NAPI poll() weight
2453 * Device drivers are strongly advised to not use bigger value
2455 #define NAPI_POLL_WEIGHT 64
2458 * netif_napi_add - initialize a NAPI context
2459 * @dev: network device
2460 * @napi: NAPI context
2461 * @poll: polling function
2462 * @weight: default weight
2464 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2465 * *any* of the other NAPI-related functions.
2467 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2468 int (*poll)(struct napi_struct *, int), int weight);
2471 * netif_tx_napi_add - initialize a NAPI context
2472 * @dev: network device
2473 * @napi: NAPI context
2474 * @poll: polling function
2475 * @weight: default weight
2477 * This variant of netif_napi_add() should be used from drivers using NAPI
2478 * to exclusively poll a TX queue.
2479 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2481 static inline void netif_tx_napi_add(struct net_device *dev,
2482 struct napi_struct *napi,
2483 int (*poll)(struct napi_struct *, int),
2486 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2487 netif_napi_add(dev, napi, poll, weight);
2491 * __netif_napi_del - remove a NAPI context
2492 * @napi: NAPI context
2494 * Warning: caller must observe RCU grace period before freeing memory
2495 * containing @napi. Drivers might want to call this helper to combine
2496 * all the needed RCU grace periods into a single one.
2498 void __netif_napi_del(struct napi_struct *napi);
2501 * netif_napi_del - remove a NAPI context
2502 * @napi: NAPI context
2504 * netif_napi_del() removes a NAPI context from the network device NAPI list
2506 static inline void netif_napi_del(struct napi_struct *napi)
2508 __netif_napi_del(napi);
2512 struct napi_gro_cb {
2513 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2516 /* Length of frag0. */
2517 unsigned int frag0_len;
2519 /* This indicates where we are processing relative to skb->data. */
2522 /* This is non-zero if the packet cannot be merged with the new skb. */
2525 /* Save the IP ID here and check when we get to the transport layer */
2528 /* Number of segments aggregated. */
2531 /* Start offset for remote checksum offload */
2532 u16 gro_remcsum_start;
2534 /* jiffies when first packet was created/queued */
2537 /* Used in ipv6_gro_receive() and foo-over-udp */
2540 /* This is non-zero if the packet may be of the same flow. */
2543 /* Used in tunnel GRO receive */
2546 /* GRO checksum is valid */
2549 /* Number of checksums via CHECKSUM_UNNECESSARY */
2554 #define NAPI_GRO_FREE 1
2555 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2557 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2560 /* Used in GRE, set in fou/gue_gro_receive */
2563 /* Used to determine if flush_id can be ignored */
2566 /* Number of gro_receive callbacks this packet already went through */
2567 u8 recursion_counter:4;
2569 /* GRO is done by frag_list pointer chaining. */
2572 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2575 /* used in skb_gro_receive() slow path */
2576 struct sk_buff *last;
2579 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2581 #define GRO_RECURSION_LIMIT 15
2582 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2584 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2587 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2588 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2589 struct list_head *head,
2590 struct sk_buff *skb)
2592 if (unlikely(gro_recursion_inc_test(skb))) {
2593 NAPI_GRO_CB(skb)->flush |= 1;
2597 return cb(head, skb);
2600 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2602 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2604 struct list_head *head,
2605 struct sk_buff *skb)
2607 if (unlikely(gro_recursion_inc_test(skb))) {
2608 NAPI_GRO_CB(skb)->flush |= 1;
2612 return cb(sk, head, skb);
2615 struct packet_type {
2616 __be16 type; /* This is really htons(ether_type). */
2617 bool ignore_outgoing;
2618 struct net_device *dev; /* NULL is wildcarded here */
2619 int (*func) (struct sk_buff *,
2620 struct net_device *,
2621 struct packet_type *,
2622 struct net_device *);
2623 void (*list_func) (struct list_head *,
2624 struct packet_type *,
2625 struct net_device *);
2626 bool (*id_match)(struct packet_type *ptype,
2628 void *af_packet_priv;
2629 struct list_head list;
2632 struct offload_callbacks {
2633 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2634 netdev_features_t features);
2635 struct sk_buff *(*gro_receive)(struct list_head *head,
2636 struct sk_buff *skb);
2637 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2640 struct packet_offload {
2641 __be16 type; /* This is really htons(ether_type). */
2643 struct offload_callbacks callbacks;
2644 struct list_head list;
2647 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2648 struct pcpu_sw_netstats {
2653 struct u64_stats_sync syncp;
2654 } __aligned(4 * sizeof(u64));
2656 struct pcpu_lstats {
2657 u64_stats_t packets;
2659 struct u64_stats_sync syncp;
2660 } __aligned(2 * sizeof(u64));
2662 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2664 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2666 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2668 u64_stats_update_begin(&tstats->syncp);
2669 tstats->rx_bytes += len;
2670 tstats->rx_packets++;
2671 u64_stats_update_end(&tstats->syncp);
2674 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2675 unsigned int packets,
2678 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2680 u64_stats_update_begin(&tstats->syncp);
2681 tstats->tx_bytes += len;
2682 tstats->tx_packets += packets;
2683 u64_stats_update_end(&tstats->syncp);
2686 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2688 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2690 u64_stats_update_begin(&lstats->syncp);
2691 u64_stats_add(&lstats->bytes, len);
2692 u64_stats_inc(&lstats->packets);
2693 u64_stats_update_end(&lstats->syncp);
2696 #define __netdev_alloc_pcpu_stats(type, gfp) \
2698 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2701 for_each_possible_cpu(__cpu) { \
2702 typeof(type) *stat; \
2703 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2704 u64_stats_init(&stat->syncp); \
2710 #define netdev_alloc_pcpu_stats(type) \
2711 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2713 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2715 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2718 for_each_possible_cpu(__cpu) { \
2719 typeof(type) *stat; \
2720 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2721 u64_stats_init(&stat->syncp); \
2727 enum netdev_lag_tx_type {
2728 NETDEV_LAG_TX_TYPE_UNKNOWN,
2729 NETDEV_LAG_TX_TYPE_RANDOM,
2730 NETDEV_LAG_TX_TYPE_BROADCAST,
2731 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2732 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2733 NETDEV_LAG_TX_TYPE_HASH,
2736 enum netdev_lag_hash {
2737 NETDEV_LAG_HASH_NONE,
2739 NETDEV_LAG_HASH_L34,
2740 NETDEV_LAG_HASH_L23,
2741 NETDEV_LAG_HASH_E23,
2742 NETDEV_LAG_HASH_E34,
2743 NETDEV_LAG_HASH_VLAN_SRCMAC,
2744 NETDEV_LAG_HASH_UNKNOWN,
2747 struct netdev_lag_upper_info {
2748 enum netdev_lag_tx_type tx_type;
2749 enum netdev_lag_hash hash_type;
2752 struct netdev_lag_lower_state_info {
2757 #include <linux/notifier.h>
2759 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2760 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2764 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2766 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2767 detected a hardware crash and restarted
2768 - we can use this eg to kick tcp sessions
2770 NETDEV_CHANGE, /* Notify device state change */
2773 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2774 NETDEV_CHANGEADDR, /* notify after the address change */
2775 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2779 NETDEV_BONDING_FAILOVER,
2781 NETDEV_PRE_TYPE_CHANGE,
2782 NETDEV_POST_TYPE_CHANGE,
2785 NETDEV_NOTIFY_PEERS,
2789 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2790 NETDEV_CHANGEINFODATA,
2791 NETDEV_BONDING_INFO,
2792 NETDEV_PRECHANGEUPPER,
2793 NETDEV_CHANGELOWERSTATE,
2794 NETDEV_UDP_TUNNEL_PUSH_INFO,
2795 NETDEV_UDP_TUNNEL_DROP_INFO,
2796 NETDEV_CHANGE_TX_QUEUE_LEN,
2797 NETDEV_CVLAN_FILTER_PUSH_INFO,
2798 NETDEV_CVLAN_FILTER_DROP_INFO,
2799 NETDEV_SVLAN_FILTER_PUSH_INFO,
2800 NETDEV_SVLAN_FILTER_DROP_INFO,
2802 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2804 int register_netdevice_notifier(struct notifier_block *nb);
2805 int unregister_netdevice_notifier(struct notifier_block *nb);
2806 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2807 int unregister_netdevice_notifier_net(struct net *net,
2808 struct notifier_block *nb);
2809 int register_netdevice_notifier_dev_net(struct net_device *dev,
2810 struct notifier_block *nb,
2811 struct netdev_net_notifier *nn);
2812 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2813 struct notifier_block *nb,
2814 struct netdev_net_notifier *nn);
2816 struct netdev_notifier_info {
2817 struct net_device *dev;
2818 struct netlink_ext_ack *extack;
2821 struct netdev_notifier_info_ext {
2822 struct netdev_notifier_info info; /* must be first */
2828 struct netdev_notifier_change_info {
2829 struct netdev_notifier_info info; /* must be first */
2830 unsigned int flags_changed;
2833 struct netdev_notifier_changeupper_info {
2834 struct netdev_notifier_info info; /* must be first */
2835 struct net_device *upper_dev; /* new upper dev */
2836 bool master; /* is upper dev master */
2837 bool linking; /* is the notification for link or unlink */
2838 void *upper_info; /* upper dev info */
2841 struct netdev_notifier_changelowerstate_info {
2842 struct netdev_notifier_info info; /* must be first */
2843 void *lower_state_info; /* is lower dev state */
2846 struct netdev_notifier_pre_changeaddr_info {
2847 struct netdev_notifier_info info; /* must be first */
2848 const unsigned char *dev_addr;
2851 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2852 struct net_device *dev)
2855 info->extack = NULL;
2858 static inline struct net_device *
2859 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2864 static inline struct netlink_ext_ack *
2865 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2867 return info->extack;
2870 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2873 extern rwlock_t dev_base_lock; /* Device list lock */
2875 #define for_each_netdev(net, d) \
2876 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2877 #define for_each_netdev_reverse(net, d) \
2878 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2879 #define for_each_netdev_rcu(net, d) \
2880 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2881 #define for_each_netdev_safe(net, d, n) \
2882 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2883 #define for_each_netdev_continue(net, d) \
2884 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2885 #define for_each_netdev_continue_reverse(net, d) \
2886 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2888 #define for_each_netdev_continue_rcu(net, d) \
2889 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2890 #define for_each_netdev_in_bond_rcu(bond, slave) \
2891 for_each_netdev_rcu(&init_net, slave) \
2892 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2893 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2895 static inline struct net_device *next_net_device(struct net_device *dev)
2897 struct list_head *lh;
2901 lh = dev->dev_list.next;
2902 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2905 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2907 struct list_head *lh;
2911 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2912 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2915 static inline struct net_device *first_net_device(struct net *net)
2917 return list_empty(&net->dev_base_head) ? NULL :
2918 net_device_entry(net->dev_base_head.next);
2921 static inline struct net_device *first_net_device_rcu(struct net *net)
2923 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2925 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2928 int netdev_boot_setup_check(struct net_device *dev);
2929 unsigned long netdev_boot_base(const char *prefix, int unit);
2930 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2931 const char *hwaddr);
2932 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2933 void dev_add_pack(struct packet_type *pt);
2934 void dev_remove_pack(struct packet_type *pt);
2935 void __dev_remove_pack(struct packet_type *pt);
2936 void dev_add_offload(struct packet_offload *po);
2937 void dev_remove_offload(struct packet_offload *po);
2939 int dev_get_iflink(const struct net_device *dev);
2940 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2941 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2942 struct net_device_path_stack *stack);
2943 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2944 unsigned short mask);
2945 struct net_device *dev_get_by_name(struct net *net, const char *name);
2946 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2947 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2948 int dev_alloc_name(struct net_device *dev, const char *name);
2949 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2950 void dev_close(struct net_device *dev);
2951 void dev_close_many(struct list_head *head, bool unlink);
2952 void dev_disable_lro(struct net_device *dev);
2953 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2954 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2955 struct net_device *sb_dev);
2956 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2957 struct net_device *sb_dev);
2959 int dev_queue_xmit(struct sk_buff *skb);
2960 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2961 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2963 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2967 ret = __dev_direct_xmit(skb, queue_id);
2968 if (!dev_xmit_complete(ret))
2973 int register_netdevice(struct net_device *dev);
2974 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2975 void unregister_netdevice_many(struct list_head *head);
2976 static inline void unregister_netdevice(struct net_device *dev)
2978 unregister_netdevice_queue(dev, NULL);
2981 int netdev_refcnt_read(const struct net_device *dev);
2982 void free_netdev(struct net_device *dev);
2983 void netdev_freemem(struct net_device *dev);
2984 int init_dummy_netdev(struct net_device *dev);
2986 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2987 struct sk_buff *skb,
2989 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2991 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2992 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2993 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2994 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2995 int netdev_get_name(struct net *net, char *name, int ifindex);
2996 int dev_restart(struct net_device *dev);
2997 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2998 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3000 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3002 return NAPI_GRO_CB(skb)->data_offset;
3005 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3007 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3010 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3012 NAPI_GRO_CB(skb)->data_offset += len;
3015 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3016 unsigned int offset)
3018 return NAPI_GRO_CB(skb)->frag0 + offset;
3021 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3023 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3026 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3028 NAPI_GRO_CB(skb)->frag0 = NULL;
3029 NAPI_GRO_CB(skb)->frag0_len = 0;
3032 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3033 unsigned int offset)
3035 if (!pskb_may_pull(skb, hlen))
3038 skb_gro_frag0_invalidate(skb);
3039 return skb->data + offset;
3042 static inline void *skb_gro_network_header(struct sk_buff *skb)
3044 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3045 skb_network_offset(skb);
3048 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3049 const void *start, unsigned int len)
3051 if (NAPI_GRO_CB(skb)->csum_valid)
3052 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3053 csum_partial(start, len, 0));
3056 /* GRO checksum functions. These are logical equivalents of the normal
3057 * checksum functions (in skbuff.h) except that they operate on the GRO
3058 * offsets and fields in sk_buff.
3061 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3063 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3065 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3068 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3072 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3073 skb_checksum_start_offset(skb) <
3074 skb_gro_offset(skb)) &&
3075 !skb_at_gro_remcsum_start(skb) &&
3076 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3077 (!zero_okay || check));
3080 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3083 if (NAPI_GRO_CB(skb)->csum_valid &&
3084 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3087 NAPI_GRO_CB(skb)->csum = psum;
3089 return __skb_gro_checksum_complete(skb);
3092 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3094 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3095 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3096 NAPI_GRO_CB(skb)->csum_cnt--;
3098 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3099 * verified a new top level checksum or an encapsulated one
3100 * during GRO. This saves work if we fallback to normal path.
3102 __skb_incr_checksum_unnecessary(skb);
3106 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3109 __sum16 __ret = 0; \
3110 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3111 __ret = __skb_gro_checksum_validate_complete(skb, \
3112 compute_pseudo(skb, proto)); \
3114 skb_gro_incr_csum_unnecessary(skb); \
3118 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3119 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3121 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3123 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3125 #define skb_gro_checksum_simple_validate(skb) \
3126 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3128 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3130 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3131 !NAPI_GRO_CB(skb)->csum_valid);
3134 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3137 NAPI_GRO_CB(skb)->csum = ~pseudo;
3138 NAPI_GRO_CB(skb)->csum_valid = 1;
3141 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3143 if (__skb_gro_checksum_convert_check(skb)) \
3144 __skb_gro_checksum_convert(skb, \
3145 compute_pseudo(skb, proto)); \
3148 struct gro_remcsum {
3153 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3159 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3160 unsigned int off, size_t hdrlen,
3161 int start, int offset,
3162 struct gro_remcsum *grc,
3166 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3168 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3171 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3175 ptr = skb_gro_header_fast(skb, off);
3176 if (skb_gro_header_hard(skb, off + plen)) {
3177 ptr = skb_gro_header_slow(skb, off + plen, off);
3182 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3185 /* Adjust skb->csum since we changed the packet */
3186 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3188 grc->offset = off + hdrlen + offset;
3194 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3195 struct gro_remcsum *grc)
3198 size_t plen = grc->offset + sizeof(u16);
3203 ptr = skb_gro_header_fast(skb, grc->offset);
3204 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3205 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3210 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3213 #ifdef CONFIG_XFRM_OFFLOAD
3214 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3216 if (PTR_ERR(pp) != -EINPROGRESS)
3217 NAPI_GRO_CB(skb)->flush |= flush;
3219 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3222 struct gro_remcsum *grc)
3224 if (PTR_ERR(pp) != -EINPROGRESS) {
3225 NAPI_GRO_CB(skb)->flush |= flush;
3226 skb_gro_remcsum_cleanup(skb, grc);
3227 skb->remcsum_offload = 0;
3231 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3233 NAPI_GRO_CB(skb)->flush |= flush;
3235 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3238 struct gro_remcsum *grc)
3240 NAPI_GRO_CB(skb)->flush |= flush;
3241 skb_gro_remcsum_cleanup(skb, grc);
3242 skb->remcsum_offload = 0;
3246 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3247 unsigned short type,
3248 const void *daddr, const void *saddr,
3251 if (!dev->header_ops || !dev->header_ops->create)
3254 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3257 static inline int dev_parse_header(const struct sk_buff *skb,
3258 unsigned char *haddr)
3260 const struct net_device *dev = skb->dev;
3262 if (!dev->header_ops || !dev->header_ops->parse)
3264 return dev->header_ops->parse(skb, haddr);
3267 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3269 const struct net_device *dev = skb->dev;
3271 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3273 return dev->header_ops->parse_protocol(skb);
3276 /* ll_header must have at least hard_header_len allocated */
3277 static inline bool dev_validate_header(const struct net_device *dev,
3278 char *ll_header, int len)
3280 if (likely(len >= dev->hard_header_len))
3282 if (len < dev->min_header_len)
3285 if (capable(CAP_SYS_RAWIO)) {
3286 memset(ll_header + len, 0, dev->hard_header_len - len);
3290 if (dev->header_ops && dev->header_ops->validate)
3291 return dev->header_ops->validate(ll_header, len);
3296 static inline bool dev_has_header(const struct net_device *dev)
3298 return dev->header_ops && dev->header_ops->create;
3301 #ifdef CONFIG_NET_FLOW_LIMIT
3302 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3303 struct sd_flow_limit {
3305 unsigned int num_buckets;
3306 unsigned int history_head;
3307 u16 history[FLOW_LIMIT_HISTORY];
3311 extern int netdev_flow_limit_table_len;
3312 #endif /* CONFIG_NET_FLOW_LIMIT */
3315 * Incoming packets are placed on per-CPU queues
3317 struct softnet_data {
3318 struct list_head poll_list;
3319 struct sk_buff_head process_queue;
3322 unsigned int processed;
3323 unsigned int time_squeeze;
3324 unsigned int received_rps;
3326 struct softnet_data *rps_ipi_list;
3328 #ifdef CONFIG_NET_FLOW_LIMIT
3329 struct sd_flow_limit __rcu *flow_limit;
3331 struct Qdisc *output_queue;
3332 struct Qdisc **output_queue_tailp;
3333 struct sk_buff *completion_queue;
3334 #ifdef CONFIG_XFRM_OFFLOAD
3335 struct sk_buff_head xfrm_backlog;
3337 /* written and read only by owning cpu: */
3343 /* input_queue_head should be written by cpu owning this struct,
3344 * and only read by other cpus. Worth using a cache line.
3346 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3348 /* Elements below can be accessed between CPUs for RPS/RFS */
3349 call_single_data_t csd ____cacheline_aligned_in_smp;
3350 struct softnet_data *rps_ipi_next;
3352 unsigned int input_queue_tail;
3354 unsigned int dropped;
3355 struct sk_buff_head input_pkt_queue;
3356 struct napi_struct backlog;
3360 static inline void input_queue_head_incr(struct softnet_data *sd)
3363 sd->input_queue_head++;
3367 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3368 unsigned int *qtail)
3371 *qtail = ++sd->input_queue_tail;
3375 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3377 static inline int dev_recursion_level(void)
3379 return this_cpu_read(softnet_data.xmit.recursion);
3382 #define XMIT_RECURSION_LIMIT 8
3383 static inline bool dev_xmit_recursion(void)
3385 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3386 XMIT_RECURSION_LIMIT);
3389 static inline void dev_xmit_recursion_inc(void)
3391 __this_cpu_inc(softnet_data.xmit.recursion);
3394 static inline void dev_xmit_recursion_dec(void)
3396 __this_cpu_dec(softnet_data.xmit.recursion);
3399 void __netif_schedule(struct Qdisc *q);
3400 void netif_schedule_queue(struct netdev_queue *txq);
3402 static inline void netif_tx_schedule_all(struct net_device *dev)
3406 for (i = 0; i < dev->num_tx_queues; i++)
3407 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3410 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3412 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3416 * netif_start_queue - allow transmit
3417 * @dev: network device
3419 * Allow upper layers to call the device hard_start_xmit routine.
3421 static inline void netif_start_queue(struct net_device *dev)
3423 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3426 static inline void netif_tx_start_all_queues(struct net_device *dev)
3430 for (i = 0; i < dev->num_tx_queues; i++) {
3431 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3432 netif_tx_start_queue(txq);
3436 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3439 * netif_wake_queue - restart transmit
3440 * @dev: network device
3442 * Allow upper layers to call the device hard_start_xmit routine.
3443 * Used for flow control when transmit resources are available.
3445 static inline void netif_wake_queue(struct net_device *dev)
3447 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3450 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3454 for (i = 0; i < dev->num_tx_queues; i++) {
3455 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3456 netif_tx_wake_queue(txq);
3460 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3462 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3466 * netif_stop_queue - stop transmitted packets
3467 * @dev: network device
3469 * Stop upper layers calling the device hard_start_xmit routine.
3470 * Used for flow control when transmit resources are unavailable.
3472 static inline void netif_stop_queue(struct net_device *dev)
3474 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3477 void netif_tx_stop_all_queues(struct net_device *dev);
3479 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3481 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3485 * netif_queue_stopped - test if transmit queue is flowblocked
3486 * @dev: network device
3488 * Test if transmit queue on device is currently unable to send.
3490 static inline bool netif_queue_stopped(const struct net_device *dev)
3492 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3495 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3497 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3501 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3503 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3507 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3509 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3513 * netdev_queue_set_dql_min_limit - set dql minimum limit
3514 * @dev_queue: pointer to transmit queue
3515 * @min_limit: dql minimum limit
3517 * Forces xmit_more() to return true until the minimum threshold
3518 * defined by @min_limit is reached (or until the tx queue is
3519 * empty). Warning: to be use with care, misuse will impact the
3522 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3523 unsigned int min_limit)
3526 dev_queue->dql.min_limit = min_limit;
3531 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3532 * @dev_queue: pointer to transmit queue
3534 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3535 * to give appropriate hint to the CPU.
3537 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3540 prefetchw(&dev_queue->dql.num_queued);
3545 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3546 * @dev_queue: pointer to transmit queue
3548 * BQL enabled drivers might use this helper in their TX completion path,
3549 * to give appropriate hint to the CPU.
3551 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3554 prefetchw(&dev_queue->dql.limit);
3558 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3562 dql_queued(&dev_queue->dql, bytes);
3564 if (likely(dql_avail(&dev_queue->dql) >= 0))
3567 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3570 * The XOFF flag must be set before checking the dql_avail below,
3571 * because in netdev_tx_completed_queue we update the dql_completed
3572 * before checking the XOFF flag.
3576 /* check again in case another CPU has just made room avail */
3577 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3578 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3582 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3583 * that they should not test BQL status themselves.
3584 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3586 * Returns true if the doorbell must be used to kick the NIC.
3588 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3594 dql_queued(&dev_queue->dql, bytes);
3596 return netif_tx_queue_stopped(dev_queue);
3598 netdev_tx_sent_queue(dev_queue, bytes);
3603 * netdev_sent_queue - report the number of bytes queued to hardware
3604 * @dev: network device
3605 * @bytes: number of bytes queued to the hardware device queue
3607 * Report the number of bytes queued for sending/completion to the network
3608 * device hardware queue. @bytes should be a good approximation and should
3609 * exactly match netdev_completed_queue() @bytes
3611 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3613 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3616 static inline bool __netdev_sent_queue(struct net_device *dev,
3620 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3624 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3625 unsigned int pkts, unsigned int bytes)
3628 if (unlikely(!bytes))
3631 dql_completed(&dev_queue->dql, bytes);
3634 * Without the memory barrier there is a small possiblity that
3635 * netdev_tx_sent_queue will miss the update and cause the queue to
3636 * be stopped forever
3640 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3643 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3644 netif_schedule_queue(dev_queue);
3649 * netdev_completed_queue - report bytes and packets completed by device
3650 * @dev: network device
3651 * @pkts: actual number of packets sent over the medium
3652 * @bytes: actual number of bytes sent over the medium
3654 * Report the number of bytes and packets transmitted by the network device
3655 * hardware queue over the physical medium, @bytes must exactly match the
3656 * @bytes amount passed to netdev_sent_queue()
3658 static inline void netdev_completed_queue(struct net_device *dev,
3659 unsigned int pkts, unsigned int bytes)
3661 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3664 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3667 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3673 * netdev_reset_queue - reset the packets and bytes count of a network device
3674 * @dev_queue: network device
3676 * Reset the bytes and packet count of a network device and clear the
3677 * software flow control OFF bit for this network device
3679 static inline void netdev_reset_queue(struct net_device *dev_queue)
3681 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3685 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3686 * @dev: network device
3687 * @queue_index: given tx queue index
3689 * Returns 0 if given tx queue index >= number of device tx queues,
3690 * otherwise returns the originally passed tx queue index.
3692 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3694 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3695 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3696 dev->name, queue_index,
3697 dev->real_num_tx_queues);
3705 * netif_running - test if up
3706 * @dev: network device
3708 * Test if the device has been brought up.
3710 static inline bool netif_running(const struct net_device *dev)
3712 return test_bit(__LINK_STATE_START, &dev->state);
3716 * Routines to manage the subqueues on a device. We only need start,
3717 * stop, and a check if it's stopped. All other device management is
3718 * done at the overall netdevice level.
3719 * Also test the device if we're multiqueue.
3723 * netif_start_subqueue - allow sending packets on subqueue
3724 * @dev: network device
3725 * @queue_index: sub queue index
3727 * Start individual transmit queue of a device with multiple transmit queues.
3729 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3731 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3733 netif_tx_start_queue(txq);
3737 * netif_stop_subqueue - stop sending packets on subqueue
3738 * @dev: network device
3739 * @queue_index: sub queue index
3741 * Stop individual transmit queue of a device with multiple transmit queues.
3743 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3745 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3746 netif_tx_stop_queue(txq);
3750 * __netif_subqueue_stopped - test status of subqueue
3751 * @dev: network device
3752 * @queue_index: sub queue index
3754 * Check individual transmit queue of a device with multiple transmit queues.
3756 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3759 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3761 return netif_tx_queue_stopped(txq);
3765 * netif_subqueue_stopped - test status of subqueue
3766 * @dev: network device
3767 * @skb: sub queue buffer pointer
3769 * Check individual transmit queue of a device with multiple transmit queues.
3771 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3772 struct sk_buff *skb)
3774 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3778 * netif_wake_subqueue - allow sending packets on subqueue
3779 * @dev: network device
3780 * @queue_index: sub queue index
3782 * Resume individual transmit queue of a device with multiple transmit queues.
3784 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3786 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3788 netif_tx_wake_queue(txq);
3792 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3794 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3795 u16 index, enum xps_map_type type);
3798 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3799 * @j: CPU/Rx queue index
3800 * @mask: bitmask of all cpus/rx queues
3801 * @nr_bits: number of bits in the bitmask
3803 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3805 static inline bool netif_attr_test_mask(unsigned long j,
3806 const unsigned long *mask,
3807 unsigned int nr_bits)
3809 cpu_max_bits_warn(j, nr_bits);
3810 return test_bit(j, mask);
3814 * netif_attr_test_online - Test for online CPU/Rx queue
3815 * @j: CPU/Rx queue index
3816 * @online_mask: bitmask for CPUs/Rx queues that are online
3817 * @nr_bits: number of bits in the bitmask
3819 * Returns true if a CPU/Rx queue is online.
3821 static inline bool netif_attr_test_online(unsigned long j,
3822 const unsigned long *online_mask,
3823 unsigned int nr_bits)
3825 cpu_max_bits_warn(j, nr_bits);
3828 return test_bit(j, online_mask);
3830 return (j < nr_bits);
3834 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3835 * @n: CPU/Rx queue index
3836 * @srcp: the cpumask/Rx queue mask pointer
3837 * @nr_bits: number of bits in the bitmask
3839 * Returns >= nr_bits if no further CPUs/Rx queues set.
3841 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3842 unsigned int nr_bits)
3844 /* -1 is a legal arg here. */
3846 cpu_max_bits_warn(n, nr_bits);
3849 return find_next_bit(srcp, nr_bits, n + 1);
3855 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3856 * @n: CPU/Rx queue index
3857 * @src1p: the first CPUs/Rx queues mask pointer
3858 * @src2p: the second CPUs/Rx queues mask pointer
3859 * @nr_bits: number of bits in the bitmask
3861 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3863 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3864 const unsigned long *src2p,
3865 unsigned int nr_bits)
3867 /* -1 is a legal arg here. */
3869 cpu_max_bits_warn(n, nr_bits);
3872 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3874 return find_next_bit(src1p, nr_bits, n + 1);
3876 return find_next_bit(src2p, nr_bits, n + 1);
3881 static inline int netif_set_xps_queue(struct net_device *dev,
3882 const struct cpumask *mask,
3888 static inline int __netif_set_xps_queue(struct net_device *dev,
3889 const unsigned long *mask,
3890 u16 index, enum xps_map_type type)
3897 * netif_is_multiqueue - test if device has multiple transmit queues
3898 * @dev: network device
3900 * Check if device has multiple transmit queues
3902 static inline bool netif_is_multiqueue(const struct net_device *dev)
3904 return dev->num_tx_queues > 1;
3907 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3910 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3912 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3915 dev->real_num_rx_queues = rxqs;
3920 static inline struct netdev_rx_queue *
3921 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3923 return dev->_rx + rxq;
3927 static inline unsigned int get_netdev_rx_queue_index(
3928 struct netdev_rx_queue *queue)
3930 struct net_device *dev = queue->dev;
3931 int index = queue - dev->_rx;
3933 BUG_ON(index >= dev->num_rx_queues);
3938 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3939 int netif_get_num_default_rss_queues(void);
3941 enum skb_free_reason {
3942 SKB_REASON_CONSUMED,
3946 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3947 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3950 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3951 * interrupt context or with hardware interrupts being disabled.
3952 * (in_irq() || irqs_disabled())
3954 * We provide four helpers that can be used in following contexts :
3956 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3957 * replacing kfree_skb(skb)
3959 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3960 * Typically used in place of consume_skb(skb) in TX completion path
3962 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3963 * replacing kfree_skb(skb)
3965 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3966 * and consumed a packet. Used in place of consume_skb(skb)
3968 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3970 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3973 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3975 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3978 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3980 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3983 static inline void dev_consume_skb_any(struct sk_buff *skb)
3985 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3988 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3989 struct bpf_prog *xdp_prog);
3990 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3991 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3992 int netif_rx(struct sk_buff *skb);
3993 int netif_rx_ni(struct sk_buff *skb);
3994 int netif_rx_any_context(struct sk_buff *skb);
3995 int netif_receive_skb(struct sk_buff *skb);
3996 int netif_receive_skb_core(struct sk_buff *skb);
3997 void netif_receive_skb_list(struct list_head *head);
3998 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3999 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4000 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4001 gro_result_t napi_gro_frags(struct napi_struct *napi);
4002 struct packet_offload *gro_find_receive_by_type(__be16 type);
4003 struct packet_offload *gro_find_complete_by_type(__be16 type);
4005 static inline void napi_free_frags(struct napi_struct *napi)
4007 kfree_skb(napi->skb);
4011 bool netdev_is_rx_handler_busy(struct net_device *dev);
4012 int netdev_rx_handler_register(struct net_device *dev,
4013 rx_handler_func_t *rx_handler,
4014 void *rx_handler_data);
4015 void netdev_rx_handler_unregister(struct net_device *dev);
4017 bool dev_valid_name(const char *name);
4018 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4019 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4020 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4021 void __user *data, bool *need_copyout);
4022 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4023 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4024 unsigned int dev_get_flags(const struct net_device *);
4025 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4026 struct netlink_ext_ack *extack);
4027 int dev_change_flags(struct net_device *dev, unsigned int flags,
4028 struct netlink_ext_ack *extack);
4029 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4030 unsigned int gchanges);
4031 int dev_change_name(struct net_device *, const char *);
4032 int dev_set_alias(struct net_device *, const char *, size_t);
4033 int dev_get_alias(const struct net_device *, char *, size_t);
4034 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4035 const char *pat, int new_ifindex);
4037 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4040 return __dev_change_net_namespace(dev, net, pat, 0);
4042 int __dev_set_mtu(struct net_device *, int);
4043 int dev_validate_mtu(struct net_device *dev, int mtu,
4044 struct netlink_ext_ack *extack);
4045 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4046 struct netlink_ext_ack *extack);
4047 int dev_set_mtu(struct net_device *, int);
4048 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4049 void dev_set_group(struct net_device *, int);
4050 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4051 struct netlink_ext_ack *extack);
4052 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4053 struct netlink_ext_ack *extack);
4054 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4055 struct netlink_ext_ack *extack);
4056 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4057 int dev_change_carrier(struct net_device *, bool new_carrier);
4058 int dev_get_phys_port_id(struct net_device *dev,
4059 struct netdev_phys_item_id *ppid);
4060 int dev_get_phys_port_name(struct net_device *dev,
4061 char *name, size_t len);
4062 int dev_get_port_parent_id(struct net_device *dev,
4063 struct netdev_phys_item_id *ppid, bool recurse);
4064 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4065 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4066 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4067 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4069 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4070 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4071 struct netdev_queue *txq, int *ret);
4073 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4074 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4075 int fd, int expected_fd, u32 flags);
4076 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4077 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4079 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4080 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4081 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4082 bool is_skb_forwardable(const struct net_device *dev,
4083 const struct sk_buff *skb);
4085 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4086 const struct sk_buff *skb,
4087 const bool check_mtu)
4089 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4092 if (!(dev->flags & IFF_UP))
4098 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4099 if (skb->len <= len)
4102 /* if TSO is enabled, we don't care about the length as the packet
4103 * could be forwarded without being segmented before
4105 if (skb_is_gso(skb))
4111 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4112 struct sk_buff *skb,
4113 const bool check_mtu)
4115 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4116 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4117 atomic_long_inc(&dev->rx_dropped);
4122 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4127 bool dev_nit_active(struct net_device *dev);
4128 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4130 extern int netdev_budget;
4131 extern unsigned int netdev_budget_usecs;
4133 /* Called by rtnetlink.c:rtnl_unlock() */
4134 void netdev_run_todo(void);
4137 * dev_put - release reference to device
4138 * @dev: network device
4140 * Release reference to device to allow it to be freed.
4142 static inline void dev_put(struct net_device *dev)
4144 #ifdef CONFIG_PCPU_DEV_REFCNT
4145 this_cpu_dec(*dev->pcpu_refcnt);
4147 refcount_dec(&dev->dev_refcnt);
4152 * dev_hold - get reference to device
4153 * @dev: network device
4155 * Hold reference to device to keep it from being freed.
4157 static inline void dev_hold(struct net_device *dev)
4159 #ifdef CONFIG_PCPU_DEV_REFCNT
4160 this_cpu_inc(*dev->pcpu_refcnt);
4162 refcount_inc(&dev->dev_refcnt);
4166 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4167 * and _off may be called from IRQ context, but it is caller
4168 * who is responsible for serialization of these calls.
4170 * The name carrier is inappropriate, these functions should really be
4171 * called netif_lowerlayer_*() because they represent the state of any
4172 * kind of lower layer not just hardware media.
4175 void linkwatch_init_dev(struct net_device *dev);
4176 void linkwatch_fire_event(struct net_device *dev);
4177 void linkwatch_forget_dev(struct net_device *dev);
4180 * netif_carrier_ok - test if carrier present
4181 * @dev: network device
4183 * Check if carrier is present on device
4185 static inline bool netif_carrier_ok(const struct net_device *dev)
4187 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4190 unsigned long dev_trans_start(struct net_device *dev);
4192 void __netdev_watchdog_up(struct net_device *dev);
4194 void netif_carrier_on(struct net_device *dev);
4195 void netif_carrier_off(struct net_device *dev);
4196 void netif_carrier_event(struct net_device *dev);
4199 * netif_dormant_on - mark device as dormant.
4200 * @dev: network device
4202 * Mark device as dormant (as per RFC2863).
4204 * The dormant state indicates that the relevant interface is not
4205 * actually in a condition to pass packets (i.e., it is not 'up') but is
4206 * in a "pending" state, waiting for some external event. For "on-
4207 * demand" interfaces, this new state identifies the situation where the
4208 * interface is waiting for events to place it in the up state.
4210 static inline void netif_dormant_on(struct net_device *dev)
4212 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4213 linkwatch_fire_event(dev);
4217 * netif_dormant_off - set device as not dormant.
4218 * @dev: network device
4220 * Device is not in dormant state.
4222 static inline void netif_dormant_off(struct net_device *dev)
4224 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4225 linkwatch_fire_event(dev);
4229 * netif_dormant - test if device is dormant
4230 * @dev: network device
4232 * Check if device is dormant.
4234 static inline bool netif_dormant(const struct net_device *dev)
4236 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4241 * netif_testing_on - mark device as under test.
4242 * @dev: network device
4244 * Mark device as under test (as per RFC2863).
4246 * The testing state indicates that some test(s) must be performed on
4247 * the interface. After completion, of the test, the interface state
4248 * will change to up, dormant, or down, as appropriate.
4250 static inline void netif_testing_on(struct net_device *dev)
4252 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4253 linkwatch_fire_event(dev);
4257 * netif_testing_off - set device as not under test.
4258 * @dev: network device
4260 * Device is not in testing state.
4262 static inline void netif_testing_off(struct net_device *dev)
4264 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4265 linkwatch_fire_event(dev);
4269 * netif_testing - test if device is under test
4270 * @dev: network device
4272 * Check if device is under test
4274 static inline bool netif_testing(const struct net_device *dev)
4276 return test_bit(__LINK_STATE_TESTING, &dev->state);
4281 * netif_oper_up - test if device is operational
4282 * @dev: network device
4284 * Check if carrier is operational
4286 static inline bool netif_oper_up(const struct net_device *dev)
4288 return (dev->operstate == IF_OPER_UP ||
4289 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4293 * netif_device_present - is device available or removed
4294 * @dev: network device
4296 * Check if device has not been removed from system.
4298 static inline bool netif_device_present(const struct net_device *dev)
4300 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4303 void netif_device_detach(struct net_device *dev);
4305 void netif_device_attach(struct net_device *dev);
4308 * Network interface message level settings
4313 NETIF_MSG_PROBE_BIT,
4315 NETIF_MSG_TIMER_BIT,
4316 NETIF_MSG_IFDOWN_BIT,
4318 NETIF_MSG_RX_ERR_BIT,
4319 NETIF_MSG_TX_ERR_BIT,
4320 NETIF_MSG_TX_QUEUED_BIT,
4322 NETIF_MSG_TX_DONE_BIT,
4323 NETIF_MSG_RX_STATUS_BIT,
4324 NETIF_MSG_PKTDATA_BIT,
4328 /* When you add a new bit above, update netif_msg_class_names array
4329 * in net/ethtool/common.c
4331 NETIF_MSG_CLASS_COUNT,
4333 /* Both ethtool_ops interface and internal driver implementation use u32 */
4334 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4336 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4337 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4339 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4340 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4341 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4342 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4343 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4344 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4345 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4346 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4347 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4348 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4349 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4350 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4351 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4352 #define NETIF_MSG_HW __NETIF_MSG(HW)
4353 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4355 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4356 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4357 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4358 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4359 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4360 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4361 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4362 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4363 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4364 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4365 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4366 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4367 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4368 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4369 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4371 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4374 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4375 return default_msg_enable_bits;
4376 if (debug_value == 0) /* no output */
4378 /* set low N bits */
4379 return (1U << debug_value) - 1;
4382 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4384 spin_lock(&txq->_xmit_lock);
4385 txq->xmit_lock_owner = cpu;
4388 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4390 __acquire(&txq->_xmit_lock);
4394 static inline void __netif_tx_release(struct netdev_queue *txq)
4396 __release(&txq->_xmit_lock);
4399 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4401 spin_lock_bh(&txq->_xmit_lock);
4402 txq->xmit_lock_owner = smp_processor_id();
4405 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4407 bool ok = spin_trylock(&txq->_xmit_lock);
4409 txq->xmit_lock_owner = smp_processor_id();
4413 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4415 txq->xmit_lock_owner = -1;
4416 spin_unlock(&txq->_xmit_lock);
4419 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4421 txq->xmit_lock_owner = -1;
4422 spin_unlock_bh(&txq->_xmit_lock);
4425 static inline void txq_trans_update(struct netdev_queue *txq)
4427 if (txq->xmit_lock_owner != -1)
4428 txq->trans_start = jiffies;
4431 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4432 static inline void netif_trans_update(struct net_device *dev)
4434 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4436 if (txq->trans_start != jiffies)
4437 txq->trans_start = jiffies;
4441 * netif_tx_lock - grab network device transmit lock
4442 * @dev: network device
4444 * Get network device transmit lock
4446 static inline void netif_tx_lock(struct net_device *dev)
4451 spin_lock(&dev->tx_global_lock);
4452 cpu = smp_processor_id();
4453 for (i = 0; i < dev->num_tx_queues; i++) {
4454 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4456 /* We are the only thread of execution doing a
4457 * freeze, but we have to grab the _xmit_lock in
4458 * order to synchronize with threads which are in
4459 * the ->hard_start_xmit() handler and already
4460 * checked the frozen bit.
4462 __netif_tx_lock(txq, cpu);
4463 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4464 __netif_tx_unlock(txq);
4468 static inline void netif_tx_lock_bh(struct net_device *dev)
4474 static inline void netif_tx_unlock(struct net_device *dev)
4478 for (i = 0; i < dev->num_tx_queues; i++) {
4479 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4481 /* No need to grab the _xmit_lock here. If the
4482 * queue is not stopped for another reason, we
4485 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4486 netif_schedule_queue(txq);
4488 spin_unlock(&dev->tx_global_lock);
4491 static inline void netif_tx_unlock_bh(struct net_device *dev)
4493 netif_tx_unlock(dev);
4497 #define HARD_TX_LOCK(dev, txq, cpu) { \
4498 if ((dev->features & NETIF_F_LLTX) == 0) { \
4499 __netif_tx_lock(txq, cpu); \
4501 __netif_tx_acquire(txq); \
4505 #define HARD_TX_TRYLOCK(dev, txq) \
4506 (((dev->features & NETIF_F_LLTX) == 0) ? \
4507 __netif_tx_trylock(txq) : \
4508 __netif_tx_acquire(txq))
4510 #define HARD_TX_UNLOCK(dev, txq) { \
4511 if ((dev->features & NETIF_F_LLTX) == 0) { \
4512 __netif_tx_unlock(txq); \
4514 __netif_tx_release(txq); \
4518 static inline void netif_tx_disable(struct net_device *dev)
4524 cpu = smp_processor_id();
4525 spin_lock(&dev->tx_global_lock);
4526 for (i = 0; i < dev->num_tx_queues; i++) {
4527 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4529 __netif_tx_lock(txq, cpu);
4530 netif_tx_stop_queue(txq);
4531 __netif_tx_unlock(txq);
4533 spin_unlock(&dev->tx_global_lock);
4537 static inline void netif_addr_lock(struct net_device *dev)
4539 unsigned char nest_level = 0;
4541 #ifdef CONFIG_LOCKDEP
4542 nest_level = dev->nested_level;
4544 spin_lock_nested(&dev->addr_list_lock, nest_level);
4547 static inline void netif_addr_lock_bh(struct net_device *dev)
4549 unsigned char nest_level = 0;
4551 #ifdef CONFIG_LOCKDEP
4552 nest_level = dev->nested_level;
4555 spin_lock_nested(&dev->addr_list_lock, nest_level);
4558 static inline void netif_addr_unlock(struct net_device *dev)
4560 spin_unlock(&dev->addr_list_lock);
4563 static inline void netif_addr_unlock_bh(struct net_device *dev)
4565 spin_unlock_bh(&dev->addr_list_lock);
4569 * dev_addrs walker. Should be used only for read access. Call with
4570 * rcu_read_lock held.
4572 #define for_each_dev_addr(dev, ha) \
4573 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4575 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4577 void ether_setup(struct net_device *dev);
4579 /* Support for loadable net-drivers */
4580 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4581 unsigned char name_assign_type,
4582 void (*setup)(struct net_device *),
4583 unsigned int txqs, unsigned int rxqs);
4584 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4585 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4587 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4588 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4591 int register_netdev(struct net_device *dev);
4592 void unregister_netdev(struct net_device *dev);
4594 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4596 /* General hardware address lists handling functions */
4597 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4598 struct netdev_hw_addr_list *from_list, int addr_len);
4599 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4600 struct netdev_hw_addr_list *from_list, int addr_len);
4601 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4602 struct net_device *dev,
4603 int (*sync)(struct net_device *, const unsigned char *),
4604 int (*unsync)(struct net_device *,
4605 const unsigned char *));
4606 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4607 struct net_device *dev,
4608 int (*sync)(struct net_device *,
4609 const unsigned char *, int),
4610 int (*unsync)(struct net_device *,
4611 const unsigned char *, int));
4612 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4613 struct net_device *dev,
4614 int (*unsync)(struct net_device *,
4615 const unsigned char *, int));
4616 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4617 struct net_device *dev,
4618 int (*unsync)(struct net_device *,
4619 const unsigned char *));
4620 void __hw_addr_init(struct netdev_hw_addr_list *list);
4622 /* Functions used for device addresses handling */
4623 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4624 unsigned char addr_type);
4625 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4626 unsigned char addr_type);
4627 void dev_addr_flush(struct net_device *dev);
4628 int dev_addr_init(struct net_device *dev);
4630 /* Functions used for unicast addresses handling */
4631 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4632 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4633 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4634 int dev_uc_sync(struct net_device *to, struct net_device *from);
4635 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4636 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4637 void dev_uc_flush(struct net_device *dev);
4638 void dev_uc_init(struct net_device *dev);
4641 * __dev_uc_sync - Synchonize device's unicast list
4642 * @dev: device to sync
4643 * @sync: function to call if address should be added
4644 * @unsync: function to call if address should be removed
4646 * Add newly added addresses to the interface, and release
4647 * addresses that have been deleted.
4649 static inline int __dev_uc_sync(struct net_device *dev,
4650 int (*sync)(struct net_device *,
4651 const unsigned char *),
4652 int (*unsync)(struct net_device *,
4653 const unsigned char *))
4655 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4659 * __dev_uc_unsync - Remove synchronized addresses from device
4660 * @dev: device to sync
4661 * @unsync: function to call if address should be removed
4663 * Remove all addresses that were added to the device by dev_uc_sync().
4665 static inline void __dev_uc_unsync(struct net_device *dev,
4666 int (*unsync)(struct net_device *,
4667 const unsigned char *))
4669 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4672 /* Functions used for multicast addresses handling */
4673 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4674 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4675 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4676 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4677 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4678 int dev_mc_sync(struct net_device *to, struct net_device *from);
4679 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4680 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4681 void dev_mc_flush(struct net_device *dev);
4682 void dev_mc_init(struct net_device *dev);
4685 * __dev_mc_sync - Synchonize device's multicast list
4686 * @dev: device to sync
4687 * @sync: function to call if address should be added
4688 * @unsync: function to call if address should be removed
4690 * Add newly added addresses to the interface, and release
4691 * addresses that have been deleted.
4693 static inline int __dev_mc_sync(struct net_device *dev,
4694 int (*sync)(struct net_device *,
4695 const unsigned char *),
4696 int (*unsync)(struct net_device *,
4697 const unsigned char *))
4699 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4703 * __dev_mc_unsync - Remove synchronized addresses from device
4704 * @dev: device to sync
4705 * @unsync: function to call if address should be removed
4707 * Remove all addresses that were added to the device by dev_mc_sync().
4709 static inline void __dev_mc_unsync(struct net_device *dev,
4710 int (*unsync)(struct net_device *,
4711 const unsigned char *))
4713 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4716 /* Functions used for secondary unicast and multicast support */
4717 void dev_set_rx_mode(struct net_device *dev);
4718 void __dev_set_rx_mode(struct net_device *dev);
4719 int dev_set_promiscuity(struct net_device *dev, int inc);
4720 int dev_set_allmulti(struct net_device *dev, int inc);
4721 void netdev_state_change(struct net_device *dev);
4722 void __netdev_notify_peers(struct net_device *dev);
4723 void netdev_notify_peers(struct net_device *dev);
4724 void netdev_features_change(struct net_device *dev);
4725 /* Load a device via the kmod */
4726 void dev_load(struct net *net, const char *name);
4727 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4728 struct rtnl_link_stats64 *storage);
4729 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4730 const struct net_device_stats *netdev_stats);
4731 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4732 const struct pcpu_sw_netstats __percpu *netstats);
4733 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4735 extern int netdev_max_backlog;
4736 extern int netdev_tstamp_prequeue;
4737 extern int netdev_unregister_timeout_secs;
4738 extern int weight_p;
4739 extern int dev_weight_rx_bias;
4740 extern int dev_weight_tx_bias;
4741 extern int dev_rx_weight;
4742 extern int dev_tx_weight;
4743 extern int gro_normal_batch;
4746 NESTED_SYNC_IMM_BIT,
4747 NESTED_SYNC_TODO_BIT,
4750 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4751 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4753 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4754 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4756 struct netdev_nested_priv {
4757 unsigned char flags;
4761 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4762 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4763 struct list_head **iter);
4764 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4765 struct list_head **iter);
4767 #ifdef CONFIG_LOCKDEP
4768 static LIST_HEAD(net_unlink_list);
4770 static inline void net_unlink_todo(struct net_device *dev)
4772 if (list_empty(&dev->unlink_list))
4773 list_add_tail(&dev->unlink_list, &net_unlink_list);
4777 /* iterate through upper list, must be called under RCU read lock */
4778 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4779 for (iter = &(dev)->adj_list.upper, \
4780 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4782 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4784 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4785 int (*fn)(struct net_device *upper_dev,
4786 struct netdev_nested_priv *priv),
4787 struct netdev_nested_priv *priv);
4789 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4790 struct net_device *upper_dev);
4792 bool netdev_has_any_upper_dev(struct net_device *dev);
4794 void *netdev_lower_get_next_private(struct net_device *dev,
4795 struct list_head **iter);
4796 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4797 struct list_head **iter);
4799 #define netdev_for_each_lower_private(dev, priv, iter) \
4800 for (iter = (dev)->adj_list.lower.next, \
4801 priv = netdev_lower_get_next_private(dev, &(iter)); \
4803 priv = netdev_lower_get_next_private(dev, &(iter)))
4805 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4806 for (iter = &(dev)->adj_list.lower, \
4807 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4809 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4811 void *netdev_lower_get_next(struct net_device *dev,
4812 struct list_head **iter);
4814 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4815 for (iter = (dev)->adj_list.lower.next, \
4816 ldev = netdev_lower_get_next(dev, &(iter)); \
4818 ldev = netdev_lower_get_next(dev, &(iter)))
4820 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4821 struct list_head **iter);
4822 int netdev_walk_all_lower_dev(struct net_device *dev,
4823 int (*fn)(struct net_device *lower_dev,
4824 struct netdev_nested_priv *priv),
4825 struct netdev_nested_priv *priv);
4826 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4827 int (*fn)(struct net_device *lower_dev,
4828 struct netdev_nested_priv *priv),
4829 struct netdev_nested_priv *priv);
4831 void *netdev_adjacent_get_private(struct list_head *adj_list);
4832 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4833 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4834 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4835 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4836 struct netlink_ext_ack *extack);
4837 int netdev_master_upper_dev_link(struct net_device *dev,
4838 struct net_device *upper_dev,
4839 void *upper_priv, void *upper_info,
4840 struct netlink_ext_ack *extack);
4841 void netdev_upper_dev_unlink(struct net_device *dev,
4842 struct net_device *upper_dev);
4843 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4844 struct net_device *new_dev,
4845 struct net_device *dev,
4846 struct netlink_ext_ack *extack);
4847 void netdev_adjacent_change_commit(struct net_device *old_dev,
4848 struct net_device *new_dev,
4849 struct net_device *dev);
4850 void netdev_adjacent_change_abort(struct net_device *old_dev,
4851 struct net_device *new_dev,
4852 struct net_device *dev);
4853 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4854 void *netdev_lower_dev_get_private(struct net_device *dev,
4855 struct net_device *lower_dev);
4856 void netdev_lower_state_changed(struct net_device *lower_dev,
4857 void *lower_state_info);
4859 /* RSS keys are 40 or 52 bytes long */
4860 #define NETDEV_RSS_KEY_LEN 52
4861 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4862 void netdev_rss_key_fill(void *buffer, size_t len);
4864 int skb_checksum_help(struct sk_buff *skb);
4865 int skb_crc32c_csum_help(struct sk_buff *skb);
4866 int skb_csum_hwoffload_help(struct sk_buff *skb,
4867 const netdev_features_t features);
4869 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4870 netdev_features_t features, bool tx_path);
4871 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4872 netdev_features_t features);
4874 struct netdev_bonding_info {
4879 struct netdev_notifier_bonding_info {
4880 struct netdev_notifier_info info; /* must be first */
4881 struct netdev_bonding_info bonding_info;
4884 void netdev_bonding_info_change(struct net_device *dev,
4885 struct netdev_bonding_info *bonding_info);
4887 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4888 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4890 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4897 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4899 return __skb_gso_segment(skb, features, true);
4901 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4903 static inline bool can_checksum_protocol(netdev_features_t features,
4906 if (protocol == htons(ETH_P_FCOE))
4907 return !!(features & NETIF_F_FCOE_CRC);
4909 /* Assume this is an IP checksum (not SCTP CRC) */
4911 if (features & NETIF_F_HW_CSUM) {
4912 /* Can checksum everything */
4917 case htons(ETH_P_IP):
4918 return !!(features & NETIF_F_IP_CSUM);
4919 case htons(ETH_P_IPV6):
4920 return !!(features & NETIF_F_IPV6_CSUM);
4927 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4929 static inline void netdev_rx_csum_fault(struct net_device *dev,
4930 struct sk_buff *skb)
4934 /* rx skb timestamps */
4935 void net_enable_timestamp(void);
4936 void net_disable_timestamp(void);
4938 #ifdef CONFIG_PROC_FS
4939 int __init dev_proc_init(void);
4941 #define dev_proc_init() 0
4944 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4945 struct sk_buff *skb, struct net_device *dev,
4948 __this_cpu_write(softnet_data.xmit.more, more);
4949 return ops->ndo_start_xmit(skb, dev);
4952 static inline bool netdev_xmit_more(void)
4954 return __this_cpu_read(softnet_data.xmit.more);
4957 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4958 struct netdev_queue *txq, bool more)
4960 const struct net_device_ops *ops = dev->netdev_ops;
4963 rc = __netdev_start_xmit(ops, skb, dev, more);
4964 if (rc == NETDEV_TX_OK)
4965 txq_trans_update(txq);
4970 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4972 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4975 extern const struct kobj_ns_type_operations net_ns_type_operations;
4977 const char *netdev_drivername(const struct net_device *dev);
4979 void linkwatch_run_queue(void);
4981 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4982 netdev_features_t f2)
4984 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4985 if (f1 & NETIF_F_HW_CSUM)
4986 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4988 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4994 static inline netdev_features_t netdev_get_wanted_features(
4995 struct net_device *dev)
4997 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4999 netdev_features_t netdev_increment_features(netdev_features_t all,
5000 netdev_features_t one, netdev_features_t mask);
5002 /* Allow TSO being used on stacked device :
5003 * Performing the GSO segmentation before last device
5004 * is a performance improvement.
5006 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5007 netdev_features_t mask)
5009 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5012 int __netdev_update_features(struct net_device *dev);
5013 void netdev_update_features(struct net_device *dev);
5014 void netdev_change_features(struct net_device *dev);
5016 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5017 struct net_device *dev);
5019 netdev_features_t passthru_features_check(struct sk_buff *skb,
5020 struct net_device *dev,
5021 netdev_features_t features);
5022 netdev_features_t netif_skb_features(struct sk_buff *skb);
5024 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5026 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5028 /* check flags correspondence */
5029 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5030 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5031 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5032 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5033 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5034 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5035 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5036 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5037 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5038 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5039 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5040 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5041 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5042 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5043 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5044 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5045 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5046 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5047 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5049 return (features & feature) == feature;
5052 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5054 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5055 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5058 static inline bool netif_needs_gso(struct sk_buff *skb,
5059 netdev_features_t features)
5061 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5062 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5063 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5066 static inline void netif_set_gso_max_size(struct net_device *dev,
5069 dev->gso_max_size = size;
5072 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5073 int pulled_hlen, u16 mac_offset,
5076 skb->protocol = protocol;
5077 skb->encapsulation = 1;
5078 skb_push(skb, pulled_hlen);
5079 skb_reset_transport_header(skb);
5080 skb->mac_header = mac_offset;
5081 skb->network_header = skb->mac_header + mac_len;
5082 skb->mac_len = mac_len;
5085 static inline bool netif_is_macsec(const struct net_device *dev)
5087 return dev->priv_flags & IFF_MACSEC;
5090 static inline bool netif_is_macvlan(const struct net_device *dev)
5092 return dev->priv_flags & IFF_MACVLAN;
5095 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5097 return dev->priv_flags & IFF_MACVLAN_PORT;
5100 static inline bool netif_is_bond_master(const struct net_device *dev)
5102 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5105 static inline bool netif_is_bond_slave(const struct net_device *dev)
5107 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5110 static inline bool netif_supports_nofcs(struct net_device *dev)
5112 return dev->priv_flags & IFF_SUPP_NOFCS;
5115 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5117 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5120 static inline bool netif_is_l3_master(const struct net_device *dev)
5122 return dev->priv_flags & IFF_L3MDEV_MASTER;
5125 static inline bool netif_is_l3_slave(const struct net_device *dev)
5127 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5130 static inline bool netif_is_bridge_master(const struct net_device *dev)
5132 return dev->priv_flags & IFF_EBRIDGE;
5135 static inline bool netif_is_bridge_port(const struct net_device *dev)
5137 return dev->priv_flags & IFF_BRIDGE_PORT;
5140 static inline bool netif_is_ovs_master(const struct net_device *dev)
5142 return dev->priv_flags & IFF_OPENVSWITCH;
5145 static inline bool netif_is_ovs_port(const struct net_device *dev)
5147 return dev->priv_flags & IFF_OVS_DATAPATH;
5150 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5152 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5155 static inline bool netif_is_team_master(const struct net_device *dev)
5157 return dev->priv_flags & IFF_TEAM;
5160 static inline bool netif_is_team_port(const struct net_device *dev)
5162 return dev->priv_flags & IFF_TEAM_PORT;
5165 static inline bool netif_is_lag_master(const struct net_device *dev)
5167 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5170 static inline bool netif_is_lag_port(const struct net_device *dev)
5172 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5175 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5177 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5180 static inline bool netif_is_failover(const struct net_device *dev)
5182 return dev->priv_flags & IFF_FAILOVER;
5185 static inline bool netif_is_failover_slave(const struct net_device *dev)
5187 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5190 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5191 static inline void netif_keep_dst(struct net_device *dev)
5193 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5196 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5197 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5199 /* TODO: reserve and use an additional IFF bit, if we get more users */
5200 return dev->priv_flags & IFF_MACSEC;
5203 extern struct pernet_operations __net_initdata loopback_net_ops;
5205 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5207 /* netdev_printk helpers, similar to dev_printk */
5209 static inline const char *netdev_name(const struct net_device *dev)
5211 if (!dev->name[0] || strchr(dev->name, '%'))
5212 return "(unnamed net_device)";
5216 static inline bool netdev_unregistering(const struct net_device *dev)
5218 return dev->reg_state == NETREG_UNREGISTERING;
5221 static inline const char *netdev_reg_state(const struct net_device *dev)
5223 switch (dev->reg_state) {
5224 case NETREG_UNINITIALIZED: return " (uninitialized)";
5225 case NETREG_REGISTERED: return "";
5226 case NETREG_UNREGISTERING: return " (unregistering)";
5227 case NETREG_UNREGISTERED: return " (unregistered)";
5228 case NETREG_RELEASED: return " (released)";
5229 case NETREG_DUMMY: return " (dummy)";
5232 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5233 return " (unknown)";
5236 __printf(3, 4) __cold
5237 void netdev_printk(const char *level, const struct net_device *dev,
5238 const char *format, ...);
5239 __printf(2, 3) __cold
5240 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5241 __printf(2, 3) __cold
5242 void netdev_alert(const struct net_device *dev, const char *format, ...);
5243 __printf(2, 3) __cold
5244 void netdev_crit(const struct net_device *dev, const char *format, ...);
5245 __printf(2, 3) __cold
5246 void netdev_err(const struct net_device *dev, const char *format, ...);
5247 __printf(2, 3) __cold
5248 void netdev_warn(const struct net_device *dev, const char *format, ...);
5249 __printf(2, 3) __cold
5250 void netdev_notice(const struct net_device *dev, const char *format, ...);
5251 __printf(2, 3) __cold
5252 void netdev_info(const struct net_device *dev, const char *format, ...);
5254 #define netdev_level_once(level, dev, fmt, ...) \
5256 static bool __print_once __read_mostly; \
5258 if (!__print_once) { \
5259 __print_once = true; \
5260 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5264 #define netdev_emerg_once(dev, fmt, ...) \
5265 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5266 #define netdev_alert_once(dev, fmt, ...) \
5267 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5268 #define netdev_crit_once(dev, fmt, ...) \
5269 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5270 #define netdev_err_once(dev, fmt, ...) \
5271 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5272 #define netdev_warn_once(dev, fmt, ...) \
5273 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5274 #define netdev_notice_once(dev, fmt, ...) \
5275 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5276 #define netdev_info_once(dev, fmt, ...) \
5277 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5279 #define MODULE_ALIAS_NETDEV(device) \
5280 MODULE_ALIAS("netdev-" device)
5282 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5283 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5284 #define netdev_dbg(__dev, format, args...) \
5286 dynamic_netdev_dbg(__dev, format, ##args); \
5288 #elif defined(DEBUG)
5289 #define netdev_dbg(__dev, format, args...) \
5290 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5292 #define netdev_dbg(__dev, format, args...) \
5295 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5299 #if defined(VERBOSE_DEBUG)
5300 #define netdev_vdbg netdev_dbg
5303 #define netdev_vdbg(dev, format, args...) \
5306 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5312 * netdev_WARN() acts like dev_printk(), but with the key difference
5313 * of using a WARN/WARN_ON to get the message out, including the
5314 * file/line information and a backtrace.
5316 #define netdev_WARN(dev, format, args...) \
5317 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5318 netdev_reg_state(dev), ##args)
5320 #define netdev_WARN_ONCE(dev, format, args...) \
5321 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5322 netdev_reg_state(dev), ##args)
5324 /* netif printk helpers, similar to netdev_printk */
5326 #define netif_printk(priv, type, level, dev, fmt, args...) \
5328 if (netif_msg_##type(priv)) \
5329 netdev_printk(level, (dev), fmt, ##args); \
5332 #define netif_level(level, priv, type, dev, fmt, args...) \
5334 if (netif_msg_##type(priv)) \
5335 netdev_##level(dev, fmt, ##args); \
5338 #define netif_emerg(priv, type, dev, fmt, args...) \
5339 netif_level(emerg, priv, type, dev, fmt, ##args)
5340 #define netif_alert(priv, type, dev, fmt, args...) \
5341 netif_level(alert, priv, type, dev, fmt, ##args)
5342 #define netif_crit(priv, type, dev, fmt, args...) \
5343 netif_level(crit, priv, type, dev, fmt, ##args)
5344 #define netif_err(priv, type, dev, fmt, args...) \
5345 netif_level(err, priv, type, dev, fmt, ##args)
5346 #define netif_warn(priv, type, dev, fmt, args...) \
5347 netif_level(warn, priv, type, dev, fmt, ##args)
5348 #define netif_notice(priv, type, dev, fmt, args...) \
5349 netif_level(notice, priv, type, dev, fmt, ##args)
5350 #define netif_info(priv, type, dev, fmt, args...) \
5351 netif_level(info, priv, type, dev, fmt, ##args)
5353 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5354 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5355 #define netif_dbg(priv, type, netdev, format, args...) \
5357 if (netif_msg_##type(priv)) \
5358 dynamic_netdev_dbg(netdev, format, ##args); \
5360 #elif defined(DEBUG)
5361 #define netif_dbg(priv, type, dev, format, args...) \
5362 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5364 #define netif_dbg(priv, type, dev, format, args...) \
5367 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5372 /* if @cond then downgrade to debug, else print at @level */
5373 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5376 netif_dbg(priv, type, netdev, fmt, ##args); \
5378 netif_ ## level(priv, type, netdev, fmt, ##args); \
5381 #if defined(VERBOSE_DEBUG)
5382 #define netif_vdbg netif_dbg
5384 #define netif_vdbg(priv, type, dev, format, args...) \
5387 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5393 * The list of packet types we will receive (as opposed to discard)
5394 * and the routines to invoke.
5396 * Why 16. Because with 16 the only overlap we get on a hash of the
5397 * low nibble of the protocol value is RARP/SNAP/X.25.
5411 #define PTYPE_HASH_SIZE (16)
5412 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5414 extern struct list_head ptype_all __read_mostly;
5415 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5417 extern struct net_device *blackhole_netdev;
5419 #endif /* _LINUX_NETDEVICE_H */