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*/
366 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
367 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
368 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
369 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
370 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
371 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
372 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
373 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
374 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
384 typedef enum gro_result gro_result_t;
387 * enum rx_handler_result - Possible return values for rx_handlers.
388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
391 * case skb->dev was changed by rx_handler.
392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
395 * rx_handlers are functions called from inside __netif_receive_skb(), to do
396 * special processing of the skb, prior to delivery to protocol handlers.
398 * Currently, a net_device can only have a single rx_handler registered. Trying
399 * to register a second rx_handler will return -EBUSY.
401 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
402 * To unregister a rx_handler on a net_device, use
403 * netdev_rx_handler_unregister().
405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
408 * If the rx_handler consumed the skb in some way, it should return
409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
410 * the skb to be delivered in some other way.
412 * If the rx_handler changed skb->dev, to divert the skb to another
413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
414 * new device will be called if it exists.
416 * If the rx_handler decides the skb should be ignored, it should return
417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
418 * are registered on exact device (ptype->dev == skb->dev).
420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
421 * delivered, it should return RX_HANDLER_PASS.
423 * A device without a registered rx_handler will behave as if rx_handler
424 * returned RX_HANDLER_PASS.
427 enum rx_handler_result {
433 typedef enum rx_handler_result rx_handler_result_t;
434 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
436 void __napi_schedule(struct napi_struct *n);
437 void __napi_schedule_irqoff(struct napi_struct *n);
439 static inline bool napi_disable_pending(struct napi_struct *n)
441 return test_bit(NAPI_STATE_DISABLE, &n->state);
444 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
446 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
449 bool napi_schedule_prep(struct napi_struct *n);
452 * napi_schedule - schedule NAPI poll
455 * Schedule NAPI poll routine to be called if it is not already
458 static inline void napi_schedule(struct napi_struct *n)
460 if (napi_schedule_prep(n))
465 * napi_schedule_irqoff - schedule NAPI poll
468 * Variant of napi_schedule(), assuming hard irqs are masked.
470 static inline void napi_schedule_irqoff(struct napi_struct *n)
472 if (napi_schedule_prep(n))
473 __napi_schedule_irqoff(n);
476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
477 static inline bool napi_reschedule(struct napi_struct *napi)
479 if (napi_schedule_prep(napi)) {
480 __napi_schedule(napi);
486 bool napi_complete_done(struct napi_struct *n, int work_done);
488 * napi_complete - NAPI processing complete
491 * Mark NAPI processing as complete.
492 * Consider using napi_complete_done() instead.
493 * Return false if device should avoid rearming interrupts.
495 static inline bool napi_complete(struct napi_struct *n)
497 return napi_complete_done(n, 0);
500 int dev_set_threaded(struct net_device *dev, bool threaded);
503 * napi_disable - prevent NAPI from scheduling
506 * Stop NAPI from being scheduled on this context.
507 * Waits till any outstanding processing completes.
509 void napi_disable(struct napi_struct *n);
511 void napi_enable(struct napi_struct *n);
514 * napi_synchronize - wait until NAPI is not running
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
521 static inline void napi_synchronize(const struct napi_struct *n)
523 if (IS_ENABLED(CONFIG_SMP))
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 * napi_if_scheduled_mark_missed - if napi is running, set the
535 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
538 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
540 unsigned long val, new;
543 val = READ_ONCE(n->state);
544 if (val & NAPIF_STATE_DISABLE)
547 if (!(val & NAPIF_STATE_SCHED))
550 new = val | NAPIF_STATE_MISSED;
551 } while (cmpxchg(&n->state, val, new) != val);
556 enum netdev_queue_state_t {
557 __QUEUE_STATE_DRV_XOFF,
558 __QUEUE_STATE_STACK_XOFF,
559 __QUEUE_STATE_FROZEN,
562 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
563 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
564 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
566 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
567 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
569 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
573 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
574 * netif_tx_* functions below are used to manipulate this flag. The
575 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
576 * queue independently. The netif_xmit_*stopped functions below are called
577 * to check if the queue has been stopped by the driver or stack (either
578 * of the XOFF bits are set in the state). Drivers should not need to call
579 * netif_xmit*stopped functions, they should only be using netif_tx_*.
582 struct netdev_queue {
586 struct net_device *dev;
587 struct Qdisc __rcu *qdisc;
588 struct Qdisc *qdisc_sleeping;
592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
595 unsigned long tx_maxrate;
597 * Number of TX timeouts for this queue
598 * (/sys/class/net/DEV/Q/trans_timeout)
600 unsigned long trans_timeout;
602 /* Subordinate device that the queue has been assigned to */
603 struct net_device *sb_dev;
604 #ifdef CONFIG_XDP_SOCKETS
605 struct xsk_buff_pool *pool;
610 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
613 * Time (in jiffies) of last Tx
615 unsigned long trans_start;
622 } ____cacheline_aligned_in_smp;
624 extern int sysctl_fb_tunnels_only_for_init_net;
625 extern int sysctl_devconf_inherit_init_net;
628 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
629 * == 1 : For initns only
632 static inline bool net_has_fallback_tunnels(const struct net *net)
634 return !IS_ENABLED(CONFIG_SYSCTL) ||
635 !sysctl_fb_tunnels_only_for_init_net ||
636 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
639 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
641 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
648 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 * This structure holds an RPS map which can be of variable length. The
658 * map is an array of CPUs.
665 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
669 * tail pointer for that CPU's input queue at the time of last enqueue, and
670 * a hardware filter index.
672 struct rps_dev_flow {
675 unsigned int last_qtail;
677 #define RPS_NO_FILTER 0xffff
680 * The rps_dev_flow_table structure contains a table of flow mappings.
682 struct rps_dev_flow_table {
685 struct rps_dev_flow flows[];
687 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
688 ((_num) * sizeof(struct rps_dev_flow)))
691 * The rps_sock_flow_table contains mappings of flows to the last CPU
692 * on which they were processed by the application (set in recvmsg).
693 * Each entry is a 32bit value. Upper part is the high-order bits
694 * of flow hash, lower part is CPU number.
695 * rps_cpu_mask is used to partition the space, depending on number of
696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
698 * meaning we use 32-6=26 bits for the hash.
700 struct rps_sock_flow_table {
703 u32 ents[] ____cacheline_aligned_in_smp;
705 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
707 #define RPS_NO_CPU 0xffff
709 extern u32 rps_cpu_mask;
710 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
712 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
716 unsigned int index = hash & table->mask;
717 u32 val = hash & ~rps_cpu_mask;
719 /* We only give a hint, preemption can change CPU under us */
720 val |= raw_smp_processor_id();
722 if (table->ents[index] != val)
723 table->ents[index] = val;
727 #ifdef CONFIG_RFS_ACCEL
728 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 #endif /* CONFIG_RPS */
733 /* This structure contains an instance of an RX queue. */
734 struct netdev_rx_queue {
736 struct rps_map __rcu *rps_map;
737 struct rps_dev_flow_table __rcu *rps_flow_table;
740 struct net_device *dev;
741 struct xdp_rxq_info xdp_rxq;
742 #ifdef CONFIG_XDP_SOCKETS
743 struct xsk_buff_pool *pool;
745 } ____cacheline_aligned_in_smp;
748 * RX queue sysfs structures and functions.
750 struct rx_queue_attribute {
751 struct attribute attr;
752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
753 ssize_t (*store)(struct netdev_rx_queue *queue,
754 const char *buf, size_t len);
757 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
766 * This structure holds an XPS map which can be of variable length. The
767 * map is an array of queues.
771 unsigned int alloc_len;
775 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
776 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
777 - sizeof(struct xps_map)) / sizeof(u16))
780 * This structure holds all XPS maps for device. Maps are indexed by CPU.
782 * We keep track of the number of cpus/rxqs used when the struct is allocated,
783 * in nr_ids. This will help not accessing out-of-bound memory.
785 * We keep track of the number of traffic classes used when the struct is
786 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
787 * not crossing its upper bound, as the original dev->num_tc can be updated in
790 struct xps_dev_maps {
794 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
797 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
798 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
800 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
801 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
803 #endif /* CONFIG_XPS */
805 #define TC_MAX_QUEUE 16
806 #define TC_BITMASK 15
807 /* HW offloaded queuing disciplines txq count and offset maps */
808 struct netdev_tc_txq {
813 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
815 * This structure is to hold information about the device
816 * configured to run FCoE protocol stack.
818 struct netdev_fcoe_hbainfo {
819 char manufacturer[64];
820 char serial_number[64];
821 char hardware_version[64];
822 char driver_version[64];
823 char optionrom_version[64];
824 char firmware_version[64];
826 char model_description[256];
830 #define MAX_PHYS_ITEM_ID_LEN 32
832 /* This structure holds a unique identifier to identify some
833 * physical item (port for example) used by a netdevice.
835 struct netdev_phys_item_id {
836 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
837 unsigned char id_len;
840 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
841 struct netdev_phys_item_id *b)
843 return a->id_len == b->id_len &&
844 memcmp(a->id, b->id, a->id_len) == 0;
847 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
849 struct net_device *sb_dev);
851 enum net_device_path_type {
852 DEV_PATH_ETHERNET = 0,
859 struct net_device_path {
860 enum net_device_path_type type;
861 const struct net_device *dev;
870 DEV_PATH_BR_VLAN_KEEP,
871 DEV_PATH_BR_VLAN_TAG,
872 DEV_PATH_BR_VLAN_UNTAG,
884 #define NET_DEVICE_PATH_STACK_MAX 5
885 #define NET_DEVICE_PATH_VLAN_MAX 2
887 struct net_device_path_stack {
889 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
892 struct net_device_path_ctx {
893 const struct net_device *dev;
900 } vlan[NET_DEVICE_PATH_VLAN_MAX];
904 TC_SETUP_QDISC_MQPRIO,
907 TC_SETUP_CLSMATCHALL,
917 TC_SETUP_QDISC_TAPRIO,
925 /* These structures hold the attributes of bpf state that are being passed
926 * to the netdevice through the bpf op.
928 enum bpf_netdev_command {
929 /* Set or clear a bpf program used in the earliest stages of packet
930 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
931 * is responsible for calling bpf_prog_put on any old progs that are
932 * stored. In case of error, the callee need not release the new prog
933 * reference, but on success it takes ownership and must bpf_prog_put
934 * when it is no longer used.
938 /* BPF program for offload callbacks, invoked at program load time. */
939 BPF_OFFLOAD_MAP_ALLOC,
940 BPF_OFFLOAD_MAP_FREE,
944 struct bpf_prog_offload_ops;
945 struct netlink_ext_ack;
947 struct xdp_dev_bulk_queue;
957 struct bpf_xdp_entity {
958 struct bpf_prog *prog;
959 struct bpf_xdp_link *link;
963 enum bpf_netdev_command command;
968 struct bpf_prog *prog;
969 struct netlink_ext_ack *extack;
971 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
973 struct bpf_offloaded_map *offmap;
975 /* XDP_SETUP_XSK_POOL */
977 struct xsk_buff_pool *pool;
983 /* Flags for ndo_xsk_wakeup. */
984 #define XDP_WAKEUP_RX (1 << 0)
985 #define XDP_WAKEUP_TX (1 << 1)
987 #ifdef CONFIG_XFRM_OFFLOAD
989 int (*xdo_dev_state_add) (struct xfrm_state *x);
990 void (*xdo_dev_state_delete) (struct xfrm_state *x);
991 void (*xdo_dev_state_free) (struct xfrm_state *x);
992 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
993 struct xfrm_state *x);
994 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
999 struct rcu_head rcuhead;
1006 struct netdev_name_node {
1007 struct hlist_node hlist;
1008 struct list_head list;
1009 struct net_device *dev;
1013 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1014 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1016 struct netdev_net_notifier {
1017 struct list_head list;
1018 struct notifier_block *nb;
1022 * This structure defines the management hooks for network devices.
1023 * The following hooks can be defined; unless noted otherwise, they are
1024 * optional and can be filled with a null pointer.
1026 * int (*ndo_init)(struct net_device *dev);
1027 * This function is called once when a network device is registered.
1028 * The network device can use this for any late stage initialization
1029 * or semantic validation. It can fail with an error code which will
1030 * be propagated back to register_netdev.
1032 * void (*ndo_uninit)(struct net_device *dev);
1033 * This function is called when device is unregistered or when registration
1034 * fails. It is not called if init fails.
1036 * int (*ndo_open)(struct net_device *dev);
1037 * This function is called when a network device transitions to the up
1040 * int (*ndo_stop)(struct net_device *dev);
1041 * This function is called when a network device transitions to the down
1044 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1045 * struct net_device *dev);
1046 * Called when a packet needs to be transmitted.
1047 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1048 * the queue before that can happen; it's for obsolete devices and weird
1049 * corner cases, but the stack really does a non-trivial amount
1050 * of useless work if you return NETDEV_TX_BUSY.
1051 * Required; cannot be NULL.
1053 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1054 * struct net_device *dev
1055 * netdev_features_t features);
1056 * Called by core transmit path to determine if device is capable of
1057 * performing offload operations on a given packet. This is to give
1058 * the device an opportunity to implement any restrictions that cannot
1059 * be otherwise expressed by feature flags. The check is called with
1060 * the set of features that the stack has calculated and it returns
1061 * those the driver believes to be appropriate.
1063 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1064 * struct net_device *sb_dev);
1065 * Called to decide which queue to use when device supports multiple
1068 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1069 * This function is called to allow device receiver to make
1070 * changes to configuration when multicast or promiscuous is enabled.
1072 * void (*ndo_set_rx_mode)(struct net_device *dev);
1073 * This function is called device changes address list filtering.
1074 * If driver handles unicast address filtering, it should set
1075 * IFF_UNICAST_FLT in its priv_flags.
1077 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1078 * This function is called when the Media Access Control address
1079 * needs to be changed. If this interface is not defined, the
1080 * MAC address can not be changed.
1082 * int (*ndo_validate_addr)(struct net_device *dev);
1083 * Test if Media Access Control address is valid for the device.
1085 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1086 * Called when a user requests an ioctl which can't be handled by
1087 * the generic interface code. If not defined ioctls return
1088 * not supported error code.
1090 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1091 * Used to set network devices bus interface parameters. This interface
1092 * is retained for legacy reasons; new devices should use the bus
1093 * interface (PCI) for low level management.
1095 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1096 * Called when a user wants to change the Maximum Transfer Unit
1099 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1100 * Callback used when the transmitter has not made any progress
1101 * for dev->watchdog ticks.
1103 * void (*ndo_get_stats64)(struct net_device *dev,
1104 * struct rtnl_link_stats64 *storage);
1105 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1106 * Called when a user wants to get the network device usage
1107 * statistics. Drivers must do one of the following:
1108 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1109 * rtnl_link_stats64 structure passed by the caller.
1110 * 2. Define @ndo_get_stats to update a net_device_stats structure
1111 * (which should normally be dev->stats) and return a pointer to
1112 * it. The structure may be changed asynchronously only if each
1113 * field is written atomically.
1114 * 3. Update dev->stats asynchronously and atomically, and define
1115 * neither operation.
1117 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1118 * Return true if this device supports offload stats of this attr_id.
1120 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1122 * Get statistics for offload operations by attr_id. Write it into the
1123 * attr_data pointer.
1125 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1126 * If device supports VLAN filtering this function is called when a
1127 * VLAN id is registered.
1129 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1130 * If device supports VLAN filtering this function is called when a
1131 * VLAN id is unregistered.
1133 * void (*ndo_poll_controller)(struct net_device *dev);
1135 * SR-IOV management functions.
1136 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1137 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1138 * u8 qos, __be16 proto);
1139 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1141 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1142 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1143 * int (*ndo_get_vf_config)(struct net_device *dev,
1144 * int vf, struct ifla_vf_info *ivf);
1145 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1146 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1147 * struct nlattr *port[]);
1149 * Enable or disable the VF ability to query its RSS Redirection Table and
1150 * Hash Key. This is needed since on some devices VF share this information
1151 * with PF and querying it may introduce a theoretical security risk.
1152 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1153 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1154 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1156 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1157 * This is always called from the stack with the rtnl lock held and netif
1158 * tx queues stopped. This allows the netdevice to perform queue
1159 * management safely.
1161 * Fiber Channel over Ethernet (FCoE) offload functions.
1162 * int (*ndo_fcoe_enable)(struct net_device *dev);
1163 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1164 * so the underlying device can perform whatever needed configuration or
1165 * initialization to support acceleration of FCoE traffic.
1167 * int (*ndo_fcoe_disable)(struct net_device *dev);
1168 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1169 * so the underlying device can perform whatever needed clean-ups to
1170 * stop supporting acceleration of FCoE traffic.
1172 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1173 * struct scatterlist *sgl, unsigned int sgc);
1174 * Called when the FCoE Initiator wants to initialize an I/O that
1175 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1176 * perform necessary setup and returns 1 to indicate the device is set up
1177 * successfully to perform DDP on this I/O, otherwise this returns 0.
1179 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1180 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1181 * indicated by the FC exchange id 'xid', so the underlying device can
1182 * clean up and reuse resources for later DDP requests.
1184 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1185 * struct scatterlist *sgl, unsigned int sgc);
1186 * Called when the FCoE Target wants to initialize an I/O that
1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1188 * perform necessary setup and returns 1 to indicate the device is set up
1189 * successfully to perform DDP on this I/O, otherwise this returns 0.
1191 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1192 * struct netdev_fcoe_hbainfo *hbainfo);
1193 * Called when the FCoE Protocol stack wants information on the underlying
1194 * device. This information is utilized by the FCoE protocol stack to
1195 * register attributes with Fiber Channel management service as per the
1196 * FC-GS Fabric Device Management Information(FDMI) specification.
1198 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1199 * Called when the underlying device wants to override default World Wide
1200 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1201 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1202 * protocol stack to use.
1205 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1206 * u16 rxq_index, u32 flow_id);
1207 * Set hardware filter for RFS. rxq_index is the target queue index;
1208 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1209 * Return the filter ID on success, or a negative error code.
1211 * Slave management functions (for bridge, bonding, etc).
1212 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1213 * Called to make another netdev an underling.
1215 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1216 * Called to release previously enslaved netdev.
1218 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1219 * struct sk_buff *skb,
1221 * Get the xmit slave of master device. If all_slaves is true, function
1222 * assume all the slaves can transmit.
1224 * Feature/offload setting functions.
1225 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1226 * netdev_features_t features);
1227 * Adjusts the requested feature flags according to device-specific
1228 * constraints, and returns the resulting flags. Must not modify
1231 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1232 * Called to update device configuration to new features. Passed
1233 * feature set might be less than what was returned by ndo_fix_features()).
1234 * Must return >0 or -errno if it changed dev->features itself.
1236 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1237 * struct net_device *dev,
1238 * const unsigned char *addr, u16 vid, u16 flags,
1239 * struct netlink_ext_ack *extack);
1240 * Adds an FDB entry to dev for addr.
1241 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1242 * struct net_device *dev,
1243 * const unsigned char *addr, u16 vid)
1244 * Deletes the FDB entry from dev coresponding to addr.
1245 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1246 * struct net_device *dev, struct net_device *filter_dev,
1248 * Used to add FDB entries to dump requests. Implementers should add
1249 * entries to skb and update idx with the number of entries.
1251 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1252 * u16 flags, struct netlink_ext_ack *extack)
1253 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1254 * struct net_device *dev, u32 filter_mask,
1256 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1259 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1260 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1261 * which do not represent real hardware may define this to allow their
1262 * userspace components to manage their virtual carrier state. Devices
1263 * that determine carrier state from physical hardware properties (eg
1264 * network cables) or protocol-dependent mechanisms (eg
1265 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1267 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1268 * struct netdev_phys_item_id *ppid);
1269 * Called to get ID of physical port of this device. If driver does
1270 * not implement this, it is assumed that the hw is not able to have
1271 * multiple net devices on single physical port.
1273 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1274 * struct netdev_phys_item_id *ppid)
1275 * Called to get the parent ID of the physical port of this device.
1277 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1278 * struct net_device *dev)
1279 * Called by upper layer devices to accelerate switching or other
1280 * station functionality into hardware. 'pdev is the lowerdev
1281 * to use for the offload and 'dev' is the net device that will
1282 * back the offload. Returns a pointer to the private structure
1283 * the upper layer will maintain.
1284 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1285 * Called by upper layer device to delete the station created
1286 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1287 * the station and priv is the structure returned by the add
1289 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1290 * int queue_index, u32 maxrate);
1291 * Called when a user wants to set a max-rate limitation of specific
1293 * int (*ndo_get_iflink)(const struct net_device *dev);
1294 * Called to get the iflink value of this device.
1295 * void (*ndo_change_proto_down)(struct net_device *dev,
1297 * This function is used to pass protocol port error state information
1298 * to the switch driver. The switch driver can react to the proto_down
1299 * by doing a phys down on the associated switch port.
1300 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1301 * This function is used to get egress tunnel information for given skb.
1302 * This is useful for retrieving outer tunnel header parameters while
1304 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1305 * This function is used to specify the headroom that the skb must
1306 * consider when allocation skb during packet reception. Setting
1307 * appropriate rx headroom value allows avoiding skb head copy on
1308 * forward. Setting a negative value resets the rx headroom to the
1310 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1311 * This function is used to set or query state related to XDP on the
1312 * netdevice and manage BPF offload. See definition of
1313 * enum bpf_netdev_command for details.
1314 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1316 * This function is used to submit @n XDP packets for transmit on a
1317 * netdevice. Returns number of frames successfully transmitted, frames
1318 * that got dropped are freed/returned via xdp_return_frame().
1319 * Returns negative number, means general error invoking ndo, meaning
1320 * no frames were xmit'ed and core-caller will free all frames.
1321 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1322 * This function is used to wake up the softirq, ksoftirqd or kthread
1323 * responsible for sending and/or receiving packets on a specific
1324 * queue id bound to an AF_XDP socket. The flags field specifies if
1325 * only RX, only Tx, or both should be woken up using the flags
1326 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1327 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1328 * Get devlink port instance associated with a given netdev.
1329 * Called with a reference on the netdevice and devlink locks only,
1330 * rtnl_lock is not held.
1331 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1333 * Add, change, delete or get information on an IPv4 tunnel.
1334 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1335 * If a device is paired with a peer device, return the peer instance.
1336 * The caller must be under RCU read context.
1337 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1338 * Get the forwarding path to reach the real device from the HW destination address
1340 struct net_device_ops {
1341 int (*ndo_init)(struct net_device *dev);
1342 void (*ndo_uninit)(struct net_device *dev);
1343 int (*ndo_open)(struct net_device *dev);
1344 int (*ndo_stop)(struct net_device *dev);
1345 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1346 struct net_device *dev);
1347 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1348 struct net_device *dev,
1349 netdev_features_t features);
1350 u16 (*ndo_select_queue)(struct net_device *dev,
1351 struct sk_buff *skb,
1352 struct net_device *sb_dev);
1353 void (*ndo_change_rx_flags)(struct net_device *dev,
1355 void (*ndo_set_rx_mode)(struct net_device *dev);
1356 int (*ndo_set_mac_address)(struct net_device *dev,
1358 int (*ndo_validate_addr)(struct net_device *dev);
1359 int (*ndo_do_ioctl)(struct net_device *dev,
1360 struct ifreq *ifr, int cmd);
1361 int (*ndo_set_config)(struct net_device *dev,
1363 int (*ndo_change_mtu)(struct net_device *dev,
1365 int (*ndo_neigh_setup)(struct net_device *dev,
1366 struct neigh_parms *);
1367 void (*ndo_tx_timeout) (struct net_device *dev,
1368 unsigned int txqueue);
1370 void (*ndo_get_stats64)(struct net_device *dev,
1371 struct rtnl_link_stats64 *storage);
1372 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1373 int (*ndo_get_offload_stats)(int attr_id,
1374 const struct net_device *dev,
1376 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1378 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1379 __be16 proto, u16 vid);
1380 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1381 __be16 proto, u16 vid);
1382 #ifdef CONFIG_NET_POLL_CONTROLLER
1383 void (*ndo_poll_controller)(struct net_device *dev);
1384 int (*ndo_netpoll_setup)(struct net_device *dev,
1385 struct netpoll_info *info);
1386 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1388 int (*ndo_set_vf_mac)(struct net_device *dev,
1389 int queue, u8 *mac);
1390 int (*ndo_set_vf_vlan)(struct net_device *dev,
1391 int queue, u16 vlan,
1392 u8 qos, __be16 proto);
1393 int (*ndo_set_vf_rate)(struct net_device *dev,
1394 int vf, int min_tx_rate,
1396 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1397 int vf, bool setting);
1398 int (*ndo_set_vf_trust)(struct net_device *dev,
1399 int vf, bool setting);
1400 int (*ndo_get_vf_config)(struct net_device *dev,
1402 struct ifla_vf_info *ivf);
1403 int (*ndo_set_vf_link_state)(struct net_device *dev,
1404 int vf, int link_state);
1405 int (*ndo_get_vf_stats)(struct net_device *dev,
1407 struct ifla_vf_stats
1409 int (*ndo_set_vf_port)(struct net_device *dev,
1411 struct nlattr *port[]);
1412 int (*ndo_get_vf_port)(struct net_device *dev,
1413 int vf, struct sk_buff *skb);
1414 int (*ndo_get_vf_guid)(struct net_device *dev,
1416 struct ifla_vf_guid *node_guid,
1417 struct ifla_vf_guid *port_guid);
1418 int (*ndo_set_vf_guid)(struct net_device *dev,
1421 int (*ndo_set_vf_rss_query_en)(
1422 struct net_device *dev,
1423 int vf, bool setting);
1424 int (*ndo_setup_tc)(struct net_device *dev,
1425 enum tc_setup_type type,
1427 #if IS_ENABLED(CONFIG_FCOE)
1428 int (*ndo_fcoe_enable)(struct net_device *dev);
1429 int (*ndo_fcoe_disable)(struct net_device *dev);
1430 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1432 struct scatterlist *sgl,
1434 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1436 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1438 struct scatterlist *sgl,
1440 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1441 struct netdev_fcoe_hbainfo *hbainfo);
1444 #if IS_ENABLED(CONFIG_LIBFCOE)
1445 #define NETDEV_FCOE_WWNN 0
1446 #define NETDEV_FCOE_WWPN 1
1447 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1448 u64 *wwn, int type);
1451 #ifdef CONFIG_RFS_ACCEL
1452 int (*ndo_rx_flow_steer)(struct net_device *dev,
1453 const struct sk_buff *skb,
1457 int (*ndo_add_slave)(struct net_device *dev,
1458 struct net_device *slave_dev,
1459 struct netlink_ext_ack *extack);
1460 int (*ndo_del_slave)(struct net_device *dev,
1461 struct net_device *slave_dev);
1462 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1463 struct sk_buff *skb,
1465 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1467 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1468 netdev_features_t features);
1469 int (*ndo_set_features)(struct net_device *dev,
1470 netdev_features_t features);
1471 int (*ndo_neigh_construct)(struct net_device *dev,
1472 struct neighbour *n);
1473 void (*ndo_neigh_destroy)(struct net_device *dev,
1474 struct neighbour *n);
1476 int (*ndo_fdb_add)(struct ndmsg *ndm,
1477 struct nlattr *tb[],
1478 struct net_device *dev,
1479 const unsigned char *addr,
1482 struct netlink_ext_ack *extack);
1483 int (*ndo_fdb_del)(struct ndmsg *ndm,
1484 struct nlattr *tb[],
1485 struct net_device *dev,
1486 const unsigned char *addr,
1488 int (*ndo_fdb_dump)(struct sk_buff *skb,
1489 struct netlink_callback *cb,
1490 struct net_device *dev,
1491 struct net_device *filter_dev,
1493 int (*ndo_fdb_get)(struct sk_buff *skb,
1494 struct nlattr *tb[],
1495 struct net_device *dev,
1496 const unsigned char *addr,
1497 u16 vid, u32 portid, u32 seq,
1498 struct netlink_ext_ack *extack);
1499 int (*ndo_bridge_setlink)(struct net_device *dev,
1500 struct nlmsghdr *nlh,
1502 struct netlink_ext_ack *extack);
1503 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1505 struct net_device *dev,
1508 int (*ndo_bridge_dellink)(struct net_device *dev,
1509 struct nlmsghdr *nlh,
1511 int (*ndo_change_carrier)(struct net_device *dev,
1513 int (*ndo_get_phys_port_id)(struct net_device *dev,
1514 struct netdev_phys_item_id *ppid);
1515 int (*ndo_get_port_parent_id)(struct net_device *dev,
1516 struct netdev_phys_item_id *ppid);
1517 int (*ndo_get_phys_port_name)(struct net_device *dev,
1518 char *name, size_t len);
1519 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1520 struct net_device *dev);
1521 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1524 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1527 int (*ndo_get_iflink)(const struct net_device *dev);
1528 int (*ndo_change_proto_down)(struct net_device *dev,
1530 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1531 struct sk_buff *skb);
1532 void (*ndo_set_rx_headroom)(struct net_device *dev,
1533 int needed_headroom);
1534 int (*ndo_bpf)(struct net_device *dev,
1535 struct netdev_bpf *bpf);
1536 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1537 struct xdp_frame **xdp,
1539 int (*ndo_xsk_wakeup)(struct net_device *dev,
1540 u32 queue_id, u32 flags);
1541 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1542 int (*ndo_tunnel_ctl)(struct net_device *dev,
1543 struct ip_tunnel_parm *p, int cmd);
1544 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1545 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1546 struct net_device_path *path);
1550 * enum netdev_priv_flags - &struct net_device priv_flags
1552 * These are the &struct net_device, they are only set internally
1553 * by drivers and used in the kernel. These flags are invisible to
1554 * userspace; this means that the order of these flags can change
1555 * during any kernel release.
1557 * You should have a pretty good reason to be extending these flags.
1559 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1560 * @IFF_EBRIDGE: Ethernet bridging device
1561 * @IFF_BONDING: bonding master or slave
1562 * @IFF_ISATAP: ISATAP interface (RFC4214)
1563 * @IFF_WAN_HDLC: WAN HDLC device
1564 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1566 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1567 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1568 * @IFF_MACVLAN_PORT: device used as macvlan port
1569 * @IFF_BRIDGE_PORT: device used as bridge port
1570 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1571 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1572 * @IFF_UNICAST_FLT: Supports unicast filtering
1573 * @IFF_TEAM_PORT: device used as team port
1574 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1575 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1576 * change when it's running
1577 * @IFF_MACVLAN: Macvlan device
1578 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1579 * underlying stacked devices
1580 * @IFF_L3MDEV_MASTER: device is an L3 master device
1581 * @IFF_NO_QUEUE: device can run without qdisc attached
1582 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1583 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1584 * @IFF_TEAM: device is a team device
1585 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1586 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1587 * entity (i.e. the master device for bridged veth)
1588 * @IFF_MACSEC: device is a MACsec device
1589 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1590 * @IFF_FAILOVER: device is a failover master device
1591 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1592 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1593 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1594 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1595 * skb_headlen(skb) == 0 (data starts from frag0)
1597 enum netdev_priv_flags {
1598 IFF_802_1Q_VLAN = 1<<0,
1602 IFF_WAN_HDLC = 1<<4,
1603 IFF_XMIT_DST_RELEASE = 1<<5,
1604 IFF_DONT_BRIDGE = 1<<6,
1605 IFF_DISABLE_NETPOLL = 1<<7,
1606 IFF_MACVLAN_PORT = 1<<8,
1607 IFF_BRIDGE_PORT = 1<<9,
1608 IFF_OVS_DATAPATH = 1<<10,
1609 IFF_TX_SKB_SHARING = 1<<11,
1610 IFF_UNICAST_FLT = 1<<12,
1611 IFF_TEAM_PORT = 1<<13,
1612 IFF_SUPP_NOFCS = 1<<14,
1613 IFF_LIVE_ADDR_CHANGE = 1<<15,
1614 IFF_MACVLAN = 1<<16,
1615 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1616 IFF_L3MDEV_MASTER = 1<<18,
1617 IFF_NO_QUEUE = 1<<19,
1618 IFF_OPENVSWITCH = 1<<20,
1619 IFF_L3MDEV_SLAVE = 1<<21,
1621 IFF_RXFH_CONFIGURED = 1<<23,
1622 IFF_PHONY_HEADROOM = 1<<24,
1624 IFF_NO_RX_HANDLER = 1<<26,
1625 IFF_FAILOVER = 1<<27,
1626 IFF_FAILOVER_SLAVE = 1<<28,
1627 IFF_L3MDEV_RX_HANDLER = 1<<29,
1628 IFF_LIVE_RENAME_OK = 1<<30,
1629 IFF_TX_SKB_NO_LINEAR = 1<<31,
1632 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1633 #define IFF_EBRIDGE IFF_EBRIDGE
1634 #define IFF_BONDING IFF_BONDING
1635 #define IFF_ISATAP IFF_ISATAP
1636 #define IFF_WAN_HDLC IFF_WAN_HDLC
1637 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1638 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1639 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1640 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1641 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1642 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1643 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1644 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1645 #define IFF_TEAM_PORT IFF_TEAM_PORT
1646 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1647 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1648 #define IFF_MACVLAN IFF_MACVLAN
1649 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1650 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1651 #define IFF_NO_QUEUE IFF_NO_QUEUE
1652 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1653 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1654 #define IFF_TEAM IFF_TEAM
1655 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1656 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1657 #define IFF_MACSEC IFF_MACSEC
1658 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1659 #define IFF_FAILOVER IFF_FAILOVER
1660 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1661 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1662 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1663 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1665 /* Specifies the type of the struct net_device::ml_priv pointer */
1666 enum netdev_ml_priv_type {
1672 * struct net_device - The DEVICE structure.
1674 * Actually, this whole structure is a big mistake. It mixes I/O
1675 * data with strictly "high-level" data, and it has to know about
1676 * almost every data structure used in the INET module.
1678 * @name: This is the first field of the "visible" part of this structure
1679 * (i.e. as seen by users in the "Space.c" file). It is the name
1682 * @name_node: Name hashlist node
1683 * @ifalias: SNMP alias
1684 * @mem_end: Shared memory end
1685 * @mem_start: Shared memory start
1686 * @base_addr: Device I/O address
1687 * @irq: Device IRQ number
1689 * @state: Generic network queuing layer state, see netdev_state_t
1690 * @dev_list: The global list of network devices
1691 * @napi_list: List entry used for polling NAPI devices
1692 * @unreg_list: List entry when we are unregistering the
1693 * device; see the function unregister_netdev
1694 * @close_list: List entry used when we are closing the device
1695 * @ptype_all: Device-specific packet handlers for all protocols
1696 * @ptype_specific: Device-specific, protocol-specific packet handlers
1698 * @adj_list: Directly linked devices, like slaves for bonding
1699 * @features: Currently active device features
1700 * @hw_features: User-changeable features
1702 * @wanted_features: User-requested features
1703 * @vlan_features: Mask of features inheritable by VLAN devices
1705 * @hw_enc_features: Mask of features inherited by encapsulating devices
1706 * This field indicates what encapsulation
1707 * offloads the hardware is capable of doing,
1708 * and drivers will need to set them appropriately.
1710 * @mpls_features: Mask of features inheritable by MPLS
1711 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1713 * @ifindex: interface index
1714 * @group: The group the device belongs to
1716 * @stats: Statistics struct, which was left as a legacy, use
1717 * rtnl_link_stats64 instead
1719 * @rx_dropped: Dropped packets by core network,
1720 * do not use this in drivers
1721 * @tx_dropped: Dropped packets by core network,
1722 * do not use this in drivers
1723 * @rx_nohandler: nohandler dropped packets by core network on
1724 * inactive devices, do not use this in drivers
1725 * @carrier_up_count: Number of times the carrier has been up
1726 * @carrier_down_count: Number of times the carrier has been down
1728 * @wireless_handlers: List of functions to handle Wireless Extensions,
1730 * see <net/iw_handler.h> for details.
1731 * @wireless_data: Instance data managed by the core of wireless extensions
1733 * @netdev_ops: Includes several pointers to callbacks,
1734 * if one wants to override the ndo_*() functions
1735 * @ethtool_ops: Management operations
1736 * @l3mdev_ops: Layer 3 master device operations
1737 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1738 * discovery handling. Necessary for e.g. 6LoWPAN.
1739 * @xfrmdev_ops: Transformation offload operations
1740 * @tlsdev_ops: Transport Layer Security offload operations
1741 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1742 * of Layer 2 headers.
1744 * @flags: Interface flags (a la BSD)
1745 * @priv_flags: Like 'flags' but invisible to userspace,
1746 * see if.h for the definitions
1747 * @gflags: Global flags ( kept as legacy )
1748 * @padded: How much padding added by alloc_netdev()
1749 * @operstate: RFC2863 operstate
1750 * @link_mode: Mapping policy to operstate
1751 * @if_port: Selectable AUI, TP, ...
1753 * @mtu: Interface MTU value
1754 * @min_mtu: Interface Minimum MTU value
1755 * @max_mtu: Interface Maximum MTU value
1756 * @type: Interface hardware type
1757 * @hard_header_len: Maximum hardware header length.
1758 * @min_header_len: Minimum hardware header length
1760 * @needed_headroom: Extra headroom the hardware may need, but not in all
1761 * cases can this be guaranteed
1762 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1763 * cases can this be guaranteed. Some cases also use
1764 * LL_MAX_HEADER instead to allocate the skb
1766 * interface address info:
1768 * @perm_addr: Permanent hw address
1769 * @addr_assign_type: Hw address assignment type
1770 * @addr_len: Hardware address length
1771 * @upper_level: Maximum depth level of upper devices.
1772 * @lower_level: Maximum depth level of lower devices.
1773 * @neigh_priv_len: Used in neigh_alloc()
1774 * @dev_id: Used to differentiate devices that share
1775 * the same link layer address
1776 * @dev_port: Used to differentiate devices that share
1778 * @addr_list_lock: XXX: need comments on this one
1779 * @name_assign_type: network interface name assignment type
1780 * @uc_promisc: Counter that indicates promiscuous mode
1781 * has been enabled due to the need to listen to
1782 * additional unicast addresses in a device that
1783 * does not implement ndo_set_rx_mode()
1784 * @uc: unicast mac addresses
1785 * @mc: multicast mac addresses
1786 * @dev_addrs: list of device hw addresses
1787 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1788 * @promiscuity: Number of times the NIC is told to work in
1789 * promiscuous mode; if it becomes 0 the NIC will
1790 * exit promiscuous mode
1791 * @allmulti: Counter, enables or disables allmulticast mode
1793 * @vlan_info: VLAN info
1794 * @dsa_ptr: dsa specific data
1795 * @tipc_ptr: TIPC specific data
1796 * @atalk_ptr: AppleTalk link
1797 * @ip_ptr: IPv4 specific data
1798 * @dn_ptr: DECnet specific data
1799 * @ip6_ptr: IPv6 specific data
1800 * @ax25_ptr: AX.25 specific data
1801 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1802 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1804 * @mpls_ptr: mpls_dev struct pointer
1806 * @dev_addr: Hw address (before bcast,
1807 * because most packets are unicast)
1809 * @_rx: Array of RX queues
1810 * @num_rx_queues: Number of RX queues
1811 * allocated at register_netdev() time
1812 * @real_num_rx_queues: Number of RX queues currently active in device
1813 * @xdp_prog: XDP sockets filter program pointer
1814 * @gro_flush_timeout: timeout for GRO layer in NAPI
1815 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1816 * allow to avoid NIC hard IRQ, on busy queues.
1818 * @rx_handler: handler for received packets
1819 * @rx_handler_data: XXX: need comments on this one
1820 * @miniq_ingress: ingress/clsact qdisc specific data for
1821 * ingress processing
1822 * @ingress_queue: XXX: need comments on this one
1823 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1824 * @broadcast: hw bcast address
1826 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1827 * indexed by RX queue number. Assigned by driver.
1828 * This must only be set if the ndo_rx_flow_steer
1829 * operation is defined
1830 * @index_hlist: Device index hash chain
1832 * @_tx: Array of TX queues
1833 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1834 * @real_num_tx_queues: Number of TX queues currently active in device
1835 * @qdisc: Root qdisc from userspace point of view
1836 * @tx_queue_len: Max frames per queue allowed
1837 * @tx_global_lock: XXX: need comments on this one
1838 * @xdp_bulkq: XDP device bulk queue
1839 * @xps_maps: all CPUs/RXQs maps for XPS device
1841 * @xps_maps: XXX: need comments on this one
1842 * @miniq_egress: clsact qdisc specific data for
1844 * @qdisc_hash: qdisc hash table
1845 * @watchdog_timeo: Represents the timeout that is used by
1846 * the watchdog (see dev_watchdog())
1847 * @watchdog_timer: List of timers
1849 * @proto_down_reason: reason a netdev interface is held down
1850 * @pcpu_refcnt: Number of references to this device
1851 * @dev_refcnt: Number of references to this device
1852 * @todo_list: Delayed register/unregister
1853 * @link_watch_list: XXX: need comments on this one
1855 * @reg_state: Register/unregister state machine
1856 * @dismantle: Device is going to be freed
1857 * @rtnl_link_state: This enum represents the phases of creating
1860 * @needs_free_netdev: Should unregister perform free_netdev?
1861 * @priv_destructor: Called from unregister
1862 * @npinfo: XXX: need comments on this one
1863 * @nd_net: Network namespace this network device is inside
1865 * @ml_priv: Mid-layer private
1866 * @ml_priv_type: Mid-layer private type
1867 * @lstats: Loopback statistics
1868 * @tstats: Tunnel statistics
1869 * @dstats: Dummy statistics
1870 * @vstats: Virtual ethernet statistics
1875 * @dev: Class/net/name entry
1876 * @sysfs_groups: Space for optional device, statistics and wireless
1879 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1880 * @rtnl_link_ops: Rtnl_link_ops
1882 * @gso_max_size: Maximum size of generic segmentation offload
1883 * @gso_max_segs: Maximum number of segments that can be passed to the
1886 * @dcbnl_ops: Data Center Bridging netlink ops
1887 * @num_tc: Number of traffic classes in the net device
1888 * @tc_to_txq: XXX: need comments on this one
1889 * @prio_tc_map: XXX: need comments on this one
1891 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1893 * @priomap: XXX: need comments on this one
1894 * @phydev: Physical device may attach itself
1895 * for hardware timestamping
1896 * @sfp_bus: attached &struct sfp_bus structure.
1898 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1899 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1901 * @proto_down: protocol port state information can be sent to the
1902 * switch driver and used to set the phys state of the
1905 * @wol_enabled: Wake-on-LAN is enabled
1907 * @threaded: napi threaded mode is enabled
1909 * @net_notifier_list: List of per-net netdev notifier block
1910 * that follow this device when it is moved
1911 * to another network namespace.
1913 * @macsec_ops: MACsec offloading ops
1915 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1916 * offload capabilities of the device
1917 * @udp_tunnel_nic: UDP tunnel offload state
1918 * @xdp_state: stores info on attached XDP BPF programs
1920 * @nested_level: Used as as a parameter of spin_lock_nested() of
1921 * dev->addr_list_lock.
1922 * @unlink_list: As netif_addr_lock() can be called recursively,
1923 * keep a list of interfaces to be deleted.
1925 * FIXME: cleanup struct net_device such that network protocol info
1930 char name[IFNAMSIZ];
1931 struct netdev_name_node *name_node;
1932 struct dev_ifalias __rcu *ifalias;
1934 * I/O specific fields
1935 * FIXME: Merge these and struct ifmap into one
1937 unsigned long mem_end;
1938 unsigned long mem_start;
1939 unsigned long base_addr;
1942 * Some hardware also needs these fields (state,dev_list,
1943 * napi_list,unreg_list,close_list) but they are not
1944 * part of the usual set specified in Space.c.
1947 unsigned long state;
1949 struct list_head dev_list;
1950 struct list_head napi_list;
1951 struct list_head unreg_list;
1952 struct list_head close_list;
1953 struct list_head ptype_all;
1954 struct list_head ptype_specific;
1957 struct list_head upper;
1958 struct list_head lower;
1961 /* Read-mostly cache-line for fast-path access */
1963 unsigned int priv_flags;
1964 const struct net_device_ops *netdev_ops;
1966 unsigned short gflags;
1967 unsigned short hard_header_len;
1969 /* Note : dev->mtu is often read without holding a lock.
1970 * Writers usually hold RTNL.
1971 * It is recommended to use READ_ONCE() to annotate the reads,
1972 * and to use WRITE_ONCE() to annotate the writes.
1975 unsigned short needed_headroom;
1976 unsigned short needed_tailroom;
1978 netdev_features_t features;
1979 netdev_features_t hw_features;
1980 netdev_features_t wanted_features;
1981 netdev_features_t vlan_features;
1982 netdev_features_t hw_enc_features;
1983 netdev_features_t mpls_features;
1984 netdev_features_t gso_partial_features;
1986 unsigned int min_mtu;
1987 unsigned int max_mtu;
1988 unsigned short type;
1989 unsigned char min_header_len;
1990 unsigned char name_assign_type;
1994 struct net_device_stats stats; /* not used by modern drivers */
1996 atomic_long_t rx_dropped;
1997 atomic_long_t tx_dropped;
1998 atomic_long_t rx_nohandler;
2000 /* Stats to monitor link on/off, flapping */
2001 atomic_t carrier_up_count;
2002 atomic_t carrier_down_count;
2004 #ifdef CONFIG_WIRELESS_EXT
2005 const struct iw_handler_def *wireless_handlers;
2006 struct iw_public_data *wireless_data;
2008 const struct ethtool_ops *ethtool_ops;
2009 #ifdef CONFIG_NET_L3_MASTER_DEV
2010 const struct l3mdev_ops *l3mdev_ops;
2012 #if IS_ENABLED(CONFIG_IPV6)
2013 const struct ndisc_ops *ndisc_ops;
2016 #ifdef CONFIG_XFRM_OFFLOAD
2017 const struct xfrmdev_ops *xfrmdev_ops;
2020 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2021 const struct tlsdev_ops *tlsdev_ops;
2024 const struct header_ops *header_ops;
2026 unsigned char operstate;
2027 unsigned char link_mode;
2029 unsigned char if_port;
2032 /* Interface address info. */
2033 unsigned char perm_addr[MAX_ADDR_LEN];
2034 unsigned char addr_assign_type;
2035 unsigned char addr_len;
2036 unsigned char upper_level;
2037 unsigned char lower_level;
2039 unsigned short neigh_priv_len;
2040 unsigned short dev_id;
2041 unsigned short dev_port;
2042 unsigned short padded;
2044 spinlock_t addr_list_lock;
2047 struct netdev_hw_addr_list uc;
2048 struct netdev_hw_addr_list mc;
2049 struct netdev_hw_addr_list dev_addrs;
2052 struct kset *queues_kset;
2054 #ifdef CONFIG_LOCKDEP
2055 struct list_head unlink_list;
2057 unsigned int promiscuity;
2058 unsigned int allmulti;
2060 #ifdef CONFIG_LOCKDEP
2061 unsigned char nested_level;
2065 /* Protocol-specific pointers */
2067 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2068 struct vlan_info __rcu *vlan_info;
2070 #if IS_ENABLED(CONFIG_NET_DSA)
2071 struct dsa_port *dsa_ptr;
2073 #if IS_ENABLED(CONFIG_TIPC)
2074 struct tipc_bearer __rcu *tipc_ptr;
2076 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2079 struct in_device __rcu *ip_ptr;
2080 #if IS_ENABLED(CONFIG_DECNET)
2081 struct dn_dev __rcu *dn_ptr;
2083 struct inet6_dev __rcu *ip6_ptr;
2084 #if IS_ENABLED(CONFIG_AX25)
2087 struct wireless_dev *ieee80211_ptr;
2088 struct wpan_dev *ieee802154_ptr;
2089 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2090 struct mpls_dev __rcu *mpls_ptr;
2094 * Cache lines mostly used on receive path (including eth_type_trans())
2096 /* Interface address info used in eth_type_trans() */
2097 unsigned char *dev_addr;
2099 struct netdev_rx_queue *_rx;
2100 unsigned int num_rx_queues;
2101 unsigned int real_num_rx_queues;
2103 struct bpf_prog __rcu *xdp_prog;
2104 unsigned long gro_flush_timeout;
2105 int napi_defer_hard_irqs;
2106 rx_handler_func_t __rcu *rx_handler;
2107 void __rcu *rx_handler_data;
2109 #ifdef CONFIG_NET_CLS_ACT
2110 struct mini_Qdisc __rcu *miniq_ingress;
2112 struct netdev_queue __rcu *ingress_queue;
2113 #ifdef CONFIG_NETFILTER_INGRESS
2114 struct nf_hook_entries __rcu *nf_hooks_ingress;
2117 unsigned char broadcast[MAX_ADDR_LEN];
2118 #ifdef CONFIG_RFS_ACCEL
2119 struct cpu_rmap *rx_cpu_rmap;
2121 struct hlist_node index_hlist;
2124 * Cache lines mostly used on transmit path
2126 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2127 unsigned int num_tx_queues;
2128 unsigned int real_num_tx_queues;
2129 struct Qdisc *qdisc;
2130 unsigned int tx_queue_len;
2131 spinlock_t tx_global_lock;
2133 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2136 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2138 #ifdef CONFIG_NET_CLS_ACT
2139 struct mini_Qdisc __rcu *miniq_egress;
2142 #ifdef CONFIG_NET_SCHED
2143 DECLARE_HASHTABLE (qdisc_hash, 4);
2145 /* These may be needed for future network-power-down code. */
2146 struct timer_list watchdog_timer;
2149 u32 proto_down_reason;
2151 struct list_head todo_list;
2153 #ifdef CONFIG_PCPU_DEV_REFCNT
2154 int __percpu *pcpu_refcnt;
2156 refcount_t dev_refcnt;
2159 struct list_head link_watch_list;
2161 enum { NETREG_UNINITIALIZED=0,
2162 NETREG_REGISTERED, /* completed register_netdevice */
2163 NETREG_UNREGISTERING, /* called unregister_netdevice */
2164 NETREG_UNREGISTERED, /* completed unregister todo */
2165 NETREG_RELEASED, /* called free_netdev */
2166 NETREG_DUMMY, /* dummy device for NAPI poll */
2172 RTNL_LINK_INITIALIZED,
2173 RTNL_LINK_INITIALIZING,
2174 } rtnl_link_state:16;
2176 bool needs_free_netdev;
2177 void (*priv_destructor)(struct net_device *dev);
2179 #ifdef CONFIG_NETPOLL
2180 struct netpoll_info __rcu *npinfo;
2183 possible_net_t nd_net;
2185 /* mid-layer private */
2187 enum netdev_ml_priv_type ml_priv_type;
2190 struct pcpu_lstats __percpu *lstats;
2191 struct pcpu_sw_netstats __percpu *tstats;
2192 struct pcpu_dstats __percpu *dstats;
2195 #if IS_ENABLED(CONFIG_GARP)
2196 struct garp_port __rcu *garp_port;
2198 #if IS_ENABLED(CONFIG_MRP)
2199 struct mrp_port __rcu *mrp_port;
2203 const struct attribute_group *sysfs_groups[4];
2204 const struct attribute_group *sysfs_rx_queue_group;
2206 const struct rtnl_link_ops *rtnl_link_ops;
2208 /* for setting kernel sock attribute on TCP connection setup */
2209 #define GSO_MAX_SIZE 65536
2210 unsigned int gso_max_size;
2211 #define GSO_MAX_SEGS 65535
2215 const struct dcbnl_rtnl_ops *dcbnl_ops;
2218 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2219 u8 prio_tc_map[TC_BITMASK + 1];
2221 #if IS_ENABLED(CONFIG_FCOE)
2222 unsigned int fcoe_ddp_xid;
2224 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2225 struct netprio_map __rcu *priomap;
2227 struct phy_device *phydev;
2228 struct sfp_bus *sfp_bus;
2229 struct lock_class_key *qdisc_tx_busylock;
2230 struct lock_class_key *qdisc_running_key;
2232 unsigned wol_enabled:1;
2233 unsigned threaded:1;
2235 struct list_head net_notifier_list;
2237 #if IS_ENABLED(CONFIG_MACSEC)
2238 /* MACsec management functions */
2239 const struct macsec_ops *macsec_ops;
2241 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2242 struct udp_tunnel_nic *udp_tunnel_nic;
2244 /* protected by rtnl_lock */
2245 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2247 #define to_net_dev(d) container_of(d, struct net_device, dev)
2249 static inline bool netif_elide_gro(const struct net_device *dev)
2251 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2256 #define NETDEV_ALIGN 32
2259 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2261 return dev->prio_tc_map[prio & TC_BITMASK];
2265 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2267 if (tc >= dev->num_tc)
2270 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2274 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2275 void netdev_reset_tc(struct net_device *dev);
2276 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2277 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2280 int netdev_get_num_tc(struct net_device *dev)
2285 static inline void net_prefetch(void *p)
2288 #if L1_CACHE_BYTES < 128
2289 prefetch((u8 *)p + L1_CACHE_BYTES);
2293 static inline void net_prefetchw(void *p)
2296 #if L1_CACHE_BYTES < 128
2297 prefetchw((u8 *)p + L1_CACHE_BYTES);
2301 void netdev_unbind_sb_channel(struct net_device *dev,
2302 struct net_device *sb_dev);
2303 int netdev_bind_sb_channel_queue(struct net_device *dev,
2304 struct net_device *sb_dev,
2305 u8 tc, u16 count, u16 offset);
2306 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2307 static inline int netdev_get_sb_channel(struct net_device *dev)
2309 return max_t(int, -dev->num_tc, 0);
2313 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2316 return &dev->_tx[index];
2319 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2320 const struct sk_buff *skb)
2322 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2325 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2326 void (*f)(struct net_device *,
2327 struct netdev_queue *,
2333 for (i = 0; i < dev->num_tx_queues; i++)
2334 f(dev, &dev->_tx[i], arg);
2337 #define netdev_lockdep_set_classes(dev) \
2339 static struct lock_class_key qdisc_tx_busylock_key; \
2340 static struct lock_class_key qdisc_running_key; \
2341 static struct lock_class_key qdisc_xmit_lock_key; \
2342 static struct lock_class_key dev_addr_list_lock_key; \
2345 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2346 (dev)->qdisc_running_key = &qdisc_running_key; \
2347 lockdep_set_class(&(dev)->addr_list_lock, \
2348 &dev_addr_list_lock_key); \
2349 for (i = 0; i < (dev)->num_tx_queues; i++) \
2350 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2351 &qdisc_xmit_lock_key); \
2354 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2355 struct net_device *sb_dev);
2356 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2357 struct sk_buff *skb,
2358 struct net_device *sb_dev);
2360 /* returns the headroom that the master device needs to take in account
2361 * when forwarding to this dev
2363 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2365 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2368 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2370 if (dev->netdev_ops->ndo_set_rx_headroom)
2371 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2374 /* set the device rx headroom to the dev's default */
2375 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2377 netdev_set_rx_headroom(dev, -1);
2380 static inline void *netdev_get_ml_priv(struct net_device *dev,
2381 enum netdev_ml_priv_type type)
2383 if (dev->ml_priv_type != type)
2386 return dev->ml_priv;
2389 static inline void netdev_set_ml_priv(struct net_device *dev,
2391 enum netdev_ml_priv_type type)
2393 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2394 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2395 dev->ml_priv_type, type);
2396 WARN(!dev->ml_priv_type && dev->ml_priv,
2397 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2399 dev->ml_priv = ml_priv;
2400 dev->ml_priv_type = type;
2404 * Net namespace inlines
2407 struct net *dev_net(const struct net_device *dev)
2409 return read_pnet(&dev->nd_net);
2413 void dev_net_set(struct net_device *dev, struct net *net)
2415 write_pnet(&dev->nd_net, net);
2419 * netdev_priv - access network device private data
2420 * @dev: network device
2422 * Get network device private data
2424 static inline void *netdev_priv(const struct net_device *dev)
2426 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2429 /* Set the sysfs physical device reference for the network logical device
2430 * if set prior to registration will cause a symlink during initialization.
2432 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2434 /* Set the sysfs device type for the network logical device to allow
2435 * fine-grained identification of different network device types. For
2436 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2438 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2440 /* Default NAPI poll() weight
2441 * Device drivers are strongly advised to not use bigger value
2443 #define NAPI_POLL_WEIGHT 64
2446 * netif_napi_add - initialize a NAPI context
2447 * @dev: network device
2448 * @napi: NAPI context
2449 * @poll: polling function
2450 * @weight: default weight
2452 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2453 * *any* of the other NAPI-related functions.
2455 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2456 int (*poll)(struct napi_struct *, int), int weight);
2459 * netif_tx_napi_add - initialize a NAPI context
2460 * @dev: network device
2461 * @napi: NAPI context
2462 * @poll: polling function
2463 * @weight: default weight
2465 * This variant of netif_napi_add() should be used from drivers using NAPI
2466 * to exclusively poll a TX queue.
2467 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2469 static inline void netif_tx_napi_add(struct net_device *dev,
2470 struct napi_struct *napi,
2471 int (*poll)(struct napi_struct *, int),
2474 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2475 netif_napi_add(dev, napi, poll, weight);
2479 * __netif_napi_del - remove a NAPI context
2480 * @napi: NAPI context
2482 * Warning: caller must observe RCU grace period before freeing memory
2483 * containing @napi. Drivers might want to call this helper to combine
2484 * all the needed RCU grace periods into a single one.
2486 void __netif_napi_del(struct napi_struct *napi);
2489 * netif_napi_del - remove a NAPI context
2490 * @napi: NAPI context
2492 * netif_napi_del() removes a NAPI context from the network device NAPI list
2494 static inline void netif_napi_del(struct napi_struct *napi)
2496 __netif_napi_del(napi);
2500 struct napi_gro_cb {
2501 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2504 /* Length of frag0. */
2505 unsigned int frag0_len;
2507 /* This indicates where we are processing relative to skb->data. */
2510 /* This is non-zero if the packet cannot be merged with the new skb. */
2513 /* Save the IP ID here and check when we get to the transport layer */
2516 /* Number of segments aggregated. */
2519 /* Start offset for remote checksum offload */
2520 u16 gro_remcsum_start;
2522 /* jiffies when first packet was created/queued */
2525 /* Used in ipv6_gro_receive() and foo-over-udp */
2528 /* This is non-zero if the packet may be of the same flow. */
2531 /* Used in tunnel GRO receive */
2534 /* GRO checksum is valid */
2537 /* Number of checksums via CHECKSUM_UNNECESSARY */
2542 #define NAPI_GRO_FREE 1
2543 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2545 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2548 /* Used in GRE, set in fou/gue_gro_receive */
2551 /* Used to determine if flush_id can be ignored */
2554 /* Number of gro_receive callbacks this packet already went through */
2555 u8 recursion_counter:4;
2557 /* GRO is done by frag_list pointer chaining. */
2560 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2563 /* used in skb_gro_receive() slow path */
2564 struct sk_buff *last;
2567 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2569 #define GRO_RECURSION_LIMIT 15
2570 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2572 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2575 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2576 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2577 struct list_head *head,
2578 struct sk_buff *skb)
2580 if (unlikely(gro_recursion_inc_test(skb))) {
2581 NAPI_GRO_CB(skb)->flush |= 1;
2585 return cb(head, skb);
2588 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2590 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2592 struct list_head *head,
2593 struct sk_buff *skb)
2595 if (unlikely(gro_recursion_inc_test(skb))) {
2596 NAPI_GRO_CB(skb)->flush |= 1;
2600 return cb(sk, head, skb);
2603 struct packet_type {
2604 __be16 type; /* This is really htons(ether_type). */
2605 bool ignore_outgoing;
2606 struct net_device *dev; /* NULL is wildcarded here */
2607 int (*func) (struct sk_buff *,
2608 struct net_device *,
2609 struct packet_type *,
2610 struct net_device *);
2611 void (*list_func) (struct list_head *,
2612 struct packet_type *,
2613 struct net_device *);
2614 bool (*id_match)(struct packet_type *ptype,
2616 void *af_packet_priv;
2617 struct list_head list;
2620 struct offload_callbacks {
2621 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2622 netdev_features_t features);
2623 struct sk_buff *(*gro_receive)(struct list_head *head,
2624 struct sk_buff *skb);
2625 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2628 struct packet_offload {
2629 __be16 type; /* This is really htons(ether_type). */
2631 struct offload_callbacks callbacks;
2632 struct list_head list;
2635 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2636 struct pcpu_sw_netstats {
2641 struct u64_stats_sync syncp;
2642 } __aligned(4 * sizeof(u64));
2644 struct pcpu_lstats {
2645 u64_stats_t packets;
2647 struct u64_stats_sync syncp;
2648 } __aligned(2 * sizeof(u64));
2650 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2652 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2654 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2656 u64_stats_update_begin(&tstats->syncp);
2657 tstats->rx_bytes += len;
2658 tstats->rx_packets++;
2659 u64_stats_update_end(&tstats->syncp);
2662 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2663 unsigned int packets,
2666 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2668 u64_stats_update_begin(&tstats->syncp);
2669 tstats->tx_bytes += len;
2670 tstats->tx_packets += packets;
2671 u64_stats_update_end(&tstats->syncp);
2674 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2676 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2678 u64_stats_update_begin(&lstats->syncp);
2679 u64_stats_add(&lstats->bytes, len);
2680 u64_stats_inc(&lstats->packets);
2681 u64_stats_update_end(&lstats->syncp);
2684 #define __netdev_alloc_pcpu_stats(type, gfp) \
2686 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2689 for_each_possible_cpu(__cpu) { \
2690 typeof(type) *stat; \
2691 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2692 u64_stats_init(&stat->syncp); \
2698 #define netdev_alloc_pcpu_stats(type) \
2699 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2701 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2703 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2706 for_each_possible_cpu(__cpu) { \
2707 typeof(type) *stat; \
2708 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2709 u64_stats_init(&stat->syncp); \
2715 enum netdev_lag_tx_type {
2716 NETDEV_LAG_TX_TYPE_UNKNOWN,
2717 NETDEV_LAG_TX_TYPE_RANDOM,
2718 NETDEV_LAG_TX_TYPE_BROADCAST,
2719 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2720 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2721 NETDEV_LAG_TX_TYPE_HASH,
2724 enum netdev_lag_hash {
2725 NETDEV_LAG_HASH_NONE,
2727 NETDEV_LAG_HASH_L34,
2728 NETDEV_LAG_HASH_L23,
2729 NETDEV_LAG_HASH_E23,
2730 NETDEV_LAG_HASH_E34,
2731 NETDEV_LAG_HASH_VLAN_SRCMAC,
2732 NETDEV_LAG_HASH_UNKNOWN,
2735 struct netdev_lag_upper_info {
2736 enum netdev_lag_tx_type tx_type;
2737 enum netdev_lag_hash hash_type;
2740 struct netdev_lag_lower_state_info {
2745 #include <linux/notifier.h>
2747 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2748 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2752 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2754 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2755 detected a hardware crash and restarted
2756 - we can use this eg to kick tcp sessions
2758 NETDEV_CHANGE, /* Notify device state change */
2761 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2762 NETDEV_CHANGEADDR, /* notify after the address change */
2763 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2767 NETDEV_BONDING_FAILOVER,
2769 NETDEV_PRE_TYPE_CHANGE,
2770 NETDEV_POST_TYPE_CHANGE,
2773 NETDEV_NOTIFY_PEERS,
2777 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2778 NETDEV_CHANGEINFODATA,
2779 NETDEV_BONDING_INFO,
2780 NETDEV_PRECHANGEUPPER,
2781 NETDEV_CHANGELOWERSTATE,
2782 NETDEV_UDP_TUNNEL_PUSH_INFO,
2783 NETDEV_UDP_TUNNEL_DROP_INFO,
2784 NETDEV_CHANGE_TX_QUEUE_LEN,
2785 NETDEV_CVLAN_FILTER_PUSH_INFO,
2786 NETDEV_CVLAN_FILTER_DROP_INFO,
2787 NETDEV_SVLAN_FILTER_PUSH_INFO,
2788 NETDEV_SVLAN_FILTER_DROP_INFO,
2790 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2792 int register_netdevice_notifier(struct notifier_block *nb);
2793 int unregister_netdevice_notifier(struct notifier_block *nb);
2794 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2795 int unregister_netdevice_notifier_net(struct net *net,
2796 struct notifier_block *nb);
2797 int register_netdevice_notifier_dev_net(struct net_device *dev,
2798 struct notifier_block *nb,
2799 struct netdev_net_notifier *nn);
2800 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2801 struct notifier_block *nb,
2802 struct netdev_net_notifier *nn);
2804 struct netdev_notifier_info {
2805 struct net_device *dev;
2806 struct netlink_ext_ack *extack;
2809 struct netdev_notifier_info_ext {
2810 struct netdev_notifier_info info; /* must be first */
2816 struct netdev_notifier_change_info {
2817 struct netdev_notifier_info info; /* must be first */
2818 unsigned int flags_changed;
2821 struct netdev_notifier_changeupper_info {
2822 struct netdev_notifier_info info; /* must be first */
2823 struct net_device *upper_dev; /* new upper dev */
2824 bool master; /* is upper dev master */
2825 bool linking; /* is the notification for link or unlink */
2826 void *upper_info; /* upper dev info */
2829 struct netdev_notifier_changelowerstate_info {
2830 struct netdev_notifier_info info; /* must be first */
2831 void *lower_state_info; /* is lower dev state */
2834 struct netdev_notifier_pre_changeaddr_info {
2835 struct netdev_notifier_info info; /* must be first */
2836 const unsigned char *dev_addr;
2839 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2840 struct net_device *dev)
2843 info->extack = NULL;
2846 static inline struct net_device *
2847 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2852 static inline struct netlink_ext_ack *
2853 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2855 return info->extack;
2858 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2861 extern rwlock_t dev_base_lock; /* Device list lock */
2863 #define for_each_netdev(net, d) \
2864 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2865 #define for_each_netdev_reverse(net, d) \
2866 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2867 #define for_each_netdev_rcu(net, d) \
2868 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2869 #define for_each_netdev_safe(net, d, n) \
2870 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2871 #define for_each_netdev_continue(net, d) \
2872 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2873 #define for_each_netdev_continue_reverse(net, d) \
2874 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2876 #define for_each_netdev_continue_rcu(net, d) \
2877 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2878 #define for_each_netdev_in_bond_rcu(bond, slave) \
2879 for_each_netdev_rcu(&init_net, slave) \
2880 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2881 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2883 static inline struct net_device *next_net_device(struct net_device *dev)
2885 struct list_head *lh;
2889 lh = dev->dev_list.next;
2890 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2893 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2895 struct list_head *lh;
2899 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2900 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2903 static inline struct net_device *first_net_device(struct net *net)
2905 return list_empty(&net->dev_base_head) ? NULL :
2906 net_device_entry(net->dev_base_head.next);
2909 static inline struct net_device *first_net_device_rcu(struct net *net)
2911 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2913 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2916 int netdev_boot_setup_check(struct net_device *dev);
2917 unsigned long netdev_boot_base(const char *prefix, int unit);
2918 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2919 const char *hwaddr);
2920 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2921 void dev_add_pack(struct packet_type *pt);
2922 void dev_remove_pack(struct packet_type *pt);
2923 void __dev_remove_pack(struct packet_type *pt);
2924 void dev_add_offload(struct packet_offload *po);
2925 void dev_remove_offload(struct packet_offload *po);
2927 int dev_get_iflink(const struct net_device *dev);
2928 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2929 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2930 struct net_device_path_stack *stack);
2931 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2932 unsigned short mask);
2933 struct net_device *dev_get_by_name(struct net *net, const char *name);
2934 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2935 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2936 int dev_alloc_name(struct net_device *dev, const char *name);
2937 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2938 void dev_close(struct net_device *dev);
2939 void dev_close_many(struct list_head *head, bool unlink);
2940 void dev_disable_lro(struct net_device *dev);
2941 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2942 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2943 struct net_device *sb_dev);
2944 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2945 struct net_device *sb_dev);
2947 int dev_queue_xmit(struct sk_buff *skb);
2948 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2949 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2951 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2955 ret = __dev_direct_xmit(skb, queue_id);
2956 if (!dev_xmit_complete(ret))
2961 int register_netdevice(struct net_device *dev);
2962 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2963 void unregister_netdevice_many(struct list_head *head);
2964 static inline void unregister_netdevice(struct net_device *dev)
2966 unregister_netdevice_queue(dev, NULL);
2969 int netdev_refcnt_read(const struct net_device *dev);
2970 void free_netdev(struct net_device *dev);
2971 void netdev_freemem(struct net_device *dev);
2972 int init_dummy_netdev(struct net_device *dev);
2974 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2975 struct sk_buff *skb,
2977 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2979 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2980 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2981 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2982 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2983 int netdev_get_name(struct net *net, char *name, int ifindex);
2984 int dev_restart(struct net_device *dev);
2985 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2986 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2988 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2990 return NAPI_GRO_CB(skb)->data_offset;
2993 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2995 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2998 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3000 NAPI_GRO_CB(skb)->data_offset += len;
3003 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3004 unsigned int offset)
3006 return NAPI_GRO_CB(skb)->frag0 + offset;
3009 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3011 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3014 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3016 NAPI_GRO_CB(skb)->frag0 = NULL;
3017 NAPI_GRO_CB(skb)->frag0_len = 0;
3020 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3021 unsigned int offset)
3023 if (!pskb_may_pull(skb, hlen))
3026 skb_gro_frag0_invalidate(skb);
3027 return skb->data + offset;
3030 static inline void *skb_gro_network_header(struct sk_buff *skb)
3032 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3033 skb_network_offset(skb);
3036 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3037 const void *start, unsigned int len)
3039 if (NAPI_GRO_CB(skb)->csum_valid)
3040 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3041 csum_partial(start, len, 0));
3044 /* GRO checksum functions. These are logical equivalents of the normal
3045 * checksum functions (in skbuff.h) except that they operate on the GRO
3046 * offsets and fields in sk_buff.
3049 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3051 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3053 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3056 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3060 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3061 skb_checksum_start_offset(skb) <
3062 skb_gro_offset(skb)) &&
3063 !skb_at_gro_remcsum_start(skb) &&
3064 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3065 (!zero_okay || check));
3068 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3071 if (NAPI_GRO_CB(skb)->csum_valid &&
3072 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3075 NAPI_GRO_CB(skb)->csum = psum;
3077 return __skb_gro_checksum_complete(skb);
3080 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3082 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3083 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3084 NAPI_GRO_CB(skb)->csum_cnt--;
3086 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3087 * verified a new top level checksum or an encapsulated one
3088 * during GRO. This saves work if we fallback to normal path.
3090 __skb_incr_checksum_unnecessary(skb);
3094 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3097 __sum16 __ret = 0; \
3098 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3099 __ret = __skb_gro_checksum_validate_complete(skb, \
3100 compute_pseudo(skb, proto)); \
3102 skb_gro_incr_csum_unnecessary(skb); \
3106 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3107 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3109 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3111 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3113 #define skb_gro_checksum_simple_validate(skb) \
3114 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3116 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3118 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3119 !NAPI_GRO_CB(skb)->csum_valid);
3122 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3125 NAPI_GRO_CB(skb)->csum = ~pseudo;
3126 NAPI_GRO_CB(skb)->csum_valid = 1;
3129 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3131 if (__skb_gro_checksum_convert_check(skb)) \
3132 __skb_gro_checksum_convert(skb, \
3133 compute_pseudo(skb, proto)); \
3136 struct gro_remcsum {
3141 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3147 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3148 unsigned int off, size_t hdrlen,
3149 int start, int offset,
3150 struct gro_remcsum *grc,
3154 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3156 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3159 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3163 ptr = skb_gro_header_fast(skb, off);
3164 if (skb_gro_header_hard(skb, off + plen)) {
3165 ptr = skb_gro_header_slow(skb, off + plen, off);
3170 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3173 /* Adjust skb->csum since we changed the packet */
3174 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3176 grc->offset = off + hdrlen + offset;
3182 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3183 struct gro_remcsum *grc)
3186 size_t plen = grc->offset + sizeof(u16);
3191 ptr = skb_gro_header_fast(skb, grc->offset);
3192 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3193 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3198 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3201 #ifdef CONFIG_XFRM_OFFLOAD
3202 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3204 if (PTR_ERR(pp) != -EINPROGRESS)
3205 NAPI_GRO_CB(skb)->flush |= flush;
3207 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3210 struct gro_remcsum *grc)
3212 if (PTR_ERR(pp) != -EINPROGRESS) {
3213 NAPI_GRO_CB(skb)->flush |= flush;
3214 skb_gro_remcsum_cleanup(skb, grc);
3215 skb->remcsum_offload = 0;
3219 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3221 NAPI_GRO_CB(skb)->flush |= flush;
3223 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3226 struct gro_remcsum *grc)
3228 NAPI_GRO_CB(skb)->flush |= flush;
3229 skb_gro_remcsum_cleanup(skb, grc);
3230 skb->remcsum_offload = 0;
3234 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3235 unsigned short type,
3236 const void *daddr, const void *saddr,
3239 if (!dev->header_ops || !dev->header_ops->create)
3242 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3245 static inline int dev_parse_header(const struct sk_buff *skb,
3246 unsigned char *haddr)
3248 const struct net_device *dev = skb->dev;
3250 if (!dev->header_ops || !dev->header_ops->parse)
3252 return dev->header_ops->parse(skb, haddr);
3255 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3257 const struct net_device *dev = skb->dev;
3259 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3261 return dev->header_ops->parse_protocol(skb);
3264 /* ll_header must have at least hard_header_len allocated */
3265 static inline bool dev_validate_header(const struct net_device *dev,
3266 char *ll_header, int len)
3268 if (likely(len >= dev->hard_header_len))
3270 if (len < dev->min_header_len)
3273 if (capable(CAP_SYS_RAWIO)) {
3274 memset(ll_header + len, 0, dev->hard_header_len - len);
3278 if (dev->header_ops && dev->header_ops->validate)
3279 return dev->header_ops->validate(ll_header, len);
3284 static inline bool dev_has_header(const struct net_device *dev)
3286 return dev->header_ops && dev->header_ops->create;
3289 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3291 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3292 static inline int unregister_gifconf(unsigned int family)
3294 return register_gifconf(family, NULL);
3297 #ifdef CONFIG_NET_FLOW_LIMIT
3298 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3299 struct sd_flow_limit {
3301 unsigned int num_buckets;
3302 unsigned int history_head;
3303 u16 history[FLOW_LIMIT_HISTORY];
3307 extern int netdev_flow_limit_table_len;
3308 #endif /* CONFIG_NET_FLOW_LIMIT */
3311 * Incoming packets are placed on per-CPU queues
3313 struct softnet_data {
3314 struct list_head poll_list;
3315 struct sk_buff_head process_queue;
3318 unsigned int processed;
3319 unsigned int time_squeeze;
3320 unsigned int received_rps;
3322 struct softnet_data *rps_ipi_list;
3324 #ifdef CONFIG_NET_FLOW_LIMIT
3325 struct sd_flow_limit __rcu *flow_limit;
3327 struct Qdisc *output_queue;
3328 struct Qdisc **output_queue_tailp;
3329 struct sk_buff *completion_queue;
3330 #ifdef CONFIG_XFRM_OFFLOAD
3331 struct sk_buff_head xfrm_backlog;
3333 /* written and read only by owning cpu: */
3339 /* input_queue_head should be written by cpu owning this struct,
3340 * and only read by other cpus. Worth using a cache line.
3342 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3344 /* Elements below can be accessed between CPUs for RPS/RFS */
3345 call_single_data_t csd ____cacheline_aligned_in_smp;
3346 struct softnet_data *rps_ipi_next;
3348 unsigned int input_queue_tail;
3350 unsigned int dropped;
3351 struct sk_buff_head input_pkt_queue;
3352 struct napi_struct backlog;
3356 static inline void input_queue_head_incr(struct softnet_data *sd)
3359 sd->input_queue_head++;
3363 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3364 unsigned int *qtail)
3367 *qtail = ++sd->input_queue_tail;
3371 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3373 static inline int dev_recursion_level(void)
3375 return this_cpu_read(softnet_data.xmit.recursion);
3378 #define XMIT_RECURSION_LIMIT 8
3379 static inline bool dev_xmit_recursion(void)
3381 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3382 XMIT_RECURSION_LIMIT);
3385 static inline void dev_xmit_recursion_inc(void)
3387 __this_cpu_inc(softnet_data.xmit.recursion);
3390 static inline void dev_xmit_recursion_dec(void)
3392 __this_cpu_dec(softnet_data.xmit.recursion);
3395 void __netif_schedule(struct Qdisc *q);
3396 void netif_schedule_queue(struct netdev_queue *txq);
3398 static inline void netif_tx_schedule_all(struct net_device *dev)
3402 for (i = 0; i < dev->num_tx_queues; i++)
3403 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3406 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3408 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3412 * netif_start_queue - allow transmit
3413 * @dev: network device
3415 * Allow upper layers to call the device hard_start_xmit routine.
3417 static inline void netif_start_queue(struct net_device *dev)
3419 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3422 static inline void netif_tx_start_all_queues(struct net_device *dev)
3426 for (i = 0; i < dev->num_tx_queues; i++) {
3427 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3428 netif_tx_start_queue(txq);
3432 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3435 * netif_wake_queue - restart transmit
3436 * @dev: network device
3438 * Allow upper layers to call the device hard_start_xmit routine.
3439 * Used for flow control when transmit resources are available.
3441 static inline void netif_wake_queue(struct net_device *dev)
3443 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3446 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3450 for (i = 0; i < dev->num_tx_queues; i++) {
3451 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3452 netif_tx_wake_queue(txq);
3456 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3458 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3462 * netif_stop_queue - stop transmitted packets
3463 * @dev: network device
3465 * Stop upper layers calling the device hard_start_xmit routine.
3466 * Used for flow control when transmit resources are unavailable.
3468 static inline void netif_stop_queue(struct net_device *dev)
3470 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3473 void netif_tx_stop_all_queues(struct net_device *dev);
3475 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3477 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3481 * netif_queue_stopped - test if transmit queue is flowblocked
3482 * @dev: network device
3484 * Test if transmit queue on device is currently unable to send.
3486 static inline bool netif_queue_stopped(const struct net_device *dev)
3488 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3491 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3493 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3497 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3499 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3503 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3505 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3509 * netdev_queue_set_dql_min_limit - set dql minimum limit
3510 * @dev_queue: pointer to transmit queue
3511 * @min_limit: dql minimum limit
3513 * Forces xmit_more() to return true until the minimum threshold
3514 * defined by @min_limit is reached (or until the tx queue is
3515 * empty). Warning: to be use with care, misuse will impact the
3518 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3519 unsigned int min_limit)
3522 dev_queue->dql.min_limit = min_limit;
3527 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3528 * @dev_queue: pointer to transmit queue
3530 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3531 * to give appropriate hint to the CPU.
3533 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3536 prefetchw(&dev_queue->dql.num_queued);
3541 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3542 * @dev_queue: pointer to transmit queue
3544 * BQL enabled drivers might use this helper in their TX completion path,
3545 * to give appropriate hint to the CPU.
3547 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3550 prefetchw(&dev_queue->dql.limit);
3554 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3558 dql_queued(&dev_queue->dql, bytes);
3560 if (likely(dql_avail(&dev_queue->dql) >= 0))
3563 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3566 * The XOFF flag must be set before checking the dql_avail below,
3567 * because in netdev_tx_completed_queue we update the dql_completed
3568 * before checking the XOFF flag.
3572 /* check again in case another CPU has just made room avail */
3573 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3574 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3578 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3579 * that they should not test BQL status themselves.
3580 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3582 * Returns true if the doorbell must be used to kick the NIC.
3584 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3590 dql_queued(&dev_queue->dql, bytes);
3592 return netif_tx_queue_stopped(dev_queue);
3594 netdev_tx_sent_queue(dev_queue, bytes);
3599 * netdev_sent_queue - report the number of bytes queued to hardware
3600 * @dev: network device
3601 * @bytes: number of bytes queued to the hardware device queue
3603 * Report the number of bytes queued for sending/completion to the network
3604 * device hardware queue. @bytes should be a good approximation and should
3605 * exactly match netdev_completed_queue() @bytes
3607 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3609 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3612 static inline bool __netdev_sent_queue(struct net_device *dev,
3616 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3620 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3621 unsigned int pkts, unsigned int bytes)
3624 if (unlikely(!bytes))
3627 dql_completed(&dev_queue->dql, bytes);
3630 * Without the memory barrier there is a small possiblity that
3631 * netdev_tx_sent_queue will miss the update and cause the queue to
3632 * be stopped forever
3636 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3639 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3640 netif_schedule_queue(dev_queue);
3645 * netdev_completed_queue - report bytes and packets completed by device
3646 * @dev: network device
3647 * @pkts: actual number of packets sent over the medium
3648 * @bytes: actual number of bytes sent over the medium
3650 * Report the number of bytes and packets transmitted by the network device
3651 * hardware queue over the physical medium, @bytes must exactly match the
3652 * @bytes amount passed to netdev_sent_queue()
3654 static inline void netdev_completed_queue(struct net_device *dev,
3655 unsigned int pkts, unsigned int bytes)
3657 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3660 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3663 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3669 * netdev_reset_queue - reset the packets and bytes count of a network device
3670 * @dev_queue: network device
3672 * Reset the bytes and packet count of a network device and clear the
3673 * software flow control OFF bit for this network device
3675 static inline void netdev_reset_queue(struct net_device *dev_queue)
3677 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3681 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3682 * @dev: network device
3683 * @queue_index: given tx queue index
3685 * Returns 0 if given tx queue index >= number of device tx queues,
3686 * otherwise returns the originally passed tx queue index.
3688 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3690 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3691 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3692 dev->name, queue_index,
3693 dev->real_num_tx_queues);
3701 * netif_running - test if up
3702 * @dev: network device
3704 * Test if the device has been brought up.
3706 static inline bool netif_running(const struct net_device *dev)
3708 return test_bit(__LINK_STATE_START, &dev->state);
3712 * Routines to manage the subqueues on a device. We only need start,
3713 * stop, and a check if it's stopped. All other device management is
3714 * done at the overall netdevice level.
3715 * Also test the device if we're multiqueue.
3719 * netif_start_subqueue - allow sending packets on subqueue
3720 * @dev: network device
3721 * @queue_index: sub queue index
3723 * Start individual transmit queue of a device with multiple transmit queues.
3725 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3727 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3729 netif_tx_start_queue(txq);
3733 * netif_stop_subqueue - stop sending packets on subqueue
3734 * @dev: network device
3735 * @queue_index: sub queue index
3737 * Stop individual transmit queue of a device with multiple transmit queues.
3739 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3741 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3742 netif_tx_stop_queue(txq);
3746 * __netif_subqueue_stopped - test status of subqueue
3747 * @dev: network device
3748 * @queue_index: sub queue index
3750 * Check individual transmit queue of a device with multiple transmit queues.
3752 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3755 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3757 return netif_tx_queue_stopped(txq);
3761 * netif_subqueue_stopped - test status of subqueue
3762 * @dev: network device
3763 * @skb: sub queue buffer pointer
3765 * Check individual transmit queue of a device with multiple transmit queues.
3767 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3768 struct sk_buff *skb)
3770 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3774 * netif_wake_subqueue - allow sending packets on subqueue
3775 * @dev: network device
3776 * @queue_index: sub queue index
3778 * Resume individual transmit queue of a device with multiple transmit queues.
3780 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3782 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3784 netif_tx_wake_queue(txq);
3788 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3790 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3791 u16 index, enum xps_map_type type);
3794 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3795 * @j: CPU/Rx queue index
3796 * @mask: bitmask of all cpus/rx queues
3797 * @nr_bits: number of bits in the bitmask
3799 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3801 static inline bool netif_attr_test_mask(unsigned long j,
3802 const unsigned long *mask,
3803 unsigned int nr_bits)
3805 cpu_max_bits_warn(j, nr_bits);
3806 return test_bit(j, mask);
3810 * netif_attr_test_online - Test for online CPU/Rx queue
3811 * @j: CPU/Rx queue index
3812 * @online_mask: bitmask for CPUs/Rx queues that are online
3813 * @nr_bits: number of bits in the bitmask
3815 * Returns true if a CPU/Rx queue is online.
3817 static inline bool netif_attr_test_online(unsigned long j,
3818 const unsigned long *online_mask,
3819 unsigned int nr_bits)
3821 cpu_max_bits_warn(j, nr_bits);
3824 return test_bit(j, online_mask);
3826 return (j < nr_bits);
3830 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3831 * @n: CPU/Rx queue index
3832 * @srcp: the cpumask/Rx queue mask pointer
3833 * @nr_bits: number of bits in the bitmask
3835 * Returns >= nr_bits if no further CPUs/Rx queues set.
3837 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3838 unsigned int nr_bits)
3840 /* -1 is a legal arg here. */
3842 cpu_max_bits_warn(n, nr_bits);
3845 return find_next_bit(srcp, nr_bits, n + 1);
3851 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3852 * @n: CPU/Rx queue index
3853 * @src1p: the first CPUs/Rx queues mask pointer
3854 * @src2p: the second CPUs/Rx queues mask pointer
3855 * @nr_bits: number of bits in the bitmask
3857 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3859 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3860 const unsigned long *src2p,
3861 unsigned int nr_bits)
3863 /* -1 is a legal arg here. */
3865 cpu_max_bits_warn(n, nr_bits);
3868 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3870 return find_next_bit(src1p, nr_bits, n + 1);
3872 return find_next_bit(src2p, nr_bits, n + 1);
3877 static inline int netif_set_xps_queue(struct net_device *dev,
3878 const struct cpumask *mask,
3884 static inline int __netif_set_xps_queue(struct net_device *dev,
3885 const unsigned long *mask,
3886 u16 index, enum xps_map_type type)
3893 * netif_is_multiqueue - test if device has multiple transmit queues
3894 * @dev: network device
3896 * Check if device has multiple transmit queues
3898 static inline bool netif_is_multiqueue(const struct net_device *dev)
3900 return dev->num_tx_queues > 1;
3903 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3906 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3908 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3911 dev->real_num_rx_queues = rxqs;
3916 static inline struct netdev_rx_queue *
3917 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3919 return dev->_rx + rxq;
3923 static inline unsigned int get_netdev_rx_queue_index(
3924 struct netdev_rx_queue *queue)
3926 struct net_device *dev = queue->dev;
3927 int index = queue - dev->_rx;
3929 BUG_ON(index >= dev->num_rx_queues);
3934 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3935 int netif_get_num_default_rss_queues(void);
3937 enum skb_free_reason {
3938 SKB_REASON_CONSUMED,
3942 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3943 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3946 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3947 * interrupt context or with hardware interrupts being disabled.
3948 * (in_irq() || irqs_disabled())
3950 * We provide four helpers that can be used in following contexts :
3952 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3953 * replacing kfree_skb(skb)
3955 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3956 * Typically used in place of consume_skb(skb) in TX completion path
3958 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3959 * replacing kfree_skb(skb)
3961 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3962 * and consumed a packet. Used in place of consume_skb(skb)
3964 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3966 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3969 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3971 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3974 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3976 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3979 static inline void dev_consume_skb_any(struct sk_buff *skb)
3981 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3984 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3985 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3986 int netif_rx(struct sk_buff *skb);
3987 int netif_rx_ni(struct sk_buff *skb);
3988 int netif_rx_any_context(struct sk_buff *skb);
3989 int netif_receive_skb(struct sk_buff *skb);
3990 int netif_receive_skb_core(struct sk_buff *skb);
3991 void netif_receive_skb_list(struct list_head *head);
3992 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3993 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3994 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3995 gro_result_t napi_gro_frags(struct napi_struct *napi);
3996 struct packet_offload *gro_find_receive_by_type(__be16 type);
3997 struct packet_offload *gro_find_complete_by_type(__be16 type);
3999 static inline void napi_free_frags(struct napi_struct *napi)
4001 kfree_skb(napi->skb);
4005 bool netdev_is_rx_handler_busy(struct net_device *dev);
4006 int netdev_rx_handler_register(struct net_device *dev,
4007 rx_handler_func_t *rx_handler,
4008 void *rx_handler_data);
4009 void netdev_rx_handler_unregister(struct net_device *dev);
4011 bool dev_valid_name(const char *name);
4012 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4013 bool *need_copyout);
4014 int dev_ifconf(struct net *net, struct ifconf *, int);
4015 int dev_ethtool(struct net *net, struct ifreq *);
4016 unsigned int dev_get_flags(const struct net_device *);
4017 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4018 struct netlink_ext_ack *extack);
4019 int dev_change_flags(struct net_device *dev, unsigned int flags,
4020 struct netlink_ext_ack *extack);
4021 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4022 unsigned int gchanges);
4023 int dev_change_name(struct net_device *, const char *);
4024 int dev_set_alias(struct net_device *, const char *, size_t);
4025 int dev_get_alias(const struct net_device *, char *, size_t);
4026 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
4027 int __dev_set_mtu(struct net_device *, int);
4028 int dev_validate_mtu(struct net_device *dev, int mtu,
4029 struct netlink_ext_ack *extack);
4030 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4031 struct netlink_ext_ack *extack);
4032 int dev_set_mtu(struct net_device *, int);
4033 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4034 void dev_set_group(struct net_device *, int);
4035 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4036 struct netlink_ext_ack *extack);
4037 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4038 struct netlink_ext_ack *extack);
4039 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4040 struct netlink_ext_ack *extack);
4041 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4042 int dev_change_carrier(struct net_device *, bool new_carrier);
4043 int dev_get_phys_port_id(struct net_device *dev,
4044 struct netdev_phys_item_id *ppid);
4045 int dev_get_phys_port_name(struct net_device *dev,
4046 char *name, size_t len);
4047 int dev_get_port_parent_id(struct net_device *dev,
4048 struct netdev_phys_item_id *ppid, bool recurse);
4049 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4050 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4051 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4052 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4054 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4055 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4056 struct netdev_queue *txq, int *ret);
4058 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4059 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4060 int fd, int expected_fd, u32 flags);
4061 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4062 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4064 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4065 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4066 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4067 bool is_skb_forwardable(const struct net_device *dev,
4068 const struct sk_buff *skb);
4070 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4071 const struct sk_buff *skb,
4072 const bool check_mtu)
4074 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4077 if (!(dev->flags & IFF_UP))
4083 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4084 if (skb->len <= len)
4087 /* if TSO is enabled, we don't care about the length as the packet
4088 * could be forwarded without being segmented before
4090 if (skb_is_gso(skb))
4096 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4097 struct sk_buff *skb,
4098 const bool check_mtu)
4100 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4101 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4102 atomic_long_inc(&dev->rx_dropped);
4107 skb_scrub_packet(skb, true);
4112 bool dev_nit_active(struct net_device *dev);
4113 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4115 extern int netdev_budget;
4116 extern unsigned int netdev_budget_usecs;
4118 /* Called by rtnetlink.c:rtnl_unlock() */
4119 void netdev_run_todo(void);
4122 * dev_put - release reference to device
4123 * @dev: network device
4125 * Release reference to device to allow it to be freed.
4127 static inline void dev_put(struct net_device *dev)
4129 #ifdef CONFIG_PCPU_DEV_REFCNT
4130 this_cpu_dec(*dev->pcpu_refcnt);
4132 refcount_dec(&dev->dev_refcnt);
4137 * dev_hold - get reference to device
4138 * @dev: network device
4140 * Hold reference to device to keep it from being freed.
4142 static inline void dev_hold(struct net_device *dev)
4144 #ifdef CONFIG_PCPU_DEV_REFCNT
4145 this_cpu_inc(*dev->pcpu_refcnt);
4147 refcount_inc(&dev->dev_refcnt);
4151 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4152 * and _off may be called from IRQ context, but it is caller
4153 * who is responsible for serialization of these calls.
4155 * The name carrier is inappropriate, these functions should really be
4156 * called netif_lowerlayer_*() because they represent the state of any
4157 * kind of lower layer not just hardware media.
4160 void linkwatch_init_dev(struct net_device *dev);
4161 void linkwatch_fire_event(struct net_device *dev);
4162 void linkwatch_forget_dev(struct net_device *dev);
4165 * netif_carrier_ok - test if carrier present
4166 * @dev: network device
4168 * Check if carrier is present on device
4170 static inline bool netif_carrier_ok(const struct net_device *dev)
4172 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4175 unsigned long dev_trans_start(struct net_device *dev);
4177 void __netdev_watchdog_up(struct net_device *dev);
4179 void netif_carrier_on(struct net_device *dev);
4181 void netif_carrier_off(struct net_device *dev);
4184 * netif_dormant_on - mark device as dormant.
4185 * @dev: network device
4187 * Mark device as dormant (as per RFC2863).
4189 * The dormant state indicates that the relevant interface is not
4190 * actually in a condition to pass packets (i.e., it is not 'up') but is
4191 * in a "pending" state, waiting for some external event. For "on-
4192 * demand" interfaces, this new state identifies the situation where the
4193 * interface is waiting for events to place it in the up state.
4195 static inline void netif_dormant_on(struct net_device *dev)
4197 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4198 linkwatch_fire_event(dev);
4202 * netif_dormant_off - set device as not dormant.
4203 * @dev: network device
4205 * Device is not in dormant state.
4207 static inline void netif_dormant_off(struct net_device *dev)
4209 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4210 linkwatch_fire_event(dev);
4214 * netif_dormant - test if device is dormant
4215 * @dev: network device
4217 * Check if device is dormant.
4219 static inline bool netif_dormant(const struct net_device *dev)
4221 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4226 * netif_testing_on - mark device as under test.
4227 * @dev: network device
4229 * Mark device as under test (as per RFC2863).
4231 * The testing state indicates that some test(s) must be performed on
4232 * the interface. After completion, of the test, the interface state
4233 * will change to up, dormant, or down, as appropriate.
4235 static inline void netif_testing_on(struct net_device *dev)
4237 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4238 linkwatch_fire_event(dev);
4242 * netif_testing_off - set device as not under test.
4243 * @dev: network device
4245 * Device is not in testing state.
4247 static inline void netif_testing_off(struct net_device *dev)
4249 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4250 linkwatch_fire_event(dev);
4254 * netif_testing - test if device is under test
4255 * @dev: network device
4257 * Check if device is under test
4259 static inline bool netif_testing(const struct net_device *dev)
4261 return test_bit(__LINK_STATE_TESTING, &dev->state);
4266 * netif_oper_up - test if device is operational
4267 * @dev: network device
4269 * Check if carrier is operational
4271 static inline bool netif_oper_up(const struct net_device *dev)
4273 return (dev->operstate == IF_OPER_UP ||
4274 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4278 * netif_device_present - is device available or removed
4279 * @dev: network device
4281 * Check if device has not been removed from system.
4283 static inline bool netif_device_present(const struct net_device *dev)
4285 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4288 void netif_device_detach(struct net_device *dev);
4290 void netif_device_attach(struct net_device *dev);
4293 * Network interface message level settings
4298 NETIF_MSG_PROBE_BIT,
4300 NETIF_MSG_TIMER_BIT,
4301 NETIF_MSG_IFDOWN_BIT,
4303 NETIF_MSG_RX_ERR_BIT,
4304 NETIF_MSG_TX_ERR_BIT,
4305 NETIF_MSG_TX_QUEUED_BIT,
4307 NETIF_MSG_TX_DONE_BIT,
4308 NETIF_MSG_RX_STATUS_BIT,
4309 NETIF_MSG_PKTDATA_BIT,
4313 /* When you add a new bit above, update netif_msg_class_names array
4314 * in net/ethtool/common.c
4316 NETIF_MSG_CLASS_COUNT,
4318 /* Both ethtool_ops interface and internal driver implementation use u32 */
4319 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4321 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4322 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4324 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4325 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4326 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4327 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4328 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4329 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4330 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4331 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4332 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4333 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4334 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4335 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4336 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4337 #define NETIF_MSG_HW __NETIF_MSG(HW)
4338 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4340 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4341 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4342 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4343 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4344 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4345 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4346 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4347 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4348 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4349 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4350 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4351 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4352 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4353 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4354 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4356 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4359 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4360 return default_msg_enable_bits;
4361 if (debug_value == 0) /* no output */
4363 /* set low N bits */
4364 return (1U << debug_value) - 1;
4367 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4369 spin_lock(&txq->_xmit_lock);
4370 txq->xmit_lock_owner = cpu;
4373 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4375 __acquire(&txq->_xmit_lock);
4379 static inline void __netif_tx_release(struct netdev_queue *txq)
4381 __release(&txq->_xmit_lock);
4384 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4386 spin_lock_bh(&txq->_xmit_lock);
4387 txq->xmit_lock_owner = smp_processor_id();
4390 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4392 bool ok = spin_trylock(&txq->_xmit_lock);
4394 txq->xmit_lock_owner = smp_processor_id();
4398 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4400 txq->xmit_lock_owner = -1;
4401 spin_unlock(&txq->_xmit_lock);
4404 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4406 txq->xmit_lock_owner = -1;
4407 spin_unlock_bh(&txq->_xmit_lock);
4410 static inline void txq_trans_update(struct netdev_queue *txq)
4412 if (txq->xmit_lock_owner != -1)
4413 txq->trans_start = jiffies;
4416 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4417 static inline void netif_trans_update(struct net_device *dev)
4419 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4421 if (txq->trans_start != jiffies)
4422 txq->trans_start = jiffies;
4426 * netif_tx_lock - grab network device transmit lock
4427 * @dev: network device
4429 * Get network device transmit lock
4431 static inline void netif_tx_lock(struct net_device *dev)
4436 spin_lock(&dev->tx_global_lock);
4437 cpu = smp_processor_id();
4438 for (i = 0; i < dev->num_tx_queues; i++) {
4439 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4441 /* We are the only thread of execution doing a
4442 * freeze, but we have to grab the _xmit_lock in
4443 * order to synchronize with threads which are in
4444 * the ->hard_start_xmit() handler and already
4445 * checked the frozen bit.
4447 __netif_tx_lock(txq, cpu);
4448 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4449 __netif_tx_unlock(txq);
4453 static inline void netif_tx_lock_bh(struct net_device *dev)
4459 static inline void netif_tx_unlock(struct net_device *dev)
4463 for (i = 0; i < dev->num_tx_queues; i++) {
4464 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4466 /* No need to grab the _xmit_lock here. If the
4467 * queue is not stopped for another reason, we
4470 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4471 netif_schedule_queue(txq);
4473 spin_unlock(&dev->tx_global_lock);
4476 static inline void netif_tx_unlock_bh(struct net_device *dev)
4478 netif_tx_unlock(dev);
4482 #define HARD_TX_LOCK(dev, txq, cpu) { \
4483 if ((dev->features & NETIF_F_LLTX) == 0) { \
4484 __netif_tx_lock(txq, cpu); \
4486 __netif_tx_acquire(txq); \
4490 #define HARD_TX_TRYLOCK(dev, txq) \
4491 (((dev->features & NETIF_F_LLTX) == 0) ? \
4492 __netif_tx_trylock(txq) : \
4493 __netif_tx_acquire(txq))
4495 #define HARD_TX_UNLOCK(dev, txq) { \
4496 if ((dev->features & NETIF_F_LLTX) == 0) { \
4497 __netif_tx_unlock(txq); \
4499 __netif_tx_release(txq); \
4503 static inline void netif_tx_disable(struct net_device *dev)
4509 cpu = smp_processor_id();
4510 spin_lock(&dev->tx_global_lock);
4511 for (i = 0; i < dev->num_tx_queues; i++) {
4512 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4514 __netif_tx_lock(txq, cpu);
4515 netif_tx_stop_queue(txq);
4516 __netif_tx_unlock(txq);
4518 spin_unlock(&dev->tx_global_lock);
4522 static inline void netif_addr_lock(struct net_device *dev)
4524 unsigned char nest_level = 0;
4526 #ifdef CONFIG_LOCKDEP
4527 nest_level = dev->nested_level;
4529 spin_lock_nested(&dev->addr_list_lock, nest_level);
4532 static inline void netif_addr_lock_bh(struct net_device *dev)
4534 unsigned char nest_level = 0;
4536 #ifdef CONFIG_LOCKDEP
4537 nest_level = dev->nested_level;
4540 spin_lock_nested(&dev->addr_list_lock, nest_level);
4543 static inline void netif_addr_unlock(struct net_device *dev)
4545 spin_unlock(&dev->addr_list_lock);
4548 static inline void netif_addr_unlock_bh(struct net_device *dev)
4550 spin_unlock_bh(&dev->addr_list_lock);
4554 * dev_addrs walker. Should be used only for read access. Call with
4555 * rcu_read_lock held.
4557 #define for_each_dev_addr(dev, ha) \
4558 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4560 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4562 void ether_setup(struct net_device *dev);
4564 /* Support for loadable net-drivers */
4565 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4566 unsigned char name_assign_type,
4567 void (*setup)(struct net_device *),
4568 unsigned int txqs, unsigned int rxqs);
4569 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4570 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4572 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4573 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4576 int register_netdev(struct net_device *dev);
4577 void unregister_netdev(struct net_device *dev);
4579 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4581 /* General hardware address lists handling functions */
4582 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4583 struct netdev_hw_addr_list *from_list, int addr_len);
4584 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4585 struct netdev_hw_addr_list *from_list, int addr_len);
4586 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4587 struct net_device *dev,
4588 int (*sync)(struct net_device *, const unsigned char *),
4589 int (*unsync)(struct net_device *,
4590 const unsigned char *));
4591 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4592 struct net_device *dev,
4593 int (*sync)(struct net_device *,
4594 const unsigned char *, int),
4595 int (*unsync)(struct net_device *,
4596 const unsigned char *, int));
4597 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4598 struct net_device *dev,
4599 int (*unsync)(struct net_device *,
4600 const unsigned char *, int));
4601 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4602 struct net_device *dev,
4603 int (*unsync)(struct net_device *,
4604 const unsigned char *));
4605 void __hw_addr_init(struct netdev_hw_addr_list *list);
4607 /* Functions used for device addresses handling */
4608 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4609 unsigned char addr_type);
4610 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4611 unsigned char addr_type);
4612 void dev_addr_flush(struct net_device *dev);
4613 int dev_addr_init(struct net_device *dev);
4615 /* Functions used for unicast addresses handling */
4616 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4617 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4618 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4619 int dev_uc_sync(struct net_device *to, struct net_device *from);
4620 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4621 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4622 void dev_uc_flush(struct net_device *dev);
4623 void dev_uc_init(struct net_device *dev);
4626 * __dev_uc_sync - Synchonize device's unicast list
4627 * @dev: device to sync
4628 * @sync: function to call if address should be added
4629 * @unsync: function to call if address should be removed
4631 * Add newly added addresses to the interface, and release
4632 * addresses that have been deleted.
4634 static inline int __dev_uc_sync(struct net_device *dev,
4635 int (*sync)(struct net_device *,
4636 const unsigned char *),
4637 int (*unsync)(struct net_device *,
4638 const unsigned char *))
4640 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4644 * __dev_uc_unsync - Remove synchronized addresses from device
4645 * @dev: device to sync
4646 * @unsync: function to call if address should be removed
4648 * Remove all addresses that were added to the device by dev_uc_sync().
4650 static inline void __dev_uc_unsync(struct net_device *dev,
4651 int (*unsync)(struct net_device *,
4652 const unsigned char *))
4654 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4657 /* Functions used for multicast addresses handling */
4658 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4659 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4660 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4661 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4662 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4663 int dev_mc_sync(struct net_device *to, struct net_device *from);
4664 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4665 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4666 void dev_mc_flush(struct net_device *dev);
4667 void dev_mc_init(struct net_device *dev);
4670 * __dev_mc_sync - Synchonize device's multicast list
4671 * @dev: device to sync
4672 * @sync: function to call if address should be added
4673 * @unsync: function to call if address should be removed
4675 * Add newly added addresses to the interface, and release
4676 * addresses that have been deleted.
4678 static inline int __dev_mc_sync(struct net_device *dev,
4679 int (*sync)(struct net_device *,
4680 const unsigned char *),
4681 int (*unsync)(struct net_device *,
4682 const unsigned char *))
4684 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4688 * __dev_mc_unsync - Remove synchronized addresses from device
4689 * @dev: device to sync
4690 * @unsync: function to call if address should be removed
4692 * Remove all addresses that were added to the device by dev_mc_sync().
4694 static inline void __dev_mc_unsync(struct net_device *dev,
4695 int (*unsync)(struct net_device *,
4696 const unsigned char *))
4698 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4701 /* Functions used for secondary unicast and multicast support */
4702 void dev_set_rx_mode(struct net_device *dev);
4703 void __dev_set_rx_mode(struct net_device *dev);
4704 int dev_set_promiscuity(struct net_device *dev, int inc);
4705 int dev_set_allmulti(struct net_device *dev, int inc);
4706 void netdev_state_change(struct net_device *dev);
4707 void __netdev_notify_peers(struct net_device *dev);
4708 void netdev_notify_peers(struct net_device *dev);
4709 void netdev_features_change(struct net_device *dev);
4710 /* Load a device via the kmod */
4711 void dev_load(struct net *net, const char *name);
4712 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4713 struct rtnl_link_stats64 *storage);
4714 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4715 const struct net_device_stats *netdev_stats);
4716 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4717 const struct pcpu_sw_netstats __percpu *netstats);
4718 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4720 extern int netdev_max_backlog;
4721 extern int netdev_tstamp_prequeue;
4722 extern int netdev_unregister_timeout_secs;
4723 extern int weight_p;
4724 extern int dev_weight_rx_bias;
4725 extern int dev_weight_tx_bias;
4726 extern int dev_rx_weight;
4727 extern int dev_tx_weight;
4728 extern int gro_normal_batch;
4731 NESTED_SYNC_IMM_BIT,
4732 NESTED_SYNC_TODO_BIT,
4735 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4736 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4738 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4739 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4741 struct netdev_nested_priv {
4742 unsigned char flags;
4746 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4747 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4748 struct list_head **iter);
4749 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4750 struct list_head **iter);
4752 #ifdef CONFIG_LOCKDEP
4753 static LIST_HEAD(net_unlink_list);
4755 static inline void net_unlink_todo(struct net_device *dev)
4757 if (list_empty(&dev->unlink_list))
4758 list_add_tail(&dev->unlink_list, &net_unlink_list);
4762 /* iterate through upper list, must be called under RCU read lock */
4763 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4764 for (iter = &(dev)->adj_list.upper, \
4765 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4767 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4769 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4770 int (*fn)(struct net_device *upper_dev,
4771 struct netdev_nested_priv *priv),
4772 struct netdev_nested_priv *priv);
4774 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4775 struct net_device *upper_dev);
4777 bool netdev_has_any_upper_dev(struct net_device *dev);
4779 void *netdev_lower_get_next_private(struct net_device *dev,
4780 struct list_head **iter);
4781 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4782 struct list_head **iter);
4784 #define netdev_for_each_lower_private(dev, priv, iter) \
4785 for (iter = (dev)->adj_list.lower.next, \
4786 priv = netdev_lower_get_next_private(dev, &(iter)); \
4788 priv = netdev_lower_get_next_private(dev, &(iter)))
4790 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4791 for (iter = &(dev)->adj_list.lower, \
4792 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4794 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4796 void *netdev_lower_get_next(struct net_device *dev,
4797 struct list_head **iter);
4799 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4800 for (iter = (dev)->adj_list.lower.next, \
4801 ldev = netdev_lower_get_next(dev, &(iter)); \
4803 ldev = netdev_lower_get_next(dev, &(iter)))
4805 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4806 struct list_head **iter);
4807 int netdev_walk_all_lower_dev(struct net_device *dev,
4808 int (*fn)(struct net_device *lower_dev,
4809 struct netdev_nested_priv *priv),
4810 struct netdev_nested_priv *priv);
4811 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4812 int (*fn)(struct net_device *lower_dev,
4813 struct netdev_nested_priv *priv),
4814 struct netdev_nested_priv *priv);
4816 void *netdev_adjacent_get_private(struct list_head *adj_list);
4817 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4818 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4819 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4820 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4821 struct netlink_ext_ack *extack);
4822 int netdev_master_upper_dev_link(struct net_device *dev,
4823 struct net_device *upper_dev,
4824 void *upper_priv, void *upper_info,
4825 struct netlink_ext_ack *extack);
4826 void netdev_upper_dev_unlink(struct net_device *dev,
4827 struct net_device *upper_dev);
4828 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4829 struct net_device *new_dev,
4830 struct net_device *dev,
4831 struct netlink_ext_ack *extack);
4832 void netdev_adjacent_change_commit(struct net_device *old_dev,
4833 struct net_device *new_dev,
4834 struct net_device *dev);
4835 void netdev_adjacent_change_abort(struct net_device *old_dev,
4836 struct net_device *new_dev,
4837 struct net_device *dev);
4838 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4839 void *netdev_lower_dev_get_private(struct net_device *dev,
4840 struct net_device *lower_dev);
4841 void netdev_lower_state_changed(struct net_device *lower_dev,
4842 void *lower_state_info);
4844 /* RSS keys are 40 or 52 bytes long */
4845 #define NETDEV_RSS_KEY_LEN 52
4846 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4847 void netdev_rss_key_fill(void *buffer, size_t len);
4849 int skb_checksum_help(struct sk_buff *skb);
4850 int skb_crc32c_csum_help(struct sk_buff *skb);
4851 int skb_csum_hwoffload_help(struct sk_buff *skb,
4852 const netdev_features_t features);
4854 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4855 netdev_features_t features, bool tx_path);
4856 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4857 netdev_features_t features);
4859 struct netdev_bonding_info {
4864 struct netdev_notifier_bonding_info {
4865 struct netdev_notifier_info info; /* must be first */
4866 struct netdev_bonding_info bonding_info;
4869 void netdev_bonding_info_change(struct net_device *dev,
4870 struct netdev_bonding_info *bonding_info);
4872 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4873 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4875 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4882 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4884 return __skb_gso_segment(skb, features, true);
4886 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4888 static inline bool can_checksum_protocol(netdev_features_t features,
4891 if (protocol == htons(ETH_P_FCOE))
4892 return !!(features & NETIF_F_FCOE_CRC);
4894 /* Assume this is an IP checksum (not SCTP CRC) */
4896 if (features & NETIF_F_HW_CSUM) {
4897 /* Can checksum everything */
4902 case htons(ETH_P_IP):
4903 return !!(features & NETIF_F_IP_CSUM);
4904 case htons(ETH_P_IPV6):
4905 return !!(features & NETIF_F_IPV6_CSUM);
4912 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4914 static inline void netdev_rx_csum_fault(struct net_device *dev,
4915 struct sk_buff *skb)
4919 /* rx skb timestamps */
4920 void net_enable_timestamp(void);
4921 void net_disable_timestamp(void);
4923 #ifdef CONFIG_PROC_FS
4924 int __init dev_proc_init(void);
4926 #define dev_proc_init() 0
4929 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4930 struct sk_buff *skb, struct net_device *dev,
4933 __this_cpu_write(softnet_data.xmit.more, more);
4934 return ops->ndo_start_xmit(skb, dev);
4937 static inline bool netdev_xmit_more(void)
4939 return __this_cpu_read(softnet_data.xmit.more);
4942 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4943 struct netdev_queue *txq, bool more)
4945 const struct net_device_ops *ops = dev->netdev_ops;
4948 rc = __netdev_start_xmit(ops, skb, dev, more);
4949 if (rc == NETDEV_TX_OK)
4950 txq_trans_update(txq);
4955 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4957 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4960 extern const struct kobj_ns_type_operations net_ns_type_operations;
4962 const char *netdev_drivername(const struct net_device *dev);
4964 void linkwatch_run_queue(void);
4966 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4967 netdev_features_t f2)
4969 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4970 if (f1 & NETIF_F_HW_CSUM)
4971 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4973 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4979 static inline netdev_features_t netdev_get_wanted_features(
4980 struct net_device *dev)
4982 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4984 netdev_features_t netdev_increment_features(netdev_features_t all,
4985 netdev_features_t one, netdev_features_t mask);
4987 /* Allow TSO being used on stacked device :
4988 * Performing the GSO segmentation before last device
4989 * is a performance improvement.
4991 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4992 netdev_features_t mask)
4994 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4997 int __netdev_update_features(struct net_device *dev);
4998 void netdev_update_features(struct net_device *dev);
4999 void netdev_change_features(struct net_device *dev);
5001 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5002 struct net_device *dev);
5004 netdev_features_t passthru_features_check(struct sk_buff *skb,
5005 struct net_device *dev,
5006 netdev_features_t features);
5007 netdev_features_t netif_skb_features(struct sk_buff *skb);
5009 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5011 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5013 /* check flags correspondence */
5014 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5015 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5016 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5017 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5018 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5019 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5020 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5021 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5022 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5023 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5024 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5025 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5026 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5027 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5028 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5029 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5030 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5031 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5032 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5034 return (features & feature) == feature;
5037 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5039 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5040 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5043 static inline bool netif_needs_gso(struct sk_buff *skb,
5044 netdev_features_t features)
5046 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5047 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5048 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5051 static inline void netif_set_gso_max_size(struct net_device *dev,
5054 dev->gso_max_size = size;
5057 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5058 int pulled_hlen, u16 mac_offset,
5061 skb->protocol = protocol;
5062 skb->encapsulation = 1;
5063 skb_push(skb, pulled_hlen);
5064 skb_reset_transport_header(skb);
5065 skb->mac_header = mac_offset;
5066 skb->network_header = skb->mac_header + mac_len;
5067 skb->mac_len = mac_len;
5070 static inline bool netif_is_macsec(const struct net_device *dev)
5072 return dev->priv_flags & IFF_MACSEC;
5075 static inline bool netif_is_macvlan(const struct net_device *dev)
5077 return dev->priv_flags & IFF_MACVLAN;
5080 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5082 return dev->priv_flags & IFF_MACVLAN_PORT;
5085 static inline bool netif_is_bond_master(const struct net_device *dev)
5087 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5090 static inline bool netif_is_bond_slave(const struct net_device *dev)
5092 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5095 static inline bool netif_supports_nofcs(struct net_device *dev)
5097 return dev->priv_flags & IFF_SUPP_NOFCS;
5100 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5102 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5105 static inline bool netif_is_l3_master(const struct net_device *dev)
5107 return dev->priv_flags & IFF_L3MDEV_MASTER;
5110 static inline bool netif_is_l3_slave(const struct net_device *dev)
5112 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5115 static inline bool netif_is_bridge_master(const struct net_device *dev)
5117 return dev->priv_flags & IFF_EBRIDGE;
5120 static inline bool netif_is_bridge_port(const struct net_device *dev)
5122 return dev->priv_flags & IFF_BRIDGE_PORT;
5125 static inline bool netif_is_ovs_master(const struct net_device *dev)
5127 return dev->priv_flags & IFF_OPENVSWITCH;
5130 static inline bool netif_is_ovs_port(const struct net_device *dev)
5132 return dev->priv_flags & IFF_OVS_DATAPATH;
5135 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5137 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5140 static inline bool netif_is_team_master(const struct net_device *dev)
5142 return dev->priv_flags & IFF_TEAM;
5145 static inline bool netif_is_team_port(const struct net_device *dev)
5147 return dev->priv_flags & IFF_TEAM_PORT;
5150 static inline bool netif_is_lag_master(const struct net_device *dev)
5152 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5155 static inline bool netif_is_lag_port(const struct net_device *dev)
5157 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5160 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5162 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5165 static inline bool netif_is_failover(const struct net_device *dev)
5167 return dev->priv_flags & IFF_FAILOVER;
5170 static inline bool netif_is_failover_slave(const struct net_device *dev)
5172 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5175 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5176 static inline void netif_keep_dst(struct net_device *dev)
5178 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5181 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5182 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5184 /* TODO: reserve and use an additional IFF bit, if we get more users */
5185 return dev->priv_flags & IFF_MACSEC;
5188 extern struct pernet_operations __net_initdata loopback_net_ops;
5190 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5192 /* netdev_printk helpers, similar to dev_printk */
5194 static inline const char *netdev_name(const struct net_device *dev)
5196 if (!dev->name[0] || strchr(dev->name, '%'))
5197 return "(unnamed net_device)";
5201 static inline bool netdev_unregistering(const struct net_device *dev)
5203 return dev->reg_state == NETREG_UNREGISTERING;
5206 static inline const char *netdev_reg_state(const struct net_device *dev)
5208 switch (dev->reg_state) {
5209 case NETREG_UNINITIALIZED: return " (uninitialized)";
5210 case NETREG_REGISTERED: return "";
5211 case NETREG_UNREGISTERING: return " (unregistering)";
5212 case NETREG_UNREGISTERED: return " (unregistered)";
5213 case NETREG_RELEASED: return " (released)";
5214 case NETREG_DUMMY: return " (dummy)";
5217 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5218 return " (unknown)";
5221 __printf(3, 4) __cold
5222 void netdev_printk(const char *level, const struct net_device *dev,
5223 const char *format, ...);
5224 __printf(2, 3) __cold
5225 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5226 __printf(2, 3) __cold
5227 void netdev_alert(const struct net_device *dev, const char *format, ...);
5228 __printf(2, 3) __cold
5229 void netdev_crit(const struct net_device *dev, const char *format, ...);
5230 __printf(2, 3) __cold
5231 void netdev_err(const struct net_device *dev, const char *format, ...);
5232 __printf(2, 3) __cold
5233 void netdev_warn(const struct net_device *dev, const char *format, ...);
5234 __printf(2, 3) __cold
5235 void netdev_notice(const struct net_device *dev, const char *format, ...);
5236 __printf(2, 3) __cold
5237 void netdev_info(const struct net_device *dev, const char *format, ...);
5239 #define netdev_level_once(level, dev, fmt, ...) \
5241 static bool __print_once __read_mostly; \
5243 if (!__print_once) { \
5244 __print_once = true; \
5245 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5249 #define netdev_emerg_once(dev, fmt, ...) \
5250 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5251 #define netdev_alert_once(dev, fmt, ...) \
5252 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5253 #define netdev_crit_once(dev, fmt, ...) \
5254 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5255 #define netdev_err_once(dev, fmt, ...) \
5256 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5257 #define netdev_warn_once(dev, fmt, ...) \
5258 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5259 #define netdev_notice_once(dev, fmt, ...) \
5260 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5261 #define netdev_info_once(dev, fmt, ...) \
5262 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5264 #define MODULE_ALIAS_NETDEV(device) \
5265 MODULE_ALIAS("netdev-" device)
5267 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5268 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5269 #define netdev_dbg(__dev, format, args...) \
5271 dynamic_netdev_dbg(__dev, format, ##args); \
5273 #elif defined(DEBUG)
5274 #define netdev_dbg(__dev, format, args...) \
5275 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5277 #define netdev_dbg(__dev, format, args...) \
5280 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5284 #if defined(VERBOSE_DEBUG)
5285 #define netdev_vdbg netdev_dbg
5288 #define netdev_vdbg(dev, format, args...) \
5291 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5297 * netdev_WARN() acts like dev_printk(), but with the key difference
5298 * of using a WARN/WARN_ON to get the message out, including the
5299 * file/line information and a backtrace.
5301 #define netdev_WARN(dev, format, args...) \
5302 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5303 netdev_reg_state(dev), ##args)
5305 #define netdev_WARN_ONCE(dev, format, args...) \
5306 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5307 netdev_reg_state(dev), ##args)
5309 /* netif printk helpers, similar to netdev_printk */
5311 #define netif_printk(priv, type, level, dev, fmt, args...) \
5313 if (netif_msg_##type(priv)) \
5314 netdev_printk(level, (dev), fmt, ##args); \
5317 #define netif_level(level, priv, type, dev, fmt, args...) \
5319 if (netif_msg_##type(priv)) \
5320 netdev_##level(dev, fmt, ##args); \
5323 #define netif_emerg(priv, type, dev, fmt, args...) \
5324 netif_level(emerg, priv, type, dev, fmt, ##args)
5325 #define netif_alert(priv, type, dev, fmt, args...) \
5326 netif_level(alert, priv, type, dev, fmt, ##args)
5327 #define netif_crit(priv, type, dev, fmt, args...) \
5328 netif_level(crit, priv, type, dev, fmt, ##args)
5329 #define netif_err(priv, type, dev, fmt, args...) \
5330 netif_level(err, priv, type, dev, fmt, ##args)
5331 #define netif_warn(priv, type, dev, fmt, args...) \
5332 netif_level(warn, priv, type, dev, fmt, ##args)
5333 #define netif_notice(priv, type, dev, fmt, args...) \
5334 netif_level(notice, priv, type, dev, fmt, ##args)
5335 #define netif_info(priv, type, dev, fmt, args...) \
5336 netif_level(info, priv, type, dev, fmt, ##args)
5338 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5339 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5340 #define netif_dbg(priv, type, netdev, format, args...) \
5342 if (netif_msg_##type(priv)) \
5343 dynamic_netdev_dbg(netdev, format, ##args); \
5345 #elif defined(DEBUG)
5346 #define netif_dbg(priv, type, dev, format, args...) \
5347 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5349 #define netif_dbg(priv, type, dev, format, args...) \
5352 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5357 /* if @cond then downgrade to debug, else print at @level */
5358 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5361 netif_dbg(priv, type, netdev, fmt, ##args); \
5363 netif_ ## level(priv, type, netdev, fmt, ##args); \
5366 #if defined(VERBOSE_DEBUG)
5367 #define netif_vdbg netif_dbg
5369 #define netif_vdbg(priv, type, dev, format, args...) \
5372 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5378 * The list of packet types we will receive (as opposed to discard)
5379 * and the routines to invoke.
5381 * Why 16. Because with 16 the only overlap we get on a hash of the
5382 * low nibble of the protocol value is RARP/SNAP/X.25.
5396 #define PTYPE_HASH_SIZE (16)
5397 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5399 extern struct list_head ptype_all __read_mostly;
5400 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5402 extern struct net_device *blackhole_netdev;
5404 #endif /* _LINUX_NETDEVICE_H */